IFPRI-Using Remote Sensing technologies to improve sampling-Mangesh Patankar
5 likes657 views
Using remote sensing technologies such as drones, satellites, and mobile apps can improve crop cutting experiments (CCEs) conducted for the Pradhan Mantri Fasal Bima Yojana (PMFBY) agricultural insurance program in India. A pilot project in Maharashtra called TOPS integrated satellite imagery, weather data, and a mobile app to automatically select sampling locations, estimate crop yields virtually, and produce high resolution yield maps. Initial results found crop yield estimates were within 8-22% of independent assessments. The project demonstrated resource optimization through intelligent sampling and yield monitoring at village-level scales. Wider adoption of such technologies could help make CCEs more efficient and yield estimates more accurate.
IFPRI-Using Remote Sensing technologies to improve sampling-Mangesh Patankar
- 1. Using Remote Sensing Technologies to Improve Sampling Opportunities in the new Pradhan Mantri Fasal Bima Yojana (PMFBY) Agriculture Insurance Program Workshop at IFPRI, New Delhi, 21st December 2016 Mangesh Niranjan Patankar
- 2. 2 Agenda • Where are we? CCEs in the current context of PMFBY • How can we improve? • Optimization, experiences from pilot • Way forward
- 3. Where are we? CCEs in the current context of PMFBY 3
- 4. CCEs in the context of PMFBY 4 Standing Crop (Sowing to Harvesting) - Area Yield Index Based Coverage Prevented Sowing Mid Season Calamity Post Harvest Coverage Localized Calamity Claim given in case of ‘failed’ sowing of major crops due to inclement weather condition. Cover triggers if more than 75% area is affected. Provision for ‘on account’ settlement of claim in case of a wide spread calamity affecting the crops. Such payment would occur only if 'estimated' loss is more than 50%, with actual claim amount capped at 25% of 'potential claim' ‘Indemnity style’ claim assessment. Farmer needs to file for the losses happening due to localized perils like hailstorm. Coverage reflects expenses occurred till the event Forms core ‘index style’ component of the product. ‘Crop Cutting Experiments’ to be carried out at the ‘notified area’ to determine the ‘representative sample yield’ based on which yield index for the given territory is decided. Such yield is then matched with the ‘threshold yield’, which is 70%/80%/90% of the average historical yield (excluding maximum 2 calamity years). Yield shortfall is paid as claim at a pre-determined rate. ‘Indemnity style’ claim assessment. Covers post harvest losses suffered from crops which are allowed to dry in cut and spread condition on the farms, within maximum 2 weeks of harvesting. Perils included – cyclone, cyclonic rains and unseasonal rainfall.
- 5. Existing Mechanism 5 Random selection of field (well before the harvesting begins) Locating and marking experimental plot in the specific field Harvesting the CCE plot Threshing and harvesting of crop Weighing, Drying and Weighing dry weights (depending on the crop) Publishing plot data, extrapolating yield index based on average yields for number of plots Granularity Costs,Resources Not much emphasis on agro-climatic seggregation
- 6. Existing Mechanism 6 Party Function Local Revenue Department Officer (Patwari) Plot Selection, Conducting CCEs, Reporting in a specific format (Form – 2) Agriculture Department User of the data for insurance and other purposes Monitoring Agency Deployed by Insurer Coordination with Patwari, Monitoring actual CCEs based on targets Insurer, Reinsurer Ultimate user of the data for calculation of claims/disbursement of claims State Level Coordination Committee on Crop Insurance First level of contact for any clarifications on notification, claim etc. Department of Agriculture and Cooperation and Farmers’ Welfare (DAC & FW) Ultimate dispute resolution authority. Decision binding on state, insurer, bank and farmers
- 7. PMFBY: Where we are… Emphasis on Use of Technology 7 • Use of drones for reliability, accuracy, speed of conducting CCEs • Online transmission of data to avoid delays in claims assessments and payouts • Use of mobile for real time transmission of CCE data with GPS, date & time stamping • Use of these technologies to settle claims based on satellite images/derived products once the correlations are established • Cost of technology including purchase of hardware devices like smart phones will be equally borne by state & central Govt., at 50:50 basis. • Technical Advisory Committee to assist state government and insurers on technology intervention • NITI Aayog deliberating on the CCE optimization mechanism Central Crop Insurance Portal http://agri-insurance.gov.in/Login.aspx
- 8. How can we improve? 8
- 9. • Minimize the number of experiments. E.g. even for a circle level assessment, state of Maharashtra needs about 55,000 crop cuttings. • Minimize time lags in publishing the crop yield statistics, generated for the cropping period • Predict sowing failure and associated economical loss • Determine accurate crop yield statistics at a granular level (micro level) • Minimize on-site loss survey expenses and delays in case of localized calamities like hailstorm • Detect early warnings from the field and implement an action plan accordingly for irrigation, agri- credit and other agri-inputs • Minimize insurance claim disputes by the use of objectively verifiable techniques • Equip the stakeholders with necessary technological know-how (especially the government machinery, insurers and agricultural research agencies) Key Requirements 9
- 10. Geographical Scope Limited to the State of Maharashtra 14 districts for production of the yield statistics All the districts for production of the weather statistics (overall 40,000 villages) Temporal Scope Kharif 2012 to Rabi 2015 Project Partners Swiss Re, Agriculture Insurance Company of India, Niruthi Climate and Ecosystems, CRIDA, Government of Maharashtra Special Assistance Recognizing the long term utility of the concept, TOPS project received an assistance from Government of Maharashtra, under its PPP-IAD programme. Rest of the support was received from Swiss Re and AICI. Initial Deliverables (set in 2012) 1. Weather statistics – Produce historical (1982 to 2011) as well as current (2012 to 2015) weather data statistics at village level (for 40000 villages) for the whole state 2. Yield statistics – Prepare historical (2001 to 2011) as well as current (2012 to 2015) yield statistics for selected regions at village level for specific crops (jowar, bajra, cotton, soybean and gram) As a bi-product, it was observed that the state government also benefited from the project as the project demonstrated that considerable resource optimization is possible using the smart CCE sampling methodologies. TOPS Project in the State of Maharashtra
- 11. • TOPS (Terrestrial Observation and Prediction System) integrates extensive libraries for data retrieval and preprocessing, algorithms for processing satellite data, interpolation of meteorological data, application programming interfaces to facilitate integration of multiple models, and numerous databases to archive and manage the metadata associated with model inputs and model outputs. • Satellite imagery is extensively used in the project to identify the crop and crop conditions. • The project also heavily uses handheld mobiles and a mobile app named CropSnap for capturing GPS tagged images, crop booking, crop stage assessment and crop yield estimation. • The project uses cloud technology for the storage of data where the images are interpreted through machine learning algorithms (artificial intelligence). The entire acquisition and processing process is automated taking into account all the communication issues in rural India. • A dedicated portal was created to indicate the locations of CCEs, yield statistics and contact information of the farmers. Software for dynamic sampling scheme was also created. 11 Further details - TOPS
- 12. • Create an intelligent sampling scheme using TOPS technology to determine the minimum number of CCEs to be conducted • Cluster villages in each circle into low, medium and high yield categories • Conduct virtual CCEs with CropSnap in each village • Conduct CCEs manually at chosen locations based on intelligent sampling and virtual CCEs • Assimilate crop yield data from the CCEs into the satellite-based yield maps to produce a final yield surface for each village • Collate crop yields for each village and submit the data to Maharashtra government Methodology 12
- 13. Manual CCE’s Camera CCE’s Panoramic Close-up Granular CROP SNAP Photo to Yield Yield MACHINELEARNING& ANALYTICS Potential Yield Calibrate High Resolution Final Yield (0.5 ac pixels) Refinement High Res Satellite Data High Res Weather Data Crowdsourcing Send Spreadsheets Historical Yield Models Village-level yield monitoring using satellite, weather and mobile data CROP SNAP Steps in crop yield estimation A unique example where several technologies (Satellites/Sensors and Data Management) put to use for tackling a complex challenge!
- 14. • Conducted nearly 6000 CCEs spread across 14 districts during Kharif and Rabi seasons of 2014 in selected crops. Further, in 2015, around 6000 CCEs were conducted in 108 villages together in Kharif and Rabi. Data for gram, jowar, bajra and soybean was found to be within 8%, 13%, 15% and 22% of the yields assessed by independent methods. • Intelligent sampling performed better during Rabi season when compared to Kharif. Lowest savings were seen in Soybean and highest in Jowar. Specific Recent Findings 14
- 15. • Use of image based crop recognition can be operationalized once the necessary models are in place – but it takes time! • Accuracy can be an issue in the initial years, given that the initial data fed to the models needs to be from the historical datasets available through manual conventional crop cuttings – which may not be the best source for such data • Need an end to end system/portal, which takes care of – smarter sampling, – rapid yield assessment using manual as well as image based inputs and – digital reporting of the yield • Smooth adaptation by government is important to ensure that the efforts are not limited to pilots Way Forward… 15
- 16. Example of potential expansion of village-level yield data to individual fields within the village. This process involves processing high- resolution satellite data at sub-meter resolution to identify and tag individual fields Future outlook: Farm-level crop yield estimation
- 17. Thank you… Contact Details: Mangesh Niranjan Patankar Client Service Manager Agriculture, Assistant Vice President Property & Specialty Underwriting Swiss Re Services India Private Ltd., Mumbai, India Direct: +91 22 6661 2153 Mobile: +91 77108 91100 E-mail: Mangesh_Patankar@swissre.com 17
- 18. 18
- 19. Legal notice 19 ©2016 Swiss Re. All rights reserved. You are not permitted to create any modifications or derivative works of this presentation or to use it for commercial or other public purposes without the prior written permission of Swiss Re. The information and opinions contained in the presentation are provided as at the date of the presentation and are subject to change without notice. Although the information used was taken from reliable sources, Swiss Re does not accept any responsibility for the accuracy or comprehensiveness of the details given. All liability for the accuracy and completeness thereof or for any damage or loss resulting from the use of the information contained in this presentation is expressly excluded. Under no circumstances shall Swiss Re or its Group companies be liable for any financial or consequential loss relating to this presentation.