Sources of temperature biases in regional climate simulations over complex terrain

In the frame of the worrying pace and effects of human induced climate change, mountainous regions are warming about twice as fast as the global average. Modeling climate and climate change scenarios over highly complex topography regions like the Alps still represents a great challenge to regional climate modeling: better characterizing the sources of models biases remains a major issue. We analyze the sources of the bias affecting the near surface temperature (TAS) of regional climate simulations, focusing on a region with complex orography, namely, the Friulian Alps. The temperature measurements consist of weather station records, while model TAS outputs belong to an ensemble of EURO-CORDEX models.

Starting from the vertical structure of the atmospheric thermal profiles, we provide a description of the origin of TAS biases in climate simulations and suggest the presence of three main bias components. The first is linked to the driving GCM ability to reproduce the free atmosphere temperatures at a given level (e.g. 500 hPa); the second bias component depends on the models performance in reproducing the thermal structure between the free atmosphere and the boundary layer top: we show that, under the environmental lapse rate approximation, this in turn depends on the orographic error and on the error in the boundary layer thickness. The last bias component accounts for the poor performance of the boundary layers parameterization schemes in reproducing land-atmosphere processes.

Applying this bias analysis to the temperature measurements and EURO-CORDEX simulations that are available for the study area, we evaluated the goodness of the common practice to remove the effects of orographic discrepancies, by means of the environmental lapse rate approximation, frequently adopted in climate projections and model evaluation studies. We show that, after the orographic correction of GCM-RCM TAS outputs, substantial biases still persist, especially during cold months. According to the existing literature, we argue that CMIP5 might contribute to the overall TAS bias, with a cold bias of about 2 degrees, explaining the free atmosphere level component of the bias, and we discuss the influence of the errors in reproducing the boundary layer thickness.