City to get two hyperlocal rain forecast models for accurate predictions
IIT Bombay team develops hyperlocal weather forecasting system for Mumbai, predicting rain hourly and daily. Flood monitoring also planned.
Mumbai: In order to get accurate and precise weather predictions for specific neighbourhoods, streets and areas across the city, a team from the climate studies department of the Indian Institute of Technology Bombay (IIT-B) have developed a hyperlocal weather forecasting system which will help citizens get localised weather data during the monsoon season.
Apart from the Indian Meteorology Department (IMD)’s main Colaba, Santacruz and Marine Lines Observatory, the city gets its weather forecast from over 60 automatic weather stations (AWS) and rainfall monitoring stations spread across the city and its suburbs.
The first of these hyperlocal weather forecasting systems is expected to go public on a website linked on climate.iitb.ac.in between mid or end of June. The model will give information on the rainfall predicted at over 25 spots on an hourly basis for the next 24 hours, and a day-to-day basis for the next three days. Behind the forecasting model is a team of ten students, faculty and staff from the IIT-B.
“We’ve been working on this for a year and a half,” said professor Subimal Ghosh, the convener behind the climate studies programme, giving the model’s first forecast for Friday, June 7 predicting the onset of monsoon over the city: “a good rain of 10-25 mm across the city.” Chincholi is predicted to receive the highest quantum, at 31.6mm.
Ghosh explained the model calculating the predictions. “We’ve built a physics-based model with a machine learning algorithm that we’ve trained on past rainfall data of the city for a substantial number of years. It will take in publicly available meteorological available data over a large 100 sq km area and output forecasts for hyperlocal spots in the city. Till we go public on the website portal, we will be posting updates on the @ClimateIITB handle on X (formerly Twitter).”
Apart from rain forecasting, the portal also plans to double as a watch for floods.
“We’re in the process of installing eight flood sensors across the city. Apart from this, we will also be crowdsourcing flooding alerts from X and have an online form where people can submit the height of the floods. This will all be automated,” said Ghosh.
The climate studies programme had attempted a crowd-sourced flood monitoring system a few years ago using responses from then-Twitter but did not succeed due to too few responses and expenses after Twitter changed to X.
“This is a completely non-profit and experimental project, which we are going to be improving as the months progress. We really wanted to do something for the city, especially in the tough time of monsoon,” said Ghosh.
The next forecast model will follow in the month of July on the website experiqs.tech, a start up incubated by IIT B. It will display forecasts for over 25 locations in Mumbai for a period of 48 hours.
“We’ve had meetings with the Regional Meteorological Centre, Mumbai, and BMC, who have agreed to share data of the past years of rainfall in the city. Once we receive that, we’ll be training our model, specifically designed for Mumbai’s distinct meteorological conditions, to generate more accurate hyperlocal forecasts,” relayed Sooraj Padinjattayil, co-founder and CEO of Experiqs.
“While it will take around six months to refine our model, which is a physics-informed deep learning model that we’ll feed with data from the RMC’s and BMC’s automatic weather system (AWS) stations, we hope to be accessible to the public in July,” he said.