As global-scale climate scenarios at high spatial detail are not yet available due to existing limitations in computational resources, regional climate models are often applied to allow fine-scale consideration of climate change at the regional scale. This chapter describes the derivation of climate variants from the simulations of the two regional climate models REMO and MM5, supplementing the climate variants generated by applying predefined climate trends in combination with a statistical climate generator (Chaps. 49 and 50). To overcome limitations arising from the fact that (1) the small-scale climatic variability, primarily in the southern Alpine regions of the study area, cannot be reproduced despite the comparatively high spatial resolution of the regional climate models and (2) systematic deviations between meteorological simulations and observations for the past exist (often referred to as “biases”), a method for refinement (downscaling) and bias correction of RCM data is described that is applied to the RCM data prior to its application as input for DANUBIA. Biases in the applied RCM data are shown exemplarily for simulated precipitation. Moreover the effect of bias correction on simulated discharge is illustrated by comparing the discharge duration curve simulated by DANUBIA with and without correction of biases in the RCM data. The climate change signal characterising the climate variants based on scaled and bias-corrected REMO and MM5 data is analysed by considering changes in annual and monthly mean temperature and precipitation as well as spatial patterns and seasonal changes in temperature and precipitation change in the Upper Danube watershed.