IIT Delhi-led Researchers Create AI/ML Model for Rainfall Prediction; Predicts 2023 to be a Normal Monsoon Year

Content Curator

Researchers from IIT Delhi, in collaboration with IIIT Delhi, MIT USA, and JAMSTEC Japan, have created an advanced machine learning model for monsoon rainfall prediction.

IIT Delhi-led Researchers Create AI/ML Model for Rainfall Prediction

New Delhi: A state-of-the-art machine learning model for predicting monsoon rainfall has been created by researchers of the DST Centre of Excellence in Climate Modeling at IIT Delhi, in partnership with IIIT Delhi, MIT USA, and JAMSTEC Japan.

Led by Saroj K. Mishra of IIT Delhi, the research team has successfully developed this advanced model. According to their findings, the monsoon rainfall in India for the year 2023 is expected to be normal, with an estimated All India Summer Monsoon Rainfall (AISMR) of around 790 mm.

The research team comprises Tanmoy Chakraborty from the Department of Electrical Engineering, PhD scholar Pankaj Upadhaya from IIT Delhi, Udit Narang and Kushal Juneja from IIIT Delhi, Swadhin Behera from JAMSTEC Japan, and Popat Salunke from MIT USA.

According to an official statement, the prediction is derived from a model trained using historical AISMR data, Nino 3.4 index data, and categorical Indian Ocean Dipole (IOD) data spanning the period between from 1901 and 2001.

During the test period from 2002 to 2022, the AI/ML model demonstrated a remarkable forecast success rate of 61.9%. This rate was determined by assessing the model's ability to predict the AISMR within a range of +/-5% in comparison to the actual observed values each year. The performance of the AI/ML model demonstrates its capacity to capture nonlinear relationships among the monsoon drivers and its adaptability to accommodate various inputs.

Saroj K Mishra, the Principal Investigator at the DST Centre of Excellence in Climate Modeling and a professor at the Centre for Atmospheric Sciences, IIT Delhi, stated that accurate monsoon forecasts, provided well in advance, are crucial for making informed decisions in various socioeconomic sectors such as agriculture, energy, water resources, disaster management, and health.

The AI/ML model developed by IIT Delhi offers several advantages. One notable advantage is its ability to operate with fewer computational resources, making it a highly efficient alternative to traditional physical models. In contrast to the resource-intensive nature of physical models, the AI/ML model can generate monsoon rainfall forecasts months ahead, contingent upon the availability of Nino3.4 index and IOD forecast data.

Moreover, the model can be easily updated as these factors evolve, as stated in the announcement.

This streamlined process allows a small team to run the model on a personal computer, resulting in the generation of highly accurate monsoon forecasts.

Mishra stated that the data-driven techniques developed in this study will be expanded to offer state-wise monsoon rainfall predictions. This extension aims to enhance the usefulness of the techniques for regional applications.

Read More: