Downscaling of daily precipitation with a stochastic weather generator for the subtropical region in South China

Abstract

Abstract. Daily precipitation series at station or local scales is a critical input for rainfall-runoff modelling which, in turn, plays a vital role in the assessment of climate change impact on hydrologic processes and many other water resource studies. Future climate projected by General Circulation Models (GCMs) presents averaged values in large scales. Therefore, downscaling techniques are usually needed to transfer GCM-derived climate outputs into station-based values. In this study, a statistical downscaling model is investigated and its applicability in generating daily precipitation series is tested in the subtropical region of South China, which has not been investigated before. The model includes the first-order Markov chain for modeling wet day probability, Gamma distribution function for describing variation of wet-day precipitation amounts, and a statistical downscaling approach to transferring large-scale (in both space and time) future precipitation series from GCM climate change scenarios to station or local scales. A set of observed daily precipitation series of 32 years from 17 rainfall stations in and around a grid of 2.5° in latitude by 3.75° in longitude in Guangdong province of China is used to evaluate the model accuracy and validate the downscaling results. The downscaled daily precipitation series and the extreme precipitation features (including maximum, maximum 3-day average and maximum 7-day average) are compared with the observed values. The results show that the proposed model is capable of reproducing the mean daily amount and model parameters of the daily precipitation series at station or local scales in the study region.