Abstract
Numerical climate models simulate future climate projections, which offer information that can be utilized for climate change adaptation. For projections at regional scales, regional climate models (RCMs) are used to downscale global climate models (GCMs). To assess the models’ ability to simulate future climate, their ability to represent observed historical climate should be evaluated first. When RCMs are driven by a reanalysis instead of a GCM, the RCMs’ performance can be isolated and evaluated separately from the influence of the GCM. This article evaluates the performance of 8 ERA-Interim reanalysis driven EURO-CORDEX RCMs, comparing their simulations of daily precipitation totals, daily temperature averages, maxima and minima over Slovakia to gridded observations CARPATCLIM. The area of Slovakia has complex orography, ranging from lowlands to high mountains (95 to 2654 m MSL). The evaluation of RCMs is detailed, describing both individual and shared RCM biases, including biases in spatial and temporal distribution. The results of the evaluation show that all 8 RCMs show bias in respect to observations, with no RCM performing generally better than other RCMs. Shared biases between the RCMs include smaller precipitation annual cycle amplitude, higher precipitation maxima, larger temperature annual cycle amplitudes and annual interquartile range, higher warm extremes and lower cold extremes of daily temperature averages, higher maxima and lower means of daily temperature maxima, more temporally clustered ‘coldest days’. For minima of daily temperature minima, RCMs show large opposing biases. In contrast to observations and all other RCMs, REMO2015 simulates precipitation fields nearly uncorrelated with elevation, RACMO22E substantially underestimates daily temperature averages.