Rainfall data analyzing using moving average (MA) model and wavelet multi-resolution intelligent model for noise evaluation to improve the forecasting accuracy

Rainfall forecasting and approximation of its magnitude have a huge and imperative role in water management and runoff forecasting. The main objective of this paper is to obtain the relationship between rainfall time series achieved from wavelet transform (WT) and moving average (MA) in Klang River basin, Malaysia. For this purpose, the Haar and Dmey WTs were applied to decompose the rainfall time series into 7, 10 different resolution levels, respectively. Several preprocessing case studies based on 2-, 3-, 5-, 10-, 15-, 20-, 25-, and 30-month MAs were carried out to discover a longer-term trend compared to a shorter-term MA. The information and data were gathered from Klang Gates Dam, Malaysia, from 1997 to 2008. Regarding the behavior, the time series of 10-, 15-, 20-, and 30-day rainfall are decomposed into approximation and details coefficient with different kind of WT. Correlation coefficient R ² and root-mean-square error criteria are applied to examine the performance of the models. The results show that there are some similarities between MA filters and wavelet approximation sub-series filters due to noise elimination. Moreover, the results obtained that the high correlation with MAs can be achieved via Dmey WT compared to Haar wavelet for rainfall data. Moreover, clean signals could be used as model inputs to improve the model performance. Therefore, signal decomposition techniques for the purpose of data preprocessing could be favorable and could be appropriate for elimination of the errors.