Using satellites to make index Index Insurance and insurance scalable* Daniel Satellites: Osgood, Lead the Scientist IRI Perspective Financial Instruments Sector Team deo@iri.columbia.edu Daniel Osgood, Lead Scientist *With support from UN-ILO Financial Instruments Sector Team Microinsurance Innovation Facility deo@iri.columbia.edu
Acknowledgements This work was commissioned by the United Nations ILO's Microinsurance Innovation Facility contract to the International Research Institute for Climate and Society at Columbia University "Using Satellites to Make Index Insurance Scalable" PG004060, ILO CU11-0558. Additional resources were provided by the generous support of the American people through the United States Agency for International Development (USAID), partially through the Joint NASA/USAID SERVIR program under contract NNH12AA54C. The US National Oceanic and Atmospheric Administration (NOAA) has also provided its support under cooperative agreement NA050AR431104. The US National Science Foundation has provided support for some of the activities related to this report through cooperative agreements NSF-SES 0345840 and NSF-SES 0951516 with the Center for Research on Environmental Decisions (CRED). Additional support has been provided by the CGIAR Research Program on Climate Change, Agriculture and Food Security (CCAFS), which is a strategic partnership of CGIAR and Future Earth. Work on the R4 Rural Resilience/Horn of Africa Risk Transfer for Adaptation (HARITA) project has been funded by sources including Oxfam America (OA) and the Rockefeller Foundation. We gratefully acknowledge the work and collaboration of project partners and stakeholders, particularly Relief Society of Tigray (REST). The contents are the responsibility of the IRI and do not necessarily reflect the views of USAID or the United States Government, CGIAR, Future Earth, Oxfam America, REST, or other funders and project partners.
Can satellites help validate large scale index insurance projects? Sparse rainfall and crop data blocks insurance scaling Hope satellites provide data for insurance at scale But potential errors in satellites big problem So must do expensive slow physical validation Increasing scale burdens the ground network of farmers, experts and partners, as it reaches new places, and updates an increasingly large number of products. Can satellites strengthen ground network to reduce the cost/improve quality for large scale insurance projects? Case study of R4/HARITA project in Ethiopia
Index insurance scaling fast CASE STUDY: R4 in Ethiopia Rainfall Estimate (scaling to Senegal) First sold (Ethiopia) 2009 2012: 77 villages, ~20,000 farms Min purchase possible: ~$3 Average premium for package chosen: ~$19 Average cash only premium : ~$9 Other satellite based indexes we have been involved in: Kilimo Salama (Kenya/Rwanda, Rainfall Estimate) Technical support for satellites, satellite contracts for 30,000+ farmers in a loan bundled satellite insurance Other KS projects have well over 100,000 insurance contracts MVP (Kenya/Ethiopia) 2007, Satellite veg index (NDVI) merged with raingauge development project protection
Lots of people exposed--need to do a good job
Case Study: Ethiopia, the R4 Rural Resilience initiative/harita Partnership: UN World Food Programme, Oxfam America R4 builds on HARITA Integrated risk management framework developed by Oxfam America, the Relief Society of Tigray (REST), together with Ethiopian farmers and several other national and global partners R4: risk reduction, risk transfer (insurance), prudent risk taking (microcredit), and risk reserves (savings) ~20,000 Individual farms purchased in 77 villages unsubsidized, nonloan bundled, nonmanditory Satellite rainfall estimate triggers payouts Can satellites help ground team provide quality product at large scales?
What is the insurance for? Climate change: more bad years Adaptation: increase productivity in normal years to cover bad year loss But strategies that increase productivity in most years face increased risk in bad years Threat of 1 drought year out of 5 prevents other 4 from being much more productive Key to adaptation is to relax risk of bad year to unlock productivity options Insurance: help reduce risk to unlock productivity
Complex system, many decision makers What chances to take? High yielding seed? Transplanting? Insurance Early season / Late season drought probability? Basis risk not simple: does insurance help reduce risk enough to take these opportunities? Must solve sophisticated problems together to design and validate insurance
On the ground network: Social Network for Index Insurance Design (SNIID) 1) Visit village with participatory design exercises Focus interaction on purpose of insurance Obtain initial parameters for design software Holistic strategy, timing of season, bad years to check, details about bad years 2) Run scripts to download data, generate indexes 3) Visit village, design exercises with initial contract Participatory exercises Check historical payouts Discussion of issues, quality of coverage Formal farmer/partner approval to move forward, how to address issues 4) Cross check across different different information sources different satellite technologies 5) Work with experts to address issues 6) Repeat (as necessary) Easier and cheaper to go to 83 villages With SNIID than initial 5
Reflected in many projects: Local partners build local processes/materials
In local languages
Games to train, elicit risks and farmer opportunities (Senegal)
Farmer risk quantification activities in Ethiopia for agriculture and insurance
Indonesia
Farmers can be comfortable with satellite triggered products CRED Ethiopia Game Experiment (2009-2013) Senegal 2013 (R4 Dry Run) Very realistic game scenarios Real money Real rainfall Real market product Real context (savings, community savings ) Tested Farmer insurance preferences (frequent) Community risk sharing (some) Time preferences (high discount rates) Impact of experiment on insurance purchase (yes) Satellite vs raingauge (either fine)
Smart use of satellites in index insurance Satellite rainfall estimates simply based on temperature of top of clouds -Need robust, simple indexes, easy to validate, communicate, improve with feedback, not sensitive to few mm measurement error eg: Few stormclouds at end of season Satellite vegetation probably not image of crop, but mix of dirt, rock shadow, trees, grasses, crops -Use satellite vegetation to proxy response of landscape to rainfall eg: Did landscape turn brown one month after end of season Similar issues with all satellite products
Validation strategy Starting point: agreement across region Index would have paid out in nearly all sites in 3 well known regional droughts in past ~10 years Maybe as good as insurance can be But can we provide meaningful payouts for more localized drought? Crosscheck across sources Confidence where agree, follow up where they don t Key to satellites as part of validation: Variety of satellite products available to users Each has strengths and weaknesses Each has a new piece of information about conditions on the ground, reducing the need for physical validation
Farmer interview comparisons with index
Regional Yield Comparisons with index
Using satellites to validate satellites (vegetation) Need to combine satellite vegetation info Hi resolution: very short history/few images Low resolution: mix of land covers/many images Used vegetative fraction analysis to try to use high resolution images to check low resolution images Then checked high resolution images with field visits Helpful finding: Satellite veg (EVI) good measure of fraction of vegetative coverage, scalable from high resolution to low resolution
Ground validation by satellite experts and farmers
Comparing bad years with other satellites
Example of 2012 payouts 2012 had Drought in many sites Normal year in many other sites Important test for index Follow up and index update 77 villages covered 11 villages complained 3 complaints were non-trivial Useful test of validation process
Merged TAMSAT/Gauge Ethiopian NMA served from NMA website
2012 as test of satellite vegetation validation of indexes All complaint indexes were flagged for verification due to low agreement with satellite vegetation Satellite vegetation agreement with index could help target project design resources to appropriate sites
Summary of key findings We have been able to successfully use satellite imagery of vegetation to improve the verification process for a large scale index insurance project. Vegetation indexes appear to work well in validating hazards that are the most widespread, but can miss some important problems Satellite vegetation verification approach: -Effective at end of season when foliage dense -Weak at season start before veg established -EVI was preferred veg index for validation -Other satellite products likely useful -Need to know what satellite pixel sees -Hi res satellites (infrequent) help validate low res (frequent) Satellite imagery of vegetation can improve the coverage quality and lower the cost of validating products offered to clients.