New Delhi: India tested an AI-based monsoon forecasting pilot for agriculturally sensitive regions ahead of Kharif 2025. The Development Innovation Lab–India partnered in the initiative, which covered parts of 13 states. The project used an open-source blended model that combined Google’s NeuralGCM, ECMWF’s Artificial Intelligence Forecasting System, and 125 years of rainfall data from the India Meteorological Department.
Mid-season decisions guided through AI-based monsoon forecasting
The pilot produced probabilistic predictions only for the local monsoon onset, a factor that directly affects sowing dates. Officials sent these localised forecasts through SMS to 3,88,45,214 farmers via the M-Kisan portal. The messages went out in five regional languages—Hindi, Odia, Marathi, Bangla and Punjabi. The project did not offer any financial support to farmers.
Kisan Call Centres gathered feedback through telephonic surveys in Madhya Pradesh and Bihar. The responses showed that 31–52% of farmers changed their planting decisions after receiving the forecasts. Many of them adjusted land preparation and sowing timelines. Their decisions also influenced crop choices and input use.
