Agricultural production is quite inequitable globally. However, in the post-pandemic era, most countries have realised the importance of increasing self-reliance to safeguard their interests. This is in addition to efforts being undertaken to improve productivity as well as the sustainability of agriculture. These efforts have gained impetus due to long-lasting factors like climate change and hopefully, shorter-duration incidents related to geo-political instability.
The agriculture sector is a sophisticated and interdisciplinary system that can be governed by a data-driven approach to increase farmers’ income while focusing on technological advancements to develop localised and predictive models which enable informed decisions about crop planning, monitoring, yield estimation and market intelligence. Widespread adoption of digital tools like Artificial intelligence (AI) and Machine learning (ML) can help transform agriculture in India.
AI interventions can be a part of the entire value chain of agricultural produce, even before cultivation starts. They can be used for better analysis of soil before planting. They can forecast the prices of commodities even before they are harvested. Farmers can take advantage of technological developments across the value chain and improve their decisions about the crops to grow, the market to sell in and the buyer to sell to.
Farms generate huge datasets on a daily basis which can be collected, tracked and trained to develop sustainable solutions from insights into plant diseases, pests and nutrition needs of crops. AI sensors can be trained to detect and target weeds for focused use of herbicides, lowering the cost of cultivation while producing better quality products. AI can also help develop ever-evolving yield maps which predict potential yield based on past data so growers can decide whether to grow a crop or alternatives or take precautionary steps to safeguard yields (through monitoring of weather, and pests and pathogens). Farmers can monitor the application of inputs to standing crops, irrigation schedules and harvest readiness. They can do likewise in livestock farming. Higher order application of ML for quality, pricing, elimination of weeds and insurance-linked to crop health will also evolve.
Currently, digital agriculture solutions are utilised by multiple agritech startups which try to solve traditional problems with new solutions backed by AI. While these solutions are innovative and indigenous, they are only able to solve specific problems and often lack technological amalgamation. The effect of clubbing AI with other technologies such as blockchain for traceability (from farm to plate) and Internet of Things (IoT) to take decisions, say, on precision spraying to ward off pest and disease attacks is multiplicative and not just additive. Further, the impact is not restricted to only on-field activities but can positively influence ease of access to finance for farmers and curtail post-harvest losses.
These futuristic agricultural solutions have some limitations which hinder their adoption by smallholder farmers. First of all, there is an added cost of services. The typical Indian farmer is aware of traditional tried and tested solutions used over generations which makes them resistant to incurring additional costs on digitalisatio of agriculture. Poor awareness among end-users and trust issues with data privacy make it difficult to understand the complex data science and algorithms that work in generating predictive solutions. Third is the fragmentation of technological infrastructure, especially in rural areas. Rural India has witnessed a digital boom with increased mobile penetration and internet connectivity, but the integration of data collection centers and sensors with ever-improving interoperability is yet to happen.
The government has been focused on developing a digital ecosystem that focuses on long-term aspects such as interoperability, data governance, data quality and data protection in the agriculture sector. The development of AgriStack is a commendable step in this direction which will digitise farmers’ data and land records. In the last budget, the government has announced the setting up of three centres of excellence in AI.
In terms of products and solutions also, there is still a long way to go. While developing an ML model for a specific use case or a limited set of data is easy the data collection infrastructure itself is quite immature in the country. Under such circumstances, developing a robust machine-learning model which will work with high and acceptable degree of accuracy becomes a challenge. This requires a lot of effort in the collection, standardisation and annotation of raw data to derive meaningful inferences.
In the long term, the private sector has to play a bigger role and innovate to make the solutions financially viable and technologically amenable for scale to be built up. But in the short term, this is likely to be driven by social returns rather than business returns. While public-private partnerships (in the form of corporates collaborating with research institutes and the public sector) can be a way forward, there is also a need to build checks to prevent data capture by a few large stakeholders and early movers.
By Srinivas Kuchibhotla
(The author is Partner, KPMG in India. The views expressed and observations made in this article are those of the author and not necessarily of this website. Top photo is of a polyhouse in Garounda, by Vivian Fernandes. Photo for representation only.).