Estimation of Tropical Cyclone Intensity from Satellite Data using GRU and DBN Approach

Abstract

This research therefore seeks to fill this gap. The ways it can be applied, concerns the examination of GRU and DBN models and satellite imagery data for the assessment of the intensity of tropical cyclones. Not only can the findings of this research help understand the physical process of rapid intensity change of a storm, but it is also valuable to the ability to forecast the tropical cyclone intensity change – a factor incredibly important for local governments to develop their long-term strategies and avoid high losses. A continuous development over the modern computational technology as well as the numerical models have been developed over the several decades which invoked the great interest and extensive research in these specific areas that further enhances the understanding of the formation of tropical cyclones; track and intensity change. As a result of these advances in NWP there has been light shed in the advancements that now allows us to predict that path of a tropical cyclone in large accuracy. Nevertheless, the ability to predict the alteration of the intensity of a tropical cyclone is still one of the tough challenges which weather forecasters are facing up to. This is because there are times a tropical cyclone can intensify within the minimum time and with such intensification not easy to capture by numerical models. Moreover, the choice of the intensity forecast and its accuracy fully depends on the provided initial background information, as well as on the enormous information growth in the weather forecast and in a plenty of other application fields, the increase of which is always associated with the need for the development of more efficient algorithms and the increase in the amount of available computational resources. This paper will be designed to give a brief outline of the results achieved and to show how our approach has been used in two different case-studies. Then it will proceed towards explaining the values obtained through the model parameters and assess the performance of the model based on different conditions. In the last place, the model with largest performance will be used for the prediction of the tropical cyclone intensity of a case data. This work will give empirical evidence to define and solve the problems relevant to the current capability of the tropical cyclone intensity change forecast.