LATIN AMERICA & 92958 CARIBBEAN REGION Environment & Water Resources OCCASIONAL PAPER SERIES Framework for Conducting Benefit- Cost Analyses of Investments in Hydro-Meteorological Systems June 2014 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems June 2014 Arun S. Malik Gregory S. Amacher Jason Russ Enos E. Esikuri Keiko Ashida Tao © 2014 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW Washington DC 20433 Telephone: 202-473-1000 Internet: www.worldbank.org This work is a product of the staff of The World Bank with external contributions. The findings, interpretations, and con- clusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denomina- tions, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. The Environment & Water Resources Occasional Paper Series is a knowledge product of the Environment Unit in the Latin America and Caribbean Region (LCSEN) of the World Bank. The publications in this Series were designed and produced under the supervision of Emilia Battaglini, Acting Sector Manager, Keiko Ashida Tao, Environmental Special- ist, and Anna-Lena Sauer, Consultant (LCSEN). A list of the most recent papers is located at the end of this publication. For electronic copies of all our LAC Environment & Water Resources Occasional Paper Series please visit our web site at http://www.worldbank.org/LAC/LCSENSeries. Rights and Permissions The material in this work is subject to copyright. Because The World Bank encourages dissemination of its knowledge, this work may be reproduced, in whole or in part, for noncommercial purposes as long as full attribution to this work is given. Any queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the Publisher, The World Bank, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2422; e-mail: pubrights@worldbank.org. Photos courtesy of Rachel Pasternack. Front cover: Doppler Weather Radar, Coopers Hill, Jamaica. Back cover: Rio Cobre Flood Warning System, St. Catherine, Jamaica. Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems Foreword The Latin America and Caribbean (LAC) region has Latin America and Caribbean Region (LCSEN) was a unique mix of qualities and challenges when it launched in 2013. The objective of the Series is comes to the environment. It is exceptionally en- to contribute to global knowledge exchange on in- dowed with natural assets—diverse ecosystems in- novation in addressing environmental issues and cluding the world’s greatest carbon sink in the Am- the pursuit of greener and more inclusive growth. azon, globally significant biodiversity such as the The papers seek to bring to a broader public—deci- Mesoamerican Barrier Reef, and valuable crops. sion-makers, development practitioners, academ- At the same time, the region registers the highest ics and other partners—lessons learned from World rates of urbanization in the developing world, wa- Bank-financed projects, technical assistance and ter and natural resources overuse, and increased other knowledge activities jointly undertaken with pollution, with detrimental consequence for the our partners. The Series highlights issues relevant environment and the health of people, especially to the region’s environmental sustainability agen- the poor. da such as biodiversity conservation, natural and water resources management, irrigation, ecosys- Over the past twenty years, the LAC region has tem services, environmental health, environmental made impressive gains in addressing these issues. policy, pollution management, environmental insti- It leads the developing world in biodiversity conser- tutions and governance, environmental financing, vation, natural and water resource management, and climate change and their linkages to develop- and is at the forefront in reducing urban pollution. ment, growth and shared prosperity. The World Bank often has the privilege to partner with countries in the region to pioneer innovative The cases presented in the Series show how the environmental policies and initiatives. Such initia- LAC region continues to make its growth more en- tives include improvement of fuel and air quality vironmentally sustainable and inclusive. We hope standards in Peru, carbon emission reduction in that this Series will make a contribution to knowl- Mexico, payment for ecosystem services in Costa edge sharing among a wider audience within the Rica, participatory and integrated water resources LAC region and globally. management in Brazil, and new approaches to irri- gation management in Mexico. Emilia Battaglini Acting Sector Manager The Environment & Water Resources Occasional Environment Unit Paper Series, a publication of the Environment Unit Sustainable Development Department of the Sustainable Development Department in Latin America and Caribbean Region Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems Table of Contents Foreword. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vi Acknowledgements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii Executive Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 2. Benefits of Hydromet Investments and Their Expected Development Impacts . . . . . . . . . . . . . . . . . . . 3 . . . . . . . . . . . . . . 7 3. Rationale for Hydromet Investments by Public Sector and World Bank’s Involvement 4. Factors That Influence Forecast Value. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 5. Approaches to Estimating Benefits of Routine Climate for Specific Users/Sectors . . . . . . . . . . . . . . 12 6. Approaches to Estimating Country-Level Net Benefits of Hydromet Investments . . . . . . . . . . . . . . . . 16 7. Costs of Hydromet Investments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 8. Proposed Framework for Estimating Net Benefits of Investments in Hydromet Systems. . . . . . . . . . . 26 9. Interim and Ex-Post Evaluations of Hydromet Investments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 10.  Conclusions and Recommendations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 Annex 1: Economic Value of a Forecast – An Illustrative Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 Annex 2: The Value of Hydromet Forecasts to the Insurance and Financial Sector . . . . . . . . . . . . . . . . . . 44 Annex 3: The Value to the Hydropower Sector of Improved Forecasts of Routine Climate. . . . . . . . . . . . . 48 Annex 4: The Value of Hydromet Forecasts to the Transport Sector. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 Annex 5: Economic Analysis of Improving Climate Data and Information Management Project—Jamaica . . . 54 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70 List of Figures Figure 1 Simplified Representation of How Hydromet Systems Generate Social Benefits . . . . . . . . . . . . . . . . . . . . 3 Figure 2  Overview of Framework for Estimating Net Benefits of Hydromet Investments Expected Net Benifits of Hydromet Investments ����������������������������������������������������������������������������������������� 28 Figure A-1 Deriving the Expected Value of a (Perfect) Forecast. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 List of Tables Table 1 Examples of Benefits from Forecasts. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Table 2 Comparison of Country-Level Approaches. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 Table 3  Benchmarking – First Stage Parameter Values. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 Table 4  Comparison of Results of Assessment of Hydromet Modernization Using Benchmarking and Sectoral Approaches��������������������������������������������������������������������������������������� 20 Table 5 Hydromet Modernization Cost Components from Selected Studies. . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 Table 6 Expected Cost Example (Euros Million). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 Table A-1 Payoffs for Alternative Harvesting Decisions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 Table A-2 Benefits to Hydropower on the Missouri River from Improvements in Forecasting Ability . . . . . . . . . . 48 Table A-3  Estimated Annual Total Benefits (Euros) Per Capita and Per Driver from Weather Forecasts and Warnings Under Two Hydromet Assumptions ������������������������������������������������� 52 Table A-4 Examples of Benefits from Improved Climate Forecasts. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 Table A-5 Major Damage-Causing Meteorological Events, 2000–2010. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 iv Table A-6 Average Damages from Hurricanes by Sector (USD Million, Constant 2010) . . . . . . . . . . . . . . . . . . . . 61 Table A-7 Estimates of Percentage Losses Avoided with Project. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .63 Table A-8  Expected Annual Benefits of Improved Forecasts of Extreme Hazards and Early Warning Systems, Baseline Year (USD Million, Constant 2010) ��������������������������������������������������� 64 Table A-9  Expected Annual Benefits of Improved Forecasts of Routine Climate for Selected Sectors (Million USD, Constant 2010) ������������������������������������������������������������������������������������������������������������������� 65 Table A-10a Project Benefit-Cost Ratios Assuming No Climate Change for Alternative Benefits Estimates. . . . . . . 68 Table A-10b Project Benefit-Cost Ratios Assuming Moderate Climate Change for Alternative Benefits Estimates������������������������������������������������������������������������������������������������������������������������������������� 68 Table A-10c Project Benefit-Cost Ratios Assuming High Climate Change for Alternative Benefits Estimates . . . . . 69 Table A-11a Project Benefit-Cost Ratios Assuming No Climate Change and Reduced Damage Due to Mitigation for Alternative Benefits Estimates������������������������������������������������������������������������������� 69 Table A-11b Project Benefit-Cost Ratios Assuming Moderate Climate Change and Reduced Damage Due to Mitigation for Alternative Benefits Estimates ��������������������������������������������������������������� 69 Table A-11c Project Benefit-Cost Ratios Assuming High Climate Change and Reduced Damage Due to Mitigation for Alternative Benefits Estimates ��������������������������������������������������������������� 69 v Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems Abbreviations ADM Administrative Cost GFDRR Global Facility for Disaster Reduction AHDL Advisors to the High Level Group on and Recovery Infrastructure Charging HLC Hospital Care Cost ASEAN Association of Southeast Asian Nations NLP Net Lost Production BFFWC Bangladesh Flood Forecasting and PDV Property and Material Damage Cost Warning Centre PLA Percentage Loss Avoided CAT Catastrophe PLC Price Level Coefficient CCRIF Caribbean Catastrophe Risk Insurance PPCR Pilot Program for Climate Resilience Facility SPCR Strategic Program for Climate Resilience DHMZ Meteorological and Hydrologic Service of UNISDR United Nations International Strategy for Croatia Disaster Reduction ENSO El Niño Southern Oscillation U.S. United States EPNF1 Expected Payoff for Option One USD United States Dollar EPNF2 Expected Payoff for Option Two VOSL Value of Statistical Lives Saved EPPF Expected Payoff with Perfect Forecast WRMA Weather Risk Management Association EUR Euro WTP Willingness To Pay EVPF Expected Value of Perfect Forecast WMO World Meteorological Organization GDP Gross Domestic Product vi Acknowledgements This paper was prepared by Arun S. Malik (Profes- task team comprised of Enos E. Esikuri, Richard sor, Department of Economics and School of Pub- Damania, and Keiko Ashida Tao. The team would lic Policy and Public Administration, George Wash- like to acknowledge the comments and the guid- ington University), Gregory S. Amacher (Julian N. ance received from the reviewers, which includ- Cheatham Professor of Natural Resource Econom- ed Dugkeun Park, Rohan Longmore, and Momoe ics, Department of Forest Resources and Environ- Kanada (World Bank). mental Conservation, College of Natural Resources, Virginia Polytechnic Institute and State University), Funding for this work was made possible through Jason Russ (Adjunct Professor, George Washing- the Jamaica Improving Climate Data and Informa- ton University), Enos E. Esikuri (Task Team Lead- tion Management Project which is financed by the er and Sr. Environmental Specialist, World Bank), Pilot Program for Climate Resilience (PPCR) under and Keiko Ashida Tao (Environmental Specialist, the Strategic Climate Fund of the Climate Invest- World Bank). This work was led by the World Bank ment Funds. vii Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems Executive Summary The importance of hydro-meteorological (hydrom- • surveys approaches that have been used to es- et) services in the form of weather, climate, and hy- timate the value of improved forecasts to spe- drological forecasts is in many ways self-evident. cific user groups or sectors of an economy, as Across the world, hundreds of thousands of weath- well as approaches that have been used to es- er forecasts, severe weather warnings, and climate timate the net benefits of improved forecasts at predictions are issued each year. These forecasts the country level; are used by myriad users, ranging from households • lays out a framework for estimating the expect- to firms to government agencies. ed net benefits of hydromet investments at a country level without onerous data or analytical Determining the appropriate level and type of in- requirements; and vestment in hydromet systems requires a compar- • describes data that can be collected to conduct ison of the social benefits and costs associated interim and ex-post evaluations of hydromet in- with different levels and types of investment. In re- vestments that can potentially enhance the net cent years, tight government budgets have result- benefits yielded by current and future hydrom- ed in renewed efforts to identify and estimate the et investments. net social benefits of the services provided by hy- dromet agencies. The call for such estimates is es- The proposed framework draws on existing ap- pecially strong when large-scale investments to proaches with some important modifications and improve or maintain hydromet services are being additions. In particular, it attempts to capture the contemplated. benefits from improved forecasts of extreme me- teorological events as well as the benefits from im- This whitepaper provides a survey of the issues in- proved forecasts of routine climate. To the extent volved in estimating the (expected) net social bene- possible, estimates of the first type of benefits are fits of investments in hydromet systems, and it pres- based on historical data on losses from extreme ents (in Section 8) a framework that can be used to events. The proposed framework makes extensive obtain first-cut estimates of these net benefits at use of benefits transfer. However, the results of a country level by those tasked with evaluating hy- benefits transfer should be subject to review and dromet investments. Specifically, the whitepaper: revision by sector experts. • provides an overview of the types of benefits Given the high degree of uncertainty regarding associated with hydromet investments, the pro- the precise magnitude of benefits and costs as- cess by which the benefits are generated, as sociated with hydromet investments, conducting a well as their expected development impacts; wide range of sensitivity analyses is an essential • explains the rationale for public sector invest- component of the proposed framework. Sensitivi- ment in hydromet systems and for involvement ty analyses should be conducted using alternative by the World Bank; estimates of benefits and costs, alternative time • discusses the wide range of factors that influ- horizons and discount rates, alternative assump- ence the magnitude of benefits generated by tions about the consequences of climate change hydromet systems; for the frequency and magnitude of extreme viii events, and alternative assumptions about the abil- Benefits of this type that are likely to be large in ity of households and enterprises to adapt to these magnitude should be identified and characterized events. even if they cannot be quantified or monetized. Not all of the benefits of hydromet investments Supplementing ex-ante evaluation of a hydromet in- can be quantified and/or monetized. Although vestment with interim and ex-post evaluations offers these benefits fall outside the framework of a tra- a range of benefits beyond the obvious ability to as- ditional economic evaluation (in which all bene- sess the validity of the ex-ante evaluation and refine fits and costs are monetized), they should not be it. In particular, priorities and allocations of funds for ignored. Examples of such benefits are improved the investment being evaluated can be revised, and budgeting and contingency planning for extreme design of future hydromet investments can be im- meteorological events, and improved ability to de- proved. Accordingly, collection of data for conduct- velop higher-resolution climate change scenarios ing interim and ex-post evaluations plays an import- that can enhance climate-resilience planning and ant role in enhancing the net benefits from hydromet decision-making. investments in the long run, in addition to enhanc- ing the ability to evaluate the investments ex-ante. ix Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems 1. Introduction The importance of hydro-meteorological (hydrom- at a country level without onerous data or analyti- et) services in the form of weather, climate, and cal requirements. hydrological forecasts is in many ways self-evi- dent. Across the world, hundreds of thousands of The whitepaper is organized as follows: weather forecasts, severe weather warnings, and climate predictions are issued each year. These • Section 2 provides an overview of the types of forecasts are used by myriad users, ranging from benefits associated with hydromet investments, households to firms to government agencies. For the process by which the benefits are generat- the U.S. alone, it is estimated that households col- ed, and their expected development impacts; lectively make use of 300 billion weather forecasts • Section 3 explains the rationale for public sec- each year (Lazo et al. 2009), and that approximate- tor investment in hydromet systems and in- ly one-seventh of the country’s economy is weather volvement by the World Bank; sensitive (National Research Council 1998). • Section 4 discusses the wide range of factors that influence the magnitude of benefits gen- These observations, striking though they are, tell us erated by hydromet systems, in particular the little about the appropriate level of investment by value of weather/climate forecasts. The dis- society in hydromet systems (Pielke and Carbone cussion is supplemented by a stylized example 2002). Determining the appropriate level and type presented in Annex 1; of investment requires a comparison of the social • Section 5 provides an overview of approaches benefits and costs associated with different levels that have been used to estimate the value of and types of investment. In recent years, tight gov- improved forecasts of routine climate to specif- ernment budgets have resulted in renewed efforts ic user groups or sectors of an economy; to identify and estimate the net social benefits of • Section 6 then turns to an overview of approach- the services provided by hydromet agencies (e.g., es that have been used to estimate the net ben- Pielke and Carbone 2002, Morss et al. 2008, Mills efits of hydromet investments at the country 2010). The call for such estimates is especially level. The primary benefits estimated by these strong when large-scale investments to maintain or approaches are those associated with improved improve hydromet services are being contemplated. forecasts of extreme meteorological events; • Section 7 contains a discussion of the costs of In this white paper we provide a survey of the issues hydromet investments, with particular atten- involved in estimating the (expected) net social tion given to the challenges faced in estimating benefits of investments in hydromet systems, and these costs in developing countries; we lay out a tentative framework that can be used • Section 8 lays out a framework for estimat- to estimate the net benefits of these investments ing the expected net benefits of hydromet 1 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems investments at a country level. The proposed • Section 9 describes data that can be collected framework builds on existing approaches and to conduct interim and ex-post evaluations of is designed to be used with data available hydromet investments that supplement and re- from secondary sources. This section will be fine ex-ante evaluations of these investments; of central interest to those tasked with con- and ducting economic evaluations of hydromet • Section 10 offers conclusions and recommen- investments; dations. 2 2. Benefits of Hydromet Investments and Their Expected Development Impacts The benefits of hydromet systems stem from the A stylized, yet informative representation of how services that they provide. Efforts to estimate the forecasts result in social benefits is presented in benefits of hydromet services date back at least 40 Figure 1. The figure is adapted from Morss et al. years. At this point there is a small, and still grow- 2008 and Lazo et al. 2008. The box labeled “Hy- ing, literature that examines the use and value of dromet System” at the top of the figure encom- weather forecasts.1 Recent assessments of the cur- passes the infrastructure of hydromet systems, rent state of knowledge reveal that despite these including weather satellites, Doppler radar, radio- efforts many challenges remain. Our understanding sondes, ocean buoys, river gauges, and weather of the value of forecasts is still patchy and incom- stations, along with the data assimilation, numer- plete (e.g., Pielke and Carbone 2002, and Morss ical modeling and other activities that are under- et al. 2008). In large part, this is a function of the taken to generate forecasts. These forecasts are very diverse set of users that make use of fore- communicated to users either directly or indirect- casts and the multitude of contexts in which fore- ly through intermediaries that repackage the fore- casts are used. To accurately determine the value casts (such as local news channels) or add value of forecasts to each of these users requires a fairly to them (such as specialized forecasts for the avi- detailed understanding of how each user interprets ation industry). forecasts and combines them with other informa- tion in its decision-making processes, and how the outcomes of the decision-making processes are al- Simplified Representation of How Figure 1  tered by the forecasts. For instance, the manner in Hydromet Systems Generate Social which a small farmer uses a forecast is far different Benefits from the manner in which a hydro-electric genera- tor does so. The farmer’s decision-making process Hydromet System is likely to be informal whereas the hydro-electric generator’s is likely to make use of sophisticated Observations/Forecasts decision support tools. Moreover, the changes in- duced by the forecasts in the outcomes of their de- Intermediary/Media Communication cision-making processes are likely to vary markedly in both character and magnitude. Observations/Forecasts Even for a single farmer, the types of forecasts Users used by the farmer and the decision contexts can vary. For instance, the farmer’s decision on which Economic, Decision crop varieties to plant will depend on seasonal fore- Social, Political Decision-Making Support Contraints Tools casts, such as forecasts of whether rainfall during a coming monsoon will be above or below average, Outcomes and Benifits whereas the farmer’s decision on when to plant will depend on near-term forecasts of precipitation. Source: Authors. 1  Surveys of this literature can be found in Katz and Murphy (2005) and Katz and Lazo (2011). 3 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems A diversity of users combine these forecasts with On an economy-wide level, the value of improved other information in their decision-making process- forecasts varies across sectors. Some sectors of es to make choices. In the case of larger users, de- the economy are more climate sensitive than oth- cisions are likely to be made with the aid of deci- ers, and are therefore more likely to benefit from sion support tools. Users’ choices together with improved forecasts. The sectors generally consid- actual meteorological conditions determine out- ered to be climate sensitive are: agriculture, avia- comes. These outcomes range from individuals not tion, construction, surface and water transporta- getting wet when it rains, to higher crop yields, to tion, water resources, energy, fisheries, forestry, fewer heat-wave casualties. These outcomes deter- health, and tourism and recreation (Houston et al. mine the change in expected net economic bene- 2004 and World Bank 2008). fits, broadly defined, enjoyed by users, or society, as a result of the forecasts. The economic value The Spectrum of Benefits. Table 1 presents ex- of (better) forecasts is the aggregate increase in amples of the use of different types of forecasts in expected (net) economic benefits enjoyed by all different sectors of the economy and the variety of users. benefits that they generate. The table emphasiz- es benefits associated with forecasts of routine cli- Classification of Benefits. A wide range of ben- mate, which are often less apparent. The examples efits have been attributed to investments in hy- are drawn from a wide variety of sources. Following dromet systems and to improving dissemination convention (e.g., Sene 2010), short-range forecasts of information about meteorological conditions generally extend 2 days from when the forecast is is- and hazards. The benefits are associated with im- sued, whereas medium-range forecasts extend from proved forecasts of extreme meteorological events 3 to 10 days into the future. Seasonal forecasts are such as hurricanes, storms, floods, and droughts, descriptions of average weather parameters over as well as improved forecasts of routine weather the next 3 to 6 months, excluding individual events. and climate conditions. The examples illustrate the pervasive importance of climate forecasts to economic activity. Improved forecasts of extreme meteorological events and effective dissemination of information Importance of Unquantifiable Benefits. An im- about their effects and appropriate responses to portant subset of the benefits of hydromet invest- them can substantially reduce economic losses ments is likely to be difficult to quantify. Ignoring caused by the events. Improved forecasts of rou- such benefits is justifiable if, and only if, they are tine climate can result in increased enterprise prof- likely to be small in relative magnitude. If they are its (or reduced costs) and improved decision-mak- likely to be large in magnitude, they should be iden- ing by households. We classify benefits into three tified and described even if they cannot be quan- categories: tified. Examples of benefits of hydromet improve- ments that are difficult to quantify but that can be • benefits from improved forecasts of extreme large in magnitude include: meteorological events; • benefits to enterprises of improved forecasts of • Improved budgeting and contingency plan- routine climate; and ning for extreme meteorological events. Better • benefits to households of improved forecasts forecasts enable governments to develop bet- of routine climate. ter plans for coping with extreme events and 4 Table 1  Examples of Benefits from Forecasts Period of Sector Forecast Benefits from Forecasts Agriculture Short-Range Information on daily precipitation is vital to pesticide application decisions, as Forecasts heavy rains can wash away recently applied pesticides. Medium-Range Accurate medium term precipitation forecasts inform farmers whether or not Forecasts they need to irrigate, and how much. Ideal seeding rates are sensitive to the weather conditions in the days and weeks following planting. Timing of planting and harvesting decisions can be improved. Seasonal Having more accurate data on the seasonal climate can aid farmers in deter- Forecasts mining which crops will yield more value. Crop insurance programs can benefit from reduced uncertainty of weather pat- terns. Agencies can anticipate food shortages earlier with better seasonal forecasts. Household Short-Range Weather forecasts are used to make everyday decisions, such as what to Forecasts wear, or whether or not to take an umbrella. Medium-Range Early warnings of major storms can signal the need for a household to stock Forecasts up on essentials in case of power outages or road closures, or for potentially life-saving evacuations in cases of extreme events. Seasonal Seasonal forecasts can inform households on many decisions, ranging from Forecasts whether or not and how much insurance to purchase, to what type of house- hold fortifications and improvements to undertake. Energy Short-Range Daily weather patterns have an effect on peak energy use patterns, and more Forecasts accurate weather forecasts can inform power plants when to increase or de- crease electricity production. Seasonal Hydro-electric generators benefit from improved streamflow forecasts. Forecasts Recreation and Golf course management, recreational fishing, and other outdoor and ma- Short-Range Tourism Forecasts rine-based activities benefit from more accurate temperature and precipita- tion forecasts. Seasonal Tourist resorts make staffing and investment decisions based on expected Forecasts tourists in a given season, which can be very sensitive to seasonal weather. Transportation Short-Range Routing decisions of trucks, ships, and airplanes can be improved with better and Shipping Forecasts forecasts of daily weather conditions. Medium-Range Cargo ships can use better forecasts to minimize costs and delivery delays in Forecasts maneuvering around unexpected storms. Water Resource Seasonal Improved forecasts can lead to more efficient reservoir operations and savings Management Forecasts from avoiding groundwater pumping to augment reservoirs. Fisheries Medium-Range Fishing vessels rely on weather forecasts to determine when to set sail, how Forecasts long to stay at sea, and where to navigate to avoid adverse weather conditions. Emergency Short-Range Adequate warning before extreme weather disaster can significantly reduce Response Forecasts losses of life. First responders and emergency rescuers can pre-position emer- gency response assets to places where they will be most effective for rescue operations. More accurate forecasts can reduce the costs associated with “false alarms,” including the costs of evacuating communities. Source: Authors. 5 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems develop better estimates of the sums needed expressed in common units. In general, monetiz- to cope with such events; ing costs is not problematic given that they are eas- • Increased ability to develop higher-resolution ily observed and measured. In contrast, monetiz- climate change scenarios that can enhance cli- ing some benefits of hydromet improvements can mate-resilient planning and decision-making; be problematic. A necessary condition for mone- • Improved ability of individuals, enterprises, and tizing benefits is the feasibility of quantifying the government agencies to adapt to changes in benefits in physical terms. For example, benefits of climate given better information about climate improved forecasts for agriculture can be quanti- trends and future changes; fied in terms of the magnitude of expected increas- • Diminished likelihood of maladaptations be- es in annual crop yields. This quantified benefit cause of poor or inaccurate information about can then be monetized by applying suitable pric- climate trends and future changes in climate; es for the crops. The examples in the preceding and paragraph indicate that in some cases it may not • Facilitated development of weather insurance be feasible to quantify benefits. In other cases, it instruments that can help users manage the may be possible to quantify benefits, but monetiz- large risks associated with extreme meteo- ing them may be difficult or controversial because rological events. Further detail on this ben- identifying a suitable price (or economic value) to efit is provided in Annex 2 at the end of this attach to the quantified change may be difficult. Ex- document. amples of such benefits are expected reductions in mortality and morbidity due to more timely, or Monetized vs. Non-Monetized Benefits. An eco- more accurate, forecasts of extreme meteorolog- nomic evaluation of a project should ideally mon- ical events. Although economic values can be at- etize all benefits and costs, i.e., all benefits and tached to these benefits, the precise values to be costs should be expressed in terms of monetary attached can be subject to disagreement. In such (e.g., dollar) values. This allows for a clear-cut cases, benefits can be expressed in their “native” comparison of benefits and costs since they are units instead of being monetized. 6 3. Rationale for Hydromet Investments by Public Sector and World Bank’s Involvement Hydromet services are typically divided into two forecast. The distinguishing characteristic of pure broad categories: (i) basic services and (ii) special public goods is that it is infeasible or prohibitive- services. Both of these services are derived from ly expensive to exclude one group from benefiting the basic data collection and processing infrastruc- from the good if it is made available to some other ture that make up a hydromet system, often re- group. For example, if basic weather forecasts are ferred to as basic systems. Basic services comprise made available to households, it is infeasible to ex- the weather and climate forecasts made available clude enterprises from learning of these forecasts to the public through mass media. In addition, ba- and benefiting from them. The pure public good na- sic services generally include forecasts designed ture of basic weather forecasts implies that it is in- for large community sectors, such as farming, fish- feasible for the private sector to provide them, or to ing and sailing. In contrast, special (or value-add- provide them at a socially desirable level. Attempts ed) services are tailored to specific users or small by a private forecast provider to collect payments user groups, such as construction, shipping and for the forecasts would be thwarted by the ability aviation. The specialized forecasts differ from ba- of users to free-ride off forecasts purchased by oth- sic forecasts in the frequency of the forecasts, ers. In general, pure public goods, such as basic their spatial location and resolution, and the set of weather forecasts, are therefore provided by the weather variables that are forecast. public sector and paid for out of general revenues. Basic services are traditionally provided by the pub- Specialized (or value-added) weather forecasts, as lic sector through a national meteorological ser- opposed to basic forecasts, are provided success- vice, as part of a government’s obligation to pro- fully by the private sector, or by the public sector, for tect the life and property of citizens and to enhance a fee. As noted, these forecasts are tailored to spe- their welfare. These basic services, and the basic cific users or small user groups. The tailored nature systems on which they are based, have long been of these forecasts implies that forecasts prepared viewed as best provided by a national government for one user group are unlikely to be of high value to avoid the wasteful duplication of infrastructure to another user group, diminishing the free-riding that would arise if the services were provided by problem. For example, specialized forecasts pre- competing private enterprises (Zillman and Free- pared for offshore oil and gas producers are likely to bairn 2001). be of little value to inland construction companies. A further rationale for public sector provision of ba- Building climate resilience is critical for the World sic services is that they have the characteristics of Bank Group’s twin goals to end extreme poverty “pure public goods”. The defining characteristic of and build shared prosperity. Climate change will a public good is that provision of the good to one have huge impacts on poor and marginalized popu- group (of individuals or enterprises) does not di- lations, who commonly live in the highest-risk areas minish the ability of another group to simultane- and have the least ability to recover from extreme ously benefit from the good. For example, if house- and/or recurrent low-intensity events. Many low in- holds benefit from a forecast of the likelihood of come countries are also those with the least ca- precipitation tomorrow, this does not diminish pacity to prepare for, and absorb, the effects of cli- the ability of enterprises to benefit from the same mate events. Unless development projects address 7 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems their preparedness and adaptation to current and supports hydro-meteorological investments close future climate and disaster risks, physical invest- to USD 500 million, including integrated support ments made would be repeatedly wasted after ex- in Central Asia, Mozambique, Nepal, and Yemen. treme climate events and disasters occur. With in- In 2011, the Global Facility for Disaster Reduction creasing change in the climate and uncertainty of and Recovery (GFDRR) launched a hydro-meteo- natural disasters, building climate and disaster re- rological initiative to support and leverage World silience must assume a central role in develop- Bank investments to strengthen weather, climate ment and in national plans. Risk reduction and bet- and hydrological services, and ensure that World ter preparedness to deal with climate and disaster Bank investments support and contribute to inter- impacts can substantially decrease the cost of di- national norms, standards, systems and efforts un- sasters. Early warning systems, better prepared- der the auspices of the World Meteorological Or- ness, and improved safety codes have proven to be ganization (WMO) and the Global Framework for cost effective, save human lives, and protect public Climate Services (World Bank/GFDRR, 2013). The and private investments. Climate and disaster re- Pilot Program for Climate Resilience (PPCR) has silient development, therefore, makes sense both helped channel more than USD 95 million in invest- from poverty alleviation and economic perspec- ments and USD 50 in co-financing through the Stra- tives (World Bank/GFDRR 2013). tegic Programs for Climate Resilience (SPCRs) into hydromet related projects in Cambodia, Grenada, Weather, climate and hydrologic monitoring and Nepal, Niger, Saint Vincent and the Grenadines, forecasting are essential to inform decision-mak- Samoa, Tajikistan, Zambia, Mozambique, Jamaica, ing for climate resilience and provide critical inputs Yemen, and Bolivia. to early warning systems. The World Bank currently 8 4. Factors That Influence Forecast Value As the examples in Table 1 earlier suggest, the mag- manner that users can understand and incorpo- nitudes of benefits associated with improved fore- rate in their decision-making processes. It has tra- casts will vary across users and decision contexts. ditionally been assumed that providing more ac- In addition, the benefits depend on the precise curate forecasts at higher spatial and temporal characteristics of the forecasts being evaluated. resolutions will enable users to make better de- The forecast characteristics that are of importance cisions, which in turn will result in larger benefits are (e.g., Mjelde et al. 1989, Stern and Easterling (Millner 2009). However, as has already been em- 1999, Blench 1999, Houston et al. 2004, and Teis- phasized, users differ widely. Some may be able to berg and Weiher 2009, Lazo and Waldman 2011): process and benefit from improved forecasts, but others may not and may instead be overwhelmed • time frame (span) of the forecast, e.g., fore- when provided with more detailed and disaggre- casts can be made of tomorrow’s weather or gate forecasts (Morss et al. 2005, 2008). the coming summer’s average weather; • lead time of the forecast, i.e., the length of time In countries that have fairly modern, well-devel- between the issuance of a forecast and the oped hydromet systems, investments in improving time of the event forecasted; how forecasts are communicated and incorporated • spatial resolution of the forecast; into users’ decision-making process may yield high- • set of weather parameters forecast, e.g., rain- er benefits than investments in more sophisticated fall, temperature, etc.; and infrastructure. As Mills (2010) notes, “…it may be • perceived and actual accuracy of the forecast— more effective to change the message or the medi- perceived accuracy can differ from actual ac- um rather than invest in a new monitoring technol- curacy, especially if past forecasts have been ogy or supercomputer. Often this involves greater wrong. consideration of users and the decisions they face”. Longer lead times for forecasts are obviously ben- In countries with outdated or poorly-developed hy- eficial if they are not accompanied by marked in- dromet systems, investments in infrastructure are creases in uncertainty. For example, a very accu- a necessary condition for providing basic meteo- rate flood forecast with limited lead time might rological services. However, investments in infra- enable households to flee a flood zone, but might structure alone will not ensure that the potential not give them time to move their belongings to benefits of providing these services are realized. In- higher ground. The importance of lead times can vestments in systems that effectively communicate be more pronounced in poor communities: individ- the forecasts to users and help them take advan- uals who do not have access to transport or are tage of the forecasts is critical. This is especially reliant on animal-powered transport cannot move true for “small” users, such as individual house- themselves and their belongings as quickly as holds or small enterprises, and it is likely to be es- those who have access to motorized transport. pecially true in the case of small users in low-in- come countries who have limited education. Ability of Users to Understand and Incorporate Forecasts in Decision-Making. For forecasts to Even “large” users may be overwhelmed by more be of value to users, they must be presented in a detailed and disaggregated forecasts. In a recent 9 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems paper, Webster (2012) observes that: “Paradoxi- will generally differ from its actual accuracy, espe- cally, as forecasts become better and their reso- cially if past forecasts have been wrong. For exam- lution grows, it becomes more difficult for devel- ple, in South Florida, there was limited response oping countries to access them.” He notes that to warnings issued in advance of Hurricane Wilma global weather forecasts prepared by the Nation- in 2005. One of the explanations offered for the al Centers for Environmental Prediction in the Unit- limited response is that almost all of the six hur- ed States are posted daily on the Internet. Howev- ricane warnings issues in the preceding two years er, downloading the large amounts of information had been false alarms (Meyer 2006). associated with these forecasts and extracting the relevant regional component in a timely manner Strikingly, most forecasts issued today are sin- requires reliable, high-speed internet access. In gle-valued: they provide a predicted value of a developing countries, this is problematic given of- meteorological parameter with no objective infor- ten costly and slow internet connections. Thus, for mation about the accuracy of the forecast. As a re- the benefits of investments in improved hydrom- sult, any information about the forecast’s accura- et systems to be fully realized, investments in as- cy must be derived by the user, and will therefore sociated communications infrastructure may well be subjective. This is true even though forecasts be needed. are invariably subject to uncertainty, with the un- certainty typically increasing with the lead time In addition to this data acquisition problem, there is of the forecast. There has been a push in recent the issue of processing the data to generate infor- years to convey forecast uncertainty to users, so mative local forecasts in a timely manner. Webster that users are able to make better use of the fore- (2012) cites the case of the 1 to 10-day flood fore- casts in their decision-making (National Research casting system for Bangladesh created in 2007 by Council 2006, American Meteorological Society the European Centre for Medium-Range Weather 2011). In principle, conveying forecast uncertain- Forecasts and other parties. In 2009, in an effort to ty is also likely to engender greater long-run confi- boost regional capacity, the flood forecast modules dence in forecasts among users, because they will were handed over to the Bangladesh Flood Fore- be explicitly aware of the probability that a fore- casting and Warning Centre (BFFWC). However, the cast is wrong. large volume of data generated by these modules proved to be too difficult for the BFFWC to handle. However, simply conveying uncertainty about fore- As a result, responsibility for using the models to casts is not sufficient. Many, perhaps most, users prepare flood forecasts was handed over to an in- may not have the skills needed to properly incor- ternational non-government entity, the Regional In- porate the uncertainty in their decision-making tegrated and Multi-Hazard Early Warning System. (Morss et al. 2008). Some users may in fact pre- fer the seeming certainty of single-valued forecasts Conveying Forecast Uncertainty. As noted in the to the explicit uncertainty of probabilistic forecasts. above list of forecast characteristics, users must Accordingly, calls for forecasts to include informa- trust the accuracy of forecasts if they are to act tion about uncertainty have been accompanied by on them (e.g., Stern and Easterling 1999, Chap- recommendations that users be educated on how ter 4; Blench 1999, World Bank 2008, Chapter to interpret and make use of this information (e.g., 2). What is of importance is the perceived accu- American Meteorological Society 2011). This is a racy of forecasts. A forecast’s perceived accuracy significant challenge given the large number of 10 users who make use of forecasts and the diversi- provide suitable shelter for this number of people ty of their decision contexts, and it is a challenge in (Meyer 2006). both developing and developed countries. In a unique study of flood risk management in Col- Costs of Forecast Inaccuracy. The inaccuracy of orado, Morss et al. (2005) note that probabilistic forecasts can result in costs being incurred that forecasts may complicate the already difficult job would not be incurred in the absence of the fore- faced by practitioners because they do not have cast. For example, a forecast of a hurricane can the time or resources for the more complex anal- prompt evacuation of communities. If the forecast yses necessitated by such forecasts. In addition, proves to be incorrect, the costs of evacuation can- Morss et al. (2008) find that local flood plain man- not be recouped (they are sunk). For the U.S., a agers “must also balance the needs and desires of back-of-the-envelope estimate of the average an- multiple constituencies, including local elected of- nual cost of false alarm evacuations is USD 1 bil- ficials, private businesses, and the diverse popula- lion. More accurate forecasts offer the potential for tions at risk, and respond to these constituencies’ reducing the costs of such false alarms (Regnier decisions and demands”. 2008). The above discussion makes clear that what hap- Ability of Users to Respond to Forecasts. In pens at the bottom of the chain from forecast gener- terms of the benefits that forecasts generate, no ation to outcomes in Figure 1 is critical to econom- less important than communicating the forecasts ic benefits being realized from forecasts. Providing effectively is ensuring that users are aware of how forecasts does not by itself ensure that net benefits to make use of the forecasts and respond to them. will be realized from them, no matter how detailed, In the case of flood forecasts for Bangladesh, Web- accurate and timely the forecasts may be. Similarly, ster (2012) observes that to promote the effective an understanding of what happens at the bottom of use of the forecasts, village and community lead- the chain is critical to accurately estimating the eco- ers were trained to interpret forecasts and to rec- nomic value of forecasts. Annex 1 presents a simple ommend appropriate actions for communities, economic example that illustrates the importance of such as harvesting crops, sheltering animals, stor- a number of the factors described above that influ- ing clean water, and securing belongings. ence the economic value of forecasts. In addition to knowing how to make use of fore- Maximizing Forecast Value. Figure 1 suggests a casts and respond appropriately to them, the abili- uni-directional flow from hydromet systems to out- ty of users to actually do so will determine the val- comes and benefits. However, in recent years ana- ue that they derive from the forecasts. Users often lysts have noted that a feedback loop from outcomes face economic, social, and political constraints that to hydromet systems is desirable. Better information limit their ability to take appropriate action. This is about how users make use of different forecasts and clearly true in developing countries, but it also true about the magnitudes of benefits derived can help in developed countries. Consider the case of Hur- guide decisions on the types of hydromet services to ricane Katrina. Despite repeated and early warn- provide or enhance, with an emphasis on high-value ings of the hurricane, approximately 100,000 res- or “high-impact” weather forecasts in order to max- idents of New Orleans did not have the means to imize the net benefits generated by hydromet sys- leave the city. Moreover, the city was unprepared to tems (Lazo et al. 2008 and Morss et al. 2008). 11 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems 5. Approaches to Estimating Benefits of Routine Climate for Specific Users/Sectors A survey of the literature indicates that there have studies attempt to replicate real-world deci- been several dozen efforts to empirically estimate sion-making contexts more faithfully. In contrast, the value of forecasts to specific users or to specif- stated preference methods rely on individuals’ re- ic industries/sectors. These studies typically esti- sponses to hypothetical survey questions to infer mate the value of current forecasts or the increase the value users place on forecasts; there is no at- in value from improving one or more dimensions tempt to simulate a user’s decision-making pro- of current forecast quality, such as accuracy of the cess. A brief description of each approach is pro- forecast, its frequency, its lead time, or the medi- vided below, along with examples of applications of um through which it is conveyed. Surveys of these each, where appropriate. studies can be found in Nichols (1996), Anaman et al. (1998), Stern and Easterling (1999, Chapter Prototype Decision-Making Models. As their 5), Houston et al. (2004), Katz and Murphy (2005), name suggests, these models are highly simplified Weiher et al. (2005), Teisberg and Weiher (2009), versions of real-world situations that capture only Rogers and Tsirkunov (2010), Mills (2010), and a few critical features of the situation being mod- Katz and Lazo (2011).2 Not surprisingly, the stud- eled. The simplification is motivated by the desire ies have focused on users in the most climate-sen- to obtain analytical results about the relationship sitive sectors of the economy, in particular agricul- between the value of a forecast and characteristics ture, but they encompass a wide range of sectors. of the forecast or the situation being modeled. The A preponderance of the studies are of the value of approach is prescriptive, or normative, in nature, forecasts in the United States. in that it prescribes how a rational decision-mak- er that wants to maximize its welfare should make A variety of analytical approaches have been use of a forecast.3 The objective is to obtain gener- used in these studies. Following Katz and Murphy al insights that ideally extend to more complex re- (2005) and Katz and Lazo (2011), the approaches al-world situations. can be divided into four categories: Prescriptive Decision-Making Studies. These • prototype decision-making models; studies share the prescriptive, or normative, char- • prescriptive (or normative) decision-making acter of prototype decision-making models, but in- studies; corporate much more detail about the specific, re- • descriptive decision-making studies; and al-world decision context being modeled. Unlike • stated preference methods. prototype decision-making models, the objective is to obtain numerical estimates of the value of fore- The first three approaches rely on simulating a us- casts (or improved forecasts) in the situation being er’s decision-making process and the role that fore- modeled rather than general analytical insights. casts play in it. Prototype decision-making models are abstract and theoretical in nature, whereas For example, Mjelde et al. (1988) study corn pro- the prescriptive and descriptive decision-making duction in Illinois and assess how characteristics 2  Details on some, more recent studies can be found online at http://www.isse.ucar.edu/staff/katz/esig.html#cases. 3  More formally, the models employ a Bayesian decision-theoretic approach and assume that the user maximizes expected utility. 12 of the agricultural producer’s environment as well assumed quality of the forecast. The authors note as characteristics of the forecast, namely accura- that these numbers may be underestimated be- cy and lead time, influence the value of a forecast. cause they are derived from the assumption that The producer decisions modeled include the timing producers alter only their cropping decisions in re- of nitrogen fertilizer application, selection of the hy- sponse to the forecasts. In reality, producers could brid variety to plant, planting density, and the tim- also take advantage of the forecasts by altering ing of both planting and harvesting. Seasonal fore- their input use and harvesting decisions. As is true casts are examined with variations in the lead time for all prescriptive studies, there is also a possibil- of the forecast and its accuracy. They find that fore- ity that forecast value is overestimated because casts are of value only if they alter the producer’s the analysis assumes that producers make optimal decisions. The estimated value of a forecast de- use of the forecasts. pends on its accuracy and lead time, as well as on the price of corn; estimates obtained range from Descriptive Decision-Making Studies. Like pre- tens of cents per acre per year to nearly USD 30 scriptive decision-making studies, descriptive deci- per acre per year. Interestingly, they also find that sion-making studies incorporate much more detail a less accurate forecast received earlier can have about the situation being modeled than prototype greater value than a more accurate forecast re- decision-making models. However, unlike prescrip- ceived later. tive decision-making studies, they attempt to mod- el how users actually behave and make use of As is true for nearly all studies of the value of fore- forecasts instead of assuming ideal, optimizing be- casts, Mjelde et al. use a “partial-equilibrium” havior. Descriptive studies can range from simple framework. Specifically, they assume that chang- surveys of forecast users, to in-depth interviews of es in yields induced by better forecasts do not af- users, to monitoring of users in the course of mak- fect the market price of corn. But improved weath- ing use of actual or hypothetical forecasts. These er forecasts are likely to benefit a large portion of a studies often do not develop numerical estimates country’s agricultural producers and are therefore of forecast value. Instead, they are used to comple- likely to result in shifts in the supply of agricultur- ment and inform prescriptive studies. al commodities, leading to changes in market pric- es. Solow et al. (1998), building on earlier work by For example, Stewart et al. (1984) conduct a de- Adams et al. (1995), take these “price effects” into scriptive study of how fruit growers in Washington account when estimating the value of the El Niño state use frost forecasts to make decisions about Southern Oscillation (ENSO) forecasts for U.S. agri- protecting their orchards. The study follows up on culture in general. The study links together scien- an earlier prescriptive study (Katz et al. 1982) of tific and economic models that trace the effect of the same set of users. The prescriptive study as- ENSO forecasts on crop yields, which in turn affect sumed that growers only used the frost forecasts producers’ cropping decisions, the supply of crops, in their decision-making. In contrast, the descrip- and thus market prices. The value of the forecast is tive study revealed that growers also made use of calculated in terms of the change in economic wel- dew point and temperature observations obtained fare as measured by the change in the sum of pro- after the frost forecasts were issued. Growers used ducers’ and consumers’ surpluses. The estimated this additional information to make a series of de- value of ENSO forecasts range from USD 240 mil- cisions on whether to protect their orchards over lion to USD 323 million per year, depending on the the course of a night, as opposed to the single, 13 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems irreversible decision early in the night that was as- Some economists have expressed skepticism of sumed in the prescriptive model. Stewart et al. the value estimates derived from stated prefer- modify the prescriptive analysis to obtain more re- ence surveys, arguing that they are likely to be bi- alistic estimates of the value of frost forecasts. The ased upwards (e.g., see Diamond and Hausman modifications resulted in a small drop of USD 45 1994). However, this criticism is based on their use per acre in the estimated value of frost forecasts in valuing environmental goods and services. In the from the original estimate of USD 808 per acre. case of valuing forecasts, the direction of bias is more ambiguous. As Smith (2008) points out, stat- Stated Preference Methods. These methods rely ed preference surveys may understate benefits on surveys of samples of users to directly or indi- if respondents are not fully aware of how they di- rectly obtain estimates of the value users place rectly and indirectly benefit from forecasts. For ex- on forecasts (or improvements to forecasts). Stat- ample, he questions whether individuals are ful- ed preference methods can take the form of stat- ly aware of how forecasts improve winter weather ed value or stated choice surveys. In the case of road maintenance. stated value surveys, users are directly asked what they are willing to pay for a forecast or an improve- Policy Relevance of Estimates. The studies de- ment to a forecast. In the case of stated choice sur- scribed above provide some foundation for under- veys, users are asked questions that require them standing the determinants of forecast benefits and to make choices among alternative forecasts with the magnitudes of these benefits. However, the different costs associated with each. Users’ prefer- studies are fragmented and vary considerably in ences indirectly reveal the value they place on the their coverage, sophistication, and realism. Millner forecast characteristics. (2009) cautions: “Traditional models that have at- tempted to gauge forecast value have focused on For example, Lazo and Chestnut (2002) used both a best-case scenario, in which forecast users are stated value and stated preference methods to assumed to be statistically sophisticated, hyper- estimate the value that U.S. households place on rational decision-makers with perfect knowledge (then) current forecasts, as well as the value they and understanding of forecast performance. These place on improvements to the forecasts. A total of models provide a normative benchmark for assess- 391 individuals in 9 cities across the U.S. were sur- ing forecast value, but say nothing about the val- veyed. The survey results indicated that on aver- ue that actual forecast users realize. Real forecast age each household was willing to pay USD 109 users are subject to a variety of behavioral effects per year for current forecasts. The average value and informational constraints that violate the as- placed by households on improving multiple di- sumptions of normative models.” Millner’s cau- mensions of current forecasts, specifically, their tion regarding the likely upward bias of existing es- frequency, accuracy, and geographic detail, to their timates based on models of user decision-making maximum possible level was estimated to be USD is a valid one, however, from a policy perspective, 16 per year. In a more recent study, Lazo et al. what is of critical importance is the magnitude of (2009) rely on stated value surveys of a much larg- the bias. Small biases are not of great concern, but er number of individuals (1,520) and estimate the large biases are. Unfortunately, the current state of median value placed by U.S. households on current knowledge does not allow for any general assess- weather forecasts to be USD 240 a year. ment of the magnitude of biases. 14 Thus, existing studies are in nearly all cases partial to account for price effects, as in the work of Adams equilibrium analyses that ignore price effects. This et al. (1995) and Solow et al. (1998), are notewor- can be a significant shortcoming if the estimates de- thy, but the gain in the economic realism of these rived from the studies are used to value the bene- studies comes at the cost of less attention paid to fits of investments in hydromet systems at the coun- modeling the effect of weather on yields and the try level. As Katz and Lazo (2011) observe, efforts manner in which producers make use of forecasts. 15 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems 6. Approaches to Estimating Country-Level Net Benefits of Hydromet Investments A number of alternative approaches have been of improved forecast information depends on the used to estimate the net benefits of hydromet in- structure by which information is disseminated and vestments or modernization at the national or used locally. Compared to formal benefit-cost anal- country level. These were developed to estimate yses of a specific sector discussed earlier, rapid ex- the benefits of hydromet investments on a large re- pert assessments take the place of model-based gional scale, in addition to being applicable where or survey-based estimation of benefits. systematic recording of damages and losses has not been undertaken for sectors and populations Referring to Table 2, it is worth noting that the three within a country. The various approaches are com- approaches are closely related. Sectoral analysis plementary, and often the recommendation is that sometimes makes use of Gross Domestic Product more than one method be used in order to corrob- (GDP) to define an initial damage reduction from orate results or obtain a range of net benefits from hydromet investment. The conditional probability hydromet investments (GFDRR 2009). approach is very similar to sectoral analysis with the added improvement of specifying a prior distri- This section discusses three such approaches, bution of weather-related damages, and then ad- the benchmarking method, the sectoral approach, justing this prior distribution (i.e., using condition- and the conditional probabilities approach.4 These al probabilities) to reflect the effects of hydromet three approaches are similar in that they attempt modernization using local information. Changes to quantify the benefits from hydromet investments in expected damages then depend on both an as- in terms of the changes in the economic well-being sessment of the reductions in losses resulting from of climate-sensitive sectors in a country. hydromet modernization, and changes in the prob- ability of losses. Thus, conditional probability ap- A general comparison of the three methods is pro- proaches allow for some uncertainty in terms of the vided in Table 2. All three approaches have a com- effects of weather when arriving at a benefit esti- mon sequence of analysis. An initial monetary loss mate. Benchmarking, on the other hand, essential- with and without investment is estimated for each ly assumes a deterministic effect of hydromet mod- industry or sector in the economy, adjustments are ernization on weather hazards through the use of made to these using expert opinions, and a final GDP-based estimates. benefit estimate is computed across all sectors in the economy. A key similarity among the methods is Benchmarking Method. The benchmarking meth- that they arrive at a country-level benefit estimate od has been the most widely applied approach, es- that has been refined using as much local infor- pecially in the Southern Caucasus (Azerbaijan, Ar- mation and expertise as possible. The refinement menia, Georgia), the Balkans (Albania and Serbia), stage is critical given that the ultimate success and other parts of the former Soviet Union, namely, 4 There are also sectoral-like approaches that rely on conjoint analysis, although these have been applied infrequently to estimate country-lev- el benefits of hydromet modernization (an example is Leviäkangas et al. 2007). Conjoint analysis uses the survey method common in contingent valuation but also an additional survey where users of forecast information and experts in selected sectors are asked to rank alternatives or attri- butes associated with different hydromet investment scenarios and outcomes. The results of these expert assessments are then used to adjust benefits estimated by contingent valuation, and the results are used to extend the contingent valuation estimates of willingness to pay for users from a specific sector to other sectors. Conjoint analysis is therefore a combination of the contingent valuation approach with further sociological surveys aimed to refine and map the use of better forecast and climate information for the country in question. 16 Table 2  Comparison of Country-Level Approaches Losses (with and without Net Benefit Estimate of Method modernization) Adjustment Hydromet Modernization Benchmarking Based on GDP Expert opinion used to adjust losses Increase in GDP Sectoral Based on benefits transfer* Expert opinion plus surveys of Decrease in estimated affected agents used to adjust losses monetary losses Conditional Based on benefits transfer* Expert opinion used to adjust losses Decrease in expected Probabilities and assign conditional probabilities monetary losses** Source: Authors. *Benefits estimated in other studies in regions judged similar to the one in question. **Decrease in expected monetary losses = [probability of losses without modernization x losses without modernization] – [conditional probability of loss with modernization x losses with modernization]. Belarus, Ukraine, and Kazakhstan (GFDRR). It has Benchmarking is undertaken in a two-stage pro- also been applied in other parts of Europe and in cess. Referring to Table 3, the two most important Asia (World Bank 2008). A similarity among these components it estimates are the level of annual di- countries is a lack of data on actual total and sec- rect economic losses (as a share of GDP) caused tor-specific economic losses from hydro-meteoro- by hydro-meteorological hazards (L), and the level logical hazards. of annual preventable losses that can be achieved through improved forecast information or warn- The benchmarking method was developed to es- ings as a result of hydromet modernization (P). Pre- timate country-level economic benefits from hy- ventable losses are an agreed-upon percentage dro-meteorological modernization. It is based on of the total direct economic losses with the exist- national official macroeconomic and sector-specific ing hydromet system. The first stage of benchmark- statistics, namely GDP, assessments of the weath- ing begins with a given set of initial values (bench- er dependence of the economy, an assessment of marks) for these components. In the second stage how vulnerable the country is to weather events these benchmarks are adjusted following assess- based on meteorological statistics for temperature ment by experts. The second stage is what distin- (minimum and maximum), relative humidity, pre- guishes one country from another, in that experts fa- cipitation and wind, the current status of hydrom- miliar with the local situation refine the initial values et equipment and facilities, and the quality of hy- using specific country parameters such as weather dromet service provision to citizens in the country. and climate, economy diversification and structure, In many cases, key parameters are derived through and status of the current hydromet system. expert opinion or studies undertaken in other coun- tries. The main advantage of the benchmarking ap- As Table 3 shows, the first stage of benchmarking proach is its simplicity and the fact that it does not involves rules for initial values that are common re- require time-consuming surveys or complicated gardless of the country being studied. Initial values quantitative analysis, yet it is able to inform policy are set for: 1) average annual losses from danger- makers concerning the value of preventable losses ous weather events (ranging from 0.1–5% of GDP); in a country from weather events and the extent to 2) average annual preventable weather losses, which these losses can be reduced with hydromet measured as 40% of the losses that would have investments. occurred without hydromet investments computed 17 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems in the first component; 3) an estimate of how cli- adjustments are made to the numbers in the sec- mate sensitive the economy under consideration ond column, the annual incremental benefit from is based on the share of sectors in the economy hydromet modernization is estimated to equal: P x that are judged to depend on weather (such as L. Then, information in the last three rows is used agriculture, transportation, electricity generation, to further adjust this incremental benefit estimate and tourism)—the initial level is set at 50% of GDP; upward or downward. 4) the share of GDP due to agriculture (set initial- ly at 15%); meteorological vulnerability (assumed Advantages and Limitations of Benchmarking. to be average initially); and the status of hydrom- The main advantage of benchmarking is that all hy- et service provision. The last component is a mea- dromet modernization benefits are based on read- sure of how effectively hydromet investments can ily available, generally reliable, estimates of GDP. be integrated into the country’s forecasting infor- Benchmarking can be accomplished quickly and mation services and is initially assumed to be “sat- for many countries in a region at once, and does isfactory”. This therefore accommodates nuances not require time consuming study of specific sec- in infrastructure and educational capital necessary tors except to understand the extent to which they to implement improved forecasts and disseminate contribute to GDP. The use of local experts is also information. In the second stage, expert opinion is appealing as a means of tailoring estimates of ben- used to adjust the initial entries in the second col- efits from hydromet investments. umn of Table 3 for local circumstances. If, for exam- ple, there are barriers to disseminating information, Given its inherent simplicity, benchmarking is not as would be the case if infrastructure or education without limitations. The main problem is com- in the country is below average, then the value in mon to all currently used country-level approach- the second row would be adjusted downward. In es: benefits that have been assessed in the histori- developing economies that are largely agrarian, the cal literature can be valued, but this is problematic value in the fourth row is adjusted upward. because the literature may not focus on losses spe- cific to local situations. Second, in many countries The annual benefit from hydromet modernization GDP may not be reliably reported or computed— is computed very simply from Table 3. After all this has in fact prevented benchmarking from be- ing applied in Turkmenistan (see below). Third, the Table 3  Benchmarking – First Stage Parameter benchmarking method is too aggregate to evaluate Values specific benefits of information that may be import- ant to hydromet targeting or location. For example, Initial the safety benefits to households or the nuances Parameter Estimate concerning infrastructure that are important fac- Average annual loss from weather haz- 0.1–5% of ards (L) GDP tors in the benefits of hydromet investments may Average annual preventable loss from 40% of differ considerably in their contribution to GDP es- weather hazards (P) annual loss timates, and yet the primary way to evaluate these Weather dependence of economy (W) 50% of GDP is with expert opinion. Share of GDP due to agriculture (A) 15% of GDP Effectiveness/ability of country to make Satisfactory Sectoral Approach. The sectoral approach aims use of hydromet service modernization to assess the economic desirability of hydromet Source: World Bank (2008). investments by comparing the costs of hydromet 18 investments to the resulting losses prevented by the best and worst case outcomes, with the mean the investments. As is true for the benchmarking value used as the base case. Second, experts as- method, the benefits are computed across all cli- sess the direct and indirect costs of hydromet mod- mate-sensitive sectors of the economy that gener- ernization with an eye toward how modernization ate a significant share of GDP. should be targeted (this determines Δ). In so far as possible, actual statistics are used to back the The sectoral approach relies on estimation of two range of values picked by experts, but this is not al- key parameters for each sector. The first is Ri, the ways possible. percentage of losses that are potentially preventable in sector i. The second is Si, the percentage of po- Advantages and Limitations of Sectoral Ap- tentially preventable losses in sector i that are avoid- proach. Nearly all of the caveats for benchmarking ed with the hydromet improvement. The product apply to sectoral approaches. However, the sectoral of these two, Ri x Si, gives the percentage of loss- approach does have the advantage of tailoring loss es that are avoided in sector i with the hydromet im- information to specific sectors, rather than simply provement. Applying these percentages to estimates linking all losses to changes in GDP as benchmark- of losses for each sector yields estimates of the re- ing does. Sectoral approaches also benefit from duction in losses from the improvement. The sum of the potential to consider the validity of statistics these reductions across sectors equals the benefits concerning weather effects using expert judgment. of the improvement. The benefits are compared to In principle, this approach should provide a more an estimate of the increase in costs, Δ, associated “local” refinement of benefits and costs of hydrom- with the improvement. All of these parameters, Ri, Si et modernization. However, again the quality of ex- and Δ, are determined through surveys of experts. pert opinion and availability of outside information that can be transferred to the country in question Expert assessment and surveys are the most time remains a major obstacle to the sectoral approach. consuming part of sectoral approaches.5 These surveys must assess both general information con- Comparisons of Benchmarking and Sectoral Ap- cerning the scale and quality of the hydromet in- proaches. A recent study compares the results of formation used, and the two key parameters need- using the benchmarking and sectoral approaches ed to estimate benefits. Thus, these assessments in the Southern Caucasus (GFDRR 2009). A sum- usually proceed in two stages. First, experts are mary of this comparison is found in Table 4. In polled to obtain estimates of the two key parame- the table, benchmarking estimates appear in pa- ters for each sector of the economy that is consid- rentheses, while the other numbers represent es- ered to be climate sensitive (Ri and Si). If a “scenar- timates using the sectoral approach. Notice that io” analysis is deemed suitable, then these experts average annual preventable losses (assessed by also recommend a range of values to represent experts) are common across the approaches. 5 Hallegate (2012) proposes a specific approach, risk management analysis, for loss transfer and modification. His work details a procedure to use losses from European countries in terms of saved lives and reduced economic losses from weather data, and then transfer this information to estimate losses in developing countries. He is therefore able to provide a layer of additional sophistication and data quality to estimating ben- efits of hydromet modernization in developing countries that lack sufficient data for analysis. His procedure is based on asset losses, and he also details several necessary modifications to making sectoral analysis more rigorous in these contexts. Finally, he argues that benefit estimation must be sensitive to five needs in developing country contexts: local observation systems, forecasting capacity, interpretation capacity, commu- nication tools, and user decision-making capabilities. Finally, this approach compares only annual benefits and costs and therefore does not de- pend on a discount rate. 19 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems Comparison of Results of Assessment of Hydromet Modernization Using Benchmarking Table 4  and Sectoral Approaches* Kyrgyz Republic of Estimated Parameter Republic Tajikistan Average annual losses from weather hazards (USD million) (24.9) 39.6 (24.9) 37.0 Average annual losses (percent of GDP) (1.0) 1.5 (1.04) 1.6 Average annual preventable losses (USD million) 10.1 5.8 Expected average annual benefits due to hydromet modernization (2.9) 3.8 (1.7) 3.1 (USD million) Investment efficiency (%)** (244) 318 (199) 357 Source: GFDRR (2009). *Benchmarking estimates are in parentheses. Turkmenistan was not judged to have reliable GDP data and was not included in the comparison. **This is an imputed internal rate of return that based on the stream of benefits and costs over time. These results show that benchmarking yields low- events. For each sector, a distribution of the likeli- er estimates of the net benefits from hydromet in- hood of occurrence of losses and their magnitude vestments. For example, in the Kyrgyz Republic, in the absence of hydromet investment is first esti- the benefits of hydromet investments range from a mated. This baseline distribution is based on avail- benchmarking estimate of USD 2.9 million per year able data on the magnitude and frequency of loss- to a sectoral approach estimate of USD 3.8 million es, as well as expert opinion. In a second step, the per year. In this study, it is worth noting that the sec- changes to this distribution resulting from improve- tors studied (water management, agriculture, and ments to the hydromet system are estimated using electricity) are responsible for about 75% of GDP available information and filling in with expert opin- and are all highly climate sensitive. In Tajikistan, ion where necessary. The benefits of hydromet im- once again benchmarking yields lower estimates provement are then calculated based on the result- than the sectoral approach. The estimates range ing change in the distribution of losses. from USD 1.7 to 3.1 million per year in benefits for benchmarking and sectoral analyses, respectively. The data required for the conditional probabilities As a result, the computed investment efficiencies approach are more extensive than for the bench- in the last row of the table are considerably differ- marking approach, but less than for formal bene- ent even though the expenditure on hydromet mod- fit-cost analyses of individual climate-sensitive sec- ernization was equal across approaches for the tors. It relies on historical data on losses suffered same country. While these two approaches yield a from extreme weather events in climate-sensitive relatively wide range of values for hydromet bene- sectors in the economy, perceptions from the com- fits, both give a consistent, bottom-line economic munity and experts concerning social, econom- efficiency assessment. ic, and governance aspects of weather events and information from improved forecasts, identifica- Conditional Probabilities Approach. This ap- tion of hydromet modernization impacts, and rec- proach estimates the relationship between the state ommendations from experts concerning preventa- of the hydromet system and the frequency distribu- tive actions arising from dissemination of improved tion of losses experienced due to meteorological forecasts. 20 The conditional probabilities approach was re- when incremental benefits and incremental costs cently used to evaluate modernization of Mexi- can be reliably obtained or inferred, the magnitude co’s hydromet system (World Bank 2012, Project of expected total loss reduction is highly depen- Appraisal Document). Estimates of the frequen- dent on the adjustment in probabilities. Further, cy distribution of baseline losses were based on the underlying incremental costs and benefits at- data on insurance payments for property damage, tributable to improved forecasts are not assumed loss of livestock and agricultural output, as well to change with hydromet investments, and so like as government expenditures on rebuilding and re- the benchmarking and sectoral approaches, the pairing infrastructure, and government assistance conditional probabilities approach does not specif- to low-income households affected by extreme ically incorporate information on changes in deci- events. The change in the distribution of losses due sions that agents make. Thus, while there is a de- to hydromet modernization was largely derived us- gree of refinement beyond both the sectoral and ing expert opinion. benchmarking approaches, this refinement is still based on expert opinion and surveys rather than Advantages and Limitations of Conditional Prob- actual information from decision-makers who use abilities Approach. The conditional probabili- forecast information. ties approach has an advantage over the bench- marking and sectoral approaches in that not only Limitation of Country-Level Approaches. All of preventable losses are assessed, but a probabili- the country-level approaches to estimating the ben- ty distribution is applied over potential reductions efits of hydromet investments focus largely on loss- in these losses from modernization. Moreover, the es avoided from extreme meteorological events. As adjustment to probabilities from prior to condition- such, these approaches generally need to be sup- al distributions takes explicit account, in principle, plemented with sector-level estimates of the ben- of changes in decision-making, barriers to dissem- efits yielded by improved forecasts of routine cli- ination of information, and opportunities to use mate. The approaches for deriving these estimates forecast information in the country in question. were described in Section 5 above. It is important, however, to ensure that benefits of hydromet invest- This said, the limitations of the conditional prob- ments are not double counted: if benefits from im- abilities approach are similar to the limitations of provements in forecasts of routine climate are cap- the other approaches, in large part because it too tured for a sector in the country-level analysis, then relies heavily on expert opinion to estimate the adding an estimate of these benefits from a sec- benefits of improved forecasts. One important dif- tor-level analysis would constitute double counting ference, however, is the added complexity of having of benefits. However, the two estimates could be to rely on expert opinion for both an assessment of compared, and if substantially different, they could preventable losses, as well as an assessment for be used as alternative values of the benefit in a how probability distributions change. Indeed, even sensitivity analysis. 21 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems 7. Costs of Hydromet Investments The costs of hydromet modernization are multidi- hydromet costs may not be foreseen with certain- mensional and include more than simply install- ty at the time of modernization. This uncertainty is ing new radar stations or upgrading aging equip- nearly always the rule given that the average ser- ment and software so that better forecasts can be vice lifetime of newly purchased radar and model- made. The total costs encompass a variety of other ing software ranges between 10–15 years. investments and activities necessary for success- fully preparing and delivering forecasts. These in- Classification of Economic Costs. Broadly de- clude implementation and service associated with fined, and regardless of whether installation or im- hydromet upgrades, additional weather service plementation of hydromet systems is considered, and information delivery infrastructure, and invest- there are two types of costs associated with hy- ments in training and hiring professionals to inter- dromet investments: fixed costs and variable costs. pret data and generate warnings and forecasts. Fixed costs of modernization are those that do not The costs of implementation and service activi- depend on the amount of information generated by ties are sometimes the largest component of total the forecasting agency but are necessary for the costs, and they are often paid over an extended pe- operation to exist. Examples are costs associated riod of time, while the bulk of equipment installa- with purchases of new equipment and software, tion costs are paid up front in the project life cycle. and the building of new roads or communication In the literature, implementation and delivery costs networks and towers to disseminate information. are sometimes referred to as capacity or institution Variable costs are those that are a function of the building costs. extent of forecasts produced and the speed of in- formation delivery using the new hydromet system, Any country-wide hydromet system includes many that is, they are a function of the volume and type components, such as automatic monitoring sta- of information generated with the new system. Ex- tions, telecommunication networks, and comput- amples of variable costs are the number of addi- ers to store software and data. The requirement of tional forecasters and other staff required to deliv- “nowcasting”, or real time monitoring and warning er new and improved weather information from an of extreme events, means that the establishment expanded or upgraded hydromet network, and new of telecommunication networks and their subse- monitoring and warning activities that are required quent management are often a large part of the given additional information generated by an ex- total cost of any modernization over the hydromet panded or upgraded hydromet network. project life cycle. Kokko and Vaisala (2005) argue that ongoing maintenance, calibration, and period- Relative Magnitudes of Costs. For complex and ic update/upgrade of hydromet systems represent multi-year investments such as hydromet modern- a significant cost that is sometimes overlooked. ization, the relative magnitude of fixed and vari- Network and maintenance costs can be thought of able costs can vary widely. In some cases, the as implementation costs. Further, a part of the life fixed costs are the larger component of total costs, cycle costs of hydromet systems are “contingency” but this is not the rule. For example, Subbiah et based, that is, they are not known with certainty al. (2008) estimate for hydromet modernization at the time hydromet systems are installed or up- in Bangladesh that up-front fixed costs equaled graded. Contingency costs imply that some future USD 2.1 million, while variable costs amounted to 22 over USD 400,000 annually over a project life cy- Cooperation and Economies of Scale. The mag- cle of more than 10 years. The largest components nitude of costs associated with upgrading a coun- of variable costs (50%) in this study were capaci- try’s hydromet systems can also depend on the ty building of national and sub-national institutions presence or absence of cooperation with neighbor- for translation, interpretation, and communication ing countries. United Nations International Strate- of probabilistic forecast information. gy for Disaster Reduction (UNISDR 2008) reports that in a seven country region of southeastern Eu- The largest fixed cost components are typically ear- rope, financing to modernize national hydromet ly warning system technology purchase and instal- services amounts to 90 million EUR without region- lation, and high performance computing system al cooperation and 63 million EUR with coopera- investment. Tsirkunov (2011) reports that in the tion, with those costs attributable to measurement, Republic of Tajikstan, improvement in hydromet upper-air sounding, radars, communication, and observation networks, an activity with a large fixed delivery of forecast warnings. Cooperation reduces cost component, account for about USD 8.9 million costs through data sharing across countries, which of the estimated USD 13 million required for mod- decreases the required extent of hydromet equip- ernization in that country. A similar pattern is found ment installation, or through sharing of weather in the Kyrgyz Republic, with upgrading of the hy- service delivery networks. dromet observation networks estimated to account for USD 3.8 million of the total USD 6 million re- The differences between cooperation and no-co- quired for modernization. operation scenarios suggest there may be econo- mies of scale in modernization for developing coun- In mountainous regions, i.e., regions with very var- tries. Economies of scale means that the costs of ied topography, a denser network of monitoring a larger integrated system can be lower per unit of stations and radar are typically needed. This can forecast information produced (however, this is de- result in fixed costs accounting for a larger frac- fined) than smaller projects that still require a large tion of total costs than in regions with less varied upfront investment in equipment. topography. Cost Estimation Caveats. There are four import- Hallegatte (2012) observes that developing country ant caveats with regard to hydromet cost estima- hydromet modernization projects can differ mark- tion relevant to evaluating hydromet investments in edly from those in developed country projects, with developing countries: many of the differences reflected in costs. He ar- gues that, in general, the total costs and experienc- • counting only incremental costs; es from developed countries should not simply be • treatment of joint costs and benefits; applied to determine modernization costs in devel- • cost uncertainty in long-lived projects; and oping countries. He notes that developing country • potential for cost savings. hydromet systems are characterized by poor con- ditions with regard to staffing, buildings, observa- Each of these caveats is discussed in turn below. tion sites, and instruments. The costs of upgrading such systems can be many times higher than mod- Counting only incremental costs. When evaluating ernization in a country that already has high quali- an investment, the relevant costs are only those ty infrastructure. that are attributable to the investment. These 23 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems incremental costs are, in principle, straightforward implies greater uncertainty about the magnitude of to estimate, as they are attributable to activities project costs. This is especially true of long-lived associated with modernization, such as purchase projects that include contingency-based costs. Es- and installation of additional equipment, networks, timating costs with precision is further complicated software, and hiring of additional staff. if there is little historical data on the types of costs associated with a project. Relying on cost transfer Joint costs and benefits. A second, more challeng- from modernization efforts in developed countries, ing issue has to do with the joint nature of some where such data are generally available, adds an of the fixed and variable costs associated with hy- additional component of uncertainty to estimates dromet improvements. These joint costs are costs of costs in developing countries. that are incurred in the course of hydromet improve- ments but they also impact other non-weather re- Potential cost savings. An upgraded hydromet sys- lated activities that have value in the economy.6 tem can reduce some of the costs of operating a Examples of joint costs are those associated with hydromet system. For example, new radar equip- improvements in infrastructure, including roads (to ment can eliminate the often high costs of main- monitoring stations) and communication networks taining old, outdated systems for which spare parts (for collecting and disseminating information from must be custom made. Further, in modern hydrom- monitoring stations as well as forecasts). In stud- et systems many monitoring and information re- ies of hydromet modernization, joint costs often porting functions are automated (Kokko and Vais- fall under the category of capacity improvements ala (2005)), reducing the need for staff resources deemed necessary for improved forecast informa- to be devoted to collecting information from moni- tion to be utilized by the population. Such improve- toring stations. ments can yield benefits that extend far beyond those related to hydromet systems. Table 5 presents a breakdown of the types of spe- cific costs that may be associated with hydromet Dealing with joint costs and benefits is challeng- modernization listed by the activity required by the ing. In principle, if all joint costs are attributed to investment. These activities are drawn from a sur- the hydromet investment, then all the joint bene- vey of selected recent hydromet projects, includ- fits should be as well. However, quantifying all the ing those described in Tsirkunov (2011), Subbi- joint benefits is likely to be difficult and beyond the ah et al. (2008), GFDRR (2009), and World Bank scope of the hydromet investment analysis. In such (2012). While all of these activities are not always cases, a fraction of the joint costs could be attribut- required in modernization efforts, depending on ed to the hydromet investment, with the fraction the current system’s capabilities and the type of based on a rough estimate of the fraction of joint improvements undertaken, they are illustrative of benefits that are hydromet related. the types of costs that may need to be estimated. The table indicates the likely nature of the costs Cost uncertainty. The common need to upgrade or in each category. Some of the categories encom- put in place associated infrastructure when improv- pass many sub-activities. For example, strengthen- ing hydromet systems in developing countries often ing institutional capacity and communication and 6 Another type of jointness in costs refers to the fact that one activity required of hydromet modernization may be synergistic to another activity, so that paying for one may reduce the cost of another. 24 information networks include activities such as ed- the degree of automation and human calibration ucational investments related to gathering and in- needed over time. Infrastructure improvements will terpreting climatological information, and develop- affect the costs associated with building communi- ing means for communications with users that fit cation networks to deliver forecast warnings. The the country’s political and geographic features. Ca- precise nature of these synergies, and therefore pacity building can include development of institu- the costs associated with all of them, is also criti- tional, legal, and regulatory authority within a coun- cally dependent on a country’s social, political, and try. This has been true, for example, in the case of economic state. This highlights the recommenda- hydromet modernization in the Caucasus (Tsirnu- tion by Hallegatte (2012) to not simply transfer cost kov 2009). Modernizing observation infrastructure information from developed countries to develop- and technical design/engineering services include ing countries. It is also for this reason that the scale rehabilitation and expansion of radar networks, of the project may matter critically in estimation of development of meteorological products and ser- total costs for hydromet modernization. vices, enhancement of computational capacity both in terms of modern high speed computer sys- Hydromet Modernization Cost Table 5  tems and software required to produce meteoro- Components from Selected Studies logical information, and establishment of quality control mechanisms for forecasts. Communication Category Fixed Variable Joint and information networks include means to devel- Technical Design, X X op a regional climate information capability and re- Engineering Services quire activities such as design and maintenance of Building Communication X X and Information Networks telecommunication systems, and development of (Regional Capacity) early warning and nowcasting procedures. Strengthening Institutional X X X Capacity As noted in Table 5, some costs can have both fixed Ongoing System X and variable components. For example, the level Maintenance/Calibration and amount of capacity building needed could de- Modernizing Observation X Infrastructure pend on the planned extent to which weather infor- Non-Meteorological X X X mation is generated and delivered. A last point is Infrastructural Investments that many of these costs are related, in that activi- Monitoring and Forecast X ties in one category can affect costs in another (i.e., Staffing Enhancement there are cost-related synergies across activities). Source: Tsirkunov (2011), Subbiah et al. (2008), GFDRR (2009), For example, the type of radar chosen will affect and World Bank (2012). 25 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems 8. Proposed Framework for Estimating Net Benefits of Investments in Hydromet Systems In this section we present an outline of a proposed total benefits of hydromet systems (and the fore- framework for estimating the expected net ben- casts they generate) and the incremental benefits efits of investments in hydromet systems at the resulting from improvements to (or investments country level. The framework draws on the exist- in) hydromet systems. This distinction between ing approaches described in the previous sections, total and incremental benefits is emphasized by but with some important changes and additions. Freebairn and Zillman (2002). Much of the exist- It bears emphasis that the proposed framework ing literature on the benefits of forecasts focuses yields first-cut estimates of net benefits, as is true on estimates of total benefits. A small subset of for existing approaches. The discussion in Section the literature focuses on estimating the benefits of 4 and Annex 1 makes clear that obtaining precise improvements in hydromet systems. However, this estimates of the benefits enjoyed by all users/sec- body of literature estimates the incremental ben- tors from investments in hydromet systems is very efits associated with improved forecasts resulting difficult given the current state of knowledge and from hydromet investments, instead of estimating data available. the total benefits of existing, or hypothetically per- fect, forecasts. Given the greater difficulty of identifying and quan- tifying the benefits of hydromet improvements, as The objective of the framework proposed here is to opposed to their costs, the focus of the proposed estimate the benefits of moving from the current framework is on benefits estimation. Considerable hydromet system to an improved hydromet sys- detail on the costs of hydromet improvements is tem. Therefore incremental benefits are the rele- provided in the preceding section. vant benefit measure. Total benefits would be the relevant measure if the objective of the analysis Benefits Transfer. The proposed approach makes were to estimate the benefits of having a hydrom- use of benefits transfer. As noted earlier, benefits et system, i.e., the benefits of going from a scenar- transfer entails taking benefits estimates derived io with no hydromet system to one with a hydrom- in one setting to estimate benefits in another, sim- et system. ilar setting. The technique has been widely used to value the benefits of environmental improve- Time Horizon of the Analysis. A key question ments. Benefits transfers must be performed with when evaluating any project or investment is the care, and benefits estimates derived using this pro- time horizon of the analysis. In principle, the ap- cedure should be viewed as indicative rather than propriate time horizon corresponds to the time pe- exact. For the transfer to be valid, the original set- riod over which the project is supposed to generate ting in which benefits were estimated and the new any costs or benefits. In the case of physical infra- setting to which they are being applied must be at structure investments in hydromet systems, the ap- least roughly comparable. propriate time horizon could be dictated by the life of the physical infrastructure. For example, invest- Incremental vs. Total Benefits An important ment in a Doppler weather radar system could call point when conducting benefits transfers in the for a time horizon of 15–20 years, which is likely context of valuing the benefits of investments in to be the typical useful life of the system. Howev- hydromet systems is the distinction between the er, the benefits of broader investments in hydromet 26 systems, such as improvement of early warning A first step in estimating the benefits of invest- systems, could well extend beyond the life of the ments in hydromet systems is identifying the types physical infrastructure. This is especially true if the of benefits the investments will likely generate. The improvements encompass better climate resilient existing literature and past efforts to estimate ben- planning and better decision support tools in cli- efits of hydromet investments clearly provide useful mate-sensitive sectors. Such improvements can guidance when identifying types of benefits. How- yield benefits that extend well beyond one to two ever, as noted in Section 2, these sources may not decades. The common problem with using long adequately consider benefits associated with more time horizons is the difficulty of estimating benefits recent hydromet investments intended to promote or costs with reliability more than two to three de- climate resilience and climate adaptation. cades into the future. The output of this first step is a comprehensive list- With these caveats in mind, we turn to a descrip- ing of benefits likely to be generated by the hydrom- tion of the proposed framework. The description et investments. is presented as a sequence of steps to be car- ried out. This approach borrows from that in Lazo Step 2: Screening Benefits et al. (2008). In practice, the analysis is likely to The benefits identified in Step 1 will invariably dif- be iterative in nature, rather than purely sequen- fer widely in character. The objective of the second tial, with earlier steps in the process being revisit- step is to screen benefits and classify them into ed. To make the proposed approach more tangible one of three categories: and concrete, in places the following description re- fers to its application to evaluate investments in hy- • Ignorable Benefits. Benefits that can be ig- dromet systems in Jamaica. A full description of the nored in the analysis because they are likely to Jamaica analysis is contained in Annex 5. An over- be small in magnitude; view of the framework is provided in Figure 2. • Unquantifiable Benefits. Benefits that are likely to be relatively large in magnitude but Step 1: Identifying the Full Range of Benefits cannot be quantified (given existing methods/ As discussed in Section 5, the benefits from im- knowledge); and proving a country’s hydromet system depend on a • Quantifiable Benefits. Benefits that are likely number of broad factors. Among them are: to be relatively large in magnitude and can be quantified. • the state of the existing hydromet system; • the nature and magnitude of the improvements There is likely to be some iteration between this being contemplated; step and the next step in which benefits are quan- • the country’s susceptibility to extreme meteo- tified and monetized. Determining which benefits rological events; and are quantifiable will inevitably require consider- • the structure of the country’s economy, and ation of the approaches that can be used to quan- its dependence on climate-sensitive sectors, tify benefits. namely, agriculture, aviation, construction, sur- face and water transportation, water resourc- Benefits that cannot be quantified should be omit- es, energy, fisheries, forestry, health, and tour- ted from further analysis only if they are likely to be ism and recreation. small in relative magnitude, i.e., they are ignorable. 27 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems Benefits that are likely to be large in magnitude effective dissemination of information about their should be identified and described even if they can- effects and appropriate responses to them, can not be quantified. Where possible, these benefits substantially reduce economic losses caused by should be described in some detail and qualitative the events. The magnitude of these benefits will be assessments of their magnitude (e.g., small, mod- especially large for countries that are susceptible erate, large) provided, as well as an indication of to extreme meteorological events, such as coun- the likelihood of the benefits being realized (e.g., tries in the Caribbean. unlikely, likely, very likely). Type II and III benefits, i.e., benefits from improved The output of this second step is a classification of forecasts of routine climate can result in increased the benefits identified in the first step. Subsequent enterprise profits (or reduced costs) and improved analysis is restricted to the benefits classified as decision-making by households. These bene- quantifiable. fits may be small at the individual household or Step 3: Selecting Approaches for Estimating Quantifiable Benefits Overview of Framework for Figure 2  Given a list of quantifiable benefits, the next step is Estimating Net Benefits of Hydromet to select methods to quantify the benefits. Follow- Investments Expected Net Benifits of ing the classification scheme described earlier in Hydromet Investments Section 2, quantifiable benefits are likely to be of Step 2: Screen Benefits and Classify as Ignorable, one of the following three types: Unquantifiable, and Quantifiable • benefits of improved forecasts, and associated Step 3: Select Approaches for Estimating Quantifiable Benefits early warning systems, of extreme meteorologi- cal hazards (Type I benefits); Step 4: Quantify Benefits Associated with Extreme • benefits to enterprises of improved forecasts Meteorological Events (Type I Benefits) of routine climate and dissemination of these forecasts (Type II benefits); and Step 5: Quantify Benefits to Enterprises Associated with • benefits to households of improved forecasts Forecasts of Routine Climate (Type II Benefits) of routine climate and dissemination of these forecasts (Type III benefits). Step 5: Quantify Benefits to Households Associated with Forecasts of Routine Climate (Type III Benefits) For brevity, these benefits are referred to as Type I, II, and III benefits. Type II and III benefits are similar Step 7: Address Uncertainty in that they are both associated with improved fore- casts of routine climate, however the techniques Step 8: Aggregate Benefit and Cost Estimates to used to estimate them differ, hence a distinction is Obtain Present Value of Net Benefits made between them. Step 9: Conduct Sensitivity Analyses Type I benefits, i.e., benefits from improved fore- casts of extreme meteorological events and Source: Lazo et al. (2008). 28 individual enterprise level, but when multiplied by in an economy is affected differently by meteoro- an entire population at risk they can be substantial. logical hazards, depending on the type of meteo- rological event, the sector’s exposure to weather, To estimate Type I benefits we prescribe an ap- and the value of its assets. Furthermore, the ben- proach that is similar to the sectoral approach de- efits yielded by the hydromet investment will differ scribed in Section 4, but differs in that it makes across sectors. For instance, a hotel (in the tour- use of historical data on losses, by sector, from ex- ism sector) in a developing country is likely to be a treme meteorological events to establish a base- sturdy structure that will experience limited dam- line for the analysis. We describe the approach age from a storm even with little advance warning. further below. In the absence of adequate histor- In contrast, a rural home will be much more vulner- ical data on losses, two alternative approaches able to strong winds and rain, and a small increase can be pursued: (i) use expert opinion to develop in the lead time of a storm warning could enable estimates of historical losses, or (ii) employ the the household to better fortify the house and move benchmarking method described in Section 4. possessions to safer ground. For estimating the benefits of improved forecasts of To a first approximation, the baseline is the current routine climate—Type II and Type III benefits—sim- level of expected losses if there is no reason to be- ple benefits transfer is likely to be the most suitable lieve that expected losses will change over the time approach, given likely time and budget constraints horizon of the analysis (absent the project). In prac- that render infeasible primary user/sector-specific tice, for the time horizons likely to be considered studies of the type described in Section 5. We illus- when evaluating hydromet investments, this is un- trate the use of benefits transfer to estimate these likely to be true, for at least four reasons: types of benefits below. • the magnitude of assets and populations at risk is likely to rise over time, increasing losses; Step 4: Quantifying Benefits Associated with • existing physical hydromet infrastructure will Extreme Meteorological Events (Type I Benefits) deteriorate/depreciate, reducing the quality of Specifying an Appropriate Baseline. The ap- services provided, thereby increasing losses; proach used to estimate Type I benefits entails • households, enterprises, and government estimating the reduction in losses resulting from agencies may autonomously adapt to extreme hydromet investment relative to the baseline. Ac- events, mitigating the losses that they cause; cordingly, the first task is to quantitatively identify and the appropriate baseline, i.e., the expected losses • for sufficiently long time horizons, the effects of that would be experienced from extreme meteoro- climate change on the frequency or severity of logical hazards in the absence of the investment extreme events may result in departures from over the time horizon of the analysis. “current” experience, likely increasing losses. This baseline should take the form of sector-by-sec- The first reason is self-explanatory; it is worth not- tor estimates of expected losses in the absence of ing though that depending on the time horizon be- the investment. By disaggregating expected losses ing considered and the pace of change in the coun- by sector, it is possible to arrive at a more accurate try’s economy, for some sectors (e.g., agriculture), estimate of total expected benefits. Each sector assets at risk may decline if the composition of the 29 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems economy is changing. For time horizons of 10–15 the absence of the investment; future conditions years or more, deterioration of existing hydromet can differ appreciably from current conditions. Giv- infrastructure can be significant and could result en the nature of extreme meteorological hazards, in a non-trivial reduction of the quality of hydromet it is impossible to reliably forecast their number services provided, increasing expected losses from and severity years into the future. To derive esti- extreme events. mates of expected losses from such hazards in the future, as a first approximation it can be assumed The possibility of autonomous adaptation merits that expected annual losses from extreme haz- some explanation. Autonomous adaptation refers ards in future years are equal in magnitude to aver- to adaptation to climate change or extreme weather age annual losses over the recent past. The period events that takes place even without government in- comprising the “recent past” must be sufficiently tervention (e.g., see Malik and Smith 2012). For ex- long to be representative. These “preliminary base- ample, in the face of increasing frequency or severi- line loss estimates” are then adjusted to capture ty of hurricanes, households in developing countries the effects of the four factors listed above. could mitigate the damages they cause by shifting from dwellings made of wood to dwellings made of In the case of the Jamaica analysis, sector-level brick or concrete. If this adaptation has already oc- data on losses from extreme events over the 11- curred to a large degree, then it is presumably al- year period 2000–2010 were used to estimate an ready reflected in estimates of current losses. How- average expected loss by sector and event type. For ever, if the adaptation is in progress, and is likely to events for which sectoral decompositions of losses continue over the time horizon of the analysis, then were not available, sector-by-sector losses were es- estimates of expected losses need to be adjusted timated by taking average loss percentages by sec- downward to reflect the effects of this adaptation. tor and hazard type for events for which loss de- The available empirical literature is largely silent on compositions were available and applying these the magnitude of the losses avoided from this type percentages to estimates of total losses. of adaptation. One notable exception is a recent study of adaptation to tropical cyclones using data A total of 10 climate-sensitive sectors were se- from across the world (Hsiang and Narita 2012). It lected for analysis and three event types (hurri- indicates that only about 3% of the losses from in- canes/storms, floods, and droughts). These prelim- cremental changes of countries’ current tropical cy- inary baseline loss estimates were then adjusted clone climates are “adapted away” in the long run. to reflect the effects of autonomous adaptation, the likely increase in the frequency and intensity Turning to the last reason in the list above, for rela- of extreme meteorological hazards due to climate tively short time horizons of 5–15 years, the conse- change, and the expected increase in the values quences of climate change for the frequency or se- of assets at risk. The latter two adjustments were verity of extreme events are likely to be fairly small. made using interpolation of available year 2030 But for longer time horizons, depending on projec- estimates for three different climate change sce- tions for the country in question, they may have a narios developed for the insurance industry (no, significant impact on expected losses. moderate, and high climate change). To correct for reductions in losses due to autonomous adap- The above discussion makes clear that the appro- tation, losses in all sectors were reduced by 10% priate baseline must reflect future conditions in each year. This liberal estimate of the effects of 30 autonomous adaptation was employed to avoid po- event type, three PLA estimates were considered— tentially overstating the benefits of the proposed in- low, middle, and high values. If only one PLA es- vestment. No adjustment was made for deprecia- timate was available in the literature for a given tion of the existing hydromet infrastructure given sector/hazard, this value was treated as a “high” its already inferior state. value. The “low value” was taken to be 50% of the high value. Where multiple PLA estimates were Estimating Percentage Loss Avoided (PLA). Hav- available, the lowest value was taken to be the low ing estimated expected annual losses in the base- value, and the highest value was taken to be the line, by sector, for each hazard type, the reduction high value. The middle value was the average of in losses with the investment must be estimated. the latter two values. Alternatively, if more than two This is accomplished by first developing estimates PLA estimates are available for a given sector/haz- of the percentage loss avoided (PLA) with the proj- ard type, the median value of the estimates can be ect. The percentage loss avoided will generally vary used as the middle value. across sectors and hazard types. Because of this heterogeneity, a separate PLA estimate should be When feasible, PLA estimates derived using bene- developed for each sector and hazard type. The PLA fits transfer should be reviewed by in-country sector is equivalent to the product Ri x Si used in the sec- experts to ensure that the estimates are plausible toral approach (see Section 5), where Ri is the per- and reflect local conditions and the specifics of the centage of losses that are potentially preventable in hydromet investments being considered. sector i, and Si is the percentage of potentially pre- ventable losses in sector i that are avoided with the Estimating Expected Losses Avoided. Multiply- hydromet improvement. The product of these two, ing the PLA estimates by the baseline expected Ri x Si, gives the percentage of losses that are avoid- annual losses for each sector/hazard yields esti- ed in sector i with the hydromet improvement. mates of the expected annual reduction in losses from extreme meteorological hazards. Summing Estimates of PLA by sector are determined by some these estimates over all sectors and hazard types combination of benefits transfer7 and expert opin- for each year yields an estimate of expected bene- ion. For example, for the Jamaica analysis, absence fits for that year. of Jamaica-specific information resulted in a reli- ance on benefits transfer, i.e. use of PLA estimates Correcting for Forecast Inaccuracy. As noted in derived in similar studies of similar countries. In Section 4, forecast inaccuracy implies that forecasts cases where PLA estimates for similar studies/ generate costs in addition to benefits. Percentage countries could not be found, available PLA es- loss avoided estimates in the literature generally do timates were reduced by 50% to be conservative not appear to account for these costs. Therefore, and to acknowledge differences in climate, geog- some correction for forecast errors is appropriate raphy, economic structure, and hydromet systems. in order to guard against overstating the benefits of hydromet investments. Subbiah et al. (2008) pres- PLA estimates drawn from the literature were also ent a simple method for doing so, absent specif- subject to sensitivity analyses. For each sector/ ic information on the costs of forecast errors. The 7 For example, estimates of percentage loss avoided for improvements in hydromet systems and associated early warning systems can be found in Subbiah et al. (2008), Bangladesh Water Development Board (2006), and Rogers et al. (2009). 31 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems key assumption underlying the approach is that the examples can be found in the Jamaica hydromet cost of an incorrect forecast is equal in magnitude study mentioned earlier. to the benefit of a correct forecast. Letting A denote the average accuracy of the relevant forecast, ex- The output from this step is a time series of an- pressed as a percentage, expected annual benefits nual benefits for the time horizon of the analysis need to be multiplied by a correction factor of: for a number of climate-sensitive sectors of the economy. Correction Factor for Forecast Inaccuracy = (Probability Forecast Correct) Step 6: Quantifying Benefits to Households – (Probability Forecast Incorrect) = A – (1–A) Associated with Routine Climate (Type III Benefits) Estimates of Type III benefits, i.e., benefits to house- This correction factor reflects the fact that benefits holds from improved forecasts of routine climate, of improved forecasts are enjoyed A% of the time, generally rely on stated preference methods. As dis- while costs of incorrect forecasts are incurred (1– cussed in Section 5, these methods rely on surveys A)% of the time. The assumption that benefits from of samples of users to directly or indirectly obtain es- correct forecasts are equal in magnitude to the timates of the value users place on improvements to costs of incorrect forecasts allows for use of this forecasts. Stated preference methods take the form simple correction factor equal to the difference in of stated value or stated choice surveys. In stated probabilities of the two events. value surveys, users are directly asked what they are willing to pay for an improvement to a forecast. The output from this fourth step is annual esti- In stated choice surveys, users are asked questions mates of Type 1 benefits from the hydromet invest- that require them to make choices among alterna- ment over the time horizon of the analysis. tive forecasts with different costs associated with each. Users’ preferences indirectly reveal the value they place on forecast characteristics. Step 5: Quantifying Benefits to Enterprises Associated with Routine Climate (Type II Benefits) Annex 5 provides an illustration of how stated— The approach used to estimate the benefits of im- preference-based estimates of the value of fore- proved forecasts of routine climate is qualitative- cast improvements in the U.S. can be transferred ly different from that used to estimate the benefits to a developing country. associated with extreme events. Instead of esti- mating reductions in losses, it relies on estimates The output from this step is a time series of annual of increases in benefits or GDP by sector. Howev- benefits for the time horizon of the analysis for the er, the approach also makes use of benefits trans- household sector. fer supplemented by expert opinion. The precise application of benefits transfer will depend on the Step 7: Dealing with Uncertainty sector being examined. Annexes 3 and 4 illustrate An inevitable feature of any economic analysis of how estimates from existing studies can be trans- large and long-lived projects such as hydromet invest- ferred to other settings. The examples presented ments is uncertainty about the magnitude of bene- in the annexes are for two important climate-sensi- fits and costs. There are a number of ways of dealing tive sectors—energy and transportation. Additional with this uncertainty. The simplest and conceptually 32 most straightforward approach is to conduct a sensi- to the estimates in UNISDR (2008). The relevant tivity analysis on benefits and costs when computing data are presented in Table 6. Assume that the lev- the present value of net benefits (described in the el of hydromet services delivered to the population next step). Expert opinion or ranges of estimates for in all countries is equivalent under the cooperation transferred benefits and costs can be used to deter- and no cooperation scenarios. Further suppose, for mine high, middle, and low values for a wide set of simplicity, that the present value of benefits from benefit and cost estimates. This is the approach ad- hydromet modernization is certain and equal to 80 vocated here and discussed further in Step 9. million EUR. The expected present value of costs in the third column of the table 6 is computed by Another approach is to adjust upwards the discount weighting each scenario by its likelihood of occur- rate used to calculate the present value of bene- rence: 0.70*63 + 0.30*90 = 71.1 million EUR. This fits and/or costs to acknowledge uncertainty about gives an expected net present value of moderniza- their magnitude. This approach is conceptually less tion equal to 80–71.1 = 8.9 million EUR. Notice that desirable since it subjects all costs and benefits to if sensitivity analysis alone is used to compute the the same adjustment for uncertainty. present value of expected net benefits without de- termining the expected present value of costs, then A third, and conceptually more defensible, ap- the net present value is positive under cooperation proach is to compute expected values for benefits (second row and column) but negative under no co- and costs using alternative estimates of the pos- operation (second row, third column). sible values of benefits and costs along with es- timates of the probability of each value being re- While the expected value approach to dealing with alized. This approach is in essence a means of uncertainty is conceptually straightforward, the aggregating the results from sensitivity analyses of challenge is determining the probabilities to assign benefits and costs using probabilities as weights. to possible values of uncertain benefits and costs. One approach is to rely on expert opinion. Another The expected value approach can be illustrated by is to simply assign equal probabilities to possible returning to an example based on the UNISDR study values of an uncertain cost (or benefit). discussed in Section 7 (under Cooperation and Economies of Scale). Recall that coordination be- tween countries was argued to lower hydromet mod- Step 8: Aggregating Benefit and Cost Estimates to ernization costs. Suppose experts determine there Obtain Present Value of Net Benefits is a 70% probability of cooperation (30% of not co- Given that the benefits and costs associated with operating) in forecast generation and warnings that hydromet investments occur over a multi-year time would reduce the present value of costs according horizon, benefits and costs must be expressed in Table 6  Expected Cost Example (euros million) Cooperation No Cooperation Expected PV PV of Total Costs 63 90 71.1 Net Present Value 17 –10 8.9 Source: Based on data from UNISDR (2008), and assuming a 70% probability of cooperation and present value of benefits equal to 80 million EUR. 33 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems present value terms using an appropriate discount Step 9: Conducting Sensitivity Analyses rate. Economic theory does not clearly specify the Sensitivity analyses are an important component of choice of discount rate. However, (annual) discount any economic evaluation. This is especially true in rates commonly used for evaluating publicly fund- the context of estimating the net benefits of invest- ed projects range from 4 percent to 12 percent ments in hydromet systems, given the high degree in real terms. Once all benefits and costs are ex- of uncertainty regarding the precise magnitude of pressed in present value terms, they can be aggre- benefits and costs. The robustness of conclusions gated to obtain an estimate of the present value of drawn can be assessed by conducting sensitivity net benefits of the project. analyses using: The choice of discount rate will generally influence • alternative values of benefits estimates; both the magnitude and sign of the present value • alternative values of cost estimates; of net benefits. As noted in Section 7, a substantial • alternative time horizons; portion of the costs of hydromet improvements are • alternative real discount rates; incurred in the initial years of a project. In contrast, • alternative assumptions about the increase in the majority of benefits are likely to accrue with expected losses from extreme meteorological a lag given the time needed to bring new equip- hazards over time due to climate change; and ment online, train personnel, establish means of • alternative assumptions about adaptation by effectively communicating forecasts, and educat- households and enterprises that reduces dam- ing users on how to make use of forecasts. This ages from more frequent, or more intense, ex- difference in the time profiles of costs and bene- treme meteorological hazards even without the fits implies that higher discount rates will result project. in a larger reduction in the present value of bene- fits than in the present value of costs, lowering the The sensitivity analyses allow the analyst to identify present value of expected net benefits. the set of values that are critical to the conclusions drawn from the economic evaluation of the invest- The output of this step is an estimate of the pres- ment. Further effort can then be devoted to refining ent value of expected net benefits from the invest- estimates of these values if desired. ment. This estimate should be accompanied by a characterization of the unquantifiable benefits de- scribed in Step 2. 34 9. Interim and Ex-Post Evaluations of Hydromet Investments The framework described in the previous section • Improved estimates of percentage loss avoid- provides an ex-ante evaluation of a proposed hy- ed (PLA). Better data on losses from extreme dromet investment. If the investment is undertak- events would allow for improved estimates of en, interim and ex-post evaluations can also be the percentage loss avoided used to estimate conducted. These evaluations can yield multiple the benefits of improved forecasts of extreme benefits. They provide an opportunity to: events. PLA estimates could be further refined using in-country assessments of the types of • assess the validity of the ex-ante evaluation losses avoided with the improved forecasts and refine it; generated by the investment, and through ag- • collect information that can be used to modify gregation of expert opinion from various user project priorities and allocation of funds in or- groups; der to increase the net benefits yielded by the • Data on the costs of hydromet investments. As investment; noted in Section 7, data on the fixed costs of • enhance the design of future hydromet invest- hydromet investments can be estimated fairly ment projects; and accurately ex-ante, however, the variable costs • improve the accuracy of future ex-ante evalua- associated with operating and maintaining im- tions of hydromet investments. proved hydromet system are more difficult to estimate, especially in developing countries. In Interim Evaluations. Data collected as part of an the course of implementing a hydromet project, interim evaluation of an investment can be used to data on both variable and fixed costs, such as improve ex-ante estimates of the expected net ben- actual costs of purchasing and installing hard- efits of the investment. Referring to the framework ware, upgrading communication networks, described in the previous section, three types of in- training staff, and hiring new staff, should be formation would be particularly valuable: collected as these costs are incurred and then compared to ex-ante estimates. • Data on losses from extreme meteorological events. If data on losses from extreme events Ex-Post Evaluations. The types of information col- have traditionally not been collected, acquir- lected to conduct interim evaluations are also use- ing data on these losses would allow for bet- ful for conducting ex-post evaluations. Ideally, an ter estimates of the magnitudes of losses that ex-post evaluation would use data on the realized could be avoided with hydromet investments. benefits and costs of the hydromet investment for Information on such losses could be drawn a few years after implementation of the investment from data on: (i) private and government insur- to derive a revised estimate of expected net bene- ance payments for property damages, loss of fits. However, given the stochastic (or random) na- livestock and agricultural output, etc.; (ii) data ture of climate and extreme events, such an anal- on private and government expenditures on ysis would need to be conducted with care. For rebuilding and repairing infrastructure; and example, a comparison of losses due to extreme (iii) government assistance to households af- events before and after the hydromet investment fected by extreme events; would be meaningful and informative only if the 35 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems pre- and post-investment extreme events were sim- forecasts and their uses have changed as a result ilar in character, in terms of the type of event, its of the hydromet investment. In particular, informa- timing and severity, and the area affected. Simi- tion could be collected on: larly, comparisons of agricultural output pre- and post-investment would allow for benefits of the hy- • Changes in forecast characteristics. Important dromet investment to be reliably estimated only if characteristics of forecasts are their accuracy, meteorological (and economic) conditions a few frequency, lead time, and spatial resolution; years after implementation were similar to condi- • Perceptions of households regarding chang- tions before the investment. es in forecast characteristics and the value of these changes. Household surveys could Meaningfully comparing losses from similar ex- be conducted to learn whether households treme events before and after an investment is have noticed changes in forecast characteris- more likely to be feasible than comparing econom- tics, and how they have made use of perceived ic productivity or output (e.g., crop yields). A com- changes in forecasts and benefited from the parison of productivity or output could reasonably changes. The surveys should ideally be con- be conducted only several years after implementa- ducted for a cross-section of socio-economic tion of the investment. Only then would there be a groups and regions. In addition, stated prefer- sufficiently long time series of data to reliably es- ence surveys could be conducted to estimate timate average changes over time. Using data for the value placed by households on the per- only one or two years post investment would ren- ceived changes; and der the comparison susceptible to bias due to the • Perceptions of enterprises regarding chang- vagaries of weather in a given year. For example, es in forecast characteristics and the value of crop production from one year to the next could these changes. As in the case of households, vary simply because of changes in the quantity or enterprises and specific user groups could be timing of rainfall. The effects of these changes in surveyed to determine whether they have ob- rainfall would be difficult to disentangle from the served changes in forecast characteristics effects of the hydromet investment. and how any perceived changes have benefit- ed them. In addition, surveys could be conduct- A pitfall of waiting several years to conduct the ex- ed to learn what additional changes in forecast post evaluation, in order for sufficient data to be characteristics and dissemination would be of collected, is that holding constant the effects of most value to various user groups. other changes in the economy becomes more diffi- cult. If other changes are not held constant, the ef- This information would allow for an assessment of fects of the hydromet investment per se cannot be the extent to which improved forecasts have been reliably identified and estimated. effectively communicated to users and the ability of users to benefit from them. As emphasized in The above issues imply that conducting a tradition- Section 4, effective communication of forecasts al ex-post evaluation of hydromet investments is to users, their ability to understand the forecasts problematic and may not always be feasible. An al- and incorporate them into their decision-mak- ternative approach, one that is more generally ap- ing, and their ability to respond to forecasts are plicable and that yields valuable information, is to critical to generating value from forecasts. Fur- assess the extent to which the characteristics of thermore, an understanding of how various user 36 groups make use of forecasts, and of the fore- As indicated earlier, information gained from inter- cast characteristics that are of most importance im and ex-post evaluations, including the findings to them, can help guide decisions on the types of the surveys described above, can be used to in- of hydromet services to provide and enhance. As form future hydromet projects in other countries, discussed in Section 4, identifying and prioritiz- thereby generating project spillovers. Specifical- ing high-value or “high-impact” forecasts enables ly, these data collection efforts and accompanying the net benefits generated by hydromet systems analyses can potentially make future hydromet in- to be increased. vestment projects cheaper and more effective. 37 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems 10. Conclusions and Recommendations Hydromet investments can generate a wide range Given the high degree of uncertainty regarding of benefits that are enjoyed by a broad swath of the precise magnitude of benefits and costs as- a country’s economy. The benefits take the form sociated with hydromet investments, conducting of reduced damages from extreme meteorological a wide range of sensitivity analyses is an essen- events because of better forecasts, and increased tial component of the proposed framework. Sen- profits, or reduced costs, due to improved deci- sitivity analyses should be conducted using alter- sion-making by users whose day-to-day econom- native estimates of benefits and costs, alternative ic welfare is affected by weather, such as house- time horizons, alternative assumptions about the holds, farmers, and energy producers. consequences of climate change for the frequency and magnitude of extreme events, and alternative Estimating the benefits, and hence the net benefits, assumptions about the ability of households and of hydromet investments with precision is a chal- enterprises to adapt to these events. lenging task given the wide variety of forecast us- ers and the very varied decision contexts in which Not all the benefits of hydromet investments can forecasts are used. The task is further complicated be quantified and monetized. Although such bene- by the fact that the magnitude of benefits depends fits fall outside the framework of a traditional eco- on the precise characteristics of the forecast, the nomic evaluation (in which all benefits and costs effectiveness with which forecasts are communi- are monetized), they should not be ignored. Bene- cated to users, and the ability of users to under- fits that are likely to be large in magnitude should stand and make appropriate use of the forecasts. be identified and characterized even if they cannot be quantified or monetized. The framework proposed here for estimating the (ex-ante) expected net benefits of hydromet invest- Supplementing ex-ante evaluation of a hydromet ments is intended to yield first-cut estimates of net investment with interim and ex-post evaluations of- benefits at the country level, without onerous data fers a range of benefits beyond the obvious abili- or analytical requirements. The framework draws ty to assess the validity of the ex-ante evaluation on existing approaches with some important modi- and refine it. In particular, priorities and allocations fications and additions. In particular, it attempts to of funds for the investment under study can be re- capture the benefits of improved forecasts of ex- vised and design of future hydromet investments treme meteorological events as well as the benefits can be improved. Accordingly, collection of data for from improved forecasts of routine climate. To the conducting interim and ex-post evaluations plays extent possible, estimates of the first type of ben- an important role in enhancing the net benefits efits are based on historical data on losses from from hydromet investments in the long run, in addi- extreme events. The proposed framework makes tion to enhancing the ability to evaluate the invest- extensive use of benefits transfer. However, the re- ments ex-ante. sults of benefits transfer should be subject to re- view and revision by sector experts. 38 Payoffs for Alternative Harvesting Table A-1  Annex 1: Economic Value of a Forecast – Decisions An Illustrative Example Weather Tomorrow (USD) To develop an understanding of some of the key de- terminants of the economic value of forecasts and Farmer’s Option No Rain Rain the types of information needed to estimate these Option 1: harvest all today 20,000 20,000 values, we present a simple example of the bene- Option 2: harvest ½ today, 24,000 12,000 ½ tomorrow fits generated by an improved, or better, forecast.8 The example is based on an agricultural setting, Source: Based on Quirk (1976) and Macauley (1997). but the insights derived are generally applicable. costs). If the farmer harvests only half the crop to- Decision Context. Consider a farmer who must day, then his payoff is only USD 12,000 if it rains decide when to harvest his crop in the face of un- tomorrow because the remaining crop will be ru- certainty about heavy rains that would damage the ined (USD 15,000 in revenues minus USD 3,000 in crop. The farmer is certain that it will not rain today costs). But if it doesn’t rain tomorrow, he will also but believes there is some chance that there will be able to harvest and sell the remaining half and be (heavy) rain tomorrow. Suppose the farmer has will earn USD 24,000 in total. two options: If we let p denote the probability that it rains tomor- 1. Option 1: harvest the entire crop today, at a to- row as perceived by the farmer, we can calculate the tal cost of USD 10,000 farmer’s expected payoff for each of the two options: 2. Option 2: harvest half the crop today and half tomorrow, at a cost of USD 3,000 per day Expected Payoff for Option 1 without (1) EPNF1 =  (Better) Forecast The higher cost associated with Option 1 could be = p(USD 20,000) + (1–p)(USD 20,000) due to having to pay labor overtime. If it rains to- = USD 20,000 morrow and the farmer has chosen Option 2, he Expected Payoff for Option 2 without (2) EPNF2 =  will only be able to harvest half his crop (the portion (Better) Forecast harvested today); the other half will be damaged = p(USD 12,000) + (1–p)(USD 24,000) and rendered worthless. Each half of the crop can = USD 24,000 – (USD 12,000)p be sold for USD 15,000. The “payoff matrix” facing the farmer, which shows the farmer’s profits for the The value of p perceived by the farmer could be de- two options with and without rain tomorrow, is pre- termined by a currently available forecast of tomor- sented in Table A-1. row’s weather, or it could be entirely subjective, i.e., based on the farmer’s prior experience and beliefs If the farmers harvests the entire crop today, wheth- regarding weather. er it rains tomorrow is irrelevant. With or without rain, the farmer’s profit or payoff is USD  20,000 If we assume, for simplicity, that the farmer is risk (USD 30,000 in revenues minus USD 10,000 in neutral, he will choose the option that yields the 8  The example builds on one in Quirk (1976) and Macauley (1997). 39 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems higher expected payoff. This preferred option clear- We have already determined the farmer’s expected ly depends on the value he attaches to the prob- payoff without the forecast (captured by the kinked, ability p. For example, if he believes with certain- hashed line in Figure A-1). Therefore, we only need ty that it will rain tomorrow (p = 1), then Option 1 to derive the farmer’s expected payoff with the fore- yields the higher expected payoff (USD 20,000 vs. cast. That we are computing the expected pay- USD  12,000). On the other hand, if he believes off with the forecast merits some explanation. The there is no chance of rain tomorrow (p = 0), then Op- farmer does not know what the forecast is (rain or no tion 2 is preferred (USD 24,000 vs. USD 20,000). rain) until he obtains it. Therefore, ex-ante, i.e., be- fore the forecast is available, we can only compute The expected payoffs for the two options are de- his expected payoff with the forecast. This expect- picted in Figure A-1. The solid, horizontal line is ed payoff is calculated using the farmer’s ex-ante the expected payoff with Option 1, and the down- assessment of the probability that it rains tomor- ward-sloping, solid line is the expected payoff with row. If the forecast reveals that it will not rain tomor- Option 2. The two lines intersect at p = 1/3. This row, the farmer will choose Option 2, since it yields is the value of p at which the expected payoffs for a larger payoff in the event of no rain (USD 24,000 the two options are equal. Notice that the prefer- vs. USD 20,000). Ex-ante the farmer believes that ence for one option over the other (i.e., the gap the probability of the forecast revealing that it will between the two lines) grows larger as the value not rain tomorrow is equal to (1–p), which is simply of p moves away from 1/3 in either direction. The his assessment of the probability that it will not rain hashed portions of the two lines show the expect- tomorrow. Correspondingly, if the forecast reveals ed payoff when choosing the preferred option for that it will rain tomorrow, the farmer will choose Op- each value of p. tion 1 (payoff of USD 20,000 vs. USD 12,000 with Option 2), and ex-ante the farmer attaches a proba- Value of a Perfect Forecast. To estimate the val- bility of p to receiving this forecast. Thus, ue of a better forecast in the simplest setting pos- sible, let us assume that the improved forecast is a (4) EPPF = Expected Payoff with Perfect Forecast perfect forecast—the forecast will specify with cer- = p(USD 20,000) + (1–p)(USD 24,000) tainty, and complete accuracy, whether or not it will = USD 24,000 – (USD 4,000)p. rain tomorrow. This is obviously a best-case scenar- io of an improvement in forecasts. What would the This expected payoff is shown by the dashed line at economic value of such a forecast be to the farm- the top of Figure A-1. Note that the expected payoff er? Following economic convention, the value of the with the perfect forecast is higher than the expect- forecast is the maximum amount the farmer would ed payoff without the forecast (hashed line) except be willing to pay for it. This amount would equal the for the extreme cases where the farmer believes difference between the farmer’s expected payoff there is no chance of rain tomorrow (p = 0) or he with the forecast and his expected payoff without it, is certain that there will be rain tomorrow (p = 1). For either of these two extreme cases, the expect- = Expected Value of Perfect Forecast (3) EVPF  ed payoff with the perfect forecast is the same as = Expected Payoff with Perfect the expected payoff without the forecast. Thus, if Forecast the farmer is certain of tomorrow’s weather, even if – Expected Payoff without his beliefs are mistaken, he will not attach any val- (Better) Forecast ue ex-ante to the forecast. 40 Deriving the Expected Value of a (Perfect) Forecast. Figure A-1  24 22 EPNF2 EPNF EPNF1 20 Expected Values (USD thousands) 18 EPNF2 16 14 12 10 8 p=1/3 6 4 2 EVPF EVPF 0 0.0 0.1 0.2 0.3 0.5 0.6 0.7 0.8 0.9 1.0 Perceived Probability of Rain (p) Source: Based on Quirk (1976) and Macauley (1997). The expected payoff to the farmer with no forecast is certain that it will (p = 1). If the farmer believes depends on the option he chooses. EPNF1 is the there is some chance of rain tomorrow (0 < p < payoff with Option 1 and EPNF2 is the payoff with 1), the forecast is of value. Notice from the figure Option 2. The option that yields the higher expect- that the expected value of the forecast is highest ed payoff depends on the probability of rain per- when p takes on a value of 1/3.9 This is the value ceived by the farmer (p). The kinked, hashed line of p for which the farmer would choose one of the represents the highest payoff without a forecast. two options by a coin toss (in the absence of the The expected payoff with a perfect forecast is giv- forecast), because they yield the same expected en by EPPF. The expected value of the perfect fore- payoff. The value of the forecast diminishes from cast is obtained by subtracting the highest payoff this maximum when the value the farmer attach- without the forecast from EPPF. The kinked line la- es to p moves away from 1/3, in either direction, beled EVPF shows the resulting expected value of inducing the farmer to have a stronger preference the perfect forecast. for one option over the other in the absence of a forecast. The vertical distance between the dashed line and the hashed line corresponds to the difference Determinants of Forecast Value. The above ob- on the RHS of equation (3). Hence it captures the servations lead us to the following insights about expected value of the perfect forecast. This value the value of better forecasts: is shown by the kinked, solid line near the bottom of Figure A-1. As the figure makes clear, the value • The expected value of a better forecast to a of the forecast depends on the value of p. As not- user depends on the user’s prior beliefs about ed above, the forecast is of no value to the farm- the uncertainty he faces. If the user does not er if he does not perceive any uncertainty—he is believe there is uncertainty about future weath- certain that it will not rain tomorrow (p = 0) or he er, better forecasts have no value to the user. The expected value of the forecast when p = 1/3 can be calculated as follows: EVPF = EPPF – EPNF1 = EPPF – EPNF2 = (24,000–4,000p) – 9  20,000 = 4,000 – (4,000)(1/3) = 2,666.67. 41 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems • The value of a better forecast is larger, the of harvesting the entire crop in one day is lower, smaller is the difference in expected payoffs and equal to the total cost of harvesting it over two (in the absence of the forecast) across a us- days, USD 6,000. Then the expected payoff with er’s alternative choices; conversely, the value Option 1 would be USD 24,000. In Figure A-1, the of a better forecast is smaller, the stronger is horizontal line labeled EPNF1 would now have a the user’s preference for one choice over an- vertical intercept of USD 24,000. The farmer would other in the absence of the improved forecast. now always choose Option 1, except when he be- lieves that there is no chance of rain tomorrow (p = Note that prior beliefs about the uncertainty will 0), in which case he would be indifferent between likely differ across farmers. Thus, even if the better the two options. A forecast would now be of no val- forecast has no, or little, value for some farmers, it ue to the farmer. The forecast is rendered value- may have high value for other farmers. Given the less because one option is preferred to the other public good nature of forecasts, the total value of a regardless of the value of p. It can therefore be con- better forecast is the sum of all users’ willingness cluded that: to pay for the forecast. • A better forecast is of no value to a user if it Although the economic value of forecasts is rou- does not alter the choices made by the user. tinely expressed in terms of their ex-ante expect- ed value, it is instructive to briefly consider the ex- If, instead, the cost of harvesting the crop in one post value of forecasts. Ex-post, a forecast is of day was somewhat higher, say, USD 7,000 instead value only if it increases the farmer’s actual (rath- of USD 6,000, then the expected payoff with Op- er than expected) payoff. For example, suppose tion 1 would be USD 23,000. It is not difficult to that the farmer perceives the probability of rain to verify that in this case the perfect forecast would be 0.2. From Figure A-1 the expected value of the have a positive value, but its value would be lower perfect forecast is seen positive. In the absence than that depicted in Figure A-1, because the dif- of the forecast, the farmer would choose Option 2 ference in the maximum payoffs for the two options since it yields a higher expected payoff. Now sup- would only be USD 1,000 instead of USD 4,000. pose that the farmer obtains the perfect forecast. This leads to: If the forecast indicates that it will rain tomorrow, he will choose Option 1 instead of Option 2 and the • The value of a forecast to a user is larger the forecast will make him better off: his payoff will be larger are the differences in the maximum pay- USD 20,000 instead of USD 12,000. However, if offs across the choices available to the user. the forecast indicates that it will not rain tomorrow, he will be no better off with the forecast: both with This can be stated somewhat differently, and without the forecast he would choose Option 1. This example demonstrates that even though a • The value of a forecast is larger the larger is forecast has a positive value ex-ante, this may not the loss in expected payoffs from making the be true ex-post. wrong choice. The economic circumstances facing the farmer The example has assumed that the farmer has will also influence the value the farmer places on two options. Suppose, instead, that it is impossi- a better forecast. For example, suppose the cost ble for the farmer to harvest the entire crop in one 42 day—labor or capital constraints prevent him from A general insight provided by the above example doing so. So the only option available to the farm- and the accompanying analysis is that the value of er is to harvest the crop over two days. In this case, a better forecast to a user is determined by the us- the forecast is of no value to the famer because he er’s beliefs about the uncertainty he faces and by is unable to modify his actions to take advantage of the agent’s ability to use the forecast to modify his the forecast. More generally, we have: choices and increase his expected payoff. This abil- ity is determined by, among other factors, the lead • A better forecast is of value to a user only if the time of the forecast, the choices available to the user is able to alter his actions in a manner that agent, and the economic circumstances facing the allows him to earn a higher payoff. user. In the above analysis, it has been implicitly as- The above example makes the simplifying assump- sumed that the better forecast is made available tions that the forecast user is risk neutral, rather to the farmer in time for him to alter his choices. than risk averse, and that the forecast, when avail- Specifically, it has been implicitly assumed that able, is perfect. A few observations are in order the farmer receives the forecast before he has de- about the effects of relaxing these assumptions. It cided whether to harvest the crop over one or two can be shown using formal and more general mod- days. Suppose, instead, that he receives the fore- eling of decision-making by a forecast user, that cast at the end of today. He would then not be able the value of a forecast is higher the better, or high- to switch from Option 2 to Option 1, and the fore- er quality, is the forecast (Katz and Lazo 2011).10 cast would have no value to him: if it rains tomor- This result is consistent with intuition. Two other re- row, there is nothing he can do, and if it does not sults are not. Intuition suggests that the value of rain, he will simply harvest the remaining crop, if a forecast should be higher the more risk averse any. The forecast does not allow him to increase his is the user. But formal analysis indicates that this payoff. Now suppose that the farmer has started is not necessarily true. Similarly, it is plausible to to exercise Option 1, and after harvesting half his think that the value of a forecast is higher if the crop he obtains a forecast indicating that it will not user has more flexibility, i.e., the user has a larger rain tomorrow. He then may still be able to switch set of options from which to choose. However, this to Option 2, which would allow him to increase his too is not necessarily true. Although additional op- payoff (given the original numbers in the example). tions always make a user better off (or at least no It can therefore be stated that: worse off), they may not increase the value of the forecast, because the additional options also make • A better forecast is of value only if it is provided the user better off in the absence of the forecast with sufficient lead time for the user to modify (see Katz and Lazo 2011). his choices and increase his payoff. 10 “Higher quality” has a precise definition in this context that is related to the statistical notions of “sufficiency” and “encompassing.” Measures of quality that do not satisfy the “sufficiency” criterion can lead to so-called quality/value reversals with value falling as quality increases. See Katz and Lazo (2011) for more information. 43 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems scale. Most of the losses are therefore completely Annex 2: The Value of Hydromet Forecasts absorbed by individuals and firms. to the Insurance and Financial Sector Types of Weather Insurance Instruments. There The multi-billion dollar global insurance and finan- are clear synergies between weather events, at- cial sector has expanded in the face of catastro- tempts to forecast these events, and contracts phes since the 1970s, but most of this growth has in the private insurance and financial sector. At a been in developed countries. Hydromet systems are countrywide scale, there are two types of insur- closely linked to the protection afforded by the fi- ance instruments closely related to hydromet sys- nancial and insurance sector, as up to date weath- tems and data: industry or individual agent based er information is needed to design contracts be- (micro), and macro based. These are also some- tween insurers and insured, verify losses, and times referred to as ‘parametric insurance’ instru- initiate compensation. A recent example is weath- ments as they are based on measurements of criti- er indexing, where compensation is provided for re- cal weather and forecast predictions and variables. corded and verified weather events that result in Macro based instruments focus on mitigation to parameters such as temperature or rainfall outside government budgets due to the adverse effects of normal limits set by insurance contracts. These of disasters by allowing governments to hedge instruments are not yet present in developing coun- against potentially large fiscal losses such as infra- tries save for a few small-scale pilot projects involv- structure repair. Examples are insurance offered ing agriculture. Thus, there is at present very little from large philanthropic investors, major insurers, quantitative information concerning the benefits of or other countries that provide development aid in hydromet modernization to the financial and insur- the event of major disasters. Micro based instru- ance sector in developing economies. As a result, ments, for producers or consumers, are risk miti- quantifying the benefits to this sector from hydrom- gation mechanisms aimed at preventing damage et improvements is difficult at present. However, to specific assets and are most commonly provided the benefits could be large for some countries. If by private insurance firms. Examples are insurance so, they should not be ignored even though they are purchased by farmers that guarantee an annual re- currently unquantifiable. turn in the event of rainfall or crop yields below an agreed-upon (by contract) level, degree-day con- Scope of Losses and Importance of Insurance. tracts with power companies that provide compen- The scale and scope of potential losses from ex- sation for temperatures above or below levels that treme weather events as well as periodic cycles and greatly increase costs of energy production, and fluctuations in climate have large effects on country wind power insurance in Europe that provides com- level GDP and per capita wealth. Between 1974– pensation to owners for periods without sufficient 2006, losses in terms of drought, flood, and trop- wind. ical storms accounted for between 0.3%–11.98% of GDP in aggregate losses over that period among Importance of Hydromet System Performance. developing economies comprising Southeastern Improved weather forecasts from hydromet sys- Europe alone (Pollner et al. 2010). Very few of these tems have a large impact on the development of mi- countries possess a well-developed insurance sec- cro and macro instruments in developing countries tor and nearly all have largely insufficient govern- in several ways. Insurers can use accurate monitor- ment budgets with which to deal with losses of this ing and weather information to target instruments 44 where the probability of risk is highest by aiding points of the amount actually needed once an ex- in the determination of premiums and underwrit- treme weather event arrives (UNISDR 2009). UNIS- ing of risk sharing contracts, sell contracts to out- DR reports that in all countries but Kazakhstan the side investment firms that include these instru- economic losses from a two hundred year cata- ments in their portfolios (e.g., reinsurance), and to strophic event are in the range of 100–200 times avoid potentially large losses that the private sec- the annual planned emergency budgetary alloca- tor has trouble compensating and that require gov- tions. Here, better insurance targeting through im- ernment intervention. Further, the insurance sec- proved weather information has great potential to tor is multilayered, with lower layers having direct increase wealth and ensure continued econom- insurance firms collecting adequate premiums to ic growth, as losses from natural disasters are at cover losses, but higher layers required for extreme present funded by borrowing or reallocating exist- events with a much larger capital level at risk—at ing budget money. higher layers, the premium to expected loss ratio can be as large as eight times higher than lower The Current State of the Insurance Sector in De- layers (WRMA 2002). The higher layers also oper- veloping Economies. There are a few examples of ate under greater uncertainty concerning arrival micro insurance instruments in parts of Asia, Ethi- of extreme weather events and the damage that opia, India, and Malawi where crop insurance con- these events cause. For both layers, but especially tracts exist based on drought (e.g., see Suarez et al. the higher ones, the contribution to profits of firms 2008 and ASEAN 2012). However, insurance pen- in the sector from more capable hydromet systems etration is still low; for example, in Central Asia, a can be considerably high. Further, there are indi- region that has garnered recent interest in hydrom- rect benefits of hydromet data that can be difficult et modernization, UNISDR (2009) finds that pene- to quantify, but are important nonetheless. For ex- tration is currently as low as 1.5 percent of GDP at ample, more risk but better information about that risk; this is roughly an average protection of only 12 risk implies new and greater opportunities for a USD per capita, most of which is not attributable to hedge fund industry to build portfolios with assets real property. The UNISDR report identifies the fol- relevant to this risk. The increased forecast accura- lowing reasons for this observation: cy from a modernized hydromet system would un- doubtedly make decisions regarding investments • insurance regulators lack expertise and infor- from this industry more informed and therefore mation to accurately assess risk exposure; more profitable. • firms lack underwriting and actuarial support needed to form contracts; Potential Benefits of Weather Insurance Instru- • there is a lack of reliable weather information ments to Developing Economies. Both micro and to assess true risk exposures to both insurers macro insurance instruments are still rarely used in and insured; and developing economies. As far as macro insurance • the majority of insurer firms do not have rein- instruments go, Central Asia is a representative ex- surance or excess-of-loss protection options. ample of the challenges faced. Here, every coun- try government aside from Turkmenistan devotes The reasons identified by UNISDR exist in part be- some part of their annual budget appropriation to cause accurate indexing against weather informa- emergency weather events. However, this budget- tion is not possible, but they also follow from low- ary protection is often as small as a few percentage er than optimal premiums. It is not surprising that 45 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems the insurance sector in developing countries is very for days above a certain temperature up to some underdeveloped with available insurers vulnerable limit on the total payout, in return for a premi- to extreme weather events. Better weather infor- um paid to the insurance company that is based mation will allow governments to obtain adequate on a percent of the limit. In a more recent discus- market based insurance in order to reduce contin- sion, ASEAN (2012) reviews and reports several gency budget exposure, receive immediate liquidity different types of insurance instruments relevant after weather damage, increase fiscal stability and to weather derivatives in Asian countries. This re- economic growth, and allow real estate owners to port notes that weather derivatives offer many ad- protect assets and firms to protect earnings. Such vantages to the typical catastrophic insurance ar- insurance will also reduce the costs of borrowing at rangements (such as CAT bonding),11 including: all levels in the economy. tailor made compensation and premium indices, multi-annual protection, flexibility with regards to A New Frontier for Hydromet Investments. Most the geography of protection, and rapid payout as of the barriers in the insurance sector could be parameters indexed are quickly and clearly mea- lessened with more accurate weather data record- sured and observed. The report mentions the main ing and forecast information with which to design disadvantage of using these instruments as insuf- contracts, for all parties involved including individ- ficient historical data measurements that impede uals, governments, insurers, and re-insurers. The their adoption in developing countries. Regardless World Meteorological Organization has called the of the type of weather derivative, hydromet infor- development of insurance sectors and markets a mation is critical, and these parametric contracts new frontier in hydromet investments (Golnaraghi cannot exist without accurate forecasts needed to 2009). Golnaraghi argues that the benefits of hy- define the premium payment schedule and event dromet modernization would derive from the avail- payout limits. ability and accessibility of historical and real-time data; data quality assurance, reliable and timely A multiplier type value of hydromet investments to data for contract design and settlement; and fore- the insurance sector can also be realized through casts for forward looking futures investments. In- greater private investor interest in weather deriva- deed, she argues that improved weather informa- tives, such as from hedge fund managers and fu- tion is the reason for the rapid development of tures traders. Futures traders could purchase con- weather-indexed micro insurance and investment tracts from insurance companies, obtaining a high instruments in developed countries. return through collected premiums in periods that extreme events do not occur. Futures trading would Dutton (2002) describes weather insurance con- serve to pool risk and generate additional returns tracts as “weather derivatives.” A common set up both in the financial sector and for those affected for these contracts in the U.S. and Europe involves by extreme weather events. When this interaction use of high/low degree or high/low wind days as a between financial investors and insurers is con- signal of climate variability-based damages. For ex- sidered, the size of the potential monetary ben- ample, a utility could arrange to receive payment efit attributable to better forecasts could be very 11  Catastrophe bonds are security instruments designed to reduce risk for potentially affected insurers by transferring that risk to private investors. First created in the U.S. in response to hurricanes and earthquakes in the 1990s, they were designed so that insurance companies would contin- ue to cover individuals in regions where the value of potential damages exceed premium collections. These bonds are issued through investment companies and sold to investors, who receive potentially high returns in periods where catastrophes do not occur. 46 large. The Weather Risk Management Association by the insurance sector. This study notes that, for (WRMA 2002) estimated that in 2001 the value of Asian cases, there are lower operating costs for weather derivatives was more than USD 4 billion weather derivatives than for traditional insurance, per year in the U.S., and in Europe weather deriva- because weather based (parametric) arrange- tives were worth, in terms of total risk transferred, ments do not require costly monitoring and there is USD 1.5 billion. no costly moral hazard or adverse selection typical in the insurance industry since weather events are Another encouraging fact for hydromet investments observed by all parties involved. is that the scope of meteorological information re- quired to structure weather derivatives may be low Summary of Expected Benefits from Hydromet in cost to obtain once hydromet systems are mod- Modernization. The collective message from the ernized. This would increase the net benefits esti- literature and the reality of weather risk exposure mated for hydromet investment through economic in developing economies—the large potential loss- evaluations. For example, Zeng (2003) and Jewson es, and a large percent of the population exposed— and Caballero (2003) have shown how forecasts imply that weather based instruments, growth in can be used in the pricing of weather derivatives the financial and insurance sector, and modern hy- for heating degree days, finding that accurate cal- dromet systems are closely linked. Although it is culations of expected payoffs in linear contracts re- conjecture, the size of the weather dependent in- quire only accurate mean temperatures over the surance sector in developed countries means that, contract period, while nonlinear contracts require depending on the size of the economy, expected only accurate forecasts of both the mean and the benefits to a developing country from modernizing distribution of temperatures, but not the depen- hydromet systems could range from hundreds of dence between temperature distributions on differ- millions USD to over 1 billion USD once the insur- ent days. Carriquiry and Osgood (2012) also find ance and corresponding financial sectors were ful- that agricultural insurance contracts when properly ly integrated around the new data generated from designed can exploit important synergies between improved forecasts. Moreover, the resulting future forecasts, insurance, and effective input use. Fur- growth in the financial sector and GDP would lead ther, as noted in ASEAN (2012), the critical need to additional government resources to ameliorate for hydromet information in weather derivatives is the effects of extreme weather events on budgets, for development of hazard rate functions for use services provided, and victims of disasters. 47 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems system allows plant operators to better align in- Annex 3: The Value to the Hydropower creased energy production with times of peak ener- Sector of Improved Forecasts of Routine gy demand. However, if a reservoir is allowed to get too full, flooding can occur, which will cause both Climate a lost opportunity for energy production and pos- Energy producers frequently operate in an environ- sible damage to the surrounding area. If the reser- ment of high uncertainly. Demand and supply of en- voir level is allowed to get too low, however, there is ergy are constantly in flux, and can be difficult to a risk of running out of energy at a time of excess predict. Many types of energy generating equipment energy demand. Careful stewardship of the reser- require significant amounts of time to “warm up” voir is therefore a key concern of plant operators. and “cool down,” delaying the speed at which plant operators can react to changes in the energy mar- In order to optimally manage a hydroelectric power ketplace. Continuously maintaining sufficient pow- plant, operators rely on various models to predict er to the grid, while also minimizing excess power future stream flows into the plant reservoir. Inputs production, therefore requires both constant mon- into these hydrological models include forecast- itoring and rebalancing of energy production, and ed weather data, such as air temperature, wind also a sophisticated model to predict short-term speed and direction (especially when snow fall is a fluctuations in energy demand. In addition, extreme factor), precipitation, and humidity. Since streams meteorological events in the form of heat waves, flowing into hydroelectric plants can be fed by droughts, tropical storms, and blizzards, can cripple sources sometimes several hundred kilometers vital portions of the energy infrastructure at a time away (and therefore many days upstream), these when having reliable power is most critical. hydrological models may be very complex and re- quire forecasts over a wide geographic and tem- While electrical energy is too expensive to store, poral range. The final predictions made by these hydroelectric plants have the ability to store wa- models are only as accurate as the data fed into ter in their reservoirs. During times of low energy them, making accurate weather forecasts vital to demand, water is collected in the reservoir. When efficient and reliable energy production. demand increases, that water is released from the reservoir and the current it generates is used Many factors must be considered when estimating to spin turbines, which generate electricity. This the benefits to hydropower from improved weather Table A-2  Benefits to Hydropower on the Missouri River from Improvements in Forecasting Ability Average annual hydropower benefits % increase in benefits from Baseline Scenario/forecast knowledge (USD millions) perfect forecast knowledge Zero forecast skill 359.8 7.1% 1.  Current climate knowledge 363.2 6.1% 2.  Current climate, perfect snow water content 364.5 5.8% 3.  Current climate, perfect and soil moisture content 366.6 5.2% Perfect forecast skill 385.5 — Source: Maurer and Lettenmaier (2004). 48 forecasting, including the current forecasting pow- identical to the Missouri River System but with a er, the incremental improvement in forecasting storage capacity only 1.2 times the annual flow, power, variability of stream flow, and characteris- they find the improvement from no predictive pow- tics of the river system. Chief amongst these char- er to perfect predictive power results in a 7.1% in- acteristics is the ratio of reservoir capacity to to- crease in hydropower benefits. In addition to the tal annual stream flow. If reservoir capacity is very ‘no predictive power’ baseline, the authors also large compared to stream flow, then the ability to calculate the increase in benefits to hydropower predict future stream flow is of little consequence, from perfect forecast skill from three other base- as there would be little opportunity cost to storing lines: (1) current climate knowledge and no knowl- or using excess water. edge of snow water content and soil moisture, (2) current climate knowledge, perfect knowledge Given limited resources, and the complexity of the of snow water content and no knowledge of soil systems involved, one would best approach the moisture, and (3) current climate knowledge, and estimation of net benefits to the hydropower in- perfect knowledge of snow water content and soil dustry from improved weather forecasts through moisture. The benefit from each of these scenari- the use of benefits transfer. Maurer and Letten- os is given in Table A-2. maier (2004) provide one of the few estimates of the benefits of improved stream flow predict- In general, how one should adapt these values to ability—via improvements in climate variable fore- apply benefits transfer depends on both the cur- casting—for hydropower along the Missouri River rent state of forecasting ability and the degree of System. They find that the difference between no improvement from hydromet investments. One forecasting ability and perfect forecasting ability should also adjust the magnitude of the expected leads to only a 1.8% improvement in hydropower benefits based on the ratio of reservoir storage ca- benefits for this region. This is mainly due to the pacity to annual flow, as this ratio can be key to fact that the reservoirs along the Missouri River determining the importance of weather forecasting System can store up to three times the annual in- to hydropower benefits. In the case of the Missouri flow to the reservoirs, making climate forecasting River System, a 60% decrease in this ratio leads to fairly inconsequential. However, when conduct- a nearly 300% increase in incremental benefits to ing the same analysis for a hypothetical system perfect forecasting power. 49 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems planning and delivery for firms who have materials Annex 4: The Value of Hydromet Forecasts moved through road-based systems. For example, to the Transport Sector Leviäkangas et al. (2007) find that the most im- portant and immediate benefits from improved hy- The highly integrated and multilayered nature of dromet systems are information and warning ser- transport sectors means that more accurate fore- vices offered to road users that cause them to alter casts, predictions of seasonal and daily trends, their driving and travel plans in ways that avoid and real time warnings of extreme events, afford harm to them and to property. benefits that are extensive in that they can impact an entire economy. Transport affects consumers Given that hydromet information is needed for an and demand, producers and supply, access to and efficient transport sector, the simplest measure of the functioning of markets, and even ambient en- the benefits of hydromet modernization is the over- vironmental quality conditions. Further, transport all contribution of the transport industry to GDP is multimodal, involving sea, air, and land resourc- and the protection of this economic value afforded es. Outdated hydromet systems are likely to yield by improved weather information. The overall con- inaccurate or untimely forecasts and render the tribution of the transportation industry to GDP in sub-population using the transport sector without many economies is as high as 50% (Han and Fang critical information to reduce susceptibility to ex- 2000). However, as Hallegatte (2012) notes, the treme events and to minimize losses. benefit-cost ratios of improved weather informa- tion in the transport sector, and therefore the im- Avoidable losses from improved hydromet systems portance of hydromet systems to GDP, are heavi- in the transport sector include: ly dependent on the state of the current hydromet system and infrastructure of a county. While the • reduction in deaths or damage to property from benefit of hydromet modernization is undoubtedly accidents that are avoided because of more positive using a GDP protection measure, this ben- accurate forecasting and weather based clos- efit may be lower in developing economies com- ings and evacuations; pared to developed ones. • avoided losses in profits by better-managed shipping and delivery according to weather Economic Value of Hydromet Modernization to variability; the Transport Sector. The preferred approach to • avoided cancellations in transport service us- estimating the economic value of better forecasts age; and in the transportation sector is to evaluate the mon- • better management of capacity within the etary value of decisions that all agents in the sector transport sector and travel decisions made by make with better information. The corresponding individuals. economic value of better decisions is the accom- panying improvement in incomes and profits that The direct benefits of hydromet modernization to follow from having improved climate information. the transport sector are highest for users of the Unfortunately, nearly all of the current evaluations road surface, rather than to the firms and consum- of the economic value of hydromet services and ers that depend on road-delivered items. The lat- transport are based on developed country data, ter may still be significant beneficiaries though be- primarily because the radar infrastructure already cause improved forecasts affect output and input exists and data recording is extensive and reliable. 50 The study by Leviäkangas (2007) is an exception; it for purchasing power and using the ratio of Croa- was carried out for Croatia and is illustrative of the tian adjusted GDP to Finnish adjusted GDP. In their scope and scale of benefits one can expect from study, PLC = 0.41. The implicit assumption here is modernization of hydromet systems in a develop- that human and environmental values are similar ing economy that lacks fully modern infrastructure across countries since the modern hydromet sys- and has lower per capita income than developed tem will be similar, and therefore the only adjust- countries. ment made is due to per capita and country wealth differences. Leviäkangas et al. (2007) use a method for evaluat- ing Croatian hydromet modernization that is based Current State of Croatian Hydromet Services. on Value of Statistical Lives Saved (VOSL) through The Meteorological and Hydrologic Service of Croa- better weather information. While this VOSL is not tia (DHMZ) maintains a working hydromet system, identical to those traditionally estimated, their but it is outdated with a preponderance of older measure is one that has been commonly used in technology and lack of professionally trained mete- transport sector evaluation, as discussed in Ad- orologists. The DHMZ oversees 41 surface meteo- visors to the High Level Group on Infrastructure rological stations, 116 climatological stations, and Charging (AHLG) (1999) and Tiehallinto (2006). It 336 precipitation stations spread across the coun- differs from the traditional measure by not includ- try. However, only 34 are automated meteorological ing changes in risk and predicted losses under stations. There are also only 2 radiosonde stations, these risks, and is therefore essentially a willing- one pilot balloon station, and 3–5 radar stations. ness to pay measure for greater safety augmented Moreover, these stations are understaffed. The by profits and costs avoided by reduced accidents. DHMZ employs a staff of roughly 400, with about Their measure is defined as: 8% of these classified as research staff. Only about 6% have graduate degrees (2% are postgraduate VOSL = WTP + NLP + HLC + ADM + PDV level). where Estimated Benefits to Road Surface Users. an estimate of human willingness to WTP =  Leviäkangas et al. (2007) find that critical aspects pay plus lost consumption of hydromet modernization to road users are relat- NLP = net lost production to firms ed to localized real time information, such as trav- HLC = hospital care cost el warnings or websites that provide updated road ADM = administrative cost conditions, with precipitation amounts, ice and PDV = property and material damage cost snow coverage, and humidity tabulated by location and road network. They find that the benefits of The elements of this VOSL are constructed for improved meteorological information to the entire each transport sub-sector based on expert opin- road transport network for Croatia is as high as 17 ion. A value from a Finnish hydromet study is used million EUR annually assuming a modern hydrom- as a baseline VOSL. This is then adjusted according et system is put in place. In Croatia improved mete- to specific differences between the Finnish econ- orological information enhances winter road main- omy and the Croatian economy using a price lev- tenance by allowing more efficient employment of el coefficient (PLC) that relies on purchasing pow- scarce resources in advance of storms. Leviankan- er parities. It is defined by comparing GDP adjusted gas et al. find that this benefit amounts to 340,000 51 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems Estimated Annual Total Benefits (Euros) Per Capita and Per Driver from Weather Table A-3  Forecasts and Warnings Under Two Hydromet Assumptions Current Hydromet System Modernized Hydromet System Sub-Sector Per Capita Per Driver Per Capita Per Driver Road Transport 0.69–1.38 1.55–3.20 1.12–2.23 2.5 Rail Transport Not Assessed 0.03 Maritime and Inland Waterway Transport 0.96–1.76 Not Assessed Aviation Transport 2.72 3.35 Source: Based on Leviankangas et al. (2007) and Energy Institute (2012). EUR annually in cost savings to local and federal maritime transport are: safety of human beings, governments, as well as additional profits to firms protection of environmental values through avoid- who are able to adjust production and delivery of ance of accidents such as spills or sinking, and services optimally as storms arrive. It is worth not- more reliable material delivery. ing that rail transport is very similar to road trans- port in terms of the types of benefits from hydrom- Aviation Transport. The aviation sub-sector is sim- et improvements. ilar to the maritime sub-sector in terms of the eco- nomic value of hydromet modernization. The main Maritime and Inland Water Transport. The mar- benefits are improved safety from ground radar and itime transport industry is arguably the transport storm radar monitoring, more efficient scheduling sub-sector that is most dependent on hydromet that reduces fuel and time costs, and increased products and services. Important hydromet ser- reliability and demand for services. Leviankangas vices include nowcasting and early warning sys- et al. find for Croatia that accident savings due to tems for storms and tides, daily condition fore- weather information services amount to over 5 mil- casts, medium term forecasts (4–10 days), and lion EUR per year, a number that will likely grow seasonal forecasts (1–6 months). Several studies considerably as the Croatian economy continues to have found that as much as 20% of all the delays in expand (in fact, in developed countries these bene- maritime service delivery are weather related, and fits can be as high as 6–7 million EUR annually due as much as half of these delays could have been to better flight management, irrespective of safe- entirely eliminated by more sophisticated weath- ty benefits). Greater forecast accuracy and up-to- er monitoring and reporting systems (Smith 1990 date weather information also afford considerable and Thornes and Davis 2002). Extreme events also fuel savings through improved decisions concern- have potential to cause great damage to ships and ing flight altitude, routes, and alternate airports cargos, leading to very large financial losses. Thus, which can be fine-tuned to changing weather and while meteorological information is critical to on- wind conditions. In Croatia, improved flight and fuel board decision-making (such as routes and sail decisions yield benefits of approximately 1 million dates), other maritime activities are also affect- EUR per year. Other smaller impacts are also at- ed. The most important decision impact areas of tributable to better forecasts, including reduction 52 in CO2 emissions that follow more informed sched- estimates assuming forecast information repre- uling and fuel management. sentative of a fully modern hydromet system. The relative differences between a developing country Estimation of Total Benefits of Hydromet Mod- hydromet investment and the current system de- ernization. Table A-3 provides a summary of the scribed for Croatia can in principle be used to ad- discussion in this annex and an estimate of the an- just the figures in Table A-3 to any local developing nual per capita total benefits to the transport sec- country hydromet system status. tor from hydromet modernization. These estimates were taken from Leviäkangas et al. (2007) and Based on the VOSL measure discussed above, the based on their estimation of the VOSL equation elements in the Table A-3 are equal to the benefits discussed above. In addition, the 2013 population added together for accident avoidance, plus ben- of Croatia, 4,480,043 and a current estimate of efits in terms of profits from better resource man- the total number of drivers of all types (commer- agement, i.e., better resource allocation, as well cial and personal use) of 2,000,000 in 2012 (Ener- as savings in rescue and mitigation activities that gy Institute 2012) are used to convert benefits to a accompany better forecasts. Aviation benefits the per capita value and, in the case of road transport, most, followed by road and then rail, at least for the to a per driver value. assumed hydromet modernization. The main ben- efits for the road category are increased road safe- The first set of columns in Table 8 under the head- ty and operational maintenance; the main benefits ing “Current Hydromet System” presents estimates for rail are time savings and track management; assuming the current (pre-investment) hydromet the main benefits of maritime and inland water- system, while the second set of columns under the ways are scheduling and accident reduction; and heading “Modernized Hydromet System” presents the main benefits for aviation are enhanced safety. 53 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems are associated with improved forecasts of extreme Annex 5: Economic Analysis of meteorological events such as hurricanes, storms, Improving Climate Data and Information floods, and droughts, as well as improved forecasts of routine weather and climate conditions.12 Im- Management Project—Jamaica proved forecasts of extreme meteorological events and effective dissemination of information about Introduction their effects and appropriate responses to them This annex presents an economic analysis of the can substantially reduce economic losses caused Improving Climate Data and Information Manage- by the events. Improved forecasts of routine cli- ment Project for Jamaica. The project has four ob- mate can result in increased enterprise profits (or jectives: (i) to upgrade Jamaica’s hydromet system reduced costs) and improved decision-making by so as to enhance climate monitoring, weather fore- households. Table A-4 presents examples of these casting and early warning systems; (ii) to assess benefits, emphasizing the benefits associated with the expected consequences of climate change for forecasts of routine climate, which are often less selected sectors of the economy using climate sce- apparent. The examples illustrate the pervasive im- narios, and to enhance climate resilient planning portance of climate forecasts to economic activity. and decision-making in these sectors; (iii) to put in place a comprehensive risk information platform; Existing Estimates of Benefits. A large number and (iv) to improve public knowledge, attitudes and of studies have attempted to estimate the econom- practices towards climate change. ic benefits of climate forecasts. Surveys of these studies can be found in Nichols (1996), Anaman et The analysis identifies the types of benefits gener- al. (1998), Stern and Easterling (1999), Houston et ated by the project and, where feasible, presents al. (2004), Katz and Murphy (2005), Weiher et al. first-cut estimates of the magnitudes of the bene- (2005), Teisberg and Weiher (2009), and Rogers fits. These estimates form the basis of a cost-ben- and Tsirkunov (2010). efit analysis. This cost-benefit analysis is likely un- derestimating the magnitude of benefits, because A small subset of this literature focuses on esti- a number of the benefits generated by the project mating the benefits of improvements in hydromet cannot be quantified given available data. In partic- sytems. Instead of estimating the total benefits of ular, benefits associated with the second and fourth existing, or hypothetical perfect, climate forecasts, objectives listed above are difficult to quantify. this body of literature estimates the incremental benefits associated with improved forecasts result- Benefits of Improved Climate Data and Information ing from upgrading hydromet systems. This distinc- Systems tion between total and incremental benefits is em- Types of Benefits. A wide range of benefits have phasized by Freebairn and Zillman (2002). Much been attributed to upgrading hydromet systems and of the work on incremental benefits has been asso- improving dissemination of information about me- ciated with evaluations of investments in upgrad- teorological conditions and hazards. The benefits ing hydromet systems in countries of the former For brevity, the term “climate” is used to refer to both “weather” and “climate” when a distinction between the two is not important. Weather 12  refers to atmospheric conditions over a period of days or weeks, whereas climate refers to atmospheric conditions over a longer period of time, such as months, seasons, years or decades. 54 Table A-4  Examples of Benefits from Improved Climate Forecasts Period of Sector Forecasta Benefits from Forecast Agriculture Short-Range Information on daily precipitation is vital to pesticide application decisions, Forecasts as heavy rains can wash away recently applied pesticides. Medium-Range Accurate medium term precipitation forecasts inform farmers whether or not Forecasts they need to irrigate, and how much. Ideal seeding rates are sensitive to the weather conditions in the days and weeks following planting. Timing of planting and harvesting decisions can be improved. Seasonal Having more accurate data on the seasonal climate can aid farmers in deter- Forecasts mining which crops will yield more value. Crop insurance programs can benefit from reduced uncertainty of weather patterns. Agencies can anticipate food shortages earlier with better seasonal forecasts. Household Short-Range Weather forecasts are used to make everyday decisions, such as what to Forecasts wear, or whether or not to take an umbrella. Medium-Range Early warnings of major storms can signal the need for a household to stock Forecasts up on essentials in case of power outages or road closures, or for potentially life-saving evacuations in cases of extreme events. Seasonal Seasonal forecasts can inform households on many decisions, ranging from Forecasts whether or not and how much insurance to purchase, to what type of house- hold fortifications and improvements to undertake. Energy Short-Range Daily weather patterns have an effect on peak energy use patterns, and Forecasts more accurate weather forecasts can inform power plants when to increase or decrease electricity production. Seasonal Hydro-electric generators benefit from improved streamflow forecasts. Forecasts Recreation and Short-Range Golf course management, recreational fishing, and other outdoor and ma- Tourism Forecasts rine-based activities benefit from more accurate temperature and precipita- tion forecasts. Seasonal Tourist resorts make staffing and investment decisions based on expected Forecasts tourists in a given season, which can be very sensitive to seasonal weather. Transportation Short-Range Routing decisions of trucks, ships, and airplanes can be improved with bet- and Shipping Forecasts ter forecasts of daily weather conditions. Medium-Range Cargo ships can use better forecasts to minimize costs and delivery delays in Forecasts maneuvering around unexpected storms. Water Resource Seasonal Improved forecasts can lead to more efficient reservoir operations and sav- Management Forecasts ings from avoiding groundwater pumping to augment reservoirs. Fisheries Medium-Range Fishing vessels rely on weather forecasts to determine when to set sail, how Forecasts long to stay at sea, and where to navigate to avoid adverse weather conditions. Emergency Short-Range Adequate warning before extreme weather disaster can significantly reduce Response Forecasts losses of life. First responders and emergency rescuers can pre-position emergency response assets to places where they will be most effective for rescue operations. Source: Authors. a Short-range forecasts are forecasts beyond 12 hours and up to 72 hours. Medium-range forecasts are forecasts beyond 72 hours and up to 240 hours. Seasonal forecasts are descriptions of averaged weather parameters over the next 3 to 6 months, excluding individual events. 55 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems Soviet Union. Salient examples of such evaluations depends on a multitude of factors. Among the can be found in SEEDRMAP (2008), World Bank most important are (e.g., Mjelde et al. 1989, Stern (2008), Rogers et al. (2009, 2010). and Easterling 1999, Blench 1999, Houston et al. 2004, Teisberg and Weiher 2009, and Lazo and A related body of literature has examined the ben- Waldman 2011): efits of early warning systems. This literature ex- amines the benefits from improved forecasts of • time frame (span) of the forecast, e.g., fore- extreme meteorological events and effective dis- casts can be made of tomorrow’s weather or semination of information about these events and next summer’s average weather; appropriate responses to them. Detailed studies • lead time of the forecast, i.e., the length of time have been conducted on early warning systems for between the issuance of a forecast and the floods, cyclones, and El Nino in various Asian coun- time of the event forecasted; tries (Bangladesh Water Development Board 2006 • spatial resolution of the forecast; and Subbiah et al. 2008), as well as on early warn- • set of weather parameters forecast, e.g., rain- ing systems for floods, hurricanes and tornadoes fall, temperature, etc.; in the U.S. (Carsell et al. 2004, Simmons and Sut- • perceived and actual accuracy of the forecast— ter 2005, Teisberg and Weiher 2009, and Lazo and perceived accuracy can differ from actual ac- Waldman 2011). curacy, especially if past forecasts have been wrong; For the purposes of this analysis, the literatures on • timely dissemination of the forecast in a format incremental benefits and early warning systems that is understandable and useful to users, i.e., are the most relevant, given that a salient objec- households, enterprises, or government agen- tive of the project is to improve Jamaica’s hydrom- cies; and et system and the dissemination and application • ability of users to benefit from modifying their of the information it generates, including improved decisions/actions in light of the information early warning systems. The literature on total bene- contained in the forecast, e.g., a very accurate fits, rather than incremental benefits, would be the flood forecast with limited lead time might en- more relevant if the objective of the analysis were able households to flee a flood zone, but might to estimate the benefits of having a hydromet sys- not give them time to move their belongings to tem, i.e., the benefits of going from a scenario with higher ground. no hydromet system to a scenario with a hydromet system. The objective of this analysis is, instead, to These factors imply that benefits estimates derived estimate the benefits of going from the current hy- using benefits transfer13, as is done here, should dromet system to an improved hydromet system. be viewed as indicative rather than exact. Factors that Influence the Magnitude of Ben- On an economy-wide level, the value of improved efits. It bears emphasis that the magnitudes of forecasts varies across sectors. Some sectors of benefits associated with improved forecasts are the economy are more climate sensitive than oth- very much setting specific. The value of a forecast ers, and are therefore more likely to benefit from Benefits transfer is a commonly used technique that makes use of benefits estimates derived in one setting to estimate benefits in another, 13  similar setting. 56 Table A-5  Major Damage-Causing Meteorological Events, 2000–2010 Damage Damage as Damage (USD Million, Percentage of GDP (Billion $J, Current Constant 2010 (Official Exchange Event Year Prices) Prices) Rates) Tropical Storm Nicole a 2010 20.60 235.75 1.76% Tropical Storm Gustav a 2008 15.50 262.35 2.01% Hurricane Dean1a 2007 23.80 519.08 4.09% Hurricanes Dennis and Emilya 2005 5.98 170.84 1.54% Hurricane Wilma a 2005 3.60 102.85 0.93% Major Drought b 2005 0.38 16.74 0.15% Hurricane Charleya 2004 0.44 14.74 0.14% Hurricane Ivan a 2004 36.99 1,239.14 12.18% Floods 5/21/02–6/4/02 c 2002 2.47 130.95 1.35% Hurricane Michelle a 2001 2.52 150.49 1.65% Floods 12/29/00–1/4/2001c 2001 0.20 11.94 0.13% Major Drought b 2000 0.25 16.74 0.19% a Planning Institute of Jamaica (2011). b Government of Jamaica Ministry of Agriculture & Fisheries (2010). c United Nations and IDB (2007). the improved forecasts. The sectors generally con- meteorological events have imposed large econom- sidered to be climate sensitive are: agriculture, avi- ic costs virtually every year of this century. The Stra- ation, construction, surface and water transporta- tegic Program for Climate Resilience (SPRC) identi- tion, water resources, energy, fisheries, forestry, fies water resources, agriculture, tourism, human health, and tourism and recreation (Houston et al. health, and human settlements as the most affect- 2004 and World Bank 2008). ed sectors in the Jamaican economy (Planning In- stitute of Jamaica 2011). The Jamaican Context. Jamaica’s location in the tropics makes its climate especially variable. This The benefits of improving a country’s hydromet sys- variability implies a greater need for, and larger tem depend both on the current state of the sys- benefits from, high quality climate forecasts. This tem and on the nature and magnitude of the im- observation is reinforced by Jamaica’s status as provements being evaluated. Evidence indicates one of the most susceptible countries to extreme that Jamaica’s hydromet system has deteriorated meteorological events. It is estimated to have the over time; the original network of 23 climatolog- second highest economic risk exposure to two or ical stations has dwindled to 6 functioning sta- more natural hazards: 96.3% of the national pop- tions and the 20-year old Doppler weather radar in ulation is exposed to two or more hazards, as is place is now obsolete and subject to periodic mal- 94.9% of the national territory and 96.3% of the functions. This suggests that there is a scope for country’s GDP (GFDRR 2010). The dominant natu- sizable benefits to be realized from even modest ral hazards are hurricanes, tropical storms, floods, investments in the hydromet system such as those and droughts. As revealed in Table A-5, extreme proposed in this project. Given the onerous debt 57 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems burden carried by the Jamaican government (cur- Forecasts are of value only if they are disseminat- rently at 140% of GDP), it is unlikely that these in- ed in a timely manner and in a format that users vestments will be made by the Jamaican govern- can understand and make use of. Users must also ment on its own. trust the accuracy of forecasts if they are to act on them (e.g., Stern and Easterling 1999, Chap- Overview of Analytical Approach ter 4; Blench 1999, World Bank 2008, Chapter 2). Estimates are developed for the three broad class- In the case of early warning systems, providing in- es of benefits described above: formation to users on how to best respond to the warnings is often also critical to the effectiveness benefits from improved forecasts, and associated of such systems, especially in poor communities early warning systems, of extreme meteorological (Subbiah et al. 2008; World Bank 2008, Chapter hazards; 2; Webster 2012). Without effective mechanisms for timely dissemination of understandable and us- i. benefits to enterprises of improved forecasts able forecasts, as well as information on appropri- of routine climate and dissemination of these ate responses, the potential benefits associated forecasts; and with upgrading hydromet systems will not be fully ii. benefits to households of improved forecasts realized. The benefits estimates presented here as- of routine climate and dissemination of these sume that these mechanisms are in place, or are forecasts; and put in place, as indicated in the objectives of the iii. particular emphasis is placed on deriving esti- Project Concept Note. mates of the first class of benefits, given Jamai- ca’s high susceptibility to extreme meteorolog- Improved Forecasts of Extreme Meteorological ical hazards. Hazards and Early Warning Systems. The ben- efits associated with extreme meteorological haz- Estimates of all three classes of benefits are de- ards are measured in terms of the expected reduc- rived using benefits transfer given the absence tion in economic losses resulting from improved of benefits studies for Jamaica itself. Because of forecasts and associated early warning systems. uncertainty about the suitability of the estimates Data on losses in Jamaica from extreme hazards transferred, the values of these estimates are var- during this century, broken down by sector, are used ied as part of a sensitivity analysis. The uncertain- to derive projections of future expected losses in ty about the estimates stems from the factors list- the without-project scenario.15 The anticipated eco- ed above.14 Throughout the analysis, a concerted nomic losses due to increase in the frequency of effort is made to be conservative when estimating natural hazards caused by climate change are esti- benefits, especially those benefits that are subject mated using predictions developed by the Caribbe- to a high degree of uncertainty, and to avoid double an Climate Risk Insurance Facility (2010) for Jamai- counting of benefits. ca. These predictions are aggregate losses from 14  This is particularly true given the absence of precise, quantitative information on the improvements in forecast quality and dissemination that would be generated by the project. 15  An alternative approach to estimating expected future losses would be to make use of information on return frequencies of different types of ex- treme meteorological events, together with estimates of the losses that each event would generate. Due to the difficulty in estimating the losses caused by each of a large set of possible extreme events, this approach was not pursued. A further difficulty with this approach is estimating, with any degree of accuracy, the return frequencies of extreme events for a country as small in area as Jamaica. 58 natural hazards (specifically winds and floods) be- Improved Forecasts of Routine Climate. To es- tween 2009 and 2030 for various climate change timate the benefits to Jamaican enterprises of im- scenarios. proved forecasts of routine climate (the second class of benefits listed above), this study relies Reductions in expected losses with the project on benefits transfer, making use of available es- are based on estimates of percentage loss avoid- timates in the literature. Useful, transferable esti- ed drawn from existing studies conducted else- mates are only available for a very small number of where that examine the benefits of improved fore- sectors. Accordingly, the estimates for this class of casts and associated early warning systems. To benefits are very conservative—they capture only a the extent possible, the estimates of percentage small subset of the benefits to all climate-sensitive loss avoided used in this analysis are drawn from sectors of the Jamaican economy. studies conducted in countries of similar profile as Jamaica; hazard susceptibility and geogra- The benefits to Jamaican households of improved phies and economies. This study assumes that forecasts of routine climate are based on the find- the percentage losses avoided remain constant ings of a detailed survey of U.S. households’ will- over time, due to lack of information on how they ingness to pay for improved forecasts (Lazo and might change. Chestnut 2002). As discussed in Section 5 (un- der Stated Preference Methods), Lazo and Chest- The approach taken to estimate reductions in ex- nut (2002) used a stated value approach to esti- pected losses is similar to that employed in previ- mate the value to U.S. households of improving the ous economic analyses of investments in hydromet quality of one-day and multi-day weather forecasts systems, however, with some important differenc- from their then-current levels to maximum possible es. Unlike the situation faced in some countries, levels. They find that, on average, U.S. households fairly detailed estimates of economic losses from were willing to pay USD 12 to USD 17 per year to extreme meteorological hazards are available for improve the quality of the forecasts, with a best es- Jamaica, and these loss estimates are central to timate of USD 16. this analysis.16 Because of the absence of percent- age loss avoided estimates for Jamaica, this study The willingness-to-pay estimates derived from these relies on benefits transfer to derive estimates of so-called contingent valuation studies are sensitive the reduction in expected losses with the project. to income levels. They are also sensitive to individ- These benefits estimates presented here could uals’ preferences, which can plausibly differ in a be refined by collecting information from Jamai- systematic manner across countries and cultures. can sector experts on the likely percentage of loss- Therefore, to transfer Lazo and Chestnut’s benefit es that would be avoided with the project. Such estimate to Jamaica, the USD 16 per household es- expert-opinion-based estimates have been devel- timate is first adjusted for differences in per capi- oped for evaluations of other, much larger, invest- ta income between the U.S. and Jamaica, and then ments in hydromet systems (e.g., World Bank 2008 expressed in constant (2010) U.S. dollars, yielding and Rogers et al. 2009, 2010). a value of USD 2.066. Absent information about In contrast, the benchmarking method employed to evaluate hydromet investments in countries of the former Soviet Union (World Bank 2008, 16  Rogers et al. 2009, 2010) derives estimates of losses for some sectors of the economy using available data on the weather-sensitivity of those sectors, the economic structure of the country, and the state of the country’s hydromet systems. 59 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems the precise nature and magnitude of the improve- • alternative assumptions about adaptation by ments in forecasts that would be achieved by the households and enterprises that reduces dam- Jamaica investment being evaluated, or about the ages from more frequent, or more intense, ex- nature of differences in household preferences for treme meteorological hazards even without the such improvements between the U.S. and Jamaica, project. it is conservatively assumed that the value to Ja- maican households of the improvement achieved by the project is one-half of this value, or USD 1.033 Estimated Benefits of Improved Forecasts of per year. Multiplying by the 881,078 households in Extreme Hazards and Early Warning Systems Jamaica yields a total annual benefit estimate of The method used to estimate the benefits of im- USD 910,154. Given projections of population and proved forecasts of extreme meteorological haz- income growth over time, this estimate can be ad- ards and associated early warning systems consists justed to reflect changes in population and income of several steps. First, the average annual econom- over the time horizon being considered. ic loss from extreme meteorological hazards in Ja- maica over the period 2000–2010 is calculated by The benefits of improved forecasts of routine cli- sector. These average losses form the basis for pre- mate are likely to increase as climate becomes dictions of expected losses from extreme events in more variable and as the number of enterprises future years given alternative assumptions about and households rises. However, there has been the severity of climate change. The percentage of no attempt yet in the literature to quantify this in- losses that would be avoided with the project in crease in benefits resulting from more variable cli- each sector is then estimated using benefits trans- mate. Absent estimates of this potential increase fer from existing studies. Expected annual benefits and projections 20 years out of per capita income are then calculated by multiplying expected annual and the number of households in Jamaica, this sector losses for each year by the estimate of per- study assumes that the benefits of improved rou- centage loss avoided for that sector. These sector tine forecasts do not increase in magnitude over benefits are aggregated for each year. Finally, the time. This is a conservative assumption that under- discounted present value of annual benefits is cal- estimates benefits. culated for 5-, 10-, 15-, and 20-year time horizons. To assess the robustness of the conclusions drawn In previous studies of improving hydromet systems, from this analysis, a variety of sensitivity analyses losses from meteorological events were not avail- are conducted using: able, so total damages were used (World Bank 2008, Rogers et al. 2009, 2010). However, since • alternative values of the transferred benefits sector losses are available for Jamaica for many estimates; recent meteorological events, these values were • alternative time horizons (5, 10, 15 and 20 used to improve the expected damage and bene- years); fits estimates. By disaggregating expected damag- • alternative real discount rates (4%, 10% and es by sector, it is possible to obtain a more accu- 12%); rate estimate of total expected benefits compared • alternative assumptions about the increase in to a case where a single, damage mitigation mul- expected losses from extreme meteorological tiplier were applied. Every sector in the economy hazards over time due to climate change; and is affected by meteorological hazards in a different 60 way, depending on the type of meteorological event, Extreme hazards are separated into four categories: sector’s exposure to weather, and value of its as- hurricanes, tropical storms, floods, and droughts. sets. In addition, this study estimates incremental Table A-5 (above) lists the 12 extreme meteorolog- benefits, rather than total benefits, and thus must ical events that occurred in Jamaica over this pe- take into consideration each sector’s current abil- riod. Accurate estimates of losses do not exist for ity to mitigate damages, and how effective an im- smaller events that also caused damages during provement in the hydromet early warning system this time period, and they are therefore excluded will be. For instance, a hotel in Jamaica’s bus- from this study. The loss estimates in this study are tling tourism industry is likely to be a sturdy struc- therefore conservative. ture that takes little damage from hurricane force winds, regardless of how much advanced warning The losses (or damages) from each event are de- is given. In contrast, a home of a Jamaican farmer composed into 10 sectors (see Table A-6). Both di- can be much more vulnerable to strong winds and rect losses, which are losses of physical and natu- rain, and a small increase in the amount of prepa- ral capital, as well as indirect losses, which consist ration time for the farmer to fortify his property and mainly of losses of profits, are included. Where move his possessions could result in a significant available, these decompositions are taken from ex- decrease in damages. isting damage assessments. For events for which damage decompositions are not available, as is In order to calculate average annual economic loss- the case for Hurricane Charley (2004) and the Jan- es from extreme meteorological hazards, this anal- uary, 2001 flooding event, this study computes av- ysis considers an 11-year period from 2000–2010. erage damage percentages by sector and hazard Table A-6  Average Damages from Hurricanes by Sector (USD Million, Constant 2010) Dennis & Percent of Sector Michelle Ivan Emily Wilma Dean Damages Agriculture, Livestock, & 32.33 286.42 22.15 7.11 204.90 25.6% Fisheries Emergency Operations 0.07 9.30 0.73 0.07 13.15 1.1% Environment - 125.77 1.97 - 2.62 6.0% Government & 1.24 27.03 0.03 91.39 17.56 6.4% Institutions Health 0.85 25.40 1.58 1.29 6.51 1.6% Housing 13.46 373.98 5.82 1.03 130.02 24.3% Industry - 108.37 1.39 - 44.27 7.1% Tourism 2.00 53.29 0.07 - 0.95 2.6% Transportation 94.21 109.21 122.37 - 44.65 17.1% Utilities 6.53 121.00 14.63 3.03 30.38 8.1% Total 150.67 1,239.77 170.75 103.92 495.01 100.0% Source: Planning Institute of Jamaica (2004, 2005a, 2005b, 2007, 2008, 2010 and 2011), Economic Commission for Latin America and the Caribbean (2001 and 2002), Government of Jamaica Ministry of Agriculture & Fisheries (2010), and United Nations and IDB (2007). Note: Damage totals here may be slightly different from totals in Table A-5 due to missing sectors and small discrepancies in totals be- tween different sources. 61 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems type for events for which damage decompositions the project are estimated by applying estimates of are available, and then applies these average per- the percentage loss avoided (PLA) with the proj- centages to events for which damage decomposi- ect. The percentage loss avoided will generally vary tions are not available. For example, in the case of across sectors and hazard types. Because of this Hurricane Charley, the average percentage of dam- heterogeneity, a separate PLA estimate is applied ages experienced by each sector for all other hur- to each sector for each hazard type. ricanes over the period 2000–2010 is calculated. These percentages are then used to decompose to- As noted earlier, given the lack of Jamaica-specif- tal damages from Hurricane Charley. Table A-6 il- ic information, the analysis relies on PLA estimates lustrates this procedure. An analogous procedure derived in similar studies of other regions. The es- was used for the other hazard types. timates used are drawn from a variety of studies as indicated in Table A-7 below. If only one PLA es- Due to the nature of extreme meteorological haz- timate is available for a given sector/hazard, this ards, it is impossible to reliably forecast their num- value is treated as the high value. The low value is ber and severity years into the future. To derive es- taken to be 50% of the high value. Where multiple timates of expected losses from such hazards in PLA estimates are available, the lowest value is tak- the future, it is assumed that absent changes in cli- en to be the low value, and the highest value is tak- mate or the values of assets at risk, expected an- en to be the high value. The middle value is the av- nual losses from extreme hazards in future years erage of the two values. are equal in magnitude to average annual losses over the period 2000–2010. These “baseline loss Given the scope of this study and limitations of the estimates” are then adjusted to reflect the likely in- existing literature, for some sectors/hazards it was crease in the frequency and intensity of extreme necessary to use PLA estimates from countries that meteorological hazards due to climate change, and are dissimilar to Jamaica. Specifically, for some the expected increase in the values of assets at sectors/hazards, estimates developed as part of risk. The adjustments are made using estimates an evaluation of a hydromet system upgrade in Ta- developed by the Caribbean Catastrophe Risk In- jikistan (Rogers et al. 2009) are used. The PLA val- surance Facility (CCRIF). ues taken from this study are reduced by 50% to be conservative and to acknowledge differences Using 2009 as a base year, CCRIF estimates that in the climate, geography, economic structure, and by 2030, economic losses in Jamaica from ex- hydromet systems of the two countries. treme hazards will increase by 28%, 46%, or 76%, given no changes in climate, “moderate” changes, Multiplying the PLA estimates by the expected an- or “high” changes, respectively (CCRIF 2011). As- nual losses for each sector/hazard yields esti- suming losses increase linearly, this translates to mates of the expected annual benefits associated loss increases of 1.3%, 2.2%, or 3.6% of the base- with improving forecasts of extreme hazards and line, per year. These percentages are applied to the early warning systems. However, these estimates baseline loss estimates described above to obtain implicitly assume that forecasts are perfectly ac- estimates of expected future losses. curate. In reality, forecasts are subject to uncer- tainty, especially those made in the tropics. It is After computing expected annual losses by sector, therefore important to acknowledge the likelihood for each hazard type, the reduction in losses with of incorrect forecasts and the costs associated 62 Table A-7  Estimates of Percentage Losses Avoided with Project Hurricane/Tropical Storm Flood Drought Sector Low Middle High Low Middle High Low Middle High Ag., Livestock, & 13.0% d 19.5% 26.0% a 10.0% a 40.0% 70.0% a 1.8% d 2.7% 3.6%c Fisheries Emergency 1.8%d 2.7% 3.6%c 1.8%d 2.7% 3.6%c Not Applicable Operations Environment 1.8%d 2.7% 3.6%c Not Applicable Data Not Available Government & 7.5%d 11.3% 15.0%a 5.0%a 10.0% 15.0%a Not Applicable Institutions Health 1.8%d 2.7% 3.6%c 1.8%d 2.7% 3.6%c Not Applicable Housing 5.0%b 7.5% 10.0%b 2.0%a 19.0% 36.0%a Not Applicable Industry 1.8%d 2.7% 3.6%c 1.8%d 2.7% 3.6%c Data Not Available Tourism 1.8%d 2.7% 3.6%c 1.8%d 2.7% 3.6%c Data Not Available Transportation 3.0%d 4.5% 6.0%c 0.0%a 6.9% 13.7%b Not Applicable Utilities 2.1%d 3.2% 4.3%c 0.0%a 6.9% 13.7%b Data Not Available a Subbiah et al. (2008). b Bangladesh Water Development Board (2006). c Rogers et al. (2009). d Calculated as 50% of the high value. with these forecasts (e.g., the costs of evacuating improved forecasts are only enjoyed 80% of the people unnecessarily).17 Following Subbiah et al. time, while costs of incorrect forecasts are in- (2008), it is assumed that the cost of an incorrect curred 20% of the time. forecast is equal in magnitude to the benefit of a correct forecast.18 Assuming that average fore- Table A-8 presents estimates of the resulting ex- cast accuracy is 80%, this implies that estimates pected annual benefits for the year 2014 (the first of expected annual benefits need to be multiplied year of the project) assuming, hypothetically, that by a correction factor of 0.6 (= 80% - 20%). This potential annual benefits are fully realized in that correction factor reflects the fact that benefits of year. 17  This correction is appropriate given that there is no evidence that the PLA estimates drawn from the literature account for forecast error. 18  This assumption is made given the absence of information on the costs of incorrect forecasts. 63 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems Expected Annual Benefits of Improved Forecasts of Extreme Hazards and Early Table A-8   Warning Systems, Baseline Year (USD Million, Constant 2010) Hurricanes Tropical Storms Floods Droughts Total Low Value 7.17 1.06 0.24 0.03 8.49 Middle Value 10.75 1.59 1.28 0.04 13.66 High Value 14.34 2.12 2.33 0.05 18.84 Source: Authors.In reality, it is unlikely that potential annual benefits will be fully realized in the first year of the project. Absent further information, it is assumed that 20% of potential annual benefits are realized at the end of the first year of the project, and an addition- al 20% of benefits are realized at the end of each of the four subsequent years. From 2018 onwards, 100% of potential annual benefits are realized in each year. Note: These estimates incorporate a correction factor that reflects the costs of inaccurate forecasts. The estimates in Table A-8 assume that as a result cross-section of countries. The assumption guards of experiencing more, or more intense, extreme against potentially overstating the benefits of the meteorological hazards, Jamaican households and proposed project as a result of ignoring the conse- enterprises do not adapt to the hazards, reducing quences of adaption measures that are undertak- the damages they cause. For example, households en even in the absence of the project. can respond to more extreme hurricanes by living in dwellings made of concrete rather than wood. It is unclear to what extent Jamaican households Estimated Benefits of Improved Forecasts of Routine and enterprises have already undertaken such ad- Climate aptation measures, and to what extent these mea- Benefits to Enterprises. A large number of stud- sures would increase over the next 5–20 years (the ies have been conducted examining the value of time horizon of this analysis). The available empir- climate forecasts to enterprises in various cli- ical literature is quite silent on this issue. A recent mate-sensitive sectors of the economy.19 A small study of adaptation to tropical cyclones using data subset of these studies examines the value of im- from across the world (Hsiang and Narita 2012) in- proved forecasts. The results of these studies are dicates that only about 3% of the losses from in- typically very setting specific and provide estimates cremental changes of countries’ current tropical that are difficult to transfer to other settings. How- cyclone climates are “adapted away” in the long ever, a few studies do provide estimates that can run. As part of the sensitivity analyses conducted be transferred with some degree of confidence. here, a scenario is considered that assumes that But these studies cover only a very small number damages in each year are reduced by 10% due to of climate-sensitive sectors. Specifically, Adams et such adaptation measures. (Accordingly, the ben- al. (2003) report increases of 0.3–2% in the val- efits estimates presented in Table A-8 would each ue of crops produced in five Mexican states as a be reduced by 10%). The assumption that damag- result of improved ENSO forecasts. Houston et al. es in all sectors are reduced by 10 % in each year (2004) and Yeh et al. (1982) report increases of due to adaptation is a liberal one given the afore- 0.04–0.1% in GDP from the U.S. electricity, gas and mentioned long-run estimate of 3% for a global sanitary services sector from improved short-term 19  See Houston et al. (2004) and Weiher et al. (2005) for a compendium of estimates. 64 Expected Annual Benefits of Improved Forecasts of Routine Climate for Selected Table A-9  Sectors (USD Million, Constant 2010) Low Value Middle Value High Value Agriculture, Forestry & Fishing 1.94 7.43 12.93 Electricity and Water Supply 0.19 0.33 0.46 Source: Authors. Note: Estimates were calculated only for sectors for which benefits transfer was feasible. forecasts of temperature and long-term forecasts is a very detailed 2002 study by Lazo and Chest- of precipitation.20 nut that estimates the value to U.S. households of improving the quality of one-day and multi-day Table A-9 presents the results of applying these weather forecasts from their then-current levels estimates to the corresponding sectors of the Ja- to maximum possible levels.22 Lazo and Chestnut maican economy. The table shows the benefits in find that, on average, U.S. households were will- these sectors for the year 2014 assuming, hypo- ing to pay USD 12 to USD 17 per year to improve thetically, that potential annual benefits are ful- the quality of the forecasts, with a best estimate ly realized in this year. When incorporating these of USD 16. To transfer this benefit estimate to Ja- benefits in the calculation of project benefit-cost maica, the USD 16 per household estimate is first ratios, it is assumed that the benefits are realized adjusted for differences in per capita income be- in 20-percentage-point increments over the first tween the U.S. and Jamaica, and then expressed in five years of the project (as is done for the benefits 2010 US dollars. Absent information about the pre- associated with extreme hazards). It bears empha- cise nature and magnitude of the improvements in sis that these estimates capture only a fraction of forecasts that would be achieved by the project be- the benefits to enterprises of improved forecasts ing evaluated, or about the nature of differences of routine climate given the large number of sec- in household preferences for such improvements tors omitted. between the U.S. and Jamaica, it is conservative- ly assumed that the value to Jamaican households Benefits to Households. A number of survey-based of the improvement achieved by the project is one- studies have estimated the benefits to households half of this value, or USD 1.033 per year.23 Multiply- of weather forecasts.21 Far less common are stud- ing by the 881,078 households in Jamaica (Statisti- ies that examine the benefits to households of im- cal Institute of Jamaica 2011) yields a total annual provements in forecast quality, which is the rele- benefit estimate of USD 910,154. As with the bene- vant measure for this analysis. A notable exception fits to enterprises, these benefits to households are 20  These values are in line with values employed by Hallegate (2012) to estimate the benefits of improved hydromet services in developing coun- tries. He assumes value added gains between 0.1% and 1% in weather-sensitive. 21  See Lazo et al. (2009) and Houston et al. (2004) for surveys of these studies, which focus on developed countries, and World Bank (2008) for a study conducted in Azerbaijan and Serbia. 22  The accuracy of one-day forecasts was assumed to increase from 80% to 95%; for multiday forecasts, 14-day forecasts were assumed to be- come as accurate as current 5-day forecasts. The study focused on valuing improved forecasts of day-to-day weather rather than extreme events, hence double counting of benefits should not be an issue. 23  This value is plausible given the results of studies conducted in Azerbaijan and Serbia, countries with GDP per capita similar to Jamaica’s, in- dicating that households are willing to pay USD 12–16 per year to support national hydromet services—a total, rather than incremental, benefit measure (World Bank 2008, p 71). 65 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems assumed to be realized in 20-percentage-point in- horizons being considered in the context of climate crements over the five years of the project. change. Estimated Costs of the Project The benefit-cost ratios for the “middle” and “high” The total cost of implementing the project is benefits estimates may seem unusually large, how- USD 6.8MM. It is conservatively assumed that this ever previous studies of the returns to investments entire cost is borne at the beginning of 2014. An- in hydromet systems have, in a number of cases, nual costs of maintaining and operating the im- yielded very large benefit-cost ratios. Subbiah et al. proved hydro-meteorological system are not antici- (2008) report a benefit-cost ratio of 40 for devel- pated to increase significantly. The new RADAR will opment of a cyclone early warning system in Ban- eliminate the relatively high maintenance costs as- gladesh, and benefit-cost ratios of 0.9 to 558 for sociated with the current 20-year old system, and flood early warning systems in Sri Lanka and Ban- the installation of automated monitoring and infor- gladesh, respectively. An evaluation of post-civil- mation transmission systems will free up staff re- war investment in Mozambique’s hydromet system sources for other purposes. estimated a benefit-cost ratio of 70 (World Bank 2008, p. 5). The large benefit-cost ratios obtained in this study reflect, in large part, Jamaica’s high Benefit-Cost Ratios for the Project susceptibility to extreme meteorological hazards. Given the above estimates and assumptions, ben- efit/cost ratios for the project are calculated for 5-, The benefit-cost ratios presented in Tables A-10a-c 10-, 15-, and 20-year time horizons, using discount assume that Jamaican households and enterprises rates of 4%, 10%, and 12%. These benefit-cost ra- do not adapt to more frequent or more intense ex- tios are presented in Tables A-10a-10c for the three treme meteorological hazards, reducing the dam- different climate change scenarios, and for the low, ages they cause in the future, even in the absence middle and high estimates of benefits. of the proposed project. As discussed in paragraph 36, to evaluate the sensitivity of the benefit-cost Examining the tables it can be seen that the ben- ratios to such adaptation, an alternative scenar- efit-cost ratio is well above one in all cases. For io is considered in which the estimated expected the most conservative case—no climate change, damages from extreme meteorological events are “low” benefits estimates, 5-year time horizon and reduced by 10% in each year as a result of adap- 12% discount rate—the benefit-cost ratio is 3.5 tation. The benefit-cost ratios for this alternative (left-hand panel of Table A-10a). The ratio rises scenario are presented in Tables A-11a-c. As can to 10.7 when the project time horizon is extend- be seen, the ratios are only reduced by a modest ed to 20 years, with all else remaining constant. amount, and each is still well above one. For the For the “middle” benefits estimates (middle pan- most conservative case considered in paragraph el of Table A-10a), the benefit-cost ratios are near- 42, the benefit-cost ratios diminish from 3.5 to 3.2 ly twice as high, and for the “high” benefits esti- and from 10.7 to 9.9. mates they are approximately two-and-a-half times higher. Comparing the values in Tables 7a, 7b, and 7c reveals modest variations in benefit-cost ra- Additional Observations and Caveats tios across the three climate change scenarios. The analysis presented is best viewed as an effort This can be attributed to the relatively short time to develop first-cut estimates of the benefits and 66 costs of the proposed project. A number of observa- analysis assumes that these avoidance (or preven- tions about the estimates and about the appropri- tion) costs are netted out of the percentage loss ate time horizon for the analysis are worth noting. avoided estimates that are employed. It is unclear from the literature whether this is always true. In Given the nature of the equipment being installed some cases the costs may be small enough, in rel- as part of the project, a 20-year time horizon may ative terms, that omitting them is inconsequential. well be too long. The current 20-year-old RADAR system is considered outdated and unreliable, and As has been emphasized, only a subset of the ben- this will likely be true of the new RADAR system 20 efits of the project has been quantified. This is es- years hence. The 10- and 15-year horizons are like- pecially true of the benefits to enterprises of im- ly to be more reasonable. proved forecasts of routine climate. The benefits are only quantified for two sector aggregates (ag- The incremental costs, both capital and recurring, riculture, forestry and fishing, and electricity and of upgrading the hydromet system itself are likely to water supply), but there are other climate-sensitive be adequately captured by the project implementa- sectors of the Jamaican economy that are likely to tion costs. As noted above, the new equipment will reap significant benefits from improved forecasts. likely result in reduced maintenance costs and in staff resources being freed up for other purposes. One class of benefits that has not been mentioned is a reduction in deaths as a result of improved However, the costs of ensuring that the benefits of warnings of extreme events. Given the relatively improved forecasts are fully realized may not be small number of deaths due to extreme meteoro- adequately captured in the project implementation logical events in Jamaica, benefits of this type are costs. For the benefits to be fully realized, citizens, likely to be relatively small.24 enterprises, civil society organizations, and govern- ment agencies must be effectively informed about An important component of the project is develop- the forecasts in a timely manner. No less impor- ing higher-resolution climate change scenarios for tantly, they must be educated on how best to in- Jamaica in order to enhance climate-resilient plan- terpret and make use of the improved forecasts. ning and decision-making. The benefits associated This is not an inconsequential task, especially giv- with this effort are very difficult to quantify, but they en that forecasts are imperfect. are likely to be substantial. It is not implausible to posit that the unquantified benefits of the project Another cost that is only implicitly considered in outweigh any unquantified costs of the project. the analysis is the cost of additional measures tak- en to reduce losses from extreme meteorological Key Conclusions events given improved forecasts. Examples are the The first-cut estimates of the benefits and costs of costs of building additional emergency shelters or the proposed project indicate that in all of the al- the costs borne by homeowners of protecting their ternative scenarios considered, the benefits of the dwellings given earlier warnings of storms. The project exceed the costs by a considerable margin. 24  The largest number of deaths that have been reported in this century are for Hurricane Ivan, with 17 direct deaths and 14 indirect deaths (Plan- ning Institute of Jamaica 2004). 67 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems Even with the most conservative set of assump- the project will be made by the Jamaican govern- tions, including an assumption that no climate ment on its own. change takes place and that the benefits of the project only last for 5 years, the benefit-cost ra- Extending the period over which the benefits of the tio for the project is 3.2, i.e., the expected bene- project accrue from 5 years to a more plausible 10 fits from the project are 3.2 times higher than the years or 15 years, increases the benefit-cost ratio expected costs of USD 6.88 MM (in present value to 6.6 and 8.7, respectively. Assuming no climate terms). change takes place and otherwise maintaining the most conservative set of assumptions. The benefits of the project are experienced by a very broad swath of the Jamaican economy, Thus, the estimates derived indicate that the proj- and stem in large part from a reduction in ex- ect is economically desirable even in the absence pected damages from extreme meteorological of climate change. Acknowledging climate change, hazards, such as storms, hurricanes and floods. and the concomitant increase in the frequency or Given the onerous debt burden carried by the Ja- severity of extreme meteorological events, simply maican government (currently at 140% of GDP), increases the benefit-cost ratios for the project, in- it is unlikely that the investments embodied in creasing its estimated economic desirability. Project Benefit-Cost Ratios Assuming No Climate Change for Alternative Benefits Table A-10a  Estimates Low Discount Rate Middle Discount Rate High Discount Rate Time Horizon 4% 10% 12% Time Horizon 4% 10% 12% Time Horizon 4% 10% 12% 5 Year 4.5 3.7 3.5 5 Year 8.7 7.1 6.7 5 Year 12.9 10.6 9.9 10 Year 11.1 8.0 7.2 10 Year 21.4 15.3 13.8 10 Year 31.7 22.6 20.4 15 Year 16.8 10.7 9.4 15 Year 32.3 20.6 18.0 15 Year 47.7 30.4 26.6 20 Year 21.7 12.5 10.7 20 Year 41.5 24.0 20.4 20 Year 61.3 35.4 30.2 Source: Authors. Project Benefit-Cost Ratios Assuming Moderate Climate Change for Alternative Table A-10b  Benefits Estimates Low Discount Rate Middle Discount Rate High Discount Rate Time Horizon 4% 10% 12% Time Horizon 4% 10% 12% Time Horizon 4% 10% 12% 5 Year 4.6 3.8 3.5 5 Year 8.8 7.2 6.8 5 Year 13.1 10.7 10.1 10 Year 11.5 8.2 7.4 10 Year 22.0 15.7 14.1 10 Year 32.5 23.2 20.9 15 Year 17.6 11.1 9.7 15 Year 33.4 21.2 18.5 15 Year 49.3 31.3 27.4 20 Year 22.9 13.1 11.1 20 Year 43.4 24.9 21.2 20 Year 63.9 36.7 31.2 Source: Authors. 68 Project Benefit-Cost Ratios Assuming High Climate Change for Alternative Benefits Table A10-c  Estimates Low Discount Rate Middle Discount Rate High Discount Rate Time Horizon 4% 10% 12% Time Horizon 4% 10% 12% Time Horizon 4% 10% 12% 5 Year 4.7 3.9 3.6 5 Year 9.0 7.4 6.9 5 Year 13.4 10.9 10.2 10 Year 12.1 8.6 7.7 10 Year 22.9 16.3 14.6 10 Year 33.7 24.0 21.6 15 Year 18.8 11.8 10.3 15 Year 35.3 22.3 19.4 15 Year 51.9 32.8 28.6 20 Year 24.8 14.0 11.9 20 Year 46.5 26.4 22.4 20 Year 68.1 38.8 32.9 Source: Authors. Project Benefit-Cost Ratios Assuming No Climate Change and Reduced Damage Due Table A-11a  to Mitigation for Alternative Benefits Estimates Low Discount Rate Middle Discount Rate High Discount Rate Time Horizon 4% 10% 12% Time Horizon 4% 10% 12% Time Horizon 4% 10% 12% 5 Year 4.2 3.4 3.2 5 Year 8.2 6.7 6.3 5 Year 12.2 10.0 9.4 10 Year 10.3 7.4 6.6 10 Year 20.1 14.3 12.9 10 Year 29.8 21.3 19.2 15 Year 15.6 9.9 8.7 15 Year 30.2 19.3 16.8 15 Year 44.9 28.6 25.0 20 Year 20.1 11.6 9.9 20 Year 38.9 22.4 19.1 20 Year 57.6 33.3 28.4 Source: Authors. Project Benefit-Cost Ratios Assuming Moderate Climate Change and Reduced Table A-11b  Damage Due to Mitigation for Alternative Benefits Estimates Low Discount Rate Middle Discount Rate High Discount Rate Time Horizon 4% 10% 12% Time Horizon 4% 10% 12% Time Horizon 4% 10% 12% 5 Year 4.3 3.5 3.3 5 Year 8.3 6.8 6.4 5 Year 12.3 10.1 9.5 10 Year 10.6 7.6 6.8 10 Year 20.6 14.7 13.2 10 Year 30.5 21.8 19.6 15 Year 16.2 10.3 9.0 15 Year 31.3 19.9 17.3 15 Year 46.3 29.5 25.7 20 Year 21.2 12.1 10.3 20 Year 40.6 23.3 19.8 20 Year 60.0 34.5 29.3 Source: Authors. Project Benefit-Cost Ratios Assuming High Climate Change and Reduced Damage Table A-11c  Due to Mitigation for Alternative Benefits Estimates Low Discount Rate Middle Discount Rate High Discount Rate Time Horizon 4% 10% 12% Time Horizon 4% 10% 12% Time Horizon 4% 10% 12% 5 Year 4.4 3.6 3.3 5 Year 8.5 6.9 6.5 5 Year 12.6 10.3 9.6 10 Year 11.1 7.9 7.1 10 Year 21.4 15.2 13.7 10 Year 31.7 22.5 20.3 15 Year 17.3 10.9 9.5 15 Year 33.0 20.8 18.2 15 Year 48.7 30.8 26.8 20 Year 22.9 12.9 11.0 20 Year 43.3 24. 20.9 20 Year 63.8 36.3 30.9 Source: Authors. 69 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems Natural Resource Perspectives, No. 47. London: References Overseas Development Institute. Adams, Richard M., Kelly J. Bryant, Bruce A. McCa- Carsell, Kim M., Nathan D. Pingel, and David T. rl, David M. Legler, James O’Brien, Andrew Solow, Ford. 2004. Quantifying the Benefit of a Flood and Rodney Weiher. 1995. “Value of Improved Warning System. Natural Hazards Review, August, Long-Range Weather Information.” Contemporary pp. 131–140. Economic Policy 13(3):10–19. Carriquiry, Miguel A., and Daniel E. Osgood. 2012. Adams, Richard M., Laurie L. Houston, Bruce A. Mc- “Index Insurance, Probabilistic Climate Forecasts, Carl, Mario L. Tiscareno, Jaime Matus, and Rodney and Production.” Journal of Risk and Insurance F. Weiher. 2003. “The Benefits to Mexican Agricul- 79(1): 287–300. ture of an El Niño-southern Oscillation (ENSO) Ear- ly Warning System.” Agricultural and Forest Meteo- Diamond, Peter A., and Jerry A. Hausman. 1994. rology 115(3–4): 183–94. “Contingent valuation: Is some number better than no number?” The Journal of Economic Perspec- Advisors to the High Level Group on Infrastructure tives 8(4): 45–64. Charging (AHLG). 1999. Calculating Transport Acci- dent Costs. April 27 1999. Dutton, J. 2002. Opportunities and Priorities in a New Era for Weather and Climate Services. Ameri- American Meteorological Society. 2011. Strategic can Meteorological Society. Implementation Plan for Generating and Communi- cating Forecast Uncertainty Information. Economic Commission for Latin America and the Caribbean. Jamaica: Assessment of the Damage Anaman, K. A., Lellyett, S. C., Drake, L., Leigh, R. J., Caused by Flood Rains and Landslides in Associa- Henderson-Sellers, A., Noar, P. F., Sullivan, P. J., and tion with Hurricane Michelle, October 2001. 7 Dec. Thampapillai, D. J. “Benefits of meteorological ser- 2001. vices: evidence from recent research in Australia.” Meteorological Applications, 5: 103–115, 1998. Economic Commission for Latin America and the Caribbean. Jamaica: Macro-socio-economic As- ASEAN. 2012. Advancing Disaster Risk Financing sessment of the Damage Done by Flood Rains and and Insurance in ASEAN Member States: Frame- Landslides May 2002. 22 July 2002. work and Options for Implementation, GFDRR, ISDR, Volume 2: Technical Appendices. Energy Institute. 2012. Energy Efficiency Policy and Measures in Croatia, Energy Institute Hrvoje Pozar, Bangladesh Water Development Board. 2006. Ban- Zagreb, available at: http://www.odyssee-indica- gladesh: Early Warning Systems Study. Ministry of tors.org/publications/PDF/croatia_nr.pdf. Water Resources, Government of the People’s Re- public of Bangladesh. Freebairn, J.W. and Zillman, J.W. “Economic bene- fits of meteorological services.” Meteorological Ap- Blench, R. 1999. Seasonal Climatic Forecasting: plications, 9(1): 33–44, 2002. Who Can Use It and How Should It Be Disseminated? 70 Global Facility for Disaster Risk Reduction (GFDRR). Jewson, Stephen, and Rodrigo Caballero. 2003. 2009. Improving Weather, Climate, and Hydrologi- The use of weather forecasts in the pricing of cal Services Delivery in Central Asia (Kyrgyz Repub- weather derivatives. Meteorological Applications lic, Republic of Tajikistan, and Turkmenistan). 10(4): 377–389. Global Facility for Disaster Risk Reduction (GFDRR). Katz, Richard W., Allan H. Murphy, and Robert L. 2010. Disaster Risk Management in Latin America Winkler. 1982. “Assessing the value of frost fore- and the Caribbean Region: GFDRR Country Notes, casts to orchardists: a dynamic decision-making Jamaica. approach.” Journal of Applied Meteorology 21(4): 518–531. Golnaraghi, M. 2009. The role of the WMO in nat- ural hazard awareness and disaster risk reduction Katz, Richard W. and Jeffrey K. Lazo. 2011. “Eco- education, presentation at the 2nd Conference of nomic Value of Weather and Climate Forecasts,” in the OECD International Network on the Financial The Oxford Handbook of Economic Forecasting, ed- Management of Large-scale Catastrophes, Sep- ited by Clements, Michael Peter, and David F. Hend- tember 24, Bangkok, Thailand. ry. Oxford University Press. Government of Jamaica Ministry of Agriculture & Katz, R.W. and A.H. Murphy. 2005. Economic Value Fisheries. 2010. Introducing Innovative Agriculture of Weather and Climate Forecasts. Cambridge Uni- Weather Risk Management Mechanisms for Small versity Press. Farmers. Pre-feasibility Assessment for St. Eliza- beth and Portland Parishes. 17 May 2010. Kokko, H., and Vaisala, O. 2005. Integrated Hydro- meteorological Monitoring Solutions and Network Hallegate, S. 2012. A Cost Effective Solution to Re- Management. In 21st International Conference duce Disaster Losses in Developing Countries. on Interactive Information Processing Systems (IIPS) for Meteorology, Oceanography, and Hydrol- Han, Xiaoli, and Bingsong Fang. 2000. Four Mea- ogy. Helsinki, Finland. sures of Transportation’s Economic Importance. Journal of Transportation and Statistics 3, no. 1: Lazo, Jeffrey K. and Lauraine G. Chestnut. 2002. 15–30. Economic Value of Current and Improved Weath- er Forecasts in the U.S. Household Sector. Report Houston, Laurie L., Richard M. Adams and Rodney prepared for Office of Policy and Strategic Planning, F. Weiher. 2004. The Economic Benefits of Weath- National Oceanic and Atmospheric Administration er Forecasts: Implications for Investments in Ruso- by Stratus Consulting Inc. hydromet Services. Report prepared for the World Bank under NOAA contract OG 1330-04-SE-0052. Lazo, Jeffrey K., Robert S. Raucher, Thomas J. Teis- berg, Carolyn J. Wagner, and Rodney F. Weiher. Hsiang, S.M. and D. Narita. 2012. “Adaptation to 2008. “Primer on Economics for National Meteoro- Cyclone Risk: Evidence from the Global Cross-Sec- logical and Hydrological Services.” tion.” Climate Change Economics, 3(2). Lazo, J.K., Morss, R.E., Demuth, J.L., 2009: 300 Bil- lion served: Sources, perceptions, uses, and values 71 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems of weather forecasts. Bulletin of the American Me- Mjelde, James. W., Steven T. Sonka and Derrell S. teorological Society, 90, 785–798. Peel. 1989. The Socioeconomic Value of Climate and Weather Forecasting: A Review. Research Re- Lazo, Jeffrey K. and Donald M. Waldman. 2011. port 89–01, Midwestern Climate Center, Climate & “Valuing Improved Hurricane Forecasts.” Econom- Meteorology Section, Illinois State Water Survey. ics Letters 111:43–46. Millner, A. 2009. “What is the true value of fore- Leviäkangas, Pekka, Raine Hautala, Jukka casts?.” Weather, Climate, and Society 1(1): 22–37. Räsänen, Risto Öörni, Sanna Sonninen, Martti Hek- kanen, Mikael Ohlström, Ari Venäläinen and Seppo Mills, B. 2010. “Societal and Economic Research Saku. 2007. Benefits of Meteorological Services in and Applications (SERA) for Adaptation to Atmo- Croatia VTT TIEDOTTEITA, 2420. spheric Hazards in Canada.” Macauley, Molly K. 1997. “Some Dimensions of the Morss, Rebecca E., Olga V. Wilhelmi, Mary W. Value of Weather Information: General Principles Downton, and Eve Gruntfest. 2005. “Flood risk, un- and a Taxonomy of Empirical Approaches,” Pro- certainty, and scientific information for decision ceedings of the Workshop on the Social and Eco- making: lessons from an interdisciplinary project.” nomic Impacts of Weather, Boulder, Colorado, April Bulletin of the American Meteorological Society 2–4, 1997. 86(11): 1593–1601. Malik, Arun S., and Stephen C. Smith. 2012. “Ad- Morss, Rebecca E., Jeffrey K. Lazo, Barbara G. aptation To Climate Change In Low-Income Coun- Brown, Harold E. Brooks, Philip T. Ganderton, and tries: Lessons From Current Research And Needs Brian N. Mills. 2008. “Societal and economic re- From Future Research.” Climate Change Econom- search and applications for weather forecasts.” ics 3(2). Bulletin of the American Meteorological Society 89(3): 335–346. Maurer, Edwin P., and Dennis P. Lettenmaier. 2004. “Potential Effects of Long-Lead Hydrologic Predict- National Research Council. 1998. The Atmospher- ability on Missouri River Main-Stem Reservoirs.” ic Sciences Entering the Twenty-First Century. Na- Journal of Climate 17(1): 174–186. tional Academy Press. Meyer, R. 2006. Why we Under Prepare for Haz- National Research Council 2006. Completing the ards. In On risk and disaster: Lessons from hurri- Forecast: Characterizing and Communicating Un- cane Katrina, edited by R.J. Daniels et al. Universi- certainty for Better Decisions Using Weather and ty of Pennsylvania Press, pp. 153–173. Climate Forecasts. National Research Council, Washington, D.C. Mjelde, James W., Steven T. Sonka, Bruce L. Dixon, and Peter J. Lamb. 1988. “Valuing forecast charac- Nichols, J.M. 1996. “Economic and social benefits teristics in a dynamic agricultural production sys- of climatological information and services: a review tem.” American Journal of Agricultural Economics of existing assessments.” World Meteorological 70(3): 674–684. 72 Organization, Geneva, Switzerland. WMO/TD-No. Smith, D. A. 2008. “The Social Value of Meteoro- 780. 38. logical Information.” SERA North: Economics of Weather, Climate and Climate Change. Pielke, R., and Richard E. Carbone. 2002. “Weath- er impacts, forecasts, and policy.” Bulletin-Ameri- Smith, K. 1990. Weather Sensitivity of Rail Trans- can Meteorological Society 83(3): 393–406. port. Proceedings of the WMO Technical Confer- ence Geneva: WMO. Pp. 236–244. Pollner, John, Jolanta Kryspin-Watson, and Sonja Nieuwejaar (undated). Disaster Risk Management Solow, Andrew R., Richard F. Adams, Kelly J. Bryant, and Climate Change Adaptation in Europe and Cen- David M. Legler, James J. O’Brien, Bruce A. McCarl, tral Asia. Global Facility for Disaster Risk Reduction William Nayda, and Rodney Weiher. 1998. “The val- (GFDRR). ue of improved ENSO prediction to US agriculture.” Climatic Change 39(1): 47–60. Quirk, James P. 1976. Intermediate Microeco- nomics. Vol. 1. Chicago, Ill: Science Research South Eastern Europe Disaster Risk Mitigation Associates. and Adaptation Programme (SEEDRMAP). 2008. Strengthening the Hydrometeorological Services in Regnier, E. 2008. “Public evacuation decisions and South Eastern Europe. hurricane track uncertainty.” Management Sci- ence 54(1): 16–28. Stern, Paul C. and William E. Easterling (eds.). 1999. Making Climate Forecasts Matter. Panel on Rogers, D, M. Smetanina, and V. Tsirkunov. 2009. the Human Dimensions of Seasonal-to-Interannu- Improving Weather, Climate and Hydrological Ser- al Climate Variability; Committee on the Human vices Delivery in Republic of Tajikistan. Dimensions of Global Change, National Research Council. Rogers, D, M. Smetanina, and V. Tsirkunov. 2010. Improving Weather, Climate and Hydrological Ser- Stewart, Thomas R., Richard W. Katz, and Allan H. vices Delivery in Central Asia. Murphy. 1984. “Value of weather information: A descriptive study of the fruit-frost problem.” Bulle- Rogers, D. and Vladimir Tsirkunov. 2010. Costs and tin of the American Meteorological Society 65(2): Benefits of Early Warning Systems. Global Assess- 126–137. ment Report on Disaster Risk Reduction. ISDR and World Bank. Suarez, Pablo, James Hansen, Miguel Carriquiry, Ashok Mishra, and Daniel Osgood. 2008. “Integrat- Sene, Kevin. 2010. Hydrometeorology: forecasting ing seasonal forecasts and insurance for adapta- and applications. Springer. tion among subsistence farmers: The case of Ma- lawi.” World Bank Policy Research Working Paper Simmons, Kevin M. and Daniel Sutter. 2005. WSR- Series. 88D Radar, Tornado Warnings, and Tornado Casu- alties. Weather and Forecasting, 20: 301–310. Subbiah, A.R., Lolita Bildan and Ramraj Narasim- han. 2008. Background Paper on Assessment of 73 Framework for Conducting Benefit-Cost Analyses of Investments in Hydro-Meteorological Systems the Economics of Early Warning Systems for Disas- United Nations and IDB. Information on Disaster ter Risk Reduction. Report prepared for The World Risk Management Case Study of Five Countries: Ja- Bank Group, Global Facility for Disaster Reduction maica. 13 Dec. 2007. and Recovery (GFDRR). Webster, Peter J. 2012. Improve Weather Forecasts Teisberg, Thomas J. and Rodney F. Weiher. 2009. for the Developing World. Nature 493(3): 17–19. Background Paper on the Benefits and Costs of Ear- ly Warning Systems for Major Natural Hazards. Re- Weiher, Rodney, Laurie Houston, and Richard Ad- port prepared for The World Bank Group Global Fa- ams. 2005. Socio-Economic Benefits of Climatolog- cility for Disaster Reduction and Recovery (GFDRR). ical Services. United States Contribution to the Up- date of WMO TN 145-No.424. Draft. Thornes, J. E. and B. Davis. 2002. Mitigating the Impact of Weather and Climate on Railway Oper- World Bank. 2008. Weather and Climate Services ations in the UK. Proceedings of the 2002 ASME/ in Europe and Central Asia. Working Paper No. 151. IEEE Joint Rail Conference, Washington DC, April 23–25, 2002. World Bank. 2012. Project Appraisal Document, Modernization of the National Meteorological Ser- Tiehallinto, 2006. Tieliikenteen ajokustannusten vices for Improved Climate Adaption Project (United yksikkˆarvot 2005 [Unit values of external costs Mexican States). in road transport 2005]. Helsinki: Finnish Road Administration. World Bank/GFDRR. 2013. Building Resilience: In- tegrating Climate and Disaster Risk into Develop- Tsirkunov, V. 2011. Strengthening Weather and Cli- ment, The World Bank Group Experience. mate Service Delivery in the Developing Countries –First Lessons and Future Challenges. World Bank WRMA. 2002. “The Weather Risk Management GFDRR. Available on http://www.gfdrr.org/sites/gf- Industry: Survey Findings for November 1997 to drr.org/files/WCDIS_First_Lessons_Future_Chal- March 2001”, Report prepared for Weather Risk lenges.pdf. Management Association by PriceWaterhouseCoo- pers, Available at http://www.wrma.org. UNISDR. 2008. Strengthening the Hydro-Me- teorological Services in South Eastern Europe, Zillman, J.W. and Freebairn, J.W. (2001). Economic available on http://www.unisdr.org/we/inform/ framework for the provision of meteorological ser- publications/7650. vices. WMO Bulletin 50 (3). UNISDR. 2009. Mitigating the Adverse Financial Ef- Zeng, Lixin. 2000. Pricing Weather Derivatives, fects of Natural Hazards on the Economies of Cen- Journal of Risk Finance 1(3):72–78. tral Asia, A Study of Catastrophe Risk Financing Options. 74 Publications from the LCSEN Environment & Water Resources Occasional Paper Series  Developing Offsets for Loss of Biodiversity: Experience from Colombia (2014)  Environmental Health Costs in Colombia: The changes from 2002 to 2010 (2014)  Metropolitan Pollution Management in Brazil: Taking Stock (2014)  Developing a Program for Contaminated Site Management in Low and Middle Income Countries (2014)  Restoring the Coastal Environment in Cartagena, Colombia (2014) (Available in English and Spanish)  Climate Change Impacts on Water Resources and Adaptation in the Rural Water Supply and Sanitation Sector in Nicaragua (2013) (Available in English and Spanish)  Climate Change Impacts on Water Resources Management: Adaptation Challenges and Opportunities in Northeast Brazil (2013)  El Futuro del Riego en el Perú: Desafíos y Recomendaciones (Volumen I: Informe de Síntesis y Volu- men II : Informe Principal) (2013)  Empowering Women in Irrigation Management: The Sierra in Peru (2012)  Environmental Health in Nicaragua: Addressing Key Environmental Challenges (Originally Published in 2010, Republished in 2012) (Available in English and Spanish)  Expanding Financing for Biodiversity Conservation: Experiences from Latin America and the Caribbean (2012) (Available in English and Spanish)  Overcoming Institutional and Governance Challenges in Environmental Management. Case Studies from Latin America and the Caribbean Region (2012)  Policy and Investment Priorities to Reduce Environmental Degradation of the Lake Nicaragua Water- shed (Cocibolca) (Originally Published in 2010, Republished in 2012) (Available in English and Spanish)  Uncertain Future, Robust Decisions; The Case of Climate Change Adaptation in Campeche, Mexico (2012) To find copies of these publications, please visit our web site: http://www.worldbank.org/LAC/LCSENSeries 75 LATIN AMERICA & CARIBBEAN REGION Environment & Water Resources OCCASIONAL PAPER SERIES