Mesoscale Model Intercomparison and Observational Evaluation for Three Contrasting Diurnal Cycles in CASES99: Focus on the Stable Boundary Layer

The boundary-layer scheme in Numerical Weather Prediction (NWP) models is an essential connection between the atmosphere and the land-surface. Despite its relevance, the representation of the boundary-layer and the diurnal cycle in NWP models is rather poor, especially for stable conditions. Following the GABLS single-column and LES-model intercomparisons, we evaluate three state-of-the-art meso-scale models: MM5 (using 4 boundary-layer schemes, and three radiation schemes), COAMPS and HIRLAM on their performance in the atmospheric boundary layer, for three contrasting diurnal cycles during CASES-99. All models use maximum possible horizontal resolution, and increased vertical resolution near the surface compared to operational settings. Soil moisture availability was modified to realistic conditions according to the observations. The first diurnal cycle contains a night with intermittent turbulence, while the second night is continuous turbulent. Finally, the last cycle contains a night with extremely small amount of turbulence and highly driven by the radiative forcing. We find that during daytime the TKE-closure schemes show too cold and shallow boundary layers. During nighttime TKE models show a more realistic vertical mixing compared to operational first-order schemes. All models show a high sensitivity of the boundary layer development to the soil moisture availability. Contrary to earlier findings, the representation of the Low-Level jet is surprisingly good. A general model deficiency is the underestimated longwave incoming radiation. Furthermore,it is found that the minimum temperature modeled with MM5 depend strongly on the chosen radiation scheme (up to 5 K during weak winds), with favorable results when using RRTM or CCM3. This sensitivity is an extra contribution to the already known strong sensitivity for turbulent mixing in stable conditions.