Classification of cyclones using machine learning techniques

Abstract

In this article, we provide a method for identifying and categorizing cyclones, both tropical and extratropical. The method is designed with the goal of producing a global labeled dataset for cyclones, and it is based on a set of rigorous criteria. The heuristics are defined from date, time, pressure, wind speed, wind directions, latitude and longitudes. Numerous researchers have confirmed that machine learning, a kind of artificial intelligence, can offer a fresh approach to overcoming the limitations of cyclone classification, whether employing a pure data-driven model or enhancing numerical models with machine learning. This article introduces progress based on machine learning in genesis classification, track records, intensities, and extreme weather forecasts associated with tropical as well as extratropical cyclones (such as strong winds and rainstorms and their disastrous impacts). The challenges of cyclones in recent years and successful cases of machine learning methods in these aspects are summarized and analyzed.