SONU AGRAWALMANAGING DIRECTOR
INDIA HEADQUARTERS
G 31 Quest OfficesDLF Golf Course RoadGurgaon 122 002 India
GLOBAL RESEARCH CENTER
SIDBI Innovation CenterIndian Institute of TechnologyKanpur 208 016 India
NORTH AMERICA
350 Fifth Avenue59th Floor, New York CityNY 10118 USA
Making Crop Insurance Work for Farmers
Discussion Agenda
2
Complaints against Crop Insurance Doesn’t pay adequately when there are losses Claims ratio are low, - not adequate returns on subsidy Doesn’t cover farm level losses Doesn’t cover small farmers
PRODUCTS DISTRIBUTION SETTLEMENT MODERN FARMING
Sub-phase01 Mar to 30 AprPayout per hect.
Daily rainfall (mm) more thanTrigger Pay out (Rs.) Variable Pay out
Per mm (Rs.)
Nature of Cover multiple event
15 0 76.220 380.9 76.225 761.8 60.144 1904.4 Exit
1904.4
Premium 2433 (23%)
Max Possible Loss Ratio due to Unseasonal Rainfall
78%
Mirzapur WBCIS - Unseasonal Rainfall Cover
Bhadohi MNAIS:
•Estimated Loss Ratio due to Unseasonal Rainfall is around 500%
•Bhadohi is adjacent district of Mirzapur
PRODUCTS DISTRIBUTION SETTLEMENT MODERN FARMING
Disease Congenial Weather (DCW)
Cover Phase, From 15-Jul-15
To 15-Sep-15
Index C
Maximum of consecutive DCW days where DCW is a day when Daily Min Temp, Daily Max Temp and Daily Average RH is within the benchmarksBenchmarks
Daily Max Temp (>,<) 29.40 35.60
Daily Min Temp (>,<) 19.50 27.30
Average RH (>,<) 63.00 87.00Max Payout (Rs.) 30,000
Disease Congenial Weather (DCW)
Date From 1-Jul
Date to 30-Sep
Index Condition Definition
Maximum of consecutive days of Rainfall > 2.5 and HumidityEve > 40
Strike (>) Payoff (Rs.)
2 5,0004 12,500 7 25,000 9 40,000
12 60,000
Max Payout (Rs.) 60,000
Two approved termsheets for Pomegranate within same districts for locations that are just 15 kms apart
Average Payoffs not Product Design is the selection criteria Even Sum Insured under each peril is not the same – Would farmer trust us with his risks?
PRODUCTS DISTRIBUTION SETTLEMENT MODERN FARMING
What do these Examples Suggest
• Flaws are in the design not the Scheme• If objective is to cover actual losses, design basis risk can be
minimized significantly• How do we do that?
• Bringing Irrigation, Soil and Sowing Period in the Weather Equation
• Proven research done by WRL on improving efficacy of product by introducing Irrigation and Soil variables.
• Designing product based on farmers’ risk perception rather than University suggested Thresholds and Strike
• Shunning the payoffs and adopting
PRODUCTS DISTRIBUTION SETTLEMENT MODERN FARMING
Weather Insurance
Data Quality – Case Study We were asked to review data of one
location in Moradabad which
reported received rainfall far greater
than what our station was showing
Data was reviewed using TRMM
rainfall charts that showed possible
rainfall within the range of 8 mm to
35 mm.
No recognized disturbance in synoptic
Charts was found by Meteorologists
Meteorologists thus could establish
correctness of data provided
Data Quality - Dispute Resolution
Stations need to be checked with
IMD referenced prototypes
Iso-lines and homogenous regions
need to be identified
Need to recognize that two stations
at the same location can differ by
as much as 10% ( or even higher in
case of rains) – more stations in the
same hom. region needs to be
checked
Data from Distant stations need to
be adjusted for distance
ABOUT US CLIENTS PLATFORM HOW IT WORKS IMPACT EDGE
Coffee - KarnatakaYield Weather Insurance
Damage Assessment - Hail / Flood / Landslide
10
100 Meters
250 Meters
Object based hierarchical image analysis to classify imagery of plots
Measured concurrently on the ground using standard rangeland monitoring procedure
Objects are further classified into vegetative groups and to species level by Rule Based Classification.
Well defined thresholds and Near Neighbor Classification Algorithm is feasible.
Use of spectral camera to enhance results and assessment. UAVs use for mid-term surveys is more feasible. However, for yield assessment, there are challenges:
Regulations: Not allowed by DGCA if allowed would be flying max at 150 mtr height
Can cover max 300 ha in an hour long flight – District like Jodhpur would require 27 UAVs
Advocating the use of helicopter in assessing losses due to localized calamities
Contd..
11
Inundation Statusand Yield EstimationMonitor Yields through Satellite images from LISS4, LANDSA and SAR
LANDSAT images of 30m x 30m resolution. For more detailed analysis, LISS4 images of 5m x 5m resolution.
Where visibility is affected due to clouds, Microwave SAR data is used
2D or 3D flood models can be prepared to estimate crop loss in Flood prone region (Ref: DST, NECTAR Project)
PRODUCTS DISTRIBUTION SETTLEMENT MODERN FARMING
Agro + Meteorology
Growth Monitoring
Phenology
Yield = f ( NDVI GDD Rain Index Ancillary data )
NDVI
PrecipitationWater Holding
PAR / GDD
Crop Yield Estimate
Empirical Method
YPA = f(xi)Xi = Meteorological IndicesVegetative indicesDrought Index
YPA = Yield per Unit Area
Multi Source InputsRemote Sensing Imagery / Weather DataHistorical Weather Data / Fertilizer and InputsSoil Detail / Irrigation Detail / Ground Truth
Robust Modeling Stasny - Goel Bayesian Method / Griffith AR Method Standard Ratio Estimation / Econometric Method Agro-Met Methodology / GIS MethodRemote Sensing Methodology
An integrated crop yield forecasting model adopting advanced remote sensing imagery, geographical information and appropriate statistical methodologies such as multivariate regression.
Yield Weather Insurance
Y = 0.78 - 0.0002*RF+ 0.00096*GDD - 0.0329*HDD + 3.006* Oct -1.21*Nov -8.61*Dec
R square : 0.67
Y = 18.76 - 0.0009*RF - 0.002*GDD + 0.025*HDD -11.63*Oct + 4.88*Nov - 4.24*Dec
R square : 0.79
Y = 2.601 - 0.0001*RF -0.00007*GDD - 0.002*HDD + 2.964*Sep - 0.023*Oct -
5.042*NovR square : 0.54
(Overall : Uttar pradesh, West Bengal and Bihar)
Model Results
Dividing crop period into stages
Vegetative : Count of consecutive unfolded leaves, until the reproductive parts are visibleReproductive : As soon as flowers / tuber / ear head are visible until all kernels / seed / tuber are physiologically mature
Damage based on parts of the crop
Crop Stand Damage : Count or % of crop stand area with no living axils / budsCrop Stem Damage : Count or % of crop stem snapped off with inability to yield or inactiveBranch Damage : Position and % of branches snapped off or damagedLeaf Damage : Count and % of leaves snapped off, shredded, de-colorized and inactiveEar / Pod / Head / Boll Damage : Count and % of yield part knocked off / chaffed / shriveled /broken or disease / pest infected
Fruit Damage Count and % of fruits / tree knocked off / malformed / disease / pest infected
Crop Yield estimation before Harvest
Locating representative sample area. Determining plant stand, row width & density ( plant / ear / fruit / pod ) sample population / 100 m2. Estimating yield based on observations.
In Season Crop Damage
PRODUCTS DISTRIBUTION SETTLEMENT MODERN FARMING
PRODUCTS DISTRIBUTION SETTLEMENT MODERN FARMING
Yield Insurance
Crop Cutting Experiments by Independent Third Parties with Government & Insurers Monitoring it on sample basis
Established Statistical & Parametric Models to estimate yield using function of NDVI & Weather Parameters as explained in product slides earlier Would minimize number of crop
PRODUCTS DISTRIBUTION SETTLEMENT MODERN FARMING
Pertinent Issues: Loanee Farmer
Our Research suggests up to 30% error in RUA / Station mapping by Banks
Loan is often offered for crop with highest Scale of Finance rather than crop actually sown
Bankers don’t keep a record of last year or season’s submission to make quick / correct MIS
VILLAGE FARMERS RUA BANK MISBALRASAR 1 CORRECTBARDADAS 79 CORRECTBHAMASI 86 CORRECTBINASAR 1 CORRECTBUNTIA 4 CORRECTDABALA 1 CORRECTDEPALSAR 1 INCORRECTDHADHAR 2 CORRECTKHANSOLI 109 INCORRECTRAMSARA 27 INCORRECTUNTWALIA 1 INCORRECT
Remote sensing can be used for crop identification of farmers’ crop An application has been developed for bankers to quickly prepare and save
farmer MIS for multiple future useApplication would also minimize human errors relating to MIS that have
caused non-payment of claims to many eligible farmers
PRODUCTS DISTRIBUTION SETTLEMENT MODERN FARMING
Secure browser based application allows easy access to bankers without compromising data security
Captures farmer information required for Insurance Database master of all the villages in and their Reference Unit Area / Weather
Station Banker doesn’t have to create new database every season
Automating Farmer MIS
PRODUCTS DISTRIBUTION SETTLEMENT MODERN FARMING
Pertinent Issues: Non-Loanee Farmer
Current Seasonality Discipline in some states exposes insurers to large adverse selection risks on non-loanee farmers (more in MNAIS).
Difficult to build team for non-loanee business if districts are allocated for one season or one year
Delay in notification in some states doesn’t allow us time to sell insurance to non-loanee farmers
Issues of ensuring KYC of non-loanee farmers
Discussion Agenda
19
Complaints against Crop Insurance Doesn’t pay adequately when there are losses Claims ratio are low, - not adequate returns on subsidy Doesn’t cover farm level losses Doesn’t cover small farmers
SONU AGRAWALMANAGING DIRECTOR
INDIA HEADQUARTERS
G 31 Quest OfficesDLF Golf Course RoadGurgaon 122 002 India
GLOBAL RESEARCH CENTER
SIDBI Innovation CenterIndian Institute of TechnologyKanpur 208 016 India
NORTH AMERICA
350 Fifth Avenue59th Floor, New York CityNY 10118 USA
Making Crop Insurance Work for Farmers