E-Paper
IMD to experiment with AI as alternative to supercomputers in weather forecasts
However, better data is needed to make the most out of artificial intelligence for weather forecasting, say experts.
The Indian Meteorological Department (IMD) is experimenting with artificial intelligence (AI) to enhance climate models for better weather forecasting.
Currently, IMD relies on supercomputers and mathematical models for forecasts. Integrating AI with an extended observation network could improve forecast accuracy and reduce costs.
Also read: CM Stalin says IMD failed to predict the rains correctly
AI to enhance forecast at lower cost
According to KS Hosalikar, head of climate research and services at IMD, AI-based climate models and advisories being developed will enhance forecasts.
The weather office, led by Hosalikar, has utilised AI for public alerts on heatwaves and diseases like malaria. There are plans to expand weather observatories, offering data at the village level for potentially more precise forecasts.
The agency aims to integrate AI into traditional models for weather and climate forecasts, and has established a centre for testing through workshops and conferences.
"An AI model doesn't require the high cost involved in running a supercomputer - you can even run it out of a good quality desktop," said Saurabh Rathore, an assistant professor at Indian Institute of Technology-Delhi.
Also read- Year Ender 2023: Biggest natural disasters that took place this year
AI models for weather forecasts
DeepMind, a part of Google, has created an AI model named GraphCast that swiftly generates precise 10-day forecasts by analysing decades of past weather data.
In a study, GraphCast excelled in approximately 90 per cent of tested metrics compared to the European Centre for Medium-Range Weather Forecasts' gold-standard model. Notably, it demonstrated strong performance in predicting severe weather conditions.
Despite its success, experts emphasise the necessity of improved data for maximising the potential of AI in weather forecasting.
"Without having high-resolution data in space and time, no AI model for location-specific magnification of existing model forecasts is feasible," said Parthasarathi Mukhopadhyay, a climate scientist at the Indian Institute of Tropical Meteorology.
(Inputs from Reuters)
