IIT Delhi’s AI model predicts that 2023 will be a normal monsoon year
A new climate model developed by a research team led by IIT Delhi professor Saroj K. Mishra has predicted that 2023 will be a normal monsoon year. It says that the All India Summer Monsoon Rainfall (AISMR) of about 790 mm for the approaching monsoon season means normal monsoon rainfall.
‘The forecast is made using a model trained on historical AISMR data, Niño3.4 index data, and categorical Indian Ocean Dipoles (IOD) data for the period 1901-2001,’ said an IIT Delhi press release. Researchers at IIT Delhi’s Center of Excellence for Climate Modeling in collaboration with IIIT Delhi, MIT USA, and JAMSTEC Japan have developed a state-of-the-art machine learning model to forecast monsoon rains.
The developed and tested AI/ML model proved to be more effective than the current physical models used in the country for monsoon forecasting. It showed a remarkable 61.9 percent prediction success rate over the 2002-2022 test period. This is based on whether the model can predict the AISMR within /-5% of the actual values observed each year.
With the AI/ML model, the forecast can be made months in advance, depending on the availability of the Niño3.4 index and the IOD forecast. It can be updated according to their progress. Thus, data-driven models are flexible concerning inputs and can better capture nonlinear relationships between monsoon drivers, while being less computationally demanding.
A handful of people running these models on a personal computer for a limited period can provide a more accurate monsoon precipitation forecast than the resource-intensive process associated with traditional physical models.
‘This research has immense significance for the entire country, as accurate monsoon forecasting on time is key to key decisions in various socio-economic sectors including agriculture, energy, water resources, disaster management, and health,’ said Prof. Saroj K. Mishra, PI, DST Center of Excellence for Climate Modeling and Professor, Center for Atmospheric Sciences, IIT Delhi.
Professor Mishra added that data-driven techniques are being extended to provide state forecasts of monsoon rains, making them more useful for regional applications.
Professor Tanmoy Chakraborty of the department is also part of the research team. Elektrotekniko erialal ja dr Pankaj Upadhaya IIT Delhist, Udit Narang kaj Kushal Juneja IIIT Delhist, prof. Swadhin Behera from JAMSTEC Japan, and Popat Salunke from MIT – USA.