Written by Amrithavarshini Venkatesh, Vineet Bhandari, Varad Pande
When the Government of India introduced the Multipurpose National ID Card (MNIC) scheme in the early 2000s, it had a limited scope. The MNIC was meant to be an ID card to “verify the citizenship of Indians and secure our borders”. In six years, the project was able to provide ID cards to a mere 12 lakh people. Then came Aadhaar, a paradigm shift, which re-imagined what a country can do with an ID system at scale — from targeting government subsidies to driving start-up business models. The rest, as they say, is history. Aadhaar is today ubiquitous, transforming service delivery and spurring innovation. Can we think of a similar paradigm shift in the Soil Health Card (SHC) Scheme?
The scheme, that was introduced in 2015, intended that every farmer receive a health card for their soils that tells them the status of the nutrients in it, and, as a result, guides them about the fertilisers they should apply to maximise their yields. The entire government agriculture extension and research system galvanised itself, collected samples, analysed them for 12 soil chemical parameters, recommended fertiliser dosages and printed these on the SHCs, which were given to farmers. The scheme delivered on the basic promise — as of June, 2.53 crore samples had been collected, and SHCs distributed to 10.74 crore farmers.
However, the well-intentioned scheme is falling short at three levels.
First, operational challenges plague the system. The current “census” approach, where soil samples are collected from every 2×2 hectare parcel of land in irrigated areas (10×10 hectare in dry areas), and transported en massefor analysis in a dated network of wet chemistry labs, has put tremendous strain on the system, and the quality of soil analysis has suffered. Studies conducted have shown a low correlation between the results generated by the SHC scheme and those generated by gold standard labs. For instance, a Harvard study in Gujarat last year found accuracy issues in 300 of the 800 plots tested. On the field, such stories abound.
Second, the scheme’s current design oversimplifies the nutrient recommendations — for example, if the health card shows that a farmer’s soil is deficient in zinc, it recommends topping up zinc. However, increasingly, research is showing that a crop’s “yield response” to a nutrient is far more complex than this. It is determined not only by the deficiency of that nutrient, but also other variables — rainfall, production practices, the presence of other nutrients, soil acidity, and temperature, to name a few. The correct yield response can be predicted from a model with data on the above parameters, a system that the scheme currently does not use. The simplistic recommendation based on deficiency of that nutrient alone is often sub-optimal, and can exacerbate the farmer’s problem, rather than solve it.
Finally, the scheme underestimates its own potential, because its large-scale collection of soil data sees little use outside of filling out a physical card. This vast repository of data, painstakingly aggregated from millions of samples, remains largely isolated from researchers, start-ups and even state governments.
These shortcomings, however, present a remarkable opportunity for Indian agriculture. What if we could move to a sampling-based soil information system that reduces the need for the tens of millions of samples that strain our lab capacity, and produces better results four times faster, at half the cost? What if we could develop predictive models using big data to provide recommendations to farmers that account for all the factors that affect a crop’s yield response? For example, a recommendation that encourages the use of a custom fertiliser blend in addition to asking the farmer to reduce sowing depth. What if we could go beyond health cards the way we went beyond mere identity cards with Aadhaar and re-imagine how to structure and use the vast repositories of agriculture-related data that currently reside within silos — soil, rainfall, cropping patterns, temperature, irrigation? Can we make these datasets available through an open API platform?
This could help start-ups to combine soil health card data with rainfall and irrigation data and deliver precision irrigation advisories to our farmers on their mobile phones. Fertiliser companies, building upon such a platform, leveraging soils data, weather data, and farmer demand patterns, can shape the distribution of fertiliser blends in different districts. Such a platform can catalyse a wave of innovations in agriculture, in much the same way as IndiaStack has done in financial services.
In data starved Tanzania, a version of such a platform already exists — the Africa Soil Information Service uses machine learning to bring together various pieces of data (soil, climate, production practices) to enable the government and fertiliser companies determine what blended fertilisers could improve soil nutrition. In India, states like Andhra Pradesh and Bihar have begun to go down this path. Andhra Pradesh, for example, is currently bringing together years’ worth of cropping pattern data, precipitation data, temperature readings, irrigation information and SHC data, and combining them with farmer production practices to determine what impact different nutrients have on yield. As a first step, this will act as a decision support system to do more targeted extension, and produce more customised fertiliser blends. Eventually, it can be used to offer recommendations to farmers to help improve yields.
Stories of farm distress make headlines almost every day. Could farm data and intelligent digital platforms that build on the SHC programme and leverage big data analytics be a solution? The answer, in our view, is a resounding yes.