With the advancement of an AI-based weather model at UC Berkeley and the University of Chicago, proactive decisions could be taken regarding farming, as they were able to forecast a 20-day stall in the Indian monsoon. About 38 million farmers in central and northeastern India received monsoon onset predictions of up to 4 weeks in advance, provided via weekly SMS forecasts by the Ministry of Agriculture and Farmers’ Welfare. The forecasts enabled farmers to shift their crop adjustments and planting schedules, making them more resilient to climate change. Probabilistic forecasts were provided through a 30-day model that combined differing data sets with Google’s NeuralGCM.

The model corresponds with India’s ECMWF AIFS, and the Indian Meteorological Department rainfall data. The Ministry of SMS facilitated messaging, while other regions, such as Odisha, received the information through the voice messaging system, with PxD constructing and verifying the messages sent. The forecasts were recalled by 55% of the farmers and almost half of them used the information to base their planting decisions. This initiative proves that AI-enabled weather forecasts, especially tailored for the tropics and global south, can be of considerable aid to agriculture. There are incubation and expansion plans in moderation for other regions as well as for predicting dry spells.