Cyclone Intensity Estimation using Infrared Satellite Imagery

Abstract

Tropical Cyclone is one of the most frequently occurring natural disaster in India. Roughly speaking, every year there are five to six occurrences of tropical cyclones forming, with a ratio of 4:1 between the coastal regions of the Bay of Bengal on the East coast and the Arabian Sea on the West coast. For a long time, experts have attempted to develop a method for analyzing and estimating strength or severity of a Tropical Cyclone (TC) real time. The lack of a method that is universally accepted and consistent for estimating intensity results causes an inherent bias in the estimation process, and the most commonly utilized technique is a manual algorithm that relies on cloud images obtained from satellites. Tropical cyclones, have the potential to generate a multitude of severe weather conditions, including storm's, extremely powerful winds, and lightning strikes. Tropical Cyclones lead to the risk of both human fatalities and substantial property damage. Hence, bias in this field has greater risks. As a part of this study, we have contributed a deep-learning-based state of the art model to aid in estimating strength or severity of a Tropical Cyclone (TC) real time. The model uses imagery captured using infrared technology from INSAT-3D & INSAT-3DR Satellites. The main aim of this study is to estimate cyclone intensity with reduced bias. This work has been done over the Indian Ocean Basin using Mercator projection images for Asian Sector from the India's first telecommunications satellite system, INSAT Satellite. Additionally, we have also supplied this model with a visualization portal that has dedicated pages to promote open-source and contextual information for end users.