Forecast of the trajectory of the center of typhoons and the Maslov decomposition

Abstract

We predict the trajectory of the center of a typhoon by using the coordinates of the first three positions of the center. From this information, we obtain the initial distribution of the wind velocity using a neural network trained for solving this inverse problem. We take the wind field at the initial time as the sum of a smooth part and a singular part, according to the Maslov theory. This form of the field ensures the stability and self-similarity of the flow. The trajectory is found by solving the shallow water equations numerically. In some cases, the resulting trajectory approximates the actual trajectory fairly well.