Image: used under license from shutterstock.com Applications of yield monitoring systems and agricultural statistics in agricultural (re)insurance 18 October 2018 Ernst Bedacht
Agenda Introduction 1. Munich Re Parametric Insurance Solutions 2. Weather 3. Yield 4. NDVI 5. Crop Models Conclusions 06 October 2017 2
Munich Re (Group) Added value within the group Diversified structure more security Munich Re (Group) 1 Reinsurance Primary insurance Asset Management 1 This listing is incomplete and provides no precise indication of shareholdings. Company Presentation Munich Re 06 October 2017 3
Munich Re World market leader in agricultural (re)insurance (Indemnity Based Agricultural Insurance, Parametric Insurance, Crop Revenue Insurance) 4 6 October 2017
Challenges for Parametric Products Goal: Provide a product that helps the farmers in adverse conditions, is objective/transparent and gives the insurer the ability to assess the inherent risk. Basis Risk Transparency Data Availability Data history Extreme scenarios Triggers Product Requirements Minimizing the basis risk by being as close to the farmers risk as possible The parametric product (index calculation) needs to be well documented and understood by the Insured Reliable data source to ensure permanent delivery and have adequate fallback procedures in place Homogeneous data history in order to assess the risk How can rare events (e.g. Natcat) influence the parametric cover? Important to understand the intention of the cover as triggers should ideally reflect the agronomical risk Challenges Basis risk can never be avoided in parametric products Complexity vs. Transparency Satellites products might depend on clouds Technical problems may take long to solve From our experience, 10 years of data provides reasonable information, but the more the better, also depending on the loss entry probability. Hard to assess if not observed Difficult task requiring data analysis and agronomical background 06 October 2017 5
3 Weather
Rainfall Data Set Classification Potential perils: drought, excessive rainfall (1) Rain gauge only products that build only on observations using different interpolation methods, e.g. Global Precipitation Climatology Centre (GPCC) monthly precipitation, the Climatic Research Unit (CRU) monthly precipitation, Climate Prediction Center (CPC) unified gauge based analysis of daily precipitation (2) Reanalysis products, i.e. products from data assimilation schemes and numerical weather predictions or atmospheric models that use a combination of atmospheric variables from various sources as inputs, e.g. the Climate Forecast System Reanalysis (CFSR) reanalysis or products from National Centers for Environmental Prediction National Center for Atmospheric Research (NCEP NCAR) and European Centre for Medium Range Weather Forecasts (ECMWF) (3) Satellite only products that use infra red (IR), microwave (MV) or IR MV combined information, e.g. CHIRP, CMORPH_RAW, CHOMPS (4) Satellite gauge products that combine gauge only and satellite only products through different bias correction or blending procedures, e.g. TRMM, CMORPH, PERSIANN, and CHIRPS 06 October 2017 7
Rainfall CHIRPS Weather index based crop protection on County Level in China Data used: CHIRPS Description CHIRPS: Climate Hazards Group InfraRed Precipitation with Station data Early research focused on combining models of terrain-induced precipitation enhancement with interpolated station data New resources of satellite observations such as gridded satellitebased precipitation estimates from NASA and NOAA have been leveraged to build high resolution (0.05 ) gridded precipitation climatologies. When applied to satellite-based precipitation fields, these improved climatologies can remove systematic bias Horizontal resolution: 0,05 x 0,05 Available from 1981 now Latency: Final CHIRPS (all station data) is available sometime in the third week of the following month http://chg.geog.ucsb.edu/data/chirps/ 8
Rainfall CHIRPS Weather index based crop protection on County Level in China Basic Concept Pros Cons Description Divide the growing season of each crop in three 6-8 weeks risk periods, starting with the planting date Protect drought in all three periods and excessive rainfall just in the third period (harvest) Calculate the county wide rainfall per period for all years available Either: Estimate reasonable rainfall triggers for drought and excessive rainfall per period. Or: Set target rate on line for each period (drought and excessive rainfall) and calculate the corresponding triggers Concept is applicable in all relevant growing areas of the world Satellite rainfall data helps a lot in areas where the weather station density is low Works good for larger scale (countywide) drought or excessive rainfall events Independent and public data source Basis risk of weather index for crop shortfall Latency of the data (data is available sometime in the third week of the following month) Does not work for single fields/farms 9
Rainfall and Temperature Ukraine Perils covered Drought and Heat stress Crops: Corn, Winter Wheat Data used Interpolated Daily Meteorological data (temp, rainfall) MeteoGroup, based on weather station data Historical data: 1991-2016 Resolution: 25x25km Weather Index Heat Stress Period: No of days with Tmax > Critical Temperature Drought Period (31/41 days): 40% of long-term average rainfall 10
2 Yield
Area Yield Potential perils: all Data Sets Public available data (based on surveys or crop cutting experiments) Examples: USDA (USA), IBGE (Brazil), SIAP (Mexico), Important considerations: Basis Risk for the single farmer (due to size of areas) Quality and availability of data 06 October 2017 12
Area Yield Example India 2016: Huge shift from weather based crop insurance (WBCIS) to area yield crop insurance (PMFBY = Pradhan Mantri Fasal Bima Yojana) Market Premium ~ 3bn US$ Rate paid by farmer (loanee and non-loanee) is not higher than 2%, rest is subsidized by state and government For some crops (e.g. Rice) on a very high resolution (Gram panchayat, ~250.000 in India) or on block level (~ 5.500 in India) Perils covered: - Prevented Sowing due to deficit rainfall/adverse weather - Standing Crop - due to Drought, Dry Spell, Flood, Inundation, Pest & Diseases, Landslides, Natural Fire and Lightening, Storm, Hailstorm, Cyclone, Typhoon, Tempest, Hurricane & Tornado - Post Harvest Losses - maximum up to 2 weeks in cut and spread condition due to cyclonic and unseasonal rains - Localized Calamities - Hailstorm, Landslide & Inundation affecting individual farms 06 October 2017 13
Area Yield Example India Every state of India manages the PMFBY State is divided in several clusters which consist of some districts The primary insurers bid for each cluster and the cheapest gets the cluster Most important pitfalls: Potential trends in yield data Change in spatial dimension of areas aggregated for pricing and settlement (Block and Gram panchayat) Challenges with Crop Cutting Experiment process 06 October 2017 14
4 Normalized Difference Vegetation Index (NDVI)
Parameter: NDVI Potential perils: Lack of Biomass production (e.g. drought) Data Sets Satellites providing NDVI values Dataset Details Temporal Coverage Temporal Resolution Latency Spatial Resolution MODIS LANDSAT moderate-resolution imaging spectroradiometer (MODIS) is built by Santa Barbara Remote Sensing that was launched into Earth orbit by NASA in 1999 on board the Terra Satellite, and in 2002 on board the Aqua satellite. Landsat, a joint initiative between the U.S. Geological Survey (USGS) and NASA, represents the world's longest continuously acquired collection of space-based moderate-resolution land remote sensing data. 2000 - present Both daily few days 250m 1972 - present 16-days up to 7 weeks 30m SENTINEL Sentinel-2 is an Earth observation mission developed by ESA as part of the Copernicus Programme to perform terrestrial observations (Sentinel 2A & 2B) 2015 - present 5-days few days 20m 06 October 2017 16
NDVI Modis Example Australia Data used: NDVI Description NDVI (MODIS): MODIS 8-day composite developed and provided by the NASA/GSFC/GIMMS group for the USDA/FAS/IPAD Global Agricultural Monitoring project https://glam1.gsfc.nasa.gov/ Use relevant gridcells and Images for the relevant riskperiod Define Index (e.g. average NDVI of the gridcells over the riskperiod per year) Define trigger value for the NDVI and Payoutfunction Horizontal resolution: 0,25 x 0,25 (is only available aggregated) Available from 2002 now Latency: few days https://glam1.gsfc.nasa.gov/ 06 October 2017 17
5 Crop Models
Crop Yield Estimation Project Outlook Processbased Crop Model Yield data Weather data Remote sensing data https://www.gaf.de/ Statistical Crop Model https://www.pik-potsdam.de/ Modelled Yield-index 06 October 2017 19
6 Conclusions
Parametric products Experience and Conclusions Unfortunately, always comes with basic risk the smaller the farmers, the larger the basis risk But, Transparent and objective Quick loss payments possible Systemic risks like drought or excessive rainfall can be covered on an aggregated scale (catastrophic risks) from a worldwide perspective, no clear preference regarding preferred parameters data availability, quality and resolution will improve over time, which will lead to more tailored products 06 October 2017 21
Image: used under license from shutterstock.com Thank you for your attention Ernst Bedacht ebedacht@munichre.com www.munichre.com/agro