A Comprehensive Investigation on Cyclone Prediction Using Deep Learning Technique

Weather forecasters have long been concerned with cyclones, and over the recent century, scientists have conducted numerous research on axially symmetric formations, operational dynamics, and methodologies. Several scientists have proven that deep learning can offer a fresh approach to overcoming the limitations of TC prediction by employing a pure data-driven system or enhancing numerical methods with DL. This review introduces development based on DL for various techniques in genesis TC predictions, track forecasting models, intensity estimates, severe weather prediction associated with TCs (including such heavy storms and thunderstorms, and their disastrous effects), and severe storm prediction, in addition to enhancing numerical prediction performance, by summarising and analysing the difficulties of tropical cyclone forecasting methods in recent times and maximizing of techniques using DL in these attributes. The chance is that, as of now, neither the DL capability nor the vast amounts of multi-source data were effectively employed to enhance the precision of tropical cyclone prediction. The difficulty is due to the fact that cyclones vary from typical climatic changes and oceanic mechanisms in their use of complex and dynamic frameworks and are easily persuaded by a wide range of factors. Hence, it might be difficult to ensure the observable duration and reliability of cyclone forecasting.