THE INTERPLAY BETWEEN ATMOSPHERE, HYDROLOGY AND LAND USE BY ENVIRONMENTAL MODELLING

Fábio Pereira

Doctoral Thesis 2013

THE

INTERPLAY BETWEEN ATMOSPHERE , HYDROLOGY AND LAND USE BY ENVIRONMENTAL

MODELLING

© 2013 Fábio Pereira All rights reserved Printed in Sweden by Media-Tryck, Lund, 2013

Department of Water Resources Engineering, Lund University John Ericsson väg 1 P.O. Box 118 SE–221 00 Lund Sweden http://www.tvrl.lth.se Cover: A picture of sugargane plantations near São Paulo. Photo Credits: Ian Allen (Fast Company magazine) ISSN 1101-9824 CODEN LUTVDG/(TVVR–1061)(2013) ISBN 978-91-7473-775-2 REPORT 1061

"Water is the driving force of all nature" Leonardo da Vinci

ACKNOWLEDGMENTS After almost four years, it’s now time to say thank to those people who contributed the most in making this thesis possible. First and foremost, I thank God for providing me this opportunity and granting me the capability to proceed successfully. I would also like to express my deep and sincere gratitude to my supervisor, Prof. Cintia Bertacchi Uvo, for providing invaluable guidance throughout my research. Her wide knowledge and logical way of thinking have been of great value to me. The way she inspired, advised and helped both in my academic and personal life has been unforgettable. She showed me different ways to tackle a research problem and the need to be persistent to accomplish my goal. At last, but not least, I thank to her for teaching me how to cook "properly". I feel I am very lucky for getting her as my supervisor. I would also like to extend my sincerest thanks to her family members (Little and Big Eriks). Especially to Big Erik, who taught me to never give up by seeing his support for Malmö FF even though they could barely beat Syrianska FC. I also wish to thank Prof. Magnus Larson for his friendship, empathy and great sense of humor. His endless optimism and kindness will always be in my mind. It was an enormous privilege and honor to share the same work environment with him. I am very grateful to Prof. Lars Bengtsson, Prof. Hans Hanson, Prof. Rolf Larson, Prof. Ronny Berndtsson, Prof. Kenneth Persson, Prof. Linus Zhang and Prof. Magnus Persson for providing a friendly and lively atmosphere at the Department of Water Resources Engineering during my stay in Sweden. I also thank to Dr. Mohammad Aljaradin, Dr. Arun Rana, Jaime Palalane, Feifei Yuan, Shuang Liu, Kean Foster, Hossein Hashemi, Johanna Sörensen, Angelica Lidén, Sofia Westergren, Lena Flyborg, Erik Nilsson, Clemêncio Nhantumbo and Estevão Pondja, my fellow present and former doctoral students. And now, Brazilians. Numerous thanks go to Ivaylo Vasilev (not Brazilian, but consider to be half Brazilian now), a great friend, even better personal trainer, for patiently coaching me at Gerdahallen. Of course I cannot forget my pub mate, Walan Grizoli for being there all Friday evenings no matter how bad the weather was. I also have to thank other Brazilians which were part of the Brazilian community in v

Lund, Bianca, Juninho, Ada, Eduardo, Ricardo, Danilo among many others. I want to take this opportunity to express my special thanks to Claudia Rivera for all her precious help, her availability, her patience and her politeness. Whatever and whenever I asked her, she always had a cheerful heart and a smile on her face. You were always there for me and I will always keep you in my thoughts. At last, my deepest gratitude goes to my family for their unflagging love and support throughout my life. I am indebted to my mother, Lysete Farias, my father, Abedias Pereira, my sister, Flávia Pereira and her daughter Maria Luísa for their care and love.

vi

ABSTRACT Interactions between land surface and atmosphere induced by human activities and natural environmental dynamics act on a time scale that varies from seconds to millions of years. It is by exchanging heat, water, energy and carbon that land surface and atmospheric processes are closely interrelated and influence each other in reciprocal ways. Among the natural interactions between land surface and atmosphere, the water cycle stands out for its complexity and relevance to all other physical processes. In this context, numerical models are largely recommended as tools capable of quantifying, predicting and assessing the soil, surface and atmospheric water budgets. Atmospheric models present a detailed and complex approach to atmospheric processes that includes estimation of carbon, heat, energy and water fluxes between surface and atmosphere based on energy, mass and momentum equations. The numerical solution of these equations include effects of sub-grid scale processes that are not resolved by numerical schemes but affect the resolved scales. These unresolved processes are described in terms of semi-empirical relationships by means of parameterizations. Despite widely acknowledged improvements in parameterizations of vegetation and soil processes in order to interpret the complexities inherent in the water cycle and its interactions with the atmosphere, parameterization schemes usually apply prescribed values of parameters based on their probability density functions which assume vegetation and soil characteristics as continuous distributions, consequently, mixtures in soil and vegetation within an area of interest are not captured. Therefore, the research described in this thesis aimed at understanding the interplay between hydrology and the atmosphere under land use changes using a two-way coupled model that incorporates a process-based approach to land surface hydrological processes to resolve both surface features and the full atmospheric response to them. A rapid expansion of the plantation of sugarcane over the Rio Grande basin, Brazil, as a response to government measures to boost ethanol production was used as a case study on this thesis. To reproduce the sugarcane expansion over the Rio Grande basin, historical land use scenarios were defined based on satellite images captured vii

Abstract

in 1993, 2000 and 2007. Further, a forth land use scenario was also generated based on the mapping of areas suitable for cultivation of sugarcane made by the Brazilian Institute for Agricultural Research EMBRAPA. Thereafter, specific model parameters for sugarcane were calibrated and validated to perform analysis of short-, medium- and long-term impacts of sugarcane expansion on the local hydrology. In the meanwhile, an atmospheric-hydrological modelling system was implemented and tested against estimates from a well-known atmospheric model, satellite imagery and observed data. Results obtained from numerical and imagery analysis revealed that most of the sugarcane expansion occurred over areas close to the outlet of the Rio Grande basin where climate and topographical conditions are more attractive for growing sugarcane. They also indicated that the amount of areas replaced with sugarcane plantations, their location within the basin, regional soil properties, and local groundwater contribution to stream flow are the main factors related to the impacts of sugarcane expansion on the water balance of the Rio Grande basin. Additionally, numerical analyses carried out in this thesis showed that the replacement of land surface parameterizations by process-based hydrological modelling implied improvement in temperature, atmospheric water content and zonal and meridional winds calculated near land surface. Finally, a conceptual evaluation of the interplay of land use changes, hydrology and atmosphere as given by the hybrid coupled model was carried out. The main goal of this evaluation was to assess whether the model behavior is in accordance with improved understanding of the hydrological cycle of the Rio Grande basin under land use changes due to sugarcane expansion over its drainage area, resulting in this work. The results obtained from four model runs using historic and possible land use scenarios showed that exchange of water between soil, land surface and atmosphere is an important factor that determines which processes will dominate the water balance during wet and dry seasons under expansion of agricultural lands.

viii

PAPERS

This thesis is based on the following papers, which will be referred to by their Roman numerals in the text. The papers are appended to the thesis.

Appended papers I Quantifying the Rapid Sugarcane Expansion for Ethanol Production in the Rio Grande basin, Brazil F. F. Pereira M. Tursunov and C. B. Uvo. Vatten 69, 83 – 86 (2013). II Effects of sugarcane expansion on runoff and evapotranspiration in the Rio Grande basin, Brazil. C. Borglin, S. Borglin, F. F. Pereira and C. B. Uvo. Vatten 69, 141 – 148 (2013). III Towards the response of water balance to sugarcane expansion in the Rio Grande basin, Brazil F. F. Pereira, M. Tursunov and C. B. Uvo. Hydrology and Earth System Sciences, Reviewed. IV Assessment of Numerical Schemes for Solving Advection-Diffusion Equation on Unstructured Grids: Case Study of River Guaíba, Brazil F. F. Pereira, C. R. Fragoso Jr, C. B. Uvo, W. Collischonn and D. M. L. M. Marques. Nonlinear Processes in Geophysics, Reviewed.

ix

V Implementation of a two-way coupled atmospheric-hydrological system for environmental modeling at regional scale F. F. Pereira, M. A. E. de Moraes and C. B. Uvo. Hydrology Research, Accepted. VI Assessment of the role of a process-based representation of land surface hydrological processes to the atmosphere F. F. Pereira and C. B. Uvo. Journal of Hydrology Submitted, Under review. VII Conceptual evaluation of the interplay of land use changes, hydrology and atmosphere by a hybrid atmospheric-hydrological coupled model F. F. Pereira and C. B. Uvo. , Manuscript.

Author’s contribution to the appended papers I. F. F. Pereira closely followed the process of selection of satellite imagery and automatic land use classification. Thereafter, F. F. Pereira analysed and discussed the reliability of the results with the co-authors, and wrote the paper. II. F. F. Pereira planned the land use scenarios used in the hydrological model, analysed the reliability of the results, and wrote the paper together with the co-authors. III. F. F. Pereira preprocessed satellite imagery prior to using them as land use scenarios. The natural response of the hydrological cycle to sugarcane expansion was estimated using simulations performed by F. F. Pereira for several temporal horizons. Finally, insights were gained from fruitful discussions with the co-authors which helped F. F. Pereira to write the paper and to draw its conclusions. IV. Together with the co-authors, F. F. Pereira evaluated different numerical schemes for solving the Advection-Diffusion Equation on unstructured grids. Additionally, F. F. Pereira outlined advantages and disadvantages of each numerical scheme analysed.

x

V. In light of several discussions with the co-authors, F. F. Pereira proposed a two-way strategy of coupling between atmospheric and hydrological models. Additionally, F. F. Pereira carried out comparisons of the coupling strategy against observed data. VI. F. F. Pereira performed numerical analyses of the process-based approach to hydrological processes given by the coupling strategy. Concepts, terminologies and results obtained from the numerical analyses were carefully discussed with the co-author and written down by F. F. Pereira. VII. F. F. Pereira performed four model runs using an atmospherichydrological coupled model. Thereafter, F. F. Pereira wrote the manuscript that describes the interplay of land use changes, hydrology and atmosphere.

Related publications not included in this thesis Scientific Journals F. F. Pereira, C. R. Fragoso Jr, W. Collischonn and D. M. L. M. Marques, 2013. Simulação do transporte de escalares em corpos d’água rasos usando um modelo de grades não estruturadas (Simulating scalar transport in shallow water bodies using an unstructured grids). Revista Brasileira de Recursos Hídricos (18), 7 – 18 (In Portuguese). F. F. Pereira, C. R. Fragoso Jr, C. B. Uvo and D. M. L. M. Marques, 2013. Pairing multivariate data analysis and ecological modeling in the biomanipulated Lake Engelsholm, Denmark. Vatten (69), 15 – 19.

xi

Conferences and workshops F. F. Pereira, R. R. Souza, C. R. Fragoso Jr, C. B. Uvo and D. M. L. M. Marques, 2010. Development of a two-dimensional unstructured hydrodynamic model for subtropical aquatic ecosystems. Workshop on Lake Ecosystem Modelling, Silkeborg, Denmark. R. R. Souza, F. F. Pereira, D. M. L. M. Marques, L.-A. Hanson and C. B. Uvo, 2010. Consequences of Warmer Future on Metabolism of a Very Shallow Boreal Lake. Workshop on Lake Ecosystem Modelling, Silkeborg, Denmark. M. A. E. Moraes, F. F. Pereira and C. B. Uvo, 2011. Two way coupling of a conceptual hydrological model to a regional atmospheric model. 3rd International Multidisciplinary Conference on Hydrology and Ecology: Ecosystems, Groundwater and Surface Water Pressures and Options, Vienna, Austria. F. F. Pereira, M. A. E. Moraes and C. B. Uvo, 2012. Implementation of a two-way coupled atmospheric-hydrological system for environmental modeling at regional scale. 27th Nordic Hydrology Conference, Oulu, Finland. F. F. Pereira and C. B. Uvo, 2013. Assessment of a rapid sugarcane expansion upon the water balance of the Rio Grande basin, Brazil. IAHS - IAPSO - IASPEI Joint Assembly, Gothenburg, Sweden.

Master and bachelor theses supervised by F. F. Pereira Claes Borglin and Sara Borglin, 2013. The effects of sugarcane expansion on the hydrology of the Rio Grande Basin, Brazil. Master thesis, Lund University, Sweden. Robert Willander, 2012. Offline coupling of a Brazilian regional climate model and a 2D hydrodynamic model. Master thesis, Lund University, Sweden. Luiza Correia, 2011. Mapping mussels (Mytella Charruana) in CELMM by mathematical modelling (in Portuguese). Bachelor thesis, Federal University of Alagoas, Brazil.

xii

ABBREVIATIONS ANA

National Water Agency (in Portuguese)

ANEEL

Brazilian Electricity Portuguese)

ADE

Advection-Diffusion Equation

bbl

Barrel

BRAMS

Brazilian Regional Atmospheric System

CPTEC

Center for Weather Forecasts and Climate Studies (in Portuguese)

CPRM

Company for Mineral Resources Research

Regulatory

Agency

(in

DE/UFRGS Department of Ecology of the Federal University of Rio Grande do Sul (in Portuguese) DEM

Digital Elevation Model

DOE

U.S. Department of Energy

EMBRAPA

Brazilian Agricultural Research Corporation (in Portuguese)

EUMETSAT European Organisation for the Exploitation of Meteorological Satellites FAO

Food and Agriculture Organization of the United Nations

GCM

General Circulation Model

GIS

Geographic Information System

GW

Gigawatts

HRUs

Hydrologic Response Units

hPa

hectopascal

IBGE

Brazilian Institute of Geography and Statistics (in Portuguese)

IEA

International Energy Agency

INMET

National Institute of Meteorology (in Portuguese)

INPE

National Institute For Space Research (in Portuguese) xiii

Abbreviations

xiv

IPH

Institute for Hydraulic Research (in Portuguese)

IPT

Institute for Technological Research

kW

Kilowatts

Landsat TM

Landsat Thematic Mapper

LARSIM

Large Area Runoff Simulation Model

LES

Large Eddy Simulations

LEAF

Land Ecosystem Atmosphere Feedback model

lge

Liter per gasoline equivalent

m.a.s.l.

meters above sea level

MK

Mann-Kendall

MW

Megawatts

MGB

Model for Large Basins (in Portuguese)

NDVI

Normalized Difference Vegetation Index

NOAA

National Oceanic and Atmospheric Administration

NS

Nash-Sutcliffe coefficient

ONS

Electric System National Operator (in Portuguese)

RCM

Regional climate model

RMSE

Root-Mean-Square Error

RVE

Relative Volume Error

SRTM

Shuttle Radar Topography Mission

UN

United Nations

UNICA

Brazilian Sugarcane Portuguese)

USD

US dollar

USGS

US Geological Survey

Industry

Association

UTC

Coordinated Universal Time

VIC-2L

2-Layer Variable Infiltration Capacity model

WWF

World Wildlife Fund

(in

CONTENTS

Acknowledgments 1

2

3

4

Introduction 1.1 Background . . . . . . . . . . . . . . . . . . . . 1.2 Renewable energy sources . . . . . . . . . 1.2.1 Water . . . . . . . . . . . . . . . . . 1.2.2 Wind . . . . . . . . . . . . . . . . . . 1.2.3 Geothermal . . . . . . . . . . . . . 1.2.4 Tide . . . . . . . . . . . . . . . . . . 1.2.5 The sun . . . . . . . . . . . . . . . . 1.2.6 Biomass . . . . . . . . . . . . . . . 1.3 Objective and scope . . . . . . . . . . . . . . 1.4 Thesis structure and appended papers

v

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

1 1 2 2 3 3 4 4 5 8 9

Theoretical background 2.1 A brief history of the sugarcane in Brazil . . . . . . . . . . . . . . . 2.2 Sugarcane and the water balance . . . . . . . . . . . . . . . . . . . . 2.3 Numerical models, the hydrologic cycle and the atmosphere 2.4 Integrated modelling systems . . . . . . . . . . . . . . . . . . . . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

11 11 12 14 15

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

Material and methods 3.1 Study area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Distributed hydrological model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3 Regional atmospheric model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4 Two-way coupling methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4.1 Strategy for solving temporal mismatches . . . . . . . . . . . . . . 3.4.2 Strategy for solving spatial mismatches . . . . . . . . . . . . . . . . 3.5 Data sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.5.1 Data collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.5.2 Data preprocessing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.6 Numerical experiments and imagery analyses . . . . . . . . . . . . . . . . . . 3.6.1 Mapping of sugarcane plantations . . . . . . . . . . . . . . . . . . . 3.6.2 MGB-IPH specific parameters for sugarcane: calibration and validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.6.3 Models runs and their assumptions . . . . . . . . . . . . . . . . . . 3.7 Evaluation of the model behaviour . . . . . . . . . . . . . . . . . . . . . . . . . .

17 18 20 21 22 24 25 26 26 26 28 28

Results and discussions 4.1 Overview of the sugarcane expansion over the Rio Grande basin . . . . 4.2 Analysis of runoff trends . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3 Short-, medium- and long-impacts of the rapid sugarcane expansion on local hydrology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4 The atmospheric-hydrological modelling system: analysis and tests . . 4.5 Conceptual evaluation of the exchange of water between soil, land surface and atmosphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

37 38 38

29 31 34

39 41 45

Contents

4.5.1 4.5.2 5

Soil-land surface interface . . . . . . . . . . . . . . . . . . . . . . . . . Land surface-atmosphere interface . . . . . . . . . . . . . . . . . .

Conclusions and Outlook

References

47 49 51 55

Papers I

II

III

Quantifying the Rapid Sugarcane Expansion for Ethanol Production in the Rio Grande basin, Brazil

67

Effects of sugarcane expansion on runoff and evapotranspiration in the Rio Grande basin, Brazil.

73

Towards the response of water balance to sugarcane expansion in the Rio Grande basin, Brazil

83

IV

Assessment of Numerical Schemes for Solving Advection-Diffusion Equation on Unstructured Grids: Case Study of River Guaíba, Brazil 109

V

Implementation of a two-way coupled atmospheric-hydrological system for environmental modeling at regional scale 135

VI

Assessment of the role of a process-based representation of land surface hydrological processes to the atmosphere 149

VII

Conceptual evaluation of the interplay of land use changes, hydrology and atmosphere by a hybrid atmospheric-hydrological coupled model 183

CHAPTER

1

INTRODUCTION 1.1

Background

Humanity has always been dependent on the natural resources provided by the biosphere and its ecosystems. From ancient times to the 21s t century, air, waterways, biomass and/or oceans are used as energy sources by human beings to satisfy their needs [WRI, 2005]. From the beginning of human history till the Industrial Revolution, the human species had its first application of energy by using biomass (firewood) and muscular sources. The latter provided mostly work in the form of manual labor whereas biomass sources were used for heating and cooking needs. During this period, energy from waterways and air was also harnessed by water- and windmills. Aiming at optimizing the efficiency in the conversion of energy to useful work, the Industrial Revolution took off in the 18t h century. The transition from cottage industry to factory system brought a major shift in energy sources since animal power, manual labor and biomass were replaced by steam engines stoked with coal and oil [UN, 1998, 2012]. Besides social improvements in health and longevity, the Industrial Revolution was also characterized by the replacement of wind, water and wood with fossil fuels for the generation of energy. However, fossil fuels are formed by natural processes which take millions of years, and reserves were/are being depleted much faster than new ones are being made as a response to the increase of the demand for all basic human needs, such as food production, medicine, housing and clothing. Thus, a massive growth in energy consumption on limited resources lead fossil fuels to be considered as finite natural resources, and hence, non-renewable energy sources. Furthermore, many of our recent environmental problems are associated with the historic and current use of fossil fuels. By burning 1

1.2 Renewable energy sources

2.2% 7.2% 6.9%

83.7% Biomass Wind Water Geothermal, solar and tide

Figure 1.1. The use of renewable energy sources in the world [IEA, 2011a].

fossil fuels, for example, carbon dioxide (CO2 ) is emitted to the atmosphere, where it traps heat contributing to global warming. In addition, during their combustion, fossil fuels produce several pollutants such as, carbon monoxide (CO), nitrogen oxides (NO, NO2 , N2 O, etc.), sulfur oxides (SO, SO2 , SO3 , etc.), and hydrocarbons. Besides increasing air pollution, these pollutants can combine to form ozone (O3 ), the major constituent of smog [Basha et al., 2009; Fenger, 2009]. In response to these concerns raised by the use of fossil fuels as an energy source, research and technological developments have been made in order to provide ecologically friendly alternatives for the replacement of non-renewable resources. Basically, these ecologically friendly alternatives propose generating energy from natural resources that are continually replenished and involve less environmental impacts (e.g. sunlight, wind, biomass), so-called renewable energy sources.

1.2

Figure 1.2. Itaipu hydroelectric power plant across the board between Brazil and Paraguay. Source: [ANEEL, 2008].

Renewable energy sources

Conceptually, renewable energy is all forms of energy that come from resources which are present in unlimited quantity in nature and will not become exhausted or depleted by human activities. All of these resources are known as renewable energy sources, and include sunlight, water, wind, tide, geothermal heat and biomass sources. According to statistics from the International Energy Agency (IEA) for 2009, the world relied on renewable sources for around 13.1% of its primary energy supply. And, by 2016, renewable sources will edge out natural gas as the second biggest source of electricity, after coal [IEA, 2013]. Among the renewable energy sources, water stands out by generating more than 80% of the world renewable energy, followed by biomass (7.24%), wind (6.90%), geothermal (1.68%), the sun (0.53%) and tide (0.01%) (Figure 1.1). To understand how energy is extracted from these renewable resources, a brief description containing the main methods is given for each of them. Moreover, it also presents their benefits, potential and current global status.

1.2.1

Water

There are two basic ways of capturing energy from falling or running water — the first one impounds water behind a dam whereas the second one diverts water into parallel structures that produces mechanical or electrical energy from masses of water moving in (e.g. watermills, run-of-the-river plants). Although diverting water usually has lower impact on upstream and downstream ecosystems than damming water, its use is marginal due to its little or no capacity for water storage which makes generation of energy dependent on hydrological regime rather than consumer demand. 2

Introduction

The energy from dammed waters is generated by hydroelectric power plants (Fig. 1.2). It became the most widely form of renewable energy because dammed waters are used not only for generating energy but also water supply and irrigation. Since there is no consensus among the international community on the definition of small, medium and/or large hydroelectric power plants, their potential is usually measured by energy production rates that vary from 500 kW — typical supply for a small factory or community — to 20 G W — supply for big cities. Studies carried out by IEA [2011a] reveal that China, Canada and Brazil are the top three countries in hydropower generation. All together accounted for about one-third of the world’s hydropower generation in 2008 (Fig. 1.3). In addition, their studies estimate that the global hydropower generation has increased by 50% over the past two decades.

1.2.2

Wind

The energy that comes from the wind is generated by wind turbines. The energy production of a wind turbine depends on the size of the turbine and the wind resource where the turbine is installed. According to analysis made in IEA [2008], for instance, one medium wind turbine of 1.8 MW capacity at a reasonable site produces enough electricity to meet the needs of over 1000 households. As the wind does not always blow, groups of wind turbines, also known as wind farms (Figure 1.4), are usually found scattered across a region or country. Thus while one region may be calm, another one is windy which keeps the energy production constant throughout the year. Ideally, a good wind power site needs to be consistently windy. Offshore installations produce up to 50% more electricity than onshore wind power plants [IEA, 2008], due to their higher and steadier wind speeds. Motivated by government polices, Denmark is the birthplace of modern wind energy and responsible for a third of all wind turbines sold worldwide. However, its share on the global installed wind capacity is lower than 4%. Statistics for 2007 revealed that Germany, United States, Spain, India and China are the top six countries in installed wind capacity with 22.2, 16.9, 15.1, 7.8 and 5.9 installed G W , respectively [IEA, 2008].

1.2.3

11% 4% 5%

30%

9%

11% China 18% Canada Brazil United States Russia Norway India Others

12%

Figure 1.3. World’s shares in hydropower generation in 2008 (Extracted from IEA [2011a]).

Figure 1.4. Wind farms located on land (onshore) and in sea or freshwater (offshore) (Source: Whitlock [2013]).

Geothermal

Geothermal energy is derived from steam and hot water trapped in reservoirs under the surface of the Earth. Since geothermal reservoirs present high temperatures (generally above 180 ◦ C), the energy captured by geothermal power plants can be directly used for heating houses, schools and farms. However, most geothermal power plants use the steam to rotate turbines, which generate electricity. 3

1.2.4 Tide

Although geothermal reservoirs are sustainable resources because power plants inject back into them any leftover water and/or condensed steam, their use is economically viable only in areas where the crust of the Earth is thin and geothermal reservoirs are close to the surface. In these areas, geothermal energy is cost effective since the land needed for geothermal power plants is smaller per megawatt than most of other types of power plant. A small geothermal power plant (i.e. less than 5 M W ) has the potential to supply electricity for a small rural village. Between 2000 and 2004, the use of geothermal energy increased by 1 G W in the world. The world’s biggest producer of geothermal electricity is United States whereas Iceland uses geothermal energy as its primary source of heat. Besides United States and Iceland, geothermal energy is commonly used in Italy, Indonesia, Mexico, New Zealand, Japan and China.

1.2.4

Tide

Tidal power is converted to electricity by capturing the energy from water masses transported by tides or open sea currents. Despite tidal energy also being extracted from the water, it is not classified as hydroelectric energy because tides and open sea currents are generated by surface wind stresses and orbital mechanisms of the moon-Earth system. Some of the alternatives used to extract energy from tides and open sea currents include tidal barrages and tidal fences. While tidal barrages are built across the interface between estuaries and the open sea to divert the water that goes in and out through turbines (same as used for hydroelectricity), tidal fences are similar to turnstiles and are usually placed in the open sea. Although it is acknowledged that tidal power has a great potential for future energy generation owing to its high predictability and low environmental impacts compared to hydroelectric power [Krohn et al., 2013], the energy production of a turbine and/or a tidal fence is relatively low compared to the costs inherent in installation and maintenance since salty environments are hostile places for turbines to operate. Only a few tidal energy sites are currently in operation across the world. Two world’s biggest tidal energy sites are located in the Rance River in France and the White Sea in Russia. Together they generate less than 250 M W , which is enough electricity to power around 60000 homes.

1.2.5

The sun

The energy that comes from the sun falls on the Earth as solar radiation. Once the solar radiation reaches any surface or material, it is instantaneously transformed into heat. As it is the most abundant re4

Introduction

1.2.6

120 100

Gigawatts (GW)

newable energy source available over the entire globe, this heat provided by solar radiation can be used for many different purposes since heating for greenhouse structures in temperate and polar zones to powering absorption chillers harnessed by heat (hot water) in tropical zones. There are several ways to collect solar energy, from solar panels that extract solar energy as heat (i.e. solar heat collectors) to those panels which convert it directly into electricity, also known as solar photovoltaic (PV) collectors [IEA, 2009]. Only the latter, though, is used for large-scale generation of electricity. Yet, according to IEA [2011a], large-scale solar power plants composed of PV collectors are currently being built in Portugal, Spain and United States. Similarly to wind energy, solar energy is not available all of the time. Thus, the usage of solar energy is only effective when it is integrated with other energy source. Nonetheless, trends in the global market of solar energy reveal a growth rate of around 15% in 2007. In addition, future prospects for 2008 onwards suggested an expansion of the solar energy market as a response to large projects of solar industrial process heat in Europe and China [IEA, 2009]. Nowadays, China (101 G W ), United States (22 G W ), Turkey (8 G W ), Germany (8 G W ), Japan (5 G W ), Australia (4 G W ), Israel (4 G W ), Brazil (1 G W ), Austria (1 G W ) and Greece (1 G W ) are respectively the top ten countries with respect to cumulative solar installed capacity (Fig. 1.5).

80 60 40 20 0

CN US TR DE JP AU IL BR ATGR

Figure 1.5. The top ten countries in cumulative solar capacity installed in 2007 (Extracted from IEA [2009]).

Biomass

A biomass energy source is defined as all plant- or animal-based matter that can be used as raw material for energy production [IEA, 2012]. Therefore, all types of energy derived from biomass resources are known as bioenergy. The main biomass resources used for generating bioenergy are listed and described as follows [DOE, 2012]: Energy crops are low-cost and low-maintenance crops generally combusted to generate heat and/or electricity. Agricultural crops are crops which can be used either for production of food or bioenergy. Agricultural crop residues include material left in an agricultural field after harvest which may also be converted to bioenergy. Forestry residues are biomass not harvested from logging sites (paper or lumber industries) that can be used for energy production. Aquatic crops are aquatic plants that have potential to grow fast in shallow lakes and have recently been used for producing biofuels. 5

1.2.6 Biomass

Biomass processing residues are residues derived from biomass processing which yield byproducts and waste streams with significant energy potential. Municipal waste include residential, and commercial waste which can be used to generate gases as methane gas from landfills, for example. Animal waste include organic materials from farms and animalprocessing operations, generally used to make many products, including bioenergy. These biomass energy sources are used to produce a large variety of energy-related products, such as electricity, heat, liquid and gaseous fuels. Forestry residues, for instance, can be compacted to wood chips or pellets for heat or power production. Similarly, the methane gas captured and burned by sewage treatment plants is also used as source of heat and power. Additionally, crops such as wheat, corn and sugarcane can be used to make ethanol whereas biodiesel is derived from castor oil and oil wastes. An world energy overview made by IEA [2011b] reveals that bioenergy could provide 7.5% of world electricity generation by 2050. In addition, they also show that bioenergy could supply from 15% to 20% of the global demand for heat in 2050, which presents an ongoing increase as result of the deployment of advanced biomass cookingstoves and biogas distribution systems in developing countries IEA [2013]. Besides heat and electricity, the energy extracted from biomass resources can also be used in the transport sector. In this case, ethanol and biodiesel are used as alternatives to gasoline and ordinary diesel for vehicles, and since they are produced either from biomass or waste feedstocks, ethanol and biodiesel are widely known as biofuels. Over the next 40 years, studies carried out by IEA [2011b] suggest that the global demand for bioenergy will achieve 11 billion dry tons of biomass per year of which 4 billion will exclusively be used for production of biofuels. Currently, a sharp increase in the demand for biofuel has already been detected between 2002 and 2005, when the global biofuel consumption had tripled and 38 billion litres of ethanol were produced mainly in Brazil, USA and China [IEA, 2007]. This rapid increase in the use of biofuels is basically addressed to the adoption of biofuels as a measure to reduce greenhouse gases and oil dependence as well as to stimulate rural development. Further, commercial biofuel production costs — which do not include agricultural subsidies or any type of grants or incentives from governments — present a cost advantage over petroleum products, such as gasoline and diesel (Figure 1.6). Figure 1.6 compares the prices applied to the most common biofuels in US dollar (USD) per liter of gasoline equivalent (lge) with gasoline and diesel in USD per barrel (USD/bbl), which were collected at 6

Introduction

Figure 1.6. Comparisons between the costs inherent in the production of the most common biofuels (in 2005 and projections to 2030) and petroleumbased products (2005-06) (Extracted from IEA [2007]).

12 different global locations between 3 January 2005 and 6 April 2006. In terms of commercial biofuel costs, comparisons shown in Figure 1.6 reveal that ethanol made from sugarcane (Brazil) appears as the cheapest biofuel option with 0.25 USD/lge followed by biodiesel from animal fats (New Zealand) with 0.42 USD/l and ethanol from corn (US) with 0.60 USD/lge. Yet ethanol from sugarcane can compete with gasoline when the crude oil price is above 40 USD/bbl, what has not happened since 2005 [Smith, 2013]. Therefore, biofuels — specially ethanol from sugarcane — have been seen as a short-term alternative to fossil fuels. Along with the use of ethanol from sugarcane, concerns over its sustainability has substantially increased over the past decade [Alonso-Pippo et al., 2008; Goldemberg et al., 2008; Ravindranath et al., 2005; Smeets et al., 2008]. To be sustainable, the use of ethanol must not lead to overall degradation of the natural resources or threaten people’s ability to meet their basic needs. Consequently, several questions have been raised regarding current practices used for ethanol production: • Since sugarcane is also considered a food crop, will ethanol production increase food prices? [Ajanovic, 2011; Mitchell, 2008] • Could a decline in food/feed production lead to undesirable impacts such as world hunger from higher food prices? [Runge and Senauer, 2007; Tenenbaum, 2008] • Will conversion to sugarcane land cause a regional atmospherehydrologic system to switch from a moist state to a drier one or 7

1.3 Objective and scope

vice versa? [de Fraiture et al., 2008; Jewitt and Kunz, 2011] • Will there be policies to regulate proper land use planning that preserve and protect biodiversity while expanding food and ethanol production? [Fletcher Jr et al., 2011; Groom et al., 2008]

1.3

Objective and scope

The research described in this thesis aimed at understanding regional changes in the environment as function of complex and reciprocal interactions between hydrology, land use and atmosphere. To do so, an integrated modelling system was implemented in a way that enables local changes in the land use, hydrologic cycle and/or atmosphere to be dynamically incorporated into calculations. Thereafter, the integrated modelling system has been used to evaluate the natural response of the atmosphere and hydrologic cycle to land use changes due to sugarcane expansion at a basin scale. To achieve the overall objective of this thesis, the following research questions were addressed: • What is the appropriate numerical approach to provide more flexibility to the different numerical schemes employed in the integrated modelling system? • How to couple physical processes that occur at different spatial and temporal scale? • Was there really a sugarcane expansion over the past decades? If so, can any trend and/or pattern be identified in time series data? • Whether all areas suitable for growing sugarcane were filled up by sugarcane crops, what is the potential impacts of this probable expansion on the local atmosphere and hydrology? The research presented in this thesis focus on sugarcane expansion over a relevant Brazilian river basin that supplies electricity and water to two of the biggest cities in Brazil — São Paulo and Minas Gerais. However, the methods and numerical models described along the thesis can be used elsewhere around the world. Although it is acknowledged that a fully coupled atmospherevegetation-hydrologic model would be a more realistic approach to represent all land surface-atmosphere interactions in the local environment, to include a module of dynamic vegetation in a coupled model is beyond the scope of this thesis. Since all simulations performed using the integrated modelling system were no longer than 1 year and its inputs include monthly normalized difference vegetation index (NDVI), effects of a dynamic local vegetation on the numerical analysis were considered marginal. 8

Introduction

1.4

Thesis structure and appended papers

This thesis is based on the research presented and discussed in the 6 appended papers. After a brief description of the current status and perspectives of renewable energy sources in chapter 1, a theoretical background for the appended papers is given in chapter 2 along with references to ongoing researches. Chapter 3 describes the study area as well as the data and methods used to obtain the results presented in chapter 4. Besides the results obtained from the model runs and numerical analyses performed as described in chapter 3, chapter 4 provides an overview of the findings presented in the appended papers. Finally, chapter 5 highlights the concluding remarks together with suggestions and issues left open for future studies. While the description of the numerical models and methods used to achieve the overall objective of this thesis are only briefly presented along the summary, a more detailed account of them can be found in the appended papers referred to by bold Roman numerals. A short introduction to each of the appended papers is given here. Paper I maps the rapid sugarcane expansion which occurred in the Rio Grande basin (Brazil) between early 90’s and late 2000’s based on the spectral signature of sugarcane plantations in Landsat satellite imagery captured in 1993, 2000 and 2007. Comparisons between these three different land use scenarios indicates a relevant significant expansion of sugarcane plantations over the basin between 1993 and 2000 whereas, from 2000 to 2007, the sugarcane expansion was more moderate compared to the previous one. Paper III explores the short-, medium- and long-term impacts of expansion of the sugarcane plantation on the water balance of the Rio Grande basin, Brazil, as estimated by changes in evapotranspiration, soil moisture content and surface runoff calculated by a hydrological model. Twenty years of simulation are made using historical land use of the basin that include areas planted with sugarcane in 1993, 2000 and 2007 as estimated by Paper I. Complementary, it is used a scenario for sugarcane plantation defined by the Brazilian Institute for Agricultural Research (EMPRAPA) as all areas suitable for sugarcane cultivation within the Rio Grande basin. An assessment of numerical schemes for solving the ADE is presented in Paper IV. It reveals that high order flux-limiter schemes are more conservative than first order upwind schemes. Numerical analyses performed in Paper IV also provided a better understanding of numerical schemes employed in numerical models which was fundamental to figure out an appropriate strategy of coupling between atmospheric and hydrological models proposed in Paper V. In the study presented in Paper V, the implementation of the twoway coupling between the regional atmospheric model – BRAMS – and the hydrological model – MGB-IPH – is described. In this two-way coupling, differences in spatial scales were avoided by using the same

9

1.4 Thesis structure and appended papers

grid size in both models, in such way that neither upscaling nor downscaling were necessary. The coupled system was compared against BRAMS estimates and observed data both in Paper V and Paper VI. The process-based approach to land surface hydrological processes employed in the two-way coupled model was evaluated for three case studies, which included a cold front, deep convective clouds and stable atmospheric conditions. Thus, the study presented in Paper VI aimed at assess the importance of a conceptual representation of hydrological processes when modelling atmospheric circulation by comparing results from a regional atmospheric model that interprets land surface hydrological processes based on parameterizations with results obtained from the two-way coupled model for the three case studies previously mentioned.

10

CHAPTER

2

THEORETICAL BACKGROUND This chapter provides a theoretical background that includes current developments in mathematical and numerical modelling of the atmosphere and hydrologic cycle as basis for the discussions presented in the following chapters. Although it partially covers the state of the art, this chapter is only intended to introduce terminologies, concepts and current research goals in order to familiarise the reader with the context in which this thesis has been conducted. Additionally, this chapter presents a brief history of sugarcane including important economic and political facts that influenced sugarcane to achieve its current status of one of the major crops grown in Brazil — along with coffee, soya bean, orange and rice.

2.1

A brief history of the sugarcane in Brazil

The search for sustainable sources of energy has found a realistic replacement to fossil fuels in ethanol and methanol. Nowadays, the production of ethanol from sugarcane is among the most effective and sustainable techniques for making ethanol from food crops, particularly when compared to the production of ethanol from other commercial crops (e.g. wheat, corn and barley). The reason for this is that sugarcane grows at a faster rate than other crops [Herrera, 1999], and can be cultivated with many different farming practices, which opens up possibilities for enhancing productivity but protecting the environment[AgSri, 2012; Maraddi, 2006; Mui et al., 1996]. Moreover, according to reports issued by the Brazilian Sugarcane Industry Association (UNICA), the ratio between energy production and consumption (i.e energy consumed during its production) of sugarcane ethanol is more than four times larger than that of ethanol from wheat and nearly seven times that of corn ethanol [UNICA, 2010, 2013]. Their reports also reveal that sugarcane ethanol also features the highest level of productivity in terms of liters of fuel per hectare 11

2.2 Sugarcane and the water balance

Area harvested (ha)

10

x 10

5

0

8

x 10 10

5

1980 1990 2000 2010

Production (tons)

6

0

Figure 2.2. Expansion of harvested area and production of sugarcane from 1975 (when the Pro-Álcool program began) to 2012 [FAOSTAT, 2013].

of land required, since Brazilian sugarcane industry produces, on average, 7500 liters of ethanol per hectare whereas European sugarbeet and U.S. corn ethanol yields an average of 5500 l/ha and 3800 l/ha, respectively. In response to these properties and its high potential to become a renewable energy source, many countries have significantly increased their sugarcane production during the last two decades (e.g. China, India, Brazil). In this context, Brazil is the country that retains the largest area of sugarcane cultivation in the world. It is responsible for approximately one third of the global harvested area and production [Zuurbier and van de Vooren, 2008]. Since 1975, when the ProÁlcool program (ethanol program) was established by the Brazilian Government in response to the 1973 oil crisis [Borges and Almeida, 1985; Rosillo-Calle and Cortez, 1998], the Brazilian area of sugarcane plantation increased by 170%, reaching 5.4 million hectares in late 2005 [Bolling and Suarez, 2001; IEA, 2006; Nitsch, 1991]. Figure 2.1 illustrates three phases that characterize the last three decades of sugarcane expansion in Brazil according to F.O.Litch [2007, 2008]. The first phase represents the first decade after launching the Pro-Álcool program (i.e. between 1975 and 1986) which shows a sharp increase in areas for cultivation of sugarcane due to the domestic feedstock demand of the ethanol program. From 1986 to 2000, a second phase indicates a stagnation in ethanol production attributed to various national and international factors. Finally, between 2000 and 2008, the third phase presents the most rapid expansion of sugarcane harvested occurred after 2000, especially from 2005 to 2008. During this period, ethanol demand to substitute for gasoline consumption became a driving force at the global level [Zuurbier and van de Vooren, 2008]. From 2009 on, the Brazilian ethanol industry is facing a crisis associated with environmental and political factors, such as bad sugarcane harvest due to poor weather conditions and mandatory price controls [Jagger, 2013; The Economist, 2010]. As sugar and ethanol share the same feedstock (i.e. sugarcane) and their industrial processing is fully integrated, the ethanol crisis did not affect the sugarcane expansion (Figure 2.2, FAOSTAT [2013]) but higher sugar prices in the world market made more attractive the production of sugar instead [Angelo, 2012; Serra et al., 2011].

2.2

Sugarcane and the water balance

One of the negative environmental consequences of increasing the amount of a particular land use type may be its possible impact on regional hydrological processes [Gedney et al., 2006; Sampaio et al., 2007]. Addressing this question, many studies have recently been developed to estimate the effects of land use changes on local water balance. Hlavcova et al. [2009], for example, showed the impacts of land use changes on maximum daily discharges in a catchment con12

Theoretical background

Figure 2.1. Use of Brazilian sugarcane land for ethanol and sugar production (Extracted from F.O.Litch [2007, 2008]).

sidering both rainfall and snowmelt as incoming water. In addition, Warburton et al. [2011] have applied a hydrological model over three catchments with different land covers. They showed that the runoff generation in the three catchments were closely related to their geographic distribution of land use. Finally, a hydrologic model was also used by Wijesekara et al. [2012] as a tool to analyze the expansion of built-up areas according to land use predictions, and, by means of these analyses, they could estimate variations in surface runoff, evapotranspiration, baseflow and infiltration for an urban catchment. Despite many insightful studies on land use changes affecting the surface hydrology, large speculations are still being made about such changes. In Brazil, for instance, impacts of the rapid expansion of sugarcane on surface runoff after the Pro-Álcool were not carefully investigated since sugarcane fields were not completely mapped [Cheesman, 2004; James, 2008]. However, evidence that land use changes due to sugarcane expansion may result to relevant changes in the water balance has recently been presented by Homdee et al. [2011], who revealed significant variations in seasonal ET — specially during the dry season — after the conversion of farmland to sugarcane in Thailand. Moreover, impacts of an intensification of land use from natural grass cover to sugarcane plantations on water balance have been investigated by Schulze [2000] in South Africa. Besides higher interception and transpiration rates, many other impacts on water balance were identified as consequences of the replacement of native vegetation by sugarcane as for example, water extraction from deeper soil layers as a consequence of deeper rooting systems and enhanced infiltration rates due to tillage. 13

2.3 Numerical models, the hydrologic cycle and the atmosphere

2.3

Numerical models, the hydrologic cycle and the atmosphere

Over the two last decades, interactions associated with natural fluxes between biosphere and atmosphere have been highly discussed [Pielke et al., 1998; Sellers et al., 1997; Sutton et al., 2007]. Among the natural exchanges between land surface and atmosphere, the hydrologic cycle stands out for its complexity and its relevance to all other physical processes [Stohlgren et al., 1998]. The hydrologic cycle incorporates a wide range of processes within soil, surface and atmosphere which are closely interconnected to each other. For a better understanding of these processes and how they are interrelated, many strategies have been developed to estimate water fluxes between soil, surface and atmosphere at different spatial and temporal scales [Balsamo et al., 2009; Liang et al., 1994; Pitman, 2003; Schaake et al., 1996; Wilson et al., 2001]. In this context, numerical models are widely recommended as tools capable of quantifying, predicting and assessing the soil, surface and atmospheric water budgets [Arnold et al., 1993; Bittelli et al., 2010; Maxwell and Miller, 2005]. In general, numerical models have successfully been used for estimating the exchange of water within the hydrosphere, and among surface water, soil water and groundwater. A groundwater model, for instance, has recently been applied by Singh [2013] to evaluate the impacts of a waterlogged area on groundwater levels and recharge rates. Besides finding that a small increase in the net recharge might implicate an expansion of waterlogged areas, the study also indicates that numerical models are an effective tool for groundwater simulations. Raza et al. [2013] used a numerical model to analyse the influence of three types of land use on soil water dynamics. Results obtained from simulations were compared to field experiments and revealed that the numerical model could satisfactorily reproduce the water content in the soil profile. Regarding surface water, numerical models have systematically been developed and upgraded [Lohmann et al., 1998; Panday and Huyakorn, 2004; Todini, 1996]. Among these models, hydrological models are frequently used for a large range of applications spanning from runoff simulation to flood forecasting [Nijssen et al., 1997; Toth et al., 2000]. Although hydrological models are able to physically represent most of the water balance by integrating soil and surface water processes, they do not have any atmospheric module to deal with exchange of water between surface and atmosphere; hence, estimates of runoff depend on how dense the rainfall gauge network is [St-Hilaire et al., 2003] or on the resolution of the atmospheric model used to estimate or forecast precipitation. On the other hand, atmospheric models present a detailed and complex approach of atmospheric processes that includes estimation

14

Theoretical background

of carbon, heat, energy and water fluxes between surface and atmosphere based on energy balance. By using a suite of regional climate model (RCM) scenarios, for instance, Morales et al. [2007] estimated climate impacts on carbon cycling across Europe. Their study found that projected changes in carbon balance depended on the choice of the general circulation model (GCM) providing boundary conditions to the RCM rather than the choice of RCM. RCM simulations were also conducted by Seneviratne et al. [2002] in order to investigate thermodynamic processes such as exchange of sensible/latent heat between land surface and atmosphere under a warmer climate. Among many other findings, their results underline the importance of land surface processes in climate integrations. RCMs were also evaluated by Hagemann et al. [2004] on their ability to estimate water budgets over Europe. Their study showed that all RCMs presented either prominent summer drying or overestimation of rainfall throughout the year except during the summer. They claimed that the model deficit and systematic errors may be related to deficiencies in the land surface parametrization. Moreover, parameterization schemes usually apply prescribed values of parameters based on their probability density functions. This assumption does not consider land use and soil characteristics as continuous distributions, and hence, mixtures in soil and vegetation within an area of interest are not captured. This may lead to errors in spatial distribution when estimating land surface processes [Molders, 2000; Vidale et al., 2003]. Concurrently, an alternative approach proposes to optimize the performance of hydrological and atmospheric numerical models by coupling them into integrated modelling systems [Seuffert et al., 2002; Walko et al., 2000].

2.4

Integrated modelling systems

In an attempt to simulate surface runoff on a daily basis, Hay et al. [2002] proposed a one-way coupling of an RCM and a distributed hydrological model. Their approach suggests that outputs of precipitation and temperature from the RCM are used as input to the hydrological model. This methodology incorporates the effects of the drainage network when calculating the soil moisture content, and replaces land surface parameterizations with a process-based method to estimate runoff; though feedback effects of land surface dynamics from the hydrological model are not included in calculations of precipitation and temperature time series provided by the RCM. The need for a better representation of land surface hydrological processes and their feedback mechanisms into RCMs has been strongly suggested by several numerical studies [Baron et al., 1998; Bartholmes and Todini, 2005; Haggag et al., 2008; Lin et al., 2006; Messager et al., 2006]. In this sense, a more sophisticated approach proposed by Walko et al. [2000] presents a two-way coupling of a RCM and a hydrological model. Their coupled system includes tur15

2.4 Integrated modelling systems

bulent and radiative exchange of heat and water between soil, vegetation, canopy air and atmosphere. However, sensitivity tests of this coupled system are only performed in idealized model simulations. Moreover, Walko et al. [2000] use a parametric model developed by Louis [1979] to represent fluxes of water vapor between land surface and atmosphere. This parametric model assumes that momentum roughness length is equal to heat transfer roughness length, and the height of the lowest model level is much larger than momentum roughness length. These assumptions are well supported by experimental evidence for smooth surfaces [Högström and SmedmanHögström, 1974; Phelps and Pond, 1971], though are not valid for rough surfaces and/or mountainous regions [Kot and Song, 1998; van den Hurk and Holtslag, 1997]. A similar approach has been used by Seuffert et al. [2002] to evaluate the influence of land surface hydrology on the predicted local weather. Unlike Walko et al. [2000], their study consists of a twoway coupling between an RCM and a hydrological model through a coupling strategy that, firstly, supposes that the hydrological model estimates the turbulent diffusion coefficients of heat and momentum more realistically than the RCM, and, secondly, replaces values of albedo from the RCM with those calculated by the hydrological model. Despite this, their results reveal that the two-way coupled model improved the predicted energy fluxes and rainfall in comparison with predictions made by the RCM during a 3-day forecast period, evapotranspiration rates are not only dependent on albedo but also on leaf area index, rooting depth and bulk stomatal resistance. It may lead to an over- or underestimation of evapotranspiration rates after longer periods of simulation.

16

CHAPTER

3

MATERIAL AND METHODS The research presented in this thesis is based on: • Numerical simulations • Geographic Information System (GIS)- and satellite-based remote sensing imagery analysis • Hydrological and meteorological observations Numerical simulations were performed using several types of models, from a two-dimensional depth-averaged circulation model (Paper IV) to a two-way coupled atmospheric-hydrological system (Paper V, Paper VI and Paper VII), including the stand-alone versions of its hydrological (Paper III) and atmospheric (Paper V and Paper VI) models. Each of these numerical simulations were carried out using different methods and for different purposes, which are outlined as follows. One of the primary purposes of this thesis was to build a numerical tool capable of evaluating impacts of land use changes on the atmosphere and local hydrology considering their feedback loops. To achieve this primary purpose, practical knowledge on advanced numerical schemes had to be acquired. Therefore, studies described in Paper IV — which assess different numerical schemes employed by Pereira [2010] to solve the Advection-Diffusion Equation (ADE) — were used to provide insights on the implementation of the two-way coupled atmospheric-hydrological system presented in Paper V. Paper III, Paper VI and Paper VII also include numerical simulations. The purpose of the numerical simulations made in Paper III was to investigate short-, medium- and long-term impacts of the sugarcane expansion quantified by Paper I on the water balance at a basin scale. On the other hand, numerical simulations described in Paper VI were purely used to demonstrate the role of a process-based representation of land surface hydrological processes — given by the inte17

3.1 Study area

grated modelling system (Paper V) — to the atmosphere. Finally, numerical simulations in Paper VII were performed to assess whether the coupled model behavior is in accordance with current understanding of the hydrological cycle of the Rio Grande basin under land use changes due to sugarcane expansion over its drainage area. Along with hydrological and meteorological time-series, GIS- and satellite-based remote sensing imagery were used to support the results and findings obtained in Paper V, Paper VI and Paper III. While GIS-based remote sensing was exhaustively employed to generate land use scenarios, to pre/post-process outputs and to evaluate the reliability of results in Paper III and Paper IV, satellite-based remote sensing was exclusively applied for comparisons between simulations and the real state of the atmosphere (Paper V and Paper VI). A description of the study area as well as numerical models and methodologies employed in this thesis are given below.

3.1

Study area

The Rio Grande basin is a sub-basin of the Paraná river basin formed by the rivers Grande, Pardo, Sapucaí, Verde, das Mortes and MogiGuaçu. It has an area of 145000 k m 2 located in the eastern upper Paraná basin (Fig. 3.1) where altitudes vary from 300 to 2700 m .a .s .l .. The classification of land use in the Rio Grande basin includes three distinct categories: Atlantic Rainforest, pasture and agriculture [IBGE, 1991]. Agricultural activity represents a large portion of the Rio Grande basin and, it was classified into sugarcane and agriculture of grain. Regarding types of soils, Rio Grande basin presents five major types: latosols, lithosols, cambisols, podzolics and alluvial soils which may be broken down into three groups: high, medium and low infiltration capacity [FAURGS, 2007]. Soils with high and medium infiltration capacity are equally distributed across the Rio Grande basin whereas soils with low infiltration capacity are concentrated along the drainage network. Although most of surface runoff in the Rio Grande basin is regulated by dams, its hydrological regime is strongly induced by land use changes due to harvesting practices and shifting cultivation [WWFBrasil, 2008]. After the flow regulation, a representative sample of daily values of discharge collected at the outlet of the basin, from 1970 to 2010, indicates that surface runoff varies from minimum values of 1000 m 3 /s (dry season) to maximum values over 12000 m 3 /s (rainy season). Locally, measurements of runoff are also monitored at hydroelectric power plants. At Funil, Camargos, Furnas, P Colômbia, Marimbondo and A Vermelha power plants, daily runoff ranges 70 – 3731 m 3 /s , 34 – 1253 m 3 /s , 174 – 7497 m 3 /s , 251 – 8367 m 3 /s , 532 – 9234 m 3 /s and 303 – 10186 m 3 /s , respectively. According to Espinosa [2011], spatial and temporal distribution of 18

Material and methods

Hydroelectric power plant Funil Camargos Furnas P Colômbia Marimbondo A Vermelha

(m 3 /s ) 34.0 34.0 174.0 251.0 1100.0 1600.0

Table 3.1. Minimum operating flow at the hydroelectric power plants used in this study [ONS, 2013].

Figure 3.1. Location and elevation map of the Rio Grande basin.

rainfall in the Rio Grande basin is highly induced by synoptic systems over the southeastern and south-central Brazil. In addition, annual rainfall analysis carried out by CPRM [2012] indicate that annual average rainfall varies from 1500 to 2000 m m in the basin. Annual average evapotranspiration ranges from 800 to 1000 m m [Ruhoff, 2011]. Throughout the year, a seasonal variability of evapotranspiration has been identified by Rocha et al. [2002]. Over the Rio Grande basin, their studies revealed that daily evapotranspiration can oscillate between 6 m m /d −1 in the rainy season and 1 m m /d −1 in the dry season. Production of electrical power is the largest water use in the Rio Grande basin [IPT, 2008]. Over 11% of the installed electric generation capacity of Brazil is at hydroelectric installations in the Rio Grande basin [ANEEL, 2005]. To meet this demand for electricity, hydroelectric power plants are constrained by a minimum operating flow, which varies from power plant to power plant. As recently proposed by [ONS, 2013], the minimum operating flow at all hydroelectric power plants used in this study are shown in table 3.1. The gauging stations located at the outlet of these hydroelectric 19

3.2 Distributed hydrological model

Figure 3.2.

The location of the six discharge stations and their respective

sub-basins.

power plants are continuous stream flow gages with valuable long periods of record, spanning from the early 30’s until present time. Unlike many other gages lost due to discontinued funding, these six gages were/are maintained and operated by the energy industry. Owing to their data consistency and data reliability, the Rio Grande basin has been divided into six smaller sub-basins in a way that each of the six gages corresponds to the outlet of a sub-basin. Figure 3.2 shows the location of the six discharge stations along with their respective subbasins.

3.2

Distributed hydrological model

MGB-IPH is a large scale distributed hydrological model [Collischonn, 2001] conceptually based on the LARSIM [Bremicker, 1998] and VIC2L [Liang et al., 1994] models. It consists of modules for calculating soil water budget, evapotranspiration estimation, surface and subsurface flow generation, which are interconnected by river routing. Full details of all hydrological processes and how these processes are incorporated into MGB-IPH can be found in Collischonn et al. [2007]. MGB-IPH provides the choice of smaller sub-basins [Paiva et al., 2011] or uniform grid cells [Paz et al., 2011] as computational units. For both methods, MGB-IPH divides each computational cell in hydrologic response units (HRUs) based on its land use/cover and soil distribution. HRUs are then defined by intersecting land use and soil 20

Material and methods

groups within a computational cell. Once all computational cells are classified into different groups with similar hydrological response, MGB-IPH calculates the soil water budget, evapotranspiration and flow propagation using adapted versions of the ones presented in LARSIM and VIC-2L models. These adaptations were made in order to facilitate its applications in large tropical basins. MGB-IPH generates surface flow by direct precipitation on saturated areas and subsurface flow comes from the non-linear relationship between texture, hydraulic conductivity and moisture of soil proposed by Rawls et al. [1993]. A linear reservoir concept is used to propagate surface and subsurface flow using different retention times along every computational cell. After passing through the linear reservoirs, surface and subsurface flows are summed and routed from cell to cell along the river network using the Muskingum-Cunge method. MGB-IPH has been tested and used in several South American basins, from rapid-response ones of southern Brazil and Uruguay to very low response ones such as the Pantanal. It has also been applied for several purposes, such as to predict runoff [Tucci et al., 2008], to estimate daily water balance in large basins [Collischonn et al., 2008] and to analyse the impacts of climate change upon river flow [Paiva and Collischonn, 2010]. By default, MGB-IPH is employed using a daily time step. However, its time step may fluctuate depending on the purpose of study. In this research, MGB-IPH was used to simulate rainfall-runoff processes on a daily basis.

3.3

Regional atmospheric model

BRAMS is part of the operational weather prediction system of the Centro de Previsão de Tempo e Estudos Climáticos (CPTEC), belonging to the Instituto National de Pesquisas Espaciais (INPE) in Brazil (http://brams.cptec.inpe.br). The BRAMS is a multipurpose, numerical prediction model designed to simulate atmospheric circulations spanning from hemispheric scales down to large eddy simulations (LES) of the planetary boundary layer ([Walko et al., 2000]; www.atmet.com). The model is equipped with a multiple grid nesting scheme which allows the model equations to be solved simultaneously on any number of interacting computational meshes of differing spatial resolution. It has a complex set of packages to simulate processes such as radiative transfer, surface-air water, heat and momentum exchanges, turbulent planetary boundary layer transport, and cloud microphysics. The initial conditions can be defined from various observational data sets that can be combined and processed with a meso-scale isentropic data analysis package. For the boundary conditions, a fourdimensional data assimilation technique described by Umeda and 21

3.4 Two-way coupling methodology

Martien [2002] is used to interpret atmospheric boundary conditions provided every six hours by global atmospheric analyses. BRAMS features used in this system include an ensemble version of a deep and shallow cumulus scheme based on the mass flux approach and soil moisture initialization data. The surface-atmosphere water, momentum and energy exchanges are simulated by the Land Ecosystem Atmosphere Feedback model (LEAF), which represents the storage and vertical exchange of water and energy in multiple soil layers, including the effects of freezing and thawing soil, temporary surface water or snow cover, vegetation, and canopy air [Lee and Pielke, 1992]. In order to merge the capabilities of several numerical weather codes, BRAMS was implemented using the concept of “plugcompatible” modules given by Pielke and Arritt [1984]. Although this concept allows the easy incorporation of improvements between the sub-routines of the model, it also stimulates the use of parameterizations by the developers and users of the model. To represent surface layer fluxes of water vapor into the atmosphere, LEAF-3 uses a parametric model developed by Louis [1979]. Similarly to any other trace gas, such as ozone (O3 ) and carbon dioxide (CO2 ), his parameterization scheme estimates the fluxes of water vapor using Businger’s profile functions [Businger et al., 1971]. Once computed, the fluxes of water vapor are interpreted by BRAMS as the lower boundary for the atmosphere.

3.4

Two-way coupling methodology

The atmospheric-hydrological modelling system is composed of a hydrological model two-way coupled to a regional atmospheric model. In this coupled modelling system, the hydrological Model for Large Basins developed at the Institute for Hydraulics Research (MGBIPH, Collischonn [2001]) and the atmospheric Brazilian Regional Modelling System (BRAMS, Freitas et al. [2009]) were employed. Besides the experience the authors had with them, the choice of these models was also based on their successful use in previous studies in the Rio Grande Basin [Bender and de Freitas, 2013; Nóbrega et al., 2011]. In this modelling system, BRAMS and MGB-IPH were coupled in a way to best exploit their respective strengths. Here, process-based estimates of the fluxes of water vapour by MGB-IPH replaced those estimated by BRAMS using parameterization schemes while daily rainfall calculated by BRAMS was provided as input to MGB-IPH. A brief description of the coupling strategy employed in this atmospherichydrological modelling system is given as follows. Process-based estimates of evapotranspiration rates by MGB-IPH replace fluxes of water, from canopy air to atmosphere, calculated using land surface parameterizations in the LEAF-3 routines of BRAMS as shown in Figure 3.3. Figure 3.3 shows a schematic of components 22

Material and methods

Figure 3.3. A schematic of components of the LEAF-3 and their interactions with MGB-IPH over an entire column (Adapted from Walko et al. [2000]).

of the LEAF-3 routine and their interactions with MGB-IPH, where atmosphere (A), vegetation cover (V), canopy air (C) and two soil layers (G1 and G2) are divided into multiple vertical layers. Moreover, vertical fluxes of heat, water and long wave radiation are represented by their subscripts h, w and r as well as the source and receptor (g for ground, s for snow, v for vegetation, c for canopy air, and a for free atmosphere). Therefore, evapotranspiration rates from MGB-IPH are equivalent to WCA (Water between Canopy air and Atmosphere) in the LEAF-3. On the other hand, daily accumulated rainfall estimated by BRAMS is provided as input to MGB-IPH. However, unlike the standalone version of BRAMS, the flux of water from canopy air to atmosphere incorporates feedback from a process-based approach to land surface hydrological processes given by MGB-IPH into calculations of daily accumulated rainfall. In this process-based approach to land 23

3.4.1 Strategy for solving temporal mismatches

Figure 3.4. Temporal coupling of BRAMS and MGB-IPH. As MGB-IPH runs on a daily basis whereas BRAMS uses shorter time steps, coupling variables are exchanged every 24 hours.

surface hydrological processes, MGB-IPH calculates fluxes of water from land surface to atmosphere based on air temperature, relative humidity, long wave radiation and wind speed, and hence, it is considered more accurate and comprehensive than land surface parameterizations used by LEAF-3. BRAMS and MGB-IPH were spatially and temporally coupled using a coupling strategy that avoids upscaling and downscaling issues by running both models on the same grid size. Since BRAMS and MGB-IPH output variables are calculated at their own computational cell centers that do not always match each other, the coupling strategy counted on an algorithm to identify the nearest computational cell for the exchange of variables between BRAMS and MGB-IPH which is described in the section 3.4.2. Due to numerical stability constraints, BRAMS runs with a smaller time step than the daily time step often used by MGB-IPH at basin scale. Therefore, a temporal coupling between BRAMS and MGB-IPH is also included in the coupling strategy, so that MGB-IPH is employed as a subroutine of BRAMS which is called every 24 simulation-hours as presented in the section 3.4.1.

3.4.1

Strategy for solving temporal mismatches

The two-way exchange of variables between BRAMS and MGB-IPH presents a temporal coupling in a way that MGB-IPH is employed as a subroutine of BRAMS which is called every 24 simulation-hours as depicted in Figure 3.4. Specifically, at this time step, calculations of flux of water between canopy air and atmosphere in LEAF-3 are switched off. Complementarily, observed daily rainfall was not given as input to MGB-IPH at any time step. Thus, errors in energy and water balance computations owing to modifications arising from the two-way coupling methodology are not incorporated into the coupled model. 24

Material and methods

Figure 3.5. Scheme of the two-way exchange of variables between BRAMS and MGB-IPH computational cell centers. An algorithm selects the nearest computational cell centers before exchanging the coupling variables.

3.4.2

Strategy for solving spatial mismatches

Although the two-way coupling methodology avoids upscaling and downscaling issues by running both models on the same grid cell size, BRAMS and MGB-IPH output variables are calculated at their own computational cell centers that do not always match each other. In order to correctly address the two-way exchange of variables between BRAMS and MGB-IPH, an algorithm has been developed to calculate and rank the distances between BRAMS and MGB-IPH computational cell centers. It also sorts these distances in ascending order, and identifies the nearest computational cell centers for the two-way exchange of variables. Figure 3.5 illustrates this two-way exchange according to the procedure carried out by the algorithm. 25

3.5 Data sets

3.5

Data sets

Usually, running numerical models requires several data sets which are collected from one or multiple databases. In the research presented in this thesis, three types of numerical models were applied to two study areas. While Paper IV presents the application of a circulation model to a particular estuary in southern Brazil, Paper V, Paper VI and Paper III apply atmospheric and hydrological models to the Rio Grande basin in southeastern Brazil. Each of these studies counted on their own data sets, which are not necessarily exclusive of each other — as daily precipitation and satellite imagery, for example. The section 3.5.1 presents how the data collection process was conducted in this thesis.

3.5.1

Data collection

All data sets employed in this research were freely downloaded from the Agência Nacional de Águas (ANA), INPE/CPTEC, Empresa Brasileira de Pesquisa Agropecuária (EMBRAPA), Instituto Nacional de Meteorologia (INMET), ONS, DE/UFRGS, RADAM, NOAA and EUMETSAT databases. Table 3.2 summarizes the data sets used in this research according to data provider, data period, types of data and their utility, and papers.

3.5.2

Data preprocessing

Some of the data sets presented in table 3.2 had to be preprocessed before they were used as input. This section is then intended to describe this very important step in the data mining process as well as to provide supplementary information to the section 3.5.1. The necessary digital elevation model (DEM) was freely obtained from the Department of Ecology of the Federal University of Rio Grande do Sul. Their DEM preprocessing includes data gap filling and mosaicking of Shuttle Radar Topography Mission (SRTM) database, not only, for the Rio Grande basin but also for all Brazilian territory [Hasenack et al., 2010]. The soil map of the Rio Grande basin was derived from a soil survey data created by RADAM Brasil project [RADAMBRASIL, 1982] at scale of 1:1000000. Although at a coarser scale than RADAM Brasil soil survey data, digitalized soil maps (1:3000000) from FAO [1974] were resampled and used to overcome missing data. The RADAM Brasil database includes over 12 different types of soils in the Rio Grande basin [FAURGS, 2007] which were reclassified into two groups as deep and shallow soils according to their hydrological behavior and provided to be used in the research described in this thesis (A. R. Paz, personal communication, 2012). 26

Material and methods

Table 3.2: Summary of the data sets employed in the research presented in this thesis. Data Data provider period (Institution or Organisation) INMET 1991

Types of data and their utility

Included in papers

Hourly wind speed and direction data for flow simulations Daily precipitation, water depth and runoff observations for calibration and validation of parameters as well as for comparisons with simulated values Meteorological time-series of air temperature, sunshine, relative humidity, wind speed and atmospheric pressure; monthly NDVI and sea surface temperature; maps of soil moisture distribution and soil texture classes; and reanalysis data were used for atmospheric simulations Maps of areas suitable for growing sugarcane over the State of São Paulo and Minas Gerais employed to generate land use scenarios Daily runoff at all hydroelectric power plants in the Rio Grande basin for trend analysis, calibration and validation of parameters Digital Elevation Model (DEM) for catchment delineation

Paper IV

ANA

1930 – 2012

INPE/CPTEC

2009 – 2010

EMBRAPA

–

ONS

1930 – 2012

DE/UFRGS

–

RADAM Brasil Project

–

Maps of soil for hydrological modelling

NOAA and EUMETSAT

–

Geostationary satellite imagery were used for comparisons with the real state of the atmosphere

Paper II, Paper IV, Paper VI and Paper III Paper V, Paper VI and Paper VII

Paper II and Paper III

Paper II, Paper III and Paper VII Paper II, Paper V, Paper VI and Paper III Paper II and Paper III Paper V and Paper VI 27

3.6 Numerical experiments and imagery analyses

Observed daily precipitation values were obtained from 483 precipitation stations over the Rio Grande basin and its surroundings. Daily rainfall depths were then spatially interpolated by the inverse distance-squared weighted method at the centroid of each MGB computational cell.

3.6

Numerical experiments and imagery analyses

Besides the numerical analyses carried out in Paper V, Paper VI and Paper III, this section also outlines criteria for selection of case studies evaluated in Paper VI; mapping of sugarcane plantations; and calibration and validation of parameters for sugarcane made in Paper III; and models runs and performed in Paper V, Paper VI and Paper III along with their assumptions.

3.6.1

Mapping of sugarcane plantations

Multi-temporal Landsat images were used for the characterization of sugarcane evolution in the Rio Grande basin. Land use maps were derived through analysis of satellite images made by Landsat TM 7 and extracted from the U.S. Geological Survey (USGS). The selection of satellite images was driven by the availability of cloud-free Landsat data over the Rio Grande basin from 1970 to 2010. Fourteen Landsat satellite images (170x183km) were captured in 1993, 2000 and 2007, and used to generate three land use maps. An automatic classification of Landsat satellite images showed in Rudorff et al. [2010] was used for mapping sugarcane fields in Paper I. This automatic classification is based on a linear spectral mixing which consists of a linear combination of spectral signatures from two or more types of land use, such as agriculture, pasture, forest etc. The particular sugarcane spectral signature as presented by Aguiar et al. [2011] has been used to identify sugarcane plantations in the Rio Grande basin. Moreover, a visual inspection was made to support this automatic land use classification. Figure 3.6 shows the results the automatic identification of areas of sugarcane plantation in 1993, 2000 and 2007. In addition, it is shown the suitable areas for sugarcane plantation in the basin as defined by EMBRAPA [BRASIL, 2009]. Each land use map was classified into five dominant soil use types: areas covered by water bodies, Atlantic Rainforest, agriculture of grain crops, pasture lands and sugarcane plantations according to their spectral signatures. Except for sugarcane plantations, all spectral signatures were adopted as defined by Mendes and Cirilo [2001] and only mapped for 1993. Although it may appear as a limitation, BRASIL [2009] and [FAURGS, 2007] revealed that land use changes due to shrinkage/expansion of areas covered by pasture, cereals or forest are 28

Material and methods

Figure 3.6. Land use distribution of each sub-basin of the Rio Grande basin for all four land use scenarios used in this study: 1993, 2000, 2007 and EMBRAPA. Colors represent different soil use and each bar represents 100% of the sub-basin area.

marginal compared to sugarcane expansion in the Rio Grande basin over the past two decades.

3.6.2

MGB-IPH specific parameters for sugarcane: calibration and validation

Table 3.3 shows fixed and adjustable parameters adopted in Paper III. Sugarcane planting and harvesting timing were defined using analysis made in BRASIL [2009] for sugarcane plantations located in the Southeast Region of Brazil. The adjustable parameters for sugarcane were estimated via calibration. Although disposal measured data spans a period of 40 years, only the 20 most recent years are used for calibration and validation. The calibration was performed for a eleven-year period (1990-2000), and consisted of fine-tuning the adjustable parameters by comparing calculated and observed discharges using relative volume error (RVE), Nash-Sutcliffe coefficient (NS) and root-mean-square error (RMSE) as efficiency criteria which are given by the following equations: P Q s i m −Q ob s P × 100 (3.1) RV E (%) = Q ob s P (Q ob s −Q s i m )2 NS = 1 − P € (3.2) Š2 Qob s −Qob s 29

3.6.2 MGB-IPH specific parameters for sugarcane: calibration and validation

Parameter Albedo Leaf Area Index (m2 /m2 ) Height of trees (m) Albedo Leaf Area Index (m2 /m2 ) Height of trees (m) Albedo Leaf Area Index (m2 /m2 ) Height of trees (m) Albedo Leaf area index (m2 /m2 ) Height of trees (m) Parameter Maximum water storage Mean percolation Residual water storage Maximum water storage Mean percolation Residual water storage Maximum water storage Mean percolation Residual water storage Maximum water storage Mean percolation Residual water storage Mean groundwater flow Upward flux of water Shape parameter Hydraulic conductivity

Jan 0.13 4.00 1.00 0.20 2.00 0.50 0.11 8.00 9.00 0.28 7.00 3.60

Fixed Parameters Feb Mar Apr May 0.13 0.13 0.13 0.16 4.00 4.00 5.00 1.00 1.00 1.00 1.00 0.50 0.20 0.20 0.21 0.21 2.00 2.00 3.00 2.00 0.50 0.50 0.50 0.50 0.11 0.11 0.11 0.11 8.00 8.00 8.00 8.00 9.00 9.00 9.00 9.00 0.28 0.29 0.31 0.31 7.00 8.00 9.00 9.00 3.60 3.80 3.80 3.80 Adjustable Parameters Unit mm mm d−1 mm mm mm d−1 mm mm mm d−1 mm mm mm d−1 mm mm d−1 mm d−1 mm d−1 Jun 0.16 1.00 0.80 0.21 2.00 0.50 0.11 8.00 9.00 0.24 3.00 0.50

Jul 0.17 2.00 0.80 0.21 2.00 0.50 0.11 8.00 9.00 0.25 5.00 1.20 Value 625.0 3.5 62.5 446.0 2.1 44.6 711.0 6.2 71.1 654.0 3.9 65.4 146.0 0.0 0.10 2268.0

Aug 0.17 2.00 0.80 0.21 2.00 0.50 0.11 8.00 9.00 0.25 5.00 1.20

Sep 0.16 2.00 0.80 0.21 2.00 0.50 0.11 8.00 9.00 0.25 5.00 1.20

Oct 0.15 2.00 0.90 0.20 2.00 0.50 0.11 8.00 9.00 0.27 6.00 2.80

Nov 0.14 3.00 0.90 0.20 2.00 0.50 0.11 8.00 9.00 0.27 6.00 2.80

Dec 0.13 3.00 0.90 0.20 2.00 0.50 0.11 8.00 9.00 0.27 6.00 2.80

Table 3.3: Fixed and adjustable parameters used (or assumed) in this study. The set of fixed and adjustable parameters for agriculture of grain crops, pasture lands and Atlantic Rainforest were assumed as defined via calibration and validation by Nóbrega et al. [2011] in the Rio Grande basin. On the other hand, for sugarcane fields, fixed parameters were adopted according to ranges obtained in the literature whereas adjustable parameters were calibrated and validated in this study.

of

Type of Land Use Agriculture grain crops Pasture lands Atlantic Rainforest Sugarcane fields

of

Type of Land Use Agriculture grain crops Pasture lands Atlantic Rainforest Sugarcane fields

Same for all types of land use

30

Material and methods

Table 3.4: Summary of results of the calibration and validation of parameters for sugarcane in the Rio Grande basin (see Paper III for location of each gauging station in the basin). Gauging station P Colômbia Marimbondo A Vermelha Gauging station P Colômbia Marimbondo A Vermelha

Calibration NS RMSE (m3 s−1 ) 0.92 270.97 0.92 370.40 0.96 130.50 Validation NS RMSE (m3 s−1 ) 0.88 301.14 0.87 436.10 0.85 508.92

€ Š RMSE m 3 s −1 =

rP

RVE(%) 6.79 -8.62 3.12 RVE(%) 12.30 12.31 13.27

(Qob s −Q s i m )2 n

(3.3)

where Q s i m and Q ob s are simulated and observed runoff data and n is their length. The set of the adjustable parameters for sugarcane defined during the calibration were validated over the seven-year period 2001 – 2007. Since the adjustable parameters were individually calibrated for each sub-basin, the calibration is made by sub-basin from the upper to the lower basin. When one sub-basin is calibrated, MGB-IPH runs in simulation mode for all other sub-basins. As performance criteria for this calibration, table 3.4 presents NS coefficients, RMSEs and RVEs computed for each sub-basin. Results from the calibration revealed that MGB-IPH could reproduce very well the hydrological regime of all sub-basins that presented areas covered by sugarcane fields over the 10-year calibration period. Although baseflow recessions were slightly overestimated by MGB-IPH, peak flows and rising and falling limbs of the simulated hydrographs closely matched the observed hydrographs (Figure 4 in Paper III), which may be noticed by minor RMSEs and RVEs. Morevover, NS coefficient values of up to 0.9 were obtained for all subbasins which indicated a good agreement between observed and simulated discharges (Table 3 in Paper III). In order to validate the parameters previously calibrated, MGBIPH was applied to the Rio Grande basin using a different meteorological data set, which spans from 01 January 2001 through 31 December 2007. Again, RVE, NS coefficient and RMSE were used to measure the quality of the fitting (see table 3.4).

3.6.3

Models runs and their assumptions

Several model runs were performed along the research presented in this thesis. As runs had different purposes, particular assumptions 31

3.6.3 Models runs and their assumptions

were made for each of them. These assumptions basically include length of warming-up periods — periods often used in simulations to let physical parameters reach realistic conditions —, the frequency which outputs are produced and the time of year when most of average annual rainfall occurs over the Rio Grande basin. In total, a set of 10 model runs were conducted in numerical experiments proposed in Paper III, Paper IV, Paper V and Paper VI using four numerical models. Initially, Paper III analysed the results obtained from four model runs, which employed the hydrological model MGB-IPH with the four land use scenarios generated in Paper I as the only difference among each runs. Another two model runs used a 2D circulation model (IPH-UnTRIM2D) to evaluate two different numerical schemes to solve the ADE in Paper IV. Lastly, four more model runs are presented in Paper V and Paper VI, which include simulations using the stand-alone version of the regional atmospheric model (BRAMS) and its version coupled with MGB-IPH. In Paper III, four land use scenarios were used as input to MGBIPH to perform four different model runs with daily time step covering the period of January 1s t 1990 to December 31s t 2010. All runs were preceded by a warming-up period of one year (January 1989 December 1989). The run, which incorporated the land use map of 1993, was considered as the control run – since it is the closest scenario to the beginning of the sugarcane expansion – and together with the runs that included land use scenarios of 2000, 2007 and from the EMBRAPA mapping will be respectively called CR1993, R2000, R2007 and REMBRAPA hereafter. Paper III evaluated short-term impacts of the sugarcane expansion using the sets of simulated daily runoff in the 4 model runs previously described. Three data sets were then generated from percentage differences in daily runoff between the scenarios of expansion (i.e. R2000, R2007 and EMBRAPA) and the CR1993 one. As each run was performed over a simulation period of 20 years, each of these sets corresponds to 7300 daily runoff differences. The statistical significance of these percentage differences were tested by means of bootstrapping, using 1000 random samples, for a significance level of 0.01 per sub-basin. In a medium-term temporal horizon, Paper III estimated the impacts of sugarcane expansion on water balance as a percentage differences in annual runoff, evapotranspiration and soil water content over 20 years of simulation. Annual runoff, evapotranspiration and soil water content were accumulated from daily values calculated in CR1993, R200, R2007 and REMBRAPA over the annual phenological cycle of sugarcane, so that from June to May. And finally, Paper III also assessed long-term changes in the hydrological regime under sugarcane expansion as cumulative differences in runoff, evapotranspiration and soil water content between the control run CR1993 and the scenarios of sugarcane expansion (i.e. R2000, R2007 and REMBRAPA) at the end of 20 years of simulation. 32

Material and methods

The two model runs of 2D circulation model performed in Paper IV were used for comparisons between mass balance computations made by numerical schemes with terms of high order. The model was applied to an estuary, located in southern Brazil, for two release scenarios under the same initial conditions. While the first scenario reproduces a deliberately release of a tracer at a steady rate of 5 m g L 1 into the estuary over 15 days, the second scenario represents a release of a tracer that remains at a constant rate of 5 m g L 1 only over the first 10 hours of simulation. As initial condition, the estuary is assumed spatially homogeneous and well-mixed with a tracer concentration equal to 1 m g L 1 . In Paper V, two different runs were carried out for a simulation period of 31 days. While the first run applied the stand-alone version of BRAMS to the Rio Grande basin, the second run employed the atmospheric-hydrological modeling system instead. By means of comparisons between results obtained from these runs, Paper V evaluated the capability of the coupled model to reproduce rainfall occurrence. Since austral summer is the rainy season in the Rio Grande basin [Nóbrega et al., 2011], the runs spanned 1s t through 31s t January 2009. Similarly to Paper V, Paper VI analyses results obtained from two short-term runs, which used BRAMS and the coupled model. Its analyses focused on the formation of clouds and rainfall as a result of convergence zones at the surface layer. Since Paper VI carefully investigated the influence of land surface hydrological processes on the atmosphere for atmospheric processes that occur at different time scales, the interval between outputs was assumed equal to three hours for both runs — usually, for this type of analyses, atmospheric simulations produce outputs every six hours [Baron et al., 1998; Onogi et al., 2005]. Also, as satellite imagery was captured every three hours, this is the shortest interval which could be adopted for comparisons with observed data. In Paper VI, outputs from both the atmospheric-hydrological modelling system and the regional atmospheric model were compared to each other and assessed towards the formation of clouds and rainfall derived from convergence at the surface layer. Here, significant atmospheric activities were selected according to the following procedure. Firstly, convergences at the surface layer (1000 hPa ) were identified by plotting zonal and meridional wind fields. Convergence was used as an indicator of atmospheric activities in the upper layers. Secondly, vertical motions higher than 1 m /s at 500 hPa over the convergence zones pointed to possible convection development. Finally, convection occurrence was validated by the visible satellite images. According to this procedure, three atmospheric characteristics, observed on 21 January at 2100 UTC, 22 January at 1800 UTC and 25 January at 2100 UTC were selected representing a cold front passage, local strong convections and stable atmospheric conditions. 33

3.7 Evaluation of the model behaviour

Thereafter, the atmospheric-hydrological modelling system was individually and spatially evaluated by analysing the performance of each model in reproducing each study case with respect to air flow, relative humidity, vertical motion, land surface temperature and instantaneous rainfall fields (see Paper V and Paper VI). Since brightness of clouds in visible satellite images is related to their top heights [Song et al., 2004], visible satellite images were used as an indicator of vertical motion upwards by the presence/absence of cumulus clouds and their reflectance. Additionally, a rainfall gauge network composed of 136 stations well-distributed across the Rio Grande basin was used for comparisons with simulated rainfall fields. Additionally, prior to each run, a warning-up period of 10 days was considered for the initialization of the physical variables [Benoit et al., 2000] in the model runs carried out in Paper V and Paper VI. Thus, all comparisons shown in Paper V and Paper VI refer to results obtained from the last 20 days of simulation, which means, between 11t h to 31s t January 2009.

3.7

Evaluation of the model behaviour

To evaluate the reciprocal actions of land use changes, hydrology and atmosphere as estimated by the hybrid coupled atmospherichydrological model, seven key response variables strongly related to the fluxes of water and heat across the atmosphere-land surface interface are monitored over one year of simulation in Paper VII. In the near-surface soil layer, the water balance is a function of the amount of water that remains in that layer and the amount of water that passes through the layer contributing to the deep drainage (i.e. horizontal fluxes of water below the root zone). Further, both of them are strongly associated with the exchange of water between soil and land surface as the soil becomes saturated or not. Therefore, soil water content and water contribution to deep drainage are chosen as key response variables. Evapotranspiration, runoff and temperature are assumed as key response variables to fluxes of water and heat in the land surface. On the other hand, precipitation and vertically-integrated water mixing ratio in the atmospheric column from the land surface to the upper troposphere are the key response variables in the atmosphere. Besides their strong relationship with fluxes of water and heat across atmosphere, land surface and soil, the choice of the key response variables also include their high sensitivity to environmental disturbance (e.g. pollution and land use changes) as shown in previous studies [Foley et al., 2005; Menut et al., 2013; Verburg et al., 2002]. To assess the model behaviour in reproducing the exchanges of water and heat between soil, land surface and atmosphere under land use changes, the hybrid coupled model is run four times, each time using different land use scenarios in the Rio Grande basin as gener34

Material and methods

ated in Paper III. These runs were made over a 1 year period, from 1s t Jan 2009 to 31s t Dec 2009. Despite the hybrid coupled model using a time step of 10 seconds, outputs are produced every three hours and aggregated on a monthly basis. This is because biophysical parameters such as leaf area index and canopy or surface resistance – resistance of vapour flow through a transpiring crop – for sugarcane plantations only present relevant changes over a month or so [André et al., 2010; Robertson et al., 1999]. All the results obtained from this set of model runs are described and discussed in the chapter 4.

35

CHAPTER

4

RESULTS AND DISCUSSIONS This chapter combines the results obtained from the model runs performed as indicated in Section 3.6.3 with an overview of the sugarcane expansion over the Rio Grande basin as estimated in Paper I and Paper II and its short-, medium- and long-impacts on the local hydrology as defined in Paper III. Additionally, it also discusses the performance of the atmospheric-hydrological modelling system implemented and developed in Paper V by means of comparisons of BRAMS results and satellite images as made in Paper V and Paper VI. Towards the end, Paper VII presents and discusses a conceptual evaluation of the fluxes of water across the soil-land-surface and land surfaceatmosphere interfaces as given by the coupled model. As an integrated analysis of the interplay between land use, hydrology and the atmosphere is the final goal of the research described in this thesis, this chapter outlines the results and discussions in the following logical order: • Overview of the sugarcane expansion over the Rio Grande basin from 1993 to 2007 and EMBRAPA (Paper I), • Trend analysis in runoff time-series — Are they stationary? (Paper III), • Short-, medium- and long-impacts of the rapid sugarcane expansion on local hydrology (Paper III), • Assessment of the atmospheric-hydrological modelling system (Paper V), • Effects of a process-based approach land surface hydrological processes to regional patterns of air circulation and rainfall (Paper VI), • Conceptual evaluation of the exchange of water between soil, land surface and atmosphere (Paper VII), 37

4.1 Overview of the sugarcane expansion over the Rio Grande basin

4.1

Overview of the sugarcane expansion over the Rio Grande basin

From 1993 to 2007, Paper I and Paper III reveal very little or no sugarcane plantations over Funil, Camargos and Furnas sub-basins (see 3.2 for location of each of these sub-basins). Characterized by high elevations, these sub-basins present low temperatures that may reach 8◦ C during the winter. Under such climate conditions, sugarcane productivity would negatively be affected by low temperatures, which induce damage to young leaves and lateral buds. This makes Funil, Camargos and Furnas less attractive to grow sugarcane. On the other hand, further downstream, areas covered by sugarcane represent up to 27.9% of the Marimbondo sub-basin already in 1993. In addition, areas for growing sugarcane have more than tripled over 14 years (e.g. A Vermelha). This sugarcane expansion has basically been observed in P Colômbia, Marimbondo and A Vermelha over areas of flat land at low elevations. These results are in accordance with what has been suggested by EMBRAPA as areas potentially suitable for cultivation of sugarcane in the Rio Grande basin (Figure 3.6). A chronological analysis indicates different rates of sugarcane expansion for P Colômbia, Marimbondo and A Vermelha sub-basins. Between 1993 and 2000, for example, P Colômbia presented an increase of 9.8% in sugarcane plantation area. During the same period, Marimbondo and A Vermelha showed an expansion of only 3.2% and 2.9%, respectively. In contrast, from 2000 to 2007, a higher sugarcane expansion was observed over Marimbondo and A Vermelha than P Colômbia. While Marimbondo and A Vermelha pointed to an increase of 10.9% and 17.8% in areas covered by sugarcane, the expansion over P Colômbia corresponded to 5.1% (Paper I). Overall, sugarcane plantations replaced mostly pasture lands and areas of agriculture of grain. Comparisons made between land use distribution in 2007 and 1993 showed that the replacement of pasture lands by sugarcane fields achieved 6.8%, 7.5% and 8.9% of the Marimbondo, P Colômbia and A Vermelha sub-basins, respectively. It is followed by the replacement of areas of agriculture of grain crops with 5.2%, 4.7% and 7.6%, and then Atlantic Rainforest with 2.1%, 1.6% and 3.8%, respectively (Paper I and Paper III).

4.2

Analysis of runoff trends

The non-parametric Mann-Kendall (MK) statistical test [Yue et al., 2002] is used to assess the significance of trend in monthly runoff data under the null hypothesis of stationarity of the Funil, Camargos, Furnas, P Colômbia, Marimbondo and A Vermelha sub-basins. The results of trend test performed by using the MK tests at 95% significance level are shown in table 4.1. 38

Results and discussions

Table 4.1: Trend test results for monthly runoff time series at 95% significance level. Sub-basin Funil Camargos Furnas P Colômbia Marimbondo A Vermelha

Z -0.39 -0.51 -0.21 -1.39 -1.33 -1.82

p-value 0.347 0.304 0.416 0.082 0.092 0.035

(%) Null Hypothesis (H) Not rejected (Stationary) Not rejected (Stationary) Not rejected (Stationary) Not rejected (Stationary) Not rejected (Stationary) Not rejected (Stationary)

Table 4.1 reveals that MK trend tests on 1970-2010 time series of monthly runoff data did not reject the null hypothesis - stationarity for all sub-basins. However, the outcome of the test also shows evidences of positive and negative trends according to the standardized MK statistic Z and the probability value P (p-value) calculated for each sub-basin. For independent sample data without trend, for instance, p-value and Z should be equal to 0.5 and 0, respectively. P-values closer to 1 and positive values for Z indicate data with positive trend whereas data with negative trend yields p-values closer to 0 and negative values for Z [Rao and Hsu, 2008]. In light of the results obtained from the mapping of sugarcane plantations (Paper I), MK trend tests show that sugarcane expansion is associated with downward trends in monthly runoff for the 40-year period. This is because negative trends are present in all sub-basins that have substantial expansion (i.e. P Colômbia, Marimbondo and A Vermelha). Despite Funil, Camargos and Furnas also presenting downward trends represented by negative values for Z and p-values lower than 0.5, their absolute values are too small to be considered as evidences for trends.

4.3

Short-, medium- and long-impacts of the rapid sugarcane expansion on local hydrology

According to bootstrap results obtained in Paper III, percentage differences in daily runoff between CR1993 and R2000 were not statistically significant at the 99% confidence level. It can be associated with the small expansion of sugarcane plantations between 1993 and 2000, which corresponded to a little over 2.5% of the Rio Grande basin (see Paper I). From 1993 to 2007, however, sugarcane (20.7%) surpassed agriculture of grain (11.8%) as the second-largest land use in the Rio Grande basin. It implied to reductions in average daily runoff from 0.25% to 1.5% at the outlets of the sub-basins (Figure 4.1). These 39

4.3 Short-, medium- and long-impacts of the rapid sugarcane expansion on local hydrology

Figure 4.1.

Results from bootstrap analysis of the percentage differences of daily surface runoff between CR1993 and R2007 (a) and CR1993 and EMBRAPA (b).

reductions monotonically increase with the area converted to sugarcane over each sub-basin. Accordingly, average daily runoff at the outlets of A Vermelha and Marimbondo were the most affected by sugarcane expansion which have been reduced by up to 1% and 1.5%, respectively (Paper II). On the other hand, percentage differences in daily runoff were significant at 0.01 level between CR1993 and REMBRAPA. Further, 40

Results and discussions

they reached up to -10% at the outlets of the headwater sub-basins. Since the headwater sub-basins are dominantly composed by shallow soils, which easily become saturated, conversion of pasture to sugarcane significantly increased evapotranspiration rates reducing runoff at their outlets. Concerning medium-term temporal analysis, annual fluctuations in runoff, evapotranspiration and soil water content derived from the sugarcane expansion proposed between CR1993 and R2007 range -0.7 to 1%. It means that despite that differences in daily runoff, for example, achieved up to -2.5%, annual accumulated differences in runoff, evapotranspiration and soil water content are affected by the sugarcane growth stages, which may smooth impacts of sugarcane expansion on water balance over longer time frames. A full section is dedicated to the description of the long-term changes in the hydrological cycle under sugarcane expansion in Paper III. There, results obtained from the model runs are locally (per sub-basin) and globally (at a basin scale) evaluated over the Rio Grande basin as cumulative differences at the end of 20 years of simulation. Despite that Paper III showed sugarcane expansion mostly affected the water balance if it occurs over the headwater areas of low soil water storage capacity and its impact on the runoff at the outlet of the Rio Grande basin was marginal, a full evaluation of the hydrological cycle over the Rio Grande basin was needed; not only to have a better understanding of how land use changes might affect the hydroelectric power generation in the Rio Grande basin but also to provide sufficient information to Brazilian decision makers, international stakeholders and trade organizations to manage changes towards a sustainable expansion of sugarcane — as Brazil is the current largest exporter of ethanol in the world [IEA, 2011a]. The reader is highly encouraged to visit Paper II and Paper III for further details regarding the impacts of sugarcane expansion on the hydrological cycle of the Rio Grande basin over 20 years.

4.4

The atmospheric-hydrological modelling system: analysis and tests

In this section, results from the simulations performed using the atmospheric-hydrological model system implemented as described in Paper V are compared to outputs from BRAMS, satellite images at visible spectrum and daily rainfall data for one of the three study cases presented in Section 3.6.3 identified as the cold front passage. For comparisons and analyses regarding the other two remaining study cases, the reader is referred to Paper VI. On January 21s t , a false-colour Meteosat-9 image captured at 2100 UTC shows a cold front passing over the Rio Grande basin (Figure

41

4.4 The atmospheric-hydrological modelling system: analysis and tests

4.2a). In the image, the front is detected by the presence of brighter cumulus clouds. Since brightness of clouds in visible satellite images is related to their top heights [Song et al., 2004], Figure 4.2a is an indicator of vertical motion upwards which may have caused high cloud tops and increased their reflectance in the image. Daily observed rainfall over the basin ranged from 0 to 80 m m . Although the satellite image indicates developing and mature cumulus clouds all over the basin due to the passage of this front, the highest values of rainfall depth were observed mostly concentrated in the eastern part of the basin as shown in Figure 4.2b. This spatial rainfall distribution is primarily associated with the orography of the Rio Grande basin. As the satellite image and observed daily rainfalls have different time scales, they were compared to results obtained from the shortterm runs in a way to enhance the performance of each model in reproducing spatial alignment and distribution of the front rather than estimating rainfall intensity values. The spatial alignment and distribution of fronts are generally characterized by shifts in wind directions, and sharp temperature and air moisture content changes over short distances [Ahrens, 2007]. Therefore, temperature, zonal and meridional wind fields calculated by the integrated system and regional atmospheric model were compared to the spatial alignment and distribution of the front observed in the visible satellite image (Figures 4.2 and 4.3). Moreover, Figures 4.2b and 4.2c show computations of vertical motion at 500 hPa (shaded) using the regional atmospheric model and integrated system, respectively. As presented in Figures 4.2b and 4.2c, zonal and meridional wind fields estimated by both models indicated convergent air flows over the Rio Grande basin. However, the convergence zone is placed by the coupled system in better accordance to the shape of the front observed in the satellite image than by the regional atmospheric model. According to Jacobson [2005], convergence zones derived from fronts related to abrupt variations of land surface temperature. In this context, the process-based representation of land surface hydrological processes proposed by the integrated system yielded a good agreement between spatial gradients of land surface temperature and spatial alignment of the front (Fig. 4.3b). In contrast, spatial distribution of abrupt variations of land surface temperature estimated by the regional atmospheric model presented neither the same pattern as the coupled system nor the spatial alignment of the front (Fig. 4.3a). Differences in spatial distribution of land surface temperature between the atmospheric model alone and the coupled system resulted also in different placement of areas of deep convection. Zones of deep convection were identified by upward vertical motion higher than 0.7 m .s −1 at 500 hPa [Jorgensen and Lemone, 1989] and are presented as shaded areas in Figures 4.2b and 4.2c. Comparing to the satellite image, Figure 4.2c shows that despite zones of deep convection by the 42

Results and discussions

Figure 4.2. The false-colour Meteosat-9 image captured on January 21s t at 2100 UTC over the Rio Grande basin (a). Compared to the regional atmospheric model (b), the spatial alignment of the air flow field calculated by the coupled model (c) showed a good agreement with the visible satellite image.

43

4.4 The atmospheric-hydrological modelling system: analysis and tests

Figure 4.3. Temperature rates showed a better spatial representation of the cold front using the coupled model (b) than the regional atmospheric model (a).

44

Results and discussions

coupled system did not fully match the alignment of the front, they were placed in accordance with the occurrence of brighter cumulus clouds in the satellite image. On the other hand, zones of deep convection indicated by the regional atmospheric model alone were neither associated with the alignment of the front nor with any cumulus cloud observed in the satellite image. Although estimates of rainfall by regional atmospheric models are yet not accurate enough for realistic representation of rainfall intensity [Kendon et al., 2012], observed daily rainfall data were compared to 24h-accumulated rainfall in order to capture the spatial rainfall distribution. Figure 4.4 presents observed daily rainfall (Fig. 4.4a) and 24h-accumulated rainfall estimated by the regional atmospheric model (Fig. 4.4b) and the coupled system (Fig. 4.4c). According to Figure 4.4, the same spatial patterns of rainfall were identified for the regional atmospheric model and integrated system. Moreover, both of them presented the same range of values, from 0 to 80 m m .d −1 , and similar location of wet and dry regions in the Rio Grande basin. Except for some local rainfall events in the dry regions, most of the rainfall gauging stations with values larger than 70 m m .d −1 were located in the wet region pointed by the models. This means that, in terms of regional hydrological behaviour of rainfall, the models had a good agreement with the daily rainfall data when representing the front.

4.5

Conceptual evaluation of the exchange of water between soil, land surface and atmosphere

In spite of the hybrid model dynamically coupling soil, land surface and atmosphere as an integrated system, this section presents and discusses its estimates of fluxes of water and heat individually for the soil-land surface and land surface-atmosphere interfaces. Initially, section 4.5.1 analyses the fluxes of water between soil and land surface by the response of monthly evapotranspiration and runoff rates to monthly average values of soil water content and accumulated flow to the deep drainage. Thereafter, section 4.5.2 presents the exchange of water and heat between land surface and atmosphere as seasonal variations in monthly air temperature, precipitation, evapotranspiration – which is closely associated with latent heat flux – and amount of water vapor stored in the atmospheric column given by the verticallyintegrated water mixing ratio. Below, the results obtained from the four model runs using HLU1993, HLU2000, HLU2007 and PFLU – hereafter known as R1993, R2000, R2007 and RPFLU – are presented and discussed. Although the hybrid coupled model divides the simulation domain into individual uniform cells of 10x10 k m , the results from R1993, R2000, R2007 and 45

4.5 Conceptual evaluation of the exchange of water between soil, land surface and atmosphere

Figure 4.4. 24-hour accumulated rainfall fields estimated by BRAMS (b) and 46

the coupled system (c) presented the same regional trend as (a) observed daily rainfall at rain gauges, where high intensities of rainfall were concentrated in the southeastern part of the Rio Grande basin.

Results and discussions

RPFLU are spatially-averaged over the Rio Grande basin.

4.5.1

Soil-land surface interface

The water balance in the soil near-surface layer as estimated by the hybrid coupled model shows that soil water content is associated with variations in flow to the deep drainage over the year. While soil water content decreases, the amount of water that contributes to the deep drainage increases. However, the magnitude of this response is neither linearly equivalent nor instantaneous. This is because, besides the withdraw of water to the deep drainage, the hybrid coupled model also includes loss of storage water due to evapotranspiration and replenishment of water from effective precipitation to estimate the soil water content (Fig. 4.5.2). Under land use changes, results obtained from R1993 and RPFLU reveals that the hybrid coupled model indicates relevant changes in the soil water balance. During the germination stage of sugarcane from June to August, for instance, the hybrid coupled model captures an increase in soil water content as a result of the reduction of the average leaf area index over the basin for RPLFU. On the other hand, estimates of soil water content from R1993 continuously decreases until August. After August, when a wet season month starts – defined here as a month where monthly accumulated precipitation is equal or greater than 100 millimetres – estimates of soil moisture content instantaneously increase for R1993 whereas the rising curve is delayed by a month for RPFLU (Fig. 4.5.2). This late rising curve in soil water content is addressed to changes related to physiological maturity in sugarcane plantations (from germination to maturation stage between August and September), since precipitation effective becomes smaller as the average leaf area index is increased. Runoff as calculated by the hybrid coupled model is a combination of precipitation, evapotranspiration and soil water content in the near-surface soil layer. At the outlet of the Rio Grande basin, accumulated monthly runoff from R1993 and RPFLU vary within a range of historical records from 1930 to 2010. Nevertheless, estimates of runoff from R1993 and RPFLU present different seasonal patterns over the year, which are closely associated with soil water content in the soil near-surface layer and evapotranspiration and precipitation over the land surface. From January to May, a falling limb of the hydrographs is explained by a significant reduction in precipitation over the basin. It is followed by a dry season period – when monthly accumulated precipitation does not exceed 100 millimetres. During this period, the runoff at the outlet of the Rio Grande basin comes basically from the soil near-surface layer and delayed flow from the deep drainage. As evapotranspiration rates significantly decreases for RPFLU, the effective precipitation increases which enables a higher replenishment of the soil near-surface layer and, thereby higher runoff during the dry period compared to estimates of runoff from R1993. 47

4.5.1 Soil-land surface interface

Figure 4.5. The results from R1993, R2000, R2007 and RPFLU (or EMBRAPA) performed over 1 year period. The water balance in the soil is presented in (a) and (d) by monthly accumulated water contribution to the deep drainage and monthly average soil water content. In the land surface and atmosphere, exchange of water is shown in (b), (c), (e) and (f ) as monthly accumulated values of evapotranspiration, precipitation, vertically-integrated water mixing ration and runoff. For runoff (f ), box plots mean records from 1930 to 2010 with a 90% confidence interval.

48

Results and discussions

4.5.2

Land surface-atmosphere interface

The coupling strategy used by the hybrid coupled model to couple MGB-IPH and BRAMS has evapotranspiration as its key variable to enhance the exchange of water from land surface to atmosphere. Practically, the results obtained from R1993, R2000, R2007 and RPFLU shows that the hybrid coupled model is very sensitive to seasonal changes in evapotranspiration rates. While the effects of these seasonal changes in evapotranspiration on precipitation are only marginal, estimates of vertically-integrated water mixing ratio from RPFLU present a slight increase compared to those from R1993, R2000 and R2007 in October and November (Figure ). For these two months, accumulated monthly rainfall reaches up to 600 m m – which indicates that the atmospheric column is close to saturation. At such conditions, the vertically-integrated water mixing ratio becomes more susceptible to variations as more water is added to the atmospheric column by evapotranspiration.

49

CHAPTER

5

CONCLUSIONS AND OUTLOOK This thesis is aimed at evaluating potential variations in land use practices as a trigger mechanism of changes in land surface-atmosphere interactions. To do so, an atmospheric-hydrological modeling system has been implemented and tested under the rapid sugarcane expansion in the Rio Grande basin, Brazil, caused by the National Alcohol program (Pró-Álcool) launched in 1975. Firstly, satellite imagery analysis was carried out to quantify the sugarcane expansion over the Rio Grande basin (Paper I). Secondly, an eventual hydrological response to this expansion was evaluated by means of trend analysis in observed data and simulating the hydrological cycle under land use scenarios of sugarcane expansion as presented in Paper I (Paper III). In parallel, an atmospheric-hydrological modelling system was implemented in Paper V and its numerical approach to land surface-atmosphere interactions was compared to satellite imagery and observed data (see Paper VI). In light of the results obtained from these numerical and imagery analysis performed within the research presented in this thesis, some findings concerning the interplay between land use, hydrology and atmosphere as well as — more specifically — the rapid sugarcane expansion due to the Pró-Álcool over the Rio Grande basin and its impacts on both hydrological cycle and the atmosphere were achieved and are separately described below. Towards the sugarcane expansion over the Rio Grande basin as estimated by land use classification of satellite imagery: • Very few or no sugarcane plantations over headwater subbasins basically addressed to adverse climate and topographical conditions to grow sugarcane (i.e. high elevations and low temperatures). • Further downstream, however, areas covered by sugarcane represented up to 27.9% of the Marimbondo sub-basin already 51

in 1993 whereas they almost tripled after 14 years in the A Vermelha sub-basin. • Sugarcane plantations replaced mostly pasture lands and areas of grain agriculture. Comparisons made between land use distribution in 2007 and 1993 showed that the replacement of pasture lands by sugarcane fields achieved 6.8%, 7.5% and 8.9% of the Marimbondo, P Colômbia and A Vermelha sub-basins, respectively. It is followed by the replacement of areas of agriculture of grain crops with 5.2%, 4.7% and 7.6%, and then Atlantic Rainforest with 2.1%, 1.6% and 3.8%, respectively. In respect to the changes in the hydrological cycle of the Rio Grande basin as a natural response to the sugarcane expansion: • Four factors could be identified as highly related to the impacts of sugarcane expansion on the water balance of the Rio Grande basin: · amount of area replaced with sugarcane plantations · their location within the basin · regional soil properties · local groundwater contribution to stream flow • Sugarcane expansion mostly affected the water balance if it occurs over the headwater areas of low soil water storage capacity. For these areas, sugarcane expansion significantly increased evapotranspiration whereas reduced runoff and soil moisture content. • In case all areas suitable for growing sugarcane as defined by EMBRAPA [BRASIL, 2009] are filled with sugarcane fields, the research presented in this thesis estimates that water loss by evapotranspiration exclusively due to sugarcane expansion might achieve up to 4 m /m 2 in the headwater sub-basins of the Rio Grande basin over 20 years. Regarding the numerical approach employed by the atmospherichydrological modelling system: • Overall, the replacement of land surface parameterizations by process-based hydrological modelling implied improvement in temperature, air water content, zonal and meridional winds calculated near land surface. • According to investigations made concerning the implications of these changes on the atmosphere, convergent flows and abrupt variations of temperature calculated by the coupled system were better related to the spatial alignment and distribution of the front than the regional atmospheric model. 52

Conclusions and Outlook

• As the positioning of the front and development of convection are highly associated with temperature gradients, it indicates that the coupled system provides a better support to weather forecasting when simulating atmospheric processes driven by local surface heating or sharp temperature gradients. Finally, the conceptual evaluation of the interplay between soil, land surface and atmosphere as interpreted by the coupled model allows to enumerate the following findings: • The coupled model was able to capture changes in seasonal patterns of evapotranspiration, soil water content and runoff arising from the expansion of sugarcane plantations over the Rio Grande basin. • During the dry season, when surface runoff is mostly driven by baseflow, results obtained from the hybrid coupled model revealed that the harvesting of sugarcane plantations significantly increased runoff rates by increasing the flow through the soil near-surface layer. • Besides runoff and soil moisture, another consequence of the harvesting of sugarcane to the water balance was the reduction of fluxes of water from land surface to atmosphere along with the averaged leaf area index over the basin. • And, finally, the results from the atmospheric-hydrological coupled model presented here shows that the exchange of water between soil, land surface and atmosphere is a key factor in the understanding of how changes in the environment induced by anthropogenic activities can affect the whole climate system.

53

REFERENCES AgSri. Sustainable Sugarcane Initiative: Producing “more with less". Technical report, Natural Resource Management Center, 2012. D. A. Aguiar, B. F. T. Rudorff, W. F. Silva, M. Adami, and M. P. Mello. Remote sensing images in support of environmental protocol: Monitoring the sugarcane harvest in são paulo state, brazil. Remote Sensing, 3:2682–2703, 2011. C. D. Ahrens. Meteorology today: An introduction to weather, climate, and the environment. Thomson Brooks / Cole, 8th edition, 2007. A. Ajanovic. Biofuels versus food production: Does biofuels production increase food prices? Energy, 36(4):2070 – 2076, 2011. W. Alonso-Pippo, C. A. Luengo, J. Koehlinger, P. Garzone, and G. Cornacchia. Sugarcane energy use: The cuban case. Energy Policy, 36(6):2163 – 2181, 2008. R. G. B. André, J. C. Mendonça, V. S. Marques, F. M. A. Pinheiro, and J. Marques. Aspectos energéticos do desenvolvimento da cana-de-açúcar. parte 1: Balanço de radiação e parâmetros derivados. Revista Brasileira de Meteorologia, 25:375–382, 2010. ANEEL. Atlas brasileiro de energia elétrica. Brazil (Government), Brasília, BR, 2nd edition, 2005. ANEEL. Atlas de energia elétrica do Brasil. 9788587491107.

CEDOC, Brasília, BR, 2008.

ISBN

C. Angelo. Growth of ethanol fuel stalls in brazil. Nature, 491(7426):646 – 647, November 2012. J. G. Arnold, P. M. Allen, and G. Bernhardt. A comprehensive surface-groundwater flow model. Journal of Hydrology, 142:47–69, 1993. G. Balsamo, P. Viterbo, A. Beljaars, B. D. Hurk, M. Hirschi, A. K. Betts, and K. Scipal. A revised hydrology for the ecmwf model: verification from field site to terrestrial water storage and impact in the integrated forecast system. Journal of Hydrometeorology, 10:623–643, 2009. J. S. Baron, M. D. Hartman, T. G. F. Kittel, L. E. Band, D. S. Ojima, and R. B. Lammers. Effects of land cover, water redistribution, and temperature on ecosystem processes in the south platte basin. Ecological Applications, 8(4):1037 – 1051, 1998. J. Bartholmes and E. Todini. Coupling meteorological and hydrological models for flood forecasting. Hydrology and Earth System Sciences, 9:333 – 346, 2005. S. A. Basha, K. R. Gopal, and S. Jebara. A review on biodiesel production, combustion, emissions and performance. Renewable and Sustainable Energy Review, 13:1628 – 1634, 2009. doi: http://dx.doi.org/10.1016/j.rser.2008.09.031.

55

References

A. Bender and E. D. de Freitas. Evaluation of brams turbulence schemes during a squall line occurrence in são paulo, brazil. American Journal of Environmental Engineering, 3:1–7, 2013. R. Benoit, P. Pellerin, N. Kouwen, H. Ritchie, N. Donaldson, P. Joe, and E. D. Soulis. Toward the use of coupled atmospheric and hydrologic models at regional scale. Monthly Weather Review, 128:1681 – 1706, 2000. M. Bittelli, F. Tomei, A. Pistocchi, M. Flury, J. Boll, E. S. Brooks, and G. Antolini. Development and testing of a physically based, three-dimensional model of surface and subsurface hydrology. Advances in Water Resources, 33:106–122, 2010. C. Bolling and N. R. Suarez. The brazilian sugar industry: Recent developments. Sugar and Sweetener Situation & Outlook, 232:14–18, 2001. J. M. M. Borges and A. N. Almeida. Pro-álcool: accomplishments and perspectives. World Sugar Journal, 7:7–13, 1985. BRASIL. Zoneamento agroecológico da cana-de-açúcar: Expandir a produção, presevar a vida, garantir futuro. Ministério da Agricultura, Pecuária e Abastecimento, Brasília, BR, 1st edition, 2009. M. Bremicker. Aufbau eines Wasserhaushaltsmodells für das Weser und das Ostsee Einzugsgebiet als Baustein eines Atmosphären-Hydrologie-Modells. PhD thesis, Geowissenschaftlicher Fakultät der AlbertLudwigs-Universität, Freiburg, DE, 1998. J. A. Businger, J. C. Wyngaard, Y. Izumi, and E. F. Bradley. Flux-profile relationships in the atmospheric surface layer. Journal of the Atmospheric Sciences, 28:181 – 189, 1971. O. D. Cheesman. The Environmental Impacts of Sugarcane Production. Cromwell Press, Trowbridge, UK, 1st edition, 2004. B. Collischonn, W. Collischonn, and C. E. M. Tucci. Daily hydrological modeling in the amazon basin using trmm rainfall estimates. Journal of Hydrology, 360:207–216, 2008. W. Collischonn. Simulação hidrológica de grandes bacias. PhD thesis, Instituto de Pesquisas Hidráulicas, Porto Alegre, RS, 2001. W. Collischonn, D. G. Allasia, B. C. Silva, and C. E. M. Tucci. The mgb-iph model for large-scale rainfall-runoff modeling. Hydrology Sciences Journal, 52:878 – 895, 2007. CPRM. Geodiversity of Brazil: Knowing the past to understand the present and predict the future. Geological Survey of Brazil, Rio de Janeiro, 2012. C. de Fraiture, M. Giordano, and Y. Liao. Biofuels and implications for agricultural water use: Blue impacts of green energy. Water Policy, 10:67 – 81, 2008. DOE. Biomass: Multi-year program plan. Technical report, U.S. Department of Energy, April 2012. A. M. Espinosa. Previsão quantitativa por conjunto da precipitação do verão 2007/2008 na Bacia de Rio Grande. PhD thesis, Universidade de São Paulo, 2011. FAO. Soil map of the world. Technical report, Food and Agriculture Organization, Paris, FR, 1974. FAOSTAT. Large time-series and cross sectional data relating to hunger, food and agriculture for 245 countries and territories and 35 regional areas, from 1961 to the most recent year, October 2013. URL http://faostat.fao.org/. ˘ S módulo 1. projeto fauFAURGS. Previsão de afluência a reservatórios hidrelétricos âA¸ rgs/finep 40.04.0094.00. Technical report, FINEP, Porto Alegre, BR, 2007.

56

References

J. Fenger. Air pollution in the last 50 years – from local to global. Atmospheric Environment, 43(1):13 – 22, 2009. doi: http://dx.doi.org/10.1016/j.atmosenv.2008. 09.061. R. J. Fletcher Jr, B. A. Robertson, J. Evans, P. J. Doran, J. R. R. Alavalapati, and D. W. Schemske. Biodiversity conservation in the era of biofuels: risks and opportunities. Frontiers in Ecology and the Environment, 9(3):161 – 168, 2011. J. A. Foley, R. DeFries, G. P. Asner, C. Barford, G. Bonan, S. R. Carpenter, F. S. Chapin, M. T. Coe, G. C. Daily, H. K. Gibbs, J. H. Helkowski, T. Holloway, E. A. Howard, C. J. Kucharik, C. Monfreda, J. A. Patz, I. C. Prentice, N. Ramankutty, and P. K. Snyder. Global consequences of land use. Science, 309(5734):570 – 574, 2005. doi: 10.1126/ science.1111772. F.O.Litch. World ethanol and biofuels report. Technical report, Informa Agra, London, UK, October 2007. F.O.Litch. World ethanol and biofuels report. Technical report, Informa Agra, London, UK, May 2008. S. R. Freitas, K. M. Longo, M. A. F. Silva Dias, R. Chatfield, P. Silva Dias, P. Artaxo, M. O. Andreae, G. Grell, L. F. Rodrigues, A. Fazenda, , and J. Panetta. The coupled aerosol and tracer transport model to the brazilian developments on the regional atmo˘ S part 1: Model description and evaluation. spheric modeling system (catt-brams) âA¸ Atmospheric Chemistry and Physics, 9:2843–2861, 2009. N. Gedney, P. M. Cox, R. A. Betts, and O. Boucher. Detection of a direct carbon dioxide effect in continental river runoff records. Nature, 439:835–838, 2006. J. Goldemberg, S. T. Coelho, and P. Guardabassi. The sustainability of ethanol production from sugarcane. Energy Policy, 36:2086 – 2097, 2008. M. J. Groom, E. M. Gray, and P. A. Townsend. Biofuels and biodiversity: Principles for creating better policies for biofuel production. Conservation Biology, 22(3):602 – 609, 2008. S. Hagemann, B. Machenhauer, R. Jones, O. B. Christensen, M. Deque, D. Jacob, and P. L. Vidale. Evaluation of water and energy budgets in regional climate models applied over europe. Climate Dynamics, 23:547–567, 2004. M. Haggag, T. Yamashita, H. Lee, and K. Kim. A coupled atmosphere and multi-layer land surface model for improving heavy rainfall simulations. Hydrology and Earth System Sciences, 5:1067 – 1100, 2008. H. Hasenack, E. Weber, S. M. C. Fernandes, R. J. Jonas, and E. H. G. Duarte. Processamento e organização de modelos digitais de elevação contínuos para os países de língua portuguesa a partir do SRTM. Editora da UFRGS, 2010. ISBN 9788563843043. E. L. Hay, M. P. Clark, R. L. Wilby, W. J. Gutowski Jr, G. H. Leavesley, Z. Pan, R. W. Arritt, and E. S. Takle. Use of regional climate model output for hydrologic simulations. Journal of Hydrometeorology, 3:571 – 590, 2002. M. C. Herrera. Cane, sugar and the environment. In Proceedings of the Cuba/FAO International Sugar Conference, 1999. K. Hlavcova, J. Szolgay, S. Kohnová, and O. Horvát. The limitations of assessing impacts of land use changes on runoff with a distributed hydrological model case study of the hron river. Biologia, 64:589–593, 2009. U. Högström and A. S. Smedman-Högström. Turbulence mechanisms at an agricultural site. Boundary-Layer Meteorology, 7:373 – 389, 1974.

57

References

T. Homdee, K. Pongput, and S. Kanae. Impacts of land cover changes on hydrological responses: A case study of chi river basin, thailand. Journal of Hydraulic Engineering, 55:31 – 36, 2011. IBGE. Classificação da vegetação brasileira, adaptada a um sistema universal. Technical report, Instituto Brasileiro de Geografia e Estatística, Rio de Janeiro, BR, 1991. IEA. álcool : Projeção da produção e exportação no período 2005/06 a 2015/16. Technical report, Governo de São Paulo, São Paulo, BR, 2006. IEA. Bioenergy project development & biomass supply. Technical report, International Energy Agency, Paris, FR, 2007. IEA. Renewable energy essentials: Wind. Technical report, International Energy Agency, Paris, FR, 2008. IEA. Renewable energy essentials: Solar heating and cooling. International Energy Agency, Paris, FR, 2009.

Technical report,

IEA. Renewable energy: Markets and prospects by technology. International Energy Agency, Paris, FR, 2011a.

Technical report,

IEA. World energy outlook 2011. Technical report, International Energy Agency, Paris, FR, 2011b. IEA. Technology roadmap: Bioenergy for heat and power. International Energy Agency, Paris, FR, 2012.

Technical report,

IEA. Medium-term renewable energy market report. Technical report, International Energy Agency, Paris, FR, 2013. IPT. Diagnóstico da situação dos recursos hídricos na bacia hidrográfica do rio grande ˘ S sp/mg. Technical Report 96581205, Instituto de Pesquisas Tecnológicas (bhrg) âA¸ do Estado de São Paulo, 2008. A. Z. Jacobson. Fundamentals of atmospheric modeling. Cambridge University Press, 2005. A. Jagger. The fluctuating fortunes of brazilian ethanol. Biofuels, Bioproducts and Biorefining, 7(2):103 – 105, 2013. G. James. Sugarcane. John Wiley & Sons, Ltd, Oxford, UK, 2nd edition, 2008. G. Jewitt and R. Kunz. The impact of biofuel feedstock production on water resources: a developing country perspective. Bioenergy and Water, 5(4):387 – 398, 2011. D. P. Jorgensen and M. A. Lemone. Vertical velocity characteristics of oceanic convection. Journal of the Atmospheric Sciences, 46(5):621 – 640, 1989. E. J. Kendon, N. M. Roberts, C. A. Senior, and M. J. Roberts. Realism of rainfall in a very high-resolution regional climate model. Journal of Climate, 25:5791 – 5806, 2012. S. C. Kot and Y. Song. An improvement of the louis scheme for the surface layer in an atmospheric modelling system. Boundary-Layer Meteorology, 88:239 – 254, 1998. D. Krohn, M. Woods, J. Adams, B. Valpy, F. Jones, and P. Gardner. Wave and tidal energy in the uk: Conquering challenges, generating growth. Technical report, Renewable UK, February 2013. T. J. Lee and R. A. Pielke. Estimating the soil surface specific humidity. Journal of Applied Meteorology and Climatology, 31:480 – 484, 1992.

58

References

X. Liang, D. P. Lettenmaier, E. F. Wood, and S. J. Burges. A simple hydrologically based model of land surface water and energy fluxes for general circulation models. Journal Geophysical Research, 99:14415–14428, 1994. C. A. Lin, L. Wen, G. Lu, Z. Wu, J. Zhang, Y. Yang, Y. Zhu, and L. Tong. Atmospherichydrological modeling of severe precipitation and floods in the huaihe river basin, china. Journal of Hydrology, 330:249 – 259, 2006. D. Lohmann, E. Raschke, B. Nijssen, and D. P. Lettenmaier. Regional scale hydrology: I. formulation of the vic-2l model coupled to a routing model. Hydrological Sciences Journal, 43:131–141, 1998. J. F. Louis. A parametric model of vertical eddy fluxes in the atmosphere. BoundaryLayer Meteorology, 17:187 – 202, 1979. G. N. Maraddi. An analysis of sustainable cultivation practices followed by sugarcane growers in Karnataka. PhD thesis, University of Agricultural Sciences, Dharwad, 2006. R. M. Maxwell and N. L. Miller. Development of a coupled land surface and groundwater model. Journal of Hydrometeorology, 6:233–247, 2005. C. Mendes and J. Cirilo. Geoprocessamento. In Geoprocessamento em recursos hídricos: Princípios, integração e aplicação. Associação Brasileira de Recursos Hídricos, Porto Alegre, BR, 2001. L. Menut, O. P. Tripathi, A. Colette, R. Vautard, E. Flaounas, and B. Bessagnet. Evaluation of regional climate simulations for air quality modelling purposes. Climate Dynamics, 40(9 – 10):2515 – 2533, 2013. doi: 10.1007/s00382-012-1345-9. C. Messager, H. Galle, O. Brasseur, B. C. C. Peugeot, L. Seguis, M. Vauclin, R. Ramel, G. Grasseau, and L. Leger. Influence of observed and rcm-simulated precipitation on the water discharge over the sirba basin, burkina faso-niger. Climate Dynamics, 27:199 – 214, 2006. D. Mitchell. A note on rising food prices. World Bank Policy Research Working Paper Series, 2008. URL http://ssrn.com/abstract=1233058. N. Molders. On the uncertainty in mesoscale modeling caused by surface parameters. Metereology and Atmosphere Physics, 76:119–141, 2000. P. Morales, T. Hickler, D. P. Rowell, B. Smith, and M. T. Sykes. Changes in european ecosystem productivity and carbon balance driven by regional climate model output. Global Change Biology, 13:108–122, 2007. N. T. Mui, T. R. Preston, D. van Binh, L. V. Ly, and I. Ohlsson. Effect of management practices on yield and quality of sugar cane and on soil fertility. Livestock Research for Rural Development, 8(3):51, 1996. B. Nijssen, D. P. Lettenmaier, X. Liang, S. W. Wetzel, and E. F. Wood. Streamflow simulation for continental-scale river basins. Water Resources Research, 33:711–724, 1997. M. Nitsch. O programa de biocombustíveis proalcool no contexto da estratégia energética brasileira. Brazilian Journal of Political Economy, 2:123–138, 1991. M. T. Nóbrega, W. Collishonn, C. E. M. Tucci, and A. R. Paz. Uncertainty in climate change impacts on water resources in the rio grande basin, brazil. Hydrology and Earth System Sciences, 15:585–595, 2011. K. Onogi, H. Koide, M. Sakamoto, S. Kobayashi, J. Tsutsui, H. Hatsushika, T. Matsumoto, N. Yamazaki, H. Kamahori, K. Takahashi, K. Kato, R. Oyama, T. Ose, S. Kadokura, ˘ Tprogress ˇ and K. Wada. Jra-25: Japanese 25-year re-analysis projectâA and status. Quarterly Journal of the Royal Meteorological Society, 131:3259 – 3268, 2005.

59

References

ONS. Inventário das restrições operativas dos aproveitamentos hidroelétricos. Technical Report 00132013, Operador Nacional do Sistema Elétrico, 2013. R. C. D. Paiva and W. Collischonn. Climate change impacts on river Quarai flow. In C. F. Fung, A. Lopez, and M. New, editors, Climate and Water Resources Modelling, page 200. John Willey & Sons, 2010. R. C. D. Paiva, W. Collischonn, and C. E. M. Tucci. Large scale hydrologic and hydrodynamic modeling using limited data and a gis based approach. Journal of Hydrology, 406:170 – 181, 2011. S. Panday and P. S. Huyakorn. A fully coupled physically-based spatially-distributed model for evaluating surface/subsurface flow. Advances in Water Resources, 27:361– 382, 2004. A. R. Paz, W. Collischonn, C. E. M. Tucci, and C. R. Padovani. Large scale modelling of channel flow and floodplain inundation dynamics and its application to the pantanal (brazil). Hydrological processes, 25:1498 – 1516, 2011. F. F. Pereira. Modelo hidrodinâmico e de transporte bidimensional de grade não estruturada para lagos rasos. Master’s thesis, Instituto de Pesquisas Hidráulicas, Porto Alegre, RS, January 2010. G. T. Phelps and S. Pond. Spectra of the temperature and humidity fluctuations and of the fluxes of moisture and sensible heat in the marine boundary layer. Journal of the Atmospheric Sciences, 28(6):918 – 928, 1971. R. A. Pielke and R. W. Arritt. A proposal to standardize models. Bulletin of the American Meteorological Society, 65:1082, 1984. R. A. Pielke, R. Avissar, M. Raupach, A. J. Dolman, X. Zeng, and A. S. Denning. Interactions between the atmosphere and terrestrial ecosystems: influence on weather and climate. Global change biology, 4:461–475, 1998. A. J. Pitman. The evolution of, and revolution in, land surface schemes designed for climate models. International Journal of Climatology, 23:479–510, 2003. RADAMBRASIL. Programa de integração nacional, levantamento de recursos naturais (Folhas SE-23 (Belo Horizonte) e SF-22 (Paranapanema)). Technical report, Ministério das Minas e Energia, Brasília, BR, 1982. A. R. Rao and E.-C. Hsu. Streamflow data analyses, volume 60 of Water Science and Technology Library, chapter 5, pages 121 – 148. Springer, 2008. N. H. Ravindranath, H. I. Somashekar, M. S. Nagaraja, P. Sudha, G. Sangeetha, S. C. Bhattacharya, and P. A. Salam. Assessment of sustainable non-plantation biomass resources potential for energy in india. Biomass & Bioenergy, 29(3):178 – 190, 2005. W. Rawls, L. Ahuja, D. Brakensiek, and A. Shirmohammadi. Handbook of Hydrology, chapter Infiltration and soil water movement. McGraw Hill, 1993. A. Raza, J. K. Friedel, A. M, M. R. Ardakani, W. Loiskandl, M. Himmelbauer, and G. Bodner. Modeling growth of different lucerne cultivars and their effect on soil water dynamics. Agricultural Water Management, 119:100–110, 2013. M. J. Robertson, N. G. Inman-Bamberb, R. C. Muchowa, and A. W. Wood. Physiology and productivity of sugarcane with early and mid-season water deficit. Field Crops Research, 64:211–227, 1999. H. R. Rocha, H. C. Freitas, R. Rosolem, R. I. N. Juárez, R. N. Tannus, M. A. Ligo, O. M. R. Cabral, and M. A. F. Silva Dias. Measurements of co2 exchange over a woodland savanna (cerrado sensu stricto) in southeast brasil. Biota Neotropica, 2(1):1 – 11, 2002.

60

References

F. Rosillo-Calle and L. A. B. Cortez. Towards proalcool ii - a review of brazilian bioethanol programme. Biomass & Bioenergy, 14(2):115 – 124, 1998. B. F. T. Rudorff, D. A. Aguiar, W. F. Silva, L. M. Sugawara, M. Adami, and M. A. Moreira. Studies on the rapid expansion of sugarcane for ethanol production in são paulo state (brazil) using landsat data. Remote Sensing, 2:1057–1076, 2010. A. L. Ruhoff. Sensoriamento remoto aplicado à estimativa da evapotranspiração em biomas tropicais. PhD thesis, Instituto de Pesquisas Hidráulicas, 2011. C. F. Runge and B. Senauer. How biofuels could starve the poor. Foreign Affairs, 86(3): 41 – 53, 2007. G. Sampaio, C. Nobre, M. H. Costa, and P. Satyamurty. Regional climate change over eastern amazonia caused by pasture and soybean cropland expansion. Geophysical Research Letter, 34:1–7, 2007. J. C. Schaake, V. I. Koren, Q.-Y. Duan, K. Mitchell, and F. Chen. Simple water balance model for estimating runoff at different spatial and temporal scales. Journal of Geophysical Research, 101:7461–7475, 1996. R. E. Schulze. Modelling hydrological responses to land use and climate change: A southern african perspective. Journal of the Human Environment, 29(1):12 – 22, 2000. P. J. Sellers, R. E. Dickinson, D. A. Randall, A. K. Betts, F. G. Hall, J. A. Berry, G. J. Collatz, A. S. Denning, H. A. Mooney, C. A. Nobre, N. Sato, C. B. Field, and A. HendersonSellers. Modeling the exchanges of energy, water, and carbon between continents and the atmosphere. Science, 275:502–509, 1997. S. I. Seneviratne, J. S. Pal, E. A. B. Eltahir, and C. Schar. Summer dryness in a warmer climate: a process study with a regional climate model. Climate Dynamics, 20:69–85, 2002. T. Serra, D. Zilberman, and J. Gil. Price volatility in ethanol markets. European Review of Agricultural Economics, 38(2):259 – 280, 2011. G. Seuffert, P. Gross, C. Simmer, and E. F. Wood. The influence of hydrologic modeling on the predicted local weather : Two-way coupling of a mesoscale weather prediction model and a land surface hydrologic model. Journal of Hydrometeorology, 3: 505–523, 2002. A. Singh. Groundwater modelling for the assessment of water management alternatives. Journal of Hydrology, 481:220–229, 2013. E. Smeets, M. Jungingera, A. Faaija, A. Walter, P. Dolzanb, and W. Turkenburga. The sustainability of brazilian ethanol - an assessment of the possibilities of certified production. Biomass & Bioenergy, 32:781 – 813, 2008. K. Smith. No peak oil really is dead. Forbes, 2013. X. Song, Y. Zhao, and Z. Liu. Cloud detection and analysis of modis image. In Proceedings of the Geoscience and Remote Sensing Symposium, September 2004. A. St-Hilaire, T. B. M. J. Ouarda, M. Lachance, B. Bobiée, J. Gaudet, and C. Gignac. Assessment of the impact of meteorological network density on the estimation of basin precipitation and runoff: a case study. Hydrological Processes, 17:3561–3580, 2003. T. J. Stohlgren, T. N. Chase, R. A. Pielke, T. G. F. Kittel, and J. S. Baron. Evidence that local land use practices influence regional climate, vegetation, and stream flow patterns in adjacent natural areas. Global change biology, 4:495–504, 1998.

61

References

M. A. Sutton, E. Nemitz, J. W. Erisman, C. Beier, K. B. Bahl, P. Cellier, W. de Vries, F. Cotrufo, U. Skiba, C. di Marco, S. Jones, P. Laville, J. F. Soussana, B. Loubet, M. Twigg, D. Famulari, J. Whitehead, M. W. Gallagher, A. Neftel, C. R. Flechard, B. Herrmann, P. L. Calanca, J. K. Schjoerring, U. Daemmgen, L. Horvath, Y. S. Tang, B. A. Emmett, A. Tietema, J. Penuelas, M. Kesik, N. Brueggemann, K. Pilegaard, T. Vesala, C. L. Campbell, J. E. Olesen, U. Dragosits, M. R. Theobald, , P. Levy, D. C. Mobbs, R. Milne, N. Viovy, N. Vuichard, J. U. Smith, P. Smith, P. Bergamaschi, D. Fowler, and S. Reis. Challenges in quantifying biosphere-atmosphere exchange of nitrogen species. Environmental Pollution, 150:125–139, 2007. D. J. Tenenbaum. Food vs. fuel: diversion of crops could cause more hunger. Environmental Health Perspectives, 116(6):254 – 257, 2008. The Economist. Ethanol’s mid-life crisis. The Economist, Energy in Brazil, September 2010. URL http://www.economist.com/node/16952914. E. Todini. The arno rainfall-runoff model. Journal of Hydrology, 175:339–382, 1996. E. Toth, A. Brath, and A. Montanari. Comparison of short-term rainfall prediction models for real-time flood forecasting. Journal of Hydrology, 239:132–147, 2000. C. E. M. Tucci, W. Collischonn, R. T. Clarke, A. R. Paz, and D. Allasia. Short- and longterm flow forecasting in the rio grande watershed (brazil). Atmospheric Science Letters, 9:53 – 56, 2008. T. Umeda and P. T. Martien. Evaluation of a data assimilation technique for a mesoscale meteorological model used for air quality modeling. Journal of Applied Meteorology, 41:12 – 29, 2002. UN. The 1998 revision of united nations population projections. Population and Development Review, 24(4):891–895, 1998. UN. World population prospects: The 2012 revision, key findings and advance tables. Technical Report ESA-P-WP-227, United Nations, New York, US, 2012. UNICA. Sustainability report 2010. Technical report, Brazilian Sugarcane Industry Association, São Paulo, Brazil, 2010. UNICA. Sugarcane industry in brazil: Ethanol, sugar and bioelectricity. Technical report, Brazilian Sugarcane Industry Association, São Paulo, Brazil, 2013. B. J. J. M. van den Hurk and A. A. M. Holtslag. On the bulk parameterization of surface fluxes for various conditions and parameter ranges. Boundary-Layer Meteorology, 82:119 – 134, 1997. P. H. Verburg, W. Soepboer, A. Veldkamp, R. Limpiadaada, V. Espaldon, and S. S. A. Mastura. Modeling the spatial dynamics of regional land use: The clue-s model. Environmental Management, 30(3):391 – 405, 2002. doi: 10.1007/ s00267-002-2630-x. P. L. Vidale, D. Luthi, C. Frei, S. I. Seneviratne, and C. Schar. Predictability and uncertainty in a regional climate model. Journal of Geophysical Research, 108(D18), 2003. R. L. Walko, L. E. Band, J. Baron, T. G. F. Kittel, R. Lammers, T. J. Lee, D. Ojima, R. A. Pielke, C. Taylor, C. Tague, C. J. Tremback, and P. L. Vidale. Coupled atmospherebiophysics-hydrology models for environmental modeling. Journal of Applied Meteorology, 39:931–944, 2000. M. L. Warburton, R. E. Schulze, and G. P. W. Jewitt. Hydrological impacts of land use change in three diverse south african catchments. Journal of Hydrology, 414–415: 118–135, 2011.

62

References

R. Whitlock. Onshore wind industry reacts to latest government guidance on community engagement. Renewable energy magazine, 2013. G. N. Wijesekara, A. Gupta, C. Valeo, J. G. Hasbani, Y. Qiao, P. Delaney, and D. J. Marceau. Assessing the impact of future land-use changes on hydrological processes in the elbow river watershed in southern alberta, canada. Journal of Hydrology, 412–413: 220–232, 2012. K. B. Wilson, P. J. Hanson, P. J. Mulholland, D. D. Baldocchi, and S. D. Wullschleger. A comparison of methods for determining forest evapotranspiration and its components: sap-flow, soil water budget, eddy covariance and catchment water balance. Agricultural and forest meteorology, 106:153–168, 2001. WRI. Ecosystems and human well-being: Synthesis. Island Press, Washington, DC, 2005. ISBN 1597260401. WWFBrasil. Monitoring alterations in vegetation and land use in the brazilian portion of the upper paraguay river basin. Technical report, Non-governmental organizations: Ecoa - Ecologia e Ação, Conservation International, Avina Foundation, 2008. S. Yue, P. Pilon, B. Phinney, and G. Cavadias. The influence of autocorrelation on the ability to detect trend in hydrological series. Hydrological Processes, 16:1807 – 1829, 2002. P. Zuurbier and J. van de Vooren. Sugarcane ethanol. Wageningen Academic Publishers., Wageningen, 2008.

63

PAPERS

PAPER I Quantifying the Rapid Sugarcane Expansion for Ethanol Production in the Rio Grande basin, Brazil F. F. Pereira M. Tursunov and C. B. Uvo. Vatten 69, 83 – 86 (2013).

Paper I

VATTEN – Journal of Water Management and Research 69: 83–86. Lund 2013

Quantifying the Rapid Sugarcane Expansion for Ethanol Production in the Rio Grande basin, Brazil Kvantifiering av hastig spridning av sockerrörsplantage på grund av etanolproduktion i Rio Grandes avrinningsområde, Brasilien by Fábio Pereira 1, Mehriddin Tursunov 2 and Cintia Uvo 1 1 Department of Water Resources Engineering, Lund University Box 118, 221 00 Lund, Sweden. 2  Tashkent Institute of Irrigation and Melioration, Qori-Niyoziy, UZ e-mail: [email protected]

Abstract The search for sustainable sources of energy has found a realistic replacement to fossil fuels in ethanol and methanol. In response to this trend, sugarcane plantations have rapidly increased in Brazil. In this study, this sugarcane expansion was mapped for the River Grande basin in the northern São Paulo, considering the real extension of the plantation area in 1993, 2000, 2007 using Landsat satellite images. Comparisons between these three different land use scenarios showed a significant expansion of sugarcane plantations from 1993 to 2000 and a slight increase from 2000 to 2007. Key words – Sugarcane expansion, Automatic land use classification, Rio Grande basin, Landsat images

Sammanfattning Sökandet efter en hållbar energikälla har gett en realistisk ersättare för fossila bränslen i form av etanol och metanol. På grund av denna nya trend har sockerrörsplantagen i Brasilien ökat kraftigt. I denna studie kartlades expansionen av sockerrörsplantagen för River Grandes avrinningsområde i norra delen av São Paulo, med tanke på den verkliga utbredningen av dessa plantager år 1993, 2000 och 2007 med hjälp av satelitbilder från Landsat. Jämförelse mellan dessa tre olika perioder visade en signifikant expansion av sockerrörsplantage från 1993 till 2000 och en liten ökning mellan 2000 till 2007.

Introduction The high demand for ethanol has made sugarcane the world largest crop, covering an area of 20 million of hectares over more than 70 countries (FAO, 2012; Galdos, 2009). In the beginning of the century, Sweden, for instance, adopted a national strategy to become an oil-independent country by 2020 (Persson, 2006). Most of the ethanol imported to Sweden comes from Brazil (Energimyndigheten, 2012). VATTEN · 2 · 13

  Brazil is the country that retains the largest area of sugarcane cultivation in the world. It is responsible for approximately one third of the global harvested area and production (Zuurbier and van de Vooren, 2008). Since 1975, when the Pro-Álcool (ethanol program) started in Brazil as a response to the 1973 oil crisis (e.g. Borges, 1985; Nitsch, 1991), up to 2005, the Brazilian area of sugarcane plantation increased 170 % and reached 5.4 million hectares (Bolling and Suarez, 2001; IEA, 2006). Projections from IEA (2006) estimate the plantation 83

69

Quantifying the Rapid Sugarcane Expansion for Ethanol Production in the Rio Grande basin, Brazil

area to reach 12.2 million hectares by 2015. Most of the expansion in sugarcane area during the last 20 years has occurred within São Paulo state and the northern parts of this state are the region where the replacement of the original vegetation called Cerrado, a tropical savannah grassland, by sugarcane plantation has been the most intense (Conab, 2012).   Prior to estimate the impacts of the sugarcane expansion on the water balance, this study aims at mapping land use changes due to sugarcane expansion occurred in the River Grande basin, Brazil, considering the real extension of the plantation area in 1993, 2000, 2007 using Landsat data.

Study area: River Grande basin The River Grande basin is located in the eastern upper Paraná basin (Figure 1). The basin is also formed by important subsidiaries rivers such as rivers Pardo and MogiGuaçu. Approximately 60 % of hydroelectric power generation in Brazil is provided by Paraná River basin of which ~ 12 % comes from the 15 hydropower plants in the River Grande basin (ANEEL, 2005). Its current vegetation cover is formed mainly by remaining Cerrado and agriculture in the low basin and grassland in the high basin. The altitude in the basin varies from 300 to 2700 m.a.s.l. and the soil is composed mostly by agriculture and pasture in the low lands and forest in the high lands.   The original vegetation of the basin has a perennial characteristic, unlike the sugar cane that has a marked annual cycle. Typically, the sugarcane start growing around June (spring in the southern hemisphere) and is harvested about one year later, most commonly after biomass burning. During the sugar cane’s one year cycle, the albedo of the field varies from 0.12 to 0.25, and the leaf area index varies from 0.2 to 6 m²/m². The height of the plants reaches up to 3 m.

Data The U.S. Geological Survey (USGS) provides free access to satellite images around the world. These satellite images have 170 x 183 kilometres with a spatial resolution of 30 m. In this work, they have been used to classify land use as agriculture, forest, sugarcane and pasture.

Methods Mapping of Sugarcane Plantations Multi-temporal Landsat images were used for the characterization of sugarcane evolution in the Rio Grande basin. Land use maps were derived through analysis of satellite images made by Landsat TM 7 and extracted from U.S. Geological Survey. The selection of satellite images was driven by the availability of cloud-free Landsat data over the Rio Grande basin from 1970 to 2010. In this study, fourteen Landsat satellite images (170x183 km) were captured in 1993, 2000 and 2007, and used to generate three land use maps.   An automatic classification of Landsat satellite images showed in Rudorff et al. (2010) was used for mapping sugarcane fields. This automatic classification is based on a linear spectral mixing which consists of a linear combination of spectral signatures from two or more types of land use, such as agriculture, pasture, forest etc. Hence, sugarcane fields present a particular spectral signature as described by Rudorff et al. (2010). The sugarcane spectral signature has been used to identify sugarcane fields in the Rio Grande basin. Moreover, a visual inspection was made to support this automatic land use classification.   Each land use map was classified into five dominant types as areas covered by water bodies, Atlantic Forest, agriculture of grain crops, pasture lands and sugarcane fields according to their spectral signatures. Except for sugarcane fields, all spectral signatures were adopted as defined by Mendes and Cirilo (2001).

Fig. 1. Location of Paraná river basin and River Grande basin.

84

70

VATTEN · 2 · 13

Paper I

Results and conclusions Results from the automatic land use classification in 1993, 2000 and 2007 are presented in figure 2. In order to represent spatial heterogeneity of the sugarcane expansion, the Rio Grande basin was divided into 10 subbasins as shown in Figure 3.   In general, terrain slope and altitude were equally important factors for sugarcane expansion in the Rio Grande basin once sub-basins with areas which pre­ sented terrain slopes lower than 12 % and altitudes varying between 300 and 700 m.a.s.l. significantly increased the concentration of sugarcane fields in their drainage areas. Table 1 shows the percentage of areas covered by sugarcane fields for each sub-basin.   According to table 1, Funil, Camargos and Furnas did not present sugarcane expansion in their drainage areas. This is because, despite having areas with terrain slope less than 12 %, Funil, Camargos and Furnas sub-basins are higher than 800 m.a.s.l.. These results agree with

Table 1. The portion of areas covered by sugarcane fields per subbasin in 1993, 2000 and 2007.

Sub-basin

1993 (%)

2000 (%)

2007 (%)

Funil   0.0   0.0   0.0 Camargos   0.0   0.0   0.0 Furnas   1.5   1.5   1.5 P Colômbia 11.0 20.8 25.9 Marimbondo 27.9 31.1 42.0 A Vermelha   9.4 12.3 30.1

Fig. 2. Evolution of sugarcane crops for 1993, 2000 and 2007 in the River Grande basin.

Fig. 3. River Grande basin divided into 10 catchments and hydropower plants.

VATTEN · 2 · 13

85

71

Quantifying the Rapid Sugarcane Expansion for Ethanol Production in the Rio Grande basin, Brazil

what has been suggested by BRASIL (2009) as areas potentially suitable for cultivation of sugarcane in Brazil.   On the other hand, a significant increase of sugarcane fields is pointed for P Colômbia, Marimbondo and A Vermelha sub-basins. One of the reasons for this rapid sugarcane expansion may be explained by effects of the Pró-Alcool (Brazilian ethanol program) when the Brazilian government provided support to the production of sugarcane (Nitsch, 1991).   A chronological analysis pointed to different expansion rates among P Colômbia, Marimbondo and A Vermelha sub-basins. Between 1993 and 2000, areas covered by sugarcane increased by almost 90 % in P Colômbia whereas, for the same period, Marimbondo and A Vermelha had a sugarcane expansion of only 11.4 % and 30.8 %. In contrast, from 2000 to 2007, sugarcane expansion rates were higher in Marimbondo (35 %) and A Vermelha (140 %) than P Colômbia (22 %). In 2007, sugarcane fields represented approximately 26 % of the P Colômbia sub-basin, 30 % of the A Vermelha sub-basin and more than 40 % of the Marimbondo subbasin.   Overall, sugarcane fields replaced mostly pasture lands and agriculture of grain crops. Comparisons made between land use distribution in 2007 and 1993 showed that the replacement of pasture lands by sugarcane fields achieved 6.8 %, 7.5 % and 8.9 % of the Marimbondo, P Colômbia and A Vermelha sub-basins. It is followed by agriculture of grain crops with 5.2 %, 4.7 % and 7.6 %, and then Atlantic Forest with 2.1 %, 1.6 % and 3.8 %, respectively.   Regarding the water balance, the influence of sugarcane expansion depended upon a combination of factors, such as amount of areas replaced with sugarcane, type of land use replaced and location of the expansion within each basin. This study showed that the sugarcane expansion mostly impacts the water balance if it happens over the headwater areas with low soil water retention. For these areas, sugarcane expansion significantly increased evapotranspiration rates and reduced soil moisture content and surface runoff.   Researches ongoing are evaluating short-, mediumand long-term impacts of sugarcane expansion on the water balance of the Rio Grande basin, Brazil, as estimated by changes in evapotranspiration, soil moisture content and surface runoff on daily, monthly and annual basis.

86

72

Acknowledgments We would like to thank to Erasmus Mundus project for funding a post doctoral fellowship to Mehriddin Tursunov (2010–2011). We are also grateful to Crafoord Foundation and the Swedish Research Council (Vetenskaprådet)

References FAO – Food and Agriculture Organization of the United Nations (2012). FAOSTAT [online] Available at:http://faostat.fao.org/site/567/ > [Accessed 21 November 2012]. Galdos, M.V., Cerri, C.C., Cerri, C.E.P. (2009) Simulation of sugarcane residue decomposition and aboveground growth. Plant and Soil, v. (326), p. 243–259. Persson, G. (2006) På väg mot ett OLJEFRITT Sverige. Kommissionen mot oljeberoende. [online] Available at: [Accessed 20 November 2012]. Zuurbier, P., van de Vooren, J. (2008) Sugarcane ethanol. Wageningen Academic Publishers. p.39. Borges, J.M.M., de Almeida, A.N. (1985) Proálcool: accomplishments and perspectives. World Sugar Journal, v. 7(12), p. 7–13. Nitsch, M. (1991) O programa de biocombustíveis Proalcool no contexto da estratégia energética brasileira. Brazilian Journal of Political Economy, v. 2 (42), p. 123–138. Bolling, C., Suarez, N.R. (2001). The Brazilian Sugar Industry: Recent Developments. Sugar and Sweetener Situation & Outlook 232, p.14–18, Economic Research Service, USDA. IEA (2006) Álcool: Projeção da Produção e Exportação no Período 2005/06 A 2015/16. São Paulo, Brazil: Instituto de Economia Agrícola (IEA). Conab (2012) Companhia Nacional de Abastecimento. Acompanhamento da safra brasileira. [Acessed Fevereiro 2012]. Available at: http://www.conab.gov.br/conabweb/. ANEEL (2005) Agência Nacional de Energia Elétrica: Brazilian Electric Energy Atlas, edition 2, Brasília. Rudorff, B.F.T., Aguiar, D.A., Silva, W.F., Sugawara, L.M., Adami, M., Moreira, M.A. (2011) Studies on the rapid expansion of sugarcane for ethanol production in São Paulo state (brazil) using landsat data. Remote Sensing, v.(3), p. 2682–2703. Mendes, C., Cirilo, J. (2001) Geoprocessamento, in: Geo­ processamento em Recursos hídricos: Princípios, Integração e Aplicação. Associação Brasileira de Recursos Hídricos, Porto Alegre, BR. BRASIL (2009) Zoneamento Agroecológico da Cana-de-Açúcar: Expandir a produção, presevar a vida, garantir futuro. Ministério da Agricultura, Pecuária e Abastecimento. ISSN 1517-2627.

VATTEN · 2 · 13

PAPER II Effects of sugarcane expansion on runoff and evapotranspiration in the Rio Grande basin, Brazil. C. Borglin, S. Borglin, F. F. Pereira and C. B. Uvo. Vatten 69, 141 – 148 (2013).

Paper II

VATTEN – Journal of Water Management and Research 69: 141–148. Lund 2013

Effects of sugarcane expansion on runoff and evapotranspiration in the Rio Grande basin, Brazil Effekter av sockerrörsplantagers utbredning på avrinning och avdunstning i Rio Grandes avrinningsområde i Brasilien by Claes Borglin 1; Sara Borglin 1; Fábio Pereira 1 & Cintia B. Uvo 1 1  Department of Water Resources Engineering, Lund University, Lund, Sweden

Abstract The demand for biofuel has increased in recent years as more countries desire to reduce their dependence on fossil fuels. Therefore, the amount of sugarcane plantations has rapidly increased in Brazil, one of the largest producers of ethanol from sugarcane in the world. This increase raises concerns of what effects this replacement of native vegetation and traditional crops to sugarcane plantations may have on local hydrology and climate. In order to fill up this gap, this study aims to evaluate the effects of sugarcane expansion on surface runoff and evapotranspiration in the Rio Grande basin, Brazil. For the numerical experiments carried out in this study, scenarios of sugarcane were generated based on topographic features and mapping of suitable areas for sugarcane plantation made by the Brazilian Institute for Agricultural Research (EMBRAPA). These land use scenarios were provided as input to a distributed hydrological model, which estimated surface runoff and evapotranspiration rates for the river basin. Results from simulations showed that sugarcane expansion implied to a reduction of up to 10.8 % of surface runoff and an increase of evapotranspiration rate by 9.0 %. Key words – Sugarcane expansion, Rio Grande basin, MGB-IPH, Land use changes, Surface runoff, Evapotranspiration

Sammanfattning Efterfrågan på biobränsle har de senaste åren ökat i takt med att allt fler länder strävar efter att minska sitt ­beroende av fossila bränslen. Som en följd har antalet sockerrörsodlingar kraftigt ökat i Brasilien vilket har medfört en oro inför vilka effekter denna omvandling av ursprunglig mark till sockerrörsplantager kan ha på hydrologin och klimatet i en regional skala. För att fylla denna kunskapslucka syftar den här studien till att utreda utbredningen av sockerrörs påverkan på ytavrinning och avdunstning i Rio Grandes avrinningsområde, Brasilien. För de numeriska experimenten i studien genererades ett flertal sockerrörsscenarion baserade på topografiska egenskaper och, enligt det agrara forskningsinstitutet EMBRAPA, lämpliga områden för framtida sockerrörsodlingar. Dessa scenarior användes sedan som indata till en distribuerad hydrologisk modell som uppskattade ytavrinning och avdunstning för avrinningsområdet. Resultaten från simuleringarna indikerade att expansionen av sockerrörsodlingar kan ge upphov till en minskad avrinning med upp till 10.8 % och en ökning av avdunstningen med upp till 9 %.

Introduction The global warming and climate change debate in recent years can hardly have escaped anyone’s notice. Therefore, motivated by our common obligation to protect the environment and fear for what consequences a higher oil price could have on economic growth, more and VATTEN · 3 · 13

more countries have adopted policies to reduce their dependence on fossil fuels (Persson, 2006).   In Sweden and several other countries, the introduction of bioethanol has become a popular method to reduce the transport sector contribution to global warming. However, even though Sweden has a fairly good possibility to produce ethanol from paper pulp and cere141

75

Effects of sugarcane expansion on runoff and evapotranspiration in the Rio Grande basin, Brazil.

als, this production is still not efficient enough to cover the total demand. Accordingly, about 70 % of the ethanol used as fuel in Swedish cars is imported, a majority of the amount originating from Brazil (Energimyn­ digheten, 2011). The number of sugarcane plantations in Brazil has rapidly increased since the government launched a national bioethanol program, Pro-Álcool, as a solution for the oil crisis in 1973 (Goldemberg, 2006). From 1973 to 2005, the sugarcane plantations in the country reached 5.4 million hectares and they are predicted to further increase up to 12.2 million hectares by the end of 2015 (Bolling and Suarez, 2001, IEA, 2006).   The quick replacement of native vegetation by sugarcane plantations raises concerns of what effects of this rapid sugarcane expansion can have on the local and regional hydrological processes (Gedney et al., 2006). Changes in vegetation and land cover can, for example, affect infiltration, runoff, evapotranspiration, interception and other hydrological variables (Sampaio et al., 2007). The hydrological changes could have major consequences for the stakeholders in the river basin, such as hydropower suppliers and farmers. Since 76.9 % of the electricity generated in Brazil comes from hydropower (EPE, 2010) the country is vulnerable to changes in the hydrology. The Brazilian energy crisis in 2001, caused by a dry summer period in combination with low water storage in the hydropower dams, demonstrated how

sensitive the system really is (Krishnaswamy and Stuggins, 2007). The precipitation deficit causing the drought was just barely larger than for earlier droughts but still it led to a considerable larger runoff deficit (Simões and Barros, 2007).   Thus, the present study investigates the effects of ­sugarcane expansion on hydrology and local climate in the Rio Grande basin using a distributed hydrological model. The sugarcane expansion was expressed in terms of land use scenarios where the native vegetation and traditional crops were replaced by sugarcane. Two more realistic land use scenarios were also analyzed, of which one was provided by the Brazilian Institute for Agricultural Research (EMBRAPA) and the other was based on sugarcanes limitations to grow at certain areas in the river basin.

Study area The Rio Grande basin (Fig. 1) is located in the southeastern part of Brazil and covers approximately 145 000 km2 (Nóbrega et al., 2011). The landscape in Rio Grande basin can be described as hilly with elevations ranging from 200 m above sea level (m.a.s.l) at the basins outlet in the west to more than 1800 m.a.s.l at the Mantiqueira Mountains in the east. The climate has two distinct seasons with hot rainy summers and cold dry winters. Of the 1400 mm average annual precipitation,

Figure 1. Maps showing the location of the Rio Grande basin.

142

76

VATTEN · 3 · 13

Paper II

85 % falls under the austral summer and the average annual evapotranspiration is around 950 mm.   The river basin has become very important for Brazil as a source of electricity from hydropower. Approximately 12 % of the total hydropower produced in Brazil is generated along Rio Grande River together with its subsidiaries Mogi-Guaçu and Pardo (ANEEL, 2005). In the basin there are 15 hydropower plants (HPPs), four of them have a capacity to generate more than 1000 MW (Nóbrega et al, 2011). The Rio Grande River and its subsidiaries are, apart from generating hydropower, extensively used for irrigation of agricultural land and as a source of drinking water for the urbanized areas in the river basin.

Methodology The MGB-IPH model In the present study, the MGB-IPH hydrological model was used to study the effects of different land use scenarios on the hydrology in the Rio Grande River basin. The model was used in seven runs of 20 years of simulation (1990–2009) with a daily time step.   The MGB-IPH model is a distributed hydrological model developed for large scale basins which is based on VIC-2L (Liang et al., 1994) and LARSIM (Bremicker, 1998). It is equipped with modules for calculating soil water budget, evapotranspiration, flow propagation within a cell, and flow routing through the drainage network (Collischonn et al., 2007).   To represent the spatial distribution of sugarcane plantations the model divides the river basin into 10 sub basins. The sub basins are in turn divided into catchment cells interconnected by channels. Each catchment cell is divided into Grouped Response Units (GRUs), areas with similar combinations of vegetation and soil. The runoff generated from the different GRUs in each cell is summed and the flow is routed to the stream network using three linear reservoirs; surface flow, subsurface flow and groundwater flow (Nóbrega et al., 2011). The Muskingum-Cunge method is used for stream flow propagation through the river network (Allasia et al., 2006). For evapotranspiration calculations the PenmanMonteith equation based on air temperature, relative humidity, solar radiation, atmospheric pressure and wind velocity was used (Nóbrega et al., 2011). For a full description of the model, see Collischonn et al. (2007).   The model has earlier been successfully applied to the Rio Grande basin. Nóbrega et al. (2011) analyzed how future runoff in Rio Grande basin could be affected by possible climate changes. Their simulations indicated that runoff could increase with between 8 % and 51 % for a 1 to 6 degrees higher global mean temperature. VATTEN · 3 · 13

Parameters To describe hydrological processes over different types of soil, each sub basin has a number of adjustable parameters related to the soil water capacity and drainage rate for the different soil-vegetation combinations in the area, such as maximum water storage in soil, mean percolation and mean groundwater flow. Based on this, the model estimates the exchange between ground and surface so that infiltration, subsuperficial flow and groundwater contributions to the base flow are calculated. The adjustable parameters were calibrated in earlier work (Pereira et al., 2013a) by trial and error method using recorded hydrographs and relative stream flow volume error. The adjustable parameters for forest, agriculture of grain, water and pasture were set according to ranges recommended by Collischonn et al. (2007) and the parameters for sugarcane were estimated via calibration by Pereira et al. (2013a).   The model also uses “fixed” parameters that were not considered in the calibration process. They describe changes in vegetation over the year, such as leaf area index and plant height, to calculate the water fluxes between the atmosphere and land surface as evapotranspiration. These parameters are the same for all sub basins, and they are leaf area index, albedo, canopy resistance and height of trees. Values for these parameters were adopted according to ranges suggested by Collischonn (2007) and Nóbrega et al. (2011).

Input data Precipitation and discharge data has been used as input for calibrating and validating MGB-IPH parameters for sugarcane (Pereira et al., 2013a). In order to consider spatial heterogeneity for precipitation, 483 gauging stations with daily measurements spread across the river basin and its surroundings were collected from the Agência Nacional de Águas (ANA). Daily discharge data for the HPPs was provided by the Operador Nacional do Sistema Elétrico. The HPPs chosen for this study were Camargos, Funil, Furnas, Porto Colombia, Marimbondo and Agua Vermelha (Fig 2).   To calculate the evapotranspiration, additional data regarding air temperature, sunshine hours, wind speed, relative humidity and atmospheric pressure is needed. Three meteorological stations were considered in the study and the data was provided by the Centro de Previsão de Tempo e Estudos Climáticos (CPTEC). Two stations were located within the river basin and one, Araxá, north of the basin. Monthly averages were calculated for all meteorological variables of the three stations in earlier works by Pereira et al. (2013) and were given as input in the hydrological model. 143

77

Effects of sugarcane expansion on runoff and evapotranspiration in the Rio Grande basin, Brazil.

Figure 2. Map of hydropower plants in the Rio Grande basin.

  Elevation data for the Rio Grande basin was collected by the Shuttle Radar Topography Mission (SRTM) and was available from the Department of Ecology, University of Rio Grande do Sul. The US Geographical survey provides free access to satellite maps around the world. These maps have been used to classify land use as pasture, sugarcane, forest, agriculture of grain and water in previous works by Pereira et al. (2013a; b). In order to classify the soil in the basin, Pereira et al. (2013a; b) used a soil type mosaic based on soil survey data created by the RADAM Brasil project (RADAMBRASIL, 1984) and FAO database (1972).

Scenarios of sugarcane expansion For the simulations, six sugarcane scenarios were generated based on satellite images and topographic features of the Rio Grande basin, four of the scenarios considered “gradual expansion” and two “realistic future expansion”. Each scenario contained information of vege-

Figure 3. Map of sugarcane distribution for the control scenario.

144

78

tation type, soil type and elevation. A land use scenario (Fig 3) representing true vegetation and soil condition in 1993, was used as a reference (control scenario) when calculating the changes in discharge and evapotranspiration for the different scenarios. The year 1993 was ­chosen as reference because it is the earliest land use scenario we could establish for the area.   The gradual expansion scenarios were generated to investigate how sensitive the hydrology of Rio Grande basin is for conversion of traditional land use to sugarcane. These scenarios were later used to analyze in which pace the discharge and evapotranspiration changes. For the four expansion scenarios, sugarcane gradually expanded to higher elevations in the river basin. The four different expansion scenarios were; 340–500 m.a.s.l, 340–700 m.a.s.l, 340–900 m.a.s.l and 340–1100 m.a.s.l.   To estimate possible future changes in runoff and evapotranspiration in the Rio Grande basin related to the ongoing sugarcane expansion, two realistic scenarios were generated. The first realistic scenario was based on data describing suitable areas for future sugarcane plantations in the river basin according to EMBRAPA. ­EMBRAPA is the Brazilian Institute for Agricultural Research with focus on developing and solve problems within agriculture (EMBRAPA, 2008).   An alternative future scenario was generated based on sugarcane limitations to grow at certain areas in the river basin. For this scenario, sugarcane plantations expanded up to 700 m.a.s.l where the slope is lower than 12 %. This limitation was chosen due to the cooler temperature and rocky landscape implied by higher altitudes, and because steeper slopes prevent mechanized sugarcane harvesting (Sparovek et al., 1997). The gradual expansion scenarios and the realistic scenarios are shown in Figure 4. VATTEN · 3 · 13

Paper II

Figure 4. Maps of sugarcane distribution for gradual expansion scenarios 340–500, 340–700, 340–900 and 340–1100 and realistic expansion scenarios Embrapa and 340–700_slope.

Output files and data adaptation Except for the land use maps that changed for the different scenarios, the same input data and parameters were used for all runs. For every scenario, the model run generated two output files which contained information on daily discharge and evapotranspiration rate for the catchment cells contributing to the hydropower plants; Camargos, Funil, Furnas, P. Colombia, Marimbondo and A. Vermelha. The relative changes between estimated discharge and evapotranspiration for each scenario and the control scenario (1993) were then analyzed and presented. These changes were calculated and the result was plotted in graphs over a 5-year period (2000–2004). ­Finally, relative changes in total discharge volume for the entire simulation period were calculated and presented in tables.   The other output file contained data on daily evapotranspiration rate for all catchment cells in the river basin. This data was processed and the average eva-

potranspiration rate for the entire area upstream each HPP was calculated. The relative changes in eva­ potranspiration rate for the scenarios were presented in graphs and the relative changes in total evapotranspiration for the entire simulation period were summarized in tables.

Results and discussion Table 1 summarizes the changes in relative discharge and evapotranspiration that resulted from the model runs for all scenarios.   The effects of sugarcane expansion were most noticeable for the severest scenario at HPP Funil where the discharge decreased with 10.8 % and the evapotranspiration for the area upstream increased with 9.0 %.   The impact on discharge for the hydropower plants situated in eastern parts of Rio Grande basin is more

Table 1. Relative changes in discharge and evapotranspiration for a 20 year period (1990–2009).

Camargos   dV20 dE20 (%) (%)

Funil     dV20 dE 20 (%) (%)

Furnas     P. Colombia  dV20 dE 20 dV20 dE20 (%) (%) (%) (%)

Marimbondo  A. Vermelha   dV20 dE 20 dV20 dE 20 (%) (%) (%) (%)

340–500   0.0 340–700   0.0 340–900   0.0 340–1100 –6.2

0.0    0.0 0.0    0.0 0.0   –0.8 5.4 –10.8

0.0   0.0 0.0   0.0 3.3 –3.9 9.0 –9.3

0.0 0.0 4.1 7.1

–0.2   0.5 –1.1   0.8 –4.9   4.2 –9.4   6.7

–0.2   0.5 –1.2   0.8 –3.7   4.2 –6.4   6.7

–0.8 –2.0 –4.3 –6.7

0.8 1.4 3.4 4.9

340–700_slope   0.0 Embrapa   0.0

0.0    0.0 0.0   –0.7

0.0   0.0 2.4 –1.8

0.0 2.6

–1.1   0.7 –2.5   2.5

–1.2   0.7 –2.0   2.5

–2.0 –2.3

1.4 2.2

VATTEN · 3 · 13

145

79

Effects of sugarcane expansion on runoff and evapotranspiration in the Rio Grande basin, Brazil.

Figure 5. Relative changes in discharge at the HPPs for the 340–1100 and Embrapa scenarios (2000–2004).

severe than for the HPPs located in the west. This is related to the land use in the different parts of the river basin. In the eastern mountainous part, the vegetation mainly consists of pasture while in the western parts agricultural land is more common. Converting areas with pasture to sugarcane fields gives a higher evapotranspiration rate, and thus less runoff, compared to conversion from agricultural land to sugarcane. The relative changes in discharge and evapotranspiration for the simulations are shown in Figures 5 and 6, respectively.   The changes in surface runoff and evapotranspiration rate for the sugarcane converted land are not constant 146

80

throughout the simulation period. Effects of the sugarcane annual cycle can be clearly seen in the evapotranspiration graphs (Fig 6). From being larger than the control scenario in the beginning of the year, the evapotranspiration rapidly decreases and becomes significantly lower after the sugarcane harvest. When new crops start to grow, the evapotranspiration rate recovers to the same level it was before harvest.   Regarding the surface runoff, the decrease in discharge is greater during the austral winter but recover most of the loss during the storms in the summer. The soil water capacity in sugarcane fields on shallow soils is the same, VATTEN · 3 · 13

Paper II

Figure 6. Relative changes in evapotranspiration at the HPPs for the 340–1100 and Embrapa scenarios (2000–2004).

or lower, than for the other vegetation types on shallow soils. This actually increases the runoff during the rainy summer months, as the soil store less water, which can be seen in the results for the HPPs in the eastern part of the basin. Consequently, the discharge increases in the austral summer months due to the heavy rainfalls, and decreases during the austral winter when the evapotranspiration increases. This seasonal variation is unfortunate as the decrease in runoff coincides with the dry winter season when the water level at the HHP reservoirs is low. Therefore, even if the decrease in runoff for the total simulation period is small, it could increase the VATTEN · 3 · 13

risk for low reservoir levels during possible summer droughts significantly.   For the realistic scenarios, the largest changes in runoff occurred at HPP P. Colombia where the discharge decreased with 2.5 %. The evapotranspiration for the realistic scenarios increased most for the area upstream HPP Furnas with 2.6 %.   Considering sensitivity of the hydrology in Rio Grande basin, it can be assumed that the ongoing sugarcane expansion will request adjustment in the management of HHP to avoid damage on the hydropower ­generation. 147

81

Effects of sugarcane expansion on runoff and evapotranspiration in the Rio Grande basin, Brazil.

Conclusions Overall, simulations showed that the replacement of native vegetation by sugarcane fields has a significant impact on the hydrology and climate of the Rio Grande basin. Gradually expanding sugarcane in the river basin resulted in a very clear trend of decreased surface runoff and increased evapotranspiration. This is because the sugarcane expansion changed the hydrological behaviour of the basin over the year according to its phenological cycle, which implied variations of surface runoff and evapotranspiration rates.   In terms of the surface runoff, the influence of sugarcane expansion depended upon a combination of factors, such as amount of areas replaced with sugarcane, type of land use replaced and location of the expansion within each basin. This study showed that the sugarcane expansion mostly impacts the surface runoff if it happens over headwater areas consisting of pasture land with low soil water retention as in the sub basins located in the eastern part of the Rio Grande basin. For this area, sugarcane expansion significantly reduced the surface runoff.   Regarding the evapotranspiration, the results of the simulations showed that sugarcane expansion mostly ­affect same areas as runoff e.g. eastern part of the Rio Grande basin. The evapotranspiration rate also increased significantly from this area when sugarcane expanded over the original pasture land. However, the changes in evapotranspiration rate were not constant throughout the year, the harvest reduced evapotranspiration considerably and it was not restored until new sugarcane crops started to grow.

Acknowledgements We would like to thank the Brazilian Institute for Agricultural Research, EMBRAPA, for the providing the scenario of possible areas for sugarcane plantation. Fabio Pereira and Cintia Bertacchi Uvo thank the Swedish Research Council (Vetenskapsrådet) for support. This work is also integrating part of the MERGE project.

References Allasia, D. (2007) Avaliação da previsão hidroclimática no Alto Paraguai. Doctoral Thesis. Instituto de Pesquisas Hidráulicas. Universidade Federal do Rio Grande do Sul, Porto Alegre. 235pp.

148

82

Bremicker, M. (1998) Aufbau eines Wasserhaushaltsmodells für das Weser und das Ostsee Einzugsgebiet als Baustein eines Atmosphären-Hydrologie-Modells. Dissertation Doktorgrad, Geowissenschaftlicher Fakultät der Albert­ Ludwigs-Universität. Freiburg. Germany. Collischonn, W., Allasia, D. G., Silva, B. C., Tucci, C. E. M. (2007) The MGB-IPH model for large-scale rainfall-runoff modeling, Hydrology Sciences Journal, v. 52, p. 878– 895. EMBRAPA (2008) Embrapa. [online] Available at: [Accessed 21 November 2012]. Energimyndgheten (2011) Analys av marknaderna för etanol och biodiesel Redovisning av uppdrag 15 i regleringsbrevet för 2011. ER 2011:13. EPE (2010) Balanço Energético Nacional 2010: Ano base 2009/Brazilian Energy Balance 2010: Year 2009. [online] Available at: [Accessed: 1 May 2013]. FAO (1974) Soil map of the world, Tech. rep., UNESCO, Paris. Goldemberg J. (2007) Ethanol for a sustainable energy future. Science, 315, 808–810. Krishnaswamy V. and Stuggins G. (2007) Closing the Electricity Supply-Demand Gap. Energy and mining sector board discussion paper. pp. 43–50. Liang, X., Lettenmaier, D.P., Wood, E.F., Burges, S.J. (1994) A simple hydrologically based model of land surface water and energy fluxes for general circulation models. Journal Geophysical Research, v. 99(7), p. 14415–14428. Nóbrega, M.T., Collischonn, W., Tucci, C.E.M., and Paz, A.R. (2011) Uncertainty in climate change impacts on water resources in the Rio Grande Basin, Brazil, Hydrol. Earth Syst. Sci., 15, 585–595, doi:10.5194/hess-15-585-2011. Pereira, F.F., Tursonov M., Uvo, C.B. (2013a) Towards the response of water balance to sugarcane expansion in the Rio Grande Basin, Brazil. Submitted to Journal of Hydrology. Pereira, F.F., Tursonov M., Uvo, C.B. (2013b) Quantifying the rapid sugarcane expansion for Ethanol production in the Rio Grande Vasin, Brazil. VATTEN – Journal of Water Management and Research 69:83–86. Persson, G. (2006) På väg mot ett OLJEFRITT Sverige. Kommissionen mot oljeberoende. [online] Available at: [Accessed 20 November 2012]. RADAMBRASIL (1982) Programa de Integração Nacional, Levantamento de Recursos Naturais (Folhas SE-23 (Belo Horizonte) e SF-22 (Paranapanema)), Tech. rep., Ministério das Minas e Energia, Brasília. Simões, S.J.C. and Barros A.P. (2007) Regional climate variability and its effects on Brazil´s 2001 energy crisis. Management Environmental Quality. 18(3). pp. 263–273. Sparovek G., Pereira J.C., Alleoni L.R.F., Rosetto R. (1997) Aptitude of lands of Piracicaba (SP) for sugarcane mechanical harvest. Stab 15:6–9.

VATTEN · 3 · 13

PAPER III Towards the response of water balance to sugarcane expansion in the Rio Grande basin, Brazil F. F. Pereira, M. Tursunov and C. B. Uvo. Hydrology and Earth System Sciences, Reviewed.

Paper III

Manuscript prepared for Hydrol. Earth Syst. Sci. with version 5.0 of the LATEX class copernicus.cls. Date: 5 August 2013

Towards the response of water balance to sugarcane expansion in the Rio Grande basin, Brazil F. F. Pereira1 , M. Tursunov2 , and C. B. Uvo1 1 2

Department of Water Resources Engineering, Lund University, Lund, SE Tashkent Institute of Irrigation and Melioration, Qori-Niyoziy, UZ

Correspondence to: F. F. Pereira ([email protected])

5

10

15

20

25

30

Abstract. This study explores the short-, medium- and long-term impacts of expansion of the sugarcane plantation on the water balance of the Rio Grande basin, Brazil, as estimated by changes in evapotranspiration, soil moisture content and surface runoff calculated by a hydrological model. Twenty years of simulation are made using historical land use of the basin that include area planted with sugarcane in 1993, 2000 and 2007 as estimated from satellite images. Complementary, it is used a scenario for sugarcane plantation defined by the Brazilian Institute for Agricultural Research (EMPRAPA) as all areas suitable for sugarcane cultivation within the Rio Grande basin. In addition, parameters for sugarcane fields were specifically defined via calibration and validation of the hydrological model for all growth phases based on the annual cycle of sugarcane phenology in the Rio Grande basin. According to results from the land use classification of satellite images, the expansion of sugarcane fields mostly replaced pasture lands. Modelling results for short-, mediumand long-term clarify that impacts of this expansion depended not only on the amount of areas planted with sugarcane, but also on the type of land use replaced, location of the expansion within the basin and regional soil properties. Largest impacts on the water balance are observed if areas located close to headwaters with low soil water capacity are planted with sugarcane. In case all areas suitable for sugarcane plantation, as defined by EMBRAPA will actually be planted, simulations showed that the annual accumulated values of evapotranspiration nearly doubled while surface runoff is reduced to 20% of the values calculated using the historical land use from 1993.

1

35

40

45

50

55

60

65

Introduction

The search for sustainable sources of energy has found a realistic replacement to fossil fuels in ethanol and methanol. Nowadays, the production of ethanol from sugarcane is among the most effective and sustainable techniques for making ethanol from food crops, particularly when compared to the production of ethanol from other commercial crops (e.g. wheat, corn and barley). The reason for this is that sugarcane grows at a faster rate than other crops (Herrera, 1999), and can be cultivated with many different farming practices, which opens up possibilities for enhancing productivity but protecting the environment(AgSri, 2012; Maraddi, 2006; Mui et al., 1996). In response to these properties and its high potential to become a renewable energy source, many countries have significantly increased their sugarcane production during the last two decades (e.g. China, India, Brazil). In this context, Brazil is the country that retains the largest area of sugarcane cultivation in the world. It is responsible for approximately one third of the global harvested area and production (Zuurbier ´ and van de Vooren, 2008). Since 1975, when the Pro-Alcool (ethanol program) began as a response to the 1973 oil crisis (Borges and Almeida, 1985), the Brazilian area of sugarcane plantation increased by 170%, reaching 5.4 million hectares in late 2005 (Nitsch, 1991; Bolling and Suarez, 2001; IEA, 2006). One of the negative environmental consequences by increasing the amount of a particular land use type in a catchment may be its possible impact on regional hydrological processes (Gedney et al., 2006; Sampaio et al., 2007). Addressing this question, many studies have recently been developed to estimate the effects of land use changes on local water balance. Hlavcova et al. (2009), for example,

85

Towards the response of water balance to sugarcane expansion in the Rio Grande basin, Brazil

2

70

75

80

85

90

showed the impacts of land use changes on maximum daily discharges in a catchment considering both rainfall and snowmelt as incoming water. In addition, Warburton et al. (2011) have applied a hydrological model over three catch- 120 ments with different land covers. They showed that the runoff generation in the three catchments were closely related to their geographic distribution of land use. Finally, a hydrologic model was also used by Wijesekara et al. (2012) as a tool to analyze the expansion of built-up areas according to 125 land use predictions, and, by means of these analyses, they could estimate variations in surface runoff, evapotranspiration, baseflow and infiltration for an urban catchment. Despite many insightful studies on land use changes affecting the surface hydrology, large speculations are still be- 130 ing made about such changes. In Brazil, for instance, impacts of the rapid expansion of sugarcane on surface runoff after ´ the Pro-Alcool were not carefully investigated since sugarcane fields were not completely mapped (Cheesman, 2004; James, 2008). In order to fill up this gap, this work aims to map sugarcane fields and their expansion during the past 20 years in a Brazilian river basin. Moreover, by using this mapping, it is also intended to estimate the response of water bal- 135 ance to such expansion. Since most of the recent sugarcane expansion occurred in S˜ao Paulo state, Rio Grande basin was chosen as a case study. 2

95

100

105

110

115

86

F. F. Pereira et al.: Towards the response of water balance to sugarcane expansion

140

Study area

Rio Grande basin is a sub-basin of the Paran´a basin formed by the rivers Grande, Pardo, Sapuca´ı, Verde, das Mortes and Mogi-Guac¸u. It has an area of 145000 km² located in the eastern upper Paran´a basin (Fig. 1) where altitudes vary from 145 300 to 2700 m.a.s.l.. The classification of land use in the Rio Grande basin includes three distinct categories: Atlantic Rainforest, pasture and agriculture (IBGE, 1991). Agricultural activity represents a large portion of the Rio Grande basin and, for this study, it was classified into sugarcane and agriculture of grain. Regarding types of soils, Rio Grande basin presents five major types: latosols, lithosols, cambisols, podzolics and al150 luvial soils which may be broken down into three groups: high, medium and low infiltration capacity (FAURGS, 2007). Soils with high and medium infiltration capacity are equally distributed across the Rio Grande basin whereas soils with low infiltration capacity are concentrated along the drainage 155 network. Figure 1 Although most of surface runoff in the Rio Grande basin is regulated by dams, its hydrological regime is strongly 160 induced by land use changes due to harvesting practices and shifting cultivation (WWFBrasil, 2008). After the flow regulation, a representative sample of daily values of discharge collected at the outlet of the basin, from 1970 to

2010, indicates that surface runoff varies from minimum values of 1000m/s (dry season) to maximum values over 12000m/s (rainy season). Locally, measurements of runoff are also monitored at hydroelectric power plants. At Funil, Camargos, Furnas, P Colˆombia, Marimbondo and A Vermelha power plants, daily runoff ranges 70 – 3731m3 /s, 34 – 1253m3 /s, 174 – 7497m3 /s, 251 – 8367m3 /s, 532 – 9234m3 /s and 303 – 10186m3 /s, respectively. Production of electrical power is the largest water use in the Rio Grande basin (IPT, 2008). Over 11% of the installed electric generation capacity of Brazil is at hydroelectric installations in the Rio Grande basin (ANEEL, 2005). To meet this demand for electricity, hydroelectric power plants are constrained by a minimum operating flow, which varies from power plant to power plant. As recently proposed by (ONS, 2013), the minimum operating flow at all hydroelectric power plants used in this study are shown in table 1. Table 1 According to Espinosa (2011), spatial and temporal distribution of rainfall in the Rio Grande basin is highly induced by synoptic systems over the southeastern and southcentral Brazil. In addition, annual rainfall analysis carried out by CPRM (2012) indicate that annual average rainfall varies from 1500 to 2000mm in the basin. Annual average evapotranspiration ranges from 800 to 1000mm (Ruhoff, 2011). Throughout the year, a seasonal variability of evapotranspiration has been identified by Rocha et al. (2002). Over the Rio Grande basin, their studies revealed that daily evapotranspiration can oscillate between 6mm/d−1 in the rainy season and 1mm/d−1 in the dry season.

3

Methods

In this study, impacts of sugarcane expansion on river basin water balance were locally and globally estimated for the Rio Grande basin using a spatially distributed hydrological model. To reproduce the sugarcane expansion in the Rio Grande basin during the latest 20 years, historical land use of the basin were defined based on satellite images and are compared to a scenario based on the mapping of areas suitable for cultivation of sugarcane made by the Brazilian Institute for Agricultural Research - EMBRAPA (BRASIL, 2009). The extension of sugarcane fields over the basin were identified from satellite images captured in 1993, 2000 and 2007. Once calibrated, the model was used in four runs of 20 years of simulation. Except for the land use map, the same input data was used for each of these runs. Finally, model estimates of evapotranspiration, soil moisture and surface runoff were evaluated at different time scales.

Paper III

F. F. Pereira et al.: Towards the response of water balance to sugarcane expansion 165

170

175

180

185

190

195

200

205

210

215

3.1 Distributed Hydrological Model MGB-IPH is a large scale distributed hydrological model (Collischonn, 2001) conceptually based on the LARSIM 220 (Bremicker, 1998) and VIC-2L (Liang et al., 1994) models. It consists of modules for calculating soil water budget, evapotranspiration estimation, surface and subsurface flow generation, which are interconnected by river routing. Full details of all hydrological processes and how these processes are in- 225 corporated into MGB-IPH can be found in Collischonn et al. (2007). MGB-IPH provides the choice of smaller sub-basins (Paiva et al., 2011) or uniform grid cells (Paz et al., 2011) as computational units. For both methods, MGB-IPH divides 230 each computational cell in hydrologic response units (HRUs) based on its land use/cover and soil distribution. HRUs are then defined by intersecting land use and soil groups within a computational cell. Once all computational cells are classified into different 235 groups with similar hydrological response, MGB-IPH calculates the soil water budget, evapotranspiration and flow propagation using adapted versions of the ones presented in LARSIN and VIC-2L models. These adaptations were made in order to facilitate its applications in large tropical basins. 240 MGB-IPH generates surface flow by direct precipitation on saturated areas and subsurface flow comes from the nonlinear relationship between texture, hydraulic conductivity and moisture of soil proposed by Rawls et al. (1993). A linear reservoir concept is used to propagate surface and subsur- 245 face flow using different retention times along every computational cell. After passing through the linear reservoirs, surface and subsurface flows are summed and routed from cell to cell along the river network using the Muskingum-Cunge method. 250 MGB-IPH has been tested and used in several South American basins, from rapid-response ones of southern Brazil and Uruguay to very low response ones as the Pantanal. It has also been applied for several purposes, such as to predict runoff (Tucci et al., 2008), to estimate daily water balance in large basins (Collischonn et al., 2008) and to anal- 255 yse the impacts of climate changes upon river flow (Paiva and Collischonn, 2010). By default, MGB-IPH is employed using a daily time step. However, its time step may fluctuate depending on the purpose of study. In this work, MGB-IPH was used to simulate 260 rainfall-runoff processes on a daily basis.

3

evation maps together with meteorological time series such as rainfall and evaporation. For the calibration mode, in contrast, discharge time series must be aggregated to the input dataset. The necessary digital elevation model (DEM) was freely obtained at Department of Ecology of the Federal University of Rio Grande do Sul. Their DEM preprocessing includes data gap filling and mosaicking of elevation data from the Shuttle Radar Topography Mission (SRTM) for the whole Brazil (Hasenack et al., 2010). The soil map of the Rio Grande basin was derived from a soil survey data created by RADAM Brasil project (RADAMBRASIL, 1982) at scale of 1:1000000. Although at a larger scale than RADAM Brasil soil survey data (1:3000000), digitalized soil maps from the Food and Agriculture Organization of the United Nations (Food and of the United Nations, 1974) were resampled and used to overcome missing data. The RADAM Brasil database includes over 12 different types of soils in the Rio Grande basin, (FAURGS, 2007) has then been reclassified into two groups as deep and shallow soils according to their hydrological behavior and provided to us (personal communication). Rainfall depths were selected using the Agˆencia Nacional ´ de Aguas (ANA) database (http://hidroweb.ana.gov.br/). Observed daily precipitation values were obtained from 483 precipitation stations over the River Grande basin and its surroundings. Daily rainfall depths were then spatially interpolated by the inverse distance-squared weighted method at the centroid of each MGB computational cell. Further meteorological data sets were taken from three meteorological stations and they include monthly time series of air temperature, sunshine, relative humidity, wind speed and atmospheric pressure (Fig. 2). These data sets were provided by Centro de Previs˜ao de Tempo e Estudos Clim´aticos (CPTEC). Figure 2 MGB-IPH was calibrated for the Rio Grande basin using calibration mode. For doing so, discharge stations were also used as input. Daily discharge data were collected from six gauging stations at six hydropower plants along the drainage network: Camargos, Funil, Furnas, P. Colˆombia, Marimbondo and A. Vermelha (Fig. 2). All discharge data were found at Operador Nacional do Sistema El´etrico (ONS) webpage (http://www.ons.org.br/home/). Both discharge and meteorological time series cover a period ranging from 1970 to 2010.

3.2 Data Collection and Processing

3.3

In general, hydrological models require plenty of data, which quite often need to be pre-processed before they are used as input. MGB-IPH may be performed in simulation and cali- 265 bration modes, each of these modes present different input data. For the simulation mode, MGB-IPH is dependent upon spatially distributed data that include land use, soil and el-

Multi-temporal Landsat images were used for the characterization of sugarcane evolution in the Rio Grande basin. Land use maps were derived through analysis of satellite images made by Landsat TM 7 and extracted from U.S. Geological Survey. The selection of satellite images was driven by the availability of cloud-free Landsat data over the Rio Grande

Mapping of sugarcane plantations

87

Towards the response of water balance to sugarcane expansion in the Rio Grande basin, Brazil

4

270

275

280

285

F. F. Pereira et al.: Towards the response of water balance to sugarcane expansion

basin from 1970 to 2010. Fourteen Landsat satellite images (170x183km) were captured in 1993, 2000 and 2007, and used to generate three land use maps. An automatic classification of Landsat satellite images showed in Rudorff et al. (2010) was used for mapping sug- 325 arcane fields. This automatic classification is based on a linear spectral mixing which consists of a linear combination of spectral signatures from two or more types of land use, such as agriculture, pasture, forest etc. The particular sugarcane spectral signature as presented by Aguiar et al. (2011) has 330 been used to identify sugarcane plantations in the Rio Grande basin. Moreover, a visual inspection was made to support this automatic land use classification. Figure 3 shows the results the automatic identification of areas of sugarcane plantation in 1993, 2000 and 2007. In addition, it is shown the suitable 335 areas for sugarcane plantation in the basin as defined by the Brazilian Institute for Agricultural Research (EMBRAPA). Figure 3

290

295

Each land use map was classified into five dominant types 340 as areas covered by water bodies, Atlantic Rainforest, agriculture of grain crops, pasture lands and sugarcane plantations according to their spectral signatures. Except for sugarcane plantations, all spectral signatures were adopted as defined by Mendes and Cirilo (2001) and only mapped 345 for 1993. Although it may appear as a limitation, BRASIL (2009) and (FAURGS, 2007) revealed that land use changes due to shrinkage/expansion of areas covered by pasture, cereals or forest are marginal compared to sugarcane expansion 350 in the Rio Grande basin over the past two decades. 3.4

300

305

310

315

320

88

MGB-IPH specific parameters for sugarcane: calibration and validation

MGB-IPH presents a set of parameters related either to 355 types of soil or land use that are adapted for different river basins during the calibration process. To represent hydrologic processes over different types of soil, MGB-IPH uses soil parameters such as maximum water storage in the soil, mean percolation and average groundwater flow. Based on 360 this, MGB-IPH estimates the exchange of water between ground and surface so that infiltration, sub-superficial flow and groundwater contributions to the baseflow are estimated. On the other hand, land use parameters such as leaf area index, canopy resistance, albedo and average height of trees are used by MGB-IPH to calculate the water fluxes between 365 atmosphere and land surface as evapotranspiration. The calibration of the MGB-IPH model is usually done by trial-anderrors adjustments of these parameters. In order to reduce the number of parameters to be calibrated and to identify the key parameters controlling the 370 model behavior when simulating surface runoff, sensitivity analyses were carried out by Collischonn (2001). Based on these analyses, seven parameters were identified as being the most important during calibration. They are called adjustable

parameters and are the maximum water storage, mean percolation, hydraulic conductivity, average groundwater flow, upward water flux and a shape parameter that regulates surface runoff based on the soil storage capacity. Although leaf area index, albedo and average height of trees were not found as sensitive as the other seven adjustable parameters previously listed, they are carefully defined for each type of land use and their seasonal variability is also taken into consideration. As these parameters are not consider in the calibration, they are referred to as fixed parameters. MGB-IPH has been successfully calibrated and validated for the Rio Grande basin by N´obrega et al. (2011). Their calibration includes sets of fixed and adjustable parameters for Atlantic Rainforest, areas covered by water bodies, pasture lands and agriculture of grain crops according to ranges recommended by Collischonn et al. (2007). Differently from previous application of the MGB-IPH, this work pays special attention to the effects of the plantation of sugarcane, as such, special parameters representing this culture were estimated taking into consideration the characteristics and timing of the sugarcane planting and harvesting in Brazil. In Australia, Robertson et al. (1999) carried out field experiments to evaluate consequences of water deficit on sugarcane productivity. Their analysis included the estimation of leaf area index for sugarcane in four different growth phases: germination (0-30 days), tillering (30-120 days), grand growth (120-270 days) and maturation (270-360 days). In addition, Andr´e et al. (2010) studied the radiation balance over sugarcane plantations and defined values for albedo and height of trees for each sugarcane growth phase. They also concluded that albedo varies proportionally to leaf area index in sugarcane fields. Table 2 shows fixed and adjustable parameters adopted in this study. Sugarcane planting and harvesting timing were defined using analysis made in BRASIL (2009) for sugarcane plantations located in the Southeast Region of Brazil. These analysis indicated that higher values of leaf area index, height of trees and albedo are predominant during the maturation stage from March to May whereas lower values are concentrated over the germination stage in June. Table 2 The adjustable parameters for sugarcane were estimated via calibration. Although disposal measured data spans a period of 40 years, only the 20 most recent years are used for calibration and validation. The calibration was performed for a eleven-year period (1990-2000), and consisted in finetuning the adjustable parameters by comparing calculated and observed discharges using relative volume error (RVE), Nash-Sutcliffe coefficient (NS) and root-mean-square error (RMSE) as efficiency criteria. Moreover, the set of the adjustable parameters for sugarcane defined during the calibration were validated over the seven-year period 2001-2007. For the calibration of sugarcane parameters, Rio Grande basin was first divided into six smaller sub-basins where each

Paper III

F. F. Pereira et al.: Towards the response of water balance to sugarcane expansion 375

380

385

390

395

400

405

sub-basin has a correspondent gauging station at its outlet. Also, as the land use map of 1993 was chosen as the control scenario, it has been used for both calibration and validation. 430 The calibration was performed only for those sub-basins where sugarcane fields represented a significant portion of their drainage area. In this study, particularly, this portion was set to 15% of the sub-basin since adjustments of parameters in sub-basins with areas covered by sugarcane fields 435 lower than 15% did not present significant changes in their hydrographs. For these sub-basins, parameters for sugarcane were set equal to parameters used to represent agriculture of grains and considering the annual cycle of sugarcane phenology instead. Thus, MGB-IPH parameters for sugarcane were calibrated for P Colˆombia, Marimbondo and A Vermelha sub-basins. Their gauging stations are approximately located 440 10 km upstream from the reservoirs in order to prevent backwater effects on measured discharges FAURGS (2007). Measured discharges were then used to evaluate estimates of the MGB-IPH using different sets of adjustable parameters. During the calibration, MGB-IPH ran simultaneously in 445 both calibration and simulation mode. Since the adjustable parameters were individually calibrated for each sub-basin, MGB-IPH ran in calibration mode when calibrating downstream sub-basins and, at the same time, it ran in simulation 450 mode at upstream sub-basins. By fine-tuning, the adjustable parameters for sugarcane fields were calibrated and are shown in table 2. In addition, values of discharge simulated by MGB-IPH using these parameters during the whole calibration period were compared to observed hydrographs at P Colˆombia, Marimbondo and A 455 Vermelha gauging stations (Fig. 4). As performance criteria for this calibration, table 3 presents NS coefficients, RMSEs and RVEs computed for each sub-basin. Figure 4 and table 3

410

415

420

425

460

Results from the calibration revealed that MGB-IPH could correctly reproduce the hydrological regime of all sub-basins which presented areas covered by sugarcane fields over the 10-year calibration period. Although baseflow recessions 465 were slightly overestimated by MGB-IPH, peak flows and rising and falling limbs of the simulated hydrographs closely matched the observed hydrographs which may be noticed by minor RMSEs and RVEs. Morevover, NS coefficient values up to 0.9 were obtained for all sub-basins which indicated a good agreement between observed and simulated discharges. In order to validate the parameters previously calibrated, MGB-IPH was applied to the Rio Grande basin using a different meteorological data set which spans from 01 January 470 2001 through 31 December 2007. Again, RVE, NS coefficient and RMSE were used to measure the quality of the fitting (Tab. 3). Figure 5 shows the simulated and observed hydrographs at P Colˆombia, Marimbondo and A Vermelha 475 gauging stations. Figure 5

5

Overall, results of the validation showed that MGB-IPH could provide a good agreement with observed data using the adjustable parameters for sugarcane defined in the calibration. Although NS coefficient values, RMSEs and RVEs pointed to a small decrease in the quality of the fit compared to the calibration, they still remained in the ranges of 0.85 to 0.88 for NS coefficient, 300 to 510 m3 /s for RMSEs and 20 to 23% for RVEs. MGB-IPH showed to be satisfactorily capable of reproducing peak flow patterns and seasonal recession for all sub-basins covered by sugarcane fields. 3.5 Land use scenarios and model runs To represent the expansion of sugarcane plantation at the Rio Grande basin, historical land use of the basin generated from Landsat satellite images in 1993, 2000 and 2007 were used. Also, an additional land use scenario was generated based on the mapping of all areas suitable for sugarcane crops as defined by EMPRAPA (BRASIL, 2009). The four land use scenarios were used as input for four different model runs which were performed with daily time step covering the period of January 1st 1990 to December 31st 2010. All runs were preceded by a warming-up period of one year (January 1989 - December 1989), which means a period often used in simulations to let physical parameters reach realistic conditions. The run, which incorporated the land use map of 1993, was considered as the control run and together with the runs that included land use scenarios of 2000, 2007 and from the EMBRAPA mapping will be respectively called CR1993, R2000, R2007 and REMBRAPA hereafter. The difference between each scenario and the CR1993 represents the expansion of the area planted with sugarcane over the basin. Surface runoff were calculated in CR1993, R2000, R2007 and REMBRAPA, and statistically compared to each other. Statistical comparisons of surface runoff were made by means of bootstrap analyses based on 1000 resamplings using a 99% confidence interval as described in Lall and Sharma (1996). In addition, a better assessment of the impacts of sugarcane expansion was achieved by comparing surface runoff generated in sub-basins with different concentrations of sugarcane plantations.

4

Results and discussion

In this section, an overview of the sugarcane expansion as estimated by Landsat satellite images captured in 1993, 2000 and 2007 is presented. Results from the land use classification of these satellite images are discussed for each subbasin of the Rio Grande basin. Moreover, short-, mediumand long-term impacts of sugarcane expansion on the water balance of the Rio Grande basin were separately evaluated. In this study, short-term impacts of sugarcane expansion on the hydrological cycle are investigated by bootstrap anal-

89

Towards the response of water balance to sugarcane expansion in the Rio Grande basin, Brazil

6

480

F. F. Pereira et al.: Towards the response of water balance to sugarcane expansion

yses on variations in surface runoff at daily temporal scale. 530 For the medium- and long-term, the variability of surface runoff, evapotranspiration and soil water content are assessed at inter-annual and decadal temporal scales, respectively. 4.1 An overview of the sugarcane expansion in the Rio Grande Basin 535

485

490

In general, altitude and terrain slope were equally important factors that drove sugarcane expansion between 1993 and 2007 in the Rio Grande basin. Since sugarcane plantations do not tolerate frosts (Eggleston et al., 2004; Tai and Miller, 1993), altitude appeared as a limiting factor which restricted 540 sugarcane expansion to areas below 700 m.a.s.l.. In addition, mechanical harvesting and transport facilities directed sugarcane expansion to regions of terrain slope less than 12%. The evolution of areas covered by sugarcane plantations is shown in table 4 for each sub-basin. 545 Table 4

495

500

505

510

515

520

525

90

From 1993 to 2007, table 4 reveals very little or no sugarcane plantations over Funil, Camargos and Furnas subbasins. Characterized by high elevations, these sub-basins 550 present low temperatures that may reach 8◦ C in the austral summer. Under such climate conditions, sugarcane productivity would negatively be affected by low temperatures, which induce damage to young leaves and lateral buds. This makes Funil, Camargos and Furnas less attractive to grow sugarcane. On the other hand, further downstream, areas covered by 555 sugarcane represent up to 27.9% of the Marimbondo subbasin already in 1993. In addition, areas for growing sugarcane have more than tripled over 14 years (e.g. A Vermelha). This sugarcane expansion has basically been observed in P Colˆombia, Marimbondo and A Vermelha over areas of flat 560 land at low elevations. These results are in accordance with what has been suggested by EMBRAPA as areas potentially suitable for cultivation of sugarcane in the Rio Grande basin (see figure 3). A chronological analysis indicates different rates of sug- 565 arcane expansion for P Colˆombia, Marimbondo and A Vermelha sub-basins. Between 1993 and 2000,for example, P Colˆombia presented an increase of 9.8% in sugarcane plantations. During the same period, Marimbondo and A Vermelha showed an expansion of only 3.2% and 2.9%, respectively. In 570 contrast, from 2000 to 2007, a higher sugarcane expansion has been observed over Marimbondo and A Vermelha than P Colˆombia. While Marimbondo and A Vermelha pointed to an increase of 10.9% and 17.8% in areas covered by sugarcane, the expansion over P Colˆombia corresponded to 5.1% 575 (see table 4). Overall, sugarcane plantations replaced mostly pasture lands and areas of agriculture of grain. Comparisons made between land use distribution in 2007 and 1993 showed that the replacement of pasture lands by sugarcane fields achieved 580

6.8%, 7.5% and 8.9% of the Marimbondo, P Colˆombia and A Vermelha sub-basins, respectively. It is followed by the replacement of areas of agriculture of grain crops with 5.2%, 4.7% and 7.6%, and then Atlantic Rainforest with 2.1%, 1.6% and 3.8%, respectively. 4.2

Short-term impacts of sugarcane expansion on runoff

Since fluctuations in daily evapotranspiration and soil moisture rates are marginal, short-term impacts of sugarcane expansion on the water balance are exclusively evaluated in terms of daily runoff. Though, effects of sugarcane expansion on evapotranspiration and soil moisture are incorporated when evaluating over longer temporal horizons (see sections 4.3 and 4.4). Three data sets are then generated from percentage differences in daily runoff between the scenarios of expansion (i.e. R2000, R2007 and EMBRAPA) and the CR1993 one. As each run was performed over a simulation period of 20 years, each of these sets corresponds to 7300 daily runoff differences. The statistical significance of these percentage differences were tested by means of bootstrap, using 1000 random samples, for a significance level of 0.01 and are presented in figure 6 per sub-basin. Figure 6 According to bootstrap results, percentage differences in daily runoff between CR1993 and R2000 were not statistically significant at the 99% confidence level. It can be associated with the small expansion of sugarcane plantations between 1993 and 2000, which corresponded to a little over 2.5% of the Rio Grande basin. From 1993 to 2007, however, sugarcane (20.7%) surpassed agriculture of grain (11.8%) as the second-largest land use in the Rio Grande basin. It implied to reductions in average daily runoff from 0.25% to 1.5% at the outlets of the sub-basins (Fig. 6a). These reductions monotonically increase with the area converted to sugarcane over each sub-basin. Accordingly, average daily runoff at the outlets of A Vermelha and Marimbondo were the most affected by sugarcane expansion which have been reduced by up to 1% and 1.5%, respectively. Figure 6a shows effects of sugarcane expansion on average daily runoff if the total area suggested by EMBRAPA as suitable for growing sugarcane is filled with sugarcane plantations. In this case, percentage differences in average daily runoff were significant at 0.01 level as tested by bootstrap considering 1000 random re-sampling with replacement. Further, percentage differences in average daily runoff were lower than -10% at the outlets of the headwater subbasins. Since the headwater sub-basins are dominantly composed by shallow soils, which easily become saturated, conversion of pasture to sugarcane significantly increased evapotranspiration rates reducing runoff at their outlets.

Paper III

F. F. Pereira et al.: Towards the response of water balance to sugarcane expansion 4.3 Medium-term impacts of sugarcane expansion on water balance 635

585

590

595

Medium-term impacts of sugarcane expansion on water balance are estimated as percentage differences in annual runoff, evapotranspiration and soil water content over 20 simulation years. Annual runoff, evapotranspiration and soil water con- 640 tent are accumulated from daily values calculated in CR1993, R200, R2007 and REMBRAPA over the annual phenological cycle of sugarcane, so that from June to May. As differences in daily runoff between R2000 and CR1993 were not significant at the 99% confidence interval (see item 4.2), accumulated daily values throughout the year are also marginal. Therefore, figure 7 shows only percentage differ- 645 ences in annual runoff, evapotranspiration and soil water content between REMBRAPA, R2007 and CR1993. Figure 7

600

605

610

615

620

625

630

A general pattern that emerges from figure 7 is that soil 650 moisture content monotonically decreases with evapotranspiration. It indicates that water loss by evapotranspiration is higher over saturated soils than unsaturated ones. Therefore, since sugarcane infiltrates more than pasture (see table 2), the replacement of pasture by sugarcane implied to more 655 humid soils and, hence, larger evapotranspiration rates. Other implications of sugarcane expansion to the water balance can be observed in figure 7, and they are separately discussed for R2007-CR1993 and REMBRAPA-CR1993 as it follows below. Over 20 years, annual fluctuations in runoff, evapotran- 660 spiration and soil water content derived from the sugarcane expansion proposed between CR1993 and R2007 range -0.7 to 1%. It means that despite differences in daily runoff, for example, achieved up to -2.5% (see item 4.2), annual accumulated differences in runoff, evapotranspiration and soil 665 water content are affected by the sugarcane growth stages, which may smooth impacts of sugarcane expansion on water balance over longer time frames. Locally, contribution from groundwater is also an influencing factor in reducing the impacts of sugarcane expansion on annual accumulated runoff. It can be seen from comparisons between Marimbondo and P Colˆombia sub-basins, which present the same area of sugar- 670 cane expansion but different fluctuation rates — P Colˆombia between 0.08 and 0.15% whereas Marimbondo from -0.16 to 0.08%. Figure 7 also shows percentage differences in annual runoff, evapotranspiration and soil water content as a natu- 675 ral response of the water balance to a possible sugarcane expansion if areas suitable for growing sugarcane, as defined by EMBRAPA (BRASIL, 2009), are all filled with sugarcane plantations. It reveals that the water balance of headwater sub-basins is very sensitive to the sugarcane expansion 680 since an expansion of only 4% of their drainage area represents a reduction of 33% and 60% in their annual soil water content. In addition, annual evapotranspiration rates nearly

7

double whereas annual runoff decreases by up to 22% at the outlet of the headwater sub-basins. For the other sub-basins, however, absolute values of annual fluctuations in runoff, evapotranspiration and soil moisture content are no greater than 12%, 40% and 10%, respectively. The combination of shallow soils (i.e. continuously saturated) and low contribution from groundwater found in the headwater sub-basins appear as the main reasons for the larger impacts of sugarcane expansion on their annual water balance. 4.4

Long-term impacts of sugarcane expansion on water balance

For a better understanding of the influence of sugarcane expansion on water balance of the Rio Grande basin, cumulative differences in evapotranspiration and surface runoff were investigated. In order to standardize comparisons across sub-basins, surface runoff and evapotranspiration are given in meters per square meter of drainage area. Thereafter, changes in the hydrological regime under sugarcane expansion were estimated as cumulative differences between the control run CR1993 and the scenarios of sugarcane expansion (i.e. R2000, R2007 and REMBRAPA). Moreover, trend analyses were applied to monthly runoff data from 1970 to 2010 for detecting ongoing response of the water balance to sugarcane expansion and for supporting results obtained from CR1993, R2000, R2007 and REMBRAPA. 4.4.1

Analysis of runoff trends

The non-parametric Mann-Kendall (MK) statistical test (Yue et al., 2002; Rao and Hsu, 2008) is used to assess the significance of trend in monthly runoff data under the null hypothesis of stationarity of the Funil, Camargos, Furnas, P Colˆombia, Marimbondo and A Vermelha sub-basins. The results of trend test performed by using the MK tests at 95% significance level are shown in table 5. Table 5 Table 5 reveals that MK trend tests on 1970-2010 time series of monthly runoff data did not reject the null hypothesis - stationarity - for all sub-basins. However, the outcome of the test also shows evidences of positive and negative trends according to the standardized MK statistic Z and the probability value P (p-value) calculated for each sub-basin. For independent sample data without trend, for instance, p-value and Z should be equal to 0.5 and 0, respectively. P-values closer to 1 and positive values for Z indicate data with positive trend whereas data with negative trend yields p-values closer to 0 and negative values for Z. In light of the results obtained from the mapping of sugarcane plantations, MK trend tests show that sugarcane expansion is associated with downward trends in monthly runoff for the 40-year period. This is because negative trends are present in all sub-basins that have substantial expansion (i.e.

91

Towards the response of water balance to sugarcane expansion in the Rio Grande basin, Brazil

8

685

F. F. Pereira et al.: Towards the response of water balance to sugarcane expansion

P Colˆombia, Marimbondo and A Vermelha). Despite Funil, Camargos and Furnas also present downward trends repre- 735 sented by negative values for Z and p-values lower than 0.5, their absolute values are small to be considered as evidences for trends. 4.4.2 Funil sub-basin

690

695

700

Funil is a headwater sub-basin of the Rio Grande basin where values of altitude are up to 900m.a.s.l.. For this reason, only the land use scenario proposed by EMBRAPA presented areas for cultivation of sugarcane in this sub-basin. EMBRAPA suggested that 4.7% of the Funil sub-basin are suitable for sugarcane fields from which 4.4% were previously classified 745 as pasture lands and 0.3% as Atlantic Rainforest. Figure 8a presents cumulative differences in surface runoff and evapotransporation (ET) between the scenarios of sugarcane expansion and the control run for Funil sub-basin. In this sub-basin, sugarcane expansion was observed in neither 750 R2000 nor R2007. Hence, cumulative differences in surface runoff and evapotranspiration are equal to 0; and therefore, the following findings only refer to comparisons between REMBRAPA and CR1993. Figure 8

705

710

715

725

730

92

755

As shown in figure 8a, replacing pasture lands with sugarcane plantations implies to runoff deficit at the outlet of the sub-basin. Further, over 20 simulation years, accumulated water loss due to sugarcane expansion represent 2m of surface runoff. In contrast, the cumulative water budget in Funil 760 indicates that evapotranspiration increases at the same rate as surface runoff decreases. Since sugarcane plantations mostly replaced pasture lands, the effects of sugarcane expansion on the water budget of Funil sub-basin are addressed to the increase of its averaged leaf area index. 765 4.4.3

720

740

Camargos sub-basin

Similarly to Funil, Camargos is a small headwater sub-basin. While sugarcane expansion was not observed in R2000, R2007 and CR1993, 2% of the Camargos sub-basin, previously classified as pasture lands, are categorized as suitable 770 to be used for cultivation of sugarcane by EMBRAPA. The natural response of the hydrological cycle to this replacement of pasture lands by sugarcane plantations is presented in terms of cumulative differences in surface runoff and evapotranspiration in figure 8b. 775 Although sugarcane plantations cover only a small portion of the sub-basin, its water budget is significantly affected over 20 years of simulation. In total, sugarcane expansion over Camargos sub-basin represents water losses by evapotranspiration of 5m and runoff deficit of 2.5m after a 20 780 year-period. Comparing to Funil, impacts of sugarcane expansion on water balance were larger in the Carmagos sub-basin; even

though the area suitable for growing sugarcane in Camargos being smaller. This is because, rather than the portion covered by sugarcane, such impacts depended upon the types of soil in the Camargos sub-basin. Predominantly composed of shallow soils and, consequently, often saturated, Camargos sub-basin presents favorable characteristics for increasing evapotranspiration rates. Accordingly, by increasing the capillarity of soil as reflection of the replacement of pasture lands by sugarcane plantations, Camargos is more sensitive to sugarcane expansion than Funil. 4.4.4

Furnas sub-basin

Furnas is the first sub-basin downstream Funil and Camargos, and already at CR1993 presents 1.5% of its drainage area covered by sugarcane plantations. This portion remained constant in R2000 and R2007, but is expanded to 17% in REMBRAPA. At REMBRAPA scenario, the expansion of sugarcane plantations basically replaced pasture lands (12.5%), followed by Atlantic Rainforest (2%) and agriculture of grain crops (1%). Unlike to Camargos, Furnas sub-basin presents a large water storage capacity in the soil since it is dominantly composed of deep soils. Due to this regional soil characteristic, cumulative differences in evapotranspiration between REMBRAPA and CR1993 are lower than 3m. (Fig. 9a). Figure 9 In respect to surface runoff, an expansion of 15.5% of sugarcane plantations means an accumulated reduction of 1.8 m for a 20-year period. Although sugarcane plantations represent almost one-fifth of the sub-basin, the runoff deficit derived from sugarcane expansion is smaller than Funil or Camargos. This is due to the fact that Furnas counts on the combination of a large water storage capacity and contributions from two subsidiary basins which makes runoff at its outlet more resistant to sugarcane expansion than Funil and Camargos. 4.4.5

P. Colˆombia sub-basin

P Colˆombia sub-basin has a drainage area of 75700 km² and is located downstream Furnas sub-basin. For P Colˆombia sub-basin, sugarcane expansion was observed in all land use scenarios and it is briefly described for each of them as follows. In CR1993, sugarcane plantations represented 11% of the sub-basin. Between CR1993 and R2000, they expanded to 20.8% and replaced areas of pasture lands (5%), agriculture of grain crops (3.2%) and Atlantic Rainforest (1.6%). From R2000 to R2007, the portion of the sub-basin covered by sugarcane plantations reached to 26% whereas REMBRAPA proposed that sugarcane replaces 16.4% of pasture lands, 3.2% of Atlantic Rainforest and 3.1% of agriculture of grain crops over one-third of the sub-basin.

Paper III

F. F. Pereira et al.: Towards the response of water balance to sugarcane expansion

785

790

795

Cumulative differences in surface runoff and evapotran- 835 spiration between CR1993, R2000, R2007 and REMBRAPA are shown in figure 9b. As agricultural practices are already ongoing in the P Colˆombia sub-basin, absolute values of cumulative differences in surface runoff and evapotranspiration over a 20-year period are lower than 1m. 840 Regarding water losses by evapotranspiration, cumulative differences between R2007 and CR1993 reveal that after 20 years, the amount of water reaches to 0.3m. This value goes up to 0.6m for comparisons between REMBRAPA and CR1993. On the other hand, cumulative differences in sur- 845 face runoff indicate neither up- nor downward trends between R2007, R2000 and the control scenario.In contrast, cumulative differences between REMBRAPA and CR1993 show a runoff deficit of 1m. 850

4.4.6 800

805

810

815

820

825

Marimbondo sub-basin

Unlike P Colˆombia, Furnas, Camargos and Funil sub-basins, contributions to surface runoff in the Marimbondo sub-basin come exclusively from rivers in the southern part of the Rio 855 Grande basin whose drainage areas are characterized by intensive agricultural activities. Here, sugarcane plantations are found in all land use scenarios. In CR1993, the land use distribution consisted of 40.8% of pasture lands, 27.9% of sugarcane plantations, 17.2% of agriculture of grain, 13.1% of Atlantic Rainforest and 1% of areas covered by water bodies. R2000 indicates a replacement of 1.1% of pasture lands, 1% 860 of agriculture of grain and 1% of Atlantic Rainforest by sugarcane whereas R2007 proposes that sugarcane plantations cover 42% of the sub-basin mostly replacing pasture lands. Finally, REMBRAPA assumes that 58% of Marimbondo is covered by sugarcane. The overall cumulative water budget over 20 simulation 865 years for Marimbondo is shown in figure 10a. While cumulative differences between R2000, R2007 and the control run range from 0 to -0.2m of surface runoff and from 0 to 0.2m of evapotranspiration, they achieve -0.4m and 2m, respectively, between REMBRAPA and the control run. 870 Even though sugarcane represents almost half of the Marimbondo sub-basin after expansion, these results reveal that such expansion is not as important to the local water balance in this sub-basin as it is to Camargos, for example. This is due to the fact that since the 60’s agriculture lands 875 have already been introduced into the Marimbondo landscape (Tucci and Clarke, 1998); hence impacts of sugarcane expansion on its water balance correspond basically to regional shifts in crops. 880

830

4.4.7

A Vermelha sub-basin

A Vermelha is the first sub-basin upstream the outlet of the Rio Grande basin and downstream Marimbondo and P Colˆombia sub-basins. Since most of its incoming water is 885 propagated from upstream sub-basins, surface runoff at the

9

outlet of A Vermelha highly depends on land use changes over upstream sub-basins. Here, areas covered by sugarcane begin from 9.4% in CR1993, expanded to 12.3% in R2000 and reach to 30% in R2007 whereas EMBRAPA suggests that 58% of the subbasin are suitable for growing sugarcane. While sugarcane plantations replace pasture lands (8%), agriculture of grain (8%) and Atlantic Rainforest (5%) between CR1993 and R2007, comparisons between CR1993 and REMBRAPA indicate that these percentage values go to 23%, 19.1% and 6.5% respectively. As a natural response to these land use changes, interannual variations in the local water balance were observed and estimated as cumulative differences in surface runoff and evapotranspiration (Fig. 10b). According to figure 10b, impacts of the sugarcane expansion from CR1993 to R2007 and REMBRAPA represent runoff deficit of 0.1 and 2.3m at the outlet of the sub-basin. This decreasing trend in runoff is supported by trend analysis on observed data performed in section 4.4.1. In contrast to runoff, cumulative differences in evapotranspiration reveal an increasing trend. It is explained by the replacement of 23% of pasture lands by sugarcane, which implies an increase in the spatially averaged leaf area index of the sub-basin. Figure 10 5

Conclusions

In this study, impacts of land use changes on water balance were investigated in a basin under ongoing sugarcane expansion. Sugarcane plantations were then tracked using satellite images captured in 1993, 2000 and 2007 and, along with the mapping of areas suitable for cultivation of sugarcane made by EMBRAPA, were used to generate historical and future land use in the Rio Grande basin. Finally, impacts of such sugarcane expansion were estimated as fluctuations in runoff, evapotranspiration and soil water content on daily, annual and decadal basis over 20 years. On a daily basis, sets of percentage differences in daily runoff were generated in order to highlight trends resulting from short-term effects of sugarcane expansion on runoff. Similarly, sets of percentage differences in annual runoff, evapotranspiration and soil water content aggregated over the entire annual phenological cycle of sugarcane were used to evaluate seasonal and inter-annual variability in the water balance of the Rio Grande basin. Thereafter, long-term impacts were estimated as differences in runoff and evapotranspiration accumulated over 20 years. As suggested by Warburton et al. (2011) as a good practice to adequately estimate impacts of land use change on water resources, this study assessed the water balance at different spatial and temporal scales. Overall, four factors could be identified as highly related to the impacts of sugarcane expansion on the water balance of the Rio Grande basin. These

93

Towards the response of water balance to sugarcane expansion in the Rio Grande basin, Brazil

10

890

895

900

905

F. F. Pereira et al.: Towards the response of water balance to sugarcane expansion

factors are the amount of areas replaced with sugarcane plan- 940 tations, their location within the basin, regional soil properties and local groundwater contribution to stream flow. This study also revealed that water loss by evapotranspiration due to sugarcane expansion achieves up to 4m/m2 over 20 years. Considering the soil characteristics of the sub- 945 basin and area of sugarcane expansion, this value is close to the ones estimated and observed by Watanabe et al. (2004) and is greater than values of evapotranspiration estimated by Marin et al. (2013). The latter, though, calculated evapotran- 950 spiration based on downscaled outputs from general climate models (GCMs), which often underestimate evapotranspiration rates (Milly, 1991; Rotstayn et al., 2006; Pereira et al., 2013). Consequently, it may have led to an underestimation of effects of climate change on water efficiency use in the 955 State of S˜ao Paulo. Finally, it is shown that sugarcane expansion mostly affected the water balance if it happens over the headwater areas of low soil water storage capacity. Since headwater basins 960 are dominated by pasture, sugarcane expansion significantly increased evapotranspiration whereas reduced runoff and soil moisture content. 965

910

Acknowledgements. We would like to thank to Erasmus Mundus project for funding a postdoctoral fellowship to Mehriddin Tursunov (2010-2011). We are also grateful to Crafoord Foundation and the Swedish Research Council (Vetenskapr˚adet) for supporting the development of this work. 970

References 915

920

925

930

935

94

AgSri: Sustainable Sugarcane Initiative: Producing “more with 975 less”, Tech. rep., Natural Resource Management Center, 2012. Aguiar, D. A., Rudorff, B. F. T., Silva, W. F., Adami, M., and Mello, M. P.: Studies on the Rapid Expansion of Sugarcane for Ethanol Production in S˜ao Paulo State (Brazil) Using Landsat Data., Remote Sensing, 3, 2682–2703, 2011. 980 Andr´e, R. G. B., Mendonc¸a, J. C., Marques, V. S., Pinheiro, F. M. A., and Marques, J.: Aspectos energ´eticos do desenvolvimento da cana-de-ac¸u´ car. Parte 1: Balanc¸o de radiac¸a˜ o e parˆametros derivados., Revista Brasileira de Meteorologia, 25, 375–382, 2010. 985 ANEEL: Atlas brasileiro de energia el´etrica, Brazil (Government), Bras´ılia, BR, 2nd edn., 2005. Bolling, C. and Suarez, N. R.: The Brazilian Sugar Industry: Recent Developments., Sugar and Sweetener Situation & Outlook, 232, 14–18, 2001. 990 Borges, J. M. M. and Almeida, A. N.: Pro-´alcool: accomplishments and perspectives., World Sugar Journal, 7, 7–13, 1985. BRASIL: Zoneamento Agroecolø’gico da Cana-de-Ac¸u´ car: Expandir a produc¸a˜ o, presevar a vida, garantir futuro, Minist´erio da Agricultura, Pecu´aria e Abastecimento, Bras´ılia, BR, first edn., 995 2009. Bremicker, M.: Aufbau eines Wasserhaushaltsmodells f¨ur das Weser und das Ostsee Einzugsgebiet als Baustein eines Atmosph¨aren-Hydrologie-Modells, Ph.D. thesis, Geowis-

senschaftlicher Fakult¨at der AlbertLudwigs-Universit¨at, Freiburg, Germany, 1998. Cheesman, O. D.: The Environmental Impacts of Sugarcane Production, Cromwell Press, Trowbridge, UK, 1st edn., 2004. Collischonn, B., Collischonn, W., and Tucci, C. E. M.: Daily hydrological modeling in the Amazon basin using TRMM rainfall estimates., Journal of Hydrology, 360, 207–216, 2008. Collischonn, W.: Hydrologic simulation of large basins (in Portuguese)., Ph.D. thesis, Instituto de Pesquisas Hidr´aulicas, Porto Alegre, BR, 2001. Collischonn, W., Allasia, D. G., Silva, B. C., and Tucci, C. E. M.: The MGB-IPH model for large-scale rainfall-runoff modeling., Hydrology Sciences Journal, 52, 878–895, 2007. CPRM: Geodiversity of Brazil: Knowing the past to understand the present and predict the future, Geological Survey of Brazil, Rio de Janeiro, 2012. Eggleston, G., Legendre, B., and Tew, T.: Indicators of freezedamaged sugarcane varieties which can predict processing problems, Food chemistry, 87, 119 – 133, 2004. Espinosa, A. M.: Previs˜ao quantitativa por conjunto da precipitac¸a˜ o do ver˜ao 2007/2008 na Bacia de Rio Grande, Ph.D. thesis, Universidade de S˜ao Paulo, 2011. FAURGS: Previs˜ao de afluˆencia a reservat´orios hidrel´etricos – m´odulo 1. Projeto FAURGS/FINEP 40.04.0094.00., Tech. rep., FINEP, Porto Alegre, BR, 2007. Food and of the United Nations, A. O.: Soil map of the world, Tech. rep., UNESCO, Paris, 1974. Gedney, N., Cox, P. M., Betts, R. A., and Boucher, O.: Detection of a direct carbon dioxide effect in continental river runoff records., Nature, 439, 835–838, 2006. Hasenack, H., Weber, E., Fernandes, S. M. C., Jonas, R. J., and Duarte, E. H. G.: Processamento e organizac¸a˜ o de modelos digitais de elevac¸a˜ o cont´ınuos para os pa´ıses de l´ıngua portuguesa a partir do SRTM., Editora da UFRGS, 2010. Herrera, M. C.: Cane, sugar and the environment, in: Proceedings of the Cuba/FAO International Sugar Conference, 1999. Hlavcova, K., Szolgay, J., Kohnov´a, S., and Horv´at, O.: The limitations of assessing impacts of land use changes on runoff with a distributed hydrological model case study of the Hron River., Biologia, 64, 589–593, 2009. IBGE: Classificac¸a˜ o da Vegetac¸a˜ o Brasileira, adaptada a um sistema universal, Tech. rep., Instituto Brasileiro de Geografia e Estat´ıstica, Rio de Janeiro, BR, 1991. ´ IEA: Alcool: Projec¸a˜ o da Produc¸a˜ o e Exportac¸a˜ o no Per´ıodo 2005/06 A 2015/16, Tech. rep., Governo de S˜ao Paulo, S˜ao Paulo, BR, 2006. IPT: Diagn´ostico da situac¸a˜ o dos recursos h´ıdricos na Bacia Hidrogr´afica do Rio Grande (BHRG) – SP/MG, Tech. Rep. 96581205, Instituto de Pesquisas Tecnol´ogicas do Estado de S˜ao Paulo, 2008. James, G.: Sugarcane., John Wiley & Sons, Ltd, Oxford, UK, 2nd edn., 2008. Lall, U. and Sharma, A.: A nearest neighbor bootstrap for resampling hydrologic time series., Water Resources Research, 32, 679–693, 1996. Liang, X., Lettenmaier, D. P., Wood, E. F., and Burges, S. J.: A simple hydrologically based model of land surface water and energy fluxes for general circulation models., Journal Geophysical Research, 99, 14 415–14 428, 1994.

Paper III

F. F. Pereira et al.: Towards the response of water balance to sugarcane expansion

1000

1005

1010

1015

1020

1025

1030

1035

1040

1045

1050

1055

Maraddi, G. N.: An analysis of sustainable cultivation practices followed by sugarcane growers in Karnataka., Ph.D. thesis, Uni1060 versity of Agricultural Sciences, Dharwad, 2006. Marin, F. R., Jones, J. W., Singels, A., Royce, F., Assad, E. D., Pellegrino, G. Q., and Justino, F.: Climate change impacts on sugarcane attainable yield in southern Brazil, Climatic change, 117, 227 – 239, 2013. Mendes, C. and Cirilo, J.: Geoprocessamento, in: Geoprocessa-1065 mento em recursos h´ıdricos: Princ´ıpios, Integrac¸a˜ o e Aplicac¸a˜ o, Associac¸a˜ o Brasileira de Recursos H´ıdricos, Porto Alegre, BR, 2001. Milly, P. C. D.: Potential evaporation and soil moisture in general 1070 climate models, Journal of climate, 5, 209 – 226, 1991. Mui, N. T., Preston, T. R., van Binh, D., Ly, L. V., and Ohlsson, I.: Effect of management practices on yield and quality of sugar cane and on soil fertility., Livestock Research for Rural Development, 8, 51, 1996. Nitsch, M.: O programa de biocombust´ıveis Proalcool no contexto1075 da estrat´egia energ´etica brasileira, Brazilian Journal of Political Economy, 2, 123–138, 1991. N´obrega, M. T., Collishonn, W., Tucci, C. E. M., and Paz, A. R.: Uncertainty in climate change impacts on water resources in the Rio Grande Basin, Brazil., Hydrology and Earth System Sciences,1080 15, 585 – 595, 2011. ONS: Invent´ario das restric¸o˜ es operativas dos aproveitamentos hidroel´etricos, Tech. Rep. 00132013, Operador Nacional do Sistema El´etrico, 2013. Paiva, R. C. D. and Collischonn, W.: Climate change impacts on1085 river Quarai flow, in: Climate and Water Resources Modelling, edited by Fung, C. F., Lopez, A., and New, M., p. 200, John Willey & Sons, 2010. Paiva, R. C. D., Collischonn, W., and Tucci, C. E. M.: Large scale hydrologic and hydrodynamic modeling using limited data and a1090 GIS based approach, Journal of Hydrology, 406, 170–181, 2011. Paz, A. R., Collischonn, W., Tucci, C. E. M., and Padovani, C. R.: Large scale modelling of channel flow and floodplain inundation dynamics and its application to the Pantanal (Brazil), Hydrolog1095 ical processes, 25, 1498–1516, 2011. Pereira, F. F., de Moraes, M. A. E., and Uvo, C. B.: Implementation of a two-way coupled atmospheric-hydrological system for environmental modeling at regional scale, Hydrology Research, p. to appear, 2013. RADAMBRASIL: Programa de Integrac¸a˜ o Nacional, Levanta-1100 mento de Recursos Naturais (Folhas SE-23 (Belo Horizonte) e SF-22 (Paranapanema))., Tech. rep., Minist´erio das Minas e Energia, Bras´ılia, BR, 1982. Rao, A. R. and Hsu, E.-C.: Streamflow data analyses, vol. 60 of Water Science and Technology Library, chap. 5, pp. 121 – 148,1105 Springer, 2008. Rawls, W., Ahuja, L., Brakensiek, D., and Shirmohammadi, A.: Handbook of Hydrology, chap. Infiltration and soil water movement, McGraw Hill, 1993. Robertson, M. J., Inman-Bamberb, N. G., Muchowa, R. C., and Wood, A. W.: Physiology and productivity of sugarcane with early and mid-season water deficit., Field Crops Research, 64, 211–227, 1999. Rocha, H. R., Freitas, H. C., Rosolem, R., Ju´arez, R. I. N., Tannus, R. N., Ligo, M. A., Cabral, O. M. R., and Silva Dias, M. A. F.: Measurements of CO2 exchange over a woodland savanna (Cer-

11

rado Sensu stricto) in southeast Brasil, Biota Neotropica, 2, 1 – 11, 2002. Rotstayn, L. D., Roderick, M. L., and Farquhar, G. D.: A simple pan-evaporation model for analysis of climate simulations: Evaluation over Australia, Geophysical research letters, 33, 2006. Rudorff, B. F. T., Aguiar, D. A., Silva, W. F., Sugawara, L. M., Adami, M., and Moreira, M. A.: Studies on the Rapid Expansion of Sugarcane for Ethanol Production in S˜ao Paulo State (Brazil) Using Landsat Data., Remote Sensing, 2, 1057–1076, 2010. Ruhoff, A. L.: Sensoriamento remoto aplicado a` estimativa da evapotranspirac¸a˜ o em biomas tropicais, Ph.D. thesis, Instituto de Pesquisas Hidr´aulicas, 2011. Sampaio, G., Nobre, C., Costa, M. H., and Satyamurty, P.: Regional climate change over eastern Amazonia caused by pasture and soybean cropland expansion., Geophysical Research Letter, 34, 1–7, 2007. Tai, P. Y. P. and Miller, J. D.: Recurrent selection for cold tolerance of sugarcane, American Society of Sugar Cane Technologists, 13, 40 – 45, 1993. Tucci, C. E. M. and Clarke, R. T.: Environmental issues in the la Plata basin, International Journal of Water Resources Development, 14, 157 – 173, 1998. Tucci, C. E. M., Collischonn, W., Clarke, R. T., Paz, A. R., and Allasia, D.: Short- and long-term flow forecasting in the Rio Grande watershed (Brazil)., Atmospheric Science Letters, 9, 53– 56, 2008. Warburton, M. L., Schulze, R. E., and Jewitt, G. P. W.: Hydrological impacts of land use change in three diverse South African catchments., Journal of Hydrology, 414–415, 118–135, 2011. Watanabe, K., Yamamoto, T., Yamada, T., Sakuratani, T., Nawata, E., Noichana, C., Sributta, A., and Higuchi, H.: Changes in seasonal evapotranspiration, soil water content, and crop coefficients in sugarcane, cassava, and maize fields in Northeast Thailand, Agricultural Water Management, 67, 133 – 143, 2004. Wijesekara, G. N., Gupta, A., Valeo, C., Hasbani, J. G., Qiao, Y., Delaney, P., and Marceau, D. J.: Assessing the impact of future land-use changes on hydrological processes in the Elbow River watershed in southern Alberta, Canada., Journal of Hydrology, 412–413, 220–232, 2012. WWFBrasil: Monitoring Alterations in Vegetation and Land Use in the Brazilian Portion of the Upper Paraguay River Basin., Tech. rep., Non-governmental organizations: Ecoa - Ecologia e Ac¸a˜ o, Conservation International, Avina Foundation., 2008. Yue, S., Pilon, P., Phinney, B., and Cavadias, G.: The influence of autocorrelation on the ability to detect trend in hydrological series., Hydrological Processes, 16, 1807 – 1829, 2002. Zuurbier, P. and van de Vooren, J.: Sugarcane ethanol., Wageningen Academic Publishers., Wageningen, 2008.

95

96 34.0 34.0 174.0 251.0 1100.0 1600.0

(m3 /s)

12

Hydroelectric power plant Funil Camargos Furnas P Colˆombia Marimbondo A Vermelha

Table 1: Minimum operating flow at the hydroelectric power plants used in this study (ONS, 2013).

Towards the response of water balance to sugarcane expansion in the Rio Grande basin, Brazil

F. F. Pereira et al.: Towards the response of water balance to sugarcane expansion

Same for all types of land use

Sugarcane fields

Atlantic Rainforest

Pasture lands

Agriculture of grain crops

Type of Land Use

Sugarcane fields

Atlantic Rainforest

Pasture lands

Agriculture of grain crops

Type of Land Use

Parameter Maximum water storage Mean percolation Residual water storage Maximum water storage Mean percolation Residual water storage Maximum water storage Mean percolation Residual water storage Maximum water storage Mean percolation Residual water storage Mean groundwater flow Upward flux of water Shape parameter Hydraulic conductivity

Parameter Albedo Leaf Area Index (m2 /m2 ) Height of trees (m) Albedo Leaf Area Index (m2 /m2 ) Height of trees (m) Albedo Leaf Area Index (m2 /m2 ) Height of trees (m) Albedo Leaf area index (m2 /m2 ) Height of trees (m)

Jan 0.13 4.00 1.00 0.20 2.00 0.50 0.11 8.00 9.00 0.28 7.00 3.60

Fixed Parameters Feb Mar Apr May 0.13 0.13 0.13 0.16 4.00 4.00 5.00 1.00 1.00 1.00 1.00 0.50 0.20 0.20 0.21 0.21 2.00 2.00 3.00 2.00 0.50 0.50 0.50 0.50 0.11 0.11 0.11 0.11 8.00 8.00 8.00 8.00 9.00 9.00 9.00 9.00 0.28 0.29 0.31 0.31 7.00 8.00 9.00 9.00 3.60 3.80 3.80 3.80 Adjustable Parameters Unit mm mm d−1 mm mm mm d−1 mm mm mm d−1 mm mm mm d−1 mm mm d−1 mm d−1 mm d−1 Jun 0.16 1.00 0.80 0.21 2.00 0.50 0.11 8.00 9.00 0.24 3.00 0.50

Jul 0.17 2.00 0.80 0.21 2.00 0.50 0.11 8.00 9.00 0.25 5.00 1.20 Value 625.0 3.5 62.5 446.0 2.1 44.6 711.0 6.2 71.1 654.0 3.9 65.4 146.0 0.0 0.10 2268.0

Aug 0.17 2.00 0.80 0.21 2.00 0.50 0.11 8.00 9.00 0.25 5.00 1.20

Sep 0.16 2.00 0.80 0.21 2.00 0.50 0.11 8.00 9.00 0.25 5.00 1.20

Oct 0.15 2.00 0.90 0.20 2.00 0.50 0.11 8.00 9.00 0.27 6.00 2.80

Nov 0.14 3.00 0.90 0.20 2.00 0.50 0.11 8.00 9.00 0.27 6.00 2.80

Dec 0.13 3.00 0.90 0.20 2.00 0.50 0.11 8.00 9.00 0.27 6.00 2.80

Table 2: Fixed and adjustable parameters used (or assumed) in this study. The set of fixed and adjustable parameters for agriculture of grain crops, pasture lands and Atlantic Rainforest were assumed as defined via calibration and validation by N´obrega et al. (2011) in the Rio Grande basin. On the other hand, for sugarcane fields, fixed parameters were adopted according to ranges obtained in the literature whereas adjustable parameters were calibrated and validated in this study.

Paper III

F. F. Pereira et al.: Towards the response of water balance to sugarcane expansion 13

97

98 RVE(%) 12.30 12.31 13.27

RVE(%) 6.79 -8.62 3.12

1993 0.0 0.0 1.5 11.0 27.9 9.4

(%) 2000 0.0 0.0 1.5 20.8 31.1 12.3 2007 0.0 0.0 1.5 25.9 42.0 30.1

Funil Camargos Furnas P Colˆombia Marimbondo A Vermelha

Sub-basin Z -0.39 -0.51 -0.21 -1.39 -1.33 -1.82

p-value 0.347 0.304 0.416 0.082 0.092 0.035

(%) Null Hypothesis (H) Not rejected (Stationary) Not rejected (Stationary) Not rejected (Stationary) Not rejected (Stationary) Not rejected (Stationary) Not rejected (Stationary)

Table 5: Trend test results for monthly runoff time series at 95% significance level.

Funil Camargos Furnas P Colˆombia Marimbondo A Vermelha

Sub-basin

Table 4: The portion of areas covered by sugarcane fields per sub-basin in 1993, 2000 and 2007.

Calibration NS RMSE (m3 s−1 ) 0.92 270.97 0.92 370.40 0.96 130.50 Validation NS RMSE (m3 s−1 ) 0.88 301.14 0.87 436.10 0.85 508.92

14

Gauging station P Colˆombia Marimbondo A Vermelha

Gauging station P Colˆombia Marimbondo A Vermelha

Table 3: Summary of results of the calibration and validation of parameters for sugarcane in the Rio Grande basin.

Towards the response of water balance to sugarcane expansion in the Rio Grande basin, Brazil

F. F. Pereira et al.: Towards the response of water balance to sugarcane expansion

Paper III

F. F. Pereira et al.: Towards the response of water balance to sugarcane expansion

15

99

Towards the response of water balance to sugarcane expansion in the Rio Grande basin, Brazil

F. F. Pereira et al.: Towards the response of water balance to sugarcane expansion

Fig. 2: Meteorological data network, discharge gauging stations and all the sub-basins of the Rio Grande basin.

16

100

93 00 07 EM

93 00 07 EM

93 00 07 EM

93 00 07 EM

93 00 07 EM

F. F. Pereira et al.: Towards the response of water balance to sugarcane expansion

Fig. 3: Land use distribution of each sub-basin of the Rio Grande basin for all four land use scenarios used in this study: 1993, 2000, 2007 and EMBRAPA.

Sugarcane fields

Pasture lands

Agriculture of grain crops

Atlantic Forest

Areas covered by water bodies

40

93 00 07 EM

93 00 07 EM

Paper III

17

101

102 Jan/1998

0.00 Oct/1989

2000.00

4000.00

6000.00

8000.00

10000.00

12000.00

P Colômbia CALC

Apr/1995

(a) P Colˆombia

P Colômbia OBS

Jul/1992

Jan/1998

0.00 Oct/1989

2000.00

4000.00

6000.00

8000.00

10000.00

A Vermelha CALC

Apr/1995

(c) A Vermelha

A Vermelha OBS

Jul/1992

Oct/2000

Discharge (m³/s)

Discharge (m³/s)

Jul/1992

Oct/2000

Jan/1998

Marimbondo CALC

Apr/1995

(b) Marimbondo

Marimbondo OBS

Oct/2000

Fig. 4: Calibration of the MGB-IPH parameters for sugarcane. Calculated and observed hydrographs at the outlets of the P Colˆombia, Marimbondo and A Vermelha sub-basins.

0.00 Oct/1989

2000.00

4000.00

6000.00

8000.00

12000.00

18

Discharge (m³/s)

10000.00

Towards the response of water balance to sugarcane expansion in the Rio Grande basin, Brazil

F. F. Pereira et al.: Towards the response of water balance to sugarcane expansion

0.00 Oct/2000

1000.00

2000.00

3000.00

4000.00

5000.00

Jun/2003

Mar/2006

0.00 Oct/2000

2000.00

4000.00

6000.00

8000.00

10000.00

12000.00

P Colômbia CALC

Nov/2004

(a) P Colˆombia

P Colômbia OBS

Feb/2002

Jun/2003

Mar/2006

0.00 Oct/2000

A Vermelha CALC

Nov/2004

Discharge (m³/s)

(c) A Vermelha

A Vermelha OBS

Feb/2002

Aug/2007

1500.00

3000.00

4500.00

6000.00

7500.00

9000.00

10500.00

Jun/2003

Aug/2007

Nov/2004

Mar/2006

Marimbondo CALC

(b) Marimbondo

Marimbondo OBS

Feb/2002

Aug/2007

F. F. Pereira et al.: Towards the response of water balance to sugarcane expansion

Fig. 5: Validation of the MGB-IPH parameters for sugarcane. Calculated and observed hydrographs at the outlets of the P Colˆombia, Marimbondo and A Vermelha sub-basins.

Discharge (m³/s)

6000.00

Discharge (m³/s)

7000.00

Paper III

19

103

(b) 1993 - EMBRAPA

20

104

Fig. 6: Results from bootstrap analysis of the percentage differences of daily surface runoff between CR1993 and R2007 (a) and CR1993 and EMBRAPA (b).

(a) 1993 - 2007

Towards the response of water balance to sugarcane expansion in the Rio Grande basin, Brazil

F. F. Pereira et al.: Towards the response of water balance to sugarcane expansion

R2007 − CR1993 (%) −80 1990

−60

−40

−20

0

20

−1 1990

−0.5

0

Funil

2000

2000

2010

2010

Camargos

Soil water content

0 1990

50

100

150

200

Furnas

−0.5 1990

0

2000

P Colombia

2000

2010

2010

−25 1990

−20

−15

−10

−5

0

2000

2000

2010

2010

A Vermelha

Surface runoff

Marimbondo

−0.2 1990

−0.1

0

0.1

1 0.5

0.2

Evapotranspiration 1.5

Fig. 7: Percentage differences of accumulated annual values of soil water content, evapotranspiration and surface runoff for all sub-basins during the entire simulation period between CR1993 and R2007, and CR1993 and REMBRAPA. Accumulated annual values were accumulated along the annual cycle of sugarcane phenology.

EMBRAPA − CR1993 (%)

0.5

Paper III

F. F. Pereira et al.: Towards the response of water balance to sugarcane expansion 21

105

Towards the response of water balance to sugarcane expansion in the Rio Grande basin, Brazil

22

F. F. Pereira et al.: Towards the response of water balance to sugarcane expansion

Cumulative Differences (m)

−3.0

−4.0

−5.0 00

Runoff (R2000−CR1993) Runoff (R2007−CR1993) Runoff (REMBRAPA−CR1993) ET (R2000−CR1993) ET (R2007−CR1993) ET (REMBRAPA−CR1993) 02

05

08 10 Time (years)

13

16

19

(a) Funil

Cumulative Differences (m)

−3.0

−4.0

−5.0 00

Runoff (R2000−CR1993) Runoff (R2007−CR1993) Runoff (REMBRAPA−CR1993) ET (R2000−CR1993) ET (R2007−CR1993) ET (REMBRAPA−CR1993) 02

05

08 10 Time (years)

13

16

(b) Camargos Fig. 8: Cumulative differences in surface runoff and evapotranspiration between CR1993, R2000, R2007 and REMBRAPA for Funil (a) and Camargos (b) sub-basins.

106

19

Paper III

F. F. Pereira et al.: Towards the response of water balance to sugarcane expansion

23

Cumulative Differences (m)

−3.0

−4.0

−5.0 00

Runoff (R2000−CR1993) Runoff (R2007−CR1993) Runoff (REMBRAPA−CR1993) ET (R2000−CR1993) ET (R2007−CR1993) ET (REMBRAPA−CR1993) 02

05

08 10 Time (years)

13

16

19

(a) Furnas

Cumulative Differences (m)

−3.0

−4.0

−5.0 00

Runoff (R2000−CR1993) Runoff (R2007−CR1993) Runoff (REMBRAPA−CR1993) ET (R2000−CR1993) ET (R2007−CR1993) ET (REMBRAPA−CR1993) 02

05

08 10 Time (years)

13

16

19

(b) P Colˆombia Fig. 9: Cumulative differences in the local water balance of Furnas (a) and P Colˆombia (b) sub-basins over a 20-year period.

107

Towards the response of water balance to sugarcane expansion in the Rio Grande basin, Brazil

24

F. F. Pereira et al.: Towards the response of water balance to sugarcane expansion

Cumulative Differences (m)

−3.0

−4.0

−5.0 00

Runoff (R2000−CR1993) Runoff (R2007−CR1993) Runoff (REMBRAPA−CR1993) ET (R2000−CR1993) ET (R2007−CR1993) ET (REMBRAPA−CR1993) 02

05

08 10 Time (years)

13

16

19

(a) Marimbondo

Cumulative Differences (m)

−3.0

−4.0

−5.0 00

Runoff (R2000−CR1993) Runoff (R2007−CR1993) Runoff (REMBRAPA−CR1993) ET (R2000−CR1993) ET (R2007−CR1993) ET (REMBRAPA−CR1993) 02

05

08 10 Time (years)

13

16

(b) A Vermelha Fig. 10: Differences in surface runoff and evapotranspiration accumulated over 20 simulation years for Marimbondo (a) and A Vermelha (b) subbasins.

108

19

PAPER IV Assessment of Numerical Schemes for Solving Advection-Diffusion Equation on Unstructured Grids: Case Study of River Guaíba, Brazil F. F. Pereira, C. R. Fragoso Jr, C. B. Uvo, W. Collischonn and D. M. L. M. Marques. Nonlinear Processes in Geophysics, Reviewed.

Paper IV

Manuscript prepared for Nonlin. Processes Geophys. with version 5.0 of the LATEX class copernicus.cls. Date: 27 August 2013

Assessment of Numerical Schemes for Solving Advection-Diffusion Equation on Unstructured Grids: Case Study of River Gua´ıba, Brazil F. F. Pereira1 , C. R. Fragoso Jr2 , C. B. Uvo1 , W. Collischonn3 , and D. M. L. Motta Marques3 1

Department of Water Resources Engineering, Lund University, Lund, SE Centro de Tecnologia, Universidade Federal de Alagoas, Macei´o, BR 3 Instituto de Pesquisas Hidr´aulicas, Universidade Federal do Rio Grande do Sul, BR

2

Correspondence to: F. F. Pereira ([email protected])

5

10

15

20

Abstract. In this work, a first-order upwind and a high order flux-limiter schemes for solving advection-diffusion equation on unstructured grids were evaluated. The numerical schemes were implemented as a module of an unstructured two-dimensional depth-averaged circulation model for shallow lakes (IPH-UnTRIM2D) and they have been applied to the River Gua´ıba in Brazil. Their performances were evaluated by comparing mass conservation balance errors for two scenarios of a passive tracer released into River Gua´ıba. The circulation model showed a good agreement with observed data collected at four water level stations along the River Gua´ıba where correlation coefficients achieved values up to 0.93. In addition, volume conservation errors were lower than 1% of the total volume of the River Gua´ıba. For all scenarios, the higher order flux-limiter scheme has been shown to be less diffusive than a first-order upwind scheme. Accumulated conservation mass balance errors calculated for the flux-limiter reached 8% whereas for a first-

25

order upwind scheme were close to 18% over a 15-day period. Although both schemes have presented mass conservation errors, these errors are assumed negligible compared to kinetic processes such as erosion, sedimentation or decay rates.

1 30

35

40

Introduction

Ecology of lakes and ponds are closely linked to physical factors, especially to hydrodynamic variables such as velocity, turbulence, and diffusion/convection of suspended material (Br¨onmark and Hansson, 2005). These aquatic ecosystems present complex circulation patterns which vary over time and space depending on density, temperature, wind among others (Reynolds, 1984). Fragoso Jr et al. (2008), for example, revealed a close relationship between spatial heterogeneity of phytoplankton and water circulation patterns — driven by gradients in viscosity and diffusivity, wind stress and bottom friction — in a shallow

111

Assessment of Numerical Schemes for Solving Advection-Diffusion Equation on Unstructured Grids: Case Study of River Guaíba, Brazil

2

45

50

55

60

65

70

75

80

85

112

F. F. Pereira et al.: Assessment of Numerical Schemes

lake. In addition, Hodges and Dallimore (2013) showed that upward and downward fluxes of water in lakes and/or ponds are associated with vertical heat exchange throughout the water column. 90 Thus, hydrodynamic processes affect transport of substances to, from and within lakes and ponds. The spatial variance of suspended particle concentrations may be great in shallow ecosystems once their distribution can be influenced by short- 95 term physical factors (e.g. sediment resuspension) (Carrick et al., 1993). Thus, a better understanding of the water circulation of these ecosystems plays an important role in their water quality dynamics (Chapra, 2005). 100 Several studies have presented numerical models which are able to represent velocity field and water level by solving the shallow water equations. Some of these models use finite difference scheme in a horizontal plane formed by uniform 105 rectangular grids (Casulli, 1990; Cheng et al., 1993). However, occasionally uniform grids are not flexible enough to represent complex geometries. Therefore, some numerical models use finite difference schemes on non-orthogonal curvi- 110 linear grids to allow greater flexibility (Ye and McCorquodale, 1997; Zhou, 1995). On the other hand, an disadvantage of using curvilinear grids lies in approximating oblique boundaries along the simulation domain which introduces errors to 115 the numerical solution (Margolin and Shashkov, 1999). Moreover, additional terms derived from curvilinear transformation are also sources of errors arising from a numerical approach using curvilinear grids (French, 1988). 120 A detailed study on the effects of grid nonorthogonality carried by Sankaranarayanan and Spaulding (2003) reveals that truncation error terms due to first and second derivative terms are functions of grid angle and aspect ratio in a way 125 that root-mean-square (RMS) errors in surface elevation and velocities sharply increase as the grid resolution decreases for curvilinear grids with a grid angle less than 45◦ (Thompson et al., 1985; Nielsen and Skovgaard, 2005). 130

In this sense, hydrodynamics models using orthogonal unstructured grids are considered an efficient tool to describe the dynamics of rivers, lakes and estuaries as they have high capacity of representing geometries and optimizing computational efforts with grid refinement in regions of interest (Casulli and Walters, 2000; Cheng and Casulli, 2001). Once conservation of volume is assured by an efficient hydrodynamic solution, a numerical solution for the advection–diffusion equation is needed to characterize the transport of a scalar variable such as salinity, temperature or any passive constituent that may represent sediment or biological communities. Analytical methodologies to solve advection-diffusion equation by mathematical substitutions and transformations (Mikhailov and Ozisik, 1984; Cotta, 2005) are widely applied to many experimental applications in heat and mass diffusion (Leij and van Genuchten, 2000; Guerrero et al., 2009). Nevertheless, a large number of parameters, coefficients, constants and functions are used in such methodologies which may reduce their range of applications to systems under strictly controlled external conditions. Several numerical models use a first-order upwind scheme to solve explicitly the advection–diffusion equation (Gross et al., 2002, 1999; Hetch et al., 1995). However, first-order upwind schemes are only effective in regions with low advection and low scalar concentration gradients (Le Veque, 1996). Greater accuracy may be achieved in first-order upwind schemes employing refined grids, which implies a higher computational effort. An alternative approach is to use high resolution schemes with flux-limiter functions. The high resolution scheme developed by Sweby (1984) has been successfully applied and improved over the past years (Casulli and Zanolli, 2005; Wang et al., 2008; Wood et al., 2008). Therefore, this work presents the formulation and application of a two-dimensional depthaveraged circulation and scalar transport model for shallow aquatic ecosystems based on unstruc-

Paper IV

F. F. Pereira et al.: Assessment of Numerical Schemes

135

140

145

tured orthogonal grids. It has been developed to maximize flexibility in grid specification ensuring a stable and robust numerical solution. A high-resolution scalar transport scheme using 175 flux-limiter function developed by Sweby (1984) was implemented and its results were compared to a first-order upwind scheme. The model was applied to River Gua´ıba to test its scalar transport schemes in solving realistic problems. Their nu- 180 merical solution were compared to estimates and measurements of sedimentation and erosion rates whereas their performances were evaluated using conservation of mass for the entire simulation domain.

2 2.1

150

155

160

Material and methods

185

Model description

IPH-UnTRIM2D is a finite-volume numerical representation of the two-dimensional depth- 190 integrated continuity and momentum equations of water motion developed at the Instituto de Pesquisas Hidr´aulicas (IPH), and conceptually based on the unstructured version of the Tidal, Residual, and Intertidal Mudflat (TRIM) model 195 proposed by Casulli and Walters (2000). It also represents integrated scalar transport (e.g. salinity, heat, suspended sediment) through advectiondiffusion equation. The governing equations and finite-volume numerical approximations are de- 200 scribed here. 2.2 Hydrodynamic module

165

170

IPH-UnTRIM2D solves the shallow water equations using a finite-volume approach, which calculates values of free surface and velocity at discrete places on an unstructured grid. As the shallow wa- 205 ter equations are derived from depth-integrating the Navier-Stokes equations — under the assumption that horizontal flow components are much greater than vertical ones — the hydrodynamic module of IPH-UnTRIM2D expresses the princi- 210 ple of conservation of mass and momentum.

3

To integrate the Navier-Stokes equation over the water column, IPH-UnTRIM2D assumes two boundary conditions corresponding to effects of wind stresses and bottom friction. From the momentum equations for an incompressible fluid (Fox et al., 2009), the boundary conditions due to effects of wind stresses at free surface and bottom friction at bed are given by prescribing the term that represents the divergence of stress along the water column as:

Av

∂u = τx − γu ∂z

and Av

∂v = τy − γv ∂z

(1)

where Av denotes the vertical coefficient of kine∂v matic eddy viscosity, ∂u ∂z and ∂z represent the gradient of the horizontal components of velocity along the water column, τx and τy are the horizontal components of wind shear stresses and γ indicates the bottom friction. Another assumption used in the hydrodynamic module of IPH-UnTRIM2D is related to the term of pressure gradient force in the momentum equations. Theoretically, the pressure gradient force can be divided into barotropic and baroclinic components by using the Leibniz integration rule (Blaise et al., 2010). And, since the baroclinic components are driven by vertical gradients of density and/or temperature over the water column, IPH-UnTRIM2D includes only barotropic components of pressure in the calculations as part of its depth-integrated approach (i.e. well mixed water column). This assumption reduces the term of pressure gradient force in the momentum equations to the following form: ∇p (x, y, z, t) = g∇η (x, y, z, t)

(2)

where p (x, y, z, t) is the pressure gradient force, g is the gravitational acceleration constant and η (x, y, z, t) means the free surface from an undisturbed reference level. The hydrodynamic module of IPH-UnTRIM2D also incorporates the continuity equation. Similarly to the momentum equations, the continu-

113

Assessment of Numerical Schemes for Solving Advection-Diffusion Equation on Unstructured Grids: Case Study of River Guaíba, Brazil

4

215

220

225

F. F. Pereira et al.: Assessment of Numerical Schemes

ity equation is integrated over the water column — from bed to free surface — and, along 245 with the shallow water equations, is used by IPH-UnTRIM2D to describe free surface flows. Thus, the momentum and continuity equations for two-dimensional circulation solved by IPHUnTRIM2D can be written as: ∂u ∂u ∂η ∂u +u +v = −g + ··· ∂t ∂x ∂y ∂x  2 2  ∂ u∂ u + τx − γu + f v Ah ∂x2 ∂y 2

250

(3)

255 ∂v ∂v ∂v ∂η +u +v = −g + ··· ∂t ∂x ∂y ∂y  2  ∂ v ∂2v Ah + + τy − γv − f u (4) ∂x2 ∂y 2 260

∂η ∂ + ∂t ∂x

Zη

udz +

−h

230

235

240

∂ ∂y

Zη

vdz = 0

(6) 275

τy = CD Wy kW k

(7)

where CD is the wind stress coefficient, Wx and Wy the horizontal components of wind speed at

114

γ=

√ g u2 + v 2 Cv2 H

(8)

here H(x, y, t) = h(h, y) + η(x, y, t) is total water depth and Cv is the Chezy coefficient. Coriolis acceleration in x and y directions used in the momentum equations are given by f v and f u, respectively, with f = 2ω sin φ where ω means angular speed of rotation of Earth about its axis in rads−1 and φ is geographic latitude in degrees. As part of the assumptions for depth-averaged circulation models (e.g. ADCIRC, Westerink et al. (1994) ; TRIM, Casulli (1990); River2D Steffler and Blackburn (2002)), both the effects of wind stresses and bottom friction are included at the same layer in the momentum equations. 2.3 Scalar Transport Module

(5)

−h

where u(x, y, t) and v(x, y, t) are the horizontal velocity components in x and y directions in ms−1 , t is time in seconds, h(x, y) is water depth 265 relative to an undisturbed reference level in m, f is the Coriolis parameter in s−1 assumed as constant and Ah is the is the horizontal eddy viscosity coefficient in m2 s−1 . Wind shear stresses (τx , τy ) and bottom-friction (γ) formulations are also calculated in this module. The wind shear stresses at free surface is pro270 portional to the wind speed and it was calculated as follows: τx = CD Wx kW k

the surface level in ms−1 , and kW k is the norm of the wind speed vector in ms−1 . The bottom-friction coefficient is calculated by:

The scalar transport module in IPH-UnTRIM2D calculates the concentration field of a tracer by means of the advection-diffusion equation deduced from Law of Conservation of Mass for incompressible fluid and it is expressed as: ∂HC ∂uHC ∂vHC + + = ··· ∂t ∂x ∂y     ∂ ∂ ∂HC ∂HC ··· Kh + Kh +sources ∂x ∂x ∂y ∂y (9) where C is the mean concentration in the water column in mgL−1 , H is total depth in m and Kh is horizontal eddy viscosity in m2 s−1 . Considering a conservative tracer, sources/sinks and reaction terms are assumed negligible. 2.4 Unstructured Grid IPH-UnTRIM2D calculates free surface, velocity and tracer distribution at the centres and sides of

Paper IV

F. F. Pereira et al.: Assessment of Numerical Schemes

280

285

290

computational cells on an unstructured orthogonal grid. An unstructured grid is a set of nonoverlapping computational cells over a simulation domain. Each of these computational cells is com- 325 posed of Np sides, Nv vertices and only one centre, which is located neither at its centroid nor geometric centre but at the intersection point of the perpendicular bisectors of Np sides of the computational cell. Since the segment that joins the 330 centres of two adjacent computational cells are orthogonal to each other, this set of non-overlapping computational cells is known as unstructured orthogonal grid (Fig 1). Figure 1

295

300

305

310

MTOOL (Two-dimensional Mesh Tool) has 335 been used as grid generation package in order to generate the unstructured orthogonal grids used in this study. It is an interactive graphic program for two-dimensional finite element mesh generation developed by TeCGraf (1992). It allows to combine on the same grid coarser and finer triangular and/or quadrilateral computational cells to repre- 340 sent transitions from broad and open regions and to regions of interest. Figure 1 illustrates an unstructured orthogonal grid. For an unstructured orthogonal grid of Np computational cells, which i = 1, 2, 3 · · · Np , 345 each i − th computational cell has area denoted by Pi and 3 or 4 sides represented by j = 1, 2, 3 − −4. And, for each side j, λj indicates its length whereas δj means the distance between the centres of two adjacent computational cells that share 350 the same side j. 2.5

315

320

Numerical Approximation

A semi-implicit finite-volume scheme was used 355 to obtain an efficient numerical algorithm where its stability is independent of free-surface gravity waves, wind stress, vertical viscosity and bottom friction. In contrast to circulation models based on structured grids which numerically solve the mo- 360 mentum equations for each computational polygon on the Cartesian axes, IPH-UnTRIM2D differentiates the momentum equations for each side

5

j of a computational polygon i with respect to its normal and tangential axes. Normal and tangential components of bottom friction and wind stresses are calculated by rotating their horizontal and vertical components. The bottom friction, wind stresses, Coriolis acceleration and horizontal viscosity are treated fully explicitly. Thus, the conservation of normal and tangential momentum for a side j take the following forms:   ∆t n n − ηi(j,1) ··· (1 − θ) ηi(j,2) δj   ∆t n+1 n+1 − g θ ηi(j,2) − ηi(j,1) (10) δj

un+1 = F unj − g j

  ∆t n n (1 − θ) ηv(j,2) − ηv(j,1) ··· λj   ∆t n+1 n+1 − g θ ηv(j,2) (11) − ηv(j,1) λj

vjn+1 = F vjn − g

where u and v are the normal and tangential components of velocity at time step n + 1 and F is an explicit operator which combines all explicit terms with the backtracked velocity at time step n. The implicitness factor for temporal discretization (θ) may vary from 0 to 1, given that values of θ equals to 0 indicate a fully explicit scheme whereas values of θ equal to 1 mean that an implicit scheme is being used. For semi-implicit schemes, the implicitness factor is in the range of 1/2 ≤ θ < 1 (Casulli, 1990; Casulli and Cheng, 1992; Zhang and Baptista, 2007). Stability numerical analysis for semi-implicit schemes carried out by Casulli and Cattani (1994) indicated highest accuracy and efficiency for an implicitness factor equals to 0.5, the same value of θ used in this study. Regarding the grid connectivity, i (j, 1) and i (j, 2) represent neighboring polygons of the jth side whereas v (j, 1) and v (j, 2) indicate its vertices. Water-surface elevations are computed for all polygons and vertices on the grid. At a generic vertex, the water-surface elevation (ηv ) is estimated by area-weighted average using elevations

115

Assessment of Numerical Schemes for Solving Advection-Diffusion Equation on Unstructured Grids: Case Study of River Guaíba, Brazil

6

365

370

F. F. Pereira et al.: Assessment of Numerical Schemes   K1 · · · 0 at its surrounding polygons. On the other hand, a  ..  G =  . Ki 0 , being that set of equations comes from the continuity equation in order to calculate the elevations at all poly0 · · · KNp gons. Thus, the continuity equation can numerically be written for each polygon as: Nj X S(i,l) λj(i,l) Hj(i,l) ··· Ki = Pi ηin − θ∆t 1 + γj(i,l) l=1   Nij     X ∆t n n n Pin+1 = Pin −θ∆t S(i,l) λ(i,l) H(i,l) un+1 ··· + ∆tτj(i,l) − ηi(j,1) F unj − g (1 − θ) ηi(j,2) (i,l) · · · δj(i,l) i=1 − (1 − θ) ∆t

375

385

S(i,l) λ(i,l) H(i,l) un(i,l)

i=1



(12) 395

where Nij represents the number of sides of the polygon i and S(i,l) is a function related to flow direction through the lth side of the polygon i and its value may be equals to 1 or -1. S(i,l) takes the 400 following form:

S(i,l) =

380

Nij  X

i [j (i, l), 2] + i [j (i, l), 1] − 2i i [j (i, l), 2] − i [j (i, l), 1]

(13)

where every j − th side of the grid presents neighbouring polygons of indexes [j (i, l), 1] and [j (i, l), 2]. Replacing 10 into 12, a linear system of Np 410 equations with a symmetric and positive definite matrix can be composed and it takes the following form: Aηin+1 = Gni

(14)

from which,   PN j P1 − j=1 Cj(1,l)   ..   .   P j   A =  Pi − N C , where j(i,l) j=1   ..     . PN j PNp − j=1 Cj(Np ,l)

−gθ2 ∆t2 λj(i,l) Hj(i,l) Cj(i,l) = δj(i,l) (1 + γj (i, l)) 390

116

405

and,

415

420

425

(15)

− (1 − θ) ∆t

Nj X

S(i,l) λj(i,l) Hj(i,l) unj(i,l)

(16)

l=1

where Nj denotes number of sides in computational cell i and Np is the total number of computational cells in the grid. The numerical solution of this linear system can be efficiently determined by conjugate gradient (Press et al., 1992). This numerical approach enables the hydrodynamic module included in IPH-UnTRIM2D to simulate unsteady flow under the effects of external flow disturbances caused by tidal forces and river inflows/outflows, for example. Such external forcings are interpreted by the hydrodynamic module as lateral boundaries, which are assumed strictly vertical. Theoretically, lateral boundary conditions are provided to the hydrodynamic module as described thereafter. Lateral land boundaries are set at all sides along the land-water interface.For these sides of the grid, normal flow and drag on tangential flow are equal to 0. For open sea boundaries, boundary conditions are assumed at the center of computational cells as water surface elevation. And, river boundaries are set at both sides and centers of computational cells by prescribing normal flow (i.e. discharge) and water elevation respectively. Once the elevation field at time step n + 1 is known, normal and tangential velocities are also updated at time step n + 1 by using 10 and 11. Elevations and velocities computed in the hydrodynamic module are used to calculate the numerical solution of advection-diffusion equation. Currently, IPH-UnTRIM2D presents two numerical schemes to solve the advection-diffusion equation. In the first one, IPH-UnTRIM2D em-

Paper IV

F. F. Pereira et al.: Assessment of Numerical Schemes 430

435

440

445

7

ploys a first-order upwind scheme to numerically discretize the advection-diffusion equation over the simulation domain based on geometric fea- 470 Pi Hin+1 Cin+1 = Pi Hin Cin · · ·   tures of the computational cells, their connectivX X ity with each other as well as elevation and wan ··· −∆t  |Qj n+θ |Cin − |Qj n+θ |Cm(i,j) ter velocity fields from the hydrodynamic modj∈Si+ j∈Si− ule. The second one appears as an alternative to     the first-order upwind scheme, and it includes a ∆t  X n+θ n n n  Φj |Qj | Cm(i,j) − Ci ··· − flux-limiter function that aggregates a higher or2 der term over regions of high velocity gradients j∈Si+ ∪Si−   X to reduce numerical diffusion in the solution of n (18) − Cin + ∆t Djn Cm(i,j) the advection-diffusion equation given by the firstj∈Si+ ∪Si− order upwind scheme (Casulli and Zanolli, 2005). The first-order upwind scheme solves temporal 475 Φnj is a flux-limiter function which can be written and spatial partial derivatives from the advectionas: diffusion equation using simple backward differences as shown in equation 17. Φ (r, φ) = max [φ, min ((1, 2r)) , min ((2, r))]

450

455

460

465

(19) Pi Hin+1 Cin+1 = Pi Hin Cin · · ·   where r is the ratio of consecutive gradients and X X n  · · · φ is chosen when a second order accuracy for the −∆t  |Qj n+θ |Cin − |Qj n+θ |Cm(i,j) 480 flux-limiter is sought. Thus, r and φ are calculated j∈Si+ j∈Si− according to:   X n Djn Cm(i,j) + ∆t − Cin (17) ! j∈Si+ ∪Si− 2Djn φnj = min 1, (20) |Qj n+θ | where m(i, j) is the index of the polygon that shares the j − th edge with i − th polygon, Qnj = h  i Kh P n+θ λj Hin uni and Djn = λj Hin δjj are advective and n | Cm(i,j) − Cin j∈Si+ |Qj 1 n diffusive fluxes, respectively. Since S function rerj = n P n+θ Cm(i,j) − Cin | turns either 1 or -1, S + denotes only grid sizes of j∈Si+ |Qj S function equals to 1 whereas S − indicates the (21) negative ones. Similarly to the upwind scheme, the higherAlthough the higher-order flux-limiter scheme order flux-limiter scheme also uses backward dif- 485 uses a few more functions than the upwind ferences to numerically represent the advectionscheme, both schemes are fully explicit. It means diffusion equation. However, the higher-order that in terms of computational efficiency, they are flux-limiter scheme incorporates a high-order equally efficient. However, finer grids may imterm, which depends on a function (Φ) that returns pose severe numerical constraints on these explicit 0 (first-order scheme), 1 (second-order scheme) or 490 time-stepping schemes (Chapra, 2005). 2 (first-order but less diffusive scheme). As proRegarding computational costs, numerical analposed by Sweby (1984), the higher-order fluxysis performed by Pereira (2010) on a Pentium®4 limiter scheme takes the following form: revealed that IPH-UnTRIM2D requires 5 minutes

117

Assessment of Numerical Schemes for Solving Advection-Diffusion Equation on Unstructured Grids: Case Study of River Guaíba, Brazil

8

495

500

505

F. F. Pereira et al.: Assessment of Numerical Schemes

to complete a 24-hour period simulation run for 535 a grid composed of 293 computational cells of 100x100 km. Over a larger simulation domain of 7527 computational cells, IPH-UnTRIM2D needed a little over than 40 minutes to complete the same 24-hour simulation run. For both runs, 540 IPH-UnTRIM2D presented the same CPU time when running with the upwind and flux-limiter schemes. In the next section, IPH-UnTRIM2D has been applied to the River Gua´ıba. Thereafter, its capa- 545 bility in predicting circulation and scalar transport patterns has been discussed. 2.6 Study Area

510

515

River Gua´ıba is a large (surface area 436 km2 ) and shallow (mean depth 6.0 m at full pool) water 550 body that runs through Porto Alegre in the south of Brazil. It plays a fundamental role for transport, irrigation, drinking water supply and wastewater discharge for the region. Its length measures 50 km whereas its width at certain points have cross 555 section up to 15 km. Moreover, River Gua´ıba comprises one of most important freshwater system in Rio Grande do Sul (fig.2) Figure 2

520

525

According to Gua´ıba directive plan, its major environmental problems are domestic sewage and industrial waste effluents from urban areas. On the other hand, rural areas present water pollution by 565 pesticides. Therefore, the River Gua´ıba is often subjected to eutrophication process due to the nutrient enrichment. It becomes more evident during the summer once high levels of temperature and sunlight provide optimal conditions for phy- 570 toplankton blooms. 2.7

530

118

560

Input data

The IPH-UnTRIM2D model requires multiple input data sets which are collected from different 575 databases. All time series data should match the time period for which calibration and predictions are being made. Thus, input data sets were ob-

tained for 1991 due to their consistency over the entire year. Hourly wind speed and direction data has been recorded by Instituto Nacional de Meteorologia (INMET) at a meteorological station located in Porto Alegre. Daily water level and discharge data ´ were provided by Agˆencia Nacional de Aguas, Brazil (ANA). The water level stations are situated at five sites along the River Gua´ıba: Ilha da Pintada, Ipanema, Cristal, Ponta dos Coatis and Ponta Grossa (tab. 1). Table 1 The discharge stations at River Jacu´ı, River Sinos, River Ca´ı and River Gravata´ı were used as upstream boundary conditions whereas the water level station at Ponta dos Coatis was assumed as downstream boundary condition. Since there are no discharge stations at the interface of Rivers Jacu´ı and Gua´ıba, continuous discharge values for River Jacu´ı were estimated based on a regionalization method (Tucci et al., 1995; Samuel et al., 2011). The regionalization method involved transferring discharge data from subsidiaries of the River Jacu´ı to its outlet by using interpolation techniques that depend on geographic location. The discharge stations used ias input to the flow regionalization method are located 20 km upstream from the interface of of Rivers Jacu´ı and Gua´ıba as shown in figure 2. Once the critical flow conditions were reaches in the main channel, a gradual variation of element size from the shoreline to the main channel was used where larger elements were placed along the main channel. MTOOL has therefore created a triangular grid composed of 4622 nodes, 7527 sides and 12156 triangles (fig. 3). A bathymetric survey of River Gua´ıba was carried out by the Diretoria de Hidrografia e Navegac¸a˜ o do Minist´erio da Marinha in 1964. Therefore, average depths at each element were estimated by a linear interpolation (fig. 3). Figure 3

Paper IV

F. F. Pereira et al.: Assessment of Numerical Schemes 2.8

580

585

590

595

600

605

610

615

Hydrodynamic module calibration

620

The numerical approximation of the momentum equations provided parameters related to fluid vis- 625 cosity and diffusivity, wind stress, bottom friction and implicitness for the temporal discretization. Many of these parameters have already been tested and calibrated for rivers and open-channels (i.e. Chezy’s roughness and wind drag coeffi- 630 cients) (French, 1986). Therefore, Chezy’s coefficient (Cz ) was adopted to be constant and equals to 44.7 m1/2 s−1 over the entire domain. A calibration of wind drag coefficient (CD ) for hydrodynamic models was carried out Escalante Estrada 635 et al. (2010). According to their numerical experiments, the wind drag coefficient was assumed to be 0.016. Regarding the eddy horizontal viscosity (Ah ) and diffusivity (Kh ), Fragoso Jr et al. (2008) has successfully used 5 m2 s−1 and 10 m2 s−1 , re- 640 spectively, for modeling spatial heterogeneity of phytoplankton in a shallow lake at South Brazil. For practical applications, the implicitness factor (θ) is recommended to be in the range 0.5 ≤ θ ≤ 1 (Cheng et al., 1993; Zhang and Baptista, 2007). 645 However, numerical analysis performed by Casulli and Cattani (1994) has showed that θ-method was stable and presented the highest efficiency and accuracy for a value of θ equals to 0.50. For calibration, the IPH-UnTRIM2D model 650 used a data network composed of a precipitation station, five water level stations and a meteorological station. A 150 day-period from January 1st to June 1st in 1991 was chosen in order to represent a wide range of hydrologic events (i.e. daily floods and droughts). The hydrodynamic module was calibrated by reproducing water level observations at Ilha da Pintada, Ipanema, Cristal and Ponta Grossa stations (Fig. 2). Afterwards, the conservation of vol- 655 ume was also tested where the balance between incoming and outgoing water fluxes must be equal to the volume due to water level variations in River Gua´ıba.

2.9

9 Assessment of advection-diffusion numerical schemes

The water quality module was tested based on the conservation of mass for both numerical schemes to solve advection-diffusion equation. The efficiency of the higher-order flux-limiter scheme was compared to a first-order upwind scheme by their conservation of mass for River Gua´ıba. In order to perform these comparisons, two scenarios were considered for development of numerical analysis. The mass balance computations for the higherorder flux-limiter and first-order upwind schemes were compared to each other at every time step. The first scenario reproduces a deliberately release of a tracer at a steady rate of 5 mgL−1 into the River Gua´ıba over 15 days. The tracer is released at all interfaces of River Gua´ıba and its subsidiaries. As an initial condition, River Gua´ıba is assumed spatially homogeneous and well-mixed with a tracer concentration equal to 1 mgL−1 over its entire domain. In the second scenario, same initial conditions of tracer concentration are adopted over the River Gua´ıba. However, the release of a tracer remains at a constant rate of 5 mgL−1 only over the first 10 simulation hours. After the first 10 simulation hours, the tracer concentration released at the interfaces of of River Gua´ıba and its subsidiaries is instantaneously reduced to 1 mgL−1 . Errors due to numerical diffusion were assessed by monitoring the tracer concentration of a computational cell located in the Delta of the River Jacu´ı.

3 3.1

Results Hydrodynamic Module

In general, the hydrodynamic module of the IPHUnTRIM2D model successfully reproduced water level observations in response to tidal and runoff daily inflows. Comparisons between observed and calculated elevations showed that IPH-

119

Assessment of Numerical Schemes for Solving Advection-Diffusion Equation on Unstructured Grids: Case Study of River Guaíba, Brazil

10 660

665

670

675

680

685

690

695

120

F. F. Pereira et al.: Assessment of Numerical Schemes

According to the velocity fields (fig. 5), the hyUnTRIM2D model captured the trends of peaks and valleys throughout 150-day period (fig. 4). 700 drodynamic behavior of River Gua´ıba was characterized by higher current velocities in the main channel (ca. 0.1 ms−1 ) and lower values along Figure 4 its shorelines (ca. 0.005 ms−1 ). On the other hand, at the Delta of the River Jacu´ı, velocities Among all the water level stations, Cristal and Ponta Grossa presented a better agreement with 705 and free surface elevations varied depending on the narrowing and widening effects of its set of the observed water levels once they were influstream channels. Although the hydrodynamic soenced by the downstream boundary condition. lution reproduced wetting and drying zones, these For measuring the agreement between observed processes were not considered during the simulaand calculated elevations, the Pearson’s correlation coefficient was calculated for each water level 710 tions due to the lack of topographic data along the shorelines of River Gua´ıba and Delta of the River station as given by equation 22. Jacu´ı.

Pn Xi − X Yi − Y (22) r = q i=1 Figure 5
2 q
2 Xi − X Yi − Y 715 3.2 Numerical schemes for Advectionwhere X and Y denotes the time series of obDiffusion Transport Equation served and simulated water level of length size n, respectively. X and Y are the series averages. TaOnce the hydrodynamic solution was calculated, ble 2 shows the Pearson’s correlation coefficient its velocities and elevations field were employed computed at each gauging station. to solve the advection-diffusion equation using the 720 higher-order flux-limiter scheme. In the first sceTable 2 nario, accumulated mass balance errors for both first-order upwind and higher-order flux-limiter The accuracy of the numerical approximation schemes were computed at every time step of the given by the hydrodynamic module was compared simulation (fig. 6). to the continuity equation for the water budget 725 of River Gua´ıba. Since the hydrodynamic module calculates free surface variations at every time Figure 6 step, the incremental volume at the time step t + 1 is estimated as the differences between free surBoth schemes have showed that their accuface variations between t and t + 1. Theoretically, mulated mass balance errors were lower than the incremental volume must be equal to the dif20% over a period of 15 days after tracer was ferences between incoming and outgoing water 730 released. However, the higher-order flux-limiter fluxes into the River Gua´ıba. scheme used by IPH-UnTRIM2D led to a less difResults obtained from the simulations indicated fusive approach comparing to a first-order upwind that percentage differences between the incremenscheme. At the end of the simulation, accumutal volume and incoming/outgoing water fluxes in lated mass balance errors reached 18% for a firstthe River Gua´ıba were lower than 1% of the to- 735 order upwind scheme whereas the higher-order tal volume of the system. As this error in volume flux-limiter scheme yielded errors below 8%. conservation is lower than uncertainties associated In terms of tracer concentration, 1-day accumuwith data collection procedures (Moradkhani and lated error in mass balance generated by the firstHsu, 2005), it suggests that global (and also local) order upwind scheme ranged from -0.12 to 0.18 volume conservation was achieved. 740 mgL−1 over the entire ecosystem. When derived

Paper IV

F. F. Pereira et al.: Assessment of Numerical Schemes

745

750

755

760

765

770

775

780

785

from the higher-order flux-limiter scheme, these values fluctuated within a range of -0.11 and 0.13 mgL−1 . Since suspended-sediment behaves as a passive tracer, the magnitude of these errors in mass balance are compared to estimates of daily 790 sedimentation rates performed by Stevenson et al. (1985); Ogston et al. (2004); Ogston and Field (2010) as well as measurements of sediment resuspension in estuaries (Bokuniewicz et al., 1991; Hill et al., 2003). Sediment fluxes measured 0.3 m above the bot- 795 tom in Long Island Sound (Bokuniewicz et al., 1991), Mersey estuary and Dover Straits (Hill et al., 2003) (i.e. estuaries located along the shore as River Gua´ıba) varied from 0.00063 to 0.00100 mgcm−2 s−1 . Under the same sediment resuspension or sedimentation rates, the total particle gain 800 or loss due to resuspension or sedimentation over 1-day period in the River Gua´ıba may achieve up to 144 mgL−1 . Estimates of sediment resuspension and sedimentation in bays and estuaries (Stevenson et al., 1985; Ogston et al., 2004; 805 Ogston and Field, 2010) showed that sediment resuspension and sedimentation rates can fluctuate from 50 to 80 mgcm−2 d−1 throughout the year that leads to minimum and maximum daily deposition rates of 83 and 133 mgL−1 , respectively. Therefore, both measured and estimated particle gain or loss due to sediment resuspension or sedimentation were by far larger than error in mass 810 balance generated by the numerical schemes. Differences between the diffusion of higherorder flux-limiter scheme and first-order upwind scheme were evaluated by plotting the evolution of tracer concentration fields over 26, 130 and 815 312 hours (fig. 7). Although both numerical solutions are similar after the first 130 hours of simulation, the higher-order flux-limiter scheme presented faster spreading of tracer concentration over pelagic zones (open water). After 312 hours, 820 tracer concentrations reached Patos Lagoon which indicates that River Gua´ıba has residence time approximately equal to 13 days in accordance with (Rosauro, 1982; Silveira, 1986). 825

11 Figure 7

In the second scenario, after 15 days, accumulated mass balance errors do not exceed 3% for both schemes (fig. 8). Moreover, a slight and constant difference between their mass balance errors was lower than 0.5% for whole simulation where the higher-order flux-limiter scheme showed to be more conservative than a first-order upwind scheme. Figure 8 As the higher-order flux-limiter scheme may employ either first or second order approaches, it leads a less diffusive solution compared to the solution calculated using a first-order upwind scheme. By analyzing the pulse tracer shapes, the higher-order flux-limiter scheme showed slight spreading of tracer and peak tracer concentrations of close to 5 mgL−1 whereas a first-order upwind scheme presented higher diffusing capacity that smoothes gradients of tracer concentration (fig. 9). Figure 9 4

Conclusions

This paper presented comparisons between two different numerical schemes for solving advection-diffusion equation on unstructured grids using IPH-UnTRIM2D. As shown in previous applications of unstructured grid hydrodynamic models Casulli and Walters (2000); Zhang and Baptista (2007), IPH-UnTRIM2D was willing to represent the time series of water level observations and the movement of suspended material throughout River Gua´ıba. Its application to the River Gua´ıba showed that unstructured grid presented higher flexibility to represent the shape of the River Gua´ıba than previous studies on uniform grid (Rosauro, 1982; Silveira, 1986) once unstructured computational cells varied depending on bathymetry, geometry and shoreline-fitting boundary of the River Gua´ıba. The efficiency of the higher-order flux-limiter scheme was tested by comparing its solution to a

121

Assessment of Numerical Schemes for Solving Advection-Diffusion Equation on Unstructured Grids: Case Study of River Guaíba, Brazil

12

830

835

840

845

850

855

F. F. Pereira et al.: Assessment of Numerical Schemes

first-order upwind scheme solution for two sce- 870 Br¨onmark, C. and Hansson, L.-A.: The Biology of Lakes and Ponds, Oxford, University Press, 2nd edn., narios in the River Gua´ıba. According to com2005. parisons, the higher-order flux-limiter scheme was Carrick, H. J., Aldridge, F. J., and Schelske, C. L.: more mass conservative than a first-order upwind Wind influences phytoplankton biomass and composcheme once accumulated mass balance errors sition in a shallow productive lake., Limnology and achieved 18% after a 15-day period for a first- 875 Oceanography, 36, 1179–1192, 1993. order upwind scheme while flux-limiter errors Casulli, V.: Semi-implicit Finite Difference Methods were below 8% in accordance with with numerical for the Two-Dimensional Shallow Water Equations., analysis performed by Casulli and Zanolli (2005) Journal of computational physics, 86, 56–74, 1990. in a curved channel under controlled boundary 880 Casulli, V. and Cattani, E.: Stability, accuracy and conditions. Moreover, accumulated mass balance efficiency of a semi-implicit method for threedimensional shallow water flow, Computers & Matherrors showed that, independently of the numeriematics with Applications, 27, 99–112, 1994. cal scheme employed, mass conservation was proCasulli, V. and Cheng, R. T.: Semi-Implicit Finite portional to the total amount of mass released Difference Methods for Three-Dimensional Shalin the system. Although both schemes have pre- 885 low Water Flow., International journal of numerical sented mass conservation errors, these errors are Methods in Fluids, 15, 629–648, 1992. assumed negligible compared to losses due to eroCasulli, V. and Walters, R. A.: An Unstructured Grid, sion, sedimentation or decay of a substance in esThree-Dimensional Model Based on the Shallow tuaries (Stevenson et al., 1985). 890 Water Equations., International journal of numerical Ongoing studies have been developed to inMethods in Fluids, 32, 331–348, 2000. clude the higher-order flux-limiter scheme on a Casulli, V. and Zanolli, P.: High resolution methods three-dimensional complex dynamic model for for multidimensional advection–diffusion problems in free-surface hydrodynamics., Ocean Modelling, aquatic ecosystems (Fragoso Jr et al., 2008, 2009) 10, 137–151, 2005. in order to create an efficient and capable tool for 895 Chapra, S. C.: Surface water-quality modeling, performing analysis of long-term ecosystem dyMcGraw-Hill, 2005. namics. Acknowledgements. The first two writers were sup- 900 ported by Conselho Nacional de Desenvolvimento Cient´ıfico e Tecnol´ogico (CNPq). Bathymetry data were kindly supplied by Prof. Alejandro Casalas whereas daily discharge and water level time series 905 ´ were freely provided by Agˆencia Nacional de Aguas (ANA).

860

References

865

Blaise, S., Comblen, R., Legat, V., Remacle, J.-F., Deleersnijder, E., and Lambrechts, J.: A discontinuous finite element baroclinic marine model on unstructured prismatic meshes, Ocean dynamics, 60, 915 1371–1393, 2010. Bokuniewicz, H., McTiernan, L., and Davis, W.: Measurement of Sediment Resuspension Rates in Long Island Sound, Geo-Marine Letters, 11, 159–161, 1991.

910

122

Cheng, R. T. and Casulli, V.: Evaluation of the UnTRIM Model for 3-D Tidal Circulation., in: 7th International Conference on Estuarine and Coast Modelling, pp. 628–642, 2001. Cheng, R. T., Casulli, V., and Gartner, J. W.: Tidal, Residual, Intertidal Mudflat (TRIM) Model and its Applications to San Francisco Bay, California., Estuarine, Coastal and Shelf Science, 36, 235–280, 1993. Cotta, R. M.: Integral Transforms in Computational Heat and Fluid Flow, McGraw-Hill, CRC Press, 2005. Escalante Estrada, M., Morales P´erez, R., and Vaquero, P. E.: Calibration of friction coefficients for a 3D hydrodynamic model, Environmental Hydraulics, pp. 165 – 168, 2010. Fox, R. W., McDonald, A. T., and Pritchard, P. J.: Introduction to Fluid Mechanics, John Wiley & Sons Canada, Limited, http://books.google.se/books?id= Bsg8PgAACAAJ, 2009. Fragoso Jr, C. R., Motta Marques, D. M., Collischonn, W., Tucci, C. E. M., and van Nes, E. H.: Modelling spatial heterogeneity of phytoplankton in Lake

Paper IV

F. F. Pereira et al.: Assessment of Numerical Schemes 920

925

930

935

940

945

950

955

960

965

Mangueira, a large shallow subtropical lake in South 970 Brazil, Ecological modelling, 219, 125–137, 2008. Fragoso Jr, C. R., van Nes, E. H., Jansen, J. H., and Motta Marques, D. M. L.: IPH-TRIM3D-PCLake: A three-dimensional complex dynamic model for subtropical aquatic ecosystems., Environmental Mod- 975 elling & Software, 24, 1347–1358, 2009. French, R. H.: Open Channels Hydraulics, McGrawHill, 1986. French, R. H.: Numerical Computation of Internal and External Flows, John Wiley and Sons, 2nd edn., 980 1988. Gross, E. S., Kose, J. R., and Monismith, S. G.: Evaluation of advective schemes for estuarine salinity simulations., Journal of hydraulic engineering, 125, 32– 985 46, 1999. Gross, E. S., Bonaventura, L., and Rosatti, L.: Consistency with continuity in conservative advection schemes for free-surface models., International Journal of Numerical Methods in Fluids, 38, 307–327, 990 2002. Guerrero, J. S. P., Pimentel, L. C. G., Skaggs, T. H., and van Genuchten, M. T.: Analytical solution of the advection–diffusion transport equation using a changeof-variable and integral transform technique., International journal of Heat Mass Transfer, 52, 3297– 995 3304, 2009. Hetch, M. W., Holland, W. R., and Rasch, P. J.: Upwindweighted advection schemes for ocean tracer transport: An evaluation in a passive tracer context., Jour1000 nal of geophysical research, 20, 763–778, 1995. Hill, D. C., Jones, S. E., and Prandle, D.: Derivation of sediment resuspension rates from acoustic backscatter time-series in tidal waters, Continental Shelf Research, 23, 19–40, 2003. Hodges, B. and Dallimore, C.: Estuary, Lake and1005 Coastal Ocean Model: ELCOM, 2nd edn., 2013. Le Veque, R. J.: High-resolution conservative algorithms for advection in incompressible flow., SIAM Journal on Numerical Analysis, 33, 627–665, 1996. Leij, F. J. and van Genuchten, M. T.: Analytical mod-1010 eling of nonaqueous phase liquid dissolution with Green’s functions., Transport in Porous Media, 38, 141–166, 2000. Margolin, L. and Shashkov, M.: Using a Curvilinear Grid to Construct Symmetry-Preserving Discretiza-1015 tions for Lagrangian Gas Dynamics, Journal of Computational Physics, 149, 389–417, 1999. Mikhailov, M. D. and Ozisik, M. N.: Unified analysis and solutions of heat and mass diffusion, Wiley In-

13

terscience, 1st edn., 1984. Moradkhani, H. and Hsu, K.-L.: Uncertainty assessment of hydrologic model states and parameters: Sequential data assimilation using the particle filter, Water Resources Research, 41, 1–17, 2005. Nielsen, P. and Skovgaard, O.: The effect of using nonorthogonal boundary-fitted grids for solving the shallow water equations, International Journal for Numerical Methods in Fluids, 11, 177–188, 2005. Ogston, A. S. and Field, M. E.: Predictions of Turbidity Due to Enhanced Sediment Resuspension Resulting from Sea-Level Rise on a Fringing Coral Reef: Evidence from Molokai, Hawaii, Journal of Coastal Research, 26, 1027–1037, 2010. Ogston, A. S., Storlazzi, C. D., Field, M. E., and Presto, M. K.: Sediment resuspension and transport patterns on a fringing reef flat, Molokai, Hawaii, Coral reefs, 23, 559–569, 2004. Pereira, F. F.: Modelo Hidrodinˆamico e de transporte bidimensional de grade n˜ao estruturada para lagos rasos, Master’s thesis, Instituto de Pesquisas Hidr´aulicas, Porto Alegre, RS, 2010. Press, W. H., Tuskolsky, S. A., Vetterling, W. T., and Flannery, B. P.: Numerical Recipes in FORTRAN, Cambridge University, 2nd edn., 1992. Reynolds, C. S.: The ecology of freshwater phytoplankton, vol. 2nd, Cambridge University Press, 1984. Rosauro, N. M. L.: A Finite Elements Application to the Study of Seiches and Water Circulation in the Jacu´ı Delta, Gua´ıba River and Patos Lagoon, Ph.D. thesis, Instituto de Pesquisas Hidr´aulicas, 1982. Samuel, J., Coulibaly, P., and Metcalfe, R. A.: Estimation of Continuous Streamflow in Ontario Ungauged Basins: Comparison of Regionalization Methods, Journal of Hydrologic Engineering, 16, 447–459, 2011. Sankaranarayanan, S. and Spaulding, M. L.: A study of the effects of grid non-orthogonality on the solution of shallow water equations in boundary-fitted coordinate systems, Journal of Computational Physics, 184, 299–320, 2003. Silveira, A. L. L.: Modelo hidrodinˆamico bidimensional com aplicac¸a˜ o ao rio Gua´ıba, Ph.D. thesis, Instituto de Pesquisas Hidr´aulicas, 1986. Steffler, P. and Blackburn, J.: River 2D: TwoDimensional Depth Averaged Model of River Hydrodynamics and Fish Habitat, 2002. Stevenson, J. C., Kearney, M. S., and Pendleton, E. C.: Sedimentation and erosion in a chesapeake bay brackish marsh system., Marine geology, 67, 213–

123

Assessment of Numerical Schemes for Solving Advection-Diffusion Equation on Unstructured Grids: Case Study of River Guaíba, Brazil

14 1020

1025

1030

1035

1040

1045

1050

1055

124

F. F. Pereira et al.: Assessment of Numerical Schemes

235, 1985. Sweby, P. K.: High resolution schemes using flux limiters for hyperbolic conservation laws., SIAM Journal on Numerical Analysis, 21, 995–2011, 1984. TeCGraf: Manual do Usu´ario, Tech. Rep. 1, Grupo de Tecnologia em Computac¸a˜ o Gr´afica/PUC Rio, 1992. Thompson, J. E., Warsi, Z. U. A., and Mastin, C. W.: Numerical Grid Generation: Foundations and Applications, Elsevier Science Publishing Co., New York, US, 1985. Tucci, C. E. M., Silveira, A. L. L., and Sanchez, J.: Flow regionalization in the upper Paraguay basin, Brazil., Hydrological sciences journal, 40, 485–497, 1995. Wang, B., Fringer, O. B., Gridding, S. N., and Fong, D. A.: High-resolution simulations of a macrotidal estuary using SUNTANS., Ocean Modelling, 26, 60– 85, 2008. Westerink, J. J., Blain, C. A., Luettich Jr, R. A., and Scheffner, N. W.: ADCIRC: An advanced threedimensional circulation model for shelves, coasts and estuaries, Tech. Rep. 2, U.S. Army Corps of Engineers Waterways Experiment Station, Mississippi, US, 1994. Wood, T. M., Cheng, R. T., Gartner, J. W., Hoilman, G. R., Lindenberg, M. K., and Wellman, R. E.: USGS Scientific Investigation : Modeling Hydrodynamics and Heat Transport in Upper Klamath Lake, Oregon, and Implications for Water Quality., Tech. Rep. 1, USGS, 2008. Ye, J. and McCorquodale, J. A.: Depth-averaged Hydrodynamic Model in Curvilinear Collocated Grid., Journal of hydraulic engineering, 123, 380–388, 1997. Zhang, Y. and Baptista, A. M.: SELFE: A semi-implicit Eulerian–Lagrangian finite-element model for crossscale ocean circulation., Ocean Modelling, 21, 71– 96, 2007. Zhou, J. G.: Velocity-depth coupling in shallow water flows., Journal of hydraulic engineering, 121, 717– 724, 1995.

Paper IV

F. F. Pereira et al.: Assessment of Numerical Schemes

Status Boundary Condition Upstream

Control Points

Boundary Condition Downstream

15

Station number 86720000 85900000 87382000 87170000 87399000 87450005 87460120 87450100 87460007 87460220

Position Lat Long -29140 0400 -51510 1800 0 00 -2959 41 -52220 3800 -29450 3200 -51090 0200 -29350 1900 -51220 5600 -29570 5200 -50580 4000 -30010 5000 -51150 0700 -30080 0200 -51140 0200 -30020 5600 -51110 4800 -30050 3200 -51150 0100 -30110 1900 -51140 3300

Station name Encantado Rio Pardo S˜ao Leopoldo Barca do Ca´ı Passo das Canoas Ilha da Pintada Ipanema Ipiranga Cristal Ponta Grossa

Jacu´ı Taquari Sinos Ca´ı Gravata´ı Gua´ıba Gua´ıba Gua´ıba Gua´ıba Gua´ıba

Available data Discharge Discharge Discharge Discharge Discharge Water level Water level Water level Water level Water level

87500020

-30150 3200

Ponta dos Coatis

Gua´ıba

Water level

-51090 2000

River

Table 1: Streamflow and water levels gauging stations of River Gua´ıba and its tributaries used in this study.

Station name Ilha da Pintada Ipanema Cristal Ponta Grossa

R-squared 0.85 0.89 0.93 0.90

Table 2: Comparisons between observed and simulated water elevation data at the gauging stations (R-squared).

125

Assessment of Numerical Schemes for Solving Advection-Diffusion Equation on Unstructured Grids: Case Study of River Guaíba, Brazil

16

126

F. F. Pereira et al.: Assessment of Numerical Schemes

Fig. 1: Illustration of an unstructured orthogonal grid (solid lines). An auxiliary grid (dashed lines) composed of segments that join the centres of each computational cell is also shown to highlight geometric variables used in the section 2.9

Paper IV

F. F. Pereira et al.: Assessment of Numerical Schemes 17 Fig. 2: Location of the River Gua´ıba and municipalities on its banks. It also shows the composed of a group of 5 water level stations (circles) and 2 discharge stations (triangles).

127

Assessment of Numerical Schemes for Solving Advection-Diffusion Equation on Unstructured Grids: Case Study of River Guaíba, Brazil

18

128

F. F. Pereira et al.: Assessment of Numerical Schemes

Fig. 3: Spatial discretization of the river Gua´ıba using an orthogonal triangular grid (a) and its bathymetry (b).

Paper IV

F. F. Pereira et al.: Assessment of Numerical Schemes Fig. 4: Calculated and observed elevation of water surface at four control points along the River Gua´ıba: (a) Ilha da Pintada; (b) Ipanema; (c) Cristal and (d) Ponta Grossa.

19

129

Assessment of Numerical Schemes for Solving Advection-Diffusion Equation on Unstructured Grids: Case Study of River Guaíba, Brazil

20

F. F. Pereira et al.: Assessment of Numerical Schemes

Fig. 5: Two dimensional surface velocity field over the River Gua´ıba.

130

Paper IV

F. F. Pereira et al.: Assessment of Numerical Schemes

20

Upwind Flux limiter

18

Accumulated error (%)

16 14 12 10 8 6 4 2 0

0

5

10

15 21

Time (days) Fig. 6: Accumulated mass balance errors for the flux limiter and upwind schemes over 15 days of simulation.

131

Assessment of Numerical Schemes for Solving Advection-Diffusion Equation on Unstructured Grids: Case Study of River Guaíba, Brazil

22

132

F. F. Pereira et al.: Assessment of Numerical Schemes

Fig. 7: Concentration field of a tracer plume calculated using the flux limiter and upwind schemes for 26, 130 and 312 hours in the River Gua´ıba where red computational cells denote values of concentration higher than 4.5 mgL−1 while blue cells mean values of concentration lower than 2.0 mgL−1 .

Paper IV

F. F. Pereira et al.: Assessment of Numerical Schemes

20

Upwind Flux limiter

18

Accumulated error (%)

16 14 12 10 8 6 4 2 0

0

5

10

15 23

Time (days) Fig. 8: Comparisons between accumulated mass balance errors considering a pulse input of tracer concentration into River Gua´ıba during 10 hours for the flux limiter and upwind schemes.

133

Assessment of Numerical Schemes for Solving Advection-Diffusion Equation on Unstructured Grids: Case Study of River Guaíba, Brazil

24

6

Upwind Flux limiter

5.5

4.5 4 3.5 3 2.5 2 1.5 1 0.5

0

5

10 15 Time (hours)

20

Fig. 9: Pulse of a passive tracer at the Delta of the River Jacu´ı.

134

25

F. F. Pereira et al.: Assessment of Numerical Schemes

Tracer concentration (mg/L)

5

PAPER V Implementation of a two-way coupled atmospheric-hydrological system for environmental modeling at regional scale F. F. Pereira, M. A. E. de Moraes and C. B. Uvo. Hydrology Research, Accepted.

Paper V

Uncorrected Proof 1

© IWA Publishing 2013 Hydrology Research

|

in press

|

2013

Implementation of a two-way coupled atmospherichydrological system for environmental modeling at regional scale Fábio Farias Pereira, Marcio A. E. de Moraes and Cintia Bertacchi Uvo

ABSTRACT This work describes the two-way coupling performed between the regional atmospheric model Brazilian Regional Atmospheric Modeling System (BRAMS) and the hydrological model MGB-IPH. As a first step of the atmosphere-hydrology coupling, only the water balance variables were coupled. Differences in temporal and spatial scales between MGH-IPH and BRAMS were analyzed. By default, MGB-IPH has a daily time step whereas BRAMS uses smaller time steps. Thus, accumulated rainfall values from BRAMS were used to feed MGB-IPH. On the other hand, daily values of evapotranspiration from MGB-IPH were provided to BRAMS. This procedure was assumed as a daily loop in the simulations. Differences in spatial scales were avoided by using the same grid size (10 ×

Fábio Farias Pereira (corresponding author) Cintia Bertacchi Uvo Department of Water Resources Engineering, Lund University, Lund, Sweden E-mail: [email protected] Marcio A. E. de Moraes National Institute for Space Research (INPE), São José dos Campos, SP, Brazil

10 km) in both models, in such a way that neither upscaling nor downscaling was necessary. The coupled system was tested for the Rio Grande basin situated in south-eastern Brazil by comparing results from BRAMS with results from the coupled system for the same period, with the same input data. Outputs from the runs were compared to water vapor satellite images. The results from the coupled model tests indicated that its predictions of rainfall distribution were more accurate than BRAMS. Key words

| atmospheric model, distributed hydrological model, two-way coupling

INTRODUCTION Interactions between land surface and atmosphere induced

among them, regional climate models (RCMs) stand out by

by human activities and natural environmental dynamics

including a wide range of transfer processes between land sur-

act on a time scale that varies from seconds to millions of

face, atmosphere and oceans, from root water uptake to

years. By exchanging heat, water, energy and carbon, land

transport of atmospheric aerosols (Liston & Pielke ).

surface and atmospheric processes are closely interrelated

One disadvantage of such models is that their local surface

and influence each other in reciprocal ways (Pielke et al.

hydrology does not consider the river routing; thereby esti-

; Betts ; Field et al. ). Therefore, estimating

mates of soil moisture content are underestimated in areas

the exchanges of heat, water, energy and carbon between

close to the drainage network (Graham et al. ). Yet another

land surface and atmosphere considering their interplay is

issue is associated with land surface parameterizations used by

an important step towards the understanding of the impacts

RCMs, since parameterization schemes usually apply pre-

of anthropogenic actions as potential forcing mechanisms

scribed values of parameters based on their probability

for climate regime shifts.

density functions. This assumption does not interpret land

In this context, several numerical models have systemati-

use and soil characteristics as continuous distributions, and

cally been developed and enhanced to better represent land

hence, mixtures in soil and vegetation within an area of interest

surface/atmosphere feedback loops (Benoit et al. ), and,

are not captured (Walko et al. ; Beven & Freer ).

doi: 10.2166/nh.2013.335

137

Implementation of a two-way coupled atmospheric-hydrological system for environmental modeling at regional scale

Uncorrected Proof 2

F. F. Pereira et al.

|

Hydrology Research

Two-way coupled atmospheric-hydrological system at regional scale

|

in press

|

2013

In an attempt to simulate surface runoff on a daily basis,

hydrological model. Despite their results revealing that the

Hay et al. () proposed a one-way coupling of a RCM and

two-way coupled model improved the predicted energy

a distributed hydrological model. Their approach suggests

fluxes and rainfall in comparison with predictions made by

that outputs of precipitation and temperature from the

the RCM during a 3-day forecast period, evapotranspiration

RCM are used as input to the hydrological model. This

rates are not only dependent on albedo but also on leaf area

methodology incorporates the effects of the drainage net-

index, rooting depth and bulk stomatal resistance. Also,

work when calculating the soil moisture content, and

values of evapotranspiration may be over- or underestimated

replaces land surface parameterizations by a process-based

after a longer period of simulation.

method to estimate daily surface runoff; though feedback

In order to bridge this gap, this study presents the

effects of land surface dynamics from the hydrological

implementation of an atmospheric-modeling system com-

model are not included in calculations of precipitation and

posed of a two-way coupling between the Brazilian

temperature time series provided by the RCM.

Regional Atmospheric Modeling System (BRAMS; Freitas

Therefore, the need for better representation of land sur-

et al. ) and the hydrological Model for Large Basins

face hydrological processes and their feedback mechanisms

(MGB-IPH; Collischonn ) in a way to optimize their

into RCMs has been strongly suggested by several numerical

respective strengths. Since MGB-IPH incorporates a pro-

studies (Baron et al. ; Bartholmes & Todini ; Lin

cess-based approach to estimate evapotranspiration rates

et al. ; Messager et al. ; Haggag et al. ). In

considering values of albedo, leaf area index, sunshine

this sense, a more sophisticated approach proposed by

hours, relative humidity, atmospheric pressure and wind

Walko et al. () presents a two-way coupling of a RCM

speed, this coupling methodology uses estimates of evapo-

and a hydrological model. Their coupled system includes

transpiration by MGB-IPH in BRAMS. On the other hand,

turbulent and radiative exchange of heat and water between

BRAMS provides daily precipitation as input to MGB-IPH.

soil, vegetation, canopy air and atmosphere. However, sensi-

Spatial and temporal mismatches associated with the coup-

tivity tests of this coupled system are only performed in

ling methodology are also presented. As a case study, the

idealized model simulations. Moreover, Walko et al. ()

atmospheric-hydrological modeling system is applied to

use a parametric model developed by Louis () to rep-

the Rio Grande basin, Brazil.

resent fluxes of water vapor between land surface and atmosphere. This parametric model assumes that momentum roughness length is equal to heat transfer roughness

MATERIAL AND METHODS

length, and the height of the lowest model level is much larger than momentum roughness length. These assump-

The implementation of the atmospheric-hydrological model-

tions are well supported by experimental evidence for

ing system consists of a two-way coupling of BRAMS and

smooth surfaces (Phelps & Pond ; Högström & Smed-

MGB-IPH models. Besides the experience of the authors

man-Högström ), though are not valid for rough

with them, the choice of these models was also based on

surfaces and/or mountainous regions (Van Den Hurk &

their successful use in previous studies in the Rio Grande

Holtslag ; Kot & Song ).

basin (Nóbrega et al. ; Bender & de Freitas ). Prior

A similar approach has been used by Seuffert et al.

to implementing this integrated modeling system, a brief

() to evaluate the influence of land surface hydrology

description of BRAMS and MGB-IPH including how land

on the predicted local weather. However, their study con-

surface hydrological processes are interpreted by each of

sists of a two-way coupling between a RCM and a

the models, and what kinds of ecosystems these models

hydrological model through a coupling strategy that, firstly,

have successfully been applied to are presented in order to

supposes that the hydrological model estimates the turbu-

figure out the best coupling approach between BRAMS

lent diffusion coefficients of heat and momentum more

and MGB-IPH.

realistically than the RCM, and, secondly, replaces values

Once a proper two-way coupling strategy has been

of albedo from the RCM with those calculated by the

defined, spatial and temporal mismatches are adequately

138

Paper V

Uncorrected Proof 3

F. F. Pereira et al.

|

Hydrology Research

Two-way coupled atmospheric-hydrological system at regional scale

|

in press

|

2013

addressed as part of the implementation procedure. As

is estimated using the Penman-Monteith equation and rout-

BRAMS and MGB-IPH present their own spatial-temporal

ing through the river network uses the Muskingum-Cunge

discretizations, which count on their respective grid gener-

method. Meteorological conditions are prescribed based

ation and time step settings, the two-way exchange of

on interpolation of nearby measurement stations. By default

variables between BRAMS and MGB-IPH is incorporated

MGB-IPH is employed using a daily time step, however

together with an algorithm that solves temporal and spatial

depending on the purpose of study, it might become smaller

mismatches.

or larger.

The atmospheric-hydrological modeling system is then

The MGB-IPH parameters are related to classes of phys-

evaluated by means of comparisons of amount of instan-

ical characteristics, such as soil type, land use, geology and

taneous rainfall between results from two short-term runs

vegetation.

of 31 days each performed for a wet period when the atmosphere is dominated by fronts, and local convection. The first

BRAMS atmospheric model

run was carried out using the regional atmospheric model whereas the atmospheric-hydrological modeling system

BRAMS is part of the operational weather prediction system

was applied in the second run. Based on atmospheric activi-

of the Center for Weather Forecast and Climate Studies

ties detected by the presence/absence of clouds in visible

(CPTEC), belonging to the National Institute for Space

satellite images, two case studies were selected representing

Research (INPE) in Brazil (http://brams.cptec.inpe.br). The

a cold front passage and local strong convection over the

BRAMS is a multipurpose, numerical prediction model

Rio Grande basin.

designed to simulate atmospheric circulations spanning from hemispheric scales down to large eddy simulations of

MGB-IPH hydrological model

the planetary boundary layer (Walko et al. ; www.

The large-scale hydrological model MGB-IPH (Collischonn

ing scheme which allows the model equations to be solved

atmet.com). The model is equipped with a multiple grid nest) is a distributed model based on the LARSIM (Bre-

simultaneously on any number of interacting computational

micker ) and VIC-2L (Liang et al. ) models, that

meshes of differing spatial resolution. It has a complex set

consist of modules for calculating soil water budget, evapo-

of packages to simulate processes such as radiative transfer,

transpiration, flow propagation within a cell, and flow

surface-air water, heat and momentum exchanges, turbulent

routing through the drainage network (Collischonn et al.

planetary boundary layer transport, and cloud microphysics.

). MGB-IPH has been tested and used in several

The initial conditions can be defined from various observa-

South American basins, from rapid-response ones of

tional data sets that can be combined and processed with a

southern Brazil and Uruguay to very low-response ones

meso-scale isentropic data analysis package. For the bound-

such as the Pantanal, the large wetland in the Upper Para-

ary

guay river basin. It also has been applied for several

technique described by Umeda & Martien () is used to

purposes, such as flow forecasting (Tucci et al. ) and

interpret atmospheric boundary conditions provided every

to estimate daily water balance in large basins (Collischonn

6 h by global atmospheric analyses. BRAMS features used

et al. ).

in this system include an ensemble version of a deep and shal-

conditions,

a four-dimensional

data

assimilation

MGB-IPH divides each computational cell into hydrolo-

low cumulus scheme based on the mass flux approach and

gic response units (HRUs) based on its land use/cover and

soil moisture initialization data. The surface-atmosphere

soil distribution. HRUs are then defined by intersecting

water, momentum and energy exchanges are simulated by

land use and soil groups within a computational cell.

the Land Ecosystem Atmosphere Feedback model (LEAF-

Once all computational cells are classified into different

3), which represents the storage and vertical exchange of

groups with similar hydrological response, MGB-IPH calcu-

water and energy in multiple soil layers, including the effects

lates the soil water budget, evapotranspiration and flow

of freezing and thawing soil, temporary surface water or snow

propagation (Collischonn et al. ). Evapotranspiration

cover, vegetation, and canopy air (Lee & Pielke ).

139

Implementation of a two-way coupled atmospheric-hydrological system for environmental modeling at regional scale

Uncorrected Proof 4

F. F. Pereira et al.

|

Hydrology Research

Two-way coupled atmospheric-hydrological system at regional scale

|

in press

|

2013

In order to merge the capabilities of several numerical weather codes, BRAMS was implemented using the concept of ‘plug-compatible’ modules given by Pielke & Arritt (). Although this concept allows the easy incorporation of improvements between the sub-routines of the model, it also stimulates the use of parameterizations by the developers and users of the model. To represent surface layer fluxes of water vapor into the atmosphere, LEAF-3 uses a parametric model developed by Louis (). Similarly to any other trace gas, such as ozone (O3) and carbon dioxide (CO2), his parameterization scheme estimates the fluxes of water vapor using Businger’s profile functions (Businger et al. ). Once computed, the fluxes of water vapor are interpreted by BRAMS as the lower boundary for the atmosphere. Two-way coupling methodology In the two-way coupling of BRAMS and MGB-IPH, processbased estimates of evapotranspiration rates by MGB-IPH replace fluxes of water, from canopy air to atmosphere, cal-

Figure 1

|

A schematic of components of the LEAF-3 and their interactions with MGB-IPH over an entire column (Adapted from Walko et al. 2000).

culated using land surface parameterizations in the LEAF-3 routines of BRAMS as shown in Figure 1. Figure 1 shows a

speed, and hence, it is considered more accurate and com-

schematic of components of the LEAF-3 routine and their

prehensive than land surface parameterizations used by

interactions with MGB-IPH, where atmosphere (A), veg-

LEAF-3.

etation cover (V), canopy air (C) and two soil layers (G1

As both models present their own surface grid and time

and G2) are divided into multiple vertical layers. Moreover,

step, the two-way exchange of variables between BRAMS

vertical fluxes of heat, water and long wave radiation are

and MGB-IPH includes two strategies to solve spatial and

represented by their subscripts h, w and r as well as the

temporal mismatches. These strategies are separately

source and receptor (g for ground, s for snow, v for veg-

described as follows.

etation, c for canopy air, and a for free atmosphere). Therefore, evapotranspiration rates from MGB-IPH are

Strategy for solving temporal mismatches

equivalent to WCA (Water between Canopy air and Atmosphere) in LEAF-3.

Due to numerical stability constraints, BRAMS runs with a

On the other hand, daily accumulated rainfall estimated

smaller time step than the daily time step often used by

by BRAMS is provided as input to MGB-IPH. However,

MGB-IPH at basin scale. Therefore, the two-way exchange

unlike the stand-alone version of BRAMS, the flux of

of variables between BRAMS and MGB-IPH presents a tem-

water from canopy air to atmosphere incorporates feedback

poral coupling in a way that MGB-IPH is employed as a

from a process-based approach to land surface hydrological

subroutine of BRAMS which is called every 24 simulation-

processes given by MGB-IPH into calculations of daily accu-

hours as depicted in Figure 2. Specifically, at this time

mulated rainfall. In this process-based approach to land

step, calculations of flux of water between canopy air and

surface hydrological processes, MGB-IPH calculates fluxes

atmosphere in LEAF-3 are switched off. Complementarily,

of water from land surface to atmosphere based on air temp-

observed daily rainfall was not given as input to MGB-IPH

erature, relative humidity, long wave radiation and wind

at any time step. Thus, errors in energy and water balance

140

Q8

Paper V

Uncorrected Proof 5

F. F. Pereira et al.

|

Hydrology Research

Two-way coupled atmospheric-hydrological system at regional scale

|

in press

|

2013

Model application The Rio Grande basin was used as a case study for evaluating the two-way coupled system. The Rio Grande basin is Figure 2

|

Temporal coupling of BRAMS and MGB-IPH. As MGB-IPH runs on a daily basis

located in the eastern upper Paraná basin (Figure 4). The

while BRAMS uses shorter time steps, coupling variables are exchanged every 24 h.

basin is also formed by important subsidiaries rivers such as rivers Pardo and Mogi-Guaçu. Approximately 60% of

computations owing to modifications arising from the twoway coupling methodology are not incorporated into the coupled model.

hydroelectric power generation in Brazil is provided by the Paraná River basin of which approximately 12% comes from the 15 hydropower plants in the Rio Grande basin (Agência Nacional de Energia Elétrica [ANEEL] ). The altitude in the basin varies from 300 to 2,700 m.a.s.l.

Strategy for solving spatial mismatches

and the land use is composed mostly of agriculture and pasture in the low lands and forest in the high lands.

Although the two-way coupling methodology avoids

The simulation domain corresponds to an area of 120 ×

upscaling and downscaling issues by running both models

100 km that covers the Rio Grande basin. The horizontal

on the same grid cell size, BRAMS and MGB-IPH output

and vertical grid spacing were defined equal to 10 km. The

variables are calculated at their own computational cell

atmospheric variables were driven at the lateral boundaries

centers that do not always match each other. In order to

using reanalysis from the Eta/CPTEC model. Weekly sea

correctly address the two-way exchange of variables

surface temperature (SST) was assumed on a one-degree

between BRAMS and MGB-IPH, an algorithm has been

grid (Reynolds et al. ) and topographic data from the

developed to calculate and rank the distances between

US Geological Survey (USGS) was interpolated to the

BRAMS and MGB-IPH computational cell centers. It

coupled model resolution.

also sorts these distances in ascending order, and identifies the nearest computational cell centers for the two-way

Model runs

exchange of variables. Figure 3 illustrates this two-way exchange according to the procedure carried out by the

Two different runs are performed for a simulation period of

algorithm.

31 days. The first run corresponds to the control run and, the stand-alone version of BRAMS is applied to the Rio Grande basin while the second run is carried out using the atmospheric-hydrological modeling system. This study evaluates the capability of each model to reproduce rainfall occurrence. For doing so, instantaneous rainfall fields from the BRAMS and the coupled model are compared to satellite images with respect to spatial distribution of rainfall and presence of clouds with high water content. Thereby, a wet weather period was chosen due to its high probability of rainfall occurrence. According to Nóbrega et al. (), austral summer is the rainy season at Q1 the Rio Grande basin, therefore the chosen simulation period spans 1st January through 31st January 2009. Prior to each run, a warming-up period of 10 days was

Figure 3

|

Scheme of the two-way exchange of variables between BRAMS and MGB-IPH computational cell centers. An algorithm selects the nearest computational

considered for the initialization of the physical variables

cell centers before exchanging the coupling variables.

(Benoit et al. ). Thus, all comparisons presented in

141

Implementation of a two-way coupled atmospheric-hydrological system for environmental modeling at regional scale

Uncorrected Proof 6

F. F. Pereira et al.

Figure 4

|

|

Hydrology Research

Two-way coupled atmospheric-hydrological system at regional scale

|

in press

|

2013

Maps of Paraná river basin and Rio Grande basin with its main tributaries (i.e. Rivers Grande, Pardo and Mogi-Guaçu) and Rio Grande basin topography (see online version for colours: http://www.iwaponline.com/nh/toc.htm).

this study refer to results obtained from the last 20 days of

presence of brighter clouds is used as indicator of precipi-

simulation, which means, from 11th to 31st January 2009.

tation occurrence.

Criteria for selection of case studies

the presence/absence of brighter clouds in the atmosphere.

In this study, visible satellite images are used to analyze These images were captured every 6 h for the period of simuOutputs from both the atmospheric-hydrological modeling

lation after the warming-up period, 10th of January 2009 to

system and BRAMS are compared to each other and ana-

31st of January 2009. Four out of 160 satellite images clearly

lyzed for the formation of clouds and precipitation

present brighter clouds over the Rio Grande basin, which

occurrence. Since brightness of clouds in satellite images

represent two distinct events. The first event consists of a

is related to their water content (Song et al. ), the

cold front passage that spans from January 22nd at 0000

142

Q9

Paper V

Uncorrected Proof 7

F. F. Pereira et al.

|

Hydrology Research

Two-way coupled atmospheric-hydrological system at regional scale

|

in press

|

2013

UTC (Coordinated Universal Time) to January 23rd at 0000

analysis defines as convective rainfalls, rainfall events

UTC. The second one is characterized by a local strong con-

higher than 3.5 mm h�1.

vection on January 15th at 1200 UTC. These two events are then selected as study cases.

RESULTS Assessment of the coupled model In this section, results from the models runs for each study The performance of each model is evaluated with respect to

case are presented. The study cases represent two particular

its capability of representing the spatial distribution of

atmospheric events driven by different physical processes.

instantaneous rainfall. Since satellite coverage is continuous

In the first study case, for instance, the ability of each

over most regions of the Earth surface and several satellite-

model to reproduce instantaneous rainfall derived from a

based rainfall estimation techniques have proved that rain-

cold front passage is evaluated. While the second study

fall occurrence is closely associated with brightness of

case assesses the models when representing local rainfalls

visible images (Adler & Andrew ; Porcú et al. ;

formed from convections. As rainfall is a response of the

Todd et al. ; Coppola et al. ; Lintner et al. ),

atmosphere to sharp variations of air temperature, atmos-

the spatial distribution of instantaneous rainfall given by

pheric pressure, relative humidity and air flow, the

BRAMS and the coupled model are compared to visible sat-

agreement between instantaneous rainfall fields and satellite

ellite images captured by a geostationary satellite focused on

images is used to evaluate the performance of the models as

South America and operated by the National Oceanic and

done by McMurdie & Katsaros () and O’Sullivan et al. Q2

Atmospheric Administration (NOAA), the geostationary

().

operational environmental satellite 10 (GOES-10).

Figure 5 shows the spatial distribution of precipitating

Although there is a time difference before the probable

clouds (first column) and instantaneous rainfall fields esti-

rain (i.e. brighter clouds in GOES-10 visible images) falls

mated by BRAMS (second column) and the coupled

to the surface (from 5 to 8 min; Jakob & Klein ), it is

model (third column) on 22nd January 2009 00:00 UTC,

assumed that precipitating clouds do not move out of com-

06:00 UTC, 12:00 UTC and on 23rd January 2009 00:00

putational cells of 10 × 10 km in this small time interval.

UTC for the first study case.

This assumption is valid for comparisons made between sat-

According to Figure 5, the presence of precipitating

ellite images and instantaneous rainfall fields (O’Sullivan

clouds indicates convergences all over the Rio Grande

et al. ). However, since rainfall observations from the

basin, which characterizes a cold front passage on 22nd Jan-

Tropical Rainfall Measurement Mission (TRMM) and rain

uary 2009 at 00:00 UTC and 23rd January 2009 at 00:00

gauge stations are respectively 3 h and daily accumulated

UTC. In addition, stable atmospheric conditions are

data, they are not used for comparisons due to their tem-

observed on 22nd January 2009 at 06:00 UTC and 18:00

poral resolution. Moreover, TRMM estimates are given at

UTC and they are used to investigate possible numerical

a 0.25-degree by 0.25-degree spatial resolution, thereby

instabilities in the coupled model when the atmosphere

quantitative comparisons between these estimates of rainfall

shifts from unstable to stable conditions, and vice versa.

and simulated values from the models would be affected by upscaling/downscaling approaches.

Under stable conditions, interactions between atmosphere and land surface slow down and the amount of

Once the spatial distribution of instantaneous rainfall

instantaneous rainfall is lower than 0.5 mm h�1 over the

fields estimated by BRAMS and the coupled model match

Rio Grande basin for both models. Accordingly, instan-

probable rains in GOES-10 visible images, the order of mag-

taneous rainfall fields calculated by BRAMS and the

nitude of instantaneous rainfall rates are compared to values

coupled model present the same low rainfall profile as

of rainfall obtained from studies on the interannual variabil-

observed in satellite images. It means that, despite passing

ity of extreme events in the same study area carried out by

through unstable atmospheric conditions due to a cold

Leibmann et al. (). For the Rio Grande basin, their

front, the coupled model converges to the same rainfall

143

Implementation of a two-way coupled atmospheric-hydrological system for environmental modeling at regional scale

Uncorrected Proof 8

F. F. Pereira et al.

Figure 5

|

|

Hydrology Research

Two-way coupled atmospheric-hydrological system at regional scale

|

in press

|

2013

Simulated precipitation fields at 00:00, 06:00, 12:00 UTC 22 January and 00:00 UTC 23 January for the BRAMS model without any changes (control run), with the two-way coupling (two-way run), and with the water vapor observations from GOES-10. The marked line indicates the Rio Grande basin (see online version for colours: http://www. iwaponline.com/nh/toc.htm).

patterns as BRAMS and satellite images. Moreover, this

Regarding the second study case, results from the

study case also shows that changes proposed by the two-

coupled model and BRAMS are examined using satellite

way coupling strategy do not affect rainfall under stable

image, instantaneous rainfall fields and differences between

atmospheric conditions.

those estimated by BRAMS and the coupled model

Nevertheless, on 22nd January 2009 at 00:00 UTC and

(Figure 6). These differences are calculated in a way that

23rd January 2009 at 00:00 UTC, estimates of maximum

negative values of instantaneous rainfall imply estimates of

instantaneous rainfall obtained from the coupled model

instantaneous rainfall by the coupled model higher than

and BRAMS vary at different rates. While estimates of maxi-

those by BRAMS.

mum values of instantaneous rainfall by BRAMS are higher

As shown in Figure 6, precipitating clouds observed in

than 5 mm h�1, instantaneous rainfall rates from the

the satellite image indicate probable convective rainfalls

coupled model vary between 0 and 4 mm h�1 over the Rio

towards the north-eastern part of the Rio Grande basin.

Grande basin. However, the spatial distribution of instan-

Also, despite both models estimating the amount of instan-

taneous rainfall calculated by the coupled model is well-

taneous rainfall within the range suggested by Leibmann

distributed over the basin and, by comparing to the satellite

et al. (), instantaneous rainfall fields present different

image, it shows a better agreement with precipitating clouds

patterns of spatial distribution over this region. Comparing

than the one proposed by BRAMS.

to the spatial distribution of precipitating clouds in the

144

Q10

Paper V

Uncorrected Proof 9

F. F. Pereira et al.

Figure 6

|

|

Hydrology Research

Two-way coupled atmospheric-hydrological system at regional scale

|

in press

|

2013

Precipitation fields calculated by BRAMS (a), the coupled model (b) and water vapor observations (c). Differences between precipitation fields simulated with BRAMS and the

Q11

coupled model (d) (see online version for colours: http://www.iwaponline.com/nh/toc.htm).

satellite image, the spatial distribution of instantaneous rain-

interactions between land surface and atmosphere consider-

fall from the coupled model reveals a better agreement than

ing their feedback loops. This coupled system aimed to use

the one from BRAMS. According to differences between

the process-based approach to estimate evapotranspiration

instantaneous rainfall fields, zones of probable convective

provided by MGB-IPH instead of evapotranspiration calcu-

rainfalls present values up to �5 mm h�1, which means

lated from land surface parameterizations in BRAMS.

that BRAMS may underestimate convective rainfall events.

Unlike Walko et al. (), results from the coupled

Since convective rainfalls are driven by local surface

system and BRAMS were compared to two atmospheric

heating, which increases evapotranspiration rates, a better

events captured by a geostationary satellite and character-

representation of this study case suggested that a process-

ized by variations of evapotranspiration in the land

based approach to land surface hydrological processes pro-

surface, namely, a cold front passage and a local surface

posed by the two-way coupling strategy improves the

heating. These results led to findings which are discussed

ability of regional atmospheric models to represent atmos-

below.

pheric events governed by the local hydrology.

Cold fronts are derived from sharp gradients of temperature and pressure over short distances; and although two different approaches to land surface processes were used

CONCLUSIONS

to reproduce the cold front passage, modifications arising from the coupled system did not have the same order of

This work presents the implementation of a two-way coup-

magnitude as interactions between land surface and atmos-

ling

Atmospheric

phere during the front passage. Therefore, BRAMS and the

Modeling System (BRAMS) and the Model for Large

coupled model presented similar patterns of spatial distri-

Basins (MGB-IPH) in order to develop an atmospheric-

bution of instantaneous rainfall. Another characteristic

hydrological

found in this atmospheric event was the capability of

between

the

Brazilian

modeling

system

Regional

capable

of

estimating

145

Implementation of a two-way coupled atmospheric-hydrological system for environmental modeling at regional scale

Uncorrected Proof 10

F. F. Pereira et al.

|

Two-way coupled atmospheric-hydrological system at regional scale

Hydrology Research

|

in press

|

2013

representing the transition from unstable to stable atmos-

University, supported by the Crafoord Foundation. The

pheric conditions. Even though the coupled system

authors also acknowledge CPTEC/INPE for providing

incorporates a two-way exchange of variables between

access to the ETA reanalysis and GOES-10 images.

BRAMS and MGB-IPH at a particular time step, this study case revealed that effects of this exchange of variables did not lead to numerical instability; since the coupled system

REFERENCES

and BRAMS converged to the same solution after unstable atmospheric conditions caused by the cold front. On the other hand, the coupled system and BRAMS indicated different areas of convective rainfall induced by the local surface heating. Differences between instantaneous rainfall fields revealed that areas of convective rainfall proposed by the coupled system matched precipitating clouds observed in satellite images rather than those suggested by BRAMS. Since convective rainfalls are responses to warmer temperatures in the land surface, the study case of the local surface heating evidenced that a process-based approach to land surface hydrological processes used by the coupled system increased BRAMS ability to reproduce convection; thereby, the coupled system appears to be an alternative method for issues related to simulating convective events at regional scale (Moncrieff ; Bernardet et al. ; Lang et al. ; Anabor et al. ). Ongoing researches are being conducted on the development of an integrated modeling system capable of assessing the interplay between land use, hydrology, and atmosphere in ecosystems under climatic and land use changes. Therefore, future improvements of the atmospheric-hydrological modeling system will incorporate air temperature, relative humidity, surface pressure, wind speed, and turbulence coefficient as well as solar and long wave radiation as coupling variables provided by BRAMS to MGB-IPH. In addition, the river routing scheme from MGB-IPH may be included in calculations of soil moisture rates by BRAMS.

ACKNOWLEDGEMENTS The development of this work has been supported by the Swedish Research Council (Vetenskaprådet) and the Crafoord Foundation. During the development of this work, Dr. de Moraes held a postdoc position at the Department

146

of

Water

Resources

Engineering,

Lund

Adler, R. F. & Andrew, J. N.  A satellite infrared technique to estimate tropical convective and stratiform rainfall. J. Appl. Meteorol. 27, 30–51. Anabor, V., Stensrud, D. J. & de Moraes, O. L. L.  Simulation of a serial upstream-propagating mesoscale convective system event over southeastern South America using composite initial conditions. Mon. Weather Rev. 137, 2144–2163. ANEEL  Agência Nacional de Energia Elétrica: Brazilian Electric Energy Atlas, 2nd edn. ANEEL, Brasília. Baron, J. S., Hartman, M. D., Kittel, T. G. F., Band, L. E., Ojima, D. S. & Lammers, R. B.  Effects of land cover, water redistribution, and temperature on ecosystem processes in the South Platte Basin. Ecol. Appl. 8, 1037–1051. Bartholmes, J. & Todini, E.  Coupling meteorological and hydrological models for flood forecasting. Hydrol. Earth Syst. Sci. 9, 333–346. Bender, A. & de Freitas, E. D.  Evaluation of BRAMS turbulence schemes during a squall line occurrence in São Paulo, Brazil. Am. J. Environ. Eng. 3, 1–7. Benoit, R., Pellerin, P., Kouwen, N., Ritchie, H., Donaldson, N., Joe, P. & Soulis, E. D.  Toward the use of coupled atmospheric and hydrologic models at regional scale. Mon. Weather Rev. 128, 1681–1706. Bernardet, L. R., Grasso, L. D., Nachamkin, J. E., Finley, C. A. & Cotton, W. R.  Simulating convective events using a high-resolution mesoscale model. J. Geophys. Res. 105, 14963–14982. Betts, R.  Implications of land ecosystem-atmosphere interactions for strategies for climate change adaptation and mitigation. Tellus 59B, 602–615. Beven, K. & Freer, J.  A dynamic TOPMODEL. Hydrolog. Processes 15, 1993–2011. Beven, K., Kirkby, M. J., Schoffield, N. & Tagg, A.  Testing a physically-based flood forecasting model (TOPMODEL) for three UK catchments. J. Hydrol. 69, 119–143. Bremicker, M.  Aufbau eines Wasserhaushaltsmodells für das Weser und das Ostsee Einzugsgebiet als Baustein eines Atmosphären-Hydrologie-Modells. Dissertation Doktorgrad, Geowissenschaftlicher Fakultät der AlbertLudwigsUniversität, Freiburg, Germany. Businger, J. A., Wyngaard, J. C., Izumi, Y. & Bradley, E. F.  Flux-profile relationships in the atmospheric surface layer. J. Atmos. Sci. 28, 181–189. Collischonn, W.  Hydrologic simulation of large basins (in Portuguese). PhD Thesis, Instituto de Pesquisas Hidráulicas, Universidade Federal do Rio Grande do Sul, Porto Alegre, Brazil.

Q3

Paper V

Uncorrected Proof 11

F. F. Pereira et al.

|

Two-way coupled atmospheric-hydrological system at regional scale

Collischonn, W., Allasia, D. G., Silva, B. C. & Tucci, C. E. M.  The MGB-IPH model for large-scale rainfall-runoff modeling. Hydrol. Sci. J. 52, 878–895. Collischonn, B., Collischonn, W. & Tucci, C. E. M.  Daily hydrological modeling in the Amazon basin using TRMM rainfall estimates. J. Hydrol. 360, 207–216. Coppola, E., Grimes, D. I. F. M. & Verdecchia, G. V.  Validation of improved TAMANN neural network for operational satellite-derived rainfall estimation in Africa. J. Appl. Meteorol. Climatol. 45, 1557–1572. Field, C. B., Lobell, D. B., Peters, H. A. & Chiariello, N. R.  Feedbacks of terrestrial ecosystems to climate change. Annu. Rev. Environ. Resour. 32, 1–29. Freitas, S. R., Longo, K. M., Silva Dias, M. A. F., Chatfield, R., Silva Dias, P., Artaxo, P., Andreae, M. O., Grell, G., Rodrigues, L. F., Fazenda, A. & Panetta, J.  The coupled aerosol and tracer transport model to the Brazilian developments on the regional atmospheric modeling system (CATT-BRAMS) – Part 1: Model description and evaluation. Atmos. Chem. Phys. 9, 2843–2861. Graham, L. P., Hagemann, S., Jaun, S. & Beniston, M.  On interpreting hydrological change from regional. Clim. Change 81, 97–122. Haggag, M., Yamashita, T., Lee, H. & Kim, K.  A coupled atmosphere and multi-layer land surface model for improving heavy rainfall simulations. Hydrol. Earth Syst. Sci. 5, 1067–1100. Hay, L. E., Clark, M. P., Pagowaski, M., Leavesley, G. H. & Gutowski Jr, W. J.  One-way coupling of an atmospheric and a hydrologic model in Colorado. J. Hydrometeorol. 7, Q4 569–589. Hay, E. L., Clark, M. P., Wilby, R. L., Gutowski Jr, W. J., Leavesley, G. H., Pan, Z., Arritt, R. W. & Takle, E. S.  Use of regional climate model output for hydrologic simulations. J. Hydrometeorol. 3, 571–590. Högström, U. & Smedman-Högström, A. S.  Turbulence mechanisms at an agricultural site. Boundary Layer Meteorol. 7, 373–389. Jakob, C. & Klein, S. A.  A parameterization of the effects of cloud and precipitation overlap for use in general-circulation models. Q. J. R. Meteorol. Soc. 126, 2525–2544. Kot, S. C. & Song, Y.  An improvement of the Louis scheme for the surface layer in an atmospheric modelling system. Boundary Layer Meteorol. 88, 239–254. Lang, S., Tao, W.-K., Simpson, J., Cifelli, R., Rutledge, S., Olson, W. & Halverson, J.  Improving simulations of convective systems from TRMM LBA: Easterly and westerly regimes. J. Atmos. Sci. 64, 1141–1164. Lee, T. J. & Pielke, R. A.  Estimating the soil surface specific humidity. J. Appl. Meteorol. Climatol. 31, 480–484. Leibmann, B., Jones, C. & de Carvalho, L. M. V.  Interannual variability of daily extreme precipitation events in the state of São Paulo, Brazil. J. Clim. 14, 208–218. Liang, X., Lettenmaier, D. P., Wood, E. F. & Burges, S. J.  A simple hydrologically based model of land surface water

Hydrology Research

|

in press

|

2013

and energy fluxes for general circulation models. J. Geophys. Res. 99, 14415–14428. Lin, C. A., Wen, L., Lu, G., Wu, Z., Zhang, J., Yang, Y., Zhu, Y. & Tong, L.  Atmospheric-hydrological modeling of severe precipitation and floods in the Huaihe River basin, China. J. Hydrol. 330, 249–259. Lintner, B. R., Christopher, E. H. & Neelin, J. D.  Column water vapor statistics and their relationship to deep convection, vertical and horizontal circulation, and moisture structure at Nauru. J. Clim. 24, 5454–5466. Liston, G. E. & Pielke, R. A.  A climate version of the regional atmospheric modeling system. Theor. Appl. Climatol. 66, 29–47. Louis, J. F.  A parametric model of vertical eddy fluxes in the atmosphere. Bound.-Layer Meteorol. 17, 187–202. McMurdie, L. A. & Katsaros, K. B.  Satellite-derived integrated water vapor and rain intensity patterns: Indicators for rapid cyclogenesis. Weather Forecast. 11, 230–245. Messager, C., Galle, H., Brasseur, O., Peugeot, B. C. C., Seguis, L., Vauclin, M., Ramel, R., Grasseau, G. & Leger, L.  Influence of observed and RCM-simulated precipitation on the water discharge over the Sirba basin, Burkina Faso-Niger. Clim. Dyn. 27, 199–214. Mölders, N. & Raabe, A.  Testing the effect of a two-way coupling of a meteorological and a hydrologic model on the predicted local weather. Atmos. Res. 45, 81–107. Moncrieff, M. W.  Mesoscale convection from a large-scale perspective. Atmos. Res. 35, 87–112. Nóbrega, M. T., Collischonn, W., Tucci, C. E. M. & Paz, A. R.  Uncertainty in climate change impacts on water resources in the Rio Grande basin, Brazil. Hydrol. Earth Syst. Sci. 15, 585–595. Onogi, K., Koide, H., Sakamoto, M., Kobayashi, S., Tsutsui, J., Hatsushika, H., Matsumoto, T., Yamazaki, N., Kamahori, H., Takahashi, K., Kato, K., Oyama, R., Ose, T., Kadokura, S. & Wada, K.  JRA-25: Japanese 25-year re-analysis project— progress and status. Q. J. R. Meteorol. Soc. 131, 3259–3268. Osborn, H. B., Renard, K. G. & Simanton, J. R.  Dense networks to measure convective rainfall in the southwestern United States. Water Resour. Res. 15, 1701–1711. O’Sullivan, F., Wash, C. H., Stewart, M. & Motell, C. E.  Rain estimation from infrared and visible GOES satellite data. Technical Report No. 142, October, Washington DC, US. Phelps, G. T. & Pond, S.  Spectra of the temperature and humidity fluctuations and of the fluxes of moisture and sensible heat in the marine boundary layer. J. Atmos. Sci. 28, 918–928. Pielke, R. A. & Arritt, R. W.  A proposal to standardize models. Bull. Am. Meteorol. Soc. 65, 1082. Pielke, R. A., Avissar, R., Raupach, M., Dolman, H., Zeng, X. & Denning, S.  Interactions between the atmosphere and terrestrial ecosystems: Influence on weather and climate. Global Change Biol. 4, 101–115. Porcú, F., Borga, M. & Prodi, F.  Rainfall estimation by combining radar and infrared satellite data for nowcasting purposes. Meteorol. Appl. 6, 289–300.

Q5

Q6

Q7

147

Implementation of a two-way coupled atmospheric-hydrological system for environmental modeling at regional scale

Uncorrected Proof 12

F. F. Pereira et al.

|

Two-way coupled atmospheric-hydrological system at regional scale

Reynolds, R. W., Rayner, N. A., Smith, T. M., Stokes, D. C. & Wang, W.  An improved in situ and satellite SST analysis for climate. J. Clim. 15, 1609–1625. Seuffert, G., Gross, P. & Simmer, C.  The influence of hydrologic modeling on the predicted local weather. J. Hydrometeorol. 3, 505–523. Song, X., Zhao, Y. & Liu, Z.  Cloud detection and analysis of modis image. In: Proceedings of the Geosciences and Remote Sensing Symposium, September 20–24th, Anchorage, US. Todd, M. C., Chris Kidd, D. K. & Tim, J. B.  A combined satellite infrared and passive microwave technique for estimation of small-scale rainfall. J. Atmos. Oceanic Technol. 18, 742–755. Tucci, C. E. M., Collischonn, W., Clarke, R. T., Paz, A. R. & Allasia, D.  Short- and long-term flow forecasting in the Rio Grande watershed (Brazil). Atmos. Sci. Lett. 9, 53–56.

Hydrology Research

in press

|

2013

Umeda, T. & Martien, P. T.  Evaluation of a data assimilation technique for a mesoscale meteorological model used for air quality modeling. J. Appl. Meteorol. 41, 12–29. Van Den Hurk, B. J. J. M. & Holtslag, A. A. M.  On the bulk parameterization of surface fluxes for various conditions and parameter ranges. Bound.-Layer Meteorol. 82, 119–134. Walko, R. L., Band, L. E., Baron, J., Kittel, T. G. F., Lammers, R., Lee, T. J., Ojima, D., Pielke, R. A., Taylor, C., Tague, C., Tremback, C. J. & Vidale, P. L.  Coupled atmosphere– biophysics–hydrology models for environmental modeling. J. Appl. Meteorol. Clim. 39, 931–944. Walko, R. L., Cotton, W. R., Meyers, M. P. & Harrington, J. Y.  New RAMS cloud microphysics parameterization Part I_the single-moment scheme. Atmos. Res. 38, 29–62.

First received 29 August 2012; accepted in revised form 9 July 2013. Available online 23 August 2013

148

|

PAPER VI Assessment of the role of a process-based representation of land surface hydrological processes to the atmosphere F. F. Pereira and C. B. Uvo. Journal of Hydrology Submitted, Under review.

Paper VI

Assessment of the role of a process-based representation of land surface hydrological processes to the atmosphere F. F. Pereira and C. B. Uvo Department of Water Resources Engineering, Lund University, P.O. Box 118, SE-221 00 Lund, Sweden

Abstract This study aims at assess the importance of a conceptual representation of hydrological processes when modelling atmospheric circulation. It compares results from a regional atmospheric model that interprets land surface hydrological processes based on parameterizations with results from a two-way coupled atmosphere-hydrological model which has a process-based approach to the land surface hydrological cycle. These numerical models were applied to a region covering the Rio Grande basin, Brazil. The same input data, initial and boundary conditions were used on a 31-day simulation period. Results obtained from these simulations were compared to visible satellite images and gauging rainfall stations for three case studies that included a cold front, deep convective clouds and stable atmospheric conditions. Both models could reproduce regional patterns of air circulation and rainfall influenced by the orography of the basin. However, atmospheric processes driven by spatial gradients of land surface temperature or local surface heating were spatially better represented by the atmospheric-hydrological modelling system rather than the regional atmospheric model. Since areas characterized by spatial gradients of land surface temperature and local surface heating were closely associated with convergent air flows near land surface and strong vertical motion in the mid troposphere, this finding enhanced the role of a good representation of land surface hydrological processes for a better modelling the atmospheric dynamics. Keywords: Process-based approaches, integrated hydrological modelling, atmosphere

Preprint submitted to Journal of Hydrology

July 15, 2013

151

Assessment of the role of a process-based representation of land surface hydrological processes to the atmosphere

1

2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37

1. Introduction Interactions associated to natural fluxes between biosphere and atmosphere have been highly discussed (Sellers et al., 1997; Pielke et al., 1998; Sutton et al., 2007). Among these natural exchanges, the water cycle stands out for its complexity and relevance to all other physical processes (Stohlgren et al., 1998). The water cycle incorporates a wide range of processes within soil, surface and atmosphere which are closely interconnected to each other. For a better understanding of these processes and how they are interrelated, many alternatives and strategies have been developed to estimate water fluxes between soil, surface and atmosphere at different spatial and temporal scales (Liang et al., 1994; Schaake et al., 1996; Wilson et al., 2001; Pitman, 2003; Balsamo et al., 2009). In this context, numerical models are widely recommended as tools capable of quantifying, predicting and assessing the soil, surface and atmospheric water budgets (Arnold et al., 1993; Maxwell and Miller, 2005; Bittelli et al., 2010). In general, numerical models have successfully been used for estimating the exchange of water within the hydrosphere, and among surface water, soil water and groundwater. A groundwater model, for instance, have recently been applied by Singh (2013) to evaluate the impacts of a waterlogged area on groundwater level and recharge rates. Besides finding that a small increase in the net recharge might implicate an expansion of waterlogged areas, that study also indicates that numerical models are an effective tool for groundwater simulations. Raza et al. (2013) used a numerical model to analyse the influence of three types of land use on soil water dynamics. Results obtained from simulations were compared to field experiments and, revealed that the numerical model could satisfactorily reproduce the water content in the soil profile. Regarding surface water, numerical models have systematically been developed and upgraded (Todini, 1996; Lohmann et al., 1998; Panday and Huyakorn, 2004). Among these models, hydrological models are frequently used for a large range of applications spanning from runoff simulation to flood forecasting (Nijssen et al., 1997; Toth et al., 2000). Although hydrological models are able to physically represent most of the water balance by integrating soil and surface water processes, they do not have any atmospheric module to deal with exchange of water between surface and atmosphere; hence, estimates of runoff depend on how dense the rainfall gauge network is (St-Hilaire et al., 2003) or on the resolution of the atmospheric model used 2

152

Paper VI

38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75

to estimate or forecast precipitation. On the other hand, atmospheric models present a detailed and complex approach of atmospheric processes that includes estimation of carbon, heat, energy and water fluxes between surface and atmosphere based on energy balance. By using a suite of regional climate model (RCM) scenarios, for instance, Morales et al. (2007) estimated climate impacts on carbon cycling across Europe. Their study found that projected changes in carbon balance depended on the choice of the general circulation model (GCM) providing boundary conditions to the RCM rather than the choice of RCM. RCM simulations were also conducted by Seneviratne et al. (2002) in order to investigate thermodynamic processes such as exchange of sensible/latent heat between land surface and atmosphere under a warmer climate. Among many other findings, their results underline the importance of land surface processes in climate integrations. RCMs were also evaluated by Hagemann et al. (2004) on their ability to estimate water budgets over Europe. Their study showed that all RCMs presented either prominent summer drying or overestimation of rainfall throughout the year except during the summer. They claimed that the model deficit and systematic errors may be related to deficiencies in the land surface parametrization. Moreover, parameterization schemes usually apply prescribed values of parameters based on their probability density functions. This assumption does not consider land use and soil characteristics as continuous distributions, and hence, mixtures in soil and vegetation within an area of interest are not captured. This may lead to errors in spatial distribution when estimating land surface processes (Vidale et al., 2003; Molders, 2000). Simultaneously, an alternative approach proposes to optimize the performance of hydrological and atmospheric numerical models by coupling them into integrated modelling systems (Walko et al., 2000; Seuffert et al., 2002). In this context, a two-way coupled atmospheric-hydrological modelling system have been implemented and presented in Pereira et al. (2013a). It proved to be able to estimate the exchange of fluxes between surface and atmosphere considering their feedback loops. In this study, comparisons between this integrated modelling system and a regional atmospheric model have been made in order to evaluate the influence of a process-based representation of land surface hydrological processes on regional atmospheric models. As the atmospheric-hydrological modelling system includes a distributed hydrological model, its domain simulation is restricted to river basins. In 3

153

Assessment of the role of a process-based representation of land surface hydrological processes to the atmosphere

77

this case, the Rio Grande basin, Brazil, was chosen as study area due to convenience in the data acquisition and previous experience with the area.

78

2. Data and Methods

76

79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94

95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110

The methodology used for assessing the role of a better representation of land surface hydrological processes into a regional atmospheric model is described here. This methodology consists of comparisons between air flow, relative humidity, vertical motion, land surface temperature and instantaneous rainfall fields calculated using a regional atmospheric model and an atmospheric-hydrological modelling system. Two short-term runs of 31 days each were performed for a wet period when the atmosphere is dominated by fronts, and local convection. The first run was carried out using the regional atmospheric model whereas the atmospheric-hydrological modelling system was applied in the second run. Based on atmospheric activities detected by the presence/absence of cumulus clouds in visible satellite images, three case studies were selected representing a cold front passage, local strong convections and stable atmospheric conditions. Finally, the influence of a better representation of land surface hydrological processes was individually evaluated by analysing the performance of each model in reproducing each study case. 2.1. Regional Atmospheric Model BRAMS is a regional atmospheric model operationally used at the Brazilian Center for Weather Prediction and Climate Studies (CPTEC). It is derived from a project between the Department of Atmospheric Science at Colorado State University and CPTEC that aimed to develop a version of the Regional Modelling Atmospheric System (RAMS) (Pielke et al., 1992) tailored to the tropics. In order to merge the capabilities of several numerical weather codes, BRAMS was implemented using the concept of ”plugcompatible” modules given by Pielke and Arritt (1984). Not only this concept allows the easy incorporation of improvements between the sub-routines of the model, but also stimulates the use of parameterizations by the developers and users of the model. To represent surface layer fluxes of water vapour into the atmosphere, BRAMS uses a parametric model developed by Louis (1979). Similarly to any other trace gas such as ozone (O3 ) and carbon dioxide (CO2 ), his parameterization scheme estimates the fluxes of water vapour using Busingers 4

154

Paper VI

111 112

113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140

141 142 143 144 145 146

profile functions (Businger et al., 1971). Once computed, the fluxes of water vapour are interpreted by BRAMS as the lower boundary for the atmosphere. 2.2. Atmospheric-Hydrological Modelling System The atmospheric-hydrological modelling system is composed of a hydrological model two-way coupled to a regional atmospheric model. In this coupled modelling system, the hydrological Model for Large Basins developed at the Institute for Hydraulics Research (MGB-IPH, Collischonn (2001)) and the atmospheric Brazilian Regional Modelling System (BRAMS, Freitas et al. (2009)) were employed. Besides the experience of the authors with them, the choice of these models was also based on their successful use in previous studies in the Rio Grande Basin (Nbrega et al., 2011; Bender and de Freitas, 2013). The MGB-IPH is a large scale distributed hydrological model conceptually based on the LARSIM (Bremicker, 1998) and VIC-2L (Liang et al., 1994) models. It consists of modules for calculating soil water budget, evapotranspiration estimation, surface and subsurface flow generation, which are interconnected by river routing. MGB-IPH does not present any atmospheric module so that the fluxes of water vapour between land surface and atmosphere are incorporated by means of evapotranspiration rates. The evapotranspiration rates are estimated using average monthly values of air temperature, sunshine (hours), relative humidity, wind speed and atmospheric pressure. In this modelling system, BRAMS and MGB-IPH were coupled in a way to best exploit their respective strengths. Here, process-based estimates of the fluxes of water vapour by MGB-IPH replaced those estimated by BRAMS using parameterization schemes while daily rainfall calculated by BRAMS was provided as input to MGB-IPH. A brief description of the coupling strategy employed in this atmospheric-hydrological modelling system is given in the following section. For a detailed description of it, the reader is referred to Pereira et al. (2013a). 2.2.1. Brief Description of the Coupling Strategy BRAMS and MGB-IPH were spatially and temporally coupled according to a two-way coupling strategy described in detail by Pereira et al. (2013a). Their coupling strategy avoids upscaling and downscaling issues by running both models on the same grid size. Since BRAMS and MGB-IPH output variables are calculated at their own computational cell centers that not 5

155

Assessment of the role of a process-based representation of land surface hydrological processes to the atmosphere

147 148 149 150 151 152 153 154 155 156

157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183

always match each other, the coupling strategy counted on an algorithm to identify the nearest computational cell for the exchange of variables between BRAMS and MGB-IPH. Due to numerical stability constraints, BRAMS runs with a smaller time step than the daily time step often used by MGB-IPH at basin scale. Therefore, a temporal coupling between BRAMS and MGB-IPH is also included in the coupling strategy, so that MGB-IPH is employed as a subroutine of BRAMS which is called every 24 simulation-hours. Additional details such as how and which variables are exchanged between the models, assumptions and coupling limitations can be found at Pereira et al. (2013a). 2.2.2. Input Data As the atmospheric-hydrological coupled modelling system is composed by BRAMS and MGB-IPH models, input data for both models were needed. For MGB-IPH, input data vary according to whether running in simulation or calibration mode. As in the integrated modelling system MGB-IPH is always ran in simulation mode, its input data include land use, soil and elevation maps together with air temperature, sunshine, relative humidity, wind speed and atmospheric pressure time series. Similarly to MGB-IPH, the BRAMS input data includes topographical and land use features. Its input data contain normalized difference vegetation indexes (NDVI), soil moisture distribution, sea surface temperature, soil texture classes, land use and elevation maps. Besides input data, BRAMS requires time-dependent boundary conditions which are derived from general circulation models. For the Rio Grande basin, input data and boundary meteorological observations were collected from different sources, from online databases to personal communication. The whole data collection procedure is described as follows. A digital elevation model (DEM) was freely obtained at Department of Ecology of the Federal University of Rio Grande do Sul. Their DEM preprocessing includes data gap filling and mosaicking of SRTM database, not only, for the Rio Grande basin but also for all Brazilian territory (Hasenack et al., 2010). The soil map of the Rio Grande basin was derived from a soil survey data created by RADAM Brasil project (RADAMBRASIL, 1982) at scale of 1:1000000. Although at a coarser scale than RADAM Brasil soil survey data, digitalized soil maps (1:3000000) from the Food and Agriculture Organization of the United Nations (Food and of the United Nations, 1974) were resampled 6

156

Paper VI

192

and used to overcome missing data. The RADAM Brasil database that includes over 12 different types of soils in the Rio Grande basin, (FAURGS, 2007) was reclassified into two groups as deep and shallow soils according to their hydrological behavior and provided to be used in this work (A. R. Paz, personal communication, 2012). Meteorological data sets from three stations were provided by CPTEC. They comprise monthly time series of air temperature, sunshine, relative humidity, wind speed and atmospheric pressure. The location of these stations may be seen in figure 1.

193

Figure 1

184 185 186 187 188 189 190 191

194 195 196

197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215

216 217 218

Finally, normalized difference vegetation indexes (NDVI) provided by MODIS as well as soil moisture distributions and weekly sea surface temperature were downloaded from http://www.cptec.inpe.br/brams. 2.2.3. Initial and Boundary Conditions Initial and boundary conditions were separately set for BRAMS and MGB-IPH models. In the BRAMS, a four-dimensional data assimilation technique described by Umeda and Martien (2002) was used to interpret atmospheric boundary conditions provided every 6 hours by global atmospheric analyses. In addition, initial soil moisture conditions were calculated using global analyses of monthly precipitation derived from satellite and surface measurements. For doing so, a hybrid model (Gevaerd and Freitas, 2006) which estimates soil moisture based on the Global Precipitation Climatology Program (GPCP) and the Tropical Rainfall Measuring Mission (TRMM) has been used. Moreover, soil temperature was considered as vertically homogeneous and its value was set equal to the value of air temperature near land surface. In the MGB-IPH, initial conditions were set by changing interception and infiltration stores. A initial hot start for interception and infiltration stores was adopted according to simulations performed by Pereira et al. (2013b). This hot start is reasonable since a warming-up period has been made in the beginning of each run and, therefore, errors due to this assumption were gradually attenuated. 2.2.4. Rainfall Gauge Network and Satellite Images Rainfall depths were selected from the Agncia Nacional de guas (ANA) database (http://hidroweb.ana.gov.br/). Daily precipitation values were ob7

157

Assessment of the role of a process-based representation of land surface hydrological processes to the atmosphere

237

tained from observations on 483 precipitation stations over the Rio Grande basin and its surroundings. Out of these 483, 136 stations had rainfall records for January 2009, the selected wet period for the model runs. The position of these stations is indicated in Figure 1. Visible satellite images were used to analyse the presence/absence of cumulus clouds in the upper and middle levels of the atmosphere. These images were captured by two different geostationary satellites that provide coverage for South America, the Geostationary Operational Environmental Satellite10 (GOES-10) and the Meteosat-9. The first one was operated by the the National Oceanic and Atmospheric Administration (NOAA) and it was used to support hurricane forecasting efforts in South America whereas Meteosat9 was designed by European Organisation for the Exploitation of Meteorological Satellites (EUMETSAT) to support weather forecasting in Europe and Africa. Despite the coverage area of Meteosat-9 partially includes South America, GOES-10 visible images are preferably used for analyses. When not available, the preference criterion for satellite images considers Meteosat-9 visible images, followed by false-color GOES-10 and Meteosat-9 images. In this study, all the visible and false-color images were freely downloaded from the CPTEC online database (http://satelite.cptec.inpe.br).

238

3. Short-Term Runs

219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236

239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254

Two different runs were performed for a simulation period of 31 days. The first run corresponds to the control run and, only BRAMS was applied to the Rio Grande basin while the second run was carried out using the atmospheric-hydrological modelling system. One of the aspects evaluated by this study was the capability of reproducing rainfall events by better representing land surface hydrological processes. In order to assess the formation of rainfalls, a wet weather period was chosen due to its high probability of rainfall occurrence. Austral summer is the rainy season at the Rio Grande basin, therefore the chosen simulation period spans 1st January trough 31st January 2009. Prior to each run, a warning-up period of 10 days was considered for the initialization of the physical variables (Benoit et al., 2000). Thus, all comparisons presented in this study refer to results obtained from the last 20 days of simulation, which means, between 11th to 31st January 2009. In general, atmospheric models provide outputs every 6 hours (Baron et al., 1998; Onogi et al., 2005). Depending upon the purpose of the study, 8

158

Paper VI

260

this value may become smaller or larger. Here, as the influence of land surface hydrological processes on the atmosphere was carefully investigated for atmospheric processes that occur at different time scales, the interval between outputs was assumed equal to 3 hours for both runs. Since satellite images were captured every 3 hours, this is the shortest interval which could be adopted for comparisons with observed data.

261

4. Criteria for Selection of Case Studies

255 256 257 258 259

276

Outputs from both the atmospheric-hydrological modelling system and the regional atmospheric model were compared to each other and analysed towards the formation of clouds and rainfall derived from convergence at the surface layer. In this study, significant atmospheric activities were selected according to the following procedure. Firstly, convergences at the surface layer (1000 hPa) were identified by plotting zonal and meridional wind fields. Convergence was used as indicator of atmospheric activities in the upper layers. Secondly, vertical motions higher than 1 m/s at 500 hPa over the convergence zones pointed to possible convection development. Finally, convection occurrence was validated by the visible satellite images. According to this procedure, three convection developments were observed on 21 January at 2100 UTC, 22 January at 1800 UTC and 25 January at 2100 UTC from model results. All these three events are described and discussed in section 5.

277

5. Results and Discussion

262 263 264 265 266 267 268 269 270 271 272 273 274 275

278 279 280 281

282 283 284 285 286 287

Here, results from the short-term simulations performed by the atmospherichydrological model system are compared to outputs from the regional atmospheric model, satellite images at visible spectrum and daily rainfall data for the specific cases identified in section 4. 5.1. Case Studies 5.1.1. 21 January at 2100 UTC On January 21st , a false-color Meteosat-9 image captured at 2100 UTC shows a cold front passing over the Rio Grande basin (Fig. 2a). In the image, the front is detected by the presence of brighter cumulus clouds. Since brightness of clouds in visible satellite images is related to their top heights 9

159

Assessment of the role of a process-based representation of land surface hydrological processes to the atmosphere

309

(Song et al., 2004), figure 2a is an indicator of vertical motion upwards which may have caused high cloud tops and increased their reflectance in the image. Daily observed rainfall over the basin ranged from 0 to 80 mm. Although the satellite image indicates developing and mature cumulus clouds all over the basin due to the passage of this front, the highest values of rainfall depth were observed mostly concentrated in the eastern part of the basin as shown in figure 2b. This spatial rainfall distribution is primarily associated with the orography of the Rio Grande basin. As the satellite image and observed daily rainfalls have different time scale, they were compared to results obtained from the short-term runs in a way to enhance the performance of each model in reproducing spatial alignment and distribution of the front rather than estimating rainfall intensity values. The spatial alignment and distribution of fronts are generally characterized by shifts in wind directions, and sharp temperature and air moisture content changes over short distances (Ahrens, 2007). Therefore, temperature, zonal and meridional wind fields calculated by the integrated system and regional atmospheric model were compared to the spatial alignment and distribution of the front observed in the visible satellite image (Fig. 2 and 3). Moreover, figures 2b and 2c show computations of vertical motion at 500 hPa (shaded) using the regional atmospheric model and integrated system, respectively.

310

Figures 2 and 3

288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308

311 312 313 314 315 316 317 318 319 320 321 322 323

As presented in figures 2b and 2c, zonal and meridional wind fields estimated by both models indicated convergent air flows over the Rio Grande basin. However, the convergence zone is placed by the coupled system in better accordance to the shape of the front observed in the satellite image than by the regional atmospheric model. According to Jacobson (2005), convergence zones derived from fronts related to abrupt variations of land surface temperature. In this context, the process-based representation of land surface hydrological processes proposed by the integrated system yielded a good agreement between spatial gradients of land surface temperature and spatial alignment of the front (Fig. 3b). In contrast, spatial distribution of abrupt variations of land surface temperature estimated by the regional atmospheric model presented neither the same pattern as the coupled system nor the spatial alignment of the front (Fig. 3a). 10

160

Paper VI

341

Differences in spatial distribution of land surface temperature between the atmospheric model alone and the coupled system resulted also in different placement of areas of deep convection. Zones of deep convection were identified by upward vertical motion higher than 0.7 m.s−1 at 500 hPa (Jorgensen and Lemone, 1989) and are presented as shaded areas in figures 2b and 2c. Comparing to the satellite image, Figure 2c shows that despite zones of deep convection by the coupled system did not fully match the alignment of the front, they were placed in accordance with the occurrence of brighter cumulus clouds in the satellite image. On the other hand, zones of deep convection indicated by the regional atmospheric model alone were neither associated with the alignment of the front nor with any cumulus cloud observed in the satellite image. Although estimates of rainfall by regional atmospheric models are yet not accurate enough for realistic representation of rainfall intensity (Kendon et al., 2012), observed daily rainfall data were compared to 24h-accumulated rainfall in order to capture the spatial rainfall distribution. Figure 4 presents observed daily rainfall (4a) and 24h-accumulated rainfall estimated by the regional atmospheric model (Fig. 4b) and the coupled system (Fig. 4c).

342

Figure 4

324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340

343 344 345 346 347 348 349 350 351

352 353 354 355 356 357 358

According to figure 4, the same spatial patterns of rainfall were identified for the regional atmospheric model and integrated system. Moreover, both of them presented the same range of values, from 0 to 80 mm.d−1 , and similar location of wet and dry regions in the Rio Grande basin. Except for some local rainfall events in the dry regions, most of the rainfall gauging stations with values larger than 70 mm.d−1 were located in the wet region pointed by the models. It means that, in terms of regional hydrological behaviour of rainfall, the models had a good agreement with the daily rainfall data when representing the front. 5.1.2. 22 January at 1800 UTC Even though cumulus clouds have been observed in the GOES-10 visible image captured in the early evening on January 22nd , they were only used as support to the presence of convective activities in the atmosphere. Therefore, one cloudy region containing developing and mature cumulus clouds was selected within the Rio Grande basin and is shown in figure 5a. This cloudy region was identified as an area of deep convection where rainfall is possible 11

161

Assessment of the role of a process-based representation of land surface hydrological processes to the atmosphere

365

or probable (hereafter referred to as rainfall potential area). Despite many other cumulus clouds may be found in the satellite image, their reflectance indicated that they have lower cloud-top heights. Although convective activities occur at short-time scale to be compared to observed daily rainfall, comparisons between the rainfall potential area and daily rainfall observations (Fig. 5b) were made with respect to extreme values and/or presence of outliers, which indicate convective rainfall events.

366

Figure 5

359 360 361 362 363 364

374

Figure 5 shows that daily rainfall observations presented values larger than 70 mm at two rain gauge stations (highlighted in Fig. 5b) over the rainfall potential area. These indicators were used as evidence of a strong convection in this area. Moreover, the performance of each model in representing the rainfall potential area were evaluated by comparing its spatial distribution to relative humidity, land surface temperature, vertical motion, instantaneous rainfall, zonal and meridional wind fields estimated by the models (Fig. 6 and 7).

375

Figures 6 and 7

367 368 369 370 371 372 373

376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392

Figure 6 reveals a close relationship between spatial gradients of land surface temperature and convergence at low level. According to results from both models, a combination of these two indicators of convection were presented over two regions within the Rio Grande basin (rectangles in figures 6a and 6b). The larger one corresponds to the same region as the rainfall potential area whereas the smaller one represents a local convection near the border of the basin. For the region characterized by the local convection, both models estimated values of vertical motion at 500 hPa higher than 0.7 m.s−1 . However, neither mature nor developing cumulus clouds were observed in the satellite image (Fig. 7). As values of relative humidity calculated by the models were lower than 70% in the region of this local convection (Fig. 7a and 7b), the absence of mature and developing cumulus clouds in the satellite image indicates that the amount of water vapour in the air was a limiting factor for the formation of cumulus clouds. On the other hand, in the first region defined as the rainfall potential area, values of relative humidity reached 100%; which, together with vertical 12

162

Paper VI

393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430

motion upwards, may have induced the formation of the mature and developing cumulus clouds observed in the satellite image and highlighted in figure 5a. The ability of each model to simulate convective rainfall events was investigated using instantaneous rainfall fields. Convective rainfall events were identified over the Rio Grande basin and are shown in figures 7c and 7d (shaded). These convective rainfalls were defined as instantaneous rainfall rates higher than 3.5 mm.h−1 according to studies on the variability of rainfall events in the State of So Paulo carried out by Leibmann et al. (2001). As shown in figures 7c and 7d, two main convective rainfall events were detected using instantaneous rainfall fields estimated by the models. The first convective rainfall event was characterized by the orography of the basin whereas the second convective rainfall event, located towards the middle of the basin, was derived from spatial gradients of land surface temperature. According to results obtained from the simulations, both models estimated the same values of relative humidity, instantaneous rainfall and vertical motion at 500 hPa for the two convective rainfall events. Moreover, values of vertical motion at 500 hPa revealed that the second convective rainfall event, exclusively induced by land surface processes, reached the upper layers of the troposphere unlike the first convective event. This explains the matching of the second convective rainfall event and the presence of cumulus clouds in the satellite image. In terms of spatial distribution and intensity of convective rainfall events within the basin, results obtained from both models showed a good agreement. In this study case, relative humidity, temperature, instantaneous rainfall, zonal and meridional wind fields estimated near land surface by the integrated system and regional atmospheric model showed roughly the same spatial distribution, thereby differences on how to represent land surface hydrological processes were not as significant as they were for the study case of the cold front. A possible reason for that is the different spatial scale between the front in the previous study case and the convection of this case. While the front covered much of the Rio Grande basin, convection only represented a small portion of it. Besides, the study case of the front presented variations of surface temperature greater than the convection, which implied to higher contributions from land surface processes to the atmosphere. For both models, zonal and meridional wind fields could capture convergent air flows derived from the orography of the basin and spatial gradients of land surface temperature. Although very few differences were noted be13

163

Assessment of the role of a process-based representation of land surface hydrological processes to the atmosphere

431 432 433 434 435 436

tween results obtained from the integrated system and regional atmospheric model, this study case showed that relative humidity, temperature, zonal and meridional wind calculated near land surface were important factors for the formation of cumulus clouds and areas where convective rainfalls are possible or probable. These results agree with what Pielke (2001) has proposed as triggering mechanisms of convection.

445

5.1.3. 25 January at 2100 UTC This study case is characterized by the absence of brighter clouds in the GOES-10 visible image captured at 25th January 2100 UTC. Among many other factors, cloud brightness is also associated with cloud thickness and cloud water content (Song et al., 2004; Love et al., 2001). Based on this concept, neither local convective rainfall events nor storms were observed in the visible image (Fig. 8a). Since observed values of daily rainfall were lower than 20 mm.d−1 over the whole basin (Fig. 8b), they corroborated with the low rainfall profile indicated by the satellite image.

446

Figure 8

437 438 439 440 441 442 443 444

452

This study case evaluates the capability of the models in representing atmospheric processes under stable atmospheric conditions. In this sense, temperature, zonal and meridional wind fields computed near land surface were investigated as indicators of atmospheric activities. In addition, vertical motion at 500 hPa was used to evidence the effects of possible atmospheric activities pointed by the models (Fig. 9).

453

Figure 9

447 448 449 450 451

454 455 456 457 458 459 460 461 462 463

According to figures 9a and 9c, results from the regional atmospheric model revealed a strong convection in the centre of the Rio Grande basin characterized by convergent air flow and abrupt variations of temperature over short distances. Since values of vertical motion were larger than 1 m.s−1 over this area, it has been interpreted by the regional model as a trigger mechanism to generate a convective rainfall event. On the other hand, evidences of atmospheric activity (as defined in section 4) were not detected based on temperature, vertical motion, zonal and meridional wind fields calculated by the integrated system (figures 9b and 9d). Although values of temperature near land surface varied in the same 14

164

Paper VI

469

range as the ones estimated by the regional atmospheric model, spatial gradients of temperature by the coupled system were lower than those from the regional model. Moreover, the integrated system suggested convergent flow in many areas across the basin. However, in accordance with the satellite image, values of vertical motion at 500 hPa below 0.5 m.s−1 did not indicate strong convective activity.

470

6. Summary and Conclusions

464 465 466 467 468

471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499

This study outlines the role of a better representation of land surface hydrological processes in regional atmospheric models. For doing so, comparisons were made between a regional atmospheric model and a two-way coupled atmospheric-hydrological modelling system. The major difference between these models is that the atmospheric-hydrological coupled system incorporates a distributed hydrological model for representing land surface hydrologic processes while the regional atmospheric model interprets hydrologic processes based on land surface parameterizations (Avissar and Pielke, 1989; Lee and Abriola, 1999). In order to assess the performance of each model in representing atmospheric processes, simulations were performed and three study cases were selected to evaluate how the models behave in three distinct situations: reproducing a cold front, representing formation of deep convection and at stable atmospheric conditions. Results obtained from these simulations led to findings which have been separately described for each study case. For the cold front, convergent flows and high gradients of land surface temperature estimated by the two-way coupled atmospheric-hydrological modelling system well represented the spatial alignment of the front observed in the visible image. Although the regional atmospheric model alone have also detected the front, convergent flows and gradients of land surface temperature did not match the spatial distribution of the front observed in the satellite image. However, in terms of daily rainfall, comparisons between simulated values of 24-hour accumulated rainfall and daily rainfall data revealed that both models regionally captured wet and dry regions over the Rio Grande basin. Concerning the study case composed of convective activities, high cloudtop observed in the satellite image, associated either with surface heating or orography of the basin, were represented by convergent flows and high instantaneous rainfall rates by both models. In this study case, both ap15

165

Assessment of the role of a process-based representation of land surface hydrological processes to the atmosphere

528

proaches presented similar patterns of land surface temperature, zonal and meridional wind over the Rio Grande basin, thereby rainfall potential areas estimated by the models were similar to each other and, were located at the areas suggested by the satellite image and gauging rainfall stations. Under stable atmospheric conditions, estimates of land surface temperature and air circulation by the regional atmospheric model indicated convergent flows and abrupt variations of temperature which resulted in convective unstable zones in the mid troposphere. On the other hand, despite air circulation estimated by the atmospheric-hydrological modelling system having presented convergent flows over the basin, high values of vertical motion in the mid troposphere were not associated to them. Moreover, in accordance to the satellite image, the integrated system did not present any strong convective activity over the basin. Overall, the replacement of land surface parameterizations by processbased hydrological modelling implied improvement in temperature, air water content, zonal and meridional winds calculated near land surface. According to investigations made towards the implications of these changes on the atmosphere, convergent flows and abrupt variations of temperature calculated by the coupled system were better related to the spatial alignment and distribution of the front than the regional atmospheric model. Under stable atmospheric conditions, results from the regional atmospheric model revealed a non-existent development of convection in response to a local surface heating. As the positioning of the front and development of convection are highly associated with temperature gradients, it indicates that the coupled system provides a better support to weather forecasting when simulating atmospheric processes driven by local surface heating or sharp temperature gradients. The authors are grateful to Crafoord Foundation and the Swedish Research Council (Vetenskapr˚ adet) for supporting the development of this work.

529

References

500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527

530 531

532 533

Ahrens, C.D., 2007. Meteorology Today: An Introduction to Weather, Climate, and the Environment. Eighth ed., Thomson Brooks / Cole. Arnold, J.G., Allen, P.M., Bernhardt, G., 1993. A comprehensive surfacegroundwater flow model. Journal of Hydrology 142, 47–69.

16

166

Paper VI

534 535 536

537 538 539 540

541 542 543 544

545 546 547

548 549 550

551 552 553 554

555 556 557 558

559 560 561

562 563

Avissar, R., Pielke, R.A., 1989. A parameterization of heterogeneous land surfaces for atmospheric numerical models and its impact on regional meteorology. Monthly weather review 114, 27519–27526. Balsamo, G., Viterbo, P., Beljaars, A., Hurk, B.D., Hirschi, M., Betts, A.K., Scipal, K., 2009. A revised hydrology for the ecmwf model: verification from field site to terrestrial water storage and impact in the integrated forecast system. Journal of Hydrometeorology 10, 623–643. Baron, J.S., Hartman, M.D., Kittel, T.G.F., Band, L.E., Ojima, D.S., Lammers, R.B., 1998. Effects of land cover, water redistribution, and temperature on ecossystems processes in the south platte basin. Ecological Applications 8, 1037–1051. Bender, A., de Freitas, E.D., 2013. Evaluation of brams turbulence schemes during a squall line occurrence in so paulo, brazil. American Journal of Environmental Engineering 3, 1–7. Benoit, R., Pellerin, P., Kouwen, N., Ritchie, H., Donaldson, N., Joe, P., Soulis, E.D., 2000. Toward the use of coupled atmospheric and hydrologic models at regional scale. Monthly Weather Review 128, 1681–1706. Bittelli, M., Tomei, F., Pistocchi, A., Flury, M., Boll, J., Brooks, E.S., Antolini, G., 2010. Development and testing of a physically based, threedimensional model of surface and subsurface hydrology. Advances in Water Resources 33, 106–122. Bremicker, M., 1998. Aufbau eines Wasserhaushaltsmodells fr das Weser und das Ostsee Einzugsgebiet als Baustein eines Atmosphren-HydrologieModells. Ph.D. thesis. Geowissenschaftlicher Fakultt der AlbertLudwigsUniversitt. Freiburg, Germany. Businger, J.A., Wyngaard, J.C., Izumi, Y., Bradley, E.F., 1971. Flux-profile relationships in the atmospheric surface layer. Journal of the Atmospheric Sciences 28, 181–189. Collischonn, W., 2001. Hydrologic simulation of large basins (in Portuguese). Ph.D. thesis. Instituto de Pesquisas Hidrulicas. Porto Alegre, BR.

17

167

Assessment of the role of a process-based representation of land surface hydrological processes to the atmosphere

564 565 566

567 568

569 570 571 572 573 574

575 576 577

578 579 580 581

582 583 584

585 586

587 588

589 590 591

592 593 594

FAURGS, 2007. Previso de afluncia a reservatrios hidreltricos mdulo 1. Projeto FAURGS/FINEP 40.04.0094.00. Technical Report. FINEP. Porto Alegre, BR. Food, of the United Nations, A.O., 1974. Soil map of the world. Technical Report. UNESCO. Paris. Freitas, S.R., Longo, K.M., Silva Dias, M.A.F., Chatfield, R., Silva Dias, P., Artaxo, P., Andreae, M.O., Grell, G., Rodrigues, L.F., Fazenda, A., , Panetta, J., 2009. The coupled aerosol and tracer transport model to the brazilian developments on the regional atmospheric modeling system (cattbrams) part 1: Model description and evaluation. Atmospheric Chemistry and Physics 9, 2843–2861. Gevaerd, R., Freitas, S., 2006. Estimativa operacional da umidade do solo para iniciao de modelos de previso numrica da atmosfera. parte i: descrio de maetodologia e validao. Revista Brasileira de Meteorologia 21, 1–15. Hagemann, S., Machenhauer, B., Jones, R., Christensen, O.B., Deque, M., Jacob, D., Vidale, P.L., 2004. Evaluation of water and energy budgets in regional climate models applied over europe. Climate Dynamics 23, 547–567. Hasenack, H., Weber, E., Fernandes, S.M.C., Jonas, R.J., Duarte, E.H.G., 2010. Processamento e organizao de modelos digitais de elevao contnuos para os pases de lngua portuguesa a partir do SRTM. Editora da UFRGS. Jacobson, A.Z., 2005. Fundamentals of Atmospheric Modeling. Cambridge University Press. Jorgensen, D.P., Lemone, M.A., 1989. Vertical veolocity characteristics of oceanic convection. Journal of the Atmospheric Sciences 46, 621–640. Kendon, E.J., Roberts, N.M., Senior, C.A., Roberts, M.J., 2012. Realism of rainfall in a very high-resolution regional climate model. Journal of Climate 25, 5791–5806. Lee, D.H., Abriola, L.M., 1999. Use of the richards equation in land surface parameterizations. JOURNAL OF GEOPHYSICAL RESEARCH 104, 27519–27526. 18

168

Paper VI

595 596 597

598 599 600

601 602 603

604 605

606 607 608

609 610

611 612

613 614 615

616 617 618

619 620 621

622 623 624 625 626

Leibmann, B., Jones, C., de Carvalho, L.M.V., 2001. Interannual variability of daily extreme precipitation events in the state of sao paulo, brazil. Journal of Climate 14, 208–218. Liang, X., Lettenmaier, D.P., Wood, E.F., Burges, S.J., 1994. A simple hydrologically based model of land surface water and energy fluxes for general circulation models. Journal Geophysical Research 99, 14415–14428. Lohmann, D., Raschke, E., Nijssen, B., Lettenmaier, D.P., 1998. Regional scale hydrology: i. formulation of the vic-2l model coupled to a routing model. Hydrological Sciences Journal 43, 131–141. Louis, J.F., 1979. A parametric model of vertical eddy fluxes in the atmosphere. Boundary layer meteorology 17, 187–202. Love, S.P., Davis, A.B., Ho, C., Rohde, C.A., 2001. Remote sensing of cloud thickness and liquid water content with wide−angle imaging lidar. Atmospheric Research , 295–312. Maxwell, R.M., Miller, N.L., 2005. Development of a coupled land surface and groundwater model. Journal of Hydrometeorology 6, 233–247. Molders, N., 2000. On the uncertainty in mesoscale modeling caused by surface parameters. Metereology and Atmosphere Physics 76, 119–141. Morales, P., Hickler, T., Rowell, D.P., Smith, B., Sykes, M.T., 2007. Changes in european ecosystem productivity and carbon balance driven by regional climate model output. Global Change Biology 13, 108–122. Nijssen, B., Lettenmaier, D.P., Liang, X., Wetzel, S.W., Wood, E.F., 1997. Streamflow simulation for continental-scale river basins. Water Resources Research 33, 711–724. Nbrega, M.T., Collishonn, W., Tucci, C.E.M., Paz, A.R., 2011. Uncertainty in climate change impacts on water resources in the rio grande basin, brazil. Hydrology and Earth System Sciences 15, 585–595. Onogi, K., Koide, H., Sakamoto, M., Kobayashi, S., Tsutsui, J., Hatsushika, H., Matsumoto, T., Yamazaki, N., Kamahori, H., Takahashi, K., Kato, K., Oyama, R., Ose, T., Kadokura, S., Wada, K., 2005. Jra-25: japanese 25-year re-analysis projectprogress and status. Quarterly Journal of the Royal Meteorological Society 131, 3259–3268. 19

169

Assessment of the role of a process-based representation of land surface hydrological processes to the atmosphere

627 628 629

630 631 632

633 634

635 636

637 638

639 640 641 642

643 644 645 646

647 648 649

650 651 652

653 654 655 656

Panday, S., Huyakorn, P.S., 2004. A fully coupled physically-based spatiallydistributed model for evaluating surface/subsurface flow. Advances in Water Resources 27, 361–382. Pereira, F.F., Moraes, M.A.E., Uvo, C.B., 2013a. Implementation of a twoway coupled atmospheric-hydrological system for environmental modeling at regional scale. Pereira, F.F., Tursunov, M., Bertacchi, C., 2013b. Towards the response of water balance to sugarcane expansion in the rio grande basin, brazil. Pielke, R.A., 2001. Influence of the spatial distribution of vegetation and soils on the cumulus convective rainfall. Reviews of Geophysics 39, 151–177. Pielke, R.A., Arritt, R.W., 1984. A proposal to standardize models. BULLETIN OF THE AMERICAN METEOROLOGICAL SocIETY 65. Pielke, R.A., Avissar, R., Raupach, M., Dolman, A.J., Zeng, X., Denning, A.S., 1998. Interactions between the atmosphere and terrestrial ecosystems: influence on weather and climate. Global change biology 4, 461–475. Pielke, R.A., Cotton, W.R., Walko, R.L., Trembaek, C.J., Lyons, W.A., Grasso, L.D., Nieholls, M.E., Moran, M.D., Wesley, D.A., Lee, T.J., Copeland, J.H., 1992. A comprehensive meteorological modeling system rams. Meteorology and atmospheric physics 49, 69–91. Pitman, A.J., 2003. The evolution of, and revolution in, land surface schemes designed for climate models. International Journal of Climatology 23, 479– 510. RADAMBRASIL, 1982. Programa de Integrao Nacional, Levantamento de Recursos Naturais (Folhas SE-23 (Belo Horizonte) e SF-22 (Paranapanema)). Technical Report. Ministrio das Minas e Energia. Braslia, BR. Raza, A., Friedel, J.K., M, A., Ardakani, M.R., Loiskandl, W., Himmelbauer, M., Bodner, G., 2013. Modeling growth of different lucerne cultivars and their effect on soil water dynamics. Agricultural Water Management 119, 100–110.

20

170

Paper VI

657 658 659

660 661 662 663 664

665 666 667

668 669 670 671

672 673

674 675

676 677 678 679

680 681 682 683

684 685 686 687 688

Schaake, J.C., Koren, V.I., Duan, Q.Y., Mitchell, K., Chen, F., 1996. Simple water balance model for estimating runoff at different spatial and temporal scales. Journal of Geophysical Research 101, 7461–7475. Sellers, P.J., Dickinson, R.E., Randall, D.A., Betts, A.K., Hall, F.G., Berry, J.A., Collatz, G.J., Denning, A.S., Mooney, H.A., Nobre, C.A., Sato, N., Field, C.B., Henderson-Sellers, A., 1997. Modeling the exchanges of energy, water, and carbon between continents and the atmosphere. Science 275, 502–509. Seneviratne, S.I., Pal, J.S., Eltahir, E.A.B., Schar, C., 2002. Summer dryness in a warmer climate: a process study with a regional climate model. Climate Dynamics 20, 69–85. Seuffert, G., Gross, P., Simmer, C., Wood, E.F., 2002. The influence of hydrologic modeling on the predicted local weather: two-way coupling of a mesoscale weather prediction model and a land surface hydrologic model. Journal of Hydrometeorology 3, 505–523. Singh, A., 2013. Groundwater modelling for the assessment of water management alternatives. Journal of Hydrology 481, 220–229. Song, X., Zhao, Y., Liu, Z., 2004. Cloud detection and analysis of modis image, in: Proceedings of the Geoscience and Remote Sensing Symposium. St-Hilaire, A., Ouarda, T.B.M.J., Lachance, M., Bobe, B., Gaudet, J., Gignac, C., 2003. Assessment of the impact of meteorological network density on the estimation of basin precipitation and runoff: a case study. Hydrological Processes 17, 3561–3580. Stohlgren, T.J., Chase, T.N., Pielke, R.A., Kittel, T.G.F., Baron, J.S., 1998. Evidence that local land use practices influence regional climate, vegetation, and stream flow patterns in adjacent natural areas. Global change biology 4, 495–504. Sutton, M.A., Nemitz, E., Erisman, J.W., Beier, C., Bahl, K.B., Cellier, P., de Vries, W., Cotrufo, F., Skiba, U., di Marco, C., Jones, S., Laville, P., Soussana, J.F., Loubet, B., Twigg, M., Famulari, D., Whitehead, J., Gallagher, M.W., Neftel, A., Flechard, C.R., Herrmann, B., Calanca, P.L., Schjoerring, J.K., Daemmgen, U., Horvath, L., Tang, Y.S., Emmett, B.A., 21

171

Assessment of the role of a process-based representation of land surface hydrological processes to the atmosphere

689 690 691 692 693 694

695 696

697 698 699

700 701 702

703 704 705

706 707 708 709

710 711 712 713 714

Tietema, A., Penuelas, J., Kesik, M., Brueggemann, N., Pilegaard, K., Vesala, T., Campbell, C.L., Olesen, J.E., Dragosits, U., Theobald, M.R., , Levy, P., Mobbs, D.C., Milne, R., Viovy, N., Vuichard, N., Smith, J.U., Smith, P., Bergamaschi, P., Fowler, D., Reis, S., 2007. Challenges in quantifying biosphere-atmosphere exchange of nitrogen species. Environmental Pollution 150, 125–139. Todini, E., 1996. The arno rainfall-runoff model. Journal of Hydrology 175, 339–382. Toth, E., Brath, A., Montanari, A., 2000. Comparison of short-term rainfall prediction models for real-time flood forecasting. Journal of Hydrology 239, 132–147. Umeda, T., Martien, P.T., 2002. Evaluation of a data assimilation technique for a mesoscale meteorological model used for air quality modeling. Journal of Applied Meteorology 41, 12–29. Vidale, P.L., Luthi, D., Frei, C., Seneviratne, S.I., Schar, C., 2003. Predictability and uncertainty in a regional climate model. Journal of Geophysical Research 108. Walko, R.L., Band, L.E., Baron, J., Kittel, T.G.F., Lammers, R., Lee, T.J., Ojima, D., Pielke, R.A., Taylor, C., Tague, C., Tremback, C.J., Vidale, P.L., 2000. Coupled atmospherebiophysicshydrology models for environmental modeling. Journal of Applied Meteorology 39, 931–944. Wilson, K.B., Hanson, P.J., Mulholland, P.J., Baldocchi, D.D., Wullschleger, S.D., 2001. A comparison of methods for determining forest evapotranspiration and its components: sap-flow, soil water budget, eddy covariance and catchment water balance. Agricultural and forest meteorology 106, 153–168.

22

172

51°0'0"W

#

48°0'0"W

45°0'0"W

# # #

#

#

±

Figure 1: Location of the rainfall and meteorological stations used in this study.

Rio Grande basin

Rainfall Stations

#

Meteorological Stations

#

#

Paper VI

23

173

20°0'0"S 22°0'0"S

24

174 st

(c)

Figure 2: The false-color Meteosat-9 image captured on January 21 at 2100 UTC over the Rio Grande basin (a). Compared to the regional atmospheric model (b), the spatial alignment of the air flow field calculated by the coupled model (c) showed a good agreement with the visible satellite image.

(b)

(a)

Assessment of the role of a process-based representation of land surface hydrological processes to the atmosphere

(b)

Figure 3: Temperature rates showed a better spatial representation of the cold front using the coupled model (b) than the regional atmospheric model(a).

(a)

Paper VI

25

175

26

176 10

20

(a)

Rainfall (mm/d) 30 40 50 60

70

80

(c)

Figure 4: 24-hour accumulated rainfall fields estimated by the models presented the same regional trend as observed daily rainfall, where high intensities of rainfall were concentrated in the southeastern part of the Rio Grande basin. This regional trend was directly associated with the orography of the basin.

(b)

0

Assessment of the role of a process-based representation of land surface hydrological processes to the atmosphere

Paper VI

(a)

(b) Figure 5: One cloudy region containing developing and mature cumulus clouds observed in the GOES-10 visible image (rectangle) captured on January 22nd at 1800 UTC (a) and daily rainfall observations at each rinafall gauging station (b). The ellipse indicates two 27 rainfall gauging stations which may have been affected by convective rainfalls derived from the developing and mature cumulus clouds in (a).

177

178

28

(d) Coupled Model

(c) Regional Atmospheric Model

Figure 6: Temperature (c and d) and air flow (a and b) fields near land surface calculated by both models. In a and b, rectangles indicate zones of convection. The larger one corresponds to the same region as the rainfall potential area whereas the smaller one represents a local convection near the border of the basin.

(b) Coupled Model

(a) Regional Atmospheric Model

Assessment of the role of a process-based representation of land surface hydrological processes to the atmosphere

(d) Coupled Model

(c) Regional Atmospheric Model

Figure 7: Relative humidity (a and b) and instantaneous rainfall (c and d) fields near land surface estimated using the regional atmospheric model and the coupled model.

(b) Coupled Model

(a) Regional Atmospheric Model

Paper VI

29

179

Assessment of the role of a process-based representation of land surface hydrological processes to the atmosphere

(a)

0

10

20

Rainfall (mm/d) 30 40 50

60

70

80

(b) Figure 8: Case study characterized by stable atmospheric conditions in the GOES-10 visible image (a) and low rainfall profile measured in the rainfall gauging stations (b).

30

180

(d) Coupled Model

(c) Regional Atmospheric Model

Figure 9: Temperature (c and d) and air flow (a and b) fields near land surface calculated by both models.

(b) Coupled Model

(a) Regional Atmospheric Model

Paper VI

31

181

PAPER VII Conceptual evaluation of the interplay of land use changes, hydrology and atmosphere by a hybrid atmospheric-hydrological coupled model F. F. Pereira and C. B. Uvo. , Manuscript.

Paper VII

Conceptual evaluation of the interplay of land use changes, hydrology and atmosphere by a hybrid atmospheric-hydrological coupled model F. F. Pereira, C. B. Uvo Department of Water Resources Engineering, Lund University, P.O. Box 118, SE-221 00 Lund, Sweden, tel: +46 46 2220435, fax: +46 46 2224435

Abstract Hybrid atmospheric-hydrological coupled models aim to simulate the physical processes in the soil, land surface and atmosphere including their feedback loops over a wide variety of soils, land uses and climates. To better reproduce this wide range of physical processes, the model conceptualisation has to be flexible and robust, while at the same time to harmonically couple physical processes in the soil, land surface and atmosphere as an unique and integrated system. In this study a conceptual evaluation of the interplay of land use changes, hydrology and atmosphere as given by a hybrid coupled model is carried out. The main goal is to assess whether the model behavior is in accordance with current understanding of the hydrological cycle of the Rio Grande basin under land use changes due to sugarcane expansion over its drainage area. The results obtained from 4 model runs using historic and possible land use scenarios show that exchange of water between soil, land surface and Email address: [email protected] (F. F. Pereira)

Preprint submitted to Journal of Hydrologic Engineering

October 27, 2013

185

Conceptual evaluation of the interplay of land use changes, hydrology and atmosphere by a hybrid atmospheric-hydrological coupled model

atmosphere is an important factor that determines which processes will dominate the water balance during wet and dry seasons under expansion of agricultural lands. Keywords: Integrated modelling system, conceptual model evaluation, Rio Grande basin, land use changes, hydrology, atmosphere 1. Introduction Human beings have altered the environment to fulfill their basic needs, such as food, shelter and clothing. These anthropogenic changes to the environment are often focused on local and global land use dynamics associated with urbanization, deforestation and agricultural practices (Grimmond, 2007; Lawrence et al., 2007; Gordon et al., 2008). Despite their crucial role as key drivers of progress and wellbeing, it is widely acknowledged that land use dynamics caused by human activities can potentially induce ecosystem regime shifts (Scheffer et al., 2001; Sitch et al., 2005; Parmesan and Yohe, 2002; Barnosky et al., 2012). In aquatic ecosystems, for example, a regime shift was triggered by the mass mortality of the sea urchin Diadema antillarum arising from the increase of nutrient loading in the Caribbean coral reefs. According to deYoung et al. (2008), the increase of nutrient loading was preceded by changes in land use practices during the early 80’s. In terrestrial ecosystems, regime shifts as a response to land use dynamics do not only have consequences on land surface processes but also on the whole climate system as discussed by Cao and Woodward (1998). This is because terrestrial ecosystems and the climate system are closely interconnected, especially by cycling of carbon and

2

186

Paper VII

water between vegetation, soil and the atmosphere (Matthews et al., 2005; Sellers et al., 1997; Betts, 2009; Ganzeveld et al., 2010). The consequences of land use dynamics to the carbon cycle have exhaustively been evaluated at several temporal scales, from days and weeks (Cao et al., 2012) to hundreds (Shevliakova et al., 2009) and thousands (Olofsson and Hickler, 2008) of years. Recently, a climate model has been used by Pongratz et al. (2009) to perform simulations over the last millennium with anthropogenic land use dynamics as the only factor affecting the climate system. Their simulations revealed that despite human influence on the carbon cycle beginning prior to industrialization, it only significantly altered the global mean temperatures after the strong population growth in the industrial period. Similarly to the carbon cycle, climate models also incorporate the water cycle when coupling processes in land surface, oceans and the atmosphere. However, hydrological feedbacks between land surface and atmosphere are still poorly represented by land surface parameterizations (Yang et al., 2012; Cr´etat and Pohl, 2012; Gianotti et al., 2012). Alternatively, hybrid coupled models already proved to have a more realistic approach to land surface hydrological processes by coupling hydrological and atmospheric models in a way to exploit their complementary strengths (Walko et al., 2000; Chen and Dudhia, 2001; Zabel and Mauser, 2013; Webb and Lock, 2012; Wilhelm et al., 2013). Since the use of hybrid coupled models has become more common as recent studies target a comprehensive approach to the interplay of hydrology, land use and atmosphere (Aufdenkampe et al., 2011; Schaefli et al., 2012; Cheruy et al., 2013; Shen et al., 2013), this study evaluates the performance

3

187

Conceptual evaluation of the interplay of land use changes, hydrology and atmosphere by a hybrid atmospheric-hydrological coupled model

of a hybrid coupled model (Pereira et al., 2013a) in representing hydrological interactions between land surface and atmosphere, including their feedback loops, over a watershed under land use changes due to anthropogenic activities. In Brazil, rapid sugarcane expansion observed by Rudorff et al. (2010) over the past 30 years as a response to government measures to boost the production of ethanol (Goldemberg, 2006; Matsuoka et al., 2011) made Brazil the current largest exporter of ethanol in the world (IEA, 2011). Thus, regarding its socio-economic relevance on both local and global scales, the sugarcane expansion over the Rio Grande basin in the State of S˜ao Paulo between early 90’s and late 2010 was chosen as case study of land use changes induced by anthropogenic activities. 2. Material and methods The numerical approach to atmosphere-land surface interactions given by a hybrid coupled atmospheric-hydrological model is assessed under four land use scenarios, which correspond to three historical land use scenarios generated from geostationary satellite imagery captured in 1993, 2000 and 2007, and a forth land use scenario where all areas suitable for cultivation of sugarcane as defined by BRASIL (2009) are filled with sugarcane plantations. The hybrid coupled model was run for each of the four land use scenarios over a simulation period of 1 year, capturing a full phenological cycle of sugarcane. During the simulations, each model run used roughly 2520 CPU hours over eight cores (approx. two weeks) and provided outputs every 3 hours of simulation. Thereafter, monthly mean values of land surface tem4

188

Paper VII

perature, rainfall, evapotranspiration and discharge were calculated out of 2920 3-hourly outputs from the hybrid coupled model. Finally, the interplay between atmosphere, hydrology and land use as given by the hybrid coupled model was assessed as changes in monthly mean values of land surface temperature, rainfall, evapotranspiration and discharge for the three historical land use scenarios and the plausible future scenario (hereafter called as HLU1993, HLU2000, HLU2007 and PFLU, respectively). This section presents a short description of the hybrid coupled model and study area as well as how comparisons between estimates of monthly mean values of land surface temperature, rainfall, evapotranspiration and discharge from the hybrid coupled model using HLU1993, HLU2000, HLU2007 and PFLU were conducted. 2.1. Hybrid coupled atmospheric-hydrological model The hybrid coupled atmospheric-hydrological model is composed of a hydrological model two-way coupled to a regional atmospheric model. It employs the Model for Large Basins developed at the Institute for Hydraulics Research (MGB-IPH, Collischonn (2001)) and the Brazilian Regional Modelling System (BRAMS, Freitas et al. (2009)) as the hydrological and atmospheric models, respectively. Further, the hybrid coupled model couples BRAMS and MGB-IPH in a way to best exploit their respective strengths. One of its main characteristics lies on the replacement of the BRAMS numerical approach to water fluxes from land surface to the atmosphere – which are calculated as a function of land surface parameterizations – by process-based estimates of evapotranspiration given by MGB-IPH. Additionally, the hybrid coupled model uses a two-way coupling strategy 5

189

Conceptual evaluation of the interplay of land use changes, hydrology and atmosphere by a hybrid atmospheric-hydrological coupled model

described in detail by Pereira et al. (2013a) to overcome spatial and temporal mismatches associated with model grid and time step employed by BRAMS and MGB-IPH. Their coupling strategy avoids upscaling and downscaling issues by running both models on the same grid size. Since BRAMS and MGB-IPH output variables are calculated at their own computational cell centers that do not always match each other, the coupling strategy counted on an algorithm to identify the nearest computational cell for the exchange of variables between BRAMS and MGB-IPH. Due to numerical stability constraints, BRAMS runs with a smaller time step than the daily time step often used by MGB-IPH at basin scale. The two-way coupling strategy employed by the hybrid coupled model deals with this temporal mismatch by assuming a single time step for both models, which has to be a multiple of the daily time step used by MGB-IPH and to satisfy numerical stability conditions related to explicit schemes employed by BRAMS to solve the conservation equations for heat and moisture (Pielke et al., 1992). Additional details regarding the hybrid coupled model such as how and which variables are exchanged between the models, assumptions and coupling limitations can be found at Pereira et al. (2013a). 3. Study area Rio Grande basin is a sub-basin of the Paran´a basin formed by the rivers Grande, Pardo, Sapuca´ı, Verde, das Mortes and Mogi-Gua¸cu. It has an area of 145000 km2 located in the eastern upper Paran´a basin (Fig. 1) where altitudes vary from 300 to 2700 m.a.s.l.. The classification of land use in 6

190

Paper VII

the Rio Grande basin includes three distinct categories: Atlantic Rainforest, pasture and agriculture (IBGE, 1991). Agricultural activity represents a large portion of the Rio Grande basin and, it was classified into sugarcane and agriculture of grain. Regarding types of soils, Rio Grande basin presents five major types: latosols, lithosols, cambisols, podzolics and alluvial soils which may be broken down into three groups: high, medium and low infiltration capacity (FAURGS, 2007). Soils with high and medium infiltration capacity are equally distributed across the Rio Grande basin whereas soils with low infiltration capacity are concentrated along the drainage network. Figure 1 Although most of surface runoff in the Rio Grande basin is regulated by dams, its hydrological regime is strongly induced by land use changes due to harvesting practices and shifting cultivation (WWFBrasil, 2008). After the flow regulation, a representative sample of daily values of discharge collected at the outlet of the basin, from 1970 to 2010, indicates that surface runoff varies from minimum values of 1000 m3 /s (dry season) to maximum values over 12000 m3 /s (rainy season). Locally, measurements of runoff are also monitored at hydroelectric power plants. At Funil, Camargos, Furnas, P Colˆombia, Marimbondo and A Vermelha power plants, daily runoff ranges 70 – 3731 m3 /s, 34 – 1253 m3 /s, 174 – 7497 m3 /s, 251 – 8367 m3 /s, 532 – 9234 m3 /s and 303 – 10186 m3 /s, respectively. Production of electrical power is the largest water use in the Rio Grande basin (IPT, 2008). Over 11% of the installed electric generation capacity of Brazil is at hydroelectric installations in the Rio Grande basin (ANEEL,

7

191

Conceptual evaluation of the interplay of land use changes, hydrology and atmosphere by a hybrid atmospheric-hydrological coupled model

2005). To meet this demand for electricity, hydroelectric power plants are constrained by a minimum operating flow, which varies from power plant to power plant. As recently presented (ONS, 2013), the minimum operating flow at all hydroelectric power plants are shown in table 1. Table 1 According to Espinosa (2011), spatial and temporal distribution of rainfall in the Rio Grande basin is highly induced by synoptic systems over the southeastern and south-central Brazil. In addition, annual rainfall analysis carried out by CPRM (2012) indicate that annual average rainfall varies from 1500 to 2000 mm in the basin. Annual average evapotranspiration ranges from 800 to 1000 mm (Ruhoff, 2011). Throughout the year, a seasonal variability of evapotranspiration has been identified by Rocha et al. (2002). Over the Rio Grande basin, their studies revealed that daily evapotranspiration can oscillate between 6 mm/d−1 in the rainy season and 1 mm/d−1 in the dry season. 4. Evaluation of the model behaviour To evaluate the reciprocal actions of land use changes, hydrology and atmosphere as estimated by the hybrid coupled atmospheric-hydrological model, 7 key response variables strongly related to the fluxes of water and heat across the atmosphere-land surface interface are monitored over 1 year of simulation. In the near-surface soil layer, the water balance is a function of the amount of water that remains in the layer and the amount of water that passes through the layer contributing to the deep drainage (i.e. horizontal fluxes of 8

192

Paper VII

water below the root zone). Further, both of them are strongly associated with the exchange of water between soil and land surface as the soil becomes saturated or not. Therefore, soil water content and water contribution to deep drainage are chosen as key response variables. Moving upwards, evapotranspiration, runoff and temperature are assumed as key response variables to fluxes of water and heat in the land surface while precipitation and vertically-integrated water mixing ratio of the atmospheric column from the land surface to the upper troposphere are the key response variables in the atmosphere. Besides their strong relationship with fluxes of water and heat across atmosphere, land surface and soil, the choice of the key response variables also include their high sensitivity to environmental disturbance (e.g. pollution and land use changes) as shown in previous studies (Verburg et al., 2002; Foley et al., 2005; Menut et al., 2013). To assess the model behaviour in reproducing the exchanges of water and heat between soil, land surface and atmosphere under land use changes, the hybrid coupled model is run for 4 land use scenarios in the Rio Grande basin over a 1 year period, from 1st Jan 2009 to 31st Dec 2009. All of these four land use scenarios represent scenarios of sugarcane expansion over the Rio Grande basin, the three first ones are historical land use scenarios of the basin in 1993, 2000 and 2007 and they were produced by Pereira et al. (2013b) whereas the forth scenario is a plausible future scenario in which all areas suitable for cultivation of sugarcane as defined by BRASIL (2009) are filled with sugarcane plantations. Despite the hybrid coupled model using a time step of 10 seconds, outputs are produced every 3 hours and aggregated on a monthly basis. This is

9

193

Conceptual evaluation of the interplay of land use changes, hydrology and atmosphere by a hybrid atmospheric-hydrological coupled model

because biophysical parameters such as leaf area index and canopy or surface resistance – resistance of vapour flow through a transpiring crop – for sugarcane plantations only present relevant changes over a month or so (Robertson et al., 1999; Andr´e et al., 2010). In the hybrid coupled model, initial and boundary conditions are separately set for BRAMS and MGB-IPH models. In the BRAMS, a fourdimensional data assimilation technique described by Umeda and Martien (2002) is used to interpret atmospheric boundary conditions provided every 6 hours by global atmospheric analyses. In addition, initial soil moisture conditions are calculated using global analyses of monthly precipitation derived from satellite and surface measurements. For doing so, a hybrid model (Gevaerd and Freitas, 2006) which estimates soil moisture based on the Global Precipitation Climatology Program (GPCP) and the Tropical Rainfall Measuring Mission (TRMM) is used. Moreover, soil temperature is considered as vertically homogeneous and its value is set equal to the value of air temperature near land surface. In the MGB-IPH, initial conditions are set by changing interception and infiltration stores. A initial hot start for interception and infiltration stores is adopted according to simulations performed by Pereira et al. (2013b). 5. Results and discussions In spite of the hybrid model dynamically coupling soil, land surface and atmosphere as an integrated system, this section presents and discusses its estimates of fluxes of water and heat individually for the soil-land surface and land surface-atmosphere interfaces. Initially, section 5.1 analyses the fluxes 10

194

Paper VII

of water between soil and land surface by the response of monthly evapotranspiration and runoff rates to monthly average values of soil water content and accumulated flow to the deep drainage. Thereafter, section 5.2 presents the exchange of water and heat between land surface and atmosphere as seasonal variations in monthly air temperature, precipitation, evapotranspiration – which is closely associated with latent heat flux – and amount of water vapor stored in the atmospheric column given by the vertically-integrated water mixing ratio. Below, the results obtained from the 4 model runs using HLU1993, HLU2000, HLU2007 and PFLU – hereafter known as R1993, R2000, R2007 and RPFLU – are presented and discussed. Although the hybrid coupled model divides the simulation domain into individual uniform cells of 10x10 km, the results from R1993, R2000, R2007 and RPFLU are spatially-averaged over the Rio Grande basin. 5.1. Soil-land surface interface The water balance in the soil near-surface layer as estimated by the hybrid coupled model shows that soil water content is associated with variations in flow to the deep drainage over the year. While soil water content decreases, the amount of water that contributes to the deep drainage increases. However, the magnitude of this response is neither linearly equivalent nor instantaneous. This is because, besides the withdraw of water to the deep drainage, the hybrid coupled model also includes loss of storage water due to evapotranspiration and replenishment of water from effective precipitation to estimate the soil water content (Fig. 2). Under land use changes, results obtained from R1993 and RPFLU reveal 11

195

Conceptual evaluation of the interplay of land use changes, hydrology and atmosphere by a hybrid atmospheric-hydrological coupled model

that the hybrid coupled model indicates relevant changes in the soil water balance. During the germination stage of sugarcane from June to August, for instance, the hybrid coupled model captures an increase in soil water content as a result of the reduction of the average of leaf area index over the basin for RPLFU. On the other hand, estimates of soil water content from R1993 continuously decreases until August. After August, when a wet season month starts – defined here as a month where monthly accumulated precipitation is equal or greater than 100 millimetres – estimates of soil moisture content instantaneously increase for R1993 whereas the rising curve is delayed by a month for RPFLU (Fig. ). This late rising curve in soil water content is addressed to changes related to physiological maturity in sugarcane plantations (from germination to maturation stage between August and September), since precipitation effectively becomes smaller as the average leaf area index is increased. Runoff as calculated by the hybrid coupled model is a combination of precipitation, evapotranspiration and soil water content in the near-surface soil layer. At the outlet of the Rio Grande basin, accumulated monthly runoff from R1993 and RPFLU vary within a range of historical records from 1930 to 2010. Nevertheless, estimates of runoff from R1993 and RPFLU present different seasonal patterns over the year, which are closely associated with soil water content in the soil near-surface layer and evapotranspiration and precipitation over the land surface. From January to to May, a falling limb of the hydrographs is explained by a significant reduction in precipitation over the basin. It is been followed by a dry season period – when monthly accumulated precipitation do not exceed 100 millimetres. During this period,

12

196

Paper VII

the runoff at the outlet of the Rio Grande basin comes basically from the soil near-surface layer and delayed flow from the deep drainage. As evapotranspiration rates significantly decreases for RPFLU, the effective precipitation increases which enables a higher replenishment of the soil near-surface layer and, thereby higher runoff during the dry period compared to estimates of runoff from R1993. Figure 5.2. Land surface-atmosphere interface The coupling strategy used by the hybrid coupled model to couple MGBIPH and BRAMS has evapotranspiration as its key variable to enhance the exchange of water from land surface to atmosphere. Practically, the results obtained from R1993, R2000, R2007 and RPFLU show that the hybrid coupled model is very sensitive to seasonal changes in evapotranspiration rates. While the effects of these seasonal changes in evapotranspiration on precipitation are only marginal, estimates of vertically-integrated water mixing ratio from RPFLU presents a slight increase compared to those from R1993, R2000 and R2007 in October and November (see figure ). For these two months, accumulated monthly rainfall reaches up to 600 mm – which indicates that the atmospheric column is close to saturation. At such conditions, the vertically-integrated water mixing ratio becomes more susceptible to variations as more water is added to the atmospheric column by evapotranspiration.

13

197

Conceptual evaluation of the interplay of land use changes, hydrology and atmosphere by a hybrid atmospheric-hydrological coupled model

6. Conclusions This study outlined a conceptual evaluation of the interplay between soil, hydrology and atmosphere under land use changes given by a hybrid atmospheric-hydrological coupled model. The results obtained from 4 model runs which represented historic and possible future land use scenarios of sugarcane expansion over the Rio Grande basin were used to conceptually analyse how the fluxes of water across the soil-land surface and land surfaceatmosphere interfaces behave under different land use scenarios. Overall, the hybrid coupled model was able to capture changes in seasonal patterns of evapotranspiration, soil water content and runoff arising from the expansion of sugarcane plantations over the Rio Grande basin. During the dry season, when surface runoff is mostly driven by baseflow, results obtained from the hybrid coupled model revealed that the harvesting of sugarcane plantations significantly increased runoff rates by increasing the flow through the soil near-surface layer. Besides runoff and soil moisture, another consequence of the harvesting of sugarcane to the water balance was the reduction of fluxes of water from land surface to atmosphere along with the averaged leaf area index over the basin. The results from a hybrid atmospheric-hydrological coupled model presented here shows that the exchange of water between soil, land surface and atmosphere is a key factor on the understanding of how changes in the environment induced by anthropogenic activities can affect the whole climate system.

14

198

Paper VII

References Andr´e, R.G.B., Mendon¸ca, J.C., Marques, V.S., Pinheiro, F.M.A., Marques, J., 2010. Aspectos energ´eticos do desenvolvimento da cana-de-a¸cu ´car. parte 1: Balan¸co de radia¸ca˜o e parˆametros derivados. Revista Brasileira de Meteorologia 25, 375–382. ANEEL, 2005. Atlas brasileiro de energia el´etrica. 2nd ed., Brazil (Government), Bras´ılia, BR. Aufdenkampe, A.K., Mayorga, E., Raymond, P.A., Melack, J.M., Doney, S.C., Alin, S.R., Aalto, R.E., Yoo, K., 2011. Riverine coupling of biogeochemical cycles between land, oceans, and atmosphere. Frontiers in Ecology and the Environment 9, 53 – 60. doi:10.1890/100014. Barnosky, A.D., Hadly, E.A., Bascompte, J., Berlow, E.L., Brown, J.H., Fortelius, M., Getz, W.M., Harte, J., Hastings, A., Marquet, P.A., Martinez, N.D., Mooers, A., Roopnarine, P., Vermeij, G., Williams, J.W., Gillespie, R., Kitzes, J., Marshall, C., Matzke, N., Mindell, D.P., Revilla, E., Smith, A.B., 2012. Approaching a state shift in earth’s biosphere. Nature 486, 52–58. Betts, A.K., 2009. Land-surface-atmosphere coupling in observations and models. Journal of Advances in Modeling Earth Systems 1, 18. BRASIL, 2009. Zoneamento Agroecolø’gico da Cana-de-A¸cu ´car: Expandir a produ¸ca˜o, presevar a vida, garantir futuro. 1st ed., Minist´erio da Agricultura, Pecu´aria e Abastecimento, Bras´ılia, BR.

15

199

Conceptual evaluation of the interplay of land use changes, hydrology and atmosphere by a hybrid atmospheric-hydrological coupled model

Cao, L., Bala, G., Caldeira, K., 2012. Climate response to changes in atmospheric carbon dioxide and solar irradiance on the time scale of days to weeks. Environmental Research Letters 7, 034015. Cao, M., Woodward, F.I., 1998. Dynamic responses of terrestrial ecosystem carbon cycling to global climate change. Nature 393, 249–252. Chen, F., Dudhia, J., 2001.

Coupling an advanced land surface-

hydrology model with the penn state-ncar mm5 modeling system. part i:

Model implementation and sensitivity.

Weather Review 129, 569–585.

Monthly

doi:http://dx.doi.org/10.1175/1520-

0493(2001)129¡0569:CAALSH¿2.0.CO;2. Cheruy, F., Dupont, A.C.J.C., Ducharne, A., Hourdin, F., Haeffelin, M., Chiriaco, M., Idelkadi, A., 2013.

Combined influence of atmospheric

physics and soil hydrology on the simulated meteorology at the sirta atmospheric observatory. Climate Dynamics 40, 2251 – 2269. doi:0.1007/s00382012-1469-y. Collischonn, W., 2001. Hydrologic simulation of large basins (in Portuguese). Ph.D. thesis. Instituto de Pesquisas Hidr´aulicas. Porto Alegre, BR. CPRM, 2012. Geodiversity of Brazil: Knowing the past to understand the present and predict the future. Geological Survey of Brazil, Rio de Janeiro. Cr´etat, J., Pohl, B., 2012. How physical parameterizations can modulate internal variability in a regional climate model. Journal of the Atmospheric Sciences 63, 714–724. doi:http://dx.doi.org/10.1175/JAS-D-11-0109.1.

16

200

Paper VII

deYoung, B., Barange, M., Beaugrand, G., Harris, R., Perry, R.I., Scheffer, M., Werner, F., 2008. Regime shifts in marine ecosystems: detection, prediction and management. Trends in Ecology & Evolution 23, 402–409. Espinosa, A.M., 2011. Previs˜ao quantitativa por conjunto da precipita¸c˜ao do ver˜ao 2007/2008 na Bacia de Rio Grande. Ph.D. thesis. Universidade de S˜ao Paulo. FAURGS, 2007. Previs˜ao de afluˆencia a reservat´orios hidrel´etricos m´odulo 1. Projeto FAURGS/FINEP 40.04.0094.00. Technical Report. FINEP. Porto Alegre, BR. Foley, J.A., DeFries, R., Asner, G.P., Barford, C., Bonan, G., Carpenter, S.R., Chapin, F.S., Coe, M.T., Daily, G.C., Gibbs, H.K., Helkowski, J.H., Holloway, T., Howard, E.A., Kucharik, C.J., Monfreda, C., Patz, J.A., Prentice, I.C., Ramankutty, N., Snyder, P.K., 2005. Global consequences of land use. Science 309, 570 – 574. doi:10.1126/science.1111772. Freitas, S.R., Longo, K.M., Silva Dias, M.A.F., Chatfield, R., Silva Dias, P., Artaxo, P., Andreae, M.O., Grell, G., Rodrigues, L.F., Fazenda, A., , Panetta, J., 2009. The coupled aerosol and tracer transport model to the brazilian developments on the regional atmospheric modeling system (cattbrams) part 1: Model description and evaluation. Atmospheric Chemistry and Physics 9, 2843–2861. Ganzeveld, L., Bouwman, L., Stehfest, E., van Vuuren, D.P., Eickhout, B., Lelieveld, J., 2010. Impact of future land use and land cover changes

17

201

Conceptual evaluation of the interplay of land use changes, hydrology and atmosphere by a hybrid atmospheric-hydrological coupled model

on atmospheric chemistry-climate interactions. Journal of Geophysical Research: Atmospheres 115, D23301. Gevaerd, R., Freitas, S., 2006. Estimativa operacional da umidade do solo para iniciao de modelos de previso numrica da atmosfera. parte i: descrio de maetodologia e validao. Revista Brasileira de Meteorologia 21, 1–15. Gianotti, R.L., Zhang, D., Eltahir, E.A.B., 2012. Assessment of the regional climate model version 3 over the maritime continent using different cumulus parameterization and land surface schemes. Journal of Climate 25, 638–656. doi:http://dx.doi.org/10.1175/JCLI-D-11-00025.1. Goldemberg, J., 2006. The ethanol program in brazil. Environmental Research Letters 1, 014008. doi:10.1088/1748-9326/1/1/014008. Gordon, L.J., Peterson, G.D., Bennett, E.M., 2008. Agricultural modifications of hydrological flows create ecological surprises. Trend in Ecology & Evolution 23, 211–219. Grimmond, S., 2007. Urbanization and global environmental change: local effects of urban warming. Cities and Global Environmental Change 173, 83–88. IBGE, 1991. Classifica¸ca˜o da vegeta¸ca˜o brasileira, adaptada a um sistema universal. Technical Report. Instituto Brasileiro de Geografia e Estat´ıstica. Rio de Janeiro, BR. IEA, 2011. Renewable energy: Markets and prospects by technology. Technical Report. International Energy Agency. Paris, FR. 18

202

Paper VII

IPT, 2008.

Diagn´ostico da situa¸ca˜o dos recursos h´ıdricos na Bacia

Hidrogr´afica do Rio Grande (BHRG) SP/MG. Technical Report 96581205. Instituto de Pesquisas Tecnol´ogicas do Estado de S˜ao Paulo. Lawrence, D., D’Odorico, P., Diekmann, L., DeLonge, M., Das, R., Eaton, J., 2007. Ecological feedbacks following deforestation create the potential for a catastrophic ecosystem shift in tropical dry forest. Proceedings of the National Academy of Sciences of the United States of America 105, 20696–20701. Matsuoka, S., Ferro, J., Arruda, P., 2011. The brazilian experience of sugarcane ethanol industry, in: Tomes, D., Lakshmanan, P., Songstad, D. (Eds.), Biofuels. Springer New York, pp. 157–172. doi:10.007/978-1-44197145-6-9. Matthews, H.D., Weaver, A.J., Meissner, K.J., 2005. Terrestrial carbon cycle dynamics under recent and future climate change. Journal of Climate 18, 1609–1628. Menut, L., Tripathi, O.P., Colette, A., Vautard, R., Flaounas, E., Bessagnet, B., 2013. Evaluation of regional climate simulations for air quality modelling purposes. Climate Dynamics 40, 2515 – 2533. doi:10.1007/s00382012-1345-9. Olofsson, J., Hickler, T., 2008. Effects of human land-use on the global carbon cycle during the last 6000 years. Vegetation History and Archaeobotany 17, 605–615.

19

203

Conceptual evaluation of the interplay of land use changes, hydrology and atmosphere by a hybrid atmospheric-hydrological coupled model

ONS, 2013.

Invent´ario das restri¸c˜oes operativas dos aproveitamentos

hidroel´etricos. Technical Report 00132013. Operador Nacional do Sistema El´etrico. Parmesan, C., Yohe, G., 2002. A globally coherent fingerprint of climate change impacts across natural systems. Nature 421, 37–42. Pereira, F.F., Moraes, M.A.E., Uvo, C.B., 2013a. Implementation of a twoway coupled atmospheric-hydrological system for environmental modeling at regional scale. Hydrology Research , To appeardoi:10.2166/nh.2013.335. Pereira, F.F., Tursunov, M., Uvo, C.B., 2013b. Towards the response of water balance to sugarcane expansion in the rio grande basin, brazil. Hydrology and Earth System Sciences 10, 5563 – 5603. doi:10.5194/hessd-10-55632013. Pielke, R.A., Cotton, W.R., Walko, R.L., Trembaek, C.J., Lyons, W.A., Grasso, L.D., Nieholls, M.E., Moran, M.D., Wesley, D.A., Lee, T.J., Copeland, J.H., 1992. A comprehensive meteorological modeling system rams. Meteorology and Atmospheric Physics 49, 69–91. Pongratz, J., Reick, C.H., Raddatz, T., Claussen, M., 2009. Effects of anthropogenic land cover change on the carbon cycle of the last millennium. Global Biogeochemical Cycles 23, GB4001. Robertson, M.J., Inman-Bamberb, N.G., Muchowa, R.C., Wood, A.W., 1999. Physiology and productivity of sugarcane with early and mid-season water deficit. Field Crops Research 64, 211–227.

20

204

Paper VII

Rocha, H.R., Freitas, H.C., Rosolem, R., Ju´arez, R.I.N., Tannus, R.N., Ligo, M.A., Cabral, O.M.R., Silva Dias, M.A.F., 2002. Measurements of co2 exchange over a woodland savanna (cerrado sensu stricto) in southeast brasil. Biota Neotropica 2, 1 – 11. Rudorff, B.F.T., Aguiar, D.A., Silva, W.F., Sugawara, L.M., Adami, M., Moreira, M.A., 2010. Studies on the rapid expansion of sugarcane for ethanol production in s˜ao paulo state (brazil) using landsat data. Remote Sensing 2, 1057–1076. doi:10.3390/rs2041057. Ruhoff, A.L., 2011. Sensoriamento remoto aplicado a` estimativa da evapotranspira¸ca˜o em biomas tropicais. Ph.D. thesis. Instituto de Pesquisas Hidr´aulicas. Schaefli, B., van der Ent, R.J., Woods, R., Savenije, H.H.G., 2012. An analytical model for soil-atmosphere feedback. Hydrology and Earth System Sciences 16, 1863 – 1878. doi:10.5194/hess-16-1863-2012. Scheffer, M., Carpenter, S., Foley, J.A., Folke, C., Walker, B., 2001. Catastrophic shifts in ecosystems. Nature 413, 591–596. Sellers, P.J., Dickinson, R.E., Randall, D.A., Betts, A.K., Hall, F.G., Berry, J.A., Collatz, G.J., Denning, A.S., Mooney, H.A., Nobre, C.A., Sato, N., Field, C.B., Henderson-Sellers, A., 1997. Modeling the exchanges of energy, water, and carbon between continents and the atmosphere. Science 275, 502–509. Shen, C., Niu, J., Phanikumar, M.S., 2013. Evaluating controls on coupled hydrologic and vegetation dynamics in a humid continental climate wa21

205

Conceptual evaluation of the interplay of land use changes, hydrology and atmosphere by a hybrid atmospheric-hydrological coupled model

tershed using a subsurface-land surface processes model. Water Resources Research 49, 2552 – 2572. doi:10.1002/wrcr.20189. Shevliakova, E., Pacala, S.W., Malyshev, S., Hurtt, G.C., Milly, P.C.D., Caspersen, J.P., Sentman, L.T., Fisk, J.P., Wirth, C., Crevoisier, C., 2009. Carbon cycling under 300 years of land use change: Importance of the secondary vegetation sink. Global Biogeochemical Cycles 23, GB2022. Sitch, S., Brovkin, V., von Bloh, W., van Vuuren, D., Eickhout, B., Ganopolski, A., 2005. Impacts of future land cover changes on atmospheric co2 and climate. Global Biogeochemical Cycles 19, GB2013. Umeda, T., Martien, P.T., 2002. Evaluation of a data assimilation technique for a mesoscale meteorological model used for air quality modeling. Journal of Applied Meteorology 41, 12–29. Verburg, P.H., Soepboer, W., Veldkamp, A., Limpiadaada, R., Espaldon, V., Mastura, S.S.A., 2002. Modeling the spatial dynamics of regional land use: The clue-s model. Environmental Management 30, 391 – 405. doi:10.1007/s00267-002-2630-x. Walko, R.L., Band, L.E., Baron, J., Kittel, T.G.F., Lammers, R., Lee, T.J., Ojima, D., Pielke, R.A., Taylor, C., Tague, C., Tremback,

C.J.,

Vidale,

P.L.,

2000.

Coupled atmospherebiophysic-

shydrology models for environmental modeling. plied Meteorology 39, 931–944.

doi:http://dx.doi.org/10.1175/1520-

0450(2000)039¡0931:CABHMF¿2.0.CO;2.

22

206

Journal of Ap-

Paper VII

Webb, M.J., Lock, A.P., 2012. Coupling between subtropical cloud feedback and the local hydrological cycle in a climate model. Climate Dynamics , 1–17doi:10.1007/s00382-012-1608-5. Wilhelm, C., Rechid, D., Jacob, D., 2013. Dynamic coupling of regional atmosphere to biosphere in the new generation regional climate system model remo-imove. Geoscientific Model Development 6, 3085–3135. doi:10.5194/gmdd-6-3085-2013. WWFBrasil, 2008. Monitoring alterations in vegetation and land use in the Brazilian portion of the Upper Paraguay river basin. Technical Report. Non-governmental organizations: Ecoa - Ecologia e A¸ca˜o, Conservation International, Avina Foundation. Yang,

B.,

Qian,

Y.,

Lin,

G.,

Leung,

R.,

Zhang,

Y.,

2012.

Some issues in uncertainty quantification and parameter tuning:

a

case study of convective parameterization scheme in the wrf regional climate model. 2409–2427.

Atmospheric Chemistry and Physics 12,

URL: http://www.atmos-chem-phys.net/12/2409/2012/,

doi:10.5194/acp-12-2409-2012. Zabel, F., Mauser, W., 2013. 2-way coupling the hydrological land surface model promet with the regional climate model mm5. Hydrology and Earth System Sciences 17, 1705–1714. doi:10.5194/hess-17-1705-2013.

23

207

Conceptual evaluation of the interplay of land use changes, hydrology and atmosphere by a hybrid atmospheric-hydrological coupled model

Table 1: Minimum operating flow at the hydroelectric power plants used in this study (ONS, 2013).

Hydroelectric power plant (m3 /s) Funil

34.0

Camargos

34.0

Furnas

174.0

P Colˆombia

251.0

Marimbondo

1100.0

A Vermelha

1600.0

24

208

Paper VII

Figure 1: Location and elevation map of the Rio Grande basin.

25

209

Conceptual evaluation of the interplay of land use changes, hydrology and atmosphere by a hybrid atmospheric-hydrological coupled model

Figure 2: The results from R1993, R2000, R2007 and RPFLU (or EMBRAPA) performed over 1 year period. The water balance in the soil is presented in (a) and (d) by monthly accumulated water contribution to the deep drainage and monthly average soil water content. In the land surface and atmosphere, exchange of water is shown in (b), (c), (e) and (f) as monthly accumulated values of evapotranspiration, precipitation, verticallyintegrated water mixing ration and runoff. For runoff (f), box plots mean records from 1930 to 2010 with a 90% confidence interval.

26

210

Paper VII

211