Precipitation and Its Future Changes in the Greater Alpine Region: High-resolution Bias-adjusted Versus Dynamically Downscaled Datasets

Weather and climate in mountainous regions are strongly affected by topography, which shapes temperature, precipitation, and wind systems on local scales. The topography can trigger and exacerbate extreme events such as downslope windstorms and heavy convective precipitation. These regions are also particularly sensitive to climate change - higher elevations generally experience faster warming. Nevertheless, our understanding of local-scale atmospheric phenomena in complex terrain remains limited, partially because reliable observations are sparse, and most climate simulations are too coarse to resolve the relevant processes. Recent advances in computational power and various downscaling techniques have partly alleviated this problem, giving rise to multiple multi-model ensembles of km-scale climate simulations (<4 km grid spacing). Such ensembles enable us to study atmospheric processes characteristic for complex terrain in unprecedented detail.

In this work, we use three km-scale datasets: the dynamically downscaled, convection-permitting CORDEX-FPS ensemble on convective phenomena over the greater Alpine region, and two statistically downscaled and bias-adjusted datasets used in the national climate scenarios of Austria and Switzerland (OeKS15 and CH2018, respectively). For comparison, we also analyze the three coarser-resolution ensembles from which these km-scale ensembles were downscaled. We assess to what degree these different ensembles are able to capture various daily precipitation indices, and their dependence on temperature and elevation. We discuss how credible these datasets are when evaluated against observations, and we examine how the precipitation characteristics are projected to change in the warming climate. 

Our findings show that the spatial patterns of the analyzed precipitation indices are fairly similar among the km-scale ensembles in the evaluation period. However, we find differences between the datasets at low temperatures - compared to observations, the dynamically downscaled ensemble strongly overestimates daily precipitation intensity and frequency, whereas the bias-adjusted datasets underestimate these. The dynamically downscaled ensemble also shows biases at high elevations. In the climate change projections, we see notable season-dependent differences between the datasets, and some of the bias-adjusted models exhibit spurious signals.