Technologies and Big Data for
Improved Crop Insurance
Pramod Aggarwal and Team
CGIAR Research Program on Climate Change, Agriculture and Food Security,
BISA, CIMMYT, New Delhi-110012, India
Challenges in Crop Insurance
• Everyone loves farmer-driven
insurance: several pilots,
hundreds of thousands, reduced
basis risks, but moral hazards,
transaction costs, agro-ecological
diversity, scalability
• Top down scheme reaching
millions: farmer’s dissatisfaction-
spatial basis risk, product design,
loss assessment, claim
settlement delays
Desired: A scalable farmer-centric insurance
scheme for large regions
Technologies and Big Data Provide New
Opportunities for Such a Scheme
Harnessing Technologies and Big Data for
Improved Crop Insurance
1. Acquiring data: Farmers practices
and crop growth (digital geo-referenced pictures)
crowdsourced by mobile Apps, sensors in farmer’s
fields, gridded data, VIs, satellite rainfall, UAVs
2. Assimilating data: models,
statistics, geospatial techniques
3. Analyzing data: Assessment models,
optimization, machine learning
4. Application: ICT, Mobile Apps, multiple
data for MRV, digital banking
A Tool for Harnessing Technologies and
Big Data for Improved Crop Insurance
Crowd sourced
information from
farmers: practices,
pests and diseases
+
Digital
knowledge:
satellites, UAVs,
cameras
+
Spatial databases:
weather, soils,
management,
historical yields
Loss assessment:
mid-season, end-
season, post-
harvest
Product design:
triggers, premiums,
loss ratio, farmer
satisfaction
Claim settlement:
immediate mobile
notification, bank
transfers
Models
and
analytics
on cloud
Pilot Evaluation for Soybean in India
Five stages of yield loss
assessment (all need
different approaches)
1. Prevented sowing
2. Sowing failure
3. Mid-season loss
estimates
4. End-season
5. Post-harvest
Prevented Sowing
Data Source: CHIRPS data from 1981 to 2016
Post-harvest Loss Assessment
Yield Loss Assessment
4 Day: 50 mm
Two day rain > 125 mm
Length of Dry Spell > 20 Days
Farmer participatory large-scale loss assessment of
soybean for insurance (work in progress)
Prevented Sowing
Sowing Failure
Data Source: CHIRPS data from 1981 to 2016
Post-harvest Loss Assessment
Yield Loss Assessment
4 Day: 50 mm
Two day rain > 125 mm
Length of Dry Spell > 20 Days
Farmer participatory large-scale loss assessment of
soybean for insurance (work in progress)
Prevented Sowing
Sowing Failure
Data Source: CHIRPS data from 1981 to 2016
Post-harvest Loss Assessment
Yield Loss Assessment
4 Day: 50 mm
Two day rain > 125 mm
Length of Dry Spell > 20 Days
Farmer participatory large-scale loss assessment of
soybean for insurance (work in progress)
Prevented Sowing
Sowing Failure
Data Source: CHIRPS data from 1981 to 2016
Post-harvest Loss Assessment
Yield Loss Assessment
4 Day: 50 mm
Two day rain > 125 mm
Length of Dry Spell > 20 Days
Farmer participatory large-scale loss assessment of
soybean for insurance (work in progress)
• Farmers satisfaction
index-payment
when due and in
right amount
• Industry: 70-80 %
claim ratio
• Government:
Premium subsidy
not to increase
Improved triggers for weather insurance:
win-win products for farmers, industry and government
Improved triggers for weather insurance:
win-win products for farmers, industry and government
(work done in collaboration with AIC, Aon Benfield and Govt of Maharashtra)
0
10
20
30
40
50
60
70
80
90
100
0
1
2
3
4
5
6
7
Existing Contract Satisfaction Index-farmer Proposed Contract
Claimfrequency,%
SatisfactionIndex,Claimsratio
Satisfaction Index Claim ratio claim frequency, %
 Technologies and Big
Data for increasing the
efficiency and efficacy of
crop insurance
 Interest of all
stakeholders crucial
 More evidence needed
 Insurance versus
comprehensive risk
management- bundling
Looking forward: Building evidence, Systematic Learning
and scaling of Crop Insurance
Address generic vulnerability issues simultaneously – poverty,
literacy, governance, etc., which limit adaptation even today and
will do so in future as well
And….

Pramod Aggarwal: Technologies and big data for improved crop insurance

  • 1.
    Technologies and BigData for Improved Crop Insurance Pramod Aggarwal and Team CGIAR Research Program on Climate Change, Agriculture and Food Security, BISA, CIMMYT, New Delhi-110012, India
  • 2.
    Challenges in CropInsurance • Everyone loves farmer-driven insurance: several pilots, hundreds of thousands, reduced basis risks, but moral hazards, transaction costs, agro-ecological diversity, scalability • Top down scheme reaching millions: farmer’s dissatisfaction- spatial basis risk, product design, loss assessment, claim settlement delays
  • 3.
    Desired: A scalablefarmer-centric insurance scheme for large regions
  • 4.
    Technologies and BigData Provide New Opportunities for Such a Scheme
  • 5.
    Harnessing Technologies andBig Data for Improved Crop Insurance 1. Acquiring data: Farmers practices and crop growth (digital geo-referenced pictures) crowdsourced by mobile Apps, sensors in farmer’s fields, gridded data, VIs, satellite rainfall, UAVs 2. Assimilating data: models, statistics, geospatial techniques 3. Analyzing data: Assessment models, optimization, machine learning 4. Application: ICT, Mobile Apps, multiple data for MRV, digital banking
  • 6.
    A Tool forHarnessing Technologies and Big Data for Improved Crop Insurance Crowd sourced information from farmers: practices, pests and diseases + Digital knowledge: satellites, UAVs, cameras + Spatial databases: weather, soils, management, historical yields Loss assessment: mid-season, end- season, post- harvest Product design: triggers, premiums, loss ratio, farmer satisfaction Claim settlement: immediate mobile notification, bank transfers Models and analytics on cloud
  • 7.
    Pilot Evaluation forSoybean in India Five stages of yield loss assessment (all need different approaches) 1. Prevented sowing 2. Sowing failure 3. Mid-season loss estimates 4. End-season 5. Post-harvest
  • 8.
    Prevented Sowing Data Source:CHIRPS data from 1981 to 2016 Post-harvest Loss Assessment Yield Loss Assessment 4 Day: 50 mm Two day rain > 125 mm Length of Dry Spell > 20 Days Farmer participatory large-scale loss assessment of soybean for insurance (work in progress)
  • 9.
    Prevented Sowing Sowing Failure DataSource: CHIRPS data from 1981 to 2016 Post-harvest Loss Assessment Yield Loss Assessment 4 Day: 50 mm Two day rain > 125 mm Length of Dry Spell > 20 Days Farmer participatory large-scale loss assessment of soybean for insurance (work in progress)
  • 10.
    Prevented Sowing Sowing Failure DataSource: CHIRPS data from 1981 to 2016 Post-harvest Loss Assessment Yield Loss Assessment 4 Day: 50 mm Two day rain > 125 mm Length of Dry Spell > 20 Days Farmer participatory large-scale loss assessment of soybean for insurance (work in progress)
  • 11.
    Prevented Sowing Sowing Failure DataSource: CHIRPS data from 1981 to 2016 Post-harvest Loss Assessment Yield Loss Assessment 4 Day: 50 mm Two day rain > 125 mm Length of Dry Spell > 20 Days Farmer participatory large-scale loss assessment of soybean for insurance (work in progress)
  • 12.
    • Farmers satisfaction index-payment whendue and in right amount • Industry: 70-80 % claim ratio • Government: Premium subsidy not to increase Improved triggers for weather insurance: win-win products for farmers, industry and government
  • 13.
    Improved triggers forweather insurance: win-win products for farmers, industry and government (work done in collaboration with AIC, Aon Benfield and Govt of Maharashtra) 0 10 20 30 40 50 60 70 80 90 100 0 1 2 3 4 5 6 7 Existing Contract Satisfaction Index-farmer Proposed Contract Claimfrequency,% SatisfactionIndex,Claimsratio Satisfaction Index Claim ratio claim frequency, %
  • 14.
     Technologies andBig Data for increasing the efficiency and efficacy of crop insurance  Interest of all stakeholders crucial  More evidence needed  Insurance versus comprehensive risk management- bundling Looking forward: Building evidence, Systematic Learning and scaling of Crop Insurance
  • 15.
    Address generic vulnerabilityissues simultaneously – poverty, literacy, governance, etc., which limit adaptation even today and will do so in future as well And….