Numerous statistical downscaling models have been applied to impact studies, but none clearly recommended the most appropriate one for a particular application. This study uses the geographically weighted regression (GWR) method, based on local implications from physical geographical variables, to downscale climate change impacts to a small-scale catchment. The ensembles of daily precipitation time series from 15 different regional climate models (RCMs) driven by five different general circulation models (GCMs), obtained through the European Union (EU)-ENSEMBLES project for reference (19601990) and future (2071-2100) scenarios are generated for the Omerli catchment, in the east of Istanbul city, Turkey, under scenario A1B climate change projections. Special focus is given to changes in extreme precipitation, since such information is needed to assess the changes in the frequency and intensity of flooding for future climate. The mean daily precipitation from all RCMs is under-represented in the summer, autumn and early winter, but it is overestimated in late winter and spring. The results point to an increase in extreme precipitation in winter, spring and summer, and a decrease in autumn in the future, compared to the current period. The GWR method provides significant modifications (up to 35%) to these changes and agrees on the direction of change from RCMs. The GWR method improves the representation of mean and extreme precipitation compared to RCM outputs and this is more significant, particularly for extreme cases of each season. The return period of extreme events decreases in the future, resulting in higher precipitation depths for a given return period from most of the RCMs. This feature is more significant with downscaling. According to the analysis presented, a new adaption for regulating excessive water under climate change in the Omerli basin may be recommended.