Evaluation of the Weather Research and Forecasting Mesoscale Model for GABLS3: Impact of Boundary-Layer Schemes, Boundary Conditions and Spin-Up

We evaluated the performance of the three-dimensional Weather Research and Forecasting (WRF) mesoscalemodel, specifically the performance of the planetary boundary-layer (PBL) parametrizations. For this purpose, Cabauw tower observations were used, with the study extending beyond the third GEWEX Atmospheric Boundary-Layer Study (GABLS3) one-dimensional model intercomparison. The WRF model (version 3.4.1) contains 12 different PBL parametrizations, most of which have been only partially evaluated. The GABLS3 case offers a clear opportunity to evaluate model performance, focusing on time series of near-surface weather variables, radiation and surface flux budgets, vertical structure and the nighttime inertial oscillation. The model results revealed substantial differences between the PBL schemes. Generally, non-local schemes tend to produce higher temperatures and higher wind speeds than local schemes, in particular, for nighttime. The WRFmodel underestimates the 2-m temperature during daytime (about 2K) and substantially underestimates it at night (about 4 K), in contrast to the previous studies where modelled 2-m temperature was overestimated. Considering the 10-m wind speed, during the night turbulent kinetic energy based schemes tend to produce lower wind speeds than other schemes. In all simulations the sensible and latent heat fluxes were well reproduced. For the net radiation and the soil heat flux we found good agreement with daytime observations but underestimations at night. Concerning the vertical profiles, the selected non-local PBL schemes underestimate the PBL depth and the low-level jet altitude at night by about 50 m, although with the correct wind speed. The latter contradicts most previous studies and can be attributed to the revised stability function in the Yonsei University PBL scheme. The local, turbulent kinetic energy based PBL schemes estimated the low-level jet altitude and strength more accurately. Compared to the observations, all model simulations show a similar structure for the potential temperature, with a consistent cold bias (approximate to 2 K) in the upper PBL. In addition to the sensitivity to the PBL schemes, we studied the sensitivity to technical features such as horizontal resolution and domain size. We found a substantial difference in the model performance for a range of 12, 18 and 24 h spin-up times, longer spin-up time decreased the modelled wind speed bias, but it strengthened the negative temperature bias. The sensitivity of the model to the vertical resolution of the input and boundary conditions on the model performance is confirmed, and its influence appeared most significant for the non-local PBL parametrizations.