The application of symbolic dynamics in rainstorm prediction

Abstract

Rainstorm is one of the major natural disasters in the world and lead huge losses of national economy and people's life and property every year. The prediction of rainstorms is very difficult using the current methods because of the atmospheric system's complexity and non-linearity. In recent years, nonlinear science has developed rapidly and nonlinear time series analysis has been widely used in many scientific and technological fields. Symbolic dynamics is a branch of nonlinear science and have gradually become a tool of time series analysis. In this paper the symbolic dynamics method was introduced in detail, the process including interpolation, de-noising, segmentation, symbolization, coding and statistical analysis. Using symbolic dynamics to explore storm event has some certain significance, the entropy curves of a large number of heavy rain events were analyzed and the entropy reached its minimum before most of the heavy rainfall. So this characteristic can be treated as the symptom the prediction of rainstorms. In this paper 158 global rainstorm events are used to test symbolic dynamics method. The results show that 107 rainstorms events appear obvious symptom and the prediction accuracy reach 67.7%. So it is valuable in monitoring and forecasting heavy rain events.