Answers are grounded in this paper and cite their sources.
In regions with complex terrain, high-resolution models are necessary to study regional climate and climate change impacts. Due to the high computational costs of dynamical downscaling, statistical downscaling techniques are often preferred to generate information at the kilometer-scale, despite their known limitations. While the benefits of kilometer-scale modeling for precipitation and wind are well-established, temperature is rarely in the focus of evaluation. Thus, here we assess the representation of minimum and maximum daily temperatures in high-resolution regional climate models (RCMs) at the kilometer-scale in the Alpine countries Austria and Switzerland. We compare dynamically downscaled, convection-permitting RCMs with model ensembles which have been downscaled by applying bias adjustment methods. Dynamically downscaled kilometer-scale RCMs exhibit considerable season- and elevation-dependent temperature biases. A percentile-based evaluation approach demonstrates that these models tend to overestimate hot and cold extremes. Furthermore, despite being designed to replicate observed statistics, bias-adjusted models still show distributional artefacts of the raw input models. While convection-permitting models have been shown to greatly improve the representation of variables with high spatio-temporal variability, temperature is still prone to substantial biases. In their current state, dynamically downscaled models do not offer a more accurate representation of temperature in these regions than bias-adjusted output. Our results thus underscore the importance of model tuning and motivate the adaptation and development of parameterization schemes tailored to complex topography.
No takes yet. Share an insight, caveat, or question.
Kohlhauser et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: