AIM for Scale helps India deliver AI-powered forecasts to 38 million farmers

September 17, 2025 –

The Indian Ministry of Agriculture and Farmers’ Welfare sent farmers AI-based forecasts that predicted when the monsoon would arrive up to 30 days in advance, allowing farmers to plan ahead. AI allowed forecasts to not only be accurate, but also be tailored specifically to farmers’ needs. With catalytic support from AIM for Scale, this first-of-its-kind program is a model for how governments can harness AI to help farmers adapt to climate change.

Thirty-eight million farmers across India received forecasts this summer accurately predicting the start of the rainy season. This forecast, powered by artificial intelligence (AI), was tailored to farmers’ needs, providing them with advance prediction of the rainy season earlier than ever before—up to four weeks ahead of the rain. This represents a paradigm shift for smallholder farmers who had to make important farming decisions like what, how much, and when to plant without this information.

With this initiative, the Indian Ministry of Agriculture and Farmers’ Welfare set a model for what the future of weather services for farming could look like for hundreds of millions of smallholder farmers across the tropics who depend on information about when the rainy season, known as the monsoon, will come each year. Nearly two-thirds of the global population live in areas with monsoon climates.

“This program harnesses the revolution in AI-based weather forecasting to predict the arrival of continuous rains, empowering farmers to plan agricultural activities with greater confidence and manage risks. We look forward to continuing to improve this effort in future years,” says Dr. Meherda, Additional Secretary at the Indian Ministry of Agriculture and Farmers’ Welfare.

The Indian Ministry partnered with an international team of researchers to select its forecast. The Human-Centered Weather Forecasts Initiative at the University of Chicago Institute for Climate and Sustainable Growth led the effort to evaluate forecasting models, recruiting researchers from IIT Bombay, IISc Bangalore, and the University of California, Berkeley. They found that Google’s Neural GCM model and the European Centre for Medium-range Weather Forecasts’s (ECMWF) Artificial Intelligence Forecasting System (AIFS) best predicted the monsoon. The research team then created a blended model which combined these with the India Meteorological Departments, historical rainfall observations to maximize the accuracy of the forecast. The effort was partially supported by catalytic funding from AIM for Scale, a global initiative backed by the Gates Foundation and the United Arab Emirates, which works to scale up evidenced-backed, cost-effective agricultural innovations for the benefit of farmers in low- and middle-income countries.

“Disseminating AI weather forecasts has an incredibly high return on investment, likely generating more than $100 for farmers for each dollar invested by the government,”

says University of Chicago economist Michael Kremer, a 2019 Nobel laureate and co-director of the Human-Centered Weather Forecasts Initiative. “India is leading the way in using AI to improve people’s lives across many sectors, including agriculture.”

This proved to be an important year for the Ministry to begin this effort. Typically, the Indian summer monsoon begins over southern India in June—the government declares the monsoon season has started—and the monsoon advances smoothly northward, bringing sustained rainfall to most of the country by July. This year, however, the monsoon was unusual. It hit the southern part of India earlier, leading most to expect an early monsoon season to come to their various communities. The monsoon progressed for about a week, and then it stopped for close to three weeks before moving again. The new AI-based forecast predicted that pause—and farmers were paying attention.

Parasnath Tiwari, a farmer from Madhya Pradesh, received the forecast on his phone. Because of what he learned, he was able to prepare earlier. He also decided to switch the types of crops to more lucrative ones because the message gave him confidence that the season would be long enough.

“Before this, I mostly relied on my own experience and local knowledge to know when the monsoon would arrive,” says Tiwari. “The forecast about the arrival of the monsoon was accurate…I have increased trust in the forecast, and I will rely on the information shared by scientists in the future.”

A Smarter Way to Forecast

 The Human-Centered Weather Forecasts Initiative’s team worked to identify which of the many open-access weather models best predicted the start of the monsoon across different regions of India. The key attribute: The model had to accurately forecast not only the physics of the atmosphere but also the information the farmers needed.

“We have been going through an AI-driven revolution since 2022 and AI models have shown promise for many one- to two-week forecasting applications. But their ability to predict complex phenomena—like the monsoon—was unclear, and frankly, unexpected,” says Pedram Hassanzadeh, Associate Professor at UChicago, who works on climate dynamics and AI, and is a co-director of the Initiative. In 2022, he co-created NVIDIA’s FourCastNet, the pioneering AI-based weather model. “Here, we started with rigorous benchmarking. We had to understand the strengths and weaknesses of each model, and more importantly, match the model with the needs. That’s where AI forecasts have the power to be revolutionary because they are more accurate, faster, and easier to adapt.”

Robust research suggests that when accurate forecasts like these are available, farmers respond by changing their decisions. In an earlier study, University of Chicago researchers Fiona Burlig, Amir Jina, Erin Kelley, Gregory Lane and Harshil Sahai shared a localized forecast with about 250 Indian villages. The team learned that farmers would hear when the monsoon started at the tip of India and then often wait to act until it reached them.

Jina, also a co-director of the Initiative, noted that, “Once the rain starts, it can be too late to make big important decisions, like changing a crop, planting more land, or forgoing the farming season and getting a job in the city instead. By providing an accurate forecast around a month in advance, farmers were able to align their decisions with the coming weather and make better choices.”

Farmers needed more advanced knowledge of when the rain would hit their farm. And, the farmers liked to know the odds of whether the forecasts would be correct. So the team set out by testing seven models over nearly 60 monsoon seasons since 1965 to uncover which were most accurate at predicting the monsoon with the largest lead time and with probabilistic odds. Google’s NeuralGCM—a probabilistic model that offered 32 different predictions—and ECMWF’s AIFS consistently outperformed other AI and conventional models at this task.

“NeuralGCM is designed to provide scientists with an open model that simulates Earth’s atmosphere more quickly and accurately,” says Olivia Graham, Product Manager at Google Research. “Seeing NeuralGCM at work in this impactful way is exactly what we envisioned when we released the model – it’s a great example of the progress we can make when we collaborate across AI experts, academia and domain experts. We’re excited to see how future applications of NeuralGCM unlock new and helpful insights for communities around the world.”

Because each model had different strengths and weaknesses, the team mathematically blended Google’s NeuralGCM, ECMWF’s AIFS and over 100 years of historical rainfall statistics from the India Meteorological Department (IMD). This blend produced a probabilistic model with a 30-day lead time—“merging multiple AI models and statistical methods to produce useful forecasts targeted at agriculture,” says Professor William Boos of the University of California, Berkeley, who specializes in atmospheric dynamics and contributed to the effort. “Forecasts of the start of sustained monsoon rains have historically been difficult or impossible to deliver locally with this much lead time, especially on such a large scale,” he said.

“This approach of blending open-source AI models that were selected based on a rigorous benchmarking process, in this case Google’s NeuralGCM and the ECMWF’s AIFS, followed by bias correction with the IMD’s extensive gridded climate data, is key to generate accurate predictions,” says Dr. Sivananda Pai, Head of Agromet Advisory division at IMD. “This is a new era where artificial intelligence and meteorological expertise converge to transform how we serve India’s farmers.”

Of the work, ECMWF Director-General Florence Rabier, says, “One of our goals is to make forecast data more accessible for all. We are really pleased that our AIFS model is being used in such a way to provide more accurate and timely weather forecasts which will help millions of farmers and other users that most need them. At ECMWF, we have 35 nations working together to spearhead the AI revolution and advance weather science to create impact, not only for the nations represented, but also beyond.”

AI models are not only more accurate but also skip the need for supercomputers for forecast generation, according to Mayank Gupta, a researcher at the Human-Centered Weather Forecasts Initiative. He says, “They can be run on desktops and can be tuned to the specific weather conditions and needs of the citizens on the ground—all at a fraction of the cost and time. AI offers a huge opportunity for technological leapfrogging in forecasting, proving already to be more accurate for many types of weather forecast, and able to be easily expanded to everyone.”

Reaching the Last Mile

The Ministry of Agriculture and Farmers’ Welfare knew a 30-day forecast would be incredibly useful to farmers, so they delivered the forecast to 38 million farmers directly using their SMS platform. The Odisha state government also reached nearly 1 million more through a voice messaging platform. Precision Development (PxD), a global nonprofit supporting smallholder farmers in digital advisory services, led message design and testing.

“A key feature was working with farmers to zero in on their needs and what types of messages would be most useful to act upon,” says Tomoko Harigaya, PxD, chief economist. “Even the most accurate forecast can fall flat if it’s not communicated clearly. This project showed the importance of co-designing messages with farmers.”

The team generated real-time forecasts for several weeks in May to July. Probabilistic forecasts then turned into messages. The farmers received the messages weeks before the monsoon hit. Over the coming months, the researchers will conduct surveys to better understand how many farmers read the messages and changed their behaviors because of it, as well as how the process could be improved. Already, the team is learning that farmers who received the forecasts may have spread the word. That was the case with Parasnath Tiwari, the farmer from Madhya Pradesh.

“I shared the monsoon arrival forecasts with other farmers in my locality. We usually talk to each other and share useful information that we come across,” says Tiwari. “Some farmers have benefited from the information I shared about the arrival of the monsoon. I feel that others will also start relying on this information and trust it for their agricultural decision-making.”

Professor Ramesh Chand, Member of Niti Aayog, the Indian Government’s public think-tank, stressed that focusing on the needs of farmers when providing weather information is essential. “This initiative is tremendously valuable because it centers specifically on the needs of farmers by providing tailored weather forecasts in easy to understand language and helps them make informed farming decisions.”

A Model for the World

“India’s model—led by the government, grounded in farmer needs, and taking cutting-edge science out into the real world—offers a compelling blueprint for other countries.” says Jina.

The researchers behind the project are working with AIM for Scale to scale similar programs in other low- and middle-income countries, using the Ministry of Agriculture and Farmers’ Welfare’s bold leap as a model for innovation. They’re also working with AIM for Scale to train government meteorologists in low- and middle-income countries on how to use AI models effectively.

“This is what practical food security innovation looks like,” said Paul Winters, Executive Director of AIM for Scale and Professor of Global Affairs at the University of Notre Dame’s Keough School of Global Affairs. “It’s not just about new technology—it’s about combining research, government delivery systems, and farmer voices to ensure that timely, useful information reaches those who need it most.”

Jina believes the opportunity expands beyond farmers.

“Our idea is to follow India’s lead and take this all over the world—not just to farmers, but to others as they encounter climate impacts. AI is reframing how we think of weather forecasting and providing a critical tool as people make decisions about how to live with and adapt to climate change.”