Abstract. For climate change impact assessment, many applications require very high-resolution, spatiotemporally consistent precipitation data on current or future climate. In this regard, stochastic weather generators are designed as a statistical downscaling tool that can provide such data. Here, we adopt the precipitation generator framework of Kleiber et al. (2012), which is based on latent and transformed Gaussian processes, and propose an extension of that framework for a mountainous region with complex topography by allowing elevation dependence in the model. The model is used to generate two-dimensional fields of precipitation with a 1 km spatial resolution and a daily temporal resolution in a small region with highly complex terrain in the Austrian Alps. This study aims to evaluate the model with respect to its ability to simulate realistic precipitation fields over the region using historical observations from a network of 29 meteorological stations as input. The model's added value over the original setup and its limitations are also discussed. The results show that the model generates realistic fields of precipitation with good spatial and temporal variability. The model is able to generate some of the difficult areal statistics useful for impact assessment, such as the areal dry and wet spells of different lengths and the areal monthly mean of precipitation, with great accuracy. The model also captures the inter-seasonal and intra-seasonal variability very well, while the inter-annual variability is well captured in summer but largely underestimated in autumn and winter. The proposed model adds substantial value over the original modeling framework, specifically with respect to the precipitation amount. The model is unable to reproduce the realistic spatiotemporal characteristics of precipitation in autumn. We conclude that, with further development, the model is a promising tool for downscaling precipitation in complex terrain for a wide range of applications in impact assessment studies.
Read full abstract