Evaluation and improvement study on the planetary boundary-layer scheme for fog-forecasting in North China

The numerical forecast of fog is still challenging at the current stage, as many models typically show poor fog-forecasting capabilities. Model predictions of fog exhibit strong sensitivity to the selection of planetary boundary layer (PBL) parameterizations, as well as horizontal grid spacing (HGS). To address this issue, this study intercompares the fog-forecasting performance of WRF model using five PBL schemes (YSU, MYJ, MYNN, ACM2 and SH) and different HGS ranging from ten-kilometric- to hectometric-scale (15-km to 500-m). Validation against available in-situ measurements indicates that the YSU scheme produces an overall superior performance in fog forecasting, followed by MYNN and MYJ. For a certain PBL scheme, the model shows distinct forecasting capabilities in terms of different types of fog. That is, the model generally shows better performance in the forecasting of advection fog episodes, compared to radiation fog episodes. Regarding different HGS, intercomparisons of mesoscale modeling (i.e., WRF) with HGS ranging from ten-kilometric- to hectometric-scale (15-, 5-, 2.5-km and 500-m) reveal that, although simulations using finer HGS could generally better represent the spatial distribution of meteorological elements and the influence of small-scale underlying surface, the fog forecasting skills (i.e., the TS scores) does not consistently improve with the refinement of HGS. Fog forecasting at different HGS behaves differently for two types of fog days which are primarily differentiated by whether or not the dissipation of fog is substantially influenced by the background synoptic wind flow. For type-Ⅰ fog days (with no obvious impact of the background synoptic wind), differences in fog forecasting skills (TS values) among different HGS simulations are relatively smaller. Whereas for type-Ⅱ fog days (with the dissipation of fog strongly affected by the invasion of cold airflow), deviations in TS values among simulations using different HGS become more evident, with 2.5-km HGS providing better performance and the coarsest-HGS (15-km) simulation shows a noticeable degradation in the fog forecasting skills. Also note that simulation using the finest HGS (500-m) shows no superiority (or somewhat degradation) in fog forecasting skills for both of the two types of fog days. Additionally, the influence of model HGS on simulating the spatial distribution of fog is more pronounced during the formation and dissipation stage, whereas rather limited during the developing stage. On the basis of comprehensive model evaluation, we further attempt to improve the model performance under stable stratification by incorporating a wind-shear term into the mixing-length formulation for a turbulence scheme. After validation against available observations, results of the sensitivity experiments show that this modification could generally improve the representation of turbulent mixing, near-surface meteorological elements, as well as vertical boundary layer structures during stable conditions.