AGRICULTURE GLOBAL PRACTICE TECHNICAL ASSISTANCE PAPER MANAGING VULNERABILITY AND BOOSTING PRODUCTIVITY IN AGRICULTURE THROUGH WEATHER RISK MAPPING A GUIDE FOR DEVELOPMENT PRACTITIONERS Carlos Arce and Edgar Uribe WORLD BANK GROUP REPORT NUMBER 92400 FEBRUARY 2015 AGRICULTURE GLOBAL PRACTICE TECHNICAL ASSISTANCE PAPER MANAGING VULNERABILITY AND BOOSTING PRODUCTIVITY IN AGRICULTURE THROUGH WEATHER RISK MAPPING A Guide for Development Practitioners © 2015 World Bank Group 1818 H Street NW Washington, DC 20433 Telephone: 202-473-1000 Internet: www.worldbank.org Email: feedback@worldbank.org All rights reserved This volume is a product of the staff of the World Bank Group. The findings, interpretations, and conclusions expressed in this volume do not necessarily reflect the views of the Executive Directors of World Bank Group or the governments they represent. The World Bank Group does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of World Bank Group concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Rights and Permissions The material in this publication is copyrighted. Copying and/or transmitting portions or all of this work without permission may be a violation of applicable law. World Bank Group encourages dissemination of its work and will normally grant permission to reproduce portions of the work promptly. For permission to photocopy or reprint any part of this work, please send a request with complete information to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA, telephone: 978-750-8400, fax: 978-750-4470, http://www.copyright.com/. All other queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the Publisher, World Bank Group, 1818 H Street NW, Washington, DC 20433, USA, fax: 202-522-2422, e-mail: pubrights@worldbank.org. Cover Images from Left to Right: Drought-stressed corn. Photo: © CraneStation 2012. Map: Based on 1971 to 2000 data from Environment Canada, Alberta Environment and the U.S. National Climate Data Center. Harvesting corn in Kampong Cham, Cambodia. Photo: © Chhor Sokunthea/World Bank. CONTENTS Acknowledgments vii List of Acronyms ix Executive Summary xi Chapter One: Introduction 1 Chapter Two: Agrometeorology and Mapping 3 Agrometeorology 3 Climate 4 Soil 4 Terrain 5 Mapping 5 Chapter Three: Databases and Crop Models 7 Weather and Climate 7 Satellite 8 Interpolation 9 Reanalysis 10 Objective Analysis (Gridded Datasets) 11 Soil 11 Topography (Digital Elevation Models) 11 Crop Models 12 Example of an Empirical Model: Water Use Efficiency Model 13 Examples of Mechanistic (Dynamic and Deterministic) Models 13 Chapter Four: Historical Analyses 17 Climatologies 17 Global Climatologies from the Climate Research Unit 17 National Climatologies Based on Gridded Datasets 17 Regional Climatologies—Alberta, Canada 18 Hazard and Risk Maps 19 Plant Hardiness Zones 20 CONUS Hurricane Strike Density 21 Flood index—Mexico 21 Flood Damage Potential Map for the European Union 22 Maps of the Return Period of Agrometeorological Catastrophic Events in Mexico 23 Climate Regionalizations 24 Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping iii Chapter Five: Diagnostic and Forecasting Analyses 27 U.S. Drought Monitor 28 Meteorological Monitor (Australia) 28 Soybean Rust Outlook (U.S.) 30 Flood: Warning System (United States and the European Union) 30 Rainfall Seasonal Forecast (Australia and International) 33 Agrometeorological Bulletins 35 Chapter Six: Agro-Ecological Zones 37 Inventory of Land Use Types 38 Compilation of Land Resources Inventory 39 Climatic and Agro-Climatic Regionalizations (Optional) 39 Land Suitability 39 Agro-Ecological Zoning 40 Assessment 40 Chapter Seven: World Bank Case of Study: Agro-Ecological Zoning in Mozambique 41 Objectives 41 Methodology 41 Compilation of Land Resources 41 Zonal Mapping 42 Assess Land Suitability 44 Crop Vulnerability 47 Additional Examples of Agro-Ecological Zoning 48 Agro-Ecological Zones, Bangladesh 48 Agro-Ecological Zoning in the Ilave-Huenque Watershed of the Andean High Plateau 49 Agro-Ecological Zones, Their Soil Resource, and Cropping Systems—India 51 Chapter Eight: Conclusions 53 References 55 Appendix A: Example of Terms of Reference for Agro-Ecological Zones 59 Appendix B: Example of Terms of Reference for Gridded Analyses of Meteorological Variables 63 BOXES Box 4.1. Principal Components Analysis and Clustering 24 Box 7.1. Universal Kriging (UK) 43 FIGURES Figure 1.1. Basic Risk Management Cycle 2 Figure 3.1. Engineering Framework and Input Categories for DNDC to Generate Yield 14 Figure 4.1. World’s Climatology (1960–90) of Annual Precipitation (mm) from CRU 18 Figure 4.2. Climatology (1979–2008) of Average Temperature (°C) of Nicaragua Based on an Objective Analysis from the World Bank 18 iv Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping Figure 4.3. Climatologies (1971–2000) 19 Figure 4.4. U.S. Plant Hardiness Zone Map (Climatologies of Average Annual Extreme Minimum Temperature Classified at 5°F Intervals) 20 Figure 4.5. Total Number of Hurricane Strikes by U.S. Counties/Parishes/Boroughs, 1900–2010 21 Figure 4.6. Flood-Prone Regions of Mexico (Uribe Alcántara et al. 2010) 22 Figure 4.7. Map of Flood Damage Potential (Millions of Euros in Purchasing Power Parities) of the European Union 23 Figure 4.8. Maps of Return Period of Catastrophic Drought 24 Figure 4.9. North American Monsoon Regionalization 25 Figure 5.1. U.S. Drought Monitor for June 5, 2012 29 Figure 5.2. Examples of Maps of Daily Temperature (Left Panel) and Cumulative Precipitation (Right Panel) for Australia 29 Figure 5.3. Soybean Rust Observation from the Integrated Pest Management 31 Figure 5.4. Flood Monitor Based on Gauge Data 31 Figure 5.5. Flood Monitor of the EU Based on Gauge Data (May 23, 2010) 32 Figure 5.6. Forecast of Chance of Exceeding Trimestral Median Rainfall in Australia (Left Panel) and Historical Consistency of the Forecasting Scheme (Right Panel) 33 Figure 5.7. Trimestral Forecast (May–July, 2012) of Probability of Most Likely Category of Precipitation from the ECMWF 34 Figure 5.8. ENSO Forecast from Different Climatic Centers Around the World Collected and Published by IRI (Left Panel) and Probabilities of Having El Niño, La Niña, or a Neutral Year (Right Panel) 34 Figure 5.9. Map from the Cover of an Agrometeorological Bulletin 35 Figure 6.1. Flowchart of a General Agro-Ecological Zoning Process 38 Figure 6.2. Yield and Gaps of Potential, Attainable, and Actual Production 39 Figure 7.1. Monthly Average Accumulated Rainfall 42 Figure 7.2. Final Map Product 44 Figure 7.3. Crop Suitability Index for Selected Food and Cash Crops with Conventional Fertilization Shown and 2010 Weather Drivers 46 Figure 7.4. Crop Vulnerability Index (CVI) for Selected Food and Cash Crops with Conventional Fertilization and Climate Scenario 2 Shown Compared to 2010 Weather Drivers and Yield Baselines 48 Figure 7.5. CVI (Bottom) for Climate Scenario 2 and CSI (Top) Under Conventional Fertilizer for Key Crops in Mozambique Compared to 2010 Weather Drivers and Yield Baselines 49 Figure 7.6. All Crops Combined for the Scenario 1 (Left) and 2 (Right) are Highlighted for Conventional (Top) and Optimized (Bottom) CVI 50 Figure 7.7. AEZs of Bangladesh 51 Figure 7.8. AEZs of Ilave-Huenque Watershed 52 Figure 7.9. AEZs of India 52 Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping v TABLES Table 3.1. Characteristics of MERRA 10 Table 3.2. Characteristics of NARR 10 Table 6.1. Limiting Crop Factors and Their Relationship with Potential, Attainable, and Actual Yields 39 Table 7.1. Major Agro-Climate Cluster Names Identified from Bioclimatic and Agricultural Factors 45 Table 7.2. DNDC Crop Model Input Parameterization for Generating Crop Suitability Index 45 Table 7.3. Climate Scenarios for Crop Vulnerability Assessment 47 Table 7.4. Vulnerability of All Crops to Climate Change by Agro-Climatic Cluster (Region or Zone) 47 vi Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping ACKNOWLEDGMENTS This document was written by Carlos E. Arce (Senior Economist) and Edgar M. Uribe-Alcántara (Consultant) from the Agricultural Risk Management Team (ARMT) at the World Bank. The authors are grateful to peer reviewers Ademola Braimoh (Senior Natural Resources Management Specialist, World Bank) and Nathan Torbick (Applied Geosolutions) for their feedback and critical review of the document. The authors would like to thank Marc Sadler and Vikas Choudhary for their valu- able guidance and support. Similarly, our gratitude to Jesús Escamilla for providing us with material illustrating risk mapping applications of vulnerabilities in Mexican agriculture. This activity would not have been possible without the generous contribution from the Netherlands Ministry of Foreign Affairs and the Swiss State Secretariat of Economic Affairs (SECO). Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping vii LIST OF ACRONYMS Acronym Definition Acronym Definition AEC Agro Ecological Cell JRC Joint Research Centre AEZ Agro Ecological Zone LGP Length of Growing Period AHPS Advanced Hydrologic Prediction Service LST Land Surface Temperature ARMT Agricultural Risk Management Team LUT Land Use Types AVHRR Advanced Very High Resolution Radiometer MAM March–April–May CCD Cloud Cover Duration MAP Modeling Analysis and Prediction CFS Canadian Forest Service MERRA Modern Era Retrospective-Analysis for Research and Applications CPC Climate Prediction Center MODIS Moderate Resolution Imaging CRU Climatic Research Unit Spectroradiometer CSI Crop Suitability Index MSPEC Model Simulation of the Ecological Potential CSSWB Crop Specific Soil Water Balance of Crops CVI Crop Vulnerability Index NARR North American Regional Reanalysis DEM Digital Elevation Model NASA National Aeronautics and Space DJF December–January–February Administration DNDC Denitrification-Decomposition NOAA National Oceanic and Atmospheric DSMW Digital Soil Map of the World NWPM Numerical Weather Prediction Models ECMWF European Centre for Medium-Range Weather NWS National Weather Service Forecasts OK Ordinary Kriging ENSO El Niño-Southern Oscillation PAR Photosynthetically Active Radiation EU European Union PCA Principal Components Analysis FAO Food and Agriculture Organization RMSE Root Mean Squared Error GHG Green House Gas SBR Soybean Rust GIS Geographical Information Systems SOI Southern Oscillation Index GPI Goes Precipitation Algorithm SON September–October–November GRDC Global Runoff Data Centre SRTM Shuttle Radar Topography Mission GTS Global Telecommunications System TAMSAT Tropical Applications of Meteorology HADS Hydrometeorological Automated Data using SATellite data and ground-based System observations HWSD Harmonized World Soil Database TMI TRMM Microwave Imager IAMS Integrated Aerobiological Modeling System TRMM Tropical Rainfall Measuring Mission IPM PIPE Pest Management Pest Information Platform UK Universal Kriging for Extension and Education U.S. United States IRI International Research Institute USDA United States Department of Agriculture JJA June–July–August UTC Universal Coordinated Time Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping ix EXECUTIVE SUMMARY WEATHER RISK IN AGRICULTURE: OCCURRENCE, IMPACTS, AND VULNERABILITY Productivity in the agricultural sector is inherently dependent on weather, such as variations in rainfall and temperature. As a result, weather risk events can cause losses in yield and production that translate into economic losses for producers, as well as other sector stakeholders that depend on income from agricultural trade, transport, processing, or export. Extreme temperatures, floods, droughts, hailstorm, and wind- storms are just a few examples of weather risk events that cause major economic losses. In developing countries, weather risk is especially significant due to the impor- tance of the agricultural sector in the overall economy and its contribution to house- hold food security. Up to 90 percent of the population in many developing countries relies on agriculture for a living, since the sector is the primary source of income, employment, and food. Absent effective risk management strategies, weather shocks to agriculture in these countries have far-reaching effects on wellbeing, development, and poverty reduction. Developing countries also tend to be more vulnerable to weather risks owing to factors that constrain stakeholders’ abilities to manage risk. Since households and companies involved in agriculture typically operate with a low level of assets and lack access to well-developed insurance and credit markets, their risk management strategies are limited compared with stakeholders in developed countries. An addi- tional constraint is that weather risk occurrence tends to affect households and businesses in the same community, minimizing the effectiveness of traditional cop- ing mechanisms, such as loans or gifts from family members, on which rural com- munities rely. At the same time, the predicted impacts of climate change have only increased the importance of effectively managing weather risk. As a result of global warming, extreme weather events are expected to become more frequent and more intense. Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping xi In many countries, weather variability has already flexible to support a variety of assessments and decision- increased, making it more difficult for producers to pre- making frameworks. dict and respond to weather patterns. This document is a guide for development practitioners as to why they should undertake risk mapping in agri- WEATHER RISK MAPPING: culture and how to do it step-by-step, including available A RISK MANAGEMENT TOOL resources with pros and cons, uses of the findings, and Weather risk mapping for agriculture, or agro-meteorol- descriptions of the practical applications for various ogy risk mapping, is one way to manage weather risk. users. The introduction places weather risk mapping Agro-meteorology involves the application of meteoro- within the broader context of agricultural risk, explain- logical information and data to agriculture. Most farm ing how mapping can enable risk identification, assess- operations are weather sensitive (for example, fertiliza- ment and management activities, and each chapter tion, planting, movement of agricultural machinery, and elaborates on one or more of the technical components. harvesting) and offer better results when executed under A basic definition of agro-meteorology is provided, along the right weather conditions. Medium- and long-range with a discussion of different mapping techniques. The forecasts are expected to allow farmers to identify in guide presents the available remote (satellite) databases advance either optimum or adverse weather conditions. of agro-meteorological variables that can be used for the When used in combination with geographical informa- purpose of weather risk mapping, assessing the advan- tion systems (GIS), agro-meteorology is also able to pro- tages and drawbacks of each database and their suitabil- vide spatially related information in map format, which ity for different purposes. The authors review current can be very useful for farmers. risk mapping analyses based on historical weather obser- vations, which are typically used for risk identification Weather risk mapping can provide historical (past), and assessment, including climatologies, hazard and diagnostic (present) and prognostic (future) analyses of risk maps, climate regionalizations and agro-ecological weather patterns in a given zone, allowing agricultural zones (AEZ). sector stakeholders to better understand weather condi- tions. When applied as part of a systemic approach to The document also reviews forward-looking mapping weather risk management, this tool can be used strate- techniques, known as diagnostic and forecasting analy- gically to manage risk and optimize farm productivity. ses. Diagnostic analyses are designed to provide a tech- Weather risk mapping techniques are expected to enable nical description of the current risk situation and its more risk-informed planning of production, to facilitate causes. Forecasts, on the other hand, predict potential improved information for supply chain stakeholders on risk events based on the study and analysis of pertinent potential production risks to crops in given production observations and simulations. The specific examples of zones, and to help inform investment. mapping techniques for diagnostic and forecast purposes cited here are drawn from the United States, the Euro- This document strategically presents a variety of map- pean Union, and Australia, and are meant to illustrate ping techniques for agricultural risk management and the potential use of similar applications in global agri- illustrates the application of these techniques for inform- culture. Finally, the guide provides instruction on how ing public and private sector development strategies. The and why to conduct agro-ecological zoning, a technique authors do not intend to present a comprehensive list of that can be used to assess land-use types, land resources, techniques. Instead, this is a synthesis document that pro- land suitability, and climatic and agro-climatic regionali- vides an overview of existing techniques and insight on zations, as well as to inform land use recommendations. the suitability of techniques for designing agricultural risk The concluding chapter demonstrates a step-by-step management strategies. Many of the examples cited are application of agro-ecological zoning in a case study of designed to be cost effective, automatable, scalable, and Mozambique. xii Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping This volume is a product of the staff of the International Bank The World Bank does not guarantee the accuracy of the data for Reconstruction and Development/the World Bank. The included in this work. The boundaries, colors, denominations, findings, interpretations, and conclusions expressed in this and other information shown on any map in this work do not paper do not necessarily reflect the views of the Executive imply any judgment on the part of the World Bank concern- Directors of the World Bank or the governments they ing the legal status of any territory or the endorsement or represent. acceptance of such boundaries. The set of weather risk mapping techniques presented more difficult for development practitioners to catch up in this guide is illustrative of an increasingly useful set with current developments. The authors hope that this of tools for informing risk management strategies and document will help practitioners interested in the sub- investment decisions. The growing number and com- ject to familiarize themselves with the technical aspects plexity of approaches and models to design risk map- involved in the design of these products and their poten- ping in agriculture are together making it more and tial practical uses. Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping xiii CHAPTER ONE INTRODUCTION The increasing variety of publications related to practical geospatial applications to ana- lyze weather risks to the agricultural sector is currently overwhelming development prac- titioners as to their nature and suitability for the purpose of informing agricultural risk management decisions. This document strategically presents a variety of mapping tech- niques for agricultural risk management and illustrates the application of these techniques for informing public and private sector development strategies. The target group is devel- opment practitioners that need to start incorporating these techniques into their analytical framework to assess weather risks to investment projects. As with all guides, this document is not comprehensive and does not intend to be, but it illustrates in a simplified way a num- ber of current applications that are generally accepted for providing geo-referenced risk information in the agricultural sector. Many of the example techniques are designed to be cost effective, automatable, scalable, and flexible to support a variety of assessments and decision-making frameworks. It is hoped that this introduction to risk mapping in agricul- ture will illustrate a representative array of applications used today in various countries. Making agricultural risk maps would be relatively simpler if reliable historical databases of yields (spatially distributed) were available, as well as the causes of losses, and applied management practices. Unfortunately, these databases are usually not available, relia- ble, or sufficiently comprehensive in developing countries. In addition, adaptation, miti- gation, and management practices vary widely by geography and in response to stress. Therefore, in order to estimate yields, the usual approach is yield simulation through crop models, which try to emulate the response of the crop to climatological, man- agement, and edaphological conditions (in short, how soils influence land use). Crop models can reproduce the vulnerabilities of crops to hazards of interest, which usually requires additional information like vulnerability functions. One additional challenge is that for crops, unlike infrastructure (for example, buildings, homes, bridges), the vul- nerability is a function of the crop stage. Therefore, robust agricultural risk mapping requires an extensive range of simulations of different crops, stages, and hazards, as well as management responses, which at times can become very complex and laborious. Risk management analyses can be classified by their target time horizon so it seems also natural to classify agricultural mapping techniques by this criterion as well. Risk Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 1 FIGURE 1.1. BASIC RISK adverse events produce shocks to stakeholders (savings, MANAGEMENT CYCLE government handouts, sale of assets, migration, reduc- tion of consumption, and so on). Weather risk mapping techniques aim at producing information that is useful for Assess designing risk management strategies either for risk miti- gation, risk transfer products, or for risk coping. This document outlines agricultural risk management mapping techniques developed around the world and clas- Measure Evaluate sifies them in the three time horizons. The next chapters present examples of historical (past), diagnostic (present), and prognostic (future) analyses for illustrative purposes. It offers particular attention to the work developed by Manage the Agricultural Risk Management Team (ARMT) at the World Bank and colleagues, who have used modeling, remote sensing data, and geospatial analysis to derive diverse mapping techniques for assessing weather risks for the agricultural sector in developing countries, where data analysis usually starts with identification and assessment, are usually very limited. The last chapter presents several which are based on past events (historical analysis). His- agro-ecological zoning cases, including an account by the torical mapping techniques include: climatologies, risk World Bank in Mozambique. maps, hazard maps, climatological regionalizations, and agro-ecological zonings. These analyses present an idea Weather risk mapping techniques are expected to enable of what has happened in the past in order to offer an more risk-informed planning of production and facilitate expectation of possible events that can occur in the future. improved information for supply chain stakeholders on potential production risks to crops in given production Weather risk management strategies are defined once zones and also to help inform investment. This docu- the risk assessment is complete (figure 1.1). Strategies ment is a guide for development practitioners on why they are usually grouped in risk mitigation, risk transfer, and/ should undertake risk mapping in agriculture, a step-by- or risk retention (or absorption). Risk mitigation implies step approach to this exercise, the available resources with the development of exante activities that will reduce pros and cons, uses of the findings, and descriptions of the the impact of risks once an adverse event happens. Risk practical applications for various users. After this intro- mitigation seeks to find a balance between the negative duction, the second chapter explains the basic concepts impacts of risk and the positive benefits of the activity. of agro-meteorology and mapping techniques. The third Examples of risk mitigation are the typical good agri- chapter introduces the available remote (satellite) data- cultural practices as they relate to risks (that is, drought bases of agro-meteorological variables that can be used tolerant seeds, supplemental irrigation, fertilizations and/ for the purpose of weather risk mapping, assessing the or mulching, soil drainage, pesticides, water harvesting, advantages and disadvantages of each database for dif- and so on). Risk transfer involves sharing with another ferent purposes. The fourth chapter presents a review of party the burden of loss and the benefit of gain (for exam- current risk mapping analyses based on historical weather ple, insurance, risk pools). Finally, there is risk retention. observations, illustrating the various products and uses, Risks that are not avoided or transferred are retained, while chapter 5 reviews numerous mapping techniques and stakeholders need to absorb or cope (expost). The for diagnostic and forecasting purposes. The concept literature also classifies risk strategies in formal and infor- and explanation of agro-ecological zoning is addressed mal risk coping mechanisms that households, communi- in chapter 6, and its step-by-step application for Mozam- ties, and governments use in order to cope with risks as bique is illustrated in chapter 7. 2 Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping CHAPTER TWO AGROMETEOROLOGY AND MAPPING AGROMETEOROLOGY It is important to differentiate between the two central concepts of weather and cli- mate. Weather refers to the atmospheric short-term conditions (hours, days, or weeks) in a given place. Climate, on the other hand, refers to the expected conditions of the weather. Climate is usually estimated as the average value of a given meteorologi- cal variable (for example, precipitation, temperature). The weather is studied by the meteorology, while the climate, by the climatology. The factors that condition the climate in a given place include: 1. Latitude 2. Altitude 3. Continentality 4. Topography 5. Vegetation 6. Land and water distribution 7. Soil type 8. Sea currents The climate and weather largely influence crop systems and their yields. Climate plays a critical role in the determination of the more appropriate zones for agricultural development. The three most important elements from the point of view of crop development are: light, temperature, and precipitation. Additional variables such as wind, hurricanes, and hail can be crucial in specific areas. Such elements, along with carbon dioxide and oxygen, integrate the climatic and weather factors that drive agri- cultural yield, and that are practically outside human control (excluding irrigation accessibility). Additional factors that influence crop yields but that can be managed by humans include: crop selection, farming practices, soil fertility, irrigation systems, and control of plagues and diseases. Agrometeorology is defined as the science that investigates the relevant meteoro- logical, bioclimatologic and hydrological conditions for the processes and elements Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 3 related to the agricultural production. Agrometeorol- Crops also require certain temperature conditions. Frosts ogy assists farmers in the efficient use of the physical or extreme heat waves, for example, can be lethal for crops. environment so the agricultural production improves in Therefore, thermal regimes can be defined for a given terms of quantity and quality while maintaining the sus- region in order to identify suitable conditions for crop tainability of natural resources. In order to achieve its development. On the other hand, crops also require certain goals, agrometeorology follows a four-stage procedure. amounts of heat. This requirement is usually expressed in Firstly, it formulates a description of the environment terms of reference temperatures, which are calculated by and its biological responses. Secondly, it interprets the accumulating daily average temperature in excess of tem- biological responses in terms of the physical environ- perature thresholds to derive growing degree day metrics. ment. In a third stage, it generates agrometeorological (Growing degree days are metrics of agricultural output, forecasts. Finally, it develops services, strategies, and also as a function of mean temperature. The computation support systems for tactical and strategic decisions in of degree days involves certain threshold temperatures, for the field. example, 65°F for heating and cooling degree days. These thresholds are referred to as base temperatures.) Agrometeorology involves the application of meteoro- logical information and data to agriculture. Most farm The length of growing period (LGP) combines the clima- operations are weather sensitive (for example, fertiliza- tological factors necessary for crop development in a single tion, planting, movement of agricultural machinery, and concept. The LGP identifies the climatic season where both harvesting). These operations offer better results when moisture and temperature conditions are suitable for crop they are executed under the right weather conditions. production. The resulting period provides a framework dur- Medium- and long-range forecasts are expected to allow ing which climate elements, such as temperature, precipita- farmers to identify in advance either optimum or adverse tion, and climatic hazards become more relevant for crop weather conditions. When used in combination with geo- production. The estimation of the beginning of the grow- graphical information systems (GIS), agrometeorology is ing period is associated with the onset of the rainy season. also able to provide spatially related information in map A general rule defines this as occurring when precipitation format, which can be very useful for farmers. However, in exceeds half of evapotranspiration. Growing period zones order to generate maps, it is required to have information can be plotted on a map at fixed intervals of mean LGP or that describes the agrometeorological system. The system at a given level of probability. In addition to these important includes climate, soil, terrain, and land use, for example climatological factors, there are also other factors to consider (these components are described in following chapters). including frost-free periods, multi-cropping suitability, and Since in many developing countries the information rela- aridity indices (precipitation divided by ET). tive to these components is quite limited, some databases and crop models (proxies for yield data) are provided in SOIL the following chapter. The soil works as a reservoir of water and nutrients for plant growth. In that sense, soil properties relevant for CLIMATE production are related to the capacity of the soil to hold Crop development depends on water availability (soil and provide water, which in turn depend upon soil param- moisture), nutrients, management, and solar radiation, eters, such as water retention capacity, rooting depth, field among other factors. Soil moisture is not usually mea- capacity, permanent wilting point, and soil water deple- sured systematically so it is indirectly estimated from tion factor. In addition to soil properties related to mois- climatic variables that help quantify components of the ture, there are additional soil qualities that affect crop water cycle. Precipitation, for example, is the most criti- production such as: nutrient availability, nutrient retention cal source of water for rain-fed crops. Evapotranspiration capacity, rooting conditions, oxygen availability to roots, (the loss of humidity) is not usually measured either so it excess salts, toxicity, and workability. These qualities are is usually estimated from other variables like temperature, measured through soil attributes like: texture, organic car- solar radiation, and wind. bon, pH, exchangeable bases, base saturation, exchange 4 Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping capacity of soil and of clay fraction, coarse fragments, vertic properties, phases, depth, volume, salinity, sodicity, MAPPING content of calcium carbonate and gypsum, and drainage Maps are a visual representation of an area with symbolic capabilities. Crop models usually take into account soil depictions highlighting relationships between elements of conditions, so simulated yields also seek to take these fac- that space such as objects, regions, and themes (for exam- tors into account. ple, crops, climatologies, soil classification, and so on). In maps, the aspects of the image are analogies of values related to the information presented. Maps can be very TERRAIN useful in agriculture because they can be related to assets Slope is strongly associated with sustainability. Strong (crops) for decision making and assessment. This docu- slopes favor erosion, which renders cropping unsustainable ment shows examples of maps that are considered useful in the long term. The application of some practices also for risk management in agriculture. However, in order to depends on terrain, such as whether mechanized equip- be able to create maps, information is required that often ment can be used on sloping land. Strong slopes also favor is not available in developing countries. Therefore, it is fertilizer loss. Since erosion also depends on precipitation, important to identify suitable databases for poorly docu- crop suitability associated with terrain also depends on mented regions. The next chapter provides examples of precipitation distribution. Usually, slope values are used to databases and models that can be used in developing divide a region into suitable and unsuitable zones. countries. Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 5 CHAPTER THREE DATABASES AND CROP MODELS If spatial databases of climate, yield, and the causes of losses were readily avail- able, perhaps map-making for risk management would be a more straightforward task. However, most developing countries lack reliable, consistent, or sufficiently extensive databases suitable for risk analysis. Therefore, risk analysis in these regions relies partially in databases proxies and surveys. Yields, for example, have to be esti- mated using crop models. In those cases, the issue of finding yield data is traded for an issue of simulation, calibration, and finding models and input data appropriate for the simulation of yield and the impact of hazards. Solving these issues is not an easy task. Most developing countries, for example, do not have reliable, shareable, and appropriate soil and climate databases suitable for crop models. However, there is growing experience in the use of climate data proxies (for example, satellites and reanalysis). Additionally, there are also national and global efforts underway to map agro-ecological resources. This information can be used along with crop models to estimate yields for agricultural risk analysis. In this section, we describe public and globally available climate, topography, and soil databases, as well as crop models. Although the discussion is focused on proxies, it is important to understand that developing countries may have useful databases. For example, most countries have a network of weather stations. These observations are usually more reliable than proxies, and they should be used whenever possible. However, the local databases usually have low coverage, a good portion of missing data, or poor quality, so their application can be very limited unless they are combined with other sources of information, like proxies. WEATHER AND CLIMATE There are several types of weather proxies. In this section, we provide examples derived from the following types: satellite, reanalysis, interpolation, and objective analysis. Satellite proxies rely on observations from space, while reanalysis is based on Numerical Weather Prediction Models (NWPM). Interpolation relies on observation, while objective analysis combines a proxy (for example, satellites) with observation. There is no single best dataset. The user needs to ensure the spatial and temporal Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 7 coverage and resolutions are suitable for the specific Imager (TMI) instrumentation to provide quantitative requirements. On the other hand, most products perform rainfall information. TMI quantifies water vapor, cloud better in some regions than others, so an evaluation of the water, and rainfall intensity in the atmosphere by assess- product should be part of the selection criteria. ing microwave energy emitted in the lower boundary layer. The TRMM Ground Validation program oper- The most important issues associated with climate prox- ates from the Goddard Space Flight Center of the U.S. ies are perhaps accuracy and resolutions. Generally, prox- National Aeronautics and Space Administration (NASA). ies use readily available observational data as input or for This team uses gauge, station, field, and modeled data calibration purposes. However, this information is usually to calibrate and validate precipitation. These ground very limited, particularly in developing countries. It is data are used to improve the rain rate interpretation therefore recommended to evaluate the accuracy of prox- algorithm. Wolff et al. (2005) reported monthly rainfall ies in the region of interest, and to calibrate or make adjust- accumulation scheme of TRMM matches ground pre- ments whenever necessary. In addition to the description cipitation measurements within ±5 percent. This dataset of some climate proxies, this document includes a brief is publically available and starts in 1998. TRMM data description of a methodology that has been implemented are available only between latitudes 50°N and 50°S. Sev- by the Agricultural Risk Managment Team (ARMT) to eral products are derived from TRMM with different calibrate proxies using weather stations (objective analyses time and space resolutions. The highest spatial resolution or gridded analysis). This methodology not only allows is 0.25 (~27 km), and the highest temporal resolution is higher accuracy but also higher resolutions. three hours. The following is by no means an exhaustive list of the Moderate Resolution Imaging Spectroradi- plethora of climatological databases but represents a selec- ometer (MODIS)—Land Surface Temperature tion of databases based upon four criteria: (1) importance (LST): Land surface thermal data can be obtained from for developing countries, (2) timeliness, (3) representative- the MODIS instrument onboard the Terra (descending) ness, and (4) availability (most are publically available). and Aqua (ascending) satellite platforms. The MODIS sensor collects measurements in 36 spectral bands every 1–2 days at native spatial resolutions of 1 km for LST SATELLITE products. MODIS products are available from the Ware- The advantage of weather satellite data is that they usually house Inventory Search Tool at the Land Processes Dis- provide complete databases for the period and domain of tributed Active Archive Center. MODIS has different interest (some missing data are infrequent but possible due products and resolutions. The MODIS MYD11C3 LST, to instrumental errors or inappropriate environmental for example, has a spatial resolution (pixel size) of 0.05 conditions). The disadvantages include inaccuracy (this degree, which is equivalent to approximately 5 km. The needs to be addressed locally), short-term records (usually MODIS LST products have been extensively refined starting in the 1980s), a small portion of missing data, and and validated to provide precise and accurate informa- changes in methodologies, which results in inconsistent tion. LST accuracy is reported to be within 1K under databases. Perhaps the most important selection criteria clear-sky conditions in a range of ecosystems (Wan for these products are temporal coverage (again, most of 2008; Wan et al. 2002). Validation across multiple sites them start in the 1980s) and accuracy, which has to be incorporating wide ranging ecosystems and atmospheric addressed locally (for example, comparing ground obser- conditions has consistently shown that MODIS LST vational data with satellite data). products are within ±1 km in the range 263-322 lm. Fur- ther, comparisons between V5 LSTs and in-situ values in Tropical Rainfall Measuring Mission (TRMM): 47 clear-sky cases (in the LST range from −10°C to 58°C TRMM precipitation estimates are based on both active and atmospheric column water vapor range from 0.4 to and passive microwave instruments. TRMM employs the 3.5 cm) indicate that the accuracy of the MODIS LST Visible and Infrared Scanner and TRMM Microwave product is better than 1 km in most cases (39 out of 47), 8 Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping and the root of mean squares of differences is less than CPC (Climate Prediction Center)—RFE (Rainfall 0.7 km for all 47 cases or 0.5 km for all but the eight cases Estimate): A combined gauge-satellite precipitation apparently with heavy aerosol loadings. Thus, the preci- dataset has been developed as part of the Famine Early sion, accuracy, and reliability of MODIS LST makes it Warning System Network (FEWS-NET) for Africa and well-suited for assessing land surface temperature zones was upgraded from version 1.0 to version 2.0 in 2001. in data poor regions. Finer scale remote sensing imag- Version 1.0 (Herman et al. 1997) covers the period 1995– ery, such as from Landsat, can provide field level ther- 2000 while version 2.0 from 2000 to the present. The mal information that when fused with MODIS provides rainfall algorithm (RFE version 1.0) uses infrared tem- a more thorough assessment of spatiotemporal trends in perature satellite data (METEOSAT), rain gauge data, indices useful for mapping agricultural conditions and and modeled wind and relative humidity data to compute crop status at key stages. 10-day rainfall estimates. METEOSAT infrared tempera- ture data are first used to compute estimated rainfall via TAMSAT TARCAT v1.0: The TAMSAT group pro- the Goes Precipitation Algorithm (GPI). Modeled rela- duces the TARCAT v1.0 data sets. It includes a 10-daily tive humidity and wind data are then compared to topo- (dekadal), monthly, and seasonal rainfall estimates for graphical data to estimate cross-terrain flow as orographic Africa. The Cloud Cover Duration (CCD) method is precipitation. These two estimates are then compared to applied using Meteosat thermal infrared channels. It is Global Telecommunications System (GTS) rain gauge based on the recognition of convective storm clouds, and measurements. Calibration is performed to remove bias is calibrated against ground-based rain gauge data. Data and create the final rainfall estimate. The merging process covers the period 1983–2010. For some annual/dekadal allows the final rainfall estimates to have the magnitude of combinations, no data are available due to corrupt sat- the station data, with the shape of the precipitation field ellite images. Over the past 15 years, a number of cali- determined by the satellites. The RFE 2.0 is generated bration workshops at regional centers have been carried from the Advanced Microwave Sounding Unit, the Spe- out by TAMSAT, resulting in good calibrations for the cial Sensor Microwave/Imager and GPI (infrared) pre- main rainy seasons for East Africa (except Somalia and cipitation estimates using a maximum likelihood method. Djibouti); for all Southern Africa (except Tanzania, Mau- They are then merged with daily rain gauge data from up ritius, Madagascar and Angola); and for all of West Africa to 1,000 GTS stations in Africa (although typically, the south of 25°N. Because of the involvement of national number of stations is closer to 500 owing to erroneous weather services, these calibrations have been performed station data and/or poor station maintenance). There are with much more data than are generally internationally significant differences between RFE version 2.0 and 1.0 available. Recent validation experiments (Dinku et al. (RFE version 2.0 uses passive microwave estimates while 2007) have shown that careful local calibration of CCD RFE version 1.0 includes a procedure to estimate warm in this manner produces more reliable rainfall estimates orographic rain), leading to possible biases in the com- than more sophisticated systems which rely on global bined operational series. calibrations or other parameterizations. The same studies also showed that these local calibrations are stable in time, INTERPOLATION1 lending weight to the idea of pre-calibrating with historic These datasets are created through the interpolation data. This methodology is used by the AGRHYMET of data from meteorological stations. This analysis can regional center and by a number of African meteorologi- be made at different temporal scales (for example, daily cal services to provide vital, up-to-the-minute information and monthly). Advantages include accuracy associated on the state of the rainy season. In a separate operational with observational data and completeness. Disadvan- service, TAMSAT provides dekadal, monthly, and sea- tages include the requirement of observational data, low sonal rainfall totals and anomalies to the European Com- resolution, and the fact that most traditional methods mission Joint Research Centre FOODSEC Action. Since May 2010, TAMSAT rainfall estimates have been avail- 1 Method of constructing new data points within the range of a discrete set of able via GEONETCast. known data points. Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 9 arbitrarily assign zeros when no nearby station is avail- TABLE 3.1. CHARACTERISTICS OF MERRA able. This methodology, however, is quite appropriate for Element Characteristics climatological products. Pixel size dx = 0.6666° (~72 km); dy = 0.5000° (~55 km) Global Climatologies from the Climate Research Time reference Universal Coordinated Unit (CRU): CRU has developed climatologies at pixel Time (UTC) sizes of 0.5° (~55 km) and 10° (~18 km). The dataset con- Temporal resolution 1hr tains eight mean (1961–90) monthly and annual values Initial Date January 1st, 1979 of precipitation, wet-day frequency, temperature, diurnal Final Date Semi-Current temperature range, relative humidity, sunshine duration, Geographical Reference Geographic ground frost frequency, and wind speed. The dataset is System created through the interpolation of station means. The Coordinates of the lower Latitude: 90°S; data are available through the International Water Man- left corner (center of Longitude: 180°W agement Institute World Water and Climate Atlas (http:// pixel) www.iwmi.org) and the Climatic Research Unit (http:// Coordinates of the Latitude: 90°N; www.cru.uea.ac.uk). upper right corner Longitude: 180°E (center of pixel) REANALYSIS TABLE 3.2. CHARACTERISTICS OF NARR Reanalysis methods rely on historical databases of simu- Element Characteristics lations with hundreds of variables in four dimensions (4D; space plus time). Latest generation reanalysis, such Pixel size dx = dy = 0.1875° (~20 km) as Modern Era Retrospective-analysis for Research and Time reference Universal Coordinated Time (UTC) Applications (MERRA; see the next section), already Temporal resolution 3hr (0–3, 3–6, 6–9, 9–12, includes satellite data as input in simulations. Reanalyses 12–15, 15–18, 18–21 and are based on highly sophisticated NWPMs. Their devel- 21:00–0:00UTC every day) opment requires large multi-disciplinary teams as well as Initial Date January 1st, 1979 powerful computational resources, particularly necessary Final Date Semi-Current (five days lag) for processing and storage. These models can be used for Geographical Reference Geographic forecasting and historical purposes. System Coordinates of the lower Latitude: 0.000°N; Modern Era Retrospective-analysis for Research left corner (center of Longitude: –220.000°E and Applications (MERRA): MERRA is the highest pixel) quality, latest generation reanalysis available worldwide. It Coordinates of the Latitude: 89.625°N; covers the modern era of remotely sensed data, from 1979 upper right corner Longitude: –0.625°E to near real-time, and the special focus of the atmospheric (center of pixel) assimilation is the hydrological cycle. MERRA devel- ops products with several frequencies and resolutions. models; Mesinger et al. 1988). NARR is only available for Table 3.1 summarizes the characteristics of the finest res- the northern hemispheric portion of the American con- olution MERRA products. tinent (including Mexico and most of Central America). NARR was developed by the National Center for Envi- North American Regional Reanalysis (NARR): ronmental Prediction of the National Oceanic and Atmo- This is a long-term, dynamically consistent, atmospheric spheric Administration (NOAA). The characteristics of and hydrologic database, with high spatial and tempo- NARR are indicated in table 3.2. The NARR dataset is ral resolutions, generated with the numerical weather freely available from: http://www.emc.ncep.noaa.gov/ model eta (eta stands for the Greek letter, which rep- mmb/rreanl/. Table 3.2 shows the characteristics of resents an alternative height variable in atmospheric NARR. 10 Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping OBJECTIVE ANALYSIS (GRIDDED SOIL DATASETS) Harmonized World Soil Database (HWSD; FAO All the previous climate datasets are proxies estimated 2009): The HWSD is the next iteration of the Digital from satellites, atmospheric models, the combination of Soil Map of the World (DSMW) from the Food and Agri- both satellites and models, or interpolation. These prox- culture Organization (FAO). It is the most updated and ies are usually calibrated with observational data available highest resolution product available worldwide. It is a in semi-real-time. However, only a small portion of the 30 arc-second raster database with over 16,000 different databases are shareable in real-time by the local national soil mapping units that combines existing regional and weather services so this information is usually very lim- national updates of soil information worldwide (SOTER, ited. However, it is possible to develop regional weather ESD, Soil Map of China, WISE) with the information data grids through the combination of reanalysis and the contained within the 1:5,000,000 scale FAO-UNESCO most updated and complete version of the local meteoro- DSMW. It contains the composition in terms of soil logical database, directly obtained from the local weather units and the characterization of selected soil parameters services. The gridded datasets can be created following (organic carbon, pH, water storage capacity, soil depth, a methodology extensively used in the atmospheric sci- cation exchange capacity of the soil and the clay fraction, ences, which is known as objective analysis; more spe- total exchangeable nutrients, lime and gypsum contents, cifically, a successive correction method called Cressman sodium exchange percentage, salinity, textural class and (Cressman 1959). The procedure allows the calibration granulometry). of a proxy based on the observational data. In addition to the reduction of errors associated with the calibration, the advantage of developing regional gridded analysis is that, unlike the global datasets presented earlier, the char- TOPOGRAPHY (DIGITAL acteristics of the grid are defined based on the coverage ELEVATION MODELS) analysis of the meteorological network. The analyses for Countries usually have topographical maps in which relief some applications have resulted in grids with resolutions is represented using contour lines. However, these charts as high as 0.06° (~7 km) while the resolutions of the prox- are usually not practical for risk analysis. The most useful ies are usually lower. Grids of precipitation, maximum format for topography and risk analysis is perhaps Digital and minimum temperatures, evapotranspiration, solar Elevation Models (DEM) that can be implemented in geo- radiation and relative humidity, have been developed by graphical information systems (GIS) for spatial analysis the World Bank for Guatemala, Honduras, Nicaragua, (integration with additional information like political divi- Mozambique, Haiti, and the Dominican Republic (Uribe sions, land use, and so on). Furthermore, DEM are also Alcántara, 2010). highly appropriate for hydrological analysis (for example, flood risk). Additionally, Cressman has been applied extensively by different national weather services and scientists around NASA Shuttle Radar Topography Mission the world, including the CPC from the National Oceanic (SRTM): The SRTM is an international research effort and Atmospheric Administration and the European Cen- that obtained DEM on a near-global scale from 56°S to tre for Medium-Range Weather Forecasts. Also, gridded 60°N, to generate one of the most complete high-reso- analyses based on Cressman (or other successive correc- lution digital topographic databases. SRTM consisted of tion methods) have been developed for different countries a specially modified radar system that flew on board the and regions, that is, for the United States, Mexico (Uribe Space Shuttle Endeavour in February 2000. The eleva- Alcántara and Arroyo Quiroz 2010), Brazil (Silva et al. tion models are arranged into tiles, each covering one 2007), India (Sinha et al. 2006), and Europe (Drusch et degree of latitude and one degree of longitude. The reso- al. 2004). The advantages of using Cressman are its high lution of the raw data is one arc-second (30m), but this reliability, the incorporation of a second predictor from has only been released over U.S. territory. For the rest of indirect sources, and worldwide recognition. the world, only three arc-second (90m) data are available. Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 11 The elevation models can be downloaded freely over the from lower level behaviors that are expressed in equations. Internet. The Shuttle Radar Topography Mission is an Finally, the integration of these equations, representing international project spearheaded by NASA and the U.S. the responses of the plant at all levels, defines the system. National Geospatial-Intelligence Agency. Dynamic models have output that varies with time. Pro- cesses are characterized using state variables (variables The elevation datasets have void data in areas of very that define the state of the system at some point in time). high relief. This amounts to no more than 0.2 percent Dynamic crop models predict changes in crop status as a of the total area surveyed. Accordingly, groups of scien- function of biogenetic parameters (Hume and Callander tists have worked on algorithms to fill the voids of the 1990). These models simulate the evolution of a real crop, original SRTM data. Two datasets offer global coverage growing of leaves, stems and roots. These models simulate void-filled SRTM data at full resolution: the CGIAR-CSI the evolution of a real crop, growing of leaves, stems and versions and the U.S. Geological Survey’s HydroSHEDS roots. dataset. The CGIAR-CSI version 4 provides the best global coverage full resolution SRTM dataset. The The development of dynamic models requires multidis- HydroSHEDS dataset was generated for hydrological ciplinary collaboration. In order to establish the general applications and is suitable for consistent drainage and specifications required in the model and the relation- water flow information. ships between plant growth and the environment, plant physiologists, agronomists and soil scientists are usually CROP MODELS involved. Entomologists and plant pathologists define By the late 1960s and early 1970s, there was sufficient lit- insect and pathogen subsystems that are an important erature documenting plant growth in relation with envi- part of the crop ecosystem. Agrometeorologists con- ronmental conditions (Decker 1994; Mavi and Tupper tribute databases of weather and energy fluxes (in and 2004) to allow for the development of crop models in the around the canopy). Computer programmers select the 1980s and 1990s. By the end of the twentieth century, programming language and develop the model’s algo- there were thousands of crop models. Crop models are rithms model’s algorithms (Ritche et al. 1986). The model generally divided into empirical and mechanistic models. then passes to verification and validation processes. The Empirical models try to relate a behavior with attributes verification tests the model’s exactitude by comparing the of the same level, without regard to any theory. The pro- output with observational data. The model may need cedure is just adjusting a set of equations to a dataset. to be functionally adjusted or the coefficients may need For example, a model that relates crop yield directly to to be calibrated to improve the exactitude. Finally, the a level of fertilizer through an equation is an empirical validation compares model predictions with results from mathematical model. independent observational databases or experiments. A model is considered valid if its output is within pro- jected confidence bands. Mechanistic models, on the other hand, are usually dynamic2 and deterministic.3 The mechanistic models try to represent a behavior, which has some understanding of The applications of crop models include: the process at lower levels. Unlike empirical models that 1. Crop Breeding: Produce high-yielding and more try to connect variables in whatever way that fits the data, resistant cultivars based on sensitivity analysis of the mechanistic models break the system down into com- genetic characteristics. ponents. This process introduces processes and properties 2. Physiological Probes: Processes that are not exper- imentally accessible can be explored using com- prehensive models for scientific purposes. 2 Dynamic models account for time-dependent changes in the state of the 3. Sequence Analysis: Optimal crop rotations can system. Dynamic models are typically represented by differential equations. be identified based on sets of simulations that are 3 In deterministic models, every set of variable states is determined by param- eters and sets of previous states of these variables. Therefore, the results in able to capture long-term water, nitrogen and car- deterministic models are the same for a given set of initial conditions. bon, as continuums. 12 Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 4. Strategic and Tactical Applications: Models and available soil water of the previous dekad (Wd–1), is are run to compare different management sce- lower than the crop demand (ETMd). It is assumed that narios. not all rainfall is effective, and, on average, 20 percent 5. Forecast Applications: Similar to strategic and tac- of the rainfall is lost by run-off. The main output of tical applications, but the main interest is in the CSSWB is the Water Satisfaction Index, which describes final yield and other final variables. the cumulated water shortage over the crop season. This 6. Spatial Analysis: The result of linking crop mod- index has become the most widely used crop yield indi- els with geographical information systems to cross- cator for the yield reduction assessment due to drought, reference with other spatial information. worldwide. The Water Use Efficiency model calculates 7. Seasonal Analysis: A model is used to support de- the end-of-season grain yield under rain-fed conditions cisions related to crop, plant density, planting date, based on potential evapotranspiration and a reduction irrigation, and fertilizer strategies, for example. of the yield, taking into account crop physiology and 8. Scenarios: Models can be used to assess per- water deficit. formance under different stressors such as heat waves and drought or changes in management practices. EXAMPLES OF MECHANISTIC (DYNAMIC 9. Forecasting: Models can be used in conjunction AND DETERMINISTIC) MODELS with weather forecasts to supply estimates of yield Denitrification-Decomposition Agro-ecological with given weather conditions. Model (DNDC): The DNDC model (Li et al. 2006; Li et al. 1992a; Li et al. 1992b) is a process-based computer Mechanistic models are usually more desirable to use due simulation model focusing on the exchanges of water, to their comprehensive reproduction of crops. Unfortu- carbon (C), and nitrogen (N) between terrestrial ecosys- nately, they usually have higher data requirements and tems and the atmosphere. DNDC can be downloaded they are more complex to use. As such, the selection of for free at http://www.dndc.sr.unh.edu/. DNDC works the model is also restricted by the data available and the at the molecular or microorganism level driven by ther- hazards of interest. It is important to make sure that the modynamic and reaction kinetic principles. DNDC has model is able to reproduce negative effects of regional been widely tested at multiple scales with accurate results. hazards on yield. In case, for example, frost has been doc- Accordingly, DNDC is a good tool for high-resolution umented as one of the regional hazards, the users need and/or regional studies such as the district or zonal level, to make sure crop’s vulnerability as a function of tem- and scaled to national levels for inventory or risk pro- perature is taken into account by the model. Examples grams. DNDC requires a suite of input parameters and of empirical and mechanistic models are provided in the predicts yield, water, C, and N exchanges between the next sections. atmosphere and the plant-soil systems at a daily time step (see figure 3.1). EXAMPLE OF AN EMPIRICAL MODEL: The DNDC model was originally developed to simu- WATER USE EFFICIENCY MODEL late the effects of major farming practices (for example, The model is based on the FAO Crop Specific crop rotation, tillage, fertilization, manure amendment, Soil Water Balance (CSSWB; Frere and Popov 1986). irrigation, flooding, weeding, grass cutting, and graz- It is a simple soil water balance model used to assess the ing) and climate change (temperature and precipitation) impact of weather conditions on crops. The soil profile on yield and C and N cycles in various ecosystems. By functions as a water reservoir. When the climatic water tracking vegetation biomass production and decomposi- balance (Pd—ETMd) exceeds the storage capacity of the tion rates, DNDC also simulates long-term soil organic soil, excess rainfall is accounted for as water surplus or carbon dynamics, predicts methane (CH(4)), nitrous oxide deep percolation. Crops may suffer from drought stress (N2O), and greenhouse gas (GHG) emissions by tracking when the combined total of dekadal effective rainfall (Pd) the reaction kinetics of nitrification, denitrification, and Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 13 FIGURE 3.1. ENGINEERING FRAMEWORK AND INPUT CATEGORIES FOR DNDC TO GENERATE YIELD The DNDC Model Ecological Climate Soil Vegetation Human activity drivers Plant growth: • Water use Decomposition: Soil climate: • Caccumulation • SOM decay • Temperature profiles • Callocation • N-mineralization • Water profiles • Root respiration • CO2 production • Water drainage • Litter production • DOC production • Redox potential profiles Soil + – environmental Temperature Moisture pH Eh Substrates: NH4 , NO3 DOC factors Denitrification: Nitrification: Fermentation: • NO3– consumption • NH4+ consumption • CH4 production • Net NO, N2O production • NO3– production • CH4 consumption • N2 production • Net NO, N2O production • CH4 transport • Net CH4 flux fermentation across climatic zones, soil types, and man- accumulation of UC is called Residual Method. For pur- agement regimes. The simulated vegetation growth can poses of standardization of the rate of development of be used to generate yield distributions, suitability indi- cereals, a scale ranging from 0 to 1 is considered for the ces, and assess climate scenarios. The application of this periods of emergency and start of grain filling, and from model is illustrated in chapter 7. 1 to 2 for the period between grain filling and physiologi- cal maturity. Model Simulation of the Ecological Potential of Crops (known by its Spanish abbreviation, Gross CO2 Assimilation: The Photosynthetically MSPEC): Since 1990, the National Institute for Forestry, Active Radiation (PAR) is estimated from the total global Agriculture, and Livestock in Mexico has promoted the radiation, which can be fed directly into the model or esti- application of dynamic simulation models in the iden- mated with the Angstrom equation if no observational tification and solution of problems in agriculture. As a records are available. Gross CO2 assimilation is estimated result of this process, a simulation model of the ecologi- using the index of Radiation Use Efficiency, which mul- cal potential of crops was developed (MSPEC; Quijano tiplies the PAR and is corrected for incomplete intercep- et al. 1998) that allows the estimation of growth and pro- tion of light through the Leaf Area Index. Leaf area is duction of crops (corn, wheat, sorghum, barley, potatoes, estimated based on specific leaf area and leaf dry weight. and beans) to determine the effects of light, temperature, moisture availability, genotype, and planting date. The Distribution of dry matter, from the stage of crop MSPEC model calculates the daily dry matter production development, determines the priority of growth. of these crops under conditions of potential output, as In this stage, the model calculates the leaf area and growth moisture limitation. The model components are: of different plant organs from the physiological age. Frac- tion partition dry matter is obtained from experimental Physiological Age of the Crop: calculated from the data of the level of growth analysis of genotypes. The res- accumulation of heat units (UC) or degree days above piration rate, organ senescence, and the effects of water a certain base temperature. The method used for the stress on grain yield, are also related to the physiological 14 Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping age and used to calculate daily gain and dry weight accu- for Monitoring Ecological Potential Crop, has been used mulation in each organ of the plant. in the state of Guanajuato since 2001 for: 1. Yield forecasts of corn, sorghum, wheat, and Since 2003, the MSPEC was supplemented with models barley. of population dynamics of harmful organisms like the 2. Development studies and risk warning bulletins corn’s rootworm (Diabrotica virgifera zeae K. and S.). The for various phytosanitary problems in collabora- applications of the model serve different levels of users tion with the Phytosanitary Alert System of the (producers, agronomy students, field technicians, profes- State of Guanajuato, which has been operating sors, researchers, and government officers) so a platform since 2003. that serves as an interface between models, databases (cli- 3. The estimation of yields and other agricultural mate, soil, genotypes, harmful organisms) and stakehold- parameters required for the design of the Weather ers was created. This platform, called Information System Index Insurance program from AGROASEMEX. Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 15 CHAPTER FOUR HISTORICAL ANALYSES This section reviews risk-mapping analyses based on historical weather observations. These mapping techniques usually target risk identification and, ultimately, assess- ment. Within this category, we include: climatologies, hazard and risk maps, climate regionalizations and agro-ecological zoning. Climatologies are defined as the expected value of the weather for a given period and region. They are usually estimated as the average value in the last 30 years. An annual climatology of precipitation, for example, is the average annual cumulative precipitation for the last 30 years. There are also climatologies associated with temperatures and other meteorological variables. Clima- tologies are useful to have for a general idea of the climate in a given region, as they offer a snapshot of what has usually happened in the past 30 years or so. Hazard and risk maps are similar to climatologies because they are also based on historical information. However, in the case of hazard and risk maps, they usually try to quantify intensity, probability of occurrence, or expected losses, while climatologies are only an expression of the expected value of a given meteorological variable. CLIMATOLOGIES GLOBAL CLIMATOLOGIES FROM THE CLIMATE RESEARCH UNIT The Climate Research Unit (CRU) has developed climatologies at pixel sizes of 0.5° (~55 km) and 10’ (~18 km). The dataset contains eight mean monthly and annual val- ues of precipitation (1961–1990), wet-day frequency, temperature, diurnal temperature range, relative humidity, sunshine duration, ground frost frequency and wind speed. The dataset is created through the interpolation of station means. The data are available through the International Water Management Institute World Water and Climate Atlas (http://www.iwmi.org) and the Climatic Research Unit (http://www.cru.uea.ac.uk). Figure 4.1 shows the world’s climatology of annual precipitation at the highest resolution. NATIONAL CLIMATOLOGIES BASED ON GRIDDED DATASETS Based on the gridded datasets developed using objective analysis techniques and applied by the World Bank (see page 11), climatologies can be obtained averaging Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 17 FIGURE 4.1. WORLD’S CLIMATOLOGY (1960–90) OF ANNUAL PRECIPITATION (MM) FROM CRU Source: CRU. the daily records for the last 30 years from the grid. For FIGURE 4.2. CLIMATOLOGY (1979–2008) OF example, figure 4.2 shows the climatology of average tem- AVERAGE TEMPERATURE (°C) perature for Nicaragua. The resolution achieved in this OF NICARAGUA BASED ON AN country was 16 km for temperature and 8 km for pre- OBJECTIVE ANALYSIS FROM cipitation. The difference with global analysis like the one from CRU is that these gridded products are developed THE WORLD BANK with databases collected locally so the analysis is devel- oped with the most updated and complete information available. Besides, the World Bank has used a secondary predictor from atmospheric models. These methodologies are compared with independent sources. An additional advantage over CRU, for example, is that these climatolo- gies usually correspond to the average of the last 30 years (for example, 1979–2009) while the latest climatologies from CRU run from 1961–1990 so there is a significant delay in the update of these climatologies. REGIONAL CLIMATOLOGIES— ALBERTA, CANADA Climatologies in general can address many different Source: World Bank. aspects of the climate’s variability, and it is important to develop climatologies that will be useful for end users. precipitation throughout the year. There are climatologies The communication between the developer and end that can be critical for the user, and if they are addressing users helps in the definition of what is needed and what characteristics of climate that can result in yield losses, is feasible. Figure 4.3, for example, shows a climatology they can be called hazard maps (described in the next sec- of the annual number of days with rainfall of more than tion). More information can be obtained from: http:// 0.2 mm, which can provide an idea of the distribution of www.agric.gov.ab.ca/acis/climate-maps.jsp. 18 Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping FIGURE 4.3. CLIMATOLOGIES (1971–2000) Left panel: Annual number of days with precipitation greater than 0.2 mm. Right panel: Frost-free period of Alberta, Canada. Sources: Crown copyright. Government of Alberta Agriculture and Rural Development, AgroClimatic Information Service. of the hazard. For example, a risk map can display frost HAZARD AND RISK MAPS in the early stage of maize growth. The map, in this case, Risk is usually defined as a function of the hazard,4 can indicate either the annual probability of this hazard, vulnerability,5 and exposure.6 Making a risk map requires average annual losses (in yield or cash), or the product the selection of the magnitude (intensity to produce losses) of the loss times the probability, which is the generally and of the return probability (frequency of occurrence) accepted numerical expression of risk. However, these maps can be built only if yield data are available and loss 4 A situation that poses a level of threat to life, property, or environment. In the causes are identified. case of agriculture, natural hazards include hurricanes, droughts, frosts, and so on. 5 Grade of ability to withstand the effects of a hazard. In the case of agricul- ture, some conditions can decrease or increase vulnerability. For example, if the In cases where yield data are not available, the commu- farmer is prepared to withstand losses associated to pests (for example, by using nity usually develops hazard maps (which sometimes are pesticides), his vulnerability decreases. erroneously presented as risk maps). Hazard maps usu- 6 Condition of being subject to loss because of some hazard. In the case of agriculture, exposed goods are crops. Their exposure is indicated by their loca- ally have an intensity or probability scale associated to tion and value. the magnitude or probability of a given hazard. Hazard Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 19 FIGURE 4.4. U.S. PLANT HARDINESS ZONE MAP (CLIMATOLOGIES OF AVERAGE ANNUAL EXTREME MINIMUM TEMPERATURE CLASSIFIED AT 5°F INTERVALS) Source: U.S. Department of Agriculture. maps, however, don’t take into account vulnerability and plants are expected to grow. Figure 4.4 shows the hardi- exposure so the scales and classifications are not necessar- ness zone map for the United States. ily related to losses (that is, risk). Since these maps require the least amount of information, they are much more Hardiness zones have to be used in combination with common than risk maps. This document contains exam- plant vulnerabilities to be practical. The zones have draw- ples of representative hazard and risk maps in this section. backs because the following factors are not taken into Unfortunately, most hazard maps commonly available are account: summer temperatures, impacts of snow cover, only for developed countries. It is to be hoped these maps soil moisture, humidity, number of days with frost, and will illustrate and encourage their generation in develop- duration of low temperatures. Arguably, instead of aver- ing countries as well. age minimum temperatures, perhaps the probabilities of specific lethal minimum temperatures would be more use- ful. More information is available at: http://planthardi- PLANT HARDINESS ZONES ness.ars.usda.gov/PHZMWeb/. A hardiness zone is a geographical area within defined ranges of average minimum temperatures during the win- Although the zones were first developed for the United ter (divided into 5°F zones) where specific categories of States by the Department of Agriculture (USDA), the 20 Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping FIGURE 4.5. TOTAL NUMBER OF HURRICANE STRIKES BY U.S. COUNTIES/PARISHES/BOROUGHS, 1900–2010 Source: National Weather Service (NWS). use of the zones has been adopted by other nations. have a general idea of their exposure to hurricanes; how- The Canadian Forest Service (CFS) from the Natural ever, these maps provide an objective point of comparison Resources Canada, for example, develops hardiness zone and reference. maps, but it is also aiming to develop potential range maps for individual species of trees, shrubs, and peren- FLOOD INDEX—MEXICO nial flowers (http://planthardiness.gc.ca/). The CFS has Floods are classified into: (1) flash-floods (intense precipi- requested support from Canadian experts and the pub- tation events that exceed the soil’s infiltration capacity); lic to identify plants that survive at their location. Once (2) riverine floods (runoff exceeds the capacity of rivers enough data are collected to develop a climatic profile, the and channels so the water overflows the stream banks); range maps will be generated. (3) coastal floods (for example, high tides associated with tropical cyclones, that is, storm surges); and (4) urban CONUS HURRICANE STRIKE DENSITY floods (due to lack of proper drainage in an urban area). The National Hurricane Center within NOAA has devel- Unfortunately, consistent national databases of floods oped several tropical cyclone climatologies, which provide are practically nonexistent; therefore, risk assessments a general idea of the typical timing, magnitude, and loca- are generally based on indirect methods such as physical tion of tropical cyclones based on the analysis of histori- mathematical modeling. In this context, the application cal records of cyclone tracks (http://www.nhc.noaa.gov/ of models generally attempts to document the magnitude climo/#uss). Figure 4.5 displays the total number of hur- and probability of flood damages based on simulations ricane strikes by counties between 1900 and 2010 in the that depend on historical records of precipitation and Atlantic basin (Atlantic Ocean, Caribbean Sea, and Gulf stream flow. of Mexico). The Atlantic hurricane season runs from June 1 to November 30, and the Eastern Pacific hurricane sea- The definition of a flood index that allows the iden- son runs from May 15 to November 30. Farmers usually tification of regions prone to floods based on limited Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 21 FIGURE 4.6. FLOOD-PRONE REGIONS OF MEXICO (URIBE ALCÁNTARA ET AL. 2010) Source: Uribe Alcántara et al. 2010. information was proposed in Mexico (Uribe Alcántara factors of flood disasters, one is the triggering natural et al. 2010). The index is based on the topographic index event in the form of extreme precipitation and conse- developed by Beven (1979), but additionally it considers quently extreme river discharge. The threatening natural soil, hydrologic, and climatologic factors. Additionally, the event represents the hazard component in the assessment. results are compared, with corresponding simulations of In addition, exposure is among the anthropogenic factors a routing model, to the floodplain of the state of Tabasco. that contribute to increasing flood risk at a given loca- The results indicate that the index is able to capture rea- tion. Advanced GIS-based techniques and datasets were sonably well the perennial and ephemeral flooded regions. fundamental elements within the approach. The issue of The methodology is applied to all the basins in Mexico for flood risk mapping is studied at continental scale. The aim the generation of a national map of flood-prone regions was to identify and map the regions prone to flood disas- (figure 4.6). ters and to quantify the potential losses with the support of stage-damage functions. This map provides support to FLOOD DAMAGE POTENTIAL MAP several European Commission initiatives including the FOR THE EUROPEAN UNION European Flood Action Programme, the Directive on the In these maps, flood risk areas are defined on the basis Assessment and Management of Flood Risks, the Solidar- of associated risk factors, that is, exposure and hazard ity Fund, and EU regional policies. More information can (figure 4.7). Geo-referenced data on land use is processed be obtained from: http://floods.jrc.ec.europa.eu/. Inci- for setting up the exposure component. The hazard fac- dentally, this is perhaps the only actual risk map shown tor is implemented by hydrological methods at different in this document because it actually shows potential eco- scales and for many return periods. Among the causal nomical annual losses due to flood. 22 Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping FIGURE 4.7. MAP OF FLOOD DAMAGE POTENTIAL (MILLIONS OF EUROS IN PURCHASING POWER PARITIES) OF THE EUROPEAN UNION Source: European Commission Joint Research Centre Floods Portal (http://floods.jrc.ec.europa.eu/ flood-risk.html). Maps of the return periods of agrometeorological MAPS OF THE catastrophic events in Mexico have been developed RETURN PERIOD7 OF (Escamilla Juárez 2012). The definition of a cata- AGROMETEOROLOGICAL strophic event was based on the Operation Rules from CATASTROPHIC EVENTS IN the programs of the Secretariat of Agriculture, Live- stock, Rural Development, Fisheries and Food, except MEXICO for flood and precipitation excess, where the definition Mexico is highly vulnerable to the occurrence of cata- from the National Disasters Fund was applied. Maps strophic natural hazards, mainly catastrophic and geologic, for drought, frost, precipitation excess, and flood were which involve 60 percent of the population and increase developed at state and county levels. Figure 4.8 shows the probability of having high economical damages. The maps of return periods for catastrophic droughts at amount of losses has been increasing in recent decades. state and county levels. Based on the maps, it was con- During 2001–03, the losses involved annual average con- cluded that drought is the most important risk for the tributions from the government around US$207.8 mil- agricultural sector. In average, the country presents cat- lion. In 2007 and 2009, the contributions increased to astrophic drought every 9.5 years, that is, 13.7 percent US$884.2 million. While in 2010, the amount increased average annual probability. Five states present annual to US$1,963 million. Ninety percent of this amount is probabilities above 15 percent (six to seven years). How- related to hydrometeorological events, mainly hurricanes ever, at county level the probabilities can reach between and extreme precipitation events. 18 and 30 percent (three to five years). More impor- tantly, 77.1 percent of these counties are located in only 7 Average time, in years, between events. nine out of the 32 Mexican states. Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 23 FIGURE 4.8. MAPS OF RETURN PERIOD OF CATASTROPHIC DROUGHT State (left panel) and counties (right panel) with return period between three and five years. Source: Jesus Escamilla. BOX 4.1. PRINCIPAL COMPONENTS ANALYSIS AND CLUSTERING Principal component analysis (PCA): PCA is a math- result in regions whose variability is different; the second, in ematical procedure that transforms a set of observations of regions whose range of values is different. possibly correlated variables into a set of uncorrelated vari- ables. The first principal component has the largest possible Cluster analysis or clustering: Cluster analysis can variance (that is, accounts for the largest portion of the vari- be achieved by utilizing one of several algorithms. Popular ability in the data), and each succeeding component, in turn, notions of clusters include groups with small distances among has the highest variance possible under the constraint that it the cluster members, dense areas of the data space, intervals be orthogonal (uncorrelated) to the preceding components. or particular statistical distributions. Clustering can therefore PCA reveals the internal structure of the data. If a multi- be formulated as a multi-objective optimization problem. The variate dataset is visualized as a set of coordinates in a high- appropriate clustering algorithm and parameter settings (for dimensional data space, PCA can afford a lower-dimensional example, distance function, density threshold, or number picture when viewed from its most informative viewpoint. of expected clusters) depend on the individual data set and This results from using only the first few principal compo- intended use of the results. nents, so that the dimensionality of the transformed data is Clustering can be used to make climate regionalizations. In reduced. this case, the options are related to the number the variables PCA can be used to make climate regionalizations. There are involved to determine similarity (small distance). Variables two possibilities: The PCA can be applied to either the cor- can include, for example, precipitation, temperature, altitude, relation matrix or the covariance matrix. The first option will and so on. and precipitation, and the seasonality of precipitation. CLIMATE REGIONALIZATIONS There are additional classifications based on cluster anal- The most widely used climate regionalization is the Kop- ysis that attempt to group unit cells with similar climato- pen Climate Classification, which relies on the principle logical values (for example, annual precipitation, length that vegetation is the best expression of climate; as such, of the rainy season, and so on) within regions. On the the boundaries between zones are selected with vegeta- other hand, there are also regionalizations that attempt tion distribution in mind. The classification is based on to group unit cells in regions with similar behavior (for the combination of annual and monthly temperatures example, variability; see box 4.1). Figure 4.9 (left panel) 24 Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping FIGURE 4.9. NORTH AMERICAN MONSOON REGIONALIZATION 200 1. Monsoon 200 2. Monsoon west south 150 150 PRECIP (mm) PRECIP (mm) 100 100 50 50 0 0 J FMAMJ J ASOND J FMAMJ J ASOND 200 3. Monsoon south 200 4. Monsoon west 150 150 PRECIP (mm) PRECIP (mm) 100 100 50 50 0 0 J FMAMJ J ASOND J FMAMJ J ASOND Precipitation (left panel) and corresponding annual cycles (right panel). Source: Comrie & Glenn 1998. shows a regionalization for the North American mon- relatively independent regions. When the extension of soon region, which defines four regions with distinctive the region of analysis is large and the spatial variabil- annual cycles of precipitation (figure 4.9, right panel). ity of climate is strong, very different annual cycles are These analyses are usually based on the rotation of the present. The regionalization can help identify relatively most important principal components of the correla- independent regions that need to be characterized. In tion (or covariance) matrix. In terms of applications for this way, the characterization simplifies from hundreds risk analysis, for example, AGROASEMEX developed a of stations/pixels to a few regions. With the advent of cluster-based regionalization for the definition of regions gridded databases, point analysis is becoming more com- sharing the same Weather Index Insurance scheme in mon than regionalizations; however, regions can still be Mexico (AGROASEMEX 2006). These regions can also useful to simplify the definition and application of poli- be useful for the identification and characterization of cies and analysis. Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 25 CHAPTER FIVE DIAGNOSTIC AND FORECASTING ANALYSES There are numerous mapping techniques for diagnostic and forecasting purposes. Diagnostic techniques are designed to provide a technical description of the current risk situation and its causes. Forecasts, on the other hand, predict potential risk events based on the study and analysis of pertinent observations and simulations. They have been included together in this chapter given that (1) many diagnostic techniques are used to make decisions in the near future, (2) applied agrometeorological forecasting techniques are still scarce, and (3) as the products become more sophisticated, the line between diagnostic and forecast products becomes thinner. The mapping techniques include, for example: monitors, outlooks, watches, warnings, and forecasts, which are formally issued in bulletins. Monitors are based on continuous or periodic measure- ments of existing, changing, agrometeorological conditions. Outlooks, watches and warnings are issued to indicate the probability of a hazardous event. Their difference is the likeliness that the event will occur. Outlooks have low certainty; watches, higher certainty; and warnings, the highest. These techniques aim to provide information in advance so stakeholders can prepare for hazardous events. On the other hand, there are forecasts, which are the most likely prediction. In atmos- pheric sciences, forecasts are classified in short- (up to 48 hours ahead), medium- (up to 7 days), and long-term (more than seven days). Forecasts are classified in this way partially because of the technical challenges associated with each lead time. Further- more, as the lead time increases, the uncertainty increases. Short- and medium-range forecasts can be particularly valuable for frosts and pests, for example, because risk mitigation activities can be implemented to reduce losses. Seasonal (long-term) fore- casts of precipitation, with lead times of a few months, can be particularly useful for events such as drought. These forecasts usually indicate the probabilities of climate anomalies of variables like precipitation and temperature (that is, probabilities of hav- ing higher than usual precipitation). Finally, with Anthropogenic Climate Change, additional analyses (scenarios), which have lead times of decades, have been broadly developed. Such scenarios are a description of a possible future state for a given base- line condition (for example, rate of emissions of greenhouse gases). Usually, a set of scenarios is developed to reflect the range of uncertainty in projections. Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 27 This chapter contains examples of diagnostic and fore- Rico. It began in 1999 as a federal, state, and academic cast mapping techniques that illustrate their potential partnership, growing out of a Western Governors’ Asso- use in the agricultural sector. The selections it contains ciation initiative to provide timely and understandable seek to provide a representative mapping technique for scientific information on water supply and drought for pol- each diagnostic and forecasting classification and hazards. icymakers. The Monitor is produced by a rotating group Regarding climate change scenarios, an example of the of authors from the USDA, the NOAA, and the National application of projections to assess crop vulnerability on Drought Mitigation Center. It incorporates reviews from regional basis for Mozambique is shown on page 47. a group of 250 climatologists, extension agents, and oth- ers across the nation. U.S. DROUGHT MONITOR The map is updated weekly by combining a variety of Drought is an insidious hazard, often referred to as a data-based drought indices and indicators, as well as local “creeping phenomenon.” Its definition is not straightfor- expert input. The map denotes four levels of drought ward because what might be considered a drought in, intensity (ranging from D1–D4) and one level of “abnor- say, Bali (six days without rain) would certainly not be mal dryness” (D0). Also depicted are areas experienc- considered a drought in Libya (annual rainfall less than ing agricultural (A) or hydrological (H) drought impacts. 180 mm). In the most general terms, droughts originate These impact indicators help communicate whether from a precipitation deficit over an extended period short- or long-term precipitation deficits are occurring of time (a season or more), which interplays with the (figure 5.1). demand of water supply for crop production, for exam- ple. Depending on the effect of the precipitation deficit, The U.S. Drought Monitor sets the standard for com- droughts are classified as: meteorological (precipitation municating the location and intensity of drought to a deficit), economical (precipitation deficit with economic broad audience. The map summarizes and synthe- consequences), and agricultural (precipitation deficit that sizes information from the local and state levels to the affects the sector). national scale, making it the most widely used gauge of drought conditions in the country. Policymakers There are conceptual and operational definitions of rely on it to allocate relief dollars, states use it to trig- drought. Conceptual definitions of drought may be ger drought response measures, and it is a frequent important in establishing drought policy. For example, in source for the media. The USDA utilizes the map to Australia, financial assistance is provided to farmers only distribute millions, and sometimes billions, of dollars under “exceptional drought circumstances.” Declara- in drought relief to farmers and ranchers each year, tions of exceptional drought are based on science-driven and the Internal Revenue Service also refers to it for assessments to prevent unjustified or unsustainable claims ranching-related tax determinations. More information (for example, frequent claims from farmers in semi-arid available at: http://drought.unl.edu/MonitoringTools/ areas). Operational definitions also help define the onset, USDroughtMonitor.aspx. severity, and end of droughts. No single operational defi- nition of drought works in all circumstances. Therefore, most drought planners now rely on mathematical indices METEOROLOGICAL MONITOR to decide when to start implementing drought response measures. To determine the onset of drought, for exam- (AUSTRALIA) ple, operational definitions specify the degree of departure Australia’s Bureau of Meteorology regularly produces a from the average of precipitation. This is usually achieved number of important, objective analyses (see page 11) by comparing the current situation with the climatological to provide a constant and rapid overview of rainfall and value (30-year average). temperature distribution across Australia. The analyses are computer generated using the Barnes successive correc- The U.S. Drought Monitor map provides a summary of tion technique. On most maps (both rainfall and tempera- drought conditions across the United States and Puerto ture), each grid point represents a square area with sides of 28 Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping FIGURE 5.1. U.S. DROUGHT MONITOR FOR JUNE 5, 2012 FIGURE 5.2. EXAMPLES OF MAPS OF DAILY TEMPERATURE (LEFT PANEL) AND CUMULATIVE PRECIPITATION (RIGHT PANEL) FOR AUSTRALIA approximately 25 km. The size of the grids is limited by temperature analyses use data collected through electronic the relatively coarse average data separation between sta- communication channels. These data have been screened tions. Daily rainfall maps are available at around 2:00 p.m. for errors, but are not yet fully quality controlled. Figure (Eastern Standard Time; EST), which is only a couple of 5.2 shows the objective analysis of minimum temperature hours after the morning observations from all time zones for a given day. The map suggests subzero temperatures are received in the central database. Daily temperature in the eastern portion of the country. More information maps for the previous day are available at around 12:30 available at: http://www.bom.gov.au/climate/austmaps/ p.m. (EST). The Bureau of Meteorology’s rainfall and mapinfo.shtml. Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 29 “ensemble maps” as well as an additional product that SOYBEAN RUST combines these three maps. A student team integrates OUTLOOK (U.S.) the composite weather maps with the “ensemble maps” A number of maps can be developed to improve man- to create an “SBR activity ensemble” that defines wait, agement of risks associated with pests and diseases. watch, and warning zones for field scouting and soybean This section presents an example for soybean rust (SBR) rust decision support. Three times a week, the student in the United States, but similar efforts can be devel- team issues 1–2 day and 3–5 day forecasts in text format oped for other pests and diseases in other regions of the on a restricted-access website. The maps of spore move- world. SBR, caused by two fungi species called Phakopsora ment and infection development both in the present and pachyrhizi and Phakopsora meibomiae, is an aggressive patho- the future are invaluable in guiding extension specialists gen that has spread from Asia to Africa, South America, as to the likelihood, location, and timing of soybean rust and North America. Yield losses associated with SBR can in their states. Figure 5.3 (left panel) shows the results of a be severe (10–80 percent). Soybean rust was first detected soybean rust scouting for July 25, 2005. Since the spread in the United States in the fall of 2004, and the national of soybean rust is strongly associated with weather condi- sentinel plot network has subsequently monitored its tions, scouting efforts can be combined with meteorologi- spread and development. cal conditions to issue spread forecasts (figure 5.3, right panel). Weather conditions are critical for the determination of spread of this disease. Soybean rust development is FLOOD: WARNING SYSTEM favored by temperatures ranging from 54°–84°F (65°– 80°F is optimum), with relative humidity above 90 per- (UNITED STATES AND THE cent for more than 12 hours. Soybean rust can be active EUROPEAN UNION) with daytime temperatures as high as 100°F (and possibly The Advanced Hydrologic Prediction Service (AHPS) is a higher) as long as night temperatures fall into the opti- component of the Climate, Water, and Weather Services mum range for disease development. In order for spores offered by the NWS from NOAA in the United States to germinate and infect the plant, 6 hours of continu- AHPS is a suite of accurate and information-rich forecast ous leaf wetness are required. Infection increases with products. They display the magnitude and uncertainty longer leaf wetness periods of up to 12 hours. In South of occurrence of floods or droughts, from hours to days America, significant rust development is associated with and months in advance. These products enable govern- rain events. ment agencies, private organizations, and individuals to make more informed decisions about risk-based policies The Penn State University Ensemble SBR Forecasting and actions to mitigate the dangers posed by floods and Program simulates the local development of soybean rust droughts. infections based on weather-driven transport of spores from infected geographic regions to downwind areas The vast majority of the observed water level data displayed with potential host vegetation. The aerobiology ensemble on the AHPS web pages originates from the Hydromete- modeling project follows the movement and development orological Automated Data System (HADS) operated by of SBR across the country with three models: (1) the Inte- the Office of Hydrologic Development. Following the pro- grated Aerobiological Modeling System (IAMS), (2) the cessing of the raw data, HADS delivers the observational HYSPLIT trajectory model (NOAA ARL), and (3) a cli- data to the Weather Forecast Offices and River Forecast matological based model. Composite Precipitation/Rela- Centers, which use the data in their hydrologic models and tive Humidity, Solar Radiation/Minimum Temperature, create the informational displays for the AHPS. and Wind Speed/Direction maps are created from model output from the National Weather Service (NWS). SBR Using sophisticated computer models and large amounts observations and IAMS model outputs are interpreted to of data from a wide variety of sources such as super produce SBR transport, deposition, and disease severity computers, automated gauges, geostationary satellites, 30 Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping FIGURE 5.3. SOYBEAN RUST OBSERVATION FROM THE INTEGRATED PEST MANAGEMENT Pest information platform for extension and education (IPM PIPE) in July 25, 2005 (left panel) and simulation of potential deposition areas for soybean rust spores (June 2010) based on weather conditions (right panel). Source: U.S. Department of Agriculture Pest Information Platform for Extension and Education. FIGURE 5.4. FLOOD MONITOR BASED ON GAUGE DATA Source: National Weather Service (NWS). U.S. gauges are shown as squares, and the colors of the squares change depending on flood status. Doppler radars, weather observation stations, and The information is presented through user-friendly the computer and communications system, called the graphical products. Figure 5.4, for example, shows the Advanced Weather Interactive Processing System, the presence of floods based on readings from gauge stations. NWS provides hydrologic forecasts for almost 4,000 loca- The information, such as the flood forecast level to which tions across the United States. a river will rise and when it is likely to reach its peak or Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 31 FIGURE 5.5. FLOOD MONITOR OF THE EU BASED ON GAUGE DATA (MAY 23, 2010) Source: European Commission Joint Research Centre Floods Portal. The symbol of the gauges changes depending on the flood status. crest, is shown through hydrographs. Additional informa- range of people, such as barge operators, power compa- tion includes: nies, recreational users, households, businesses, and envi- 1. The chance or probability of a river exceeding ronmentalists. More information can be obtained from minor, moderate, or major flooding, http://water.weather.gov/ahps/index.php#. 2. The chance of a river exceeding certain level, vol- ume, and flow of water at specific points on the A similar overview of the current floods in Europe is made river during 90 day periods, and through the European Terrestrial Network for River Dis- 3. A map of areas surrounding the forecast point charge, based on a close collaboration with European that provides information about major roads, rail- Hydrological Services and the Global Runoff Data Centre ways, landmarks, and so on, likely to be flooded, (GRDC) in Koblenz, Germany. An overview of the cur- the levels of past floods, and so on. rent floods in Europe is provided through the European Terrestrial Network for River Discharge. The overview AHPS forecast products are a basis for operation and map is based on near-real-time river measurements, auto- management of flood-control structures. Emergency matically transferred by the National Hydrological Cen- management officials at local and state levels use these tres, via the GRDC, to the Joint Research Centre. The forecasts to fight floods, evacuate residents, and to take map shows the locations where river levels exceed criti- other measures to mitigate the impact of flooding. In cal thresholds (figure 5.5). More information available at: addition to farmers, these products can be used by a wide http://floods.jrc.ec.europa.eu/. 32 Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping FIGURE 5.6. FORECAST OF CHANCE OF EXCEEDING TRIMESTRAL MEDIAN RAINFALL (JUNE TO AUGUST, 2012) IN AUSTRALIA (LEFT PANEL) AND HISTORICAL CONSISTENCY OF THE FORECASTING SCHEME (RIGHT PANEL) Right panel: Color scheme indicates percentage of times the scheme has forecasted correctly. consistency means that tests of the model on historical RAINFALL SEASONAL data show a high correlation between the most likely out- FORECAST (AUSTRALIA AND look category (above/below median) and the verifying INTERNATIONAL) observation (above/below median). In this situation rela- The Australian Bureau of Meteorology issues general tively high confidence can be placed in the outlook prob- statements about the probability of wetter- or drier-than- abilities. On the other hand, low consistency means the average weather over future 3-month periods (figure 5.6, historical relationship, and therefore outlook confidence, left panel). The Bureau also issues average, maximum, is weak. More information available at: http://www.bom and minimum temperatures for the entire 3-month out- .gov.au/climate/outlooks/#/overview/summary. look period. These outlooks are based on the sea surface temperature records for the tropical Pacific and Indian Forecast systems developed regionally, such as that from oceans and the Southern Oscillation Index (SOI), which Australia, usually allow higher resolution, accuracy, is calculated using the barometric pressure difference and flexibility in the presentation of the results for both between Tahiti and Darwin. The SOI is one indicator of developers and users. However, technical challenges usu- the stage of El Niño or La Niña events in the tropical ally prevent their emergence in developing countries. In Pacific Ocean. For example, a moderate to strongly nega- those cases, the users can rely on third-party forecasts tive SOI (persistently below −10) is usually characteristic from recognized institutions like the European Centre of El Niño, which is often associated with below-average for Mid-Range Weather Forecasts (ECMWF), which rainfall over eastern Australia, and a weaker-than-normal perhaps has the best simulation systems in the world. monsoon in the north. Figure 5.7 shows the precipitation forecasts for May, June, and July in 2012. In this case, unlike the case of Australia, An important part of the forecasts is the evaluation. Fore- the forecast is estimated based on late-generation numeri- casts carry a certain amount of uncertainty. It is important cal weather prediction models. These forecasts are availa- inform stakeholders as to the historical performance of the ble online in a low-resolution format, but national weather forecast system. In the case of Australia, consistency maps services can have access to these high-resolution fore- are generated regularly (figure 5.6, right panel). Strong casts in the original data format so they can focus on the Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 33 FIGURE 5.7. TRIMESTRAL FORECAST (MAY–JULY, 2012) OF PROBABILITY OF MOST LIKELY CATEGORY OF PRECIPITATION FROM THE ECMWF Source: International Resource Institute, Columbia University. FIGURE 5.8. ENSO FORECAST FROM DIFFERENT CLIMATIC CENTERS AROUND THE WORLD COLLECTED AND PUBLISHED BY IRI (LEFT PANEL) AND PROBABILITIES OF HAVING EL NIÑO, LA NIÑA, OR A NEUTRAL YEAR (RIGHT PANEL) Mid-May IRI/CPC Plume-Based Probabilistic ENSO Forecast ENSO state based on NINO3.4 SST Anomaly 90 Neutral ENSO: –0,45°C to 0.45°C 80 El Niño Neutral 70 La Niña Probability (%) 60 50 Climatological probability: 40 El Niño 30 Neutral 20 La Niña 10 0 MJJ JJA JAS ASO SON OND NDJ DJF JFM 2012 2013 Time period Source: Consensus Forecast of the International Resource Institute, Columbia University, and the Climate Prediction Center, National Weather Service. processing of these forecasts for their specific needs. International Research Institute (IRI), which also issue One of the potential and expected applications of these regular forecasts. Finally, the Australian outlook relies forecasts is downscaling, which allows local calibration upon on a numeric estimation of probabilities that is and resolution increase. More information available at: based on climate modes like El Nińo-Southern Oscilla- http://www.ecmwf.int/en/forecasts/charts. tion (ENSO). ENSO is an important climate regulator in many parts of the world. IRI collects ENSO forecasts In addition to the European Centre for Mid-Range from all over the world from recognized institutions, and Weather Forecasts, there are also institutions like the distribute these forecasts in probabilistic form (figure 5.8). 34 Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping FIGURE 5.9. MAP FROM THE COVER OF AN AGROMETEOROLOGICAL BULLETIN PLANT AVAILABLE SOIL MOISTURE 4th June 2012 0 75 150 225 300 375 Kilometres LEGEND Shire boundary Wheatbelt boundary BOM rainfall station DAFWA rainfall station Soil moisture (mm) <5 5–10 10–20 20–30 30–40 40–50 50–60 60–70 70–80 >80 Source: Copyright © Western Australian Agriculture Authority. These forecasts can be very useful for regional clima- An agrometeorological bulletin is issued for a user com- tologists once the impact of ENSO in the region has munity, so it must meet their needs. A good practice is to been documented. More information available at: prepare the bulletin jointly between national agrometeo- http://iri.columbia.edu/climate/ENSO/currentinfo rological services, extension agents, researchers, and so on. /QuickLook.html. The bulletin should be published regularly (every 7–10 days). A good approach is to publish detailed monthly bul- AGROMETEOROLOGICAL letins, with updates in between (for example, every week). BULLETINS Such a bulletin should have a one-page summary, with Agrometeorological bulletins are perhaps the most appro- two components: (1) a box with text, and a map of the priate means to distribute most of the products presented country showing areas where problems occurred or are in this document. Figure 5.9 shows the cover of an agrome- likely to occur, and (2) a weather analysis summarizing the teorological bulletin from the Department of Agriculture current weather for the period of interest. It is also impor- and Food from the state government of Western Australia. tant to compare the current period with the long-term Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 35 average and the previous season, for comparison purposes. Prospects describing what will happen between the cur- A “snapshot” of agriculture is needed to understand the rent time and harvest, including considerations of actual impact of weather. and future conditions, should be included. This may be based on climatology, weather forecasts, or seasonal fore- Qualitative information about crops, livestock, pests, and casts, plus additional relevant information (for example, diseases should also be included (for example, crop stages, market and road conditions). development stages of pests, estimated soil moisture). A number of useful maps could be included to cover these The back cover could contain glossaries or generic topics, and a description of how weather has affected information, for instance a map showing the main agro- agriculture is important. Yield maps can be included, if ecological zones of the country, or explain which crops are available. The bulletin should focus on the effect of the “native” and which are “exotic,” and so on. More informa- weather. For example: tion about bulletins and examples can be obtained from: » Lack of rainfall at the time of flowering of maize is http://www.fao.org/NR/climpag/pub/ likely to reduce yields. AgrometeorologicalBulletinsTips.pdf » Unusually high temperatures at the time of flowing of rice will probably reduce pollination. http://www.agric.wa.gov.au/PC_90647.html » The combined effect of high moisture and low temperature has certainly favored the development http://www.hydromet.gov.bz/monthly-argro-met-bulletin of black spot disease. http://63.175.159.26/~cimh/cami/files/dombulletin/ » Abundant rain in the northern region has cre- sep2012.pdf ated good breeding grounds for desert locusts but rangeland production has improved. http://www.metmalawi.com/bulletins/bulletins.php 36 Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping CHAPTER SIX AGRO-ECOLOGICAL ZONES Agro-ecological zoning divides an area of land into smaller units, which have simi- lar characteristics related to land suitability, potential production, and environmental impact (FAO/IIASA 1991). An agro-ecological zone (AEZ) is a mapping unit defined in terms of climate, soil, landform, and land cover with a specific range of potential and constraints for land use. The FAO Agro-ecological Zones Project (FAO 1978–81) was an early exercise in land evaluation at a continental scale. Results of the FAO AEZ project include land suitability, potential production, and population support capacity for 117 developing nations. Agro-ecological zoning provides an assessment of the following issues: 1. Distribution of land with different potentials and constrains 2. Response to improvements in inputs and management 3. Balance between population demand and land availability 4. Land use recommendations In terms of land use recommendations, they benefit from policy formulations that include: 1. Specific extension support 2. Specific inputs and relief programs 3. Established research priorities 4. Defined development programs The regionalization process depends on the information available, modeling capabilities, and objectives. Therefore, the methodologies applied regionally can have differences. However, in this section, we attempt to list and describe the common stages and order Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 37 FIGURE 6.1. FLOWCHART OF A models are being proposed involving an ever-increasing GENERAL AGRO- flow of complexities, climate change implications are not addressed in this paper, which is restricted to short- ECOLOGICAL ZONING term weather risk assessment techniques. Each one of PROCESS the methodology’s stages is described individually in Land use types (LUT) the following sub-sections. Subsequently, the stages are inventory illustrated based on an agro-ecological zoning exercise implemented by the World Bank in Mozambique. Sev- Land resources inventory eral examples of AEZs are provided in the last sections. The description in these cases is not exhaustive. The examples are only used to highlight regional challenges Climatic and agro-climatic and solutions so the reader can appreciate that usually regionalizations (optional) there are adaptations to solve regional challenges and needs. However, FAO Global Agro Ecological Zones (GAEZ) latest products have managed to achieve excel- Land suitability lent quality, higher resolution, and cover a wide diversity of crops. This is an application that is readily available Land vulnerability to and useful. climate change Agro-ecological zones INVENTORY OF LAND (optional) USE TYPES LUTs are representative combinations of crop(s), manage- Assessment ment system (operations and inputs), and socioeconomic settings. An inventory of the climatic, edaphic, and land- form requirements for each LUT is created. The level of necessary to develop an AEZ. In general terms, the zoning detail depends on the databases available. This informa- approach involves the following stages (see also figure 6.1): tion is used to assess the climatic and edaphic suitability, 1. Land use types (LUT) as well as the potential yield calculation. In this activity, 2. Land resource inventory current or projected representative LUTs are selected for 3. Climatic and agro-climatic regionalizations (optional) the analysis. 4. Land suitability 5. AEZ (optional)8 The crop climatic inventory should contain phonologi- 6. Analysis and recommendations cal requirement, such as thermal ranges, day length, and growth cycles (adaptation to climatic conditions). The The vulnerability of crops to climate change is becom- soil inventory, on the other hand, should summarize the ing an important assessment of agro-ecological zoning. soil requirements of crops related to internal (for exam- However, given that even more methodologies and ple, soil temperature, soil moisture soil aeration, natural fertility, soil depth) and external soil requirements (for 8 As gridded datasets are becoming more popular, agro-ecological zoning has example, slope, occurrence and depth of flooding, acces- evolved into point analysis so that regionalizations and zonifications are no lon- sibility). Therefore, although in some AEZ reports an ger essential. The topic has historically been named agro-ecological zoning; LUT inventory is listed after the land resources inventory, even though zonifications are not longer the final objective, the name is still this document provides it first because the climate and soil preserved. The objective is evolving toward the assessments of land suitability and vulnerability to climate change. Regionalizations can still be useful for plan- information required will depend on the characteristics of ning purposes and policy definition and implementation, but they are optional. the selected LUTs. 38 Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping TABLE 6.1. LIMITING CROP FACTORS AND THEIR RELATIONSHIP WITH POTENTIAL, ATTAINABLE, AND ACTUAL YIELDS Factor Potential Yield Attainable Yield Actual Yield CO2 X X X Solar Radiation X X X Temperature X X X Genotype X X X Plant Density X X X Water Supply X X Any other limiting factor (pests, weeds, diseases, and so on) COMPILATION OF LAND FIGURE 6.2. YIELD AND GAPS OF RESOURCES INVENTORY POTENTIAL, ATTAINABLE, AND ACTUAL PRODUCTION At this stage, all datasets relevant for production are com- Yield (%) Gap (%) piled. The most important information is related to cli- 100 mate, soil, and topography. However, additional datasets can be useful like land use and protected areas to identify Yield (percent) and exclude urban areas or water bodies, for example, which are not suitable for agricultural use. CLIMATIC AND AGRO- 0 CLIMATIC REGIONALIZATIONS Potential Attainable Yield type Actual (OPTIONAL) Based on the factors previously described, climatic and limiting factors such as pests and diseases are not taken agro-climatic regionalizations can be generated. The into account to estimate the potential yield. Attainable differences between these two types of regionalizations yield, on the other hand, adds available water supply to are basically the input data. Climatic regionalizations the factors implemented for the estimation of potential depend only on climatic information while agro-climatic yield. Finally, actual yield adds all additional limiting regionalizations add agronomy or farming systems to the factors like pests, diseases, nutrients, weeds, and so on. climatic database. The statistical analyses used to define Clearly, potential and attainable yields are more easily climatic regions have been described on pages 24–25. estimated from models, while actual yields are derived The same statistical methods can be used to develop from observational information (for example, national agro-climatic regionalizations with the corresponding production statistics). input data. Potential and attainable yields are used to benchmark actual yields and, therefore, suitability. Potential yield is LAND SUITABILITY the maximum possible yield, while attainable yield is the The most recent agro-ecological zoning exercises focus second-highest due to the impact of water supply, and on land suitability, which is defined in terms of potential, actual yield is the lowest. The differences between yields attainable, and actual yields. Potential yield is defined are known as gaps (figure 6.2). The larger the attainable by the amount of CO2, solar radiation, temperature, yield, the more suitable the region; the larger the gap, the genotype, and plant density (table 6.1). Nutrients and more unsuitable the region. Thus, attainable yields and water are assumed to be nonlimiting, and additional gaps can be used to identify the most suitable crops for a Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 39 given region. In addition, the analysis can be performed (b) Formulate crop rotations: This is achieved by for several crops to identify the most profitable combina- taking into account fallow requirements of tions and distributions. the selected cropping pattern. It is calculated based on the requirement of humus levels, and expressed as the percentage of time the land AGRO-ECOLOGICAL ZONING is under fallow. At intermediate input levels, The differences between climatic, agro-climatic, and fallow requirements are 33 percent of those agro-ecological zonings (or regionalizations) are basi- at low input levels while at high inputs, fallow cally characterized by the input data, which increase for requirements are only 10 percent. each one of these regionalizations. In addition to climate (c) Impact of soil erosion on productivity: The and farming systems, agro-ecological zoning includes impact is addressed in three stages: (1) estima- additional environmental factors. The objective is to tion of the potential soil erosion, (2) net soil group units with similar land suitability, potential produc- loss estimated by taking the difference between tion, and environmental impact. It is important to note potential soil erosion and the rate of soil forma- that although the name of all the products described in tion, and (3) estimated limits of tolerable soil this chapter fall under the term agro-ecological zoning, loss under different options of cropping pat- the latest exercises by FAO do not actually define zones. terns and define measures for soil conservation. The use of gridded datasets allows point analyses (that 2. Estimation of Potential Rain-fed Arable is, pixel based) that have rendered regionalizations less Land: Whenever possible or appropriate, combi- common and useful. However, regionalizations can still nations of crops have been constructed. Suitabil- be useful for planning purposes and policy definition and ity classes are defined relating average single crop implementation. suitability to maximum attainable yield. Arable extents estimation is applied in three stages: (a) Determination of the crop combinations ASSESSMENT that perform best under the worst climatic The result from the land suitability assessment is a clas- conditions; sification of crops grown in different land units. Each land (b) Selection of crop combinations with pro- suitability class reflects a range of anticipated yields. This duction stability; and information generates the following assessments, which (c) Among all qualifying crops, a selection are the final products expected from the AEZ analysis: of the combination that maximizes the sum 1. Potential Land Productivity: The assessment of extensions weighted by productivity to allows the selection of crops for each agro-eco- describe the arable land potential. logical cell (AEC) based on potential yield. The 3. Spatial resource allocation (optimizing determination of land productivity requires the land use): The optimization can address single following steps: or multiple factors simultaneously: (a) Formulate cropping pattern options: Sequen- (a) Food output (average yield or minimum tial cropping increases land productivity but yield in bad years) it is only possible when the available growing (b) Net revenue period exceeds the duration of the growth (c) Production costs cycle of a single crop. If the length of growing (d) Gross value period (LGP) is less than 120 days, for exam- (e) Arable land area ple, a single short duration crop is feasible. On (f) Harvested area the other hand, if the LGP is greater than 270 (g) Maximum or net erosion days, crop mixtures with different maturation (h) Self-sufficiency ratio (minimum of the individ- periods are common. ual commodity group self-sufficiency ratios) 40 Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping CHAPTER SEVEN WORLD BANK CASE OF STUDY: AGRO- ECOLOGICAL ZONING IN MOZAMBIQUE9 OBJECTIVES The overarching aim of this project was to develop initial analysis and mapping products to begin to build a framework for addressing agricultural risks from adverse weather and climate variability. A central goal was to identify crop risks from weather and map vulnerability. The technical objectives were: 1. Map climate and agroclimate zones 2. Characterize crop suitability for major cash and food crops 3. Assess crop vulnerability to weather conditions 4. Carry out a field campaign to integrate calibration information METHODOLOGY The methodology is, in general terms, in agreement with the general steps described in chapter 6. The following tasks were performed: (1) mapped climatic and agro-climatic zones based on remote sensing and geographic methods, (2) compiled agronomic and bioclimate datasets for use with Geographic Information Science (GISc) and biogeo- chemical modeling to generate crop suitability indices, (3) modeled yields of major cash and food crops using climate scenarios to generate crop vulnerability indices and risk maps, and (4) carried out field validation and surveys and iteratively utilized feed- back to improve the precision of the crop risk maps. COMPILATION OF LAND RESOURCES Satellite-observed rainfall and temperature data were obtained and analyzed to map spatiotemporal trends and homogenous weather and agroclimate zones. Precipitation This entire chapter is a summary from the following report: Mozambique: Agriculture Weather Risk Mapping. Final Report, 9 World Bank Project # 7159460. Project PI: Nathan M. Torbick; World Bank POC: Carlos E. Arce. Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 41 FIGURE 7.1. MONTHLY AVERAGE ACCUMULATED RAINFALL Left panel: Observed from TRMM and monthly average land surface temperature. Right panel: Observed from MODIS in Mozambique from 2002 to 2010. Generally, the remotely sensed climate products cap- ture distinct seasons across Mozambique; a relatively wet and cooler season from November to April and a relatively dry and hot season from May to September. data were derived from the Tropical Rainfall Measuring MODIS LST at corresponding time periods (annual or Mission (TRMM) and temperature data from MODIS seasonal 10-year averages). Slope and aspect data were MYD11C3. Both datasets were processed to generate combined to calculate the eastern and northern compo- gridded monthly averages at 0.05° (~6 km). Diverse cli- nents of the unit normal vector for inclusion in the regres- matological products were also developed: for example, sion analysis following established guidelines (Hutchinson monthly climatologies of precipitation and temperatures 1998). These components were calculated at multiple (figure 7.1), wettest and driest months, and hottest and scales (see Digital Elevation Model derivatives, Chapter 3, coolest months. page 11). Weather station data were received and quality con- trolled. Secondary quality control processing eliminated ZONAL MAPPING stations with more than 50 percent of missing data. This Zonal mapping relied on the available input parameters data was used to calibrate and validate the remote sensing to delineate homogenous weather zones and agrocli- data. The TRMM datasets were aggregated over time to mate zones for Mozambique. A suite of zonal products reflect 10-year average annual precipitation and 10-year was generated using climate and agricultural spatio- average seasonal precipitation (months DJF, MAM, JJA, temporal information. Variables were primarily derived SON). Station data sets were also aggregated. Coefficients from the remotely sensed weather information (that is, of correlation between TRMM and stations ranged from TRMM PPT and MODIS LST), soils, and topographic 0.75 to 0.89. Regression modeling was completed to iden- modeling. tify the best models for predicting average annual and average seasonal precipitation measurements (box 7.1). To integrate soils information into the zonal mapping and Independent variables that were considered included: to parameterize the crop model, the Harmonized World elevation, aspect, slope, longitudinal convexity, maximum Soil Database (FAO/IIASA/ISRIC/ISS/CAS/JRC curvature, and remotely sensed TRMM precipitation and 2009; see page 11) was utilized. Spatial soil information 42 Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping BOX 7.1. UNIVERSAL KRIGING (UK) 3. Ran K-means clustering algorithm on selected subset of components for annual and each season Overall R2 were relatively high, indicating that regression 4. Mapped cluster identifications for each models using the remotely sensed TRMM data are able to represent a large portion of the variability observed in precipitation at station locations. Finally, universal krig- The summary procedures for the zonal mapping of ing (UK) (interpolation methodology) of precipitation homogenous agroclimate zones were as follows: data was completed using the independent identified vari- 1. Ran PCA for soil variables only and identified ables. The ability to predict station observations with the three components universal kriging was evaluated using root mean squared 2. Ran PCA for climate and soil variables together, errors (RMSEs) and maps of significant error terms. UK using only those soil variables that loaded with the provided the lowest RMSE for average annual precipita- tion over the 10-year time period. RMSEs for the seasonal three components identified above data sets were lowest in comparison directly to the TRMM 3. Ran K-means clustering algorithm on two subsets data and predictions obtained through UK. When using an of components: 1234 and 1245 “average annual” for modeling, UK has the optimal results 4. Mapped cluster identifications for both for predicting station data. The MODIS LST data fol- lowed a similar calibration process. The maps of residuals To smooth out the agro-climatic maps, a generalization for UK consistently identify fewer significant error terms than the maps for ordinary kriging, at least for station loca- scheme was run by applying a majority filter and small tions. However, both the ordinary kriging and UK do a pixel grouping sieve filter. poor job at capturing the spatial patterns observed by the remote sensing observation due to the sparsely populated The homogenous weather maps and the agroclimate maps coverage of the station locations. have similarities and differences. The PCA analysis found that elevation, rainfall, location, and temperature were the most statistically important variables used to determine the homogenous weather regions by explaining 35 percent, 23 was extracted and gridded to identical modeling units as percent, 12 percent, and 11 percent of the variation in the the remotely sensed cells. Topographic information was component loadings. The PCA factors across seasons were obtained from the SRTM (see page 11). As input into the similar except for LST becoming more influential in the zonal mapping procedures, a suite of topographic indi- drier periods (although elevation was always the predomi- ces (for example, slope, aspect, and so on) were generated nant factor). As expected, the elevation gradients can be that were scaled to the identical modeling units for further seen in both the weather and agroclimate zonal map prod- analyses. ucts, which was shown in the PCA results to be a driver of zones. Soil composition, soil moisture attributes, and soil A set of hierarchical zonal maps was generated to iden- conductivity/mobility were the major component loadings tify homogenous climate zones and identify agroclimate for the soil dataset. The drier seasonal climate has some zones. The zonal mapping procedure used Principal influence on the agroclimate zonal maps as seen by the rep- Components Analysis (PCA; see page 24) as a method resentation of southern features in Gaza and Inhambane to reduce data dimensionality and extract unique maxi- relative to the weather zonal map with a stronger coastal mum information. The summary procedures for the factor present likely influenced by the Inter Tropical Con- zonal mapping of homogenous climate zones were as vergence Zone and Indian Ocean temperatures. The final follows: zonal map products were smoothed using a generalization 1. Ran PCA for 10-year average annual, DJF, MAM, majority filter. The maps were generated in a hierarchical JJA, and SON climate variables with elevation fashion with 375 individual zones and 11 major agrocli- derivatives mate clusters that are statistically unique (figure 7.2). To 2. Identified five components for each. Component simplify and provide general details, short, interpretive definitions varied by season titles are provided in table 7.1. Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 43 FIGURE 7.2. FINAL MAP PRODUCT (Left panel) For homogenous weather zones for Mozambique utilizing 10 year monthly average satellite remote sensing weather inputs and geostatistical methodologies. Final map product illustrated for agro-climate clusters map product (right panel) were designed in a hierarchical (scaled) approach with 375 individual zones and eleven major agro-climate clusters that are statistically unique. ASSESS LAND SUITABILITY by agroclimate zone and scaled up to the cluster within Mozambique via DNDC crop yield modeling results. A process-based agricultural productivity model was uti- To answer the question “how well does this crop per- lized to generate crop yield distributions and develop crop form, on average, within this zone and within this suitability index maps. In preparation for quantitative cluster?” for each crop of concern it assumed a wall- analyses of crop susceptibility to climate variability and to-wall distribution of crops in that respective mod- drought, crop suitability was modeled across Mozambique eling unit and ran DNDC using 2010 climate data. using the Denitrification-Decomposition agro-ecological Then DNDC was performed in two modes: one using model (DNDC; see page 13). Suitability was modeled and conventional fertilization (that is, best information on mapped for all major crops including: beans, cassava, cot- actual fertilizer N applications) and one using “opti- ton, groundnut, maize, millet, potato, paddy rice, rain-fed mal” fertilization (that is, fertilizer N is applied when- rice, sorghum, and tobacco. ever soil N falls below crop demand). The first mode addresses the likely situation that crops are farmed A Crop Suitability Index (CSI) for Mozambique was using standard N inputs (which are typically low in developed by using DNDC to understand agricul- Mozambique). The second mode sidesteps the vexing tural risks due to weather fluctuations and climate question of whether suitability is actually related to change. It was achieved by modeling crop suitability underlying geography or low N inputs, and facilitates 44 Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping TABLE 7.1. MAJOR AGRO-CLIMATE interpretation of other important variables (for exam- CLUSTER NAMES IDENTIFIED ple, soil texture or annual precipitation). FROM BIOCLIMATIC AND AGRICULTURAL FACTORS The cropping calendar was based on expert knowledge and FAO Crop Calendar input (http://www.fao.org/ Cluster # Descriptive Name agriculture/seed/cropcalendar/). For soil inputs, the Har- 1 Dry hot semiarid southern lowlands monized World Soils Database was applied. For each 2 Tropical central coast with dry unit in the modeling scheme, the area-weighted mean periods value was calculated for four soil attributes: clay fraction 3 Dry seasonally hot southlands (percent of weight, a proxy for soil texture), bulk density 4 Semi-tropical wet season with dry (g/cm(3), pH, and organic matter content (percent of periods weight). Nitrogen deposition data were based on the values 5 Tropical wet coastal and wet north-central in the original DNDC embedded grid. For 2010 climate data lowlands we relied on daily meteorological data (maximum and mini- 6 Northern mid-elevations with cooler rain season mum temperature in oC and precipitation in cm) derived 7 West-central mid-elevation with cooler wet from the NASA Modern Era Retrospective-Analysis for sea son Research and Applications dataset (MERRA; see page 10) 8 Seasonal valley regions to drive daily climate input requirements. MERRA reanaly- 9 Semi-tropical wet highlands sis is considered among the best and most current multidi- mensional climate data available. MERRA was qualitatively 10 Northern moist cool season assessed using the TRMM and LST products and was found 11 High elevation with cooler tropical wet to be satisfactory and acceptable for this analysis. TABLE 7.2. DNDC CROP MODEL INPUT PARAMETERIZATION FOR GENERATING CROP SUITABILITY INDEX Input Parameters Output Parameters Climate – Daily max and min air temperature Crop Scheme – Photosynthesis – Precipitation – Respiration – Solar radiation – Water demand and transpiration – Atmospheric N deposition – N demands/uptake Soil – Bulk density Soil – C allocation – Texture (clay fraction) – Yield and litter production – Organic C content – Temperature profile – Moisture profile Human Activities or – pH – pH profile Managements – Tillage – Eh profile – Irrigation – Evaporation – Runoff/Fertilization – Water leaching and runoff – Manure amendment – SOC dynamics – Grass cutting – N leaching – Emissions of N2O, NO, N2, NH3, CH4 and CO2 Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 45 FIGURE 7.3. CROP SUITABILITY INDEX FOR SELECTED FOOD AND CASH CROPS WITH CONVENTIONAL FERTILIZATION SHOWN AND 2010 WEATHER DRIVERS Millet and sugarcane tend to have relatively moderate to low suitability. Peanut, sugarcane, and irrigated rice have clusters in the north with higher suitability. To express the results of this analysis, a simple index was and coastal properties. This pattern is highlighted in the used to indicate how well a crop grows relative to “best” figure for beans and cassava. These suitability maps gener- yield within Mozambique: ally agree with actual production and mean yields across Nampula and Zembezi, and portions of Cabo Delgado, yieldcluster / high yield Niassa, and Tete. Beans tend to have a smaller range of suitability as indicated by the CSI maps when compared to where “high yield” is the 95th percentile of yield within cassava with similar high suitability regions but lower CSI Mozambique. in the south and coastal areas of Inhambane and Maputo. The increase in drought tolerant varieties of beans has Maps of crop suitability index were developed for all improved the suitability of beans across Mozambique. crops and with conventional and optimal fertilization. Figure 7.3 shows some of the suitability maps. The CSI readings for maize and sorghum have very simi- lar suitability based on the DNDC simulations with 2010 The selected suitability maps by top areal crop coverage data inputs. The highest suitability categories tend to occur show ranging crop suitability based on crop type and agro- in patchy locations primarily in the central and northern climate clusters. The suitability indices provide a powerful regions of Mozambique. These maize patterns generally tool to compare across crop types, management practices, match agricultural census information for Mozambique. and geography. In general, CSI is higher in the northern There is a notable increase in suitability in the north and 46 Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping coastal regions with optimized fertilizations methods, with TABLE 7.3. CLIMATE SCENARIOS FOR CROP the south remaining less suited for maize. VULNERABILITY ASSESSMENT Annual Annual It is worth noting that the crop suitability index is not Scenario Temperature (°C) Precipitation production-based and does not imply that a location with “more optimal” or “higher” suitability delineates a loca- Baseline no change no change 1 +1 –5% tion where high yields will automatically occur nor zones 2 +2.5 –10% where high yields do occur in reality. Rather, this is a nor- malized map showing the relative mean difference from a 3 +5 –15% highly suitable growing possibility (“higher”) and the rela- tive mean difference for a given crop at a given location. based on generally accepted climate projections from “Other factors” (that is, road networks, distance to market, dynamic and downscaled models, the geostatistical rela- conflicts, market price, inflation, management, adaptations, tionships constructed in this project between temperature fertilizers, and technology) will influence cropping suitabil- and precipitation, and historical and current trends in ity decisions and ultimate suitability. In Mozambique most Mozambique and the region. This approach integrates farmers do not achieve optimal yields due to a variety of the strengths of models, expert knowledge, and current factors (AFTS 2006; Coughlin 2006; INGC 2009; Loening findings. See table 7.3 for the scenarios. and Perumalpillai-Essex 2005; PEDSA 2010). The crop suitability map that was generated is a relative quantita- The crop model was run in both conventional and optimized tive index to provide guidance on relative yields based on fertilization (where fertilizer N is applied when available soil agroclimate conditions with all “other factors” being equal. N drops below crop demand) modes. To express the results of this analysis we created a simple vulnerability index which indicates each crop’s performance relative to the 2010 base- CROP VULNERABILITY line within each agroclimate cluster (see table 7.4): Crop vulnerability and risk to weather was assessed at the 100 × mean rateij / mean baseline rate agroclimate zone scale in Mozambique using yield from the crop model. Other vulnerabilities were also run and where i is the climate scenario and j is the fertilization stored in a GIS. Three strategic scenarios were simulated mode. TABLE 7.4. VULNERABILITY OF ALL CROPS TO CLIMATE CHANGE BY AGRO-CLIMATIC CLUSTER (REGION OR ZONE) Conventional Fertilization Optimal Fertilization 2010 +1°/ +2.5°C/ +5°C/ 2010 +1°/ +2.5°C/ +5°C/ Climate –5%ppt –10%ppt –15%ppt Climate –5%ppt –10%ppt –5%ppt Cluster 1 100% 84% 84% 93% 100% 82% 84% 105% Cluster 2 100% 92% 90% 84% 100% 92% 91% 92% Cluster 4 100% 98% 94% 91% 100% 97% 88% 78% Cluster 5 100% 96% 90% 85% 100% 97% 93% 91% Cluster 6 100% 96% 89% 83% 100% 90% 79% 71% Cluster 7 100% 107% 104% 100% 100% 105% 97% 92% Cluster 8 100% 128% 127% 135% 100% 122% 125% 134% Cluster 9 100% 100% 93% 83% 100% 103% 93% 86% Cluster 10 100% 107% 135% 160% 100% 105% 131% 159% Cluster 11 100% 99% 89% 82% 100% 100% 92% 86% Cluster 12 100% 97% 106% 98% 100% 98% 112% 107% Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 47 FIGURE 7.4. CROP VULNERABILITY INDEX (CVI) FOR SELECTED FOOD AND CASH CROPS WITH CONVENTIONAL FERTILIZATION AND CLIMATE SCENARIO 2 SHOWN COMPARED TO 2010 WEATHER DRIVERS AND YIELD BASELINES Millet, tobacco, soybean, sugarcane, and fruit tree have the largest area of increase in relatively higher vulnerability. Climate Scenario 2 = +2.5 degrees Celsius and –10% total PPT. The vulnerability maps indicate that, in general, clusters increase of 2.5°C and a 10 percent reduction in precipita- in the south-central region of Mozambique and covering tion compared to the 2010 baseline. This does not imply large regions of Gaza and Inhambane are highly vulner- that the optimized yields are lower than the conventional able, with 15–30 percent reductions in yield compared to yields for the same climate scenario; rather, that yields are baselines. In the most extreme scenario, these same clus- lower compared to the relative baseline and in fact the ters show less vulnerability due to the complexity of the highly vulnerable optimized fertilizations yields are typi- processes that are occurring. In addition, these areas have cally higher than the conventional fertilizations yields for a relatively small amount of crop area and not much abso- the same zone (figures 7.5 and 7.6). lute change (although there was some relative change), thereby causing an artifact in the map result (figure 7.4). ADDITIONAL EXAMPLES OF With an increase of 2.5°C in temperature, a reduction AGRO-ECOLOGICAL ZONING of 10 percent in total rainfall, and using conventional AGRO-ECOLOGICAL ZONES, fertilization, the agroclimate zones along the coast and BANGLADESH central Mozambique have a decrease of 5 percent to 15 The most important factors in the definition of agro- percent in yield compared to baseline yields and therefore ecological zoning in Bangladesh are: physiography, soils, become vulnerable. This represents a large sector of the and land level in relation to flooding. Floods represented agricultural industry in Mozambique. In the optimized a particular challenge, so they played an important part fertilization scenarios, the coastal agroclimate clusters and of the regionalization. Flood levels were categorized as south-central clusters become highly vulnerable, with an follows: (1) highlands (land normally above flood level 48 Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping FIGURE 7.5. CVI (BOTTOM) FOR CLIMATE SCENARIO 2 AND CSI (TOP) UNDER CONVENTIONAL FERTILIZER FOR KEY CROPS IN MOZAMBIQUE COMPARED TO 2010 WEATHER DRIVERS AND YIELD BASELINES Spatial comparisons show relatively high suitability cotton clusters become highly vulnerable. Tobacco tends to have substantial area to become highly vulnerable throughout Mozambique. Climate Scenario 2 = +2.5 degrees C and –10 Percent Total PPT. during the flood season), (2) medium highland (land nor- purposes, technology transfer, and specific bio-physical mally flooded up to 90 cm), (3) medium lowland (land resource utilization activities. More information available normally flooded between 90–180 cm), (4) lowland (land at: http://www.banglapedia.org/HT/A_0083.htm. normally flooded between 180-300 cm), and (5) very low- land (land normally flooded deeper than 300 cm during the flood season) and bottomland (sites that remain wet AGROECOLOGICAL ZONING IN THE throughout the year). ILAVE-HUENQUE WATERSHED OF THE ANDEAN HIGH PLATEAU Highland, for example, is considered appropriate for per- This regionalization had the challenge of addressing a ennial dryland crops if the soils are permeable. Imperme- mountainous area, requiring a high density of meteoro- able soils may be suitable for transplanted varieties of rice logical stations to capture the spatial variability. In this if bunds retain precipitation on the fields. The analysis particular case, the analysis relied on several satellite identified 30 agro-ecological zones (figure 7.7), which are sources of information (Geostationary Satellite System, subdivided into 88 agro-ecological subregions, which have Landsat TM, and NOAA/Advanced Very High Resolu- been further subdivided into 535 agro-ecological units. tion Radiometer [AVHRR]). Four agro-ecological zones Agroecological zoning is used extensively for planning resulted: (1) aptitude for crop and pasture, (2) livestock Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 49 FIGURE 7.6. ALL CROPS COMBINED FOR THE SCENARIO 1 (LEFT) AND 2 (RIGHT) ARE HIGHLIGHTED FOR CONVENTIONAL (TOP) AND OPTIMIZED (BOTTOM) CVI In general, clusters in the south and coastal regions are most susceptible to becoming highly vulnerable with warming temperatures and decreases in rainfall. with potential increase, (3) extensive livestock produc- include, for example: substituting bofedales for alpaca; tion, and (4) barren land and areas under grazing with using sheep as a flexible alternative; increasing produc- very shallow soils (figure 7.8). For the first zone, crop tivity through good quality pasture (alfalfa, ryegrass and simulations suggest that productivity can be significantly white clover), which could feasibly generate increases of increased with the help of technology. For example, rain- 40–50 percent in gross income. The study assessed sev- fed potato production could be increased from 5–12t/ eral alternatives to improve the management of natural ha to 18t/ha with irrigation. Suggestions for zone two resources. The involvement of local experts in the analysis 50 Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping FIGURE 7.7. AEZs OF BANGLADESH Source: Banglapedia. should improve the chances that these suggestions will (4) determine crop suitability for optimization of land be finally implemented. More information available at: use. The methodology was based on four basic maps: http://inrm.cip.cgiar.org/home/publicat/01cpb028 soil, physiography, LGP, and bioclimate. The country is .pdf. grouped in 20 agro-ecological regions (figure 7.9) and 60 agro-ecological sub-regions. Constraints and potentials AGRO-ECOLOGICAL ZONES, THEIR were described for each region. Cropping systems were SOIL RESOURCE, AND CROPPING planned to minimize deterioration of land quality (soil SYSTEMS—INDIA physical conditions, nutrient availability, and organic The objectives were to (1) assess yield potentials of differ- carbon pool). More information available at http:// ent crops and combinations; (2) plan crop diversification; agricoop.nic.in/Farm%20Mech.%20PDF/05024-01 (3) plan research and technological dissemination; and .pdf. Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 51 FIGURE 7.8. AEZs OF ILAVE-HUENQUE WATERSHED Source: CGIAR. FIGURE 7.9. AEZs OF INDIA Source: National Bureau of Soil Survey and Land Use Planning, Government of India. 52 Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping CHAPTER EIGHT CONCLUSIONS As stated at the introduction to this document, agro-ecological zoning and, conse- quently, weather risk mapping are becoming increasingly useful techniques in the design of agricultural risk management strategies, and in guiding informed invest- ment decisions. However, the proliferation of different applications and the attendant degree of sophistication are making it increasingly difficult for development practi- tioners to keep abreast of current developments. This overview of agriculture risk mapping techniques aims to serve as an illustrative introduction of the current state of knowledge and practice. There are an increasing number of approaches and models to design risk mappings in agriculture. These are becoming rather complex, given the challenges in modeling the behavior of living plants as they are submitted to various weather conditions. Suc- cess in achieving accurate results largely depends on the quality of the data used, the predicting ability of the models, and the chosen time horizon. Scientists and organiza- tions are undertaking serious efforts to improve the predictability of the applications, hopefully leading to a rapid evolution of such applications in the near future. This document offers a deliberately circumscribed illustration of the various methods that are available, as well as the sources of data that can be used with some degree of confidence, and those products that are highly valuable for the purposes they were designed to serve. The picture that emerges is by no means exhaustive, but hopefully it captures the array of applications in the agricultural sector. The authors hope that this document will help those development practitioners inter- ested in this subject become familiar with the technical aspects involved in the design of these products, and their potential practical uses. Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 53 REFERENCES AFTS. 2006. Mozambique Agricultural Development Strategy. Stimulating Small- holder Agriculture Growth. 32416-MZ. AGROASEMEX. 2006. La experiencia mexicana en el desarrollo y operación de seguros paramétricos orientados a agricultura. Beven, K. J., and M. J. Kirkby. 1979. “A Physically Based, Variable Contributing Area Model of Basin Hydrology.” Hydrol. Sci. Bull. 24: 43–69. Comrie, A. C., and E. C. Glenn. 1998. “Principal components-based regionalization of precipitation regimes across the southwest United States and northern Mexico, with an application to monsoon precipitation variability.” Climate Research 10.3: 201–15. Coughlin, P. 2006. Agricultural Intensification in Mozambique. Infrastructure, Policy, and Institutional Framework. Cressman, G. P. 1959. “An Operational Objective Analysis System.” Monthly Weather Review 87(10): 367–74. Decker, W. L. 1994. “Developments in Agricultural Meteorology as a Guide to Its Potential for the Twenty-first Century.” Agricultural and Forest Meteorology 103: 43–58. Dinku, T., et al. 2007. “Validation of Satellite Rainfall Products over East Africa’s Complex Topography.” International Journal of Remote Sensing 28: 1503–26. Drusch, M., D. Vasiljevic, and P. Viterbo. 2004. “ECMWF’s Global Snow Analysis: Assessment and Revision Based on Satellite Observations.” Journal of Applied Mete- orology 43(9): 1282–94. Escamilla Juárez, J. 2012. Vulnerabilidad del Sector Agrícola en México ante Fenó- menos Climáticos. FAO. 1978–81. Report on the Agro-ecological Zones Project. Vol.1, Methodology and results for Africa; Vol.2, Results for Southwest Asia; Vol.3, Methodology and results for South and Central America; Vol.4, Results for Southeast Asia. FAO/IIASA. 1991. Agro-ecological land resources assessment for agricultural devel- opment planning: A case study of Kenya: Resources database and land productiv- ity, Rome. FAO/IIASA/ISRIC/ISS/CAS/JRC. 2009. Harmonized World Soil Database (ver- sion 1.1), Rome, Italy and Laxenburg, Austria. Frere, M., and G. F. Popov. 1986. Early Agrometeorological Crop Yield Assessment. 73, FAO, Rome. Herman, A., V. Kumar, P. A. Arkin, and J. V. Kousky, 1997. “Objectively Determined 10-day African Rainfall Estimates Created for Famine Early Warning.” International Journal of Remote Sensing 18: 2147–59. Hume, C. J., and B. A. Callander. 1990. “Agrometeorology and Model Building.” Outlook on Agriculture 19: 25–30. Hutchinson, M. F. 1998. “Interpolation of Rainfall Data with Thin-Plate Smoothing Splines ii: Analysis of Topographic Dependence.” Journal of Geographic Information and Decision Analysis 2: 168–85. Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 55 INGC. 2009. Main report: INGC Climate Change Report: Study on the Impact of Climate Change on Disaster Risk in Mozambique. [Asante, K., et al. (eds.)]., Mozambique. Li, C., et al. 2006. “Modeling Nitrate Leaching with a Biogeochemical Model Modi- fied Based on Observations in a Row-Crop Field in Iowa.” Ecological Modelling 196(1–2): 116–30. Li, C., S. Frolking, and T. A. Frolking. 1992a. “A Model of Nitrous Oxide Evolution from Soil Driven by Rainfall Events: 1. Model Structure and Sensitivity.” Journal of Geophysical Research: Atmospheres 97(D9): 9759–76. Li, C., S. Frolking, and T. A. Frolking. 1992b. “A Model of Nitrous Oxide Evolution from Soil Driven by Rainfall Events: 2. Model Applications.” Journal of Geophysical Research: Atmospheres 97(D9): 9777–83. Loening and Perumalpillai-Essex. 2005. “Agriculture and Rural Poverty in Mozam- bique: Dimenions, Profiles, and Trends.” World Bank, Washington, DC. Mavi, H. S., and G. J. Tupper. 2004. “Agrometeorology: Principles and Applications of Climate Studies in Agriculture.” Mesinger, F., et al. 1988. “The Step-Mountain Coordinate: Model Description and Performance for Cases of Alpine Lee Cyclogenesis and for a Case of an Appala- chian Redevelopment.” Monthly Weather Review 116(7): 1493–518. Mozambique: Agricultural Weather Risk Mapping Final Report World Bank Project #: 7159460 Project PI: Nathan M. Torbick World Bank POC: Carlos E. Arce. PEDSA. 2010. “Strategic plan for Agricultural Development.” Republic of Mozam- bique: Ministry of Agriculture. Quijano, J. A., M. R. Paredes, and F. E. Villarreal. 1998. MSPEC.IM, Modelo de Simulación del Potencial Ecológico de los Cultivos, Congreso nacional de Fito- genética, Acapulco, Guerrero, p. 460. Ritche, J. T., et al. 1986. “Model Inputs.” In CERES-Maize: A Simulation Model for Maize Growth and Development. C. A. Jones, J. R. Kiniry, and P. T. Dyke. TAMU Press (Editor). Silva, V. B. S., et al. 2007. “An Improved Gridded Historical Daily Precipitation Anal- ysis for Brazil.” Journal of Hydrometeorology 8(4): 847–61. Sinha, S. K., S. G. Narkhedkar, and A. K. Mitra. 2006. “Barnes Objective Analysis Scheme of Daily Rainfall over Maharashtra (Indi(a) on a Mesoscale Grid.” Atmós- fera 19: 109–26. Uribe Alcántara, E. M. 2010. “Gridded Analysis of Meteorological Variables for Gua- temala and Honduras.” World Bank, Washington, DC. Uribe Alcántara, E. M., and M. D. C. Arroyo Quiroz. 2010. “On the Use of Regu- lar Grids for the Estimation of Synthetic Meteorological Records of Precipita- tion and Their Application on the Assessment of Drought Risk.” Submitted to Atmosphere. Uribe Alcántara, E. M., M. A. L. Montes-León, and E. García-Celis. 2010. “Mapa Nacional de Índice de Inundación.” Tecnología y Ciencias del Agua, antes Ingeniería Hidráulica en México I(2): 73–85. Wan, Z. 2008. “New Refinements and Validation of the MODIS Land-Surface Temperature/Emissivity Products.” Remote Sensing of Environment 112(1): 59–74. 56 Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping Wan, Z., Y. Zhang, Q. Zhang, and Z. Li. 2002. “Validation of the Land-Surface Temperature Products Retrieved from Terra Moderate Resolution Imaging Spectroradiometer Data.” Remote Sensing of Environment 83(1–2):163−80. Wolff, D. B., et al. 2005. “Ground Validation for the Tropical Rainfall Measuring Mis- sion (TRMM).” Journal of Atmospheric and Oceanic Technology 22(4): 365–80. Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 57 APPENDIX A EXAMPLE OF TERMS OF REFERENCE FOR AGRO-ECOLOGICAL ZONES INTRODUCTION These terms of reference detail the objectives, scope of work, and products for hiring a consulting firm (the Firm) to conduct a risk mapping assessment of the agricultural sector. The findings of the tasks detailed here will serve as inputs for the Government of Mozambique (GOM) for the structuring of a risk management strategy to protect small farmers. Findings of this exercise will also be used by the donor community to identify possible areas of development investment opportunities to promote agricul- tural productivity. BACKGROUND AND OBJECTIVE The Agricultural Risk Management Team (ARMT) of the Agriculture and Rural Development Department (ARD) of the World Bank has agreed with the Ministry of Agriculture (MOA) of Mozambique to provide technical assistance in various aspects related to agriculture risk management. The technical assistance that the World Bank is planning to provide to the MOA needs to serve as the basis for the government and private sectors to identify and put in place measures to start managing agricultural risks in a more informed manner and under an agreed framework. The Agricultural Census 2008 (TIA 2008) shows that farmers face production risks related to floods, droughts, cyclones, and wild animals, and it provides some dimen- sion of importance for various agricultural regions. It does not, however, delve into the details to specify those risks by crops. A recent risk assessment on the cotton supply chain also revealed that pest and diseases are important risks facing smallholder farm- ing in the country. Despite the recognized importance of identified production risks, preliminary discussions with various stakeholders in the agricultural sector revealed that there is hardly any technical analysis done in risk identification and risk exposure in agriculture, nor any systematic attempt at quantifying the losses produced by the occurrences of those risks. This situation makes it difficult for the authorities and even the private sector to begin introducing appropriate risk management practices. Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 59 Whenever an adverse catastrophic event has taken place » Define agronomic thresholds to classify Mozam- in the past, the MOA has relied on an ad-hoc system to bique into agricultural homogenous weather zones. collect financing from Ministry of Finance and donors to » Define homogenous weather zones. make available a pool of resources to deliver support to affected farmers. The MOA is therefore much interested COMPONENT 2: CROP SUITABILITY in identifying and designing a risk management frame- The Firm will define and classify crop suitability zones for work that will allow the public sector to implement risk major crops (cash crops and food crops). The result of this management measures in a more planned manner and component is a suitability map that will identify optimal in partnership with the private sector to start managing agricultural investments. The Firm will provide estimates identified agricultural risks. of potential productivity for each land unit. This compo- nent involves but is not limited to the following activities: The World Bank seeks to hire a firm that will undertake » Collect and analyze the following information: a risk mapping assessment of the agricultural sector in soil types; bioclimatic and soil requirements; water Mozambique with the objective to aid policy makers requirements indices (WRSI); topography; and and planners to identify the key crops and exposures to marginal and optimal conditions. loss and which might be selected for designing a national » Define suitability zones for crops based on the pre- agricultural risk management strategy and for future pilot vious information. crop insurance programs. » Implement information, results, and degrees of suitability for each crop in a (Geographic Informa- SCOPE OF WORK/ACTIVITIES tion System (GIS). The Firm will conduct a risk assessment mapping of pro- duction risks for major cash and food crops at the district Identify areas for optimum productivity for food and cash level, taking into account the 10 agro-ecological zones crops based on bioclimatic (that is, altitude, average tem- used by the MOA. The assessment will provide spatially perature, average rainfall, rainfall seasonality, others), soil referenced information on production systems, produc- (that is, effective soil depth, soil texture, slope), and water tion hazards, location of agricultural assets (crops), and availability. This activity should be based on the following farmers’ characteristics. The highest spatial and temporal information: soil, topography, climate, and productivity resolution (that is, the lowest possible level of aggregation) of crops. permitted by the availability and quality of data will be used for the hazard mapping. In order to proceed with the COMPONENT 3: WEATHER required assignment, the Firm will undertake the follow- VULNERABILITY ANALYSIS ing key activities: AND MAPPING A vulnerability map will be defined based on the com- COMPONENT 1: MAPPING HOMOGENOUS parison of suitability zones with actual land use. For the WEATHER ZONES conduction of this exercise, the Firm will need to: The Firm will analyze meteorological time-series, particu- » Analyze actual land use and land cover data. larly rainfall and temperature, in order to propose relevant » Contrast actual land use with suitability zones. homogenous weather zones for agricultural production » Define vulnerability zones per crop. in Mozambique. The classification of the homogeneous » Propose and apply a methodology to estimate vari- weather zones should be based on agronomic criteria. To ations in relation to the average historical yield. conduct this activity, the Firm will also need to conduct » Quantify the areas exposed, number of producers the following activities: per crop exposed, and value at risk per crop season. » Acquire long-term historical rainfall and tempera- ture databases (minimum of 20 years). After the completion of this exercise, the use of GIS must » Perform quality control to the database. be used to present the results. 60 Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping COMPONENT 4: FIELD VALIDATION Chapter 3. Weather Homogenous Zoning The objective for this component is to validate findings by Chapter 4. Crop Suitability Mapping obtaining feedback with samples of producers in various Chapter 5. Vulnerability parts of Mozambique. Some of the activities that may be carried out by the Firm include the following: Chapter 6. Conclusions » Conduct interviews and focus groups with farmers and local experts to identify climatic hazards, grow- PRODUCTS/DELIVERABLES ing periods, crops potential productivity, crops’ The deliverables from this contract are: best and worst yields, crops’ production systems, » A methodology and work plan for organizing the and production areas. weather risk analysis to accomplish the objectives » Identify those food and cash crops that are most based on the key components described (15 days suitable to produce on each homogeneous zone. after contract signature). The Firm is expected Adjusted crop yield values are expected to be esti- to present the methodology and technical details mated by the Firm as well as yield variation due to related to the agrometeorological zoning. changes on optimal agro-climatic conditions (that » An Intermediate Report including the zoning of is, unreliability of rainfall). weather homogenous areas and crop suitability mapping (3 months after contract signature). DATA FORMATS AND » A Final Report, at the satisfaction of the World REQUIREMENTS Bank (5 months after contract signature). All data products are required to have detailed metadata following the World Bank’s metadata standards (ISO TIMESCALE AND 19115 metadata standard). COSTS/BUDGET All spatial data formats shall comply with the Open Geo- The Firm will conduct the activities for this assignment spatial Consortium standards. It is strongly preferred that in a period of 5 months counted from contract signature. the vector data be delivered as shape files with associated The contract costs will be decided after submission of a OpenGIS® Styled Layer Descriptor (SLD) and the Raster technical and financial proposal to the World Bank. data be delivered in the GeoTiff format. All data should be geo referenced and projected in WGS 84 UTM zones. RESPONSIBILITY AND All databases and catalogues shall be delivered as either CONTRACT PAYMENTS Excel 2007 files or Post GIS databases where appropriate. Marc Sadler, Leader of the ARMT, will be responsible on behalf of the World Bank for managing and super- All data shall be delivered on hard disk to the World Bank vising this contract. in a format that will allow their transfer to the Govern- ment of Mozambique. The World Bank will schedule three payments for delivery of the products in the following manner: The findings will be presented in a single document that 1. 20 percent at the delivery of a work plan agreed contains the following structure: with the World Bank 2. 50 percent at the delivery of the Draft Report in Contents reference to 4.3 above, at the satisfaction of the Executive Summary World Bank 3. 30 percent at the delivery of the Final Report in Chapter 1. Introduction reference to 4.4 above, at the satisfaction of the Chapter 2. Methodology World Bank Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 61 » Capacity and experience in using GIS tech- FIRM QUALIFICATION niques for analyzing agriculture/rural-related The World Bank seeks to contract a Firm that has the activities following qualities: » Capacity to conduct risk analysis to estimate » Proven experience to conduct agriculture produc- value-at-risk and expected losses, as used by the in- tion analysis in East and Southern African region, surance industry preferably Mozambique 62 Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping APPENDIX B EXAMPLE OF TERMS OF REFERENCE FOR GRIDDED ANALYSES OF METEOROLOGICAL VARIABLES BACKGROUND AND DESCRIPTION OF THE CONSULTANCY Guided by their technical experience the Consultant must (1) assess the feasibility and (2) create a gridded (that is, mesh-based) product of rainfall, maximum and mini- mum temperature, potential evapotranspiration (Hargreaves), and relative humidity in Nicaragua that could be used by the local insurance industry there to develop an index-based weather insurance market for agriculture. The objective of such a product would be to enable better risk mapping and greater access to risk transfer products in areas with inadequate weather infrastructure. The Consultant should assess the feasibility of creating the data grid to address these spe- cific needs, outlining the steps that would be required to produce such a product if considered feasible. The methodology to assess this feasibility and ultimately to create such a product should be based, but not limited to, the methodology already used to develop a gridded climatological database for index-based insurance purposes in Mexico (Cressman 1959). The consultant will assess the feasibility of creating gridded weather data products in Nicaragua (a feasibility study has already been developed for Guatemala and Hon- duras) based on a blend of existing station data and existing gridded data products (for example, NARR, NOAA’s Climate Prediction Centre datasets, the NCEP/NCAR Reanalysis) to support the weather station-based data observations, and on the perfor- mance of a quality control process to weather datasets. The Consultant must describe the steps that would be required to perform a quality control process and to identify valid records for Nicaragua weather datasets. The methodology to be applied for the detection of discrepancies on climate datasets and a consistent weather datasets, based Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 63 on valid records should be structured, described and pro- and what investments could be made to do so, vided. The methodology will allow the elimination of out- is requested. Deliverable should be submitted in liers, climatic inconsistencies, negative precipitation and Spanish. minimum temperature greater than maximum tempera- ture based on technical and subjective considerations of The characteristics of the gridded product to be defined the expert. include: » Interpolation methodology and technical details The World Bank (The Bank) will provide the Consultant » Temporal resolution with an inventory of weather stations and station data in » Spatial resolution, which is determined based on the countries and any other information that is required the spatial distribution of the meteorological sta- by the Firm to complete the feasibility study. tions » Geographic domain The minimum information for the study includes: » Initial and Final date » Station Catalogue: Station ID, latitude and longi- tude of the stations Under these Terms of Reference (ToR), the consultancy » Dataset: ID, date and readings of precipitation, related to the First Stage is expected to require ___ con- maximum and minimum temperatures and, if fea- sultant-days. sible, any available readings of relative humidity ii. In case the gridded product is determined to be feasible, the second step will consist of creating the DELIVERABLES AND WORK DAYS product for Nicaragua. It will be attempted to re- The deliverables of the study are the following: produce Cressman Modified methodology (1959) i. A (1) report that: using NARR observations in the generation of a. Describes the procedure followed by the Con- their gridded dataset and evaluating its precision. sultant to conduct the process of data qual- The consultant will provide grounds for the use of ity control and identification of valid records Cressman products based on scientific literature in Nicaragua for rainfall, temperature (mini- and its application around the world. mum and maximum temperature). Examples to support the consultant’s findings will be The gridded analysis used by the Consultant will be required. based on the Cressman methodology (Cressman 1959). The estimation of the percentage of valid The methodology consists of correcting a preliminary records for individual stations should be deliv- field based on observations. The preliminary fields used ered in a shapefile, geo referenced and pro- by the Consultant will be the North American Regional jected to WGS 84. A database with the valid Reanalysis (NARR; Messinger et al. 2006) developed records should be structured, described, and by the National Oceanic and Atmospheric Administra- delivered in text format. tion (NOAA). The Consultant will include a gridded b. Summarizes the feasibility of creating gridded analysis for the following variables: (1) Precipitation, (2) analysis (precipitation, maximum and mini- Maximum Temperature, (3) Minimum Temperature, (4) mum temperatures, and evapotranspiration) Potential Evapotranspiration by Hargreaves Method, suitable for weather risk index insurance devel- and (5) Relative Humidity. The last two variables will be opment. Evidence and explanation to support estimated directly from NARR (that is, no application of this conclusion will be required. If deemed fea- the Cressman analysis since the meteorological records sible, the methodology to be used to construct of these variables are scarce in Nicaragua). The only pro- the gridded product and its characteristics cess involved in these particular cases is the estimation should be outlined. If deemed infeasible, sug- of daily data (from NARR’s three hourly reports) and gestions on how the spatial coverage of weather interpolation to match the spatial resolution of the other information in Nicaragua could be improved, grids. 64 Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping The evaluation will consist of comparing the gridded grid and Graphic User Interface) is of ___. The number dataset with climatological dataset from meteorological of days will start to run once the meteorological dataset stations to estimate the error associated with the interpo- is delivered. lation. A discussion on how the interpolation introduces data artifacts (for example, “smoothing” of the original Deliverables are to be submitted in Spanish. values) will be provided by the Consultant. The World Bank and the Inter-American Federation of Insurance Companies (FIDES, acronym in Spanish) can select up to DURATION AND PAYMENTS two temporal resolutions (for example, daily, monthly, and The consultancy expected start date and end date is ___ so on) for the evaluation. and ___ respectively. Finally, a Graphic User Interface (GUI) will be created to A trip to Nicaragua to present the analysis and procedures acquire individual time series from the gridded dataset for followed in the generation of the gridded products and Nicaragua. The user will be able to define interactively results will be conducted by the Consultant. the following parameters: (a) The pixel of interest by geographic coordinates, Payments will be done as per number of days worked with averaged over a geographic area defined by a corresponding evidence of accepted reports and docu- GIS shapefile or an ASCII file with geographic ments. coordinates (b) Variable of interest FIDES will support the review of the products. (c) Period of interest Additionally, the GUI will provide the following mean CONSULTANT PROFILE statistics: The consultancy requires the candidate to have an edu- » Basic Statistics: Mean, Median, Minimum value, cational background in Statistics, Agricultural Insur- Maximum value, Standard Deviation, Variance, a ance, or in a related field. The Consultant should have user defined percentile, a time series plot at least eight years of experience and outstanding exper- tise in the use and applications of gridded data made The GUI will be designed so it can be installed and exe- by NOAA, and must have experience in weather data cuted on any PC with Windows XP without the acquisi- analysis and managing extensive weather stations data- tion of any additional software by the Bank. The selection sets and weather data grids. The Consultant should have of the development environment for the GUI depends worked in Latin American countries, preferably in Cen- entirely on the Consultant. tral America, and have a working knowledge of Spanish and English (spoken and written, particularly). Finally, The deliverables of the second stage include the gridded the Consultant should have knowledge of computer dataset in text format (ASCII), and a report, installation applications related to the legible presentation of com- discs, and tutorial of the GUI. No source codes are part plex weather data that will serve for analytic purposes by of the deliverables. insurance companies. The consultancy related to the Second Stage is expected to require __ consultant-days work. However, total con- RESPONSIBILITY sultant-days work for completing the consultancy (First Technical responsibility and supervision for this consul- Stage: Feasibility Study; and Second Stage: Generation of tancy will be done by ___. Managing Vulnerability and Boosting Productivity in Agriculture Through Weather Risk Mapping 65 A G R I C U LT U R E G L O B A L P R A C T I C E T E C H N I C A L A S S I S TA N C E P A P E R W O R L D B A N K G R O U P R E P O R T N U M B E R 92400 1818 H Street, NW Washington, D.C. 20433 USA Telephone: 202-473-1000 Internet: www.worldbank.org/agriculture