Earth, Atmosphere & Environment Sciences
Title : | Prediction of Indian Summer Monsoon Active and Break Cycles using Deep Learning. |
Area of research : | Earth, Atmosphere & Environment Sciences |
Focus area : | Atmospheric Science |
Principal Investigator : | PI: Prof. Hariprasad Kodamana, Assistant Professor, Department of Chemical Engineering and School of Artificial Intelligence , IIT Delhi, Hauz Khas, New Delhi-110016 |
Timeline Start Year : | 2022 |
Timeline End Year : | 2022 |
Details
Executive Summary : | The main focus of this study is modelling of India summer monsoon rainfall (ISMR) by deep dynamic recurrent neural network such as LSTM and its variants. As per various studies reiterated, ISMR exhibits wide range of variability in time scales, ranging from daily to centennial. Due to effect of multitude of factors that affect variability, flotations in the monsoon have resulted in unprecedented spells of draught and floods. As it is widely accepted that rainfall is the most important exhibition of the monsoon given its direct socio-economic impact, it is imperative to have a model that accurately predicts monsoon rainfall. Traditional approach to achieve this is means of performing ensemble simulations using Generalized Circulation Models. Nevertheless, historical data repositories carry signatures of past monsoon events and their latent dynamics. In this work, firstly, departing from conventional approaches, the historical datasets of about past 25 years to develop data-driven models, using advanced machine learning (ML) techniques that uses deep learning techniques which are capable for long range predictions. Secondly, the outputs from GCM and ML-based models, are combined by various statistical techniques. |
Co-PI: | Sandeep Sukumaran, Assistant Professor, Center for Atmospheric Sciences, IIT Delhi, Hauz Khas, New Delhi-110016 |
Total Budget (INR): | 25,41,108 |
Organizations involved