Cloudburst prediction in the Indian Himalaya using artificial neural network

This study focuses on developing a machine learning model capable of predicting impending cloudburst events by analyzing key climatological features. We examine 31 historical cloudburst incidents that occurred between 2004 and 2019 in the Indian Himalayas region. These regions are of particular interest due to their unique orographic patterns, which make them highly susceptible to extreme weather phenomena. The raw data is meticulously curated from NASA’s Power Data Access Viewer. Later, it is processed using various feature selection techniques to identify the most influential parameters. The research outlines the data extraction process, feature selection methods, and the preparation of multiple machine learning models. These models range from standalone methods to ensemble, hybrid, and neural network-based approaches. Our experimental results indicate that the neural network performs better as compared to other machine learning based methods on the newly created dataset. The proposed neural network achieves an accuracy of 89.67% in 5-fold cross-validation on this dataset. The study facilitates meteorological agencies, local administrations, disaster response teams, and urban development planners by supporting the enhancement of early warning mechanisms and allowing prompt action in areas vulnerable to cloudburst events.