Evaluation of Minimum and Maximum Temperatures in Convection-Resolving Climate Models (CORDEX-FPS Convection)

In the evaluation of high-resolution climate model output, most research focuses on the variables and time-scales where added value is usually expected, e.g. short-term precipitation extremes. Simple temperature characteristics such as means of minimum and maximum temperatures are rarely evaluated, even though shortcomings in the representation of temperature might negatively influence multiple physical processes in the models.

In our research we analyse the representation of minimum and maximum temperatures in convection-resolving models in Austria. We make use of the ERA-Interim driven CORDEX-FPS Convection ensemble in the period 2000-2009, covering the greater Alpine region, and compare it against several available observation-based datasets - ERA5, EOBS, EMO-5 and SPARTACUS. 

Using the Austrian climate dataset SPARTACUS as the main reference, we compute the seasonal means of maximum and minimum temperatures and identify season dependent biases. We find notable differences between the CORDEX-FPS simulations and SPARTACUS, however the observations exhibit a certain spread as well. While maximum temperatures are underestimated in winter and spring, minimum temperatures are overestimated in summer and autumn. Consequently, we find that the diurnal temperature range is underestimated throughout the year. We presume that these biases are caused by parameterizations of radiation and cloud-related properties.

Additionally, we investigate the elevation-temperature relationship in the model ensemble and the observations. We identify an elevation-dependent bias in the convection-resolving models for both minimum and maximum temperatures. The difference between the model ensemble and SPARTACUS becomes more negative with higher elevations. As a consequence, the near-surface temperature lapse rate is constantly overestimated in the model ensemble. We assume this might be caused by inadequate parameterizations as well, and potentially the representation of the annual snow cover.

We explore the correlations between the aforementioned properties and parameterized processes like radiation and cloud cover, in order to get a deeper understanding of the physical processes in the models.