Efficacy of machine learning in simulating precipitation and its extremes over the capital cities in North Indian states

Efficacy of machine learning in simulating precipitation and its extremes over the capital cities in North Indian states

Abstract Climate change-induced precipitation extremes are a pressing global concern. This study investigates the predictability of precipitation patterns and extremes across North Indian states from 1984 to 2023 using NASA’s Prediction of Worldwide Energy Resources (POWER) datasets and machine lear...

Full description

Saved in:
Bibliographic Details
Main Authors: Aayushi Tandon, Amit Awasthi, Kanhu Charan Pattnayak
Format: Article
Language:English
Published: Nature Portfolio 2025-03-01
Series:Scientific Reports
Subjects:
Climate change
Machine learning
North Indian states
Precipitation patterns
Random forest
Support vector machine
Online Access:https://doi.org/10.1038/s41598-024-84360-w
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Abstract Climate change-induced precipitation extremes are a pressing global concern. This study investigates the predictability of precipitation patterns and extremes across North Indian states from 1984 to 2023 using NASA’s Prediction of Worldwide Energy Resources (POWER) datasets and machine learning (ML) models. The current ML model builds on the relationship between rainfall and key climatic parameters such as dew point temperature and relative humidity, showing a strong positive correlation (CC = 0.4) significant at the 0.05 level. In simulating precipitation, Random Forest Classifier (RFC) achieved the highest accuracy (~ 83%) for Rajasthan and Uttar Pradesh, while Support Vector Classifier (SVC) performed best (79–83% accuracy) in other states. However, ML models exhibited approximately 5% lower skill in higher elevated stations as compared to lower ones, due to differing atmospheric mechanisms. For extreme precipitation events (10th and 95th percentiles of intensity), RFC consistently outperformed SVC across all states showing superior ability to distinguish extreme from non-extreme events (Area Under Curve ~ 0.90) and better model calibration (Brier Scores ~ 0.01). The developed ML models effectively simulated precipitation and extreme patterns, with RFC excelling at classifying extreme events. These findings can aid disaster preparedness and water resource management in regions with varied topography and complex terrain.
ISSN:2045-2322

Similar Items