Examination of flow-dependent errors in near-surface temperature and winds from WRF numerical simulations over complex terrain

Accurate forecasting of near surface atmospheric conditions, such as temperature and winds, is of interests in many applications. However, the problem itself, especially the forecasting over complex terrain presents a challenge in numerical weather prediction. This study evaluates the performance of modern weather research and forecasting (WRF) model in predicting near surface atmospheric temperature and wind conditions under various terrain and weather regimes. Three individual cases under strong synoptic forcing (i.e., a frontal system, a low-level jet and a persistent cold air pool) are first verified against observations over both flat and complex terrain. It is found that WRF model is able to reproduce reasonable simulations of weather phenomena. Verification for near surface conditions (i.e., temperature at 2-m height and winds at 10-m height) indicated that forecasts of surface variables over flat terrain generally agree well with the observations while the errors could also depend on the model's predictability to the atmospheric boundary layer. Over complex terrain, forecasting of near surface atmospheric conditions not only suffers from the representative issues due to mismatches between model and realistic terrain, but also struggles with the model's ability in reproducing the accurate atmospheric conditions in lower troposphere. In addition, over complex terrain, simulations at finer resolution do not over-perform those at coarser resolutions. The statistic analysis has also performed for 120 forecasts during one-month period to further investigate the forecasting error characteristics over complex terrain. Results illustrated that the forecast errors in surface variables remarkably depend on the diurnal cycle of surface conditions, especially when the synoptic-forcing is weak. Under the strong synoptic forcing, the diurnal patterns in the errors are broken while the flow-dependent errors are clearly shown.