Improving a stochastic multi-site generation model of daily rainfall using discrete wavelet de-noising: a case study to a semi-arid region

Abstract

This paper presents a combined model of wavelet transform analysis and multi-site stochastic generation model of daily rainfall data. The followed methodology demonstrates the feasibility of the discrete wavelets transform for de-noising process as part of hydrological data processing. The classical two-part stochastic model has been investigated at the basin scale to simulate multi-site daily rainfall time series, which is among the most useful information used in water resources management. However, despite the importance of such information, as far as we know, no previous contribution has been conducted to tackle the issue for Algerian watershed. Such requirement has spurred the authors to investigate the performance of a multi-site stochastic process in this study area. Therefore, to develop this investigation, a modified stochastic model based on Wilks approach was adopted. According to the literature, the adopted model led to good results in solving the adverse effect of random noise problem. Overall, the validation of the statistical characteristics of the obtained generated series demonstrates that the model performs very well with noisy data as well as with the de-noised ones. Furthermore, the use of the pre-treated daily rainfall data in the stochastic multi-site model has contributed to improve substantially the spatial dependency results for the semi-arid study area.