MPM-Net: a Data-Driven Approach for Forecasting Indian Heatwave and Cold Wave Events Using Dehazing and Ensemble Learning Technique

摘要

Abstract This paper proposes a data-driven approach for forecasting Indian heatwave and cold wave events by dehazing high-resolution multispectral remote sensing images. The proposed method utilizes the Multi-Path Multi-Scale Dehazing Network (MPM-Net). This well-trained network provides a precise calculation of atmospheric light and transmission maps, resulting in high-resolution quality restoration of hazy images. The denoised images extract dynamic climatic features integrated with an ensemble learning global protective intelligent algorithm. This approach identifies the complex relationship between daily and past events on the measuring timeline, leading to a solid prediction of the heatwave and cold wave events. Furthermore, it facilitates solid control measures based on the prediction results. Evaluation outcomes reveal that the proposed technique notably reduces prediction errors compared to existing deep learning models. This research significantly contributes to the weather forecasting niche and emphasizes the vital role of dehazing in improving the accuracy of predictions.