Farmers conducting a crop cutting experiment in Boudh, Odihsa.

Agricultural and space scientists have joined efforts in India to create a technology-backed smart way to quickly estimate the country’s crop yields, overriding time-consuming traditional sampling methods.

Scientists at the Mahalanobis National Crop Forecast Centre of the Ministry of Agriculture and the Indian Space Research Organisation (ISRO) rolled out the first such smart sampling in the autumn of 2019 in nine states of India, undertaking crop cutting experiments (CCE) that were a major shift from the random sampling procedures being followed until now.

The CCE metric – based on pioneering work of statisticians such as John Hubback, P. C. Mahalanobis, V. G. Panse and P. V. Sukhatme – has been the traditional basis of crop yield estimation in India. In CCE, samplers harvest crops from a particular size and shape of plot (square, rectangle, triangle or circle), and extract, dry and weigh them to arrive at a final yield per hectare figure. These experiments have been part of India’s General Crop Estimation Survey (GCES), in which around 1.23 million CCEs were planned every year for different crops in randomly selected villages and plots. Each major crop growing district conducted 80-120 experiments and the minor ones about 44-46.

In 2016, India launched a nationwide crop insurance programme, the Pradhan Mantri Fasal Bima Yojana (PMFBY), which increased the number of CCEs required to be carried out manifold. The programme required at least four CCEs for a major crop in a village or village panchayat. This immediately increased the number of CCEs to around 7-8 million per year from the earlier about a million.

Since the harvest period is very short, carrying out so many CCEs with the limited manpower became extremely difficult. Using technology to optimize the number of CCEs and making them more representative of the overall crop situation of the insurance unit (village, village panchayat, block, revenue circle, mandal or taluk) offered a potential solution. During kharif (monsoon) 2018 and rabi (winter) 2018-19, India’s agriculture ministry commissioned pilot studies to develop satellite and other advanced technologies for CCE optimisation. The studies showed that using satellite data, modeling approaches and other related parameters, it is possible to reduce the number of CCEs by 30-70%, while maintaining accuracy.

Overcoming estimation discrepancies

India’s agriculture ministry is hoping to cover large number of districts and crops with the new methodology. Pilot studies are underway for technology-based direct yield estimation.

Smart sampling is a stratified random sampling method where a crop yield proxy variable is used to stratify the block or village into different strata of high to low yield. CCE locations are then selected randomly and proportionately from each stratum. In this protocol, the yield estimated from a productivity efficiency model (PEM) using multi-resolution and multiple sources of satellite data, weather data and previous year’s CCE data is used as the yield proxy. High resolution crop maps and crop sowing date maps derived from satellite data are also used to make the yield proxy more representative.

The CCE points (latitude/longitude) selected for each Insurance unit are overlaid on digital and geo-referenced cadastral maps to generate the survey numbers of the fields where CCEs are to be conducted. These CCE locations are shared with village level officials 5-10 days before harvest. For each CCE location (primary), two additional back up points (secondary and tertiary) are earmarked. The CCE data is finally collected using a mobile app (CCE-Agri) developed by the agriculture ministry. This app collects CCE and crop data along with location information (geographical coordinates) and photographs.

There are many advantages of the smart sampling approach. Since the samples are selected based on the current season, they are more representative of the crop situation in the whole Insurance unit. CCE locations shared with the primary workers at the very last moment means the human bias or adverse selection can be avoided. It is also possible to optimize the number of CCEs by carrying out stratified random sampling – by either reducing the CCE number while maintaining accuracy or increasing accuracy without changing the CCE number.

The smart sampling procedure was implemented in 96 districts across Andhra Pradesh, Assam, Haryana, Jharkhand, Karnataka, Madhya Pradesh, Odisha, Telangana and Uttar Pradesh states for estimation of rice crop yield during kharif 2019. Around 320,000 points were generated for carrying out CCEs in more than 100,000 locations. During 2019-20, for crops of wheat, rice, mustard and sorghum, 635,000 CCE points were generated for conducting 189,000 CCEs using smart sampling procedures in these states.

The massive sampling work, carried out jointly by India’s agriculture ministry, MNCFC, Space Applications Centre (ISRO), State Remote Sensing Centres and agriculture departments, will have implications for the future crop estimation process. This will help farmers receive their rightful claims at the right time. Technology-based yield estimation will bring a paradigm shift in the current sampling process and will help in all domains of agricultural planning related to pricing, storage, import/export and risk management.

(*Director, Mahalanobis National Crop Forecast Centre, New Delhi, India. The author acknowledges inputs from Ashish K. Bhutani, CEO, PMFBY and Sunil K. Dubey, Assistant Director, MNCFC.)