Improved Treatment of Model Prediction Uncertainty: Estimating Rainfall using Discrete Wavelet Transform and Principal Component Analysis

It is necessary to select appropriate rainfall series as input to the hydrologic model to access more accurate hydrologic predictions and estimate reliable parameters in the modeling process. For achieving this aim, in the present study, the rainfall multipliers with a combination of Discrete Wavelet Transform (DWT) and principal component analysis (PCA) are applied to select effective rainfall series for modeling. DREAM(ZS) algorithm based on the Markov chain Monte Carlo (MCMC) scheme is used to estimate posterior parameters and investigate prediction uncertainties of a five-parameter hydrologic model, HYMOD. The model's results are then compared to those obtained from the other methods that use only the rainfall multipliers or the raw rainfall data. This study reveals the advantages of using a combined application of DWT and PCA methods to estimate hydrological prediction uncertainty and model parameters accurately. Considering the occasional flash flood incident that occurred in the study region (Tamar basin, which is situated in the Gorganroud river basin, Golestan province, Iran), the results of this research can be useful for forecasting floods accurately and planning for flood control management.