Machine Learning-based Post-processing of Sub-seasonal to Seasonal Predictions of Subdivisional Rainfall over India

Lead PI: Dr. Andrew Robertson

Unit Affiliation: International Research Institute for Climate and Society (IRI)

May 2023 - May 2026
Asia ; India
Project Type: Research

DESCRIPTION: The goal of this project is to develop and improve the skill of sub-seasonal (Weeks 3-4; Week 2, 3, 4) probabilistic multi-model precipitation predictions at a subdivisional level for homogenous regions in India. The project datasets will be derived from the MPMME (Multi-Model MultiPhysics Ensemble, IITM Extended Range Prediction System), the S2S (Sub-seasonal to Seasonal Prediction Project), and the SubX (Sub-seasonal Experiment Improving NOAA's week 3-4 forecasts). With the proposed machine learning/artificial intelligence (ML/AI) techniques (Quantile regression Forests: QRF; Convolution Neural network: CNN; and Extended Logistic Regression: ELR), calibration and multi-model ensembling will be performed, forecast skill will be determined, windows forecast of opportunity will be identified, and probabilistic forecast guidance will be developed for sector-specific outlooks with terciles (below normal, near normal and above, normal) and probability of exceedance or non-exceedance of a specific threshold. The study will leverage the outcomes and computational framework of the recent WMO SubSeasonal to Seasonal Prediction Project (S2S) Prize Challenge to improve sub-seasonal to seasonal predictions using Artificial Intelligence.

BROADER IMPACTS: The project outcomes will include a methodology for improved S2S probabilistic predictions of rainfall at a sub-divisional scale for homogenous regions over India for implementation in real-time at IITM and IMD for the benefit of the Indian Society. The methodology will be coded into a general software package for calibration, generation of deterministic and probabilistic forecasts, forecast verification, and visualization of MPMME forecast at S2S timescale, including documentation for straightforward implementation at IITM and IMD. The new methodologies tailored to India will be developed in close collaboration with Dr. Susmitha Joseph, head of the Extended Range Prediction (ERPAS) group at IITM.


Indian Institute of Tropical Meteorology




Susmitha Joseph