Comparison of tropical thunderstorm estimation between multiple linear regression, Dvorak, and ANFIS

Abstract

Thunderstorms are dangerous and it has increased due to highly precipitation and cloud cover density in the Mesoscale Convective System area. Climate change is one of the causes to increasing the thunderstorm activity. The present studies aimed to estimate the thunderstorm activity at the Tawau area of Sabah, Malaysia based on the Multiple Linear Regression (MLR), Dvorak technique, and Adaptive Neuro-Fuzzy Inference System (ANFIS). A combination of up to six inputs of meteorological data such as Pressure (P), Temperature (T), Relative Humidity (H), Cloud (C), Precipitable Water Vapor (PWV), and Precipitation (Pr) on a daily basis in 2012 were examined in the training process to find the best configuration system. By using Jacobi algorithm, H and PWV were identified to be correlated well with thunderstorms. Based on the two inputs that have been identified, the Sugeno method was applied to develop a Fuzzy Inference System. The model demonstrated that the thunderstorm activities during intermonsoon are detected higher than the other seasons. This model is comparable to the thunderstorm data that was collected manually with percent error below 50%.