IIT Bombay unveils AI‑powered rain prediction system

Indian scientists at the Indian Institute of Technology, Bombay have created a cutting-edge AI system that forecasts rainfall up to 90 minutes ahead in Mumbai. This hyperlocal nowcasting tool analyses radar data to predict where rain will fall and issues updates every eight minutes.

How the System Works

The system uses machine learning and weather radar data from the India Meteorological Department (IMD). It looks at rainband patterns, cloud movements, wind speed, and past rainfall to predict upcoming precipitation.

Developed by IIT Bombay’s Centre for Climate Studies and led by post‑doctoral researcher Akshay Sunil, the model has already performed well during tests.

At Mumbai’s Regional Meteorological Centre in Colaba, the system will soon be fully operational and integrated with local IMD radar systems.

Rain Alerts for Mumbai

The forecasts are updated every eight minutes, providing rain predictions up to 90 minutes ahead, offering significantly more timely information than IMD’s current nowcast system (which covers 3–4 hours or 24-hour forecasts).

Mumbai residents can access alerts through the Mumbaiflood.in portal, which already provides detailed rain forecasts and flood information.

Who’s Behind the Initiative

This project is a joint effort by IIT Bombay, IMD, and the Brihanmumbai Municipal Corporation (BMC), supported financially by HDFC ERGO.

The system builds on earlier tools like the Mumbaiflood.in portal and mobile app launched previously, which offer rainfall and waterlogging updates for Mumbai.

Why It Matters

Mumbai’s varied microclimates and sudden rainfall changes often lead to dangerous flooding and commuting issues. With 90-minute lead time and frequent updates, the system helps citizens and authorities respond quickly.

It supports better preparation for flash floods, travel disruptions, and urban waterlogging situations.

What the Future Holds

Right now, the focus is on giving qualitative alerts, that is, warning where rain may occur and when. The team plans to add quantitative predictions (exact rainfall amounts in millimetres) later on, boosting accuracy and utility.