The weather generator is an essential process in water resource assessment. Most weather generators focus on extreme rainfall events and rainfall amounts in a relatively short time scale. However, drought events often last more than several months, which conventional weather generators hardly generate. Conventional weather generators assume that monthly rainfalls are independent, skewing drought event generation. The purpose of this study is to construct a weather generator with improved drought property generation, combining with monthly rainfall data from conventional weather generators and characteristics of standardized precipitation indices. The proposed weather generators employs four drought parameters, namely starting month, duration, average, and minimum standardized precipitation indices, generated using a copula method. Analytical results show that the four parameters generated by the copula method are consistent with historical records. The proposed weather generator overcomes the limitation of conventional weather generators and can generate both rainfall and drought properties. The results also indicate that the assumption of monthly independence in drought generation can cause underestimated occurrence and duration of drought events. The proposed generator is also demonstrated for climate change assessment. The analytical results show that the uncertainties from the selection of weather generators are even higher than those from the selections of global circulation models while applying to water shortage assessment. We therefore suggest that weather generators should consider drought characteristics which can be measured using the standardized precipitation index to reduce the uncertainty.
Keywords: Climate change; Downscaling; Drought characteristics; SPI; Water shortage.