Examining Patterns of Temperature and Precipitation Biases in a Dynamic Downscaling Process across Europe

Abstract

The Weather Research and Forecasting (WRFv4.0 ARW) mesoscale meteorological model dynamically downscales data from the NASA Goddard Institute for Space Studies (GISS) GCM ModelE2 atmospheric general circulation model (GCM) CMIP5 version (Model E2-R) over Europe to replicate the observed bias and temporal variability and the previously observed impact of climate change on critical meteorological parameters. Given that RCM outputs are still susceptible to climate model errors from the RCM's structure and the driving GCM/ESM's initial and boundary conditions, examining a model's ability to replicate changes in climatic factors over the last several decades is an intriguing task. This activity identifies systematic biases and examines their sources. WRFv4.0 ARW is single-nested with 0.75°–0.25° grid resolutions. The past and current periods are represented by two 30-year datasets, 1951–1980 and 1981–2010. Model biases for mean temperature, mean precipitation, the number of days with temperatures exceeding 25 °C, and days with precipitation surpassing 5 mm (following WMO guidelines) were compared against E-OBS observations at a 0.25° spatial resolution. The analysis revealed consistent underprediction of mean temperature changes across all subregions, indicating a colder climatology assessment, especially during winter in southern and eastern subregions. Most subregions have the strongest bias in winter. The southern and eastern subregions show the largest bias due to land-atmosphere interactions. Conversely, spring or fall simulate the lowest biases. The lower snow cover during these seasons counters the overestimation of surface albedo in winter, and the radiation scheme has a gentler effect than in summer. In terms of days with a mean temperature above 25 °C, Southern Europe experiences an increased number of them. Model biases in mean precipitation displayed a general negative trend across subregions, a known issue in WRF simulations influenced by regional weather patterns that do not show a geographically regular trend. Over the two time slices, northwestern Europe had more days per month with mean daily precipitation above 5 mm than the middle and southern regions. Model simulated bias is quite small everywhere, demonstrating competence in predicting changes beyond the threshold. Model seasonal findings were always comparable to observations, despite minor regional differences. This study underscores the importance and need for more observations, particularly from southern Europe, for more credible evaluation studies.