Climate data with high spatial and temporal resolution were an important basis for predicting regional climate and water environment changes. The current downscaling targets for GCM data were mainly medium-resolution (MR) reanalysis data, which were still coarse for small basins. We designed a two-step downscaling method in this study, aiming to perform 100 × resolution enhancements of GCM monthly temperature and precipitation data (i.e., target spatial resolution of 1 km × 1 km). We obtained a predictable set of MR climate data by adding a first-step downscaling process, which was then combined with high-resolution (HR) topographic data as auxiliary data in the second-step downscaling process to supplement the spatial and temporal details. The results showed that the performances of two-step downscaling method were better than the one-step downscaling method using only HR topographic data as auxiliary data. Then the two-step downscaling method was applied to downscale the future climate data of four scenarios (SSP1-2.6, SSP2-4.5, SSP3-7.0, SSP5-8.5) during 2015–2099 for the target small basin. We fed the HR climate prediction data into a Bayesian water quality model, which had been successfully applied in this basin, to simulate the responses of hydrology and water quality to climate change. The results demonstrated that the downscaled climate data with high spatial resolution could be successfully used in the water quality model for spatially differentiated simulations of climate change, streamflow change and total nitrogen load change in a small basin.