CNN based cyclone intensity estimation using satellite imagery

Abstract

Hurricanes are commonly situated over enormous waterways where weather conditions stations are scanty, meteorologists frequently need to appraise the breeze speed of hurricanes. They as a rule use float perceptions, microwave satellite symbolism, and infrared satellite symbolism to make these appraisals. There is developing interest in applying AI and AI procedures to work on the precision of functional meteorological errands, including assessing hurricane wind speed. We started investigating profoundly figuring out how to control hurricane wind speed using convolutional neural networks (CNN) and deep learning techniques. This report presents a framework for evaluating parametric probabilistic models of cyclone wind speeds from existing data on assessed breeze speed with different mean repeat scans (MRIs). The core of the model is a convolutional neural network in which both the residual learning and attention processes are integrated into the version in order to optimise the model's form and validated using k-fold validation. The result shows that our model spot structural changes that occur when tropical cyclones develop, intensify, and approach land. By using CNN modelling, it also provides Tropical Cyclone intensity estimates for the use of satellite TV for PC bands that previously didn't play a major part in quantitative methods.