United Nations Environment Programme

Climate Early Warning System Feasibility Report: Early Warning Systems and Hazard Prediction

March 2012 Dr. Zinta Zommers University of Oxford

1

Table of Contents 1   INTRODUCTION AND PURPOSE ....................................................................................................................... 3   2.   METHODOLOGY................................................................................................................................................. 6   3.   CURRENT EARLY WARNING SYSTEMS.......................................................................................................... 8  

3.1.   COMPONENTS OF EWS ...................................................................................................................... 8   3.2.   KEY ACTORS IN EWS......................................................................................................................... 9   3.3.   EWS BY NATION .............................................................................................................................. 10   3.4.   EWS BY HAZARD ............................................................................................................................. 15   3.4.1.   3.4.2.   3.4.3.   3.4.4.   3.4.5.  

Drought .......................................................................................................................................... 15   Famine........................................................................................................................................... 20   Fire................................................................................................................................................. 22   Floods ............................................................................................................................................ 26   Cyclones/Hurricanes...................................................................................................................... 30  

3.5.   EWS EVALUATION ........................................................................................................................... 33   4.   HAZARD PREDICATION CAPABILITIES ........................................................................................................ 40  

4.1.   BACKGROUND ON WEATHER AND CLIMATE FORECASTING TECHNIQUES ........................................... 40   4.1.1.   Weather Forecasting ..................................................................................................................... 40   4.1.2.   Climate Forecasting....................................................................................................................... 41  

4.2.   SEASONAL FORECASTS ................................................................................................................... 42   4.2.1.   Infrastructure.................................................................................................................................. 42   4.2.2.   Standard products ......................................................................................................................... 44   4.2.3.   Verification ..................................................................................................................................... 45  

4.3.   MULTI-YEAR AND DECADAL FORECASTS .......................................................................................... 46   4.4.   CONCLUSION ................................................................................................................................... 48   5.   STEPS FORWARD............................................................................................................................................ 49  

5.1.   KEY FINDINGS .................................................................................................................................. 49   5.2.   POSSIBLE ACTIONS .......................................................................................................................... 50   5.3.   POSSIBLE CLIM-WARN DESIGN ...................................................................................................... 52   5.3.1.   Biome or ecosystem targeted approach ........................................................................................ 52   5.3.2.   Risk assessment rather than warning............................................................................................ 53   5.3.3.   Seamless integrated system.......................................................................................................... 53   6.   ACRONYMS ...................................................................................................................................................... 55   7.   REFERENCES................................................................................................................................................... 56   8.   APPENDIX 1. LIST OF INDIVIDUALS CONSULTED....................................................................................... 60  

2

1 Introduction and Purpose “No nation, however large or small, wealthy or poor, can escape the impact of climate change. Rising sea levels threaten every coastline. More powerful storms and floods threaten every continent. More frequent drought and crop failures breed hunger and conflict in places where hunger and conflict already thrive. The security and stability of each nation and all peoples…are in jeopardy.” President Barack Obama addressing the United Nations (New York Times 2009) 2012 will be remembered as “the year that winter was cancelled” in Canada”1. It was the year that grass fires raged in the Prairies in mid winter, and farmers already began planting crops in March (Hirtzer 2012). Canadians experienced, on average, temperatures 3.6 degrees oC higher than normal and 18 percent less precipitation (Gulli 2012). Around the world these trends are not uncommon. The International Panel on Climate Change has concluded that, “warming of the climate system is unequivocal” (IPCC 2007). In addition to changes in temperature, observations since 1950 indicate increases in extreme weather events (IPCC 2011). The recent IPCC Special Report on Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation (IPCC SREX) predicts further increases in extreme events in the 21st century, including a growing frequency of heat waves, rising wind speed of tropical cyclones, and increasing intensity of droughts. A one-in- 20 years “hottest day” event is likely to occur every other year by the end of the 21st century. Heavy precipitation events are also on the rise, potentially impacting the frequency of floods and almost certainly affecting landslides (IPCC 2011). These trends have significant impacts on human lives, national economies and even on national identities (Figure 1). For example, winter, and the extreme cold, have been embraced in Canada as something distinctive to the country (Gulli 2012). Canada without winter may shake national identity to the core (Gulli 2012). For First nations communities in the Canadian North, “warm winters are a tragedy, a catastrophe,” writes Adam Gopnik. As the World Conference on Disaster Reduction (A/CONF.206/6) concluded, “Disaster loss has grave consequences for the survival, dignity and livelihoods of individuals, particularly the poor, and hard-won development gains.” Over the period 1991-2005 3,470 million people were affected by disasters globally, 960,000 people died, and economic losses totalled US$ 1,193 billion (UNISDR 2008). Losses from extreme weather are expected to continue to increase in future (IPCC 2012). However regions differ in vulnerability. Total economic losses from natural disasters are greatest in developed countries, while deaths from natural disasters are highest in developing countries (IPCC 2011). In fact, from 1970 to 2008, more than 95% of deaths from natural disasters occurred in developing countries (IPCC 2012). Mortality risk is approximately 225 times greater in low-income countries than in OECD countries when similar numbers of people are exposed to tropical cyclones (Peduzzi et al. 2012, from UNISDR 2011).

1

According to David Phillips, senior climatologist at Environment Canada (Gulli 2012).

3

Figure 1. Total number of people affected by disasters of natural origin

Nearly all current efforts to cope with climate change focus on either mitigation to reduce emissions or on long-term adaptation to adjust to changes in climate. Although it is imperative to continue with these efforts, the ongoing pace of climate change and the slow international response suggests that a third option is becoming increasingly important: the creation of climate change early warning systems to protect populations against the immediate threat of climate-related extreme events, including heat waves, forest fires, floods and droughts. Already in 2005, United Nations Secretary-General Kofi Annan called for the establishment of a worldwide early warning system (EWS) for all natural hazards (UN 2005). The 2010 Cancun Agreements specifically invite “all Parties to enhance action on adaptation…by…enhancing climate change related disaster risk reduction strategies (such as) early warning systems.” IPCC SREX Report (2012) concludes that, “The implementation of early warning systems does reduce loss of lives and, to a lesser extent, damage to property and was identified by all the extreme event case studies (heat waves, wildfires, drought, cyclones, floods and epidemic disease) as key to reducing impacts from extreme events.” This research, conducted in collaboration with the United Nations Environment Programme (UNEP) examines the feasibility of creating a global early warning system for climate change, “CLIM-WARN”. The purpose of the CLIM-WARN system will be to provide timely and actionable warnings to institutions, businesses, governments and the general public about the imminence of climate related extreme events. At the initiative of Prof. Joseph Alcamo, Chief Scientist for the Executive Office of the United Nations Environment Programme, an expert meeting was organised to develop an initial concept for CLIMWARN. The meeting was held from Monday 1 November to Tuesday 2 November 2010 at the Château de Penthes, in Geneva, Switzerland. Sixteen experts attended the meeting from several agencies and research institutions, including the National Oceanic and Atmospheric Administration (NOAA), UNEP’s

4

Division of Early Warning and Assessment (DEWA), the United States Geological Survey (USGS), the Famine Early Warning Systems Network (FEWSNET), the World Glacier Monitoring System, the Global Climate Observing System, the Global Fire Monitoring Centre, the International Research Institute for Climate and Society at Columbia University, the University of California at Berkeley, and the University of Greenwich. The meeting concluded that: 1. CLIM-WARN would focus on hydrometeorological hazards and “climate extremes”, rather than weather extremes. This would include heat waves, droughts, floods and wildfires. 2. CLIM-WARN would focus on users (people, communities and institutions) . Mechanisms for engaging with local communities must be at the heart of the system. As the IPCC recently asserted, “In some contexts, a more decentralized, sustainable early warning system that is well integrated into existing, local development … and that includes a strong focus on user needs, local knowledge, and practice can result in more effective communication and uptake of information required to reduce climate-related disasters” (IPCC 2009). 3. CLIM-WARN should build on existing systems. This could range from the creation of networks with very loose linkage to a very tight linkage by which some or all of these systems are brought under a single coordinated umbrella. Different options will have different costs and different chances of successfully delivering actionable warnings. 4. CLIM-WARN would provide two kinds of warnings: (1) short-term, operational warnings (with a time horizon of six months to two years); (2) medium- to long-term warnings (with a time horizon of two to ten years) concerning hot spots of increasing risks of climate impacts. 5. It would be useful to elaborate CLIM-WARN by conducting a specific case study or case studies, possibly in the Sahel or along a transect in Nepal and India from mountains to the ocean. The motivation for this report comes from conclusions 3 and 4, related to the use of existing systems and the timescale of warnings. The purpose of this report is: 1) to review early warning systems for hydrometeorological hazards, indicating availability, quality and accessibility, and gaps; 2) to assess current scientific ability to model hazards on the short-term and medium to long-term timescales.

5

2. Methodology Early warning systems (EWS) have evolved considerably during the past two decades (Glantz 2004, UN 2006). EWS systems can be formal with readily visible structures that are part of the national government bureaucracy or they can be informal aspects of local, cultural traditions (Glantz 2004). Formal famine early warning systems were first created in response to extended droughts and famines in the West African Sahel and in the Horn of Africa in the 1970s and 1980s (Glantz 2004). Today, there are countless EWS in operation, covering most types of natural hazards, conflict, ecological changes, health-related and complex humanitarian crises (Glantz 2004, UN 2006). For example, the Disaster Risk Reduction Portal for Asia lists 107 projects related to early warning for fires, floods, cyclones, heat waves or droughts in Asia, run by over 50 organizations2. The diversity of definitions of EWS systems and the diversity of actors makes a comprehensive review of EWS around the world nearly impossible. As a result, this report will focus on EWS that are offered by national governments. This is an important area for analysis as national governments are charged with the provision of public goods such as ensuring the economic and social well-being, safety and security of their citizens from disasters. They also control budgetary allocations as well as the legislative process that guides actions by other actors (IPCC 2012). According to the United Nations (2006), “national governments are responsible for policies and frameworks that facilitate early warning and usually also for the technical systems for preparing and issuing timely warnings. They have responsibility to ensure that warnings and related responses address all the population.” National meteorological services (NMS) act as the official providers of warning. Approximately 70 countries have legislation that sets out NMS warning responsibilities. In Australia, for example, the 1906 Meteorological Act of Parliament established that there should be a single Federal Meteorological Department responsible for both science and services. China’s 1999 Meteorological Law actually prohibits organizations, other than NMS, from issuing warnings. The World Meteorological Organization has encouraged members to enact such legislation. Resolution 26 of the 13th Meteorological Congress, “Urges Members to mandate the NMS as the official voice in issuing weather warnings for public safety to help minimize risks to the health and safety of citizens as well as the primary national authority and official source of information and policy advice on the present and future state of the atmosphere and other aspects of national weather and climate, in support of policy development.” This report will also focus on early warning systems for droughts, famines, fires, floods, and hurricanes/cyclones. This selection is relatively arbitrary. Heat waves, cold spells, blizzards, tornados, hail and lightning, extra tropical storms are other hydrometeorological hazards that could be included in the study. However, the selected hazards were identified in the previous CLIM-WARN expert meeting as areas of interest. Research for this report was conducted between November 2011 and March 2012. Results are based on an academic literature review and discussions with scientists and experts at UN agencies, academic institutions and meteorological agencies. Detailed interviews were conducted with over 30 individuals across four continents on a non-for-attribution basis. See Appendix 1 for list of these participants. Sections 3 and 4 draw heavily from research reports by two scientists, Dr. Veronica Grasso

2

See http://www.drrprojects.net/drrp/drrpp/project/list. Data accessed on March 18, 2012. The majority of lead agencies were non-governmental organizations. Only four government and nine inter-governmental organizations were listed as lead agencies.

6

and Dr. Simon Mason. Presentations by scientists at the Planet Under Pressure Conference in London, March 26 - 29, 2012, also contributed to this report. Stakeholder discussions took place at the World Meteorological Organization’s Climate observation community dialogue, “Addressing Gaps and Opportunities for improved observations and monitoring data in support of the Global Framework for Climate Service (GFCS)”, held in Geneva on December 15, 2011.3 In order to assess the state of EWS, official government progress reports on the implementation of the Hyogo Framework for Action (Priority 2, core indicator 2.3) were reviewed. The 2009 – 2011 progress reports assess current national strategic priorities with regard to disaster risk reduction actions and identify progress towards core indicator 2.3 (“early warning systems are in place for all major hazards”). National statements were reviewed and any mention of EWS for hydrometeorological hazards recorded.4 Preliminary findings of this report were presented at the “United Nations and Climate Vulnerability” session of the Association of American Geographers Annual Meeting, held in New York from February 24 – 28, 2012. Audience feedback was incorporated into the report. Collectively, this process generated a wealth of input and significant interest from experts wishing to collaborate further on CLIM-WARN in future.

3

Details about the meeting can be found here: http://www.wmo.int/pages/prog/wcp/wcdmp/gfcs_obs.php National statements can be downloaded from PreventionWeb (http://www.preventionweb.net/english/hyogo/progress/priority2/?pid:222) 4

7

3. Current Early Warning Systems “There is no perfect Early Warning System, except on paper, in governmental plans, or in a PowerPoint presentation.” Dr. Michael Glantz, Consortium for Capacity Building, University of Colorado Boulder

3.1.

Components of EWS

There are many definitions of early warning systems. According to the UN International Strategy for Disaster Reduction an early warning system is, “the set of capacities needed to generate and disseminate timely and meaningful warning information to enable individuals, communities and organizations threatened by a hazard to prepare and to act appropriately and in sufficient time to reduce the possibility of harm or loss.”5 However, others have argued for a much broader conception of early warning systems. Consensus does not exist over the meaning of the terms “early”, “warning” or “system”. Different sectors of society need different amounts of time to respond. What may be “early” for some populations is still insufficient time for others. “Warnings” may range from broad outlooks to specific alerts. It has been argued that historical trends offer a form of warning. Others claim that warnings must give future projections. A “system” may be composed of subsystems that include mechanisms for forecasting, transmission and reception. Many societies rely on systems of informal “local knowledge”. These do not involve formal bureaucratic structures but nevertheless allow individuals to cope with certain hazards in a timely fashion (Box 1). Finally, the different EWS systems may have different goals. EWS may serve to educate, to reduce infrastructure damage and economic costs, or to save lives. Operational goals may include creating and maintaining credibility, identifying the appropriate warning level, minimizing political interference and, maintaining transparency (Glantz 2004). Given these differences, Glantz (2004) therefore argues that the definition of an EWS must be “broad enough as to allow for a wide range of interpretations and flexible enough to accommodate in time and space the societal recognition of new hazards and the development of new EWS technologies.” Box 1. Indigenous EWS According to Mercer et al (2010) the use of indigenous knowledge alongside scientific knowledge is increasingly advocated but there is as yet no clearly developed framework demonstrating how the two may be integrated to reduce community vulnerability to environmental hazards. The experiences of the members of Singas village, situated in Morobe Province, Papua New Guinea (PNG), illustrate how indigenous knowledge can contribute to disaster risk reduction. Singas village is a small community situated along the banks of one of PNG’s major rivers, the Markham River. As a consequence, it is affected by yearly flooding following heavy rains experienced during the rainy season. The example holds particular significance as not only does the river represent a potential hazard but it is also the source of the community’s livelihood and therefore holds great importance within the community. As a result the community is extremely pro-active in its efforts to mitigate the consequences of flooding. Indigenous knowledge in five specific areas, namely building methods, social linkages, land use planning, food strategies and environmental strategies, has proven to help contribute to the community’s ability to mitigate the impact of regular flooding events (Mercer and Kelman 2008).

Regardless of definitions, all EWS must address “five Ws” (Glantz 2004): 1) What is happening with respect to the hazard(s) of concern?; 2) Why is this a threat in the first place (i.e., what are the underlying causes for potential adverse impacts)?; 3) When is it likely to impact (providing as much lead 5

http://www.unisdr.org/we/inform/terminology

8

time as possible to at-risk populations)?; 4) Where are the regions most at risk?; 5) Who are the people most at risk (i.e., who needs to be warned)? To be effective, EWS must be people-centred and must integrate four elements: 1) knowledge of risks; 2) technical monitoring and warning service; 3) dissemination of warnings to those at risk; and 4) public awareness and preparedness to act (UN 2006). Failure in any one of these four elements may cause failure of the entire EWS (Grasso 2012) Figure 2. Four elements of people-centred EWS (UN 2006)

3.2.

Key Actors in EWS

Early warning systems require contributions from a wide range of actors and institutions, including local communities, national governments, regional organizations, non-governmental organizations, as well as the private sector and the science community (UN 2006, IPCC 2012). As mentioned in Section 2, this report focuses on national actors. However the roles of three key agencies, the World Meteorological Organization (WMO), the United Nations International Strategy for Disaster Reduction (UNISDR) and the Office for Coordination of Humanitarian Affairs (OCHA), are highlighted below. Their activities help support many national EWS and their involvement will be critical to the future development of CLIMWARN. WMO The World Meteorological Organization (WMO) helps coordinate the efforts of national governments and supports climate monitoring through the Global Observing System, the Global Telecommunications System and the Global Data Processing and Forecasting System. The WMO’s observing system includes 14 satellites, more than 10 000 manned and automatic surface weather stations, 1 000 upperair stations, some 7 000 ships, 100 moored and over 1 000 drifting buoys, and hundreds of weather radars measure daily key parameters of the atmosphere, land and ocean surface. In addition, over 3 000 commercial aircraft provide more than 150 000 observations each day. Through establishment of

9

standards, guidelines and procedures for collection, quality control, data formatting and archiving data, the WMO has further helped national governments enhance EWS capacity (Jones and Mason 2012). The WMO is currently trying to strengthen climate observation, research and information management systems by creating a Global Climate Services Framework. The primary goal of the Framework will be to ensure greater availability of access to and use of climate services for all countries (WMO 2011). While this information can be used to support EWS, strengthening EWS is not a specific goal of the Framework. UNISDR The United Nations International Strategy for Disaster Reduction (UNISDR) serves as the focal point in the United Nations system for the coordination of disaster reduction. It also supports the implementation of the “Hyogo Framework for Action 2005-2015: Building the Resilience of Nations and Communities to Disasters” (HFA). HFA was adopted by the World Conference on Disaster Reduction in January 2005 and has been endorsed by 168 member states. The framework identifies five priority areas for action that will allow countries to strengthen risk governance capacities. The second of these priorities relates specifically to EWS: “identify, assess and monitor disaster risks and enhance early warning”. HFA areas for action include: A) Develop early warning systems that are people-centred, in particular systems whose warnings are timely and understandable to those at risk, which take into account the demographic, gender, cultural and livelihood characteristics of the target audiences, including guidance on how to act upon warnings, and that support effective operations by disaster managers and other decision makers. B) Establish, periodically review, and maintain information systems as part of early warning systems with a view to ensuring that rapid and coordinated action is taken in cases of alert/emergency. C) Establish institutional capacities to ensure that early warning systems are well integrated into governmental policy and decision-making processes and emergency management systems at both the national and the local levels, and are subject to regular system testing and performance assessments. D) Implement the outcome of the Second International Conference on Early Warning held in Bonn, Germany, in 2003, including the strengthening of coordination and cooperation among all relevant sectors and actors in the early warning chain in order to achieve fully effective early warning systems. E) Implement the outcome of the Mauritius Strategy for the further implementation of the Barbados Programme of Action for the sustainable development of small island developing States, including establishing and strengthening effective early warning systems as well as other mitigation and response measures. OCHA The Office for Coordination of Humanitarian Affairs (OCHA) is also involved in early warning and disaster risk reduction. Specifically, OCHA assists in the operation of the Humanitarian Early Warning Service (HEWS) that serves as a common platform for humanitarian early warnings for natural hazards and socio-political developments worldwide, also under the framework.

3.3.

EWS by Nation

As national meteorological and hydrological services are responsible for continuously observing, and forecasting hazards, national submissions to the Hyogo Framework for Action (HFA) Progress Review give an indication of the state of EWS. National governments lead the HFA submission process,

10

although this may include the input of intergovernmental organizations as well as local institutions (UNIDSR 2011). Countries are asked to rate achievements towards HFA from 1 to 5, with 1 representing minor achievement and 5 representing comprehensive achievement. As ratings are not triangulated, submissions reflect how governments perceive progress, rather than actual outcomes (UNISDR 2011). Nevertheless, submissions can be a useful indicator of gaps and current challenges. One hundred and thirty-three countries participated in the 2009-2011 HFA Progress Review: 58 percent of the countries in the Americas, 72 percent in Asia, 61 percent in Africa, 53 percent in Europe and 28 percent in Oceania. Eighty-six countries submitted reports on progress towards HFA Priority 2, core indicator 2.3 (early warning systems). Table 1. The number of countries from different regions that submitted reports on HFA Priority 2, core indicator 2.3 (EWS) and their self assigned rankings. Ranking Region Africa Americas Asia Europe Oceania Grand Total

1 1 1

2 1 4 3

2

2 10

3 9 6 7 3 3 28

4 7 18 5 6 2 38

5 2 1 1 3 1 8

Grand Total 20 29 17 12 8 86

The majority of countries assigned themselves a rank of 3 or 4, indicating “institutional commitment attained, but achievements are neither comprehensive nor substantial” or “substantial achievement attained but with recognized limitations in key aspects, such as financial resources and/ or operational capacities.” Guinea-Bissau and Lebanon both only reported minor progress with few signs of forward action (rank 1). Countries which reported comprehensive achievement with sustained commitment and capacities at all levels (rank 5) included Australia, Botswana, Cuba, the Czech Republic, Italy, Kenya, Malaysia and Poland. Despite apparent progress, detailed analysis of country submissions indicates great variability in EWS for hydrometeorological hazards and large gaps in coverage. Country submissions were reviewed and any mention of EWS for floods, droughts, famine, fire, hurricanes and cyclones recorded. Thirty-three countries did not provide any information on EWS for hydrometeorological hazards, raising doubts about the existence of such EWS. Six countries clearly did not have EWS, as indicated by the following comments: Lebanon – “The work is still in progress for building the capacity of the early warning systems to detect unusual hazards” Yemen – “There are no effective early warning systems, or a local readiness” Five of the eight countries with self-assigned ranks of 5 did not provide any details of EWS related to hydrometeorological hazards. The Czech Republic and Kenya only mentioned the existence of a flood EWS, while Australia mentioned the existence of both flood and fire EWS. Floods were the most common hazard for which EWS were reported, followed by cyclones/ hurricanes. Few countries reported EWS for drought, fire, famine or heat waves. Famine EWS were only reported in Africa. Europe was the only region not to report any cyclone/hurricane EWS. Two EWS examples from

11

the HFA Progress Reports are highlighted below. Pakistan’s submission (Box 2) was uncommon because specific details about the EWS were provided. Table 2. Hydrometeorological hazards for which early warning systems (EWS) were reported by nation governments in Hyogo Framework for Action (HFA) Progress Review submissions. Region Hazard Cyclone or Hurricane Drought Famine Fire Flood Heat wave No EWS No EWS details provided Grand Total

Africa 4 1 4 2 8 1 2 5 27

Americas 9

Asia

Europe 4 3

1 12

2 8

1 14 37

2 6 25

Grand Total

Oceania 5 2 1 4 1

1 3 1

8 14

12

22 6 4 7 35 2 6 33 115

Box 2. Pakistan’s Flood Early Warning System As described in HFA Progress Report: “Pakistan has good institutional capacities for monitoring and warning of flood hazards. Following the floods of 1992, a comprehensive Indus Forecasting Flood Forecasting Division (FFD), Lahore, which is part of Pakistan Meteorological Department (PMD), undertakes dissemination of flood early warning to national stakeholders through an institutionalized process that connects inputs down to vulnerable communities using multiple channels. Flood forecasting occurs through a four fold input system which includes: • Network of weather radars • Telemetric system which sends real time inputs on water flows • Satellite coverage includes both indigenous capacity and through WMO network • Ground observation through PMD ground station deployed across the country. Among weather radars deployed across the country, the more significant are the Doppler radars that furnish quantified inputs and are deployed in Lahore, Sialkot and Mangla to cover the catchments region. Water and Power Development Authority (WAPDA) has installed telemetry gauges along the rim of rivers in the catchments region and along some major rivers and it monitors water flows in these channels and provides real time information to FFD. Provincial Irrigation Departments also monitor river flows in respective provinces and they also communicate inputs to FFD. Indus Water Commission (IWC) receives flood information from India and its inputs also end up with FFD. FFD (PMD) in Lahore constitutes the nerve centre for flood early warning.”

12

Box 3. Fiji’s Tropical Cyclone (TC) Early Warning System As described in HFA Progress Report: Fiji’s “cyclone EWS is the most developed EWS in the country and warnings are received in a timely manner; the public is aware of what action to take; also curfew has helped to protect people from injuries and heighten security. Records indicate much less injuries and deaths in recent years compared to decades ago e.g. 10 deaths in TC Bebe (Category 4, 1972) and one death TC Tomas (Category 4, Mar 2010).”

Multi-hazard early warning systems were rare. Only 23 out of 88 countries reported warnings for two or more hazards6. It is unclear whether or not these constituted one unified multi-hazard system or simply different EWS for different hazards. As Pakistan reported, “The current early warning capacities encompass only a few hazard risks while institutional capacities need to be developed to cover other major risks such as landslides, drought, forest fires etc. Besides in the absence of an integrated multihazard early warning system, institutional preparedness to make an integrated and multi-hazard response remains far from the desirable levels.” Others, such as Saint Lucia, called for the creation of multi-hazard system, “A comprehensive multi-hazard early warning system needs to be established to address all deficiencies in current systems.” Coverage of EWS was often not uniform within countries. EWS systems were often only functional in specific regions. For example: Mozambique – “Accurate flood early warning system is heavily dependent on hydrological and meteorological gauge stations to provide timely data (localized) on river flow levels and rainfall. So far, the limited territorial coverage of meteorological stations is the major challenge for rapid flood risk assessment for small river basins.” Barbados – “An EWS for floods has been initiated in one of the flood prone communities where the population has been severely impacted. Plans are in train to replicate this in other flood prone communities in support of other flood mitigation options.” Only four countries (Vanuatu, Germany, Panama and the United States) provided information about the timescale of warnings. All of these EWS systems issue warnings on a very short-term timescale. Vanuatu currently issues 3-day outlooks for cyclones. Germany provides a weather-based 3-day prognosis for fire risk. Panama provides a weekly outlook of flood risks. In its’ report to the UNISDR, Germany called for improved prediction tools, “The DWD should receive the necessary financial support to develop medium-term (1 to 2 weeks) fire-danger forecast capabilities.” Funding was the most frequently cited constraint to EWS system development, followed by coordination between local, national and regional actors, lack of human capital and lack of infrastructure (Table 3). For example: Indonesia – “Collaboration with other parties such as the private sector in matters related to media and telecommunication needs to be built. The civil society needs to be empowered to participate in risk information dissemination and the development of community-based EWS.”

6

The following countries reported EWS for two hazards: Australia, Barbados, Columbia, Fiji, Guatemala, India, Indonesia, Jamaica, Japan, Laos, Mexico, New Zealand, Nicaragua, Syria, Thailand, Vanuatu, Zambia. Six countries reported EWS for three hazards: Algeria, Bangladesh, Germany, Mozambique, Tanzania, the United States.

13

Bangladesh – “Bangladesh flood warning information cannot be improved without establishment of regional data sharing and cooperation, considering flooding (and other hazards) as common hazard in the Ganges, Brahmaputra and Meghna basins” Italy – “ The main challenges concerning the future of early warning refer to systems integration. The National warning system provides an extensive coverage of risks, but a number of independent systems and networks exist as well.” Tanzania – “The Early Warning System in the country is inefficient due to lack of enough skilled manpower, equipment, technology and financial resources. These hinder the capacity for accurate and timely collection, process and release of early warning data and information.” Anguilla – “Extremely limited staff and limited technical knowledge of systems outside the director, communications officer and an IT technician. Defective equipment has been highlighted as a concern but budget constraints make improvements difficult at this time.” Some governments mentioned the need for improved predictions: Solomon Islands – “More accurate forecasts on rainfall and their dissemination are needed to improve warning of potential flood risks.” Numerous countries also indicated lack of public understanding about risks. Conversely, where knowledge of risks is common, anxiety or apathy may result: Barbados – “Some of the key hazards, such as earthquakes and tsunamis, are virtually unknown to the general public, thus the effort required to bring the population to an acceptable level of awareness represents a significant challenge.” Marshall Islands – “Again, the absence of severe disasters in recent decades has led to considerable levels of apathy towards the importance of early warning systems.” Japan – “Adverse effect of an overflow of information as highly-advanced information society could lead to excessive social anxiety” Finally, several countries in Oceania mentioned the geographic spread as a major obstacle to EWS development: Solomon Islands – “The location of remote communities limits access in order to reduce challenges relating to warning dissemination.” Marshall Islands – “A further constraint is the isolated nature of outer islands, which makes communication difficult at all times.”

Table 3. Obstacles to EWS development, as reported by national governments for the HFA review Problems Summarized Coordination Funding Region Africa Americas Asia Europe Oceania Grand Total

3 3 1 7

4 7 3 1 15

Human Capital

Prediction Abilities

1 2 2 1 6

Infrastructure

1 1 1 1 2 3

2 5

Public Apathy or Anxiety 1

Public Education

1

1 1

1 3

2

Remoteness

Grand Total 7 14 11 2 10 44

3 3

14

Reports also indicate that local or traditional early warning systems are often used to compliment more formal systems, particularly in low-income countries or in countries with large geographic spread. Guatemala, for example, has conducted work, “in the rescue, recovery and promotion of ancestral knowledge and wisdom management to reduce disaster risk." Zambia reported that non-governmental organizations “are using indigenous knowledge for early warnings at the local level and are using established structures at that level for disseminating such early warning information.” Fiji noted that, “The existing traditional knowledge on early warning signs and disaster preparedness should be documented and shared, particularly to urban dwellers dislocated from traditional settings and most likely their knowledge has been eroded.” However, concerns were raised about the applicability of traditional knowledge in the context of a changing climate, “The relevance and applicability of traditional knowledge in view of changing hazard characteristics due to the impacts of climate change will need to be analysed.” The report from the Solomon Islands concludes, “Challenges remain in terms of getting warnings from Honiara to remote communities in a timely and appropriate manner. People use and often rely on traditional EW information and local knowledge related to hazards and preparedness based on past experience (running to the hills, observing animal behaviour, changes in flora and fauna etc). However, current changes in weather patterns may challenge the use and effectiveness of some traditional knowledge.”

3.4.

EWS by Hazard

The section below moves from a focus on national governments to a focus on EWS for specific hydrometeorological hazards. Between 1988 and 2007, 76% of all disaster events were hydrological, meteorological or climatological in nature, indicating the pressing needs for EWS for droughts, floods, fires and hurricanes (UNISDR 2008). Fortunately, good models of EWS exist for such hazards.

3.4.1. Drought “Drought is the most obstinate and pernicious of the dramatic events that Nature conjures up. It can last longer and extend across larger areas than hurricanes, tornadoes, floods and earth quakes…causing hundreds of millions of dollars in losses, and dashing hopes and dreams.” — US National Drought Policy Commission Report, May 2000 Evidence indicates that the world has become increasingly dry during the past century (UNISDR 2011). General drying trends have occurred in the Sahel and southern Africa, central Brazil, southern Europe, Iran, Indonesia, north-east China and north-east Asia (Trenberth et al. 2007). The IPCC Fourth Assessment Report further noted that the land area under drought at a global scale may increase by the 2090s from 10 per cent in the present day to 30 per cent for extreme drought and 50 per cent for moderate drought, with major implications for both fire and land degradation (Vera et al 2010).

15

Figure 3. Location of drought events between 2000 and 2010 (Source: Global Risk Data Platform). Areas of drought are highlighted in brown.

© PREVIEW 2011, UNEP, UNISDR

Droughts are considered slow, hazards (Grasso 2012). Service (NWS) defines a drought as "a period of abnormally dry weather sufficiently prolonged for the lack of water to cause serious hydrologic imbalance in the affected area." There are at least three different types of droughts. Meteorological drought refers to precipitation deficits. Agricultural drought refers to abnormally low soil moisture and hydrological droughts reflect below-average water levels in lakes, rivers and streams (UNISDR 2009). Unlike other hydrometeorological hazards, drought risk remains poorly understood, without a credible global drought risk model (UNIDSR 2011). Until 2009, when the WMO adopted the Standard Precipitation Index as the global standard to measure droughts, there was no agreed standard measure of meteorological droughts (UNISDR 2011). As a result drought EWS are relatively less developed globally than other EWS (UN 2006).

16

Figure 4. Sequence in drought occurrence and impact of commonly accepted drought types. All droughts originate from a deficiency of precipitation or meteorological drought but other types of drought and impacts cascade from this deficiency (WMO 2006).

Data needs Although all types of droughts originate from a precipitation deficiency, it is insufficient to monitor solely this parameter (WMO 2006). Effective drought early warning systems must integrate precipitation and other climatic parameters with information such as stream flow, snow pack, groundwater levels, reservoir and lake levels, and soil moisture into a comprehensive assessment of current and future water supply conditions (WMO 2006). Examples Local knowledge has been a critical component of EWS for droughts in the past. Very few formal drought early warning systems exist around the world (Grasso 2012), although in recent years a variety of countries have developed national centres or strategies7. In Africa, regional centres such as the Intergovernmental Authority on Development (IGAD) Climate Prediction and Applications Centre (ICPAC) and the Drought Monitoring Centre (DMC) in Harare, provide current data, develop monthly drought monitoring bulletins, and 10-day weather advisories. The World Meteorological Organization (WMO) is currently working to expand DMC to Central Asia. FEWS Net for Eastern Africa, Afghanistan, and Central America reports on current famine conditions, including droughts, by providing monthly bulletins that are accessible on the FEWS Net webpage. For the United States, the U.S. Drought Monitor (USDM) and the National Integrated Drought Information System, offers the best available product for forecasting droughts (Svoboda et al. 2002). USDM is a joint effort between US Department of Agriculture (USDA), NOAA, Climate Prediction Center, and University of Nebraska Lincoln. It has a unique approach 7

These include Australia, South Africa, Canada, the United States, Slovenia, Spain, Portugal, South Korea, China, India, Pakistan, Morocco, Syria, Brazil, Jordan and Iran (Svoboda 2009).

17

that integrates multiple drought indicators with field information and expert input, and provides information through a single easy-to-read map of current drought conditions and short notes on drought forecast conditions. USDM relies on several key indicators and indices such as the Palmer Drought Severity Index (PDSI), the Standardized Precipitation Index, stream flow, vegetation health, soil moisture and impacts, as well as Keetch-Byram Drought Index, reservoir levels, Surface Water Supply Index, river basin snow water equivalent, and pasture and range conditions. In China, the Beijing Climate Centre (BCC) of the China Meteorological Administration (CMA) monitors drought development. Based on precipitation and soil moisture monitoring from an agricultural meteorological station network and remote-sensing-based monitoring from CMA’s National Satellite Meteorological Centre, a drought report and a map on current drought conditions are produced daily and made available on their website. In Europe, the European Commission Joint Research Centre (EC-JRC) provides publicly available droughtrelevant information through the following real-time online maps: daily soil moisture maps of Europe; daily soil moisture anomaly maps of Europe; and daily maps of the forecasted top soil moisture development in Europe (seven-day trend).8 Other Actors While national governments are now working together to monitor drought, global monitoring of impacts is virtually non-existent (Svoboda 2009). In this area, a variety of agencies may play a role in early warning development (UN 2006): WMO - Droughts often result from El Niño Southern Oscillation (ENSO) conditions. The World Meteorological Organization (WMO) coordinates the development of a global scientific consensus, which results in an El Niño Update, a unified global statement on the expected evolution of ENSO for months ahead. FAO – The Food and Agricultural Organization (FAO) gathers information on factors that influence planted areas and yields, as well as social, economic and market data at the international, national and sub-national levels. The FAO-GIEWS provides information on countries facing food insecurity through monthly briefing reports on crop prospects and food situation, including drought information, together with an interactive map of countries in crisis, available through the FAO website. UNCCD – The United Nations Convention to Combat Desertification seeks, "to forge a global partnership to reverse and prevent desertification/land degradation and to mitigate the effects of drought in affected areas in order to support poverty reduction and environmental sustainability". UNEP/ UNISDR – The United Nations Environment Programme (UNEP) and the United Nations International Strategy for Disaster Reduction (UNISDR) support national and regional drought monitoring centres. Both have collaborated to produce PREVIEW Global Risk Data Platform, a website which shares spatial data on natural hazards including droughts. HEWS – The Humanitarian Early Warning Service (HEWS) by the World Food Programme (WFP) and Benfield Hazard Research Center of the University College London collects drought status information from several sources including FAO-GIEWS, WFP, and Famine Early Warning System (FEWSNET), and packages this information into short notes and a map (from FAO-GIEWS) which is then provided, on a monthly basis, through the HEWS website. 8

This section is based on Grasso (2012), which reflects the results of a gap analysis of EWS for each hazard type. To analyse geographical coverage, Grasso compared risk maps for specific hazards (from Natural Disaster Hotspots: A Global Risk Analysis, a report from the World Bank by Dilley M. et al., 2005) against coverage of existing EWS for that specific hazard. Afterwards, each EWS was analyzed to see if each system had covered all the aspects that are essential for an EWS. This analysis helped to identify major geographical and functional/operational gaps of existing EWS.

18

Steps Forward Given the increasing intensity and frequency of droughts, there is an urgent need to expand drought early warning systems (EWS). For large parts of the world suffering from severe droughts, EWS are not yet in place, including: parts Spain, parts of France, southern Sweden, and northern Poland, India, parts of Thailand, Turkey, Iran, Iraq, eastern China, areas of Ecuador, Colombia, and the south-eastern and western parts of Australia (Grasso 2012, also see section 3.2). The US Drought Monitor could be used as a model for other countries to follow (WMO 2006). The creation of regional networks should also be encouraged, as illustrated in Figure 5. Figure 5. Potential Drought EWS regional networks, as identified by Svoboda (2009)

Although all types of droughts originate from a precipitation deficiency, monitoring only this parameter is insufficient to assess a drought’s severity and resultant impacts. Drought monitoring systems should be integrated, coupling multiple climate, water and soil parameters and socio-economic indicators to fully characterize drought magnitude, spatial extent and potential impact (WMO 2006). However, a variety of obstacles remain and will need to be addressed in order to establish effective EWS (WMO 2006): 1. Meteorological and hydrological data networks are often inadequate in terms of the density of stations for all major climate and water supply parameters. Data quality is also a problem because of missing data or inadequate long-term records. 2. Data sharing is inadequate between government agencies and research institutions, and the high cost of obtaining data limits their application in drought monitoring, preparedness mitigation and response. 3. Information delivered through early warning systems is often too technical and detailed, limiting its use by decision makers. Delivery systems are often not well developed. 4. Forecasts are often unreliable on the seasonal timescale and lack specificity, reducing their usefulness for agriculture and other sectors. Drought indices are sometimes inadequate for

19

detecting the early onset and end of drought. There is a pressing need to recommend standard indices of droughts. 6. Impact assessment methodologies, a critical part of drought monitoring and early warning systems, are not standardized or widely available, hindering impact estimates and the creation of regionally appropriate mitigation and response programmes.

3.4.2. Famine “We live in a world with widespread hunger and undernourishment and frequent famines. It is often assumed…that we can do little to remedy these desperate situations…There is little factual basis for such pessimism…” - Amartya Sen As extreme weather conditions can contribute to mass starvation, famine EWS systems are included in this report. For example, starvation in the Bangladesh famine of 1974 was initiated by regional unemployment and income deprivation caused by floods (Sen 1999). While the number of hungry people has declined around the world, undernourishment remains unacceptably high. Nine hundred twenty five million people do not have enough to eat and 98 percent of them live in developing countries (FAO 2010). Two-thirds live in just seven countries (Bangladesh, China, the Democratic Republic of the Congo, Ethiopia, India, Indonesia and Pakistan) and over 40 percent live in China and India (FAO 2010). Hunger however is not equivalent to famine. For the UN to officially declare a famine, three important conditions must be met. First, 20 per cent of the population must have fewer than 2100 kilocalories of food available per day. Secondly, more than 30 per cent of children must be acutely malnourished. And finally, two deaths per day in every 10,000 people, or four deaths per day in every 10,000 children, must be the result of lack of food (Prucell 2011). Famine is caused not only by factors affecting food production, but also by factors related to the functioning of the entire economy and the operation of the political and social systems that influence people’s ability to acquire food (Sen 1999). As Sen (1999) explains, “What is critical in analyzing hunger is the substantive freedom of the individual and the family to establish ownership over an adequate amount of food, which can be done either by growing the food oneself…or buying it in the market… A person may be forced into starvation even when there is plenty of food around if he loses his ability to buy food in the market through a loss of income.” Data Needs Data needs for a famine EWS are quite complex. While weather data, such as precipitation, is important for famine EWS, snow pack conditions and satellite vegetation index data are also needed. Normalized Difference Vegetation Index (NDVI) from AVHRR satellites has a long history of use in famine early warning. More recently, NDVI from MODIS satellites has proved better suited for the characterization of crop conditions. EWS must also incorporate socio-economic data. In many cases famine results from increased vulnerability rather than the severity of climate hazards. While socio-economic data exists, it is generally not known or used within climate communities (WMO 2011b). Required data include information on markets and trade, as well as health and nutrition. Household livelihoods also need to be assessed through the creation of livelihood zone maps (geographic areas in which people share the same patterns of access to food, income, and markets), livelihood profiles (description of the food and income sources and market access of wealth groups and the hazards to which they are vulnerable) and livelihoods baselines (detailed, quantified breakdowns of household livelihood options that permit assessment of ability to meet basic survival requirements and protect livelihoods). Examples FEWSNET is perhaps the best-known famine EWS. It is a collaborative effort of the United States Geological Survey (USGS), United States Agency for International Development (USAID), NASA, and

20

NOAA. FEWSNET can be described as a food security network that has developed its own “climate service”. The programme attempts to determine how many people are vulnerable to famine, where, for how long and if intervention is needed. FEWSNET focuses on Africa, Central America, Haiti and Afghanistan, and works through a variety of local partners. Local social scientists participate in food security and livelihoods assessments, while regional physical scientist integrate results from climate modelling. Chemonics International assists with vulnerability analysis. Satellite models are used to fill in gaps where no local monitoring is possible. FEWSNET produces special reports and food security alerts, a monthly Food Assistance Outlook and a monthly Price Watch (Figure 6). Other famine early warning systems include GIEWS and HEWS (described in following sections). The Food Security Nutrition Working Group (FSNWG), a collaboration between NGOs, UN agencies, and the Red Cross/ Crescent Movement, the Integrated Food Security Phase Classification, also work the field of famine warning primarily in East Africa (Ververs 2012).

Figure 6: Food Security map produced by The Famine Early Warning System (FEWSNET) for the Horn of Africa.

A recent report by Ververs (2012) cites massive failures of famine EWS during the 2011 famine in the Horn of Africa. The United Nations declared famine in Somalia on July 20, 2011. Nearly 13 million people were in need of humanitarian assistance by September 2011 as a result of widespread drought (FEWSNET 2011). The drying trend in East Africa has been linked to changing sea surface temperatures (Williams and Funk 2011). Scientists have long warned of possible drought and famine (UNEP 2011). However, three out of five early warning systems failed to provide timely warnings. Only FEWSNET and the FSNWG were sufficiently adequate to timely predict (at least 6 months ahead) the food security crisis in East Africa as a region. It is unclear why GIEWS did not provide timely warnings

21

about the upcoming humanitarian disasters (Ververs 2012). It appears that the system needs improvement. Other Actors Additional global actors working in famine EWS include: FAO - The FAO supports the Global Information and Early Warning System on Food and Agriculture (GIEWS). GIEWS produces special reports, often after food security assessment missions, a biannual Food Outlook, a monthly Global Food Price Monitor and a trimestrial Food Situation report. GIEWS also promotes collaboration and data exchange with other organizations and governments. As noted by Ververs (2012) the frequency of briefs and reports, which are released monthly or bimonthly,may not be adequate for early warning purposes. WFP – The World Food Programme is also involved in disseminating reports and news on famine crises through its web-service and the Humanitarian Early Warning Service (HEWS). HEWS collects drought status information from several sources including GIEWS and FEWSNET and packages this information into short notes and a map which is then provided, on a monthly basis, through the HEWS website. A seasonal hazards calendar is also produced, indicating food security-related hazards (Ververs 2012). Steps Forward FEWSNET can act as a successful model for famine EWS and should be expanded further. To expand coverage, data collection needs further improvement. Precipitation observations, often collected using the WMO GTS network, are too sparse. Many food insecure regions, such as pastoral areas, lack data altogether. Existing stations also need to increase the frequency of reporting. Forty percent of the 1232 GTS rainfall stations did not report precipitation during 2008. Additional snow telemetry (SNOTEL) stations are needed to monitor snow melt runoff. There are fewer SNOTEL stations in Canada and Asia than in the Western US. Furthermore, agencies that perform analyses on a national basis (such as WFP and United Nations High Commissioner for Refugees) should add stronger regional analysis. A national approach is likely too myopic and thus misses the implications of a disaster at the sub regional and regional levels (Ververs 2012).

3.4.3. Fire "All you see is bricks, burnt-out timber, twisted iron. And I know for those families when they do return, it's going to be absolutely heartbreaking." - Western Australia Premier Colin Barnett describing 2011 bush fires

The global increase in wildfires following the 1997-98 El Niño event emphasized the urgent need for more accurate and timely information about fire hazards. Approximately 80% of global fire occurs in grasslands and savannas, primarily in Africa and Australia, but also in South Asia and South America; the remaining 20% occurs in the world’s forests (Goldammer and de Groot 2011). From a global perspective, fires are generally absent pole ward of 70oN and 70oS, progressively more frequent towards the tropics, but dropping sharply at the equator (Mouillot and Field 2005). However anthropogenic fires occur everywhere (Figure 7), and are increasing in the tropics, driven largely by deforestation and agricultural development in South America and South-east Asia (Flannigan et al. 2009). Northern and southern hemispheres show distinct fire seasons. In addition to strong annual periodicity, there is also a substantial inter-annual variation in both hemispheres that follows a 4-year pattern and correlates well with El Niño–Southern Oscillation events (Flannigan et al. 2011). Global climate change will influence the fire patterns. Numerous studies suggest that temperature is the most important variable affecting wildland fire, with warmer temperatures leading to increased fire activity. With increasing

22

temperatures, ground moisture will decrease and lightning strikes, a common ignition source, will increase. For example, a two-degree change in temperature is expected to increase lightning strikes in the tropics by 72%. In a warmer world, fire seasons are expected to lengthen in temperate and boreal regions (Flannigan et al. 2009). Changes in fire intensity and severity are more difficult to summarize. Figure 7 Average number of high temperature events (mostly Fires) as detected by ATSR World Fire Atlas (Sources: Global Risk Data Platform). Wildland fires are cleary a hazard that affects nearly all regions of the globe.

© PREVIEW 2011, UNEP, UNISDR Data Needs Wildland fire early warning involves use of fire danger rating to identify in advance critical time periods of extreme fire danger. Fire danger rating (FDR) is the systematic assessment of fire risk and potential impact, and it is the cornerstone of contemporary fire management programs. It is used to determine suppression resource levels and mobilization (fire fighters, equipment, helicopters, air tankers), to define safe and acceptable prescribed burn prescription criteria, and to establish fire management budgets. Fire activity is strongly influenced by four factors: fuels, climate weather, ignition agents and people. In order to assess fire risk, data is needed on these four factors (Flannigan et al. 2005). Information necessary for modelling includes topography and data on vegetation cover, meteorological variables such as 3-hour cumulated rainfall, air temperature, dew point temperature, and wind speed/direction. Data obtained from satellites (MODIS products) such as Relative Greenness, Moisture Stress Index, and Normalized Difference Snow Index, are used to assess the state of live and dead fuels. Some indexes have lower data needs however. For example, the daily severity rating (DSR) index, part of the Canadian forest fire weather index (FWI) system used to reflect the amount of effort required to suppress a fire, is derived only from measurements of precipitation, air temperature, humidity, and wind. Examples Most industrial countries have EW capabilities in place, while most developing countries have neither fire early warning nor monitoring systems (Goldammer et al. 2003). Canada has one of the best fire monitoring and warning systems. It is composed of the Canadian Forest Fire Danger Rating System (CFFDRS), the Canadian Forest Fire Weather Index (FWI) System, the Canadian Forest Fire Behaviour

23

Prediction (FBP) System, Fire Monitoring, Mapping, and Modelling (Fire M3), Forecast Fire Weather, Atmospheric Dispersion Index (ADI) and Its Component. Each system has subsystems. For example, the Canadian Forest Fire Weather Index (FWI) System consists of six components that account for the effects of fuel moisture and wind on fire behaviour. The first three components of FWI are fuel moisture codes: a numeric rating of the moisture content of litter and other fine fuels, the average moisture content of loosely compacted organic layers of moderate depth, and the average moisture content of deep, compact organic layers. The remaining three components of FWI are fire behaviour indices, representing the rate of fire spread, the fuel available for combustion, and the frontal fire intensity. Their values rise as the fire danger increases. Australia also has a well-developed fire rating system based on empirical data. However, Australia’s models are based predominantly on conditions specific to the country. In the United States, NOAA has developed a fire warning system, using theoretical, rather than empirical models. Since 2008, Brazil has offered an interactive mapping service based on Google maps and EO imagery from INPE the Brazilian Space Research Institute. Individuals can contribute with information from the ground. In only 3 months the service has received 41 million reports on forest fires and illegal logging, making it one of the most successful web sites in Brazil (Grasso 2012). Italy, Alabania, Spain, Lebanon, Japan, Mexico and South Africa also have fire monitoring systems in place (Grasso 2012). Globally, wildland fire information is available through the Global Fire Monitoring Center (GFMC). GFMC provides a global portal for wildland fire data products, information, and monitoring. Fire products include: fire danger maps, near real-time fire events information, an archive of global fire information, and assistance and support in the case of a fire emergency. GFMC also is involved in the Global Early Warning System for Wildland Fires9 which can produce up to a two week warning of fire weather conditions. The Webfire Mapper, a collaboration between the University of Maryland and NASA, provides near real-time information on active fires worldwide, detected by MODIS rapid response system. Webfire Mapper integrates satellite data with GIS technologies for active fire information. This information is available to the public through the website and email alerts.

9

See http://www.fire.uni-freiburg.de/gwfews/index.html

24

Figure 8. Example of Global Fire Map. Each map accumulates the locations of the fires detected by MODIS on board the Terra and Aqua satellites over a 10-day period. Each coloured dot indicates a location where MODIS detected at least one fire during the compositing period. Colour ranges from red where the fire count is low to yellow where number of fires is large.

Other Actors In addition to the organizations listed above, key global actors that can contribute to fire EWS include: WMO – the World Meterological Organization helps supply data and analysis necessary for fire EWS. FAO – the Food and Agricultural Organization helps support the Global Observation of Forest Cover and Land Dynamics (GOFC-GOLD), a programme for observations, modelling, and analysis of terrestrial ecosystems to support sustainable development. One of the primary goals of GOFC-GOLD is to establish an operational network of fire validation sites and protocols, providing accuracy assessment for operational products and a test bed for new or enhanced products, leading to standard products of known accuracy. Activities of the fire program include: 1) availability of observations, 2) harmonization and standardization, 3) validation, 4) adequacy and advocacy of products, 5) regional networks and capacity building, 6) shared data, information and knowledge. UNEP – The United Nations Environment Programme supports GOFC-GOLD through the Global Terrestrial Observing System (GTOS). Steps Forward An internationally standardized approach is required to create a globally comprehensive fire early warning system. Great strides can be made by sharing information with developing countries, and by teaching them how to implement fire EWS. The Global Early Warning System for Wildland Fire offers a model that should be supported and expanded further. The distribution of warnings produced by this system should be improved. For example, information must reach the village level. Warnings could be communicated through means other than the Internet, such as TVs or radios. The timescale of current predictions should be extended and seasonal fire forecasts created for all regions. It may even be possible to create multi year (1 to 4 year) forecasts. However, in order to do so, further validation of

25

models is required. Current fire models, such as the Canadian Terrestrial Ecosystem Model (CTEM) DGVM, which estimates the probability of fire occurrence depending on fuel availability, fuel moisture and presence of an ignition source, have been validated only for a handful of locations. Substantial work will be required to validate models on a global scale (Pechony and Shindell 2009). In order to produce better estimates of fire risk, comprehensive global socio-economic data on sources of anthropogenic ignitions, fire suppression policies and resources, and degrees of fire management are also needed.

3.4.4. Floods "The floodwaters have devastated towns and village, downed power and communications lines, washed away bridges and roads and inflicted major damage to buildings and houses." - UN Humanitarian Chief John Holmes describing 2010 floods in Pakistan The 2010 flooding in the Indus River basin resulted in more than 1,600 people dead and disrupting the lives of about 14 million people (Fakhruddin 2012). Flood disasters have been increasing over the last 30 years, with over 182 incidents worldwide in 2010 alone (EM-DAT database, 2010). Severe flooding affects more countries than tropical cyclones (UNISDR 2009). Floods have far-reaching socio-economic and environmental implications, including loss of life, loss of property, mass migration, environmental degradation and shortages of food, energy and water (WMO 2009). While floods occur around the world, flood mortality risk is heavily concentrated in Asia, particularly India, Bangladesh and China (UNISDR 2009). GDP exposure is also heavily concentrated in Asia (UNISDR 2009). Figure 9. Flood events between 2000 and 2010 (Source: Global Risk Data Platform). Areas of flooding are highlighted in blue.

© PREVIEW 2011, UNEP, UNISDR Data Needs Flood prediction is notoriously difficult. It depends not only on the amount and intensity of precipitation, but also on the characteristics of the river catchment area. Some rivers respond very quickly to increases

26

in water levels while others do not (UK Environmental agency 2012). Traditionally data from weather radar and rain gauges have been used to monitor rainfall. However flood warnings issued based on observations of upstream river gauges offer limited reaction time. Longer lead times rely on rainfall-runoff models, which depend not only on the provision of accurate rainfall information, but also on the correct reproduction of the state and dynamic behaviour of the hydrological system (Schroter et al. 2009). Routing models route the flow from upstream to downstream river monitoring stations, incorporating inflows from other tributary streams. Hydrodynamic models rely on detailed topographic survey data and use standard model software packages to forecast levels at set locations (UK Environment Agency 2012). In summary, accurate flood forecasting depends on access to weather data (meteorological information), data on the river basin (geographical and hydrological information) and data on the socioeconomic conditions (Figure 10). Figure 10. Flood forecast methodology used by RIMES to model floods in Bangladesh. Long lead flood forecast technology involves two distinct hydrologic modelling approaches i) data-based modelling, and ii) distributed modelling. A multi-model ensemble approach is taken on a daily basis for each forecast lead-time. This involves calculation of historic simulated discharges of each model (data-based and distributed) separately, using observed weather variables (precipitation, wind-speed, etc) as inputs (i.e. not using forecast data) (Subbiah and Fakhruddin 2012).

Examples Some countries monitor flash floods through their national meteorological services, while others monitor flash floods and river floods through environmental agencies and hydrology services separately. According to the UN (2006), dedicated systems to monitor and forecast river basin floods are well established in developed countries. The UK has one of the leading flood forecast systems, based not only on meteorological data but also on catchment areas. NOAA provides observed hydrologic conditions of US major river basins and predicted values of precipitation. The Hydrometeorological Prediction Service produces a five-day River Flood Outlook based on future rainfall forecasts (Figure 11). Other initiatives that use ensemble flood prediction systems include Georgia-Tech/Bangladesh project, Finnish Hydrological Service , Swedish Hydro-Meteorological Service, MAP D-PHASE (Alpine

27

region) / Switzerland, Vituki (Hungary), Rijkswaterstaat (The Netherlands), Royal Meteorological Institute of Belgium, Météo France, Land Bayern (Germany), CSIR (South Africa)10. Deltares is an independent, institute for applied research in the field of water, which is also active in flood EWS. Between 2001 and 2005, Deltares contributed to the creation of a flood EWS along four river basins in Taiwan. Stand-alone warning systems also exist in Guatemala, Honduras, El Salvador, Nicaragua, Zimbabwe, South Africa, Belize and Bangladesh. However, gap analysis by Grasso (2012) shows inadequate coverage of flood warning and monitoring systems in many developing or least developed countries, including China, India, Nepal, and Brazil.

Figure 11. Flood outlook produced by US Hydrometeorological Center.

Most flood warning systems are stand-alone national operations, but warning systems have been developed covering several international rivers, such as those for the Rhine, Danube, Elbe and Mosel in Europe, the Mekong, Indus and Ganges- Brahmaputra-Meghna basins in Asia and the Zambezi in Southern Africa (UN 2006). Regionally, the European Flood Alert System (EFAS), which is under development by EC- JRC, provides early flood warnings to National Hydrological Services in order to mitigate flood impact on population. EFAS testing is being performed for the Danube river basin, focusing on the system’s calibration and validation. EFAS provides national institutes with information on possible fiver flooding within a 3 day timescale (Grasso 2012). The Regional Integrated Multi-Hazard Early Warning System for Africa and Asia (RIMES) is an international and intergovernmental institution established for the generation and application of early warning information. Thirteen member states participate in RIMES along with an additional sixteen collaborating countries. RIMES has developed long lead flood forecasts (1-15 days) for riverine flooding, and has demonstrated the system successfully in Bangladesh (Subbiah and Fakhruddin 2012). RIMES also issues warnings 24-72 hours in advance of flash floods. Globally, Floods are monitored worldwide from the Dartmouth flood observatory. Based at the University of Colorado the observatory has a mission to acquire, publish, and preserve for public access a digital 10

Additional information can be found here: http://www.floodrisk.net/

28

map record of the Earth’s changing surface water, including changes related to floods and droughts. Its products are based on various remote sensing sources. Satellite microwave sensors can monitor, at a global scale and on a daily basis, increases of floodplain water surface without cloud interference. The Dartmouth flood observatory provides estimated discharge and satellite images of major floods worldwide but does not provide forecasts of flood conditions or precipitation amounts that could allow flood warnings to be issued days in advance of events. IFnet Global Flood Alert System (GFAS) uses global satellite precipitation estimates for flood forecasting and warning. The GFAS website publishes useful public information for flood forecasting and warning, such as precipitation probability estimates, but the system is currently running on a trial basis (Grasso 2012). Other Actors Additional global actors in this area include: UNESCO– UNESCO currently coordinate an operational flood warning system with WMO and National Meterological Services for river flooding. The International Flood Initiative/Programme (IFI/P), launched during the World Conference on Disaster Reduction in January 2005, is a joint programme of UNESCO and WMO to be operated by the International Centre on Water Hazard and Risk Management (ICHARM), which is hosted by the Public Works Research Institute (PWRI) in Japan. The International Flood Network, through the Global Flood Alert System, provides information on precipitation based on satellite data to global subscribers for free. Such initiatives enhance the services provided by national authorities. WMO – The World Meteorological Organization’s Hydrology and Water Resources Programme works with countries to enhance flood forecasting capacity and the short and long-term Global Flood Forecasting Project. The Associated Programme on Flood Management has been developing strategies for effective community preparedness. Steps Forward

At a global scale flood monitoring systems are more developed than flood early warning systems (Grasso 2012). The 2010-2011 floods in Pakistan demonstrate the need for reliable and improved longer-term outlooks. In order to generate longer-term outlooks, both data collection (e.g. increased placement of rainfall gauges) and modelling needs to be improved. Larger and larger data sets are now being used for increasingly advanced modelling. IBM, in collaboration with Willis Ltd, the UK Met Office, Esri, and Deltares is working to create a Global Flood Model (GFM). The GFM will assess flood risks and devise long-term mitigation strategies such as land use changes and infrastructure improvements (UNISDR 2012). This model, due to be launched in March 2012, will "comprise an integrated set of modules, each composed of models and data. For each module there are two core elements: a specification, and a live 'reference version' (a worked example). Users will be able to work with the reference version, or substitute their own models and data." Each model will be self-contained but also integrated into a complete “stack” covering causal scope. If successful, perhaps this model could be used internationally.

29

Figure 12. Information used for proposed Global Flood Model

3.4.5. Cyclones/Hurricanes “The tragedy of Hurricane Katrina is in some small way mitigated by the fact that we now have more people talking about it, thinking about it and working on it, so that we will be more vigilant and ready.” - John Hickenlooper, American politician Tropical cyclones, also known as hurricanes or typhoons, are amongst the most powerful and destructive meteorological systems on earth. Globally, 80 to 100 cyclones develop over tropical oceans each year. Many of these make landfall and can cause considerable damage to property and loss of life (UNISDR 2009, Met Office 2012). Tropical cyclones form between approximately 5° and 30° latitude and initially move westward (owing to easterly winds) and slightly towards the poles (Met Office 2012). There are seven tropical cyclone "basins" where storms occur on a regular basis: 1) Atlantic basin (including the North Atlantic Ocean, the Gulf of Mexico, and the Caribbean Sea); 2) Northeast Pacific basin (from Mexico to about the dateline); 3) Northwest Pacific basin (from the dateline to Asia including the South China Sea); 4) North Indian basin (including the Bay of Bengal and the Arabian Sea); 5) Southwest Indian basin (from Africa to about 100°E); 6) Southeast Indian/Australian basin (100°E to 142°E); 7) Australian/Southwest Pacific basin (142°E to about 120°W) (NOAA 2012). With climate change, global frequency of cyclones is likely to decrease or remain unchanged, but maximum wind speed is predicted to increase (Knutson et al. 2010). In fact, Bender et al. (2010) predict nearly a doubling of the frequency of category 4 and 5 storms by the end of the 21st century, despite a decrease in the overall frequency of tropical cyclones. The largest increase is projected to occur in the Western Atlantic, north of 20°N.

30

Figure 13. Location of tropical cyclones between 1970 and 2010 (Source: Global Risk Data Platform). Wind intensity and tracks are indicated from light to dark green, with dark green indicating a 5 SS Cat.

© PREVIEW 2011, UNEP, UNISDR Between 1970 and 2009 Tropical cyclones claimed around 789,000 lives (Peduzzi et al. 2012). In 2010 an estimated 1.53 billion people were living in cyclone prone areas in 81 countries and territories (Peduzzi et al. 2012). Geographically, tropical cyclone mortality is concentrated in Bangladesh and India (UNISDR 2009). However economic losses (in absolute terms) are concentrated in Japan, the United States and Australia (UNISDR 2011). A recent study of tropical cyclone risk indicates that coastal populations and remote rural populations are particularly vulnerable to loss from high category cyclones, while poverty levels are significant predictor of loss when facing lower intensity hurricanes (Peduzzi et al. 2012). Data Needs Over the past 50 years, significant correlation has existed between Atlantic tropical cyclone power dissipation and Sea Surface Temperatures (SST) (Knutson et al. 2010). Data on SST is critical for tropical cyclone prediction, particularly data on SST above 27 oC. Information on wind activity is also needed. Until the mid-1940s, tropical cyclone observations were limited to ships at sea coastal weather stations. Since the mid-1940s, aircraft reconnaissance has allowed a more accurate data collection. However a major change in the ability to monitor tropical cyclone intensities occurred with the in the Atlantic basin, about 70 percent of the monitoring of tropical cyclones is currently done via satellite methods (Knutson et al. 2010). Examples Tropical cyclones are monitored and forecasted globally by centres affiliated with the WMO (see next section). At the national level a variety of meteorological agencies have well developed tropical cyclone monitoring and early warning systems. The UK Met Office produces dynamical seasonal prediction of tropical storms forming over the North Atlantic using the Met Office (GloSea) and ECMWF models. In Cuba, the Institute of Meteorology is responsible for hurricane prediction and monitoring, using a hurricane EWS is considered effective due to high network of 120 weather levels of public awareness, public policy commitment and a highly centralized government. Broadcasting

31

radio and TV networks, for example, are fully controlled by the government and can quickly be used to disseminate warnings (Naranjo 2005). However, the United States NOAA has the most sophisticated tropical cyclone warning system. Once a tropical depression has been identified, a series of forecast advisories, which detail the expected track and likely strength of the tropical cyclone, are broadcast. When there are definite indications that a tropical cyclone is approaching land, watches and warnings along coastal regions are raised, which aim to give information to the local authorities of places likely to be in the tropical cyclone's path, so that they can make preparations to protect public safety. In addition to warnings on a 2 -5 day timescale, NOAA issues seasonal forecasts of hurricanes. Other Actors Other actors involved in cyclone and hurricane EWS include: TSR - The Tropical Storm Risk (TSR) is a consortium of experts on insurance, risk management and seasonal climate forecasting. TSR is a leading resource for predicting and mapping tropical storm activity worldwide. The public TSR website provides forecasts, up to 120 hours in advance, and information to benefit basic risk awareness and decision making from tropical storms. The new TSR Business service and website offers real-time products for the detailed mapping and prediction of tropical storm impacts worldwide. TSR also computes seasonal probabilistic forecasts of basin and land falling tropical cyclones worldwide. Forecasts are updated monthly to provide skilful outlooks for assessing the likelihood of upcoming damage and disruption (UN 2006). UNEP and UNISDR – UNEP and UNISDR have contributed to the development of methodologies and vulnerability assessments activities related to cyclones. Data on tropical cyclones wind speed profiles can be downloaded from PREVIEW. UNEP has also prepared physical exposure, risk and vulnerability assessment for the UNDP report “Reducing Disaster Risk: A Challenge for Development”. WMO - The World Meteorological Organization has developed the Global Tropical Cyclone Warning System and Tropical Cyclone Programme (TCP). This is a global network for observations, data exchange and regional forecasting and analysis capabilities, and includes six Regional Specialized Meteorological Centres (Nadi, New Dehli, Miami National Hurricane Centre, La Reunion, Honolulu, Tokyo Typhoon Centre) that provide around-the clock forecasts, alerts and bulletins on the severity, project path and estimated landfall to the National Meteorological Services of countries at risk11. TCP seeks to promote and coordinate efforts to mitigate risks associated with tropical cyclones. Regional bodies worldwide have adopted standardized WMO-TCP operational plans and manuals, which promote internationally accepted procedures in terms of units, terminology, data and information exchange, operational procedures, and telecommunication of cyclone information. Steps Forward Although comprehensive coverage of early warning systems for tropical cyclones is available, recent disasters such as Hurricane Katrina of 2005 have highlighted inadequacies in early warning system technologies for enabling effective and timely emergency response. There is a pressing need to improve communication between the sectors involved by strengthening the links between scientific research, 11

Tropical Cyclones Centres (TCC) include the National Hurricane Center, Miami (Caribbean Sea, Gulf of Mexico, North Atlantic and eastern North Pacific oceans east of 140°W), Japan Meteorological Agency, Tokyo (Western North Pacific Ocean from Malay peninsula to 180°E), the Indian Meteorological Department, New Delhi (Bay of Bengal and the Arabian Sea), the Central Pacific Hurricane Center, Honolulu, Hawaii (North Pacific Ocean 140 – 180°W), Météo France de La Réunion (South Indian Ocean from African coast to 90°E), Meteorological Service, Nadi, Fiji (South Pacific Ocean east of 160°E and north of 25°S).

32

organizations responsible for issuing warnings, and authorities in charge of responding to these warnings. National capacities in most developing countries need improvements in order to effectively issue and manage early warnings (Grasso 2012). Improvements can be made to observing systems. Knutson et al. (2010) highlight the need for, “a comprehensive analysis ….to determine the best mix of tropical cyclone observations in support of climate studies, forecasting, and other needs. For example, is a resumption or initiation of manned or unmanned aircraft reconnaissance in various basins now justifiable in terms of costs, benefits, and alternative measurement techniques?...Another promising, lowcost technique is a tethered blimp that was successfully deployed into the eye of a severe tropical cyclone for several days. This potentially could provide continuous central pressure measurements for tropical cyclones around the world for days at a time.” With greater observations, perhaps the timescale of warnings could also be extended to the seasonal level.

3.5.

EWS Evaluation

Evaluations of early warning systems (EWS) conducted during the last decade have consistently reached the same conclusion: EWS are still far from providing the coverage and scope that are needed and technically feasible (UN 2006, Grasso 2012). In 2006 a UN survey of EWS found gaps in each of the four components of EWS: 1. Risk Knowledge – There is inadequate emphasis on social, economic and environmental vulnerability, data gaps, difficulties accessing data where it is collected, lack of early warning indicators. 2. Monitoring and Warning – There are many gaps in observing systems for hydrometeorological hazards. There is a lack of systems for flash floods and storm surges, particularly in developing countries. There is insufficient multi-disciplinary or multi -agency coordination. 3. Dissemination – Warning systems are limited in many developing countries with no formal structures to issue warnings. Failure to take action may also result from political considerations. Warnings are often unclear with varying standards of language. “There is a need for a single, consistent, easily understandable, global nomenclature to be used as a standard.” There is also a need for the development of standards, protocols and procedures for exchange of data, bulletins and alerts. The proliferation of communication technologies has resulted in the loss of a single authoritative voice. Furthermore, warnings often are not targeted at users. 4. Response capacity – Response is hindered by lack of multi-agency collaboration and clarity of roles, as well as by lack of public awareness about response plans. Few places practice simulation exercises or evacuation drills. There is a need for greater participation and the inclusion of traditional knowledge, and there needs to be greater focus on long term risk reduction strategies. This review finds that while progress has been made, many of the problems identified in 2006, specifically related to risk knowledge and monitoring, remain. Most EWS systems only deal with one climate-related hazard, such as heat waves or droughts, and only cover a limited geographic range. Large parts of the most vulnerable regions of the world are not included. As concluded in the 2006 UN report, “Developed countries and disaster exposed areas of the developing world operate more hazard forecasting and observing systems than African countries and other developing countries with less disaster exposure.” In the 2009-2011 HFA Progress Review, national governments mentioned funding problems, lack of human capital and infrastructure as major obstacles to the development of EWS. Even where the capability exists to reliably generate warnings,

33

greater coordination is needed between the local, national and regional scale, and between actors in the private and public sector. Warnings are often issued on an inadequate time scale (a few hours or days) and fail to reach those who must take action (UN 2006). Seasonal or decadal warnings are needed to mobilize resources or change behaviours. In fact, the decadal timescale is widely recognised as the key planning horizon for governments, businesses and other societal entities (WMO 2011). However, the extent to which predictability can be harnessed on decadal timescales remains uncertain and is the subject of the next section. Data Gaps The World Meterological Organization (2011) recently acknowledged that climate services, including early warning systems, are fundamentally limited by the availability of observations and analyses. While the WMO runs a Global Climate Observing System, at any given time there are approximately 100 silent surface stations and 10 to 15 silent upper air stations. Most data gaps occur in the developing world (WMO 2011). Even where data is collected, limited spatial coverage, coarse spatial resolution, short period of record, cost, and copyright restrictions, all constitute additional obstacles frequently encountered (WMO 2011b). In order to better predict hazards, there is a need for gridded datasets. Gridded datasets generally have the advantage of providing complete spatial coverage and complete temporal coverage (Jones and Mason 2012). Needed gridded data products for operational climate monitoring include: precipitation, temperature, evapotransporation, snow water equivalent, soil moisture, vegetation indices (WMO 2011b). Data should have the following characteristics: daily time step (or better); spatial resolution ≤ 5 km, global coverage, availability in near real time (next day), long, homogeneous historical time-series, availability free of charge, no IP encumbrances (public domain). Investments in knowledge and information, including observational and monitoring systems, will allow agencies to better assess risks and ensure that response strategies are adequate (IPCC 2012). Fortunately, the WMO is currently attempting to improve data collection through the Climate Services Framework. This process should help fill data gaps. However additional socio-economic information is also needed on human vulnerability, human development indicators, distribution of poverty, and livelihood sources (IPCC 2012). EWS Gaps Early warning systems technologies appear to be mature in certain fields but not yet in others (Grasso 2012). In the 2009-2011 HFA Progress Review, national governments mentioned the existence of EWS for floods most frequently, followed by cyclones and hurricanes. EWS for fires, droughts or famines were rarely reported. However, HFA statistics are based on self-assessments and are unverified. Some countries did not list EWS that are known to exist. Canada, for example, did not mention the existence of fire EWS, even though it is a leader in this area. Furthermore, the quality of EWS is unknown. Countries may report flood EWS, but these may only be based on precipitation levels from rainfall gauges rather than on ensemble models that take into account catchment areas. The format of warnings is also unclear. Many EWS appear to produce maps. However, these may not reach all sectors. Public awareness or apathy remains an additional problem. Citizens of countries where disasters are rare may be unfamiliar with hazards. Awareness will remain a major challenge in the context of a changing climate, as countries may be exposed to previously unknown hazards. Lack of Multi-Hazard EWS A major problem across all countries is the lack of multi-hazards EWS. In 2007 the Third International Early Warning Conference concluded that early warning systems must be developed with a multi-hazard and multi-sectoral approach. Progress remains slow. Only six countries reporting on the HFA mentioned EWS for more than two hazards. A review by Singh and Grasso (2008) found only five global multihazard warning systems: WFP (World Food Programme), HEWS (Humanitarian Early Warning Service),

34

AlertNet (service by Reuters), ReliefWeb (service by UN OCHA), and GDACS (Global Disaster Alert and Coordination System). Details about several of the systems are provided below. HEWS - The IASC (Inter-Agency Standing Committee) Humanitarian Early Warning Service (HEWS)12 is an inter-agency partnership project aimed at establishing a common platform for humanitarian early warnings and forecasts for natural hazards. The main objective of HEWS is to bring together and make accessible in a simple manner the most credible early warning information available at the global level from multiple specialized institutions. HEWS includes dedicated pages for floods, storms, locust, volcanoes, earthquakes, weather and other hazards. It depends fundamentally on the early warning information currently made available worldwide through a variety of specialized agencies and institutions. IASC includes many partners internal to the UN framework. Partnerships with external, non-UN specialized institutions and sources have been or are being established. AlertNet - AlertNet13 is an emergency information service run by Thompson Reuters focused on natural hazards. It addresses critical gaps in the information chain by deploying AlertNet reporting teams to disaster zones to disseminate fast, reliable information to affected populations in local languages. First launched after the 2010 Haiti earthquake, the Emergency Information service provides actionable information on everything from how to minimise disease risks to where to get medical help and how to trace missing relatives. ReliefWeb - ReliefWeb14 provides for timely, reliable and relevant humanitarian information and analysis. It is administered by the United Nations Office for the Coordination of Humanitarian Affairs (OCHA). The goal of ReliefWeb is to make sense of humanitarian crises worldwide. Its staff scan the websites of international and non-governmental organizations, governments, research institutions and the media for news, reports, press releases, appeals, policy documents, analysis and maps related to humanitarian emergencies worldwide. It then ensures the most relevant content is available on ReliefWeb, or delivered through a preferred channel (RSS, email, Twitter or Facebook). GDACS - The Global Disaster Alert and Coordination System (GDACS)15 provides near real-time alerts about natural disasters around the world and tools to facilitate response coordination, including media monitoring, map catalogues and Virtual On-Site Operations Coordination Centre. The Global Disaster Alert and Coordination System (GDACS) is a cooperation framework under the United Nations umbrella with the aim to consolidate and strengthen the network of providers and users of disaster information worldwide in order to provide reliable and accurate alerts and impact estimations after sudden-onset disasters and to improve the cooperation of international responders in the immediate aftermath of major natural, technological and environmental disasters. To date, GDACS has more than 9000 subscribers to automatic disaster alerts and users of the Virtual OSOCC, and has become an integral part of international disaster response to sudden-onset disasters. Table 4 compares the different multi-hazard systems, including types of events covered, outputs and users served. Only ReliefWeb covers all hazards. Most EWS are web based. Warnings are addressed to international organizations, rescue teams or aid agencies. Clearly web based information may not be accessible to the most vulnerable sectors of society. Furthermore, most multi-hazard EWS do not engage in actual data analysis and early warning creation. Rather, they focus on disseminating 12

http://www.hewsweb.org http://www.trust.org/alertnet/ 14 http://www.reliefweb.int 15 http://www.gdacs.org 13

35

information and providing information as events occur. As Ververs (2012) concluded in a study of the recent famine in the Horn of Africa, these multi-hazards EWS have yet to be proven effective. Table 4. Summary of global multi-hazard early warning systems (EWS) based on Singh and Grasso (2008). Shaded area indicates that the respective EWS covers that hazard, provides that form of output, or targets warnings towards that user.

Event Severe Weather El Nino Storms Floods Droughts Cyclones Fires Famine Output Web service Email SMS Briefing Notes Report Fax Users International Organizations Humanitarian Aid Decision Makers Civil Society

System AlertNet

ReliefWeb

HEWS

WFP

GDACS

Progress

Nevertheless, great strides have been made in developing climate-related early warning systems. According to a new book by Golnaraghi (2012), countries such as Bangladesh, Cuba, and France have rapidly improved their EWS. Improved systems have ten common characteristics, which have contributed to their success, irrespective of the political, social, institutional, and economic factors in each country (WMO 2011c): 1. Political recognition. There is a strong political recognition of the benefits of early warning systems, reflected in harmonized national and local disaster risk management policies, planning, legislation and budgeting. 2. Common operational components. Each effective system is built upon four components: hazard detection, monitoring and forecasting; risk analysis and incorporation of risk information in emergency planning and warnings; dissemination of timely and authoritative warnings; and community planning and preparedness with the ability to activate emergency plans to prepare and respond, coordinated across agencies at national to local levels. 3. Role clarification. Stakeholders are identified, their roles and responsibilities and coordination mechanisms are clearly defined and then they are documented within national and local plans, legislation, directives and memoranda of understanding,

36

including those of technical agencies such as National Meteorological and Hydrological Services. 4. Resource allocation. EWS capacities are supported by adequate resources (human, financial, equipment, etc.) across national and local levels, and the system is designed and implemented for long-term sustainability. 5. Risk assessment. Hazard, exposure and vulnerability information are used to carry out risk assessments at different levels, as critical input into emergency planning and development of warning messages. 6. Appropriate warnings. Warning messages are: clear, consistent and include risk information; designed to link threat levels to emergency preparedness and response actions (using colour, flags, etc.); understood by authorities and the population; and issued from a single (or unified), recognized and authoritative source. 7. Timely dissemination. Warning dissemination mechanisms are able to reach the authorities, other stakeholders and the population at risk in a timely and reliable fashion. 8. Integration into response planning. Emergency response plans are developed with consideration for hazard/risk levels, characteristics of the exposed communities (urban, rural, ethnic populations, tourists and particularly vulnerable groups such as children, the elderly and the hospitalized), coordination mechanisms and various stakeholders. 9. Integration in relevant educational programmes. Training in risk awareness, hazard recognition and related emergency response actions is integrated in various formal and informal educational programmes and linked to regularly conducted drills and tests across the system to ensure operational readiness at any time. 10. Feedback. Effective feedback and improvement mechanisms are in place at all levels to provide systematic evaluation and ensure system improvement over time. 16

Golnaraghi (2012) also documents the successful creation of the Multi-Hazard Early Warning System of the United States National Weather Service and the Shanghai Multi-Hazard Emergency Preparedness Programme. These systems should be reviewed further and possibly used as future models for countries that require multi-hazard risk management.

This review indicates that are areas where progress could quickly be made. Asia and Africa appear to be vulnerable to most hazards while agencies in North America and Europe (e.g. NOAA and the UK Met Office) have developed a range of effective EWS. Information sharing and capacity building is needed to allow developing countries to follow these successful models. Fire EWS is an area where rapid progress could be made. Few countries have fire EWS but almost all regions are affected by fires. Risk of fire will only increase with increasing temperatures and heat waves during the next century. A global early warning system is already established for fires and could easily be expanded and refined further. Likewise FEWSNET is a very good model for famine EWS. It could be expanded further into Asia, a region with high levels of hunger. Drought EWS systems, reported most often in Asia, should be expanded into Sub-Saharan Africa. Finally, this review indicates that is should be possible to develop a multi-hazard early warning system. Many of the EWS have common data requirements. Temperature and vegetation data are needed in almost all EWS. Topography, wind, snow pack and socio-economic data are necessary for several EWS. Clearly if data are collected for one hazard, the same information can be used to help develop warnings for other hazards (Table 5). 16

Unfortunately this book is still with the publisher and it was not possible to access PDF copies. Additional information can be found here: http://www.wmo.int/pages/publications/meteoworld/mhew_en.html.

37

At the same time, it is critical to ask challenging questions. What is the purpose of CLIM-WARN? To save lives? Reduce economic losses? Are some risks from disaster loss acceptable? Should warnings be issued if we cannot offer tools for response? Who should be involved in data collection and warning production? This review indicates that private sector actors such as Reuters, IBM, Chemonics International, and Deltares are actively playing a role in EWS creation and in the dissemination of warnings. Given government funding constraints, perhaps EWS should be delegated to private agencies or non-governmental organizations? Indeed, reporting to the 20092011 HFA Review, some countries mentioned that non-governmental organizations already manage EWS at the local level. At the same time, coordination is recognized as an increasing challenge to the proper functioning of EWS. Clearly greater assessment of these tradeoffs is needed.

38

Table 5. Summary of regions at risk to different hazards and projected changes to these hazards, as well as the distribution of the associated EWS and their data needs. While famine is not itself a hazard, it is listed as famine EWS are part of the review.

Areas of observed increase since 1950* Predicted areas of increase to 2100*

Drought Asia (West) Africa (East) Europe (Southern)

Famine

Africa (Southern)

Fire

Floods

Hurricanes

***

****

*****

Africa, Asia (Southeast), South America, Australia

Asia

Coastal Populations

Asia (India, Bangladesh, China)

Asia (Bangladesh and India)

Asia (but also OECD countries and countries in Sub Saharan Africa)

Asia (Japan), Australia, North America (US)

Europe (Med. + Central) North America (Southern) South America (North East Brazil)

Areas of high exposure Areas of high mortality Areas of high economic losses Key Data Needs

Sub-Saharan Africa

Africa, Asia**

Ground water and lake levels Livelihoods Humidity Hydrology (run off) Precipitation

Hydrology Precipitation

Precipitation

Precipitation SST (and ocean data)

Soil moisture Stream flow Snow pack

Snow pack

Snow pack

Socioeconomic

Socio-economic Temperature Topography

Topography

Vegetation conditions (NDVI)

Vegetation conditions (NDVI)

Vegetation conditions

Other relevant information # countries with EWS Model EWS

El Nino******

El Nino******

El Nino******

El Nino******

El Nino******

6 (50% in Asia)

4 (100% Africa)

7 (across all Regions)

35 (across all Regions)

US National Integrated Drought Information System

FEWS NET

Global Early Warning System for Wildlife Fires

UK, RIMES (Bangladesh)

22 (across all Regions) NOAA

Warning time

Monthly, Seasonal

Monthly, Quarterly

Hours to 15 days

Hours to 15 days

Wind

Wind

Evapo transpiration

2 -5 days, Seasonal

*From Chapter 3 IPCC SREX Report (2012) ** Note these are areas of hunger rather than famine *** Will be influenced by virtually certain increase in warm days and very likely increase in heat wave length, intensity, frequency **** Medium confidence of increases in heavy precipitation that would contribute to rain-generated local flooding *****Likely increase in mean maximum wind speed, possibly not in all basins. Likely increase in heavy rainfall with tropical cyclones ****** There is medium confidence of more frequent central Pacific ENSO since 1950, and insufficient agreement on future changes

39

4. Hazard Predication Capabilities “Climate modelling is coming of age. It is time to put climate modelling to service of society.” - Julia Slingo, UK Met Office Chief Scientist, at Planet Under Pressure Conference, 2012 Effective early warning systems require accurate monitoring, but equally important are prediction capabilities. Prediction capabilities vary for different hazards, regions and timescales. To date most early warning systems have been based on weather predictions, which provide less than 24 hours’ notice of an impending extreme weather event. Response options may be limited, and although lives may be saved, livelihoods can be destroyed (IPCC 2012). Fortunately, recent scientific advances offer the hope of extending warnings to the seasonal and perhaps even decadal timescale. This section summarizes current thinking on hazard prediction, based on discussions with scientists and drawing largely from the World Meteorological Organization’s paper “Climate and Weather Information Services for Humanitarian Operations” by Jones and Mason (2012).

4.1.

Background on Weather and Climate Forecasting Techniques

4.1.1. Weather Forecasting To forecast the weather accurately scientists need to know the state of current weather conditions and the physics of how these conditions interact and evolve over time (WMO 2011). Weather and climate observations come in many types: surface observations of temperature, wind speed and direction, dew point, the visibility, precipitation type, cloud type and height and the precipitation amount. Some are automated and some are taken by observers. At sea, observations are taken on ships and automated ones on buoys. To these basic observations are added Light Detection and Ranging (LIDAR), wind profilers and other methods of detecting conditions at or near the surface. Two or four times a day, weather balloon mounted radiosondes take measures aloft of the wind, temperature and humidity, as do aircraft and other airborne measurement systems. Remote sensing techniques (radar and polar orbiting and geostationary satellites in many parts of the EM spectrum) supplement the view of the atmosphere. These observations are then assimilated into a 3D analysis of the atmosphere using various techniques on grid points in the horizontal and in the vertical. Analyses are fed to a variety of different numerical weather prediction (NWP) models to predict the future state of the atmosphere using the equations of motion and others. If there is one future state of the atmosphere predicted, these models are classified as deterministic. NWP is not an exact science, there are errors in forecasting (since NWP uses approximations in the equations) and there are also errors in the initial analysis due to errors in observations and lack of observations to cover all of the earth. The atmosphere is an unstable system and these initial errors will grow without bound after several days leaving the forecast without skill. This problem can be partly addressed by making multiple predictions, each with slightly different estimates of the current weather (and possibly with slightly different representations of the physics of how the current weather will evolve). The results are ensemble forecasts. The mean of the ensemble forecast extends predictability to 10 days and is superior to the deterministic model after 3-4 days. The variability in the forecasts is related to how unstable and therefore how unpredictable the atmosphere is. Models differ in resolution and accuracy. The resolution determines the details that one can forecast in future. Global models tend to be coarser in resolution and are run to longer projections than regional (limited area) models. Finer grid scales are used in nested meso-scale models to cover very small areas allowing prediction of specific hazards such as thunderstorms. In general, weather forecasts are more

40

accurate in the mid-latitudes than they are in the tropics, due to greater numbers of observations in the mid-latitudes, but also because the physics of the mid-latitude weather systems are simpler and easier to model (WMO 2011, Jones and Mason 2012).

4.1.2. Climate Forecasting Despite a maximum limit to weather prediction of about 2 weeks, the ability to predict the climate at different timescales is possible because of different sources of predictability (Table 6). For example, scientists rely on large-scale weather systems such as the Madden Julian Oscillation (MJO)17 to make predictions on a monthly timescale. This does not enable accurate predictions for any specific day, but may provide some indication of future dry or wet periods. In such longer-range forecasts, scientists communicate what they think will be the general state of the atmosphere over the next few months or years without making any claims about what the precise weather will be at any specific time during that period (Jones and Mason 2012). Table 6. Sources of predictability at different timescales (Jones and Mason 2012) Timescale

Source of predictability

Scale of predictability

Weather (0-14 days)

Current weather

Specific locations, and specific timing

Monthly (2 – 4 weeks)

Large scale weather patterns

TBD; likely to be predominantly tropical areas only, and weekly summaries

Seasonal (1-6 months)

Sea-surface temperatures (and other surface features such as snow cover, and soil moisture)

Some parts of the tropics, and a few areas beyond, for a few months of the year only; three-month summaries

Multi-annual to decadal (6 months – 10? years)

Sub-surface oceanic conditions

TBD; likely to be sub continental scale, multi-year summaries

Century (>100 years)

Atmospheric composition

Sub continental scale, multi-decadal summaries

Millennial (1000s years)

Orbital parameters

Sub continental scale, millennial summary

Beyond a few weeks, ocean conditions form the basis of prediction (WMO 2012). The atmosphere gets much of its heat and moisture from the surface of the oceans. If the sea surface is unusually hot or cold weather patterns can be affected, and since sea temperatures change fairly slowly, this affect can last for several months. The best known example of this is the El Niño phenomenon. Although the El Niño phenomenon is reasonably well understood, it only provides accuracy in predictions about a year in advance (Jones and Mason 2012). In other areas, scientists’ understanding of the oceanic circulation is limited, largely because of a lack of subsurface observations. There are very few direct measurements, and satellites can provide very limited information of what is happening beneath the surface of the ocean. However, recent deployment of thousands of small buoys is providing data that may enable improvements in the prediction of the oceans, and thus in sea-surface temperatures, a year or more into the future (Jones and Mason 2012). In general, the influence of the ocean is much stronger in the tropics 17

The Madden-Julian Oscillation (MJO) is a tropical disturbance that propagates eastward around the global tropics with a cycle on the order of 30–60 days. The MJO has wide-ranging impacts on the patterns of tropical and extra-tropical precipitation, atmospheric circulation and surface temperature around the global tropics and subtropics (WMO 2011).

41

than in the mid-latitudes. As a result, whereas weather forecasts are more accurate in the mid-latitudes than in the tropics, the opposite is true of seasonal forecasts (Jones and Mason 2012). The spatial scale of predictability requires more detailed research. In the IPCC Fourth Assessment, for example, predictability was thought to be realizable at sub-continental scales. There are considerable efforts to provide more detailed information using downscaling techniques, but in the absence of any clear way of verifying this information, the ability to which projections can be reliably downscaled is an area of some dispute. There is considerable effort being placed into downscaling of climate change projections for input to the Fifth Assessment, and similar downscaling outputs have been considered fairly extensively in some national adaptation planning activities. However, it should be noted that many scientists are concerned about the unreliability of such detailed information. This question needs urgent research attention. Even at the seasonal timescale, most seasonal forecasts are presented as areaaverages, with areas typically of the order of 10,000s km2, or larger (Jones and Mason 2012). To date, only forecasts at the seasonal timescale are considered routine. Even at this timescale forecasts are not produced by all countries due to lack of predictability. For other timescales, there are as yet no standard product sets or standard procedures. For multi-annual forecasting, despite strong interest in information at this timescale, usable skill has yet to be demonstrated convincingly, and it would be premature at this time to move any such experimental forecasts into the mainstream for routine uptake or application (Jones and Mason 2012). That does not preclude their application, however, within a controlled, research-driven context; nor does it preclude the possibility that improvements in skill in the future might be realized.

4.2.

Seasonal Forecasts

4.2.1. Infrastructure The WMO has begun to implement a climate forecasting infrastructure in which a number of global, regional and national centres run climate prediction systems that adhere to a fixed production cycle, generate a standard set of prediction products, and routinely exchange, and disseminate predictions and related information in an operational environment similar to that operating for weather forecasting, albeit on longer production cycles (Jones and Mason 2012). Currently this infrastructure applies to only seasonal forecasts (1 to 6 months), although under the Implementation Plan for the Global Framework for Climate Services there are proposals to extend the system to cover all other timescales for which climate prediction information can be supplied. The structure for seasonal timescales involves a network of information providers at global, regional and national scales. These elements are discussed below. Global Producing Centres (GPCs) In 2006, the WMO began a process of identifying a network of Global Producing Centres (GPCs) for Long-Range Forecasts18 that make and distribute global seasonal forecasts. The current, officially designated WMO Global Producing Centres (GPCs) are shown in Figure 14. The GPCs are expected to adhere to certain well-defined standards including fixed production cycles, minimum suite of products, verification of long-range forecasts, up to date methodologies and accessible products (Jones and Mason 2012). Products currently set down as the minimum requirement for any designated GPC include: 1) Predictions for averages, accumulations, or frequencies, over 1-month periods or longer, of temperature, precipitation, sea-surface temperature (SST), MSLP, 500hPa height, 850hPa temperature; 2) Lead time between 0 and 4 months; 3) Monthly or at least quarterly issue frequency.

18

http://www.wmo.int/pages/prog/wcp/wcasp/clips/producers_forecasts.html

42

Figure 14. Current distribution of Global Producing Centres for Long Range Forecasts.

Lead Centres (LCs) The WMO has also designated two Lead Centres among the GPCs, namely the Lead Centre for LongRange Forecast Multi-model Ensembles (LC-LRFMME)19 hosted by the Korean Meteorological Agency in collaboration with the US National Oceanic and Atmospheric Administration, and the Lead Centre for Standard Verification System for Long-Range Forecasts (SVSLRF)20 hosted by the Australian Bureau of Meteorology in collaboration with the Meteorological Service of Canada. LC-LRFMME collects all the GPC real-time LRF products as well as the available hindcast data, and provides the same to NMHSs and other users in uniform formats and with common graphic displays. LC-SVSLRF is the authoritative source for mandatory verification information for all the GPCs, providing a single source for all information on the skills of the GPC products for any specific region/country in the world. The SVSLRF is a comprehensive set of standard measures for verifying seasonal forecasts and communicating their skill (Jones and Mason 2012). Regional Climate Centres (RCCs) At a regional level, the WMO is encouraging the establishment of a number of Regional Climate Centres (RCCs)21 that will generate and deliver more regionally focused, high-resolution data and products (not just predictions) as well as offer training support on the use of their products. The aim is for RCCs to assist WMO members in a given region or a defined sub-region to deliver better climate services and products including long-range forecasts, and to strengthen their capacity to meet national climate information needs (Jones and Mason 2012)

19

http://www.wmolc.org/ http://www.bom.gov.au/wmo/lrfvs/ 21 http://www.wmo.int/pages/prog/wcp/wcasp/RCCs.html 20

43

Regional Climate Outlook Forums (RCOFs) Regional Climate Outlook Forums seek to reach agreement among participants on current and expected seasonal conditions and to deliver a range of regional climate monitoring and outlook products. Using a predominantly consensus based approach, the RCOFs have an overarching responsibility to produce and disseminate a regional assessment of the state of the regional climate for the upcoming season. The forums bring together national, regional, and international climate experts, on an operational basis, to produce regional climate outlooks based on input from NMHSs, regional institutions, RCCs, and GPCs. They also facilitate enhanced feedback from the users to climate scientists, and catalyse the development of user specific products. They review impediments to the use of climate information, share successful lessons regarding applications of the past products, and enhance sector specific applications. The eleven major RCOFs currently in action are indicated in Figure 15.

Figure 15. Current distribution of Global Producing Centres for Long Range Forecasts CCOF FCCA FOCRAII GHACOF

Caribbean Central America WMO RA2 Greater Horn of Africa PICOF Pacific Islands PRESAC Central Africa PRESAO West Africa SARCOF South Africa SEECOF South East Europe SSACOF Southeast South America WCSACOF West Coast South America

National Meteorological Services (NMSs) The NMS is the authoritative centre for issuing forecasts at national levels. Around 50 to 60 percent of National Meteorological Services offer seasonal forecasts except those in Europe. The explanation for the European reticence to offer such services may be because of the weak predictability of the seasonal climate in that region (WMO 2011).

4.2.2. Standard products Seasonal forecasts are usually presented in one of two formats: 1. One or more “deterministic” predictions of a seasonally averaged or integrated meteorological variable (for example, 3-month mean temperature of total precipitation), usually as output from a statistical or dynamical prediction model. This value may represent an area-average or may be for a specific location, such as a meteorological station. Other deterministic forecast formats include counts (e.g. hurricane frequencies or rain-day frequencies), and dates (e.g. monsoon onset dates). The uncertainty in the prediction may be represented by some measure of the ensemble spread, or as a probabilistic forecast, or by using prediction intervals. 2. A set of probabilities for the observation to fall within each of two or more pre-defined ranges. Most commonly, three equi-probable categories are used (including at the RCOFs), but five categories (e.g. UK Met Office), and two categories (e.g. Australian BoM) are not uncommon. Some centres issue three-category forecast probabilities for categories that are not equi-probable: usually the outer

44

categories have climatological probabilities of less than 33% to indicate the chances of extreme climate events occurring. Forecasts are usually for 3-month periods, but some RCOFs issue 4-month averages. Lead-times can vary from 0 up to 9 months (NCEP). Since the first RCOF review meeting, in 1998, users have been requesting tailored presentation of forecast formats in place of the tercile probabilities and broad regional averages. Unfortunately, to date there has been little apparent response to this request in terms of changes to operational products generated at the RCOFs, but there has been considerable research activity to explore the predictability of more tailored information (Jones and Mason 2012). In many areas, rainfall frequencies, for example, are more predictable than rainfall totals, and can be seen as an important contribution to requests for information about the characteristics of rainfall during the season. The agricultural community has been active in making such requests. For EWSs, however, light rainfall events are generally not of much interest, and there is more concern about the risk of heavy rainfall events. Unfortunately, partly because of sample size problems, and partly because of the unpredictability of weather noise at seasonal timescales, it is unlikely to be viable to provide skillful forecasts of heavy rainfall totals at specific locations, but there may be some useful predictability of changes in heavy rainfall risks over large areas.

4.2.3. Verification Predictability of seasonal forecasts varies by climate variable and region. Analysis of seasonal temperature and precipitation forecasts by the Research Institute for Climate and Society, for example, shows that prediction skill is highest in Indonesia, eastern equatorial Africa, and southeastern South America during the last few months of the calendar year; in portions of southern Africa from November to March; and in India and the Sahel from June through September (Barston et al. 2010). Predictability on seasonal timescales is much lower for rainfall than for temperature, but also depends on geographical location (WMO 2011). Relatively high temperature skill is noted in much of the tropics and in some extra tropical regions. Skill for precipitation, while lower than that for temperature, is also generally highest in the tropics. In general skill is concentrated in the seasons and regions having known responses to ENSO. Seasonal forecasts are therefore more accurate for hazards that are influenced by ENSO (such as droughts and floods22). Standardized Verification System for Long-Range Forecasts (SVSLRF) Under the auspices of the WMO’s Commission for Basic Systems (CBS) a recommended set of procedures for the verification of “long-range” (seasonal) forecasts has been defined. This so-called Standardized Verification System for Long-Range Forecasts (SVSLRF)23 has the specific objective of providing verification information for Global Producing Centre (GPC) products that are used as inputs to seasonal forecasting processes, including RCOFs. RCOF verification Although it is standard practice in most of the RCOFs to verify the forecast products from the previous forum, there is little to any systematic tracking of these verification results, and none of the results seem to be available online. In addition, the forecasts are verified as if they were deterministic: some form of “percentage correct” is calculated taking the category with the highest forecast probability as the 22

Generally, in El Niño years droughts are experienced in South-east Asia and over large areas of Australia and Southern Africa, heavy rainfall and flooding in arid areas of South America and East Africa and failure of the monsoons in India and West Africa. In temperate regions, El Niño is associated with wet winters in southern United States and mild winters over western Canada and part of Northern United States (WMO 2012).

23

http://www.wmo.int/pages/prog/www/DPS/LRF/ATTACHII-8SVSfrom%20WMO_485_Vol_I.pdf

45

category predicted and making adjustments for tied highest probabilities. While such a score is informative, Jones and Mason (2012) conclude that: One problem is that the actual probabilities are ignored in this procedure, and only the relative probabilities are considered. There is likely to be interest in knowing how the forecasts verify when probabilities are highly inflated (e.g., probabilities of 50% or higher), compared to when only partially inflated (e.g., a 40% probability). There may also be some interest in knowing whether decreased probabilities do successfully indicate a decreased chance of the respective category occurring. A second problem is that the verification of the RCOF products is in terms of seasonal rainfall totals, not necessarily in terms of flooding events, or drought occurrence (or, more pertinently, in terms of disaster occurrence). Although the RCOFs are, appropriately, verifying what was forecast, the verification results still do not give an unequivocal indication to how successful the forecasts were in predicting events…. As a first step towards producing more user-targeted verification information, it would be a helpful practice for the RCOFs to at least provide a bit more of a detailed diagnostic of the previous target season to supplement the standard verification results. For example, instead of only reporting the seasonal rainfall total, information on the occurrence of dry spells and of heavy rainfall events may be of interest. Detailed diagnostics (which consider the reliability of the probabilities) of the performance of the RCOF forecasts over time is only possible once a reasonable number of forecasts have been produced. The African RCOFs have been running longer than in any other part of the world, and so it has been possible to perform some analyses of the reliability of these forecasts. Through collaboration between ACMAD and the IRI, the first 10 years of the SARCOF, GHACOF, and PRESAO forecasts have been verified. The primary conclusion for all three regions is that there has been some skill, but in general the reliability of the forecasts has been poor. Similar conclusions were drawn from a preliminary assessment of the forecasts from the Southeast South America forum. Part of the poor reliability is a result of hedging by the forecasters, who have a tendency to assign unrealistically high probabilities to the normal category, and is partly due to imperfect calibration of the models used as input to the consensus building process. Unfortunately, these problems mean that it is difficult for the users of the RCOF forecasts to realize benefit from the information provided. Verification activities in RCOFs should be taken more seriously, especially given the poor reliability of RCOF forecasts to date (Jones and Mason 2012).

4.3.

Multi-year and Decadal Forecasts

During recent decades there has been considerable effort to understand the decadal variability of the climate system. In particular, the Pacific decadal variability (PDV) and its possible impacts on regional and global climate has been the subject of much research (Vera et al. 2010). Other areas of interest include the North Atlantic Oscillation (NAO), which has been clearly linked to surface temperature and precipitation variations over Europe and North America on decadal and multi-decadal timescales. Large multi-decadal variations in sea-surface temperature of the Atlantic have also been linked to precipitation anomalies in Northeast Brazil, the African Sahel, the North American and European summer climate and the formation of major Atlantic hurricanes (Vera et al. 2010, Van Oldenborgh et al. 2012). Decadal forecasting is complicated by climate change itself. In order to make decadal forecasts, it is important not only to consider the initial state of the atmosphere and ocean but also to include natural external changes (e.g. solar radiation) and anthropogenic forcing (e.g., greenhouse gas emissions) (WMO 2012). Several groups around the world are working on developing decadal forecasts (MetOffice, NOAA, Japanese Meterological Research Institute, European Consortium for Midterm Weather Forecasting, Max Planck Institute for Meteorology). The Fifth IPCC report will also feature a section on decadal monitoring. Initial decadal prediction efforts in the last few years show predictive skill of global average

46

temperature up to a decade in advance. Many of the recent decadal predictability and prediction studies focused on the North Atlantic region and this may be the region where we have most skill on a 2 to 5 year timescale, including some hurricane prediction possibilities (Metha et al. 2011, Van Oldenborgh et al. 2012). These initial predictability studies also suggest that there may be very small skill in predicting means and extremes in temperature over land, that extra tropical regions may be more predictable than tropical regions, and that there may be some skill in the predictability of extreme weather event statistics (Metha et al. 2011). In a recent paper, Van Oldenborg et al. (2012) attempt to verify SST, temperature and precipitation skill of a 4-model 12-member ensemble of 10-yr hindcasts. In temperature, most of the skill in multi-model ensembles comes from externally forced trends (i.e. projections of greenhouse gas emissions) rather than from initial conditions (i.e. ocean trends). However, regionally there appears to be skill beyond the forced trend in the two areas of well-known low-frequency variability: SST in parts of the North Atlantic and Pacific Oceans. A comparison with the CMIP3 ensemble shows that the skill in the northern North Atlantic and eastern Pacific is most likely due to the initialisation, whereas the skill in the subtropical North Atlantic and western North Pacific are probably due to the forcing. Skill in decadal ENSO prediction is lower and not statistically significant. There is also an indication of skill in hindcasting decadal Sahel rainfall variations, which are known to be teleconnected to North Atlantic and Pacific SST. As a result of such efforts, decadal-scale products may become more widely available in time. Products will typically be provided as maps and tables of expected anomalies, and will most likely be in probabilistic formats. Information related to the predictions will include consensus summary assessments of key features and, at national levels, may include advisories and warnings (Jones and Mason 2012). It will be important, however, to develop and make accessible appropriate verification measures for such forecasts. In order to improve predication it is necessary to increase data collection in the oceans. The quality of observed ocean data is a major concern. Argo floats are now capable of measuring ocean temperature down to depths of 800 meters, which will help improve data collection. However, the lack of salinity data before the deployment of Argo floats hampers ability to describe decadal climate variations (DCV). More accurate characterization, better mechanistic understanding, and identification of sources of potential predictability of DCV are necessary. Model errors will need to be reduced by increasing resolution, including vertical resolution into the atmosphere (Scaife et al. 2011). This will require large increases in computing power, and sustained programs by model development centres to identify, understand, and correct principal model errors (Metha et al. 2011) The computing power now becoming available means that it will be possible to run global models with a resolution of a few kilometres (as required for many practical applications) as well as very large model ensembles to assess uncertainty (WMO 2011). The general consensus emerging from discussions with scientists, is that decadal climate predictions may become useful to society sometime in the future but remain highly experimental now (Mehta et al 2011). While it may not be possible to offer clear decadal predictions to policy makers, reviews of decade long historical trends can nevertheless be used to help plan for the future. For example, the Sahel and many parts of Southern Africa show slow variability of climate. Dry years cluster together and wet years cluster together. Trends can be used to help plan for the future. Until the science of decadal prediction improves, Wilco Hazeleger of the Koninklijk Netherlands Meteorogisch encourages the creation of “story lines of future weather”. Given lack of accurate quantitative estimates, efforts could on “foreseeability”.

47

4.4.

Conclusion

In recent years, rapid advances have improved forecasts and increased the timescale of prediction. A well-developed industry produces forecasts along the 6 month to 1 year timescale. Decadal forecasting is also advancing quickly, with improved data from oceans and increasing computing power. However continued improvements can be made at all timescales. Verification indicates that we are better at forecasting temperature and making long-term predictions in the tropics. Prediction of other climatic variables, including extreme events and hazards, needs to improve, as does the resolution of the models. More research and investment is needed to translate information about large-scale decadal climate variations, tercile probabilities and broad regional averages into hazard predictions on the local scales required for decision making. Improved prediction efforts by the scientific community alone are insufficient for effective EWS design. The scientific community and policy-makers should outline a strategy for timely decision making by indicating what information is needed by decision makers, how predictions will be used, how reliable the prediction must be to produce an effective response, and how to communicate this information to authorities and the public. A miscommunicated or misused prediction can result in costs to the society and also increase public apathy towards warnings. However, it is important to note that improved reliability and forecast skill may not be as important to the public as the general timing and form of climatic information and forecasts (IPCC 2012). Decision makers typically manage risks holistically, while scientific information is generally derived using reductionist approaches (Meinke et al. 2006). The net outcome can be a ‘disconnect’ between scientists and decision makers. Climate and hydrometeorological information can be developed that, although scientifically sound, may lack relevance to the decision-maker (IPCC 2012). Thus, decadal projections and seasonal forecasts may still need scientific improvements, but even rough outputs (e.g. “story lines” or “forseeability”) can be of relevance to certain user groups.

48

5. Steps Forward “A change in the weather is sufficient to recreate the world and ourselves.” - Marcel Proust The creation of CLIM-WARN is an ambitious proposal. This review indicates that there is a basis for the future development of CLIM-WARN. Numerous early warning systems (EWS) already exist for hydrometeorological hazards. There is solid capability to create seasonal climate forecasts, and the science of decadal modelling is rapidly developing. However, significant challenges will have to be overcome. Multi-hazard EWS remain rare and a great deal of progress will have to be made to unify different hazard EWS. Predication capabilities are better for some parts of the world, and for some meteorological variables, than for others. It is unclear how to translate probabilities produced by seasonal or mid to long-term forecasts into operational warnings. Due to the diversity of actors, the creation of a CLIM-WARN system will face considerable operational problems. Significant intellectual and financial investment will be required to overcome these challenges. The roles of different actors need to be clarified, thorough risk assessments need to be conducted, and future users of CLIM-WARN need to be identified.

5.1.

Key Findings

Early warning systems for hydrometeorological hazards are well developed. Gains can be made by supporting knowledge transfer between countries and by expanding successful models into regions without EWS. As data needs for hazards are often similar, the creation of multi-hazard EWS is also possible. At the CLIM-WARN Geneva meeting, it was suggested that it may be possible to stitch together a network of EWS by identifying shared requirements among them that could be the substance of the platform holding common ways of using satellite observation or atmospheric modelling. This review indicates that several EWS systems have common data needs, such as accurate temperature, precipitation, vegetation and socio-economic indicators (Table 5). A common platform could be established for data collection. The WMO Climate Services Framework offers an opportunity to expand and standardize data collection, filling data gaps necessary for effective EWS development. Potential exists to extend the timescale of warnings Most EWS issue warnings on a very short time frame, from several hours to several days. However, flood and fire forecasts are now available for up to 14-day periods, and seasonal forecasts are currently produced for hurricanes and droughts. Capacity exists to expand forecasts to the seasonal level for hazards influenced by temperature extremes or ENSO. Wildland fire models, for example, are well developed and could easily be extended to the seasonal level. Greater support should be given to research in this area. Future efforts should focus on Africa or Asia Major gaps in EWS exist in both Africa and Asia. According to the IPCC SREX Report (2012) Africa and Asia will be exposed to increases in climate hazards in future. These regions are highly vulnerable to losses from hazards. Fortunately, seasonal prediction is most accurate for tropical regions. There is therefore both a need and a rationale to focus the development of CLIM-WARN on parts of Africa and Asia.

49

5.2.

Possible Actions

A variety of steps can be taken to further development of CLIM-WARN: Comprehensively map business, civil-society and community activities supporting EWS Given the diversity of local, national, regional and global actors working in EWS, identifying ways to create or design a unified global network will be an operational challenge. A mapping exercise is an essential starting point, a way to identify organizations that need to be involved and their potential synergies, overlaps and gaps. This report has focused on national actors. Due to the limited timescale, it was not possible to comprehensively map all actors involved in different EWS, including private actors such as insurance companies, civil society and community-based organizations. However, it is clear that private companies such as IBM or Munich RE are active in the area of EWS. Faith-based organisations are also influential in assisting local communities in disaster risk management (IPCC 2012). A comprehensive understanding of all the actors involved in EWS is critical to building a unified EWS network. Conduct a detailed evaluation of national EWS. Work with National Meteorological organizations and UNISDR to improve national Hyogo Framework for Action reporting Government reporting for the Hyogo Framework for action offers valuable information regarding the state of national EWS. However, details provided by governments on Priority 2, core indicator 2.3, are currently insufficient to fully evaluate EWS. EWS for each hazard should be categorized along a spectrum which makes clear the quality of the EWS system and areas for action. For example, RIMES currently categorizes flood forecasting compatibilities of its member countries according to five levels: Level 0: No flood forecasting system, warning dissemination and community level response exits. Level 1: No lead time forecast available. Flood forecasting and warning services are limited or not operational. Significant upgrading and strengthening of the basic data collection and transmission networks is required. Insufficient network coverage and data exchange for hydrological forecasting. Very poor coordination among warning providers and users. Warning dissemination infrastructure is very weak. Level 2: Upgrading and strengthening of the basic data collection and transmission networks is required. Network coverage and data exchange for hydrological forecasting need to be upgraded. Experimental or very limited lead time forecasts exist. Flood forecasting and warning services need to be upgraded for operational purposes. Coordination among warning providers and users need to be upgraded. Risk communication system needs to be upgraded. Level 3: 1-2 days lead time available for flood forecast. Basic infrastructures for flood forecasting and warning services are in place. However, upgraded data management procedures (telemetric stations) and improved methodologies and models for flood forecasting are required. Regression or other simple models are generally used to forecast the peak of the water level and approximate time of transit. Some coordination exists but community level understanding and interpretation is poor. Level 4: Observation and telemetric stations are available for all river basins. Long lead forecasts available experimentally. Established flood forecasting and warning services with quality product development in place. Opportunities exist for further improvement through the use of new and innovative technology. Some hydrological tools and methods are used to produce the flood

50

forecasts and warnings (Rainfall routing model, MIKE-11, etc). Coordination among national level users and providers exits. Community level interpretation exists on pilot basis. Level 5: Well-established observation and real time data communication system with backup services. Long lead forecasts available and operational. Well-established flood forecasting and warning services with high quality products and opportunities for further improvement through the use of new technology. The systems normally combine products and information from both meteorological and hydrological sources. Identify ways to translate seasonal forecasts into warnings relevant to end users, and support further scientific research into decadal forecasts According to the IPCC (2012), “while there are potential benefits of early warning systems that span a continuum of timescales, for much of the disaster risk management community the idea of preparedness based on predictions is a new concept. Most communities have largely operated in a reactive mode, either to disasters that have already occurred or in emergency preparedness for an imminent disaster predicted with high confidence.” The possibility of using long-term weather and climate predictions to provide advanced warning of extreme conditions is a recent development. Despite over a decade of operational seasonal predictions in many parts of the globe, examples of the use of such information by the disaster risk management community are scarce, due to the uncertainty of predictions and comprehension of their implications (Meinke et al., 2006). Most seasonal rainfall predictions, for example, are presented as probabilities that total rainfall over the coming few (typically three) months will be amongst the highest or lowest third of rainfall totals as measured over a historical period and these are averaged over large areas (typically tens of thousands of square kilometres). Not only are the probabilities lacking in precision (highest probabilities are most frequently around 40% or 45%, compared to the climatologically expected probability of 33%), but also the target variable, seasonal rainfall total, does not necessarily map well onto flood occurrence. Although higher than normal seasonal rainfall will often be associated with a higher risk of floods, it is possible for the seasonal rainfall total to be unusually high with no flooding. Alternatively, the total may be unusually low, yet flooding might occur because of the occurrence of an isolated heavy rainfall event. Thus even when seasonal predictions are understood properly, it may not be obvious how to utilise them (IPCC 2012). These problems emphasize the need for the development of tools to translate such information into quantities directly relevant to users. Conduct a climate hazard vulnerability analysis Additional work is needed to identify an analytic framework for EWS. FEWSNET, for example, bases work on livelihood zones and conducts significant livelihood analysis. A review could be conducted to identify the impact of climate related threats on different livelihood types, or ecosystems, around the world. Indeed, it may be useful to produce a global climate impact inventory both for countries and species and use this as a basis for the creation of the CLIM-WARN system. Review EWS dissemination and response, and identify users’ needs This review has focused on the first two components of EWS, risk knowledge and monitoring. Additional study is needed to clarify the final two elements of an EWS, dissemination and response. It is critical to identify how to deliver warnings to groups and sectors most vulnerable to climate change. Specific areas of future research should include: 1) identifying potential users of CLIM-WARN on a local, national and regional scale, and on the seasonal and multi-year timescale; and 2) calibrating climate warnings so that they better suit the needs of users.

51

Research could be based on India, an illuminating case study country that faces a diversity of hazards from climate change. Andhra Pradesh, for example, is expected to suffer from droughts, while regions in the Himalaya face floods from melting glaciers (Narain et al. 2009). Coastal areas will be vulnerable to damage from more powerful cyclones. Local organizations, such as RIMES, can assist study. Alternatively, a country in the African Sahel could be used as a case study site. This area is highly vulnerable to climate change due to its geographic location at the southern edge of the Sahara desert and the strong dependence of its population on rain fed agriculture (Tacko Kandji 2006). Preliminary decadal models show some skill in forecasting precipitation in the Sahel. Study could explore the feasibility of creating and disseminating decadal forecasts to users. Meetings at the local, national and regional levels could then be held to identify groups vulnerable to climate change and to clarify sectors and individuals that may use warnings. Contacts for meetings could be derived from UNEP’s large climate adaptation network. Interviews would then be held to generate a list of user needs and concerns. Convene multi-stakeholder and expert group discussions to review and identify CLIM-WARN design It is also critical to further identify the design of a comprehensive EWS, perhaps through a high level panel or a series of design workshops with potential users as well as technical experts. Clear disagreement remains about how best to design CLIM-WARN and its possible areas of focus (see Section 5.3). The objective of these workshops should be to develop a unified plan for an “end-to-end” warning system.

5.3.

Possible CLIM-WARN Design

During discussions with scientists, three different possible approaches for CLIM-WARN design were repeatedly mentioned. These are briefly highlighted here.

5.3.1. Biome or ecosystem targeted approach Several scientists interviewed for this report called for the creation of an ecosystem based global EWS. Ecosystem degradation influences risks of hazards such as landslides (IPCC 2012). It is critical to quantify the impact of ecosystem degradation on disaster risk and bring dynamic ecosystems’ perspective into risk models. According to the IPCC (2012), “Ecosystem-based solutions in the context of changing climate risks can offer ‘triple-win’ solutions, as they can provide cost-effective risk reduction, support biodiversity conservation, and enable improvements in economic livelihoods and human wellbeing, particularly to the poor and vulnerable.” Systematic biosphere monitoring is not included in the WMO’s Climate Services Framework. Several monitoring systems have independently been established to record changes in ecosystems (e.g. Tropical Ecology Assessment and Monitoring Network24 and the Global Observation Research Initiative25 in Alpine Environments). Composite indicators of environmental sustainability, such as the Living Planet Index, are also available. However, few monitoring systems integrate multiple threats or drivers into calculations and none can be described as formal EWS. Several scientists called for a global effort to examine the impact of climate change on species and ecosystems, and suggested the creation of an EWS for ecosystems. An EWS could be established in ecosystems that are vulnerable to climate change, such as high altitude or polar regions (Arctic tundra and boreal forests). Alternatively, it may be possible to focus on human populated ecosystems: slopes that are threatened by landslides, coastal areas threatened by cyclones. 24

The TEAM network currently comprises 17 sites in Africa, Asia and Latin America, with the goal of expanding to 40 sites by 2013. Sites span a range of environmental and anthropogenic gradients. TEAM monitors tropical mammal and bird communities using extensive camera trap arrays, following a standardized protocol.   25 GLORIA aims to establish a worldwide network of permanent plots to monitor the impact of climate change on mountain ecosystems. GLORIA is currently represented in 77 mountain regions across five continents.

52

Box 4. Ecological Early Warnings It has been argued that rates of change in ecosystems, or changes in ecological time series data (quantified by changing skewness, increased variability or autocorrelation) can provide warning of impending collapse and can be used as the basis of EWS (Guttal and Jayaprakash 2008). The spatial variance of eutrophic water regions in a lake increases as the lake approaches a eutrophic state. It is unclear, however, whether such indicators provide sufficient warning to avert regime shifts (Biggs et al. 2009, Donangelo et al. 2010). Additional study is needed to identify tipping points in different ecosystems and to evaluate the possible applicability to CLIM-WARN.

5.3.2. Risk assessment rather than warning Given uncertainties associated with seasonal and decadal climate forecasts, several scientists suggested focusing on characterization of risks and how these are changing as a result of climate. While hazard prediction may be difficult, it is possible to obtain an accurate depiction of current changes in climate. Efforts should focus on reviewing the current state, trends and drivers of risk. It is possible to attribute changes in state and trends to drivers. By then describing changes to drivers, which are often more predictable than changes in climate itself, one can characterize how risks may change in the future. In recent years, the science of event attribution has developed considerably (Stott et al. 2011). The approach of using model experiments to calculate how a particular climate driver has changed the probability of an event occurring, as proposed by Allen and Stott (2003), has been applied to a number of different cases (eg Christidis et al, 2010, Pall et al, 2011). The probability of a particular event happening in an ensemble of model simulations representing current conditions is compared with a parallel ensemble of model simulations representing an alternative world that might have occurred had the particular driver or drivers of interest been absent (Stott et al. 2011). Scientists at the University of Cape Town have implemented a pilot version of an attribution forecast system that runs in parallel with an existing seasonal forecasting service.26 Along with the real seasonal forecast, a parallel forecast is run of a non-greenhouse gas world in which human activities had never released greenhouse gases to the atmosphere and the ocean had not warmed in response to those emissions. Currently this simple implementation mainly serves as a demonstration to aid in ascertaining the requirements and characteristics that potential users of such a system might demand. It has however revealed seasonal and regional variations in attributable risk, as well as some apparently robust similarities and differences with seasonal forecasting products. For example, the relative predictability of temperature versus precipitation events is very different in an attribution statement than in a standard seasonal forecast (Stott et al. 2011).

5.3.3. Seamless integrated system According to the IPCC (2012), “Developing resiliency to weather and climate involves developing resiliency to its variability on a continuum of timescales, and in an ideal world early warnings would be available across this continuum.” Several scientists believed the creation of a seamless integrated prediction system, which produces warnings at different timescales, was possible. In a “seamless” system, warnings extend from “minutes for tornadoes, to days for winter storms and air pollution episodes, to weeks for floods and droughts, to decades for climate variations” (McBean 1999). An 26

http://www.csag.uct.ac.za/~daithi/forecast 53

“integrated” system is an optimized observing network designed to provide comprehensive atmospheric, hydrologic, land surface and oceanic data to meet the expanding needs of diverse client communities. Such a system uses multiple observing platforms to observe common parameters and integrates the results for ingestion into a variety of display and analysis systems. In terms of CLIM-WARN, decadal modelling could be used to provide a global overview, indicating regions suffering from an increased number of hazards in future. These areas could then be monitored further, either by satellite or by a team of experts. More specific regional predictions could be created and vulnerable groups or sectors of the economy identified well in advance. The long-term forecasts could be constantly updated, and more accurate warnings issued on a seasonal timescale. Warnings may be issued to different actors (businesses, government, local communities) at different times according to needs. For example, local councils may want to change planning processes years in advance of hazards. However farmers may only need warnings on a seasonal scale, and households in the path of a hurricane may only need warnings two weeks in advance. A seamless integrated CLIM-WARN system could meet the needs of a broad diversity of users, allowing action to be taken to both modify long-term policies and change shortterm behaviours. Final Thoughts Winter did not arrive in southern Canada in 2012. Such changes in weather extremes have already affected significant aspects of Canadian society. As reported by Gulli (2012): Construction starts have increased, while demand for natural gas for heating has decreased. Retailers of winter apparel and sports equipment have laid off employees to offset excess inventory. ATVs have supplanted snowmobiles as the recreational rental vehicle of choice. Sales of cottages have risen because buyers have been able to use seasonal roads to see properties. Car sales have risen because good road conditions make for enthusiastic shoppers. Cities have spent less on snow removal and salting, but more on fixing potholes. According to the IPCC (2012) we should expect changes in extreme events and hydrometeorological hazards in the future. This poses a massive challenge to sustainable development. However, as Gordon McBean (1999), a climatologist and former Assistant Deputy Minister in Environment Canada stated, “Our improved ability to make confident forecasts of future states of the atmosphere or an aquatic system provides a new relationship with the future.” The creation of CLIM-WARN will require significant research and investment. However, if CLIM-WARN can be turned into a reality, it will “represent a considerable step towards a sustainable future, a 21st century destination, reached via a partnership of ‘international cooperation’, a consistent ‘environmental vision’ and ‘science’” (McBean 1999).

54

6. Acronyms ADI - Atmospheric Dispersion Index CLIM-WARN - proposed name for a global early warning system for climate change DCV - Decadal Climate Variations EC - European Commission EC-JRC - European Commission Joint Research Centre ECPC - Experimental Climate Prediction Centre EFAS - European Flood Alert System ENAC - Emergency Notification and Assistance Convention ENSO - El Niño Southern Oscillation EO - Earth observation ESA - European Space Agency EOS - Earth Observing Satellites EWS - Early Warning System FAO - Food and Agriculture Organization of the United Nations FDR - Fire Danger Rating FEWSNET- Famine Early Warning System FWI - Fire Weather Index GDACS - Global Disaster Alert and Coordination System GDP - Gross Domestic Product GDPS - Global Data Processing System GEOSS - Global Earth Observation System of Systems GFAS -Global Flood Alert System GFM - Global Flood Model GFMC - Global Fire Monitoring Centre GIEWS - Global Information and Early Warning System on Food and Agriculture GPC - Global Producing Centres GTS - Global Telecommunications System HEWS - Humanitarian Early Warning Early Warning Service HFA - Hyogo Framework for Action IPCC - Intergovernmental Panel on Climate Change JMA - Japan Meteorological Agency LC - Lead Centres MODIS - Moderate Resolution Imaging Spectroradiometer NAO - North Atlantic Oscillation NGO - Non-governmental Organization NMS - National Meteorological Services NOAA - National Oceanic and Atmospheric Organization NWP - Numerical Weather Prediction NWS - National Weather Service OCHA - United Nations Office for the Coordination of Humanitarian Affairs RCC - Regional Climate Centres RCOF - Regional Climate Outlook Forums RIMES - Regional Integrated Multi-Hazard Early Warning System for Africa and Asia SST - Sea surface temperature UN - United Nations UNCCD - United Nations Convention to Combat Desertification UNDP - United Nations Development Programme UNEP - United Nations Environment Programme UNEP-DEWA - United Nations Environment Programme-Division of Early Warning and Assessment UNESCO - United Nations Educational, Scientific and Cultural Organization UNISDR - International Strategy for Disaster Reduction USGS - U.S. Geological Survey WCP - World Climate Programme WFP - World Food Programme WMO - World Meteorological Organization WMO-TCP - World Meteorological Organization-Tropical Cyclone Programme

55

7. References Allen, M., and P.A. Stott. (2003). Estimating signal amplitudes in optimal fingerprinting, part I: theory. Climate Dynamics, 21, 477-491. Barnston, A., Li., S., Mason, S., deWitt, D., Goddard, L. and X. Gong. (2010). Verification of the first 11 years of IRI’s seasonal climate forecasts. Journal of Applied Meteorology and Climatology, 49, 493 520. Bender, M., Knutson, T., Tuleya ,R., Sirutis, J., Vecchi, G., Garner, S., and IM Held. (2010). Modelled impact of anthropogenic warming of the frequency of intense Atlantic hurricanes. Science, 327, 454– 458. Biggs, R., Carpenter, S., and W. Brock. (2009). Turning back from the brink: Detecting an impending regime shift in time to avert it. Proceedings National Academy of Sciences, 106 (3), 826 – 831. Christidis, N., Stott, P.A., Zwiers, F.W., Shiogama, H., and T. Nozawa. (2010). Probabilistic estimates of recent changes in temperature: A multi-scale attribution analysis. Climate Dynamics, 34, 1139-1156. Donangelo, R., Fort, H., Dakos, V., Scheffer, M. and E. Van Nes. (2010). Early warnings for catastrophic shifts in ecosystems: comparison between spatial and temporal indicators. International Journal of Bifurcation and Chaos, 20 (2), 315 – 321. Fakhruddin, S. (2012). Enhancing Flash Flood Forecasting and Warning System in Pakistan. [Online]. Available: http://www.shmfakhruddin.com/main/index.php?type=review&area=1&p=articles&id=71 FAO. (2010). Global hunger declining but still unacceptably high. [Online]. Available: www.fao.org/docrep/012/al390e/al390e00.pdf FEWS NET. (2011). East Africa Food Security Update, August 7, 2011. [Online]. Available: http://www.fews.net/docs/Publications/East_Regional_FSOU_2011_08_07_final.pdf Flannigan, M., Logan, K., Amiro, B., Skinner, W., and B. Stocks. (2005). Future area burned in Canada. Climatic Change, 72, 1–16. Flannigan, M., Krawchuk, M., de Groot, W., Wotton, B., and L. Gowman. (2009). Implications of changing climate for global wildland fire. International Journal of Wildland Fire, 18, 483 – 507. Glantz, M. (2004). Usable Science 8: Early Warning Systems: Do’s and Don’ts. (Report of Workshop held 20-23 October 2003). Shanghai, China. Goldammer J., Sukhinin A, and I. Csiszar. (2003). Russian Federation. Int. For. Fire News, 29, 89–111. Goldammer, J. and W. de Groot. (2011). Implementing a Global Early Warning System for Wildland Fire. Paper presented at the First International Conference on Wildfire in Natural Resources Lands, 26 – 28 October, Gorgan, Iran. Golnaraghi, M. (2012). Institutional Partnerships in Multi-Hazard Early Warning Systems. Geneva: Springer Verlag GmbH. Grasso, V. (2012). Early Warning Systems: A State of the Art Analysis and Future Directions. Nairobi: Division of Early Warning and Assessment (DEWA), United Nations Environment Programme (UNEP). Gulli, C. (2012, March 26). The Winter That Never Was. Maclean’s Magazine, 20 -24.

56

Guttal, V. and Jayaprakash, C. (2008). Changing skewness: an early warning signal of regime shifts in ecosystems. Ecology Letters, 11, 450 – 460. Held, I., Delworth,T., Lu, J., Findell, K., and T. Knutson. (2005). Simulation of Sahel drought in the 20th and 21st centuries. PNAS, 102, 17891-17896. Hirtzer, M. (2012, March 22). Warmest March Ever Drives Farmers to Plant Early. Reuters. [Online]. Available: http://www.reuters.com/article/2012/03/22/us-usa-corn-planting-idUSBRE82L0IW20120322 IPCC. (2007). Summary for policymakers. In Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M., Avery, KB, Tignor, M., Miller, HL. (Eds.), Climate Change 2007: The Physical Science Basis. Contribution of Working Group 1 to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge: Cambridge University Press. IPCC. (2011). Special Report on Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation: Fact Sheet. [Online]. Available: http://ipcc-wg2.gov/SREX/ IPCC. (2012). Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation. A Special Report of Working Groups I and II of the Intergovernmental Panel on Climate Change. Field, C.B., Barros, V., Stocker, B., Qin, D., Dokken, D., Ebi, K.L., Mastrandrea, M.D., Mach, K.J., Plattner, G.K., Allen, S.K., Tignor, M., and P.M. Midgley (Eds.) Cambridge: Cambridge University Press. Jones, R., and S. J. Mason. (2012). Climate and weather information services for Humanitarian Operations, unpublished World Meteorological Organization report. Knutson, T., McBride, J., Chan, J., Emanuel, K., Holland, G., Landsea, C., Held, I., Kossin, J., Srivastava, A. and M. Sugi. (2010). Tropical cyclones and climate change. Nature Geoscience, 3, 157 – 163. Mehta, V., Meehl, G., Goodard, L., Knight, J., Kumar, A., Latif, M., Lee, T., Rosati, A. and D. Stammer. (2011). Decadal Climate Predictability and Prediction: Where are We? BAMS, May, 637 – 640. Meineke, H., Nelson, R., Kokic, P., Stone, R., Selvaraju, R., and W. Baethgen. (2006). Actionable climate knowledge: from analysis to synthesis. Climate Res., 33, 101–110. Mercer, J., Kelman, I., Taranis, L. and S. Suchet-Pearson. (2010). Framework for integrating indigenous and scientific knowledge for disaster risk reduction. Disasters, 34 (1), 214 – 239. Mercer, J. and I. Kelman. (2008). Living with Floods in Singas, Papua New Guinea. In Indigenous Knowledge for Disaster Risk Reduction: Good Practices and Lessons Learned from Experiences in the Asia-Pacific Region. Bangkok: UNISDR. Met Office. (2012). Tropical Cyclones. [Online]. Available: http://www.metoffice.gov.uk/weather/tropicalcyclone McBean, G. (1999). Forecasting in 21st Century. Lecture given to the International Meteorological Organization on behalf of the Meteorological Service of Canada. McBean, G. and C. Rodgers. (2010). Climate hazards and disasters: the needs for capacity building. Wiley Interdisciplinary Reviews – Climate Change, 1, 871-885. Mouillot, F., and C. Field. (2005). Fire history and the global carbon budget: a 1◦° 1◦ fire history reconstruction for the 20th century. Global Change Biology, 11, 398–420. Narain, S., Ghosh, P., Saxena, N., Parikh, J. and P. Soni. (2009). Climate Change Perspectives from India. Delhi: UNDP.

57

Naranjo, L. (2005). An Uncommon Experience in EWS: Hurricane Warning in Cuba. Presentation at World Conference on Disaster Reduction, 18 – 22 January, Kobe, Japan. New York Times. (2009). Obama’s Speech on Climate Change. [Online]. Available: http://www.nytimes.com/2009/09/23/us/politics/23obama.text.html?pagewanted=all NOAA. 2012. Tropical Cyclone Forecasting. [Online]. Available: http://www.aoml.noaa.gov/hrd/tcfaq/tcfaqF.html Pall, P., Aina, T., Stone, D.A., Stott, P.A., Nozawa, T., Hilberts, A., Lohmann, D., and M. Allen. (2011). Anthropogenic greenhouse gas contribution to flood risk in England and Wales in autumn 2000. Nature, 470, 382-385 Pechony, O. and Shindell, D. (2009). Fire parameterization on a global scale. Journal of Geophysical Research, 114, doi:10.1029/2009JD011927. Peduzzi, P., Chatenoux, B., Dao, H., De Bono, A., Herold, C., Kossin, J., Mouton, F., and O. Nordbeck. (2012). Global trends in tropical cyclone risk. Nature Clim. Change, 2, 289 – 294. Purcell, A. (2011). Somalia: How the UN Defines a Famine. New Scientist. [Online]. Available: http://www.newscientist.com/blogs/shortsharpscience/2011/07/how-the-un-defines-a-famine.html Scaife, A., Spangehl, T., Fereday, D., Cubasch, U., Langematz, U., Akiyoshi, H., Bekki, S., Braesicke, P. Butchart, N. Chipperfield, M., Gettelman, A., Hardiman, S., Michou, M., Rozanov, E. and T.G. Shepherd. (2011). Climate Change Projections and Stratosphere-Troposphere Interaction. Climate Dynamics, doi 10.1007/s00382-011-1080-7 Sen, A. (1999). Development as Freedom. Oxford: Oxford University Press. Singh, A. and V.F. Grasso. (2008). Global Environmental Alert Service. Advances in Space Research, 41 (11), 1836-1852. Stott, P., Myles, A., Christidis, N., Dole, R., Hoerling, M., Huntingford, C., Pall, P., Perlwitz, J., and D. Stone. (2011). Attribution of Weather and Climate-Related Extreme Events. Paper presented at WCRP OSC, Climate Research in Service to Society, 24-28 October, Sheraton Denver Downtown Hotel, Denver, CO, USA. Subbiah, A. and S. Fakhruddin. (2012). Long Lead Flood Forecast Technology for Disaster Preparedness. Thailand: RIMES. Svoboda, M., LeComte, D., Hayes, M., Heim, R., Gleason, K., Angel, J., Rippey, B., Tinker, R., Palecki, M., Stooksbury, D., Miskus, D., and S. Stephens. (2002). The Drought Monitor. Bulletin of the American Meteorological Society, 83 (8), 1181–1190. Svoboda, M. (2009). Drought Early Warning Systems. Presented at Inter-Regional Workshop on Indices and Early Warning Systems for Drought, 8 – 11 December, Lincoln, NE, USA. Tacko Kandji, S., Verchot, L., and J. Mackensen. (2006). Climate Change and Variability in the Sahel Region: Impacts and Adaptation Strategies in the Agricultural Sector. Nairobi: UNEP and ICRAF. Trenberth, K.E., Jones, P.D., Ambenje, P., Bojariu, R., Easterling, D., Klein Tank, A., Parker, D., Rahimzadeh, F., Renwick, J.A., Rusticucci, M., Soden, B. and P. Zhai. (2007). Observations: Surface and atmospheric climate change. Climate change 2007: The physical science basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge: Cambridge University Press.

58

UK Environment Agency. (2012). Flood Detection and Forecasting. [Online]. Available: http://www.environment-agency.gov.uk/research/policy/107118.aspx UN. (2005). In Larger Freedom: Towards Security, Development and Human Rights For All. New York: United Nations. UN. (2006). Global Survey of Early Warning Systems: An assessment of capacities, gaps and opportunities towards building a comprehensive global early warning system for all natural hazards. New York: United Nations. UNEP. (2011). Food Security in the Horn of Africa: The Implications of a Drier, Hotter and More Crowded Future. [Online]. Available: http://na.unep.net/geas/newsletter/Nov_11.php UNISDR. (2008). Climate Change and Disaster Risk Reduction, Briefing Note 1. [Online] Available: http://www.unisdr.org/eng/risk-reduction/climate-change/docs/Climate-Change-DRR.pdf. UNISDR. (2009).Global Assessment Report on Disaster Risk Reduction. Geneva: United Nations. UNISDR. (2011). Global Assessment Report on Disaster Risk Reduction. Geneva: United Nations International Strategy for Disaster Risk Reduction. UNISDR. (2012). PPP Days 2012 hosts first business forum. [Online] Available: http://www.unisdr.org/archive/25316 Van Oldenborgh, G., Doblas-Reyes, F., Wouters, B., and Hazeleger, W. (2012). Skill in the trend and internal variability in a multi-model decadal prediction ensemble, Climate Dynamics, in press. Available: http://www.knmi.nl/publications/ Vera, C., Barange, M., Dube, O., Goddard, L., Griggs, D., Kobysheva, N., Odada, E., Parey, S., Polovina, J., Poveda, G., Seguin, B. and K. Trenberth. (2010). Needs assessment for climate information on decadal timescales and longer, Procedia Environ. Sci., 1, 275-286. Ververs, M. (2012). The East African food crisis: Did regional early warning systems function? The Journal of Nutrition, 142, 131-133. Williams, A., and C. Funk. (2011). A westward extension of the warm pool leads to a westward extension of the Walker circulation, drying eastern Africa. Climate Dynamics, in press. Available: http://www.geog.ucsb.edu/~williams/publications/WilliamsAndFunk_2011_ClimateDynamics.pdf WMO. (2006). Drought monitoring and early warning: concepts, progress and future challenges. Geneva: World Meteorological Organization. WMO. (2009). Integrated Flood Management: Concept Paper. Geneva: World Meteorological Organization. WMO. (2011). Climate Knowledge for Action: a Global Framework for Climate Services – empower the most vulnerable. Geneva: World Meteorological Organization. WMO. (2011b). Key Issues and outcomes. Meeting on Sector Applications, 15 December 2011, WMO Headquarter, Geneva. WMO. (2011c). Good practices for multi-hazard Early Warning Systems (EWS). MeteoWorld. [Online] Available: http://www.wmo.int/pages/publications/meteoworld/mhew_en.html

59

8. Appendix 1. List of Individuals Consulted The following experts were contacted for this report: o Dr. Rob Allan (Atmospheric Circulation Reconstructions over the Earth, UK Met Office) o Professor Myles Allen (Head of Climate Dynamics Group, University of Oxford) o Professor Jonathan Baillie (Conservation Programmes Director at Zoological Society London) o Dr. Chris Bone (Department of Geography, University of Oregon) o Dr. Ben Collen (Head of Indicators & Assessments Unit, Institute of Zoology, Zoolgical Society of London) o Dr. Paul Davies (UK Met Office) o Dr. Bill DeGroot (Canadian Forest Service) o Dr. S.H.M. Fakhruddin (Team Leader, Hydrology, RIMES) o Dr. Paolo Fiorucci (Wildland Fire, CIMA Research Foundation) o Dr. Mike Flannigan (Canadian Forest Service) o Dr. Steve Foreman (UK Met Office) o Dr. Justin Ginnetti (UNISDR) o Dr. Micky Glantz (Consortium for Capacity Building, INSTAAR/ University of Colorado) o Dr. Richard Graham (UK Met Office) o Dr. Johann Goldammer (Global Fire Monitoring Center) o Dr. Claire Goddess (University of East Anglia) o Dr. John Harding (UNISDR) o Dr. Chris Huntingford (Centre for Ecology and Hydrology, UK) o Dr. Ilan Kelman (CICERO) o Dr. Jim Kossin (NOAA) o Paul Kovacs (Executive Director, Institute for Catastrophic Loss Reduction) o Dr. Thom Knutson (Geophysical Fluid Dynamics Laboratory, Princeton Univeristy) o Dr. Chris Little (UK Met Office) o Mark Lynas (Author and advisor on climate change to the President of the Maldives) o Professor Gordon McBean (Director, Policy Studies, Institute for Catastrophic Loss Reduction, University of Western Ontario) o Dr. Simon Mason (IRI, Columbia University) o Elizabeth L. McLean (University of Rhodes Island) o Dr. Lino Meranjo (Cuban meteorologist) o Professor Tim Palmer (Royal Society Research Professor in Climate Physics and Professorial Fellow, University of Oxford) o Professor Richard Peltier (Director of the Centre for Global Change Science, University of Toronto) o Dr. Iván J. Ramirez (Visiting Assistant Professor of Geography/Environmental Studies, New College of Florida) o Professor Ted Shepherd (Atmospheric Physics Group, University of Toronto) o Dr. Jim Verdin (USGS, NOAA, FEWS NET) o Dr. David Walland (Bureau of Meteorology, Australia) o Dr. Francis Zwiers (Pacific Climate Impacts Consortium, University of Victoria)

60