Artificial intelligence is playing a key role in improving weather prediction for farmers in India. This spring, an AI-based model successfully predicted a delayed onset of the monsoon in northeastern India, allowing millions of smallholder farmers to adjust their planting strategies. A preliminary phone survey found that many farmers changed their planting plans based on the forecast.

The project is a collaboration between the University of California, Berkeley, and the University of Chicago. It aims to provide better forecasts for hundreds of millions of farmers across tropical regions whose livelihoods depend on the timing of monsoon rains. Climate change has made these annual patterns less predictable.

“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,” said Pramod Kumar Meherda, additional secretary at the Indian Ministry of Agriculture and Farmers’ Welfare.

The initiative involved atmospheric scientists, AI experts, India’s Ministry of Agriculture and Farmers’ Welfare, and Precision Development (PxD), a nonprofit supporting smallholder farmers. Daily climate data from the U.S. National Oceanic and Atmospheric Administration was crucial for these predictions.

To make accurate forecasts, Pedram Hassanzadeh from UChicago worked with William Boos from UC Berkeley to evaluate global AI weather models developed by Google and ECMWF (European Centre for Medium-range Weather Forecasts). These models were trained on 40 years of global climate data. The team used 100 years of rainfall statistics from the India Meteorological Department to adapt predictions specifically for India.

Forecasts were sent weekly to about 38 million farmers across 13 states in central and northeastern India — most within the core monsoon zone — providing up to four weeks’ notice on when rains would arrive in each region. Traditional models usually offer reliable rainfall predictions only five days ahead.

When southern India’s monsoon began in early June, the AI model predicted a temporary stall that lasted 20 days — something no other forecast had indicated.

“Demonstrating that the long lead-time precipitation forecasts made by these AI models are of practical use in a tropical region where people live is a major step forward — no one really knew that before we did this work,” said Boos, professor at UC Berkeley.

Parasnath Tiwari, a farmer from Madhya Pradesh who received forecasts via phone messages, said he switched crops after gaining confidence from the prediction: “Before this, I mostly relied on my own experience and local knowledge to know when the monsoon would arrive,” Tiwari said. “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.”

Farmers received updates mostly weekly between May and July about when steady rain would likely begin. The team communicated probabilistic forecasts so that recipients understood both what was being predicted and how certain it was.

Boos explained why accurate timing matters: “The classic catastrophe scenario is that you get a wet spell...they plant their seeds...and then there’s 15 days of dryness afterward and all the seeds dry out and die...that’s a huge loss.”

Michael Kremer, Nobel Prize-winning economist at UChicago and project leader, concluded that improved rain predictions could bring economic benefits for rural Indian farmers. “We have been going through an AI-driven revolution since 2022...But their ability to predict complex phenomena — like the monsoon — was unclear, and frankly unexpected,” said Hassanzadeh.

After testing several models capable of month-long forecasts with rainfall predictions, Boos and Hassanzadeh selected Google’s NeuralGCM model along with ECMWF’s AIFS system as best suited for this task. They combined these tools with historical data using mathematical blending techniques.

This produced probabilistic models able to give locally relevant guidance up to 30 days ahead — something not previously possible at such scale or lead time.

The Indian Ministry used its SMS platform for delivery; Odisha state government reached nearly one million more through voice messaging; PxD led message design efforts. Early results showed up to 55% recalled receiving weather messages; nearly half who remembered specific onset forecasts reported changing planting decisions accordingly.

“I shared the monsoon arrival forecasts with other farmers in my locality...Some farmers have benefited from the information I shared about the arrival of the monsoon...” Tiwari added.

“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,” Kremer said. “India is leading the way in using AI to improve people’s lives across many sectors, including agriculture.”

Funding came partly from AIM for Scale — backed by organizations like the Gates Foundation — which works globally to expand cost-effective agricultural innovations among low- and middle-income countries.

Researchers now plan similar programs elsewhere and aim to train meteorologists across developing nations on effective use of AI models for weather forecasting (University of Chicago announcement).

“One of things we would like to do for future years...is be able to predict dry spells throughout entire summer...” Boos said.