Wavelet application for improving annual rainfall prediction and investigation of monthly rainfall distribution over a year

Abstract

The annual rainfall prediction is significant in planning for reservoir operation and irrigation areas. In recent years, the technique of wavelet has been widely applied to various water resources research categories because of its time-frequency representation. This study was undertaken to improve annual rainfall prediction of the conventional autoregressive (AR) model by applying wavelet transformation. The 52 year rainfall records of 4 stations distributed over the northeastern part of Thailand were analyzed by the proposed wavelet-AR model (WARM). Two rainfall variables, the number of rainy days per year and the annual rainfall were analyzed in this study. Comparing the obtained R square from WARM with conventional AR, it can be seen that WARM can improve the result of annual rainfall prediction. By applying wavelet to monthly rainfall series, the technique of Unit Disaggregation Curve (UDC) has been developed in this study in order to investigate the pattern of rainfall distribution within a year. The UDC is expressed in the unit curve that is able to represent that pattern of rainfall distribution graph for a year.