BALWOIS 2004
Ohrid, FY Republic of Macedonia, 25-29 May 2004
System of Prediction & Warning of Floods in the Water Basin of Strymonas/Struma River Theologos Mimides, Spyros Rizos, Konstantinos Soulis, Panagiotis Karakatsoulis Agricultural University of Athens Athens, Greece Dobri Dimitrov National Institute of Meteorology & Hydrology Sofia, Bulgaria
Abstract Strymonas/Struma is collecting waters from four countries: Bulgaria, Serbia, FYROM and Greece. Most of its basin area is located in Bulgaria and Greece, while the upper part of its basin is in Bulgaria. There are important hydrotechnical structures just below the Bulgarian – Greek border, and the floods generated in the Bulgarian part of the basin could significantly affect the security of those structures and their operational rules. That is why several years ago a project related to flood warning at Strymonas/Struma river basin was formulated and its first phase was completed in 2000. The main objective of the project was to demonstrate the principal possibility for issuing reliable warnings for hazardous flood events with sufficient lead-time to organize flood mitigation measures. Key-words: Strymonas/Struma river basin; flood; hydrograph; routing; thematic maps; Alladin; Crocus; satellite prediction; early warning.
Introduction The Strymonas/Struma River is exhibiting a quite important role for Greece, due to its size and the fact that it irrigates many fertile lands, while it supplies water to many important Greek cities like Serres. On the other hand Strymonas/Sruma River represents a big threat for the Greek part of its basin due to its very small slope along all the way from the Greek-Bulgarian borders to the sea leading to very important flood hazards. The Strymonas/Struma river basin lie between the coordinates 22o13'43'' and 24o33'06' East longitude and 40o38'30' and 42o49'58' North latitude. It has a total area of 16780 Km2 from which 5987 Km2 belongs to the Greek part, 8510 Km2 to the Bulgarian part and 1612 Km2 to the FYROM part and 671 Km2 of the Strymonas/Struma river basin (see Figure 1).
Droughts and Floods
1
BALWOIS 2004
Ohrid, FY Republic of Macedonia, 25-29 May 2004
Figure 1. Strymonas/Struma River basin position. The elevation of the watershed ranges between 0 and 2900 m. The terrain is characterized by steep slopes at the Bulgarian part of the watershed and smooth slopes at the Greek part. That’s why floods happen mainly at the Greek part of the watershed and why the floods at the Greek part and caused from the rainfall and the snowmelt at the Bulgarian part of the watershed (see Figure 2). A key role on the control of Strymonas/Struma flow at the Greek part of the river basin is Kerkini Lake and the hydrotechnical structures around it. Kerkini Lake is located near the Greek-Bulgarian borders at the beginning of a large plain, extending from the borders to the sea. In Figure 1 is shown the position of Strymonas/Struma River basin. The larger amount of the flow reaching Kerkini comes from the Bulgarian part of the river basin while an important part comes from the FYROM. A small amount of the flow comes from the Serbian part of the river Basin.
Droughts and Floods
2
BALWOIS 2004
Ohrid, FY Republic of Macedonia, 25-29 May 2004
Figure 2. Strymonas/Struma River basin relief and Kerkini Lake position. Kerkinis structures are regulating Strymonas/Struma River flow in order to achieve different goals like water supply, irrigation and flood prevention. At the same time Kerkini Lake is a valuable and protected wetland making very difficult the construction off new structures in order to eliminate flood risk. This is the main reason why a system of floods prediction and warning is essential in order to optimize the management of Kerkini Lake and to minimize the impacts of a possible flood (Mimides et al 2001) (Burrell et al 2002). (See also Sorensen 2000 and Krzyztofowict 1995).
Objectives The System of Prediction & Warning of Floods in the Hydrological Basin of Strymonas/Struma River was designed to facilitate an early warning system. The main goal of the system is to provide: First Warning: Based on predictions made from weather models and using appropriate hydrological models a first warning appears. The prediction of the weather models concerning to precipitation duration and intensity and to snow melting. The target is to have prediction of at least 3 days. Second Warning: Based on the data coming from the automatic telemetric meteorological stations network and using the appropriate hydrological models the second warning appears. Alarm: Based on the data coming from the automatic telemetric hydrological stations network and using the appropriate runoff routing model the final alarm appears. The final alarm will be accompanied with maps of the high flood risk areas. Droughts and Floods
3
BALWOIS 2004
Ohrid, FY Republic of Macedonia, 25-29 May 2004
Further more the system will be able to present information about the river regime in almost real time or to give access, to the historical data gathered at the relational data base.
Approach The final system will comprise a GIS application linked with: A telemetric meteorological and hydrological gauging stations network. A relational database containing the historical meteorological and hydrological data. In the relational database will also collect, pre-process and store the real time data from automatic meteorological and hydrological stations. A weather model, which will provide predictions of precipitation intensity duration and a snowmelt model. A rainfall runoff model calibrated and validated for the upper part of the hydrological basin. A routing model which will provide real time flood forecasting by routing flood hydrographs through the system of the river and Kerkini lake. A Graphical User Interface, which will allow non GIS or computer experts to easily use the system. An operational representation of the complete system is shown in Figure 1. The completed project will consists an integrated Geoinformational system capable to manage the linkage between the components and to visualize the final results (Linda 2001). Additionally the system is designed to facilitate the storage and the management of all the relative data either in tabular or spatial format (see Figure 3).
Figure 3. Flow chart of the system operation.
Droughts and Floods
4
BALWOIS 2004
Ohrid, FY Republic of Macedonia, 25-29 May 2004
Telemetric meteorological and hydrological gauging stations network The network consists of 15 automatic telemetric stations. The stations are divided in three categories. Each category has different configuration concerning the sensors type and the measurements parameters while all the stations has the same data logger and telemetry devices. The parameters measured are: Rainfall-Snow, Wind speed, Wind direction, Air temperature, Humidity, Solar radiation, Barometric pressure, Water level and Soil moisture. The data collected from the data loggers are transmitted with GSM modems to the Central Computer once a day or when precipitation depth overruns a certain threshold. All the stations are placed at the Bulgarian part of the river basin. That happens, because there is interest mainly for the pick flows reaching the Greek-Bulgarian borders. The positions of the meteorological and hydrological stations are shown at Figure 4.
Figure 4. Division of the upper part of the Strymonas/Struma River basin to 14 sub-basins. Positions of meteorological and hydrological gouging stations.
Droughts and Floods
5
BALWOIS 2004
Ohrid, FY Republic of Macedonia, 25-29 May 2004
Weather models The weather forecast model used is ALADIN. The model was adapted for the study area in order to achieve forecasts of precipitation duration and intensity for at least three days (Golding B. W., December 2000). The output of the model is in a mesh of regular grid points and the distance between the points is about 10Km. The output of ALADIN is directly transferred to the Geoinformational system in order to prepare the input data package for the rainfall-runoff model. The output also is transferred to the CROCUS model to calculate possible snow melting so as to pass the additional runoff to the rainfall-runoff model. The CROCUS model is precisely describes the evolution off the snow and the run-off. The data needed provided from ALADIN model while the detailed distribution of snow cover derives from satellite imagery using special techniques.
Rainfall runoff model (HEC-1) When predicted or measured rainfall depth overruns the specified threshold, a rainfall-runoff model is used to calculate the hydrograph at the Greek-Bulgarian borders. The model used is HEC-1. HEC-1 is a program written by the United States Army Corps of Engineers to perform hydrologic routing. It Generates hydrographs from both rainfall and snowmelt and adds, diverts, or routes them to stream reaches, reservoirs, or ponds. The model adapted to the upper part of the Strymonas\Struma River basin. The first step in this direction was the division of the Hydrological basin to 14 sub-basins. The division’s criteria were the uniformity of sub-basins relief and size and the meteorological and the hydrological gouging stations positions. In figure 4 illustrated the sub-basins division. Next step was the preliminary calibration and validation of the model based on historical data. The inputs of HEC-1 can be divided in two main categories. The first category concerns to the data that cannot change or rarely changing. These are related with the relief, the soil, the geology and the land cover. (Soil Conservation Service U.S. Department of Agriculture, 1972, 1973) The above data enter at the model as a set of parameters that are calculated when needed from the Geoinformational System (Mack 1994). The calculation of these parameters happens in an automatic way using specially developed methodologies. That makes easy to change the input data set of the HEC-1 model when major changes occur (e.g. changes to land cover due to a big fire). The second category consists of the data that are different every time the model runs. These data are mainly the rainfall and the snow melting and prepared from the geoinformational system as a package with the data of the first category to consist the input dataset of the model each time there is a need for it to run. Finally, a methodology designed that allows the interface between the geoinformational system and the model in an automatic way.
Routing model (DAMBRK) The hydrographs resulting from the HEC-1 model either from predicted storm events or from measured heavy rain events as well as the measured pick discharges at the Greek-Bulgarian borders, are routed through Strymonas River to Kerkini Lake and from the Lake Kerkini to the Strymonas outflow at the Strymonas bay, so as to determine if these discharges are able to cause a flood and when that happens to determine the areas that are going to be covered with water. The model used for this purpose is DAMBRK. The basic version of this program was developed by D.Fread at the Hydrologic Research Laboratory, Office of Hydrology, National Weather Service (NWS), USA and it is a state of the art flood wave routing model in addition to the computation capabilities of the characteristics of the following a dam failure, downstream flood wave. To adapt DAMBRK model to Strymonas River, an exact Determination of the geometry of the cross sections of Strymonas River, from the Greek–Bulgarian borders up to the sea outflow is needed. Moreover much information on Kerkini’s Lake dam characteristics is required. The input of the model is the hydrograph to be routed in the form of a time series and is prepared by the Geoinformation system. It can origin even from HEC-1 output or from a measured hydrograph. The output of the model turns back to the geoinformational system, which gives the warning or the alarm and also visualizes the results in the form of maps.
Droughts and Floods
6
BALWOIS 2004
Ohrid, FY Republic of Macedonia, 25-29 May 2004
Geoinformational system The Geoinformational system consists of a Spatial Database, a Relational Database and a set of modules responsible for the communication of the parts of the integrated project. Furthermore a Graphical User Interface was designed to allow non GIS or computer experts to easily use the system and to make easy the visualization of the results in the form of thematic maps through the internet. Relational Database The main purpose of the relational database is to store, manage and retrieve all the tabular data concerning weather parameters coming from the meteorological and hydrological stations or from the forecasting models. The relational database was designed to allow the processing of non-spatial queries to be completed independent of the GIS. Key relational tables are linked to the Geographical database to promote the communication of the two databases. The main concept in the Database design is that data arriving from various data sources must be validated and converted to the appropriate form prior to be stored or manipulated. In general, all data arriving either from the telemetry network or from the forecast models ΑLΑDIN, CROCUS, HEC-1 (1990) and are less than a critical ‘threshold' Limit, are stored in the DBMS part called ‘Historical’ and are being kept there only for future reference (statistics, further calibration and validation of the models or any other use). Data that exceed the critical ‘threshold’ limit are stored also in the DBMS part called ‘Productive’ and are being processed as the system continues its operation. In this way the productive DBMS part holds only the information necessary for the real time hydrological modelling. As a result the disk space required is significantly smaller and the runtime performance of the GeoInformational System much better. A very difficult task in the relation database operation is to make compatible the data from different sources. In fact the data from different sources are stored in separate tables due to the need to compare the parameters values for each time step. On the other hand, for the same reason, the separate tables for the different data sources have to be in the same exactly format in order to allow the comparison of the data and to make possible the processing of these data and the preparation of the input data for the hydrological models in an automatic way. That’s why the data coming from the ALADIN model, which are in Grid format are converted to tables containing the same parameters and for the same positions as these that are measured at the meteorological stations. The value of its parameter and for each time step is set us the value of the nearest grid cell to each position. Geographical Database The geographical database was planned to comprise all the required data for the application of the spatial analysis procedures needed and the calculation of the hydrological parameters. (Dercas et al 2002) The format of the data was designed to promote the effective and efficient running of the analysis routines. Special care was taken for the digital geographical database to be easily managed and updated. Also the spatial database was planned in a appropriate way in order to afford data for input in the hydrological model HEC-1 and DAMBRK and also in order to be able to visualize the spatial distribution of any possible flood. Based on the above, the required data layers of the geographical database were selected (see Figure 5), as well as the data model and the topology of each data layer in parallel to their attributes. Finally because of the spatial data come from different countries (Greece, Bulgaria, FYROM), with different geographic reference systems, we decided to transform all of the spatial data in the common Hellenic Geographic Reference System EGSA87, which is in use from the Hellenic Military Geographical Service for the last years.
Droughts and Floods
7
BALWOIS 2004
Ohrid, FY Republic of Macedonia, 25-29 May 2004
Figure 5. Schematic representation of the data layers of the Geographical database. This reference system is similar with the Universal Transverse Mercador (UTM) system and was chosen for the reason that the Strymonas/Struma hydrological basin extends mainly from the north to south and secondary from west to east, so the most appropriate projection system is the Transverse Mercador. Also because the origin of the geographic reference system is 240 east of Greenwich that coinciding with the vertical central axis of the study area, this system is the most appropriate for the data projection. More specific the data layers can be divided to those, which were based on original data, and those, which were derived from other data layers. Original data layers: Digital Elevation Model The Digital Elevation Model (DEM) has ArcInfo GRID format. The pixel size, which depends on the required detail of the results, is 30m by 30m. The detail of the grid is very important for the hydrological calculations such as the definition of the watersheds. The DEM has also to be hydrologically correct because sinks can beguile the automatic hydrological calculations, leading to totally false results. For this reason the DEM was checked and corrected with special functions (Jenson et al 1988). The DEM was interpolated from the original data, which were contours with 20 m interval, peaks, hydrographic network, lakes and the coast. All the above data was vectorized on screen from scanned topographic maps in 1/50.000 scale.
Droughts and Floods
8
BALWOIS 2004
Ohrid, FY Republic of Macedonia, 25-29 May 2004
Hydrographic network The Hydrographic Network has vector format with line topology. Its attributes are the stream name (when it is possible) and the stream order. The river network can be extracted from the DEM but there is an error risk at flat areas or in places artificially altered. So it was vectorized on screen from scanned topographic maps in 1/50.000 scale and after that combined with the automatically created one (Tarboton 1991).. Geology The geology has vector format and polygon topology. Required attributes concern permeability categories, but the geological formation name was also included. The categorization depends on the methodology used to calculate necessary hydrological parameters. For the methodology that was used four categories of permeability are required. The original data come from geological maps in scale 1:50.000 which were provided from the Geology and Mineral Exploration (IGME) for the Greek part, from geological maps in scale 1:100.000 for the Bulgarian part and from geological maps in scale 1:50.000 for the FYROM part which afforded from relative services for the other two countries, which were digitized and vectorized. A very laborious work was to assign uniform geological symbols and descriptions to the whole area while identical geological formation appears with different symbols in each region, due to differences in the methodology followed by the geologists worked in each region. This work was mainly based on bibliographical and empirical data. Soil type The Soil type layer has vector format and polygon topology. The required attributes here again is soil type (according FAO) and the permeability category. The information sources for the of soil data layer come from a soil map in scale 1:500.000 for the Greek part of Strymonas/Struma river hydrological basin. This map was received from a in recognizing scale soil study of the E. E. For the Bulgarian part the data was taken from a soil map in scale 1:200.000 and for the part that belongs to FYROM from a soil map in scale 1:200.000. Land cover The data sources of the Land Cover Digital Model, is LandSat5 satellite imagery. The pixel size of the Land Cover grid is 30m x 30m. The land cover was categorized in six classes (Forests, Pasture, Bare soil, Urban areas, Crops, Surface waters). The categorization classes selected to support the methodology used to calculate necessary hydrological parameters. Meteorological stations This data layer has vector format and point topology. The attributes comprises information describing the stations and the parameters that they measure. There is also a also a key field allowing the relation of this data layer with the relational data base containing the meteorological and hydrological data. Derived data layers: From this original data layers derived many other data layers, which were required for the analysis operations such as the Slope and the Flow-direction data layers which derived from Digital Elevation Model using special routines and the Permeability categories and the Land cover categories which derived from the Land cover, Soil type and Geology data layers. All the above data layers has ArcInfo Grid format. The Flow-direction data layer needed to define the watersheds of the major hydrological basins of the Hydrological Basin while the Permeability categories and the Land cover categories needed to calculate SCS Curve Numbers for each sub-basin. The Geographical Information System Software that was used to build, manage and update this geographical database was ESRI ArcGIS. The data model used was ESRI Geodatabase data model. This data model ensures efficiency in management and update of the database, while it allows the building of special multi user applications in the future.
Droughts and Floods
9
BALWOIS 2004
Ohrid, FY Republic of Macedonia, 25-29 May 2004
In the Geographical data base was also developed routines and methodologies able to automatically calculate the necessary parameters for the hydrological models to run. This was of great importance in order to make possible the automation of the hydrological model execution when it is needed. Automation modules The Geoinformational system accomplished with different modules responsible for the communication of the projects components. One of the most important modules is the one that makes possible the update of the relational database with the data coming from the telemetric network and from the forecast models. This module is responsible to connect to the telemetric stations once a day and retrieve the data stored there or to receive the alert of the telemetric stations when a storm event occurs and then pass it to the relational database. It is also responsible to update the relational database with the forecast models’ outputs after formatting it in the standard format. An other very important module is the automation module of the Geoinformational system that facilitates the recycling of the data between the relational database, the hydrological models and the Geographical database, achieving this way the automatic execution in a sequence of the deferent parts of the system and integrating the final Early Warning System. This module takes the meteorological data from the relational database and combines it with the data coming from the GIS so as to prepare the input data file for the HEC-1 model. After that it is processing the output data of the HEC-1 model and passes it to the DAMBRK model. Finally it takes the output of the DAMBRK model and passes it to the Geoinformationla system in order to give the alarm, in case that there is a possibility of flood and to visualize the threatened areas. Graphical User Interface The Graphical User Interface is designed to be build in the frame of the GIS. In this way beyond the visualization of the warnings and the alarms it will be capable to visualize in the form of thematic maps the threatened areas of a possible flood. It is an essential part of the completed system because it will allow non GIS or computer experts to easily manage the databases, retrieve the data stored in the databases, visualize the warnings and use the system in general.
Conclusions The first phase of the project concerning the implementation of a System of Prediction & Warning of Floods in the Water Basin of Strymonas/Struma River was completed in 2000. Since that time many modifications and improvements have been made and many other have to be done in the future in order to make it operational. Some of these improvements are: Further calibration and validation of the hydrological models taking in to account the new data gathered from the automatic meteorological stations. Solving of problems concerning the telemetric meteorological and hydrological network and the automatic data receiving. Integration of some parts of the Geoinformational System. On the other hand the most important achievements of the project were: Cross border cooperation between neighboring countries facilitating the optimization of the management of their common river. Creation of an integrated and very detailed geographical database from data coming from different countries. Real time data acquisition, using GSM technology, from automatic meteorological and hydrological stations located in different countries giving the possibility to dramatically increase the response time to flood events. Application of hydrological models based on forecasts of precipitation duration and intensity and snow melting making possible the further increment of the response time. The completion of the project will make possible the optimization of the management of the Strymonas/Struma River, will minimize the losses from floods and will give information for many other applications.
Droughts and Floods
10
BALWOIS 2004
Ohrid, FY Republic of Macedonia, 25-29 May 2004
Reference Burrell E. Montz and Eve Gruntfest, March 2002, Flash flood mitigation: recommendations for research and applications, Global Environmental Change Part B: Environmental Hazards, Volume 4, Issue 1, , Pages 15-22 Charles Obled, Guillaume Bontron and Rémy Garçon, August 2002, Quantitative precipitation forecasts: a statistical adaptation of model outputs through an analogues sorting approach, Atmospheric Research, Volume 63, Issues 3-4, Pages 303-324 Cuena, J., Molina, M., Garrote, L., Blain, W.R., Cabrera, E., 1992. Combining simulation models and knowledge bases for real time flood management. In: (Eds.), Hydraulic Engineering Software IV. Computational Mechanics, London co-published with Elsevier Science, Southampton, pp. 587–598. Golding B. W., December 2000, Quantitative precipitation forecasting in the UK, Journal of Hydrology, Volume 239, Issues 1-4, 20 Pages 286-305 HEC – 1, Flood Hydrograph Package, 1990. Hydrologic Engineering Center, U.S. Army Corps of Engineers, Davis, CA., U.S.A. Jenson S. K. and Domingue J. O., 1988. Extracting Topographic Structure from Digital Elevation Data for Geographic Information System Analysis, Photogrammetric Engineering and Remote Sensing. Klemes, V., 1983, Conceptual and scale in hydrology. J. of Hydrology, 65. 1-23. Krzysztofowicz, R., 1995. Recent advances associated with flood forecast and warning systems. Review of Geophysics 33 (Supp B), 1129. Kuittinen, R. 1992. Remote sensing for Hydrology: Progress and Prospects. Operational Hydrology Report, No 36, WMO – No 773, Geneva, Switzerland. Linda See and Robert J. Abrahart, October 2001, Multi-model data fusion for hydrological forecasting, Computers & Geosciences, Volume 27, Issue 8, Pages 987-994 Mack M. J., 1994, HER – Hydrological Evaluation of Runoff; the Soil Conservation Service Curve Number technique as an interactive computer model, Computers & Geosciences Mimides Th., Karakatsoulis P., Rizos S., Soulis K. 2001, New Perspectives in the management of water resources. The case of Tranboundary hydrological basin of Strymonas/Struma River between Greece and Bulgaria. “Proceedings of the 7th International Conference on Environmental Science and Technology”, Vol. II, pp. 831-839. Nicholas Dercas, Konstantinos Soulis, Spyros Kyritsis, 2002, A GIS Application for the Integrated Water Resources Management of Naxos Island (Greece). “Proceedings of the 1st Conference of Hellenic Association of ICT in Agriculture, Food and Environment, pp. 347-356 Soil Conservation Service U.S. Department of Agriculture, 1972, National Engineering Handbook, v.4. Hydrology. U.S. Government Printing Office. Soil Conservation Service U.S. Department of Agriculture, 1973, A method for estimating volume and rate of runoff in small watersheds, Tech. Paper 149, U.S. Government Printing Office. Sorensen, J.H., 2000. Hazard warning systems: review of 20 years of progress. Natural Hazards Review 1 (2), 119–125. Tarboton D. G., R. L. Bras, I. Rodriguez-Iturbe, 1991, On the Extraction of Channel Networks from Digital Elevation Data, Hydrological Processes Ven ten Chow, Maidment, D.R., and Mays L.M. 1988. Applied Hydrology, Mc Graw - Hill, N.Y., N.Y., U.S.A.
Droughts and Floods
11
