Can digital crop survey solve India's problem of incoherent estimates?

A few weeks back, Food Secretary Sanjeev Chopra expressed concern over the mismatch of official production estimates and what has been flagged by the industry regarding wheat, saying that such divergence needs to be brought down to a minimum.

Before then, none other than Finance Minister Nirmala Sitharaman, in an address, said that inaccurate production numbers hampers decision-making and urged officials to expand digital crop survey pilots to crops other than wheat and rice to accurate estimates.

The question of production numbers has been a puzzle that most governments have found difficult to solve.

A big part of the problem as a senior official explained a few days back, was that data collection in the states is very primitive.

"We in Delhi have to rely on state administration to get estimates of production and acreage," a senior official said.

Of late, though, there has been some forward movement with the government launching a pilot on Digital Crop Survey (DCS) in a few states on an experimental basis and also awarded contracts to a few start-ups and companies to undertake those surveys.

The results of these surveys will provide the framework for further more profound technology-based crop estimates in agriculture, which are accurate and reliable.

So how do the crop estimates are currently collected?

As a senior industry official explained presently, production estimates are broadly collated through two methods. First is the aggregation method by which every patwari has to send the sown area crop-wise to the SDM. The SDM aggregates it and sends it upwards. The District Commissioner or DC then sends it to the state.

The state then sends it to the Centre. This is how, through this pyramid, the numbers get aggregated.

Yields are also collated through crop cutting experiments which typically companies adopt to get accurate yields for settling insurance claims.

The Reliability Question

A couple of years back, the Central government's second advance estimate for foodgrains production for 2022-23 (July-June crop year) had scaled up the wheat production for 2021-22 crop year to 107.7 million tonnes.

This production was estimated at 106 million tonnes per the previous estimate.

While the reasons for the scaling up weren't spelt out, market players and traders were dismayed.

They felt that if wheat production in the 2021-22 crop year was indeed almost 108 million tonnes, there was no need to ban exports, nor would the Centre's procurement fall drastically that season to just around 19 million tonnes from almost 44 million tonnes a year before.

And most importantly, prices in the market would not have climbed to record highs throughout the year.

This is because if actual wheat production in the 2021-22 season was just 1.9 million tonnes less than the production of 2020-21, then India would not have gone through the crisis in the crop it has been experiencing since then.

The trade estimates that actual wheat output was much less, somewhere around 95-96 million tonnes in 2021-22.

The disparity between what the trade estimates to be the actual crop and the Centre's estimates has been a long-standing and recurring problem in Indian agriculture that has spanned several years across several crops.

In 2022-23, trade feels wheat production is around 100-101 million tonnes, while the government estimate is almost 113 million tonnes, a wide gap of almost 10 million tonnes.

Similarly, in cotton, in 2021-22, the Centre's first advance estimate of cotton production released on September 21, 2021 had pegged the output at around 36.21 million tonnes.

But, by the time the fourth estimates were released on August 17, 2022, the production had been revised downward to 31.20 million bales.

A massive drop down of almost 14 per cent.

The revision also meant that India, a country looking to export cotton, had to scramble for supplies to meet the rising yarn and textile industry demand within seven to eight months.

Ultimately, the government had to slash the import duty of 11 per cent in April 2022 to enable yarn and textile makers to replenish their inventories.

Such revisions have been frequent in many other crops, such as pulses or horticulture items such as potatoes and tomatoes.

All these make any firm price projections and inflation control strategies inadequate, apart from raising big questions on the credibility of these production projections.

This is the reason that more often than not, the government is found on the wrong foot when it comes to adequately predicting price signals.

To fire-fight a looming crisis, it ends up using harsh and draconian measures such as export bans, stock limits, deregistering futures etc.

These do more harm than good and lead to fueling market speculation.

The Digital Crop Survey

The Centre launched pilots on Digital Crop Survey (DCS) in the 12 states from the current kharif season of 2023. The survey's reference application has been developed as an open source, open standard and interoperable public good. Further, Geo-referenced cadastral maps with Geographic Information System (GIS) and Global Positioning System (GPS) Technologies are used to ensure the Farmland position.

The survey would be done by as many as six start-ups and private companies who have expertise in this field.

According to an official statement issued a few months back, the 12 states selected for the pilot are Madhya Pradesh, Karnataka, Telangana, Andhra Pradesh, Uttar Pradesh, Rajasthan, Tamil Nadu, Maharashtra, Odisha, Assam and Telangana.

The states have been selected based on the preparedness with respect to pre-requisite criteria for DCS, that is, geo-referencing of Village Map and digitised Record of Right (RoR) with ownership extent.

"The project aims are to create a single and verified source of truth about the crop sown data which is useful for accurate crop area estimation and development of various farmers' centric solutions," the official statement said.

Sources said for some companies the primary source of the data for the survey is the satellite imagery including satellites from India, Europe and America.

"So when you look at the satellite image, the first ability is to say, to identify the farmland, right? Which means separating the farmland from buildings, water bodies, roads, plantations, you know, other structures. And then, you know, just narrowing down on the agricultural land. That itself is a lot of skill. It's not easy," the industry player said.

"The second then is within the farmland to identify specific crops. So when you look from the top, I mean if you look at a Google image, everything from the top in the farmland will look green. And of course you know you can't work with Google images because Google images are old and they, you know, it's not optimal for agriculture applications.

So, the satellites come in different resolutions. So the second step is to be able to identify crops within the farmland, to be able to say this is Paddy, this is soybean, this is cotton, this is maize, this is sugar cane.

Whether the digital crop survey manages to put this age-old practice finally at rest and whether the government accepts even unfavorable production numbers is a million-dollar question that will be answered in times to come.