At the new system’s core are two forecasting models whose predictions are “blended,” to sharpen accuracy. File
| Photo Credit: The Hindu
Ahead of the monsoon this year, the India Meteorological Department (IMD) on Tuesday unveiled a new forecast system that will, for the first time ever, generate ‘block’ level forecasts of the monsoon’s arrival over 15 States and comprising about half of India’s roughly 7,200-odd blocks.
Historically such estimates are available, at best, over States or at the level of districts. For instance, it is known that the monsoon arrives in Mumbai around the 10 of June and Delhi, the 29 of June. However the inherent variation of the monsoon is such that even within the same district, several blocks and villages in them will be rainless despite the monsoon having officially ‘arrived’ at the district borders.
It has been a long-standing aim of the IMD to provide hyper local forecasts to address this shortcoming to enable farmers to time their sowing precisely.
At the new system’s core are two forecasting models whose predictions are “blended,” to sharpen accuracy. From the date of the monsoon’s onset in Kerala, it can use AI-based analysis, IMD’s trove of nearly a century of detailed meteorological data and global weather models to give the monsoon’s itinerary with unprecedented granularity, Science Minister Jitendra Singh said at a press briefing.
Forecasts for 4 weeks
This was a system specifically developed at the request of the Ministry of Agriculture and Farmers’ Welfare, whose existing advisory system is built to deliver forecasts in a roughly weekly format. The blending framework, developed by the Indian Institute of Tropical Meteorology, a research institute of the Ministry of Earth Sciences, is designed to feed directly into the Ministry’s pipeline and issue probabilistic forecasts for the next four weeks.
Currently, this system can be used to provide forecasts to 3,196 blocks across 15 States and one Union Territory. Two trial runs have already been completed successfully, according to a press statement. “These States are part of the monsoon core zone, which are the regions that are largely rainfed and are most sensitive to southwest monsoon dynamics,” said M. Ravichandran, Secretary, MoES at a press briefing. “Of course, going ahead we aim to extend this all over India but that requires more observational data.”
Mr. Ravichandran told The Hindu that the system would face a formidable test this year given that the IMD as well as global models were expecting “below normal” rainfall in the light of a developing El Nino — frequently causing weak monsoon rain in India — from the month of July.
On Tuesday, the IMD also launched a monsoon forecast model specifically for Uttar Pradesh with a 1-km resolution (indicative of granularity) that is valid for 10 days. This, Mr. Singh said, was because of a very extensive coverage of automatic weather stations in the State that allowed a weather model called Mithuna (which works at a 12.5 km resolution) to be “downscaled” to 1 km. “We are encouraging other States to share their data with us that will allow their forecasts to be generated with higher resolution,” Mr. Ravichandran added.
Published – May 12, 2026 10:46 pm IST
