One major challenge in climate‐related issues is the lack of accurate high‐resolution precipitation data on the part of observations and climate model simulations, imposing profound constraints on various impact studies, for instance in the fields of hydrology, glaciology, agronomy and hazard assessment. Here, we present a physical model approach for orographic precipitation based on linear theory and combined with a stochastic weather generator which takes an intermediate position between computationally expensive high‐resolution regional climate models and purely statistical downscaling techniques. The model is applied to a mountainous region in central Germany and run over a decade in a 0.005° resolution (≈500 m). The resulting precipitation fields are spatially and temporally compared with station data from a very dense meteorological network and with a 6 km regional reanalysis.
The spatial correlation between local station data and corresponding model grid boxes is between 0.72 and 0.80. The presented linear model also agrees well with the observed spatial pattern of day‐to‐day variability, mean intensity, weak and extreme precipitation events. It is close to the performance of the high‐resolution regional reanalysis COSMO‐REA6, yet computationally far more efficient. Combined with the weather generator, the linear model matches the statistical properties of the station data almost perfectly and offers a consistent precipitation dataset at 0.005° resolution. The model can be applied to any larger‐scale model or observational dataset, including climate change projections.