A Case Study of Orographic Airflow and Precipitation in a Wildfire Hazard Area of British Columbia, Canada: Toward a Higher-Resolution NWP Model over Complex Terrain

In the midst of a record-breaking wildfire season in British Columbia (BC), Canada, Environment and Climate Change Canada ran an experimental convective-scale numerical weather prediction (NWP) model to support wildfire suppression efforts. This high-resolution model with a 1-km horizontal grid spacing over southern BC (SBC-1.0km) is piloted by the operational High-Resolution Deterministic Prediction System with a 2.5-km grid spacing (HRDPS-2.5km), using land surface initial conditions provided by the Canadian Land Data Assimilation System. It was run twice daily (initialized at 0000 and 1200 UTC) for 36 hours. This study summarizes a one-day evaluation of these two NWP models for 24 August 2017, when a weak frontal system moved across western Canada. The front produced some strong valley winds over the complex terrain of southern BC. Some of these strong winds were well predicted by the models at useful lead times. In particular, a phenomenon locally known as the “Trepanier Split”, which refers to the occasional strong winds funneling out of the Trepanier Creek and splitting into a northerly and a southerly flow in the Okanagan Valley as they encounter Okanagan Mountain, was correctly predicted by the SBC-1.0km with a model lead time of up to 18 hours. The HRDPS-2.5km model was unable to predict the detailed structure of this orographic airflow due to resolution limitations. Both models failed to predict some afternoon thunderstorms, which could have been initiated by wildfire-induced surface hot spots. Observations suggest that the diabatic cooling effect associated with the thunderstorms produced a meso-scale cold high pressure center, which forced a severe evening windstorm through the Okanagan Valley. Both models failed to predict this windstorm with a reasonable lead time: the strong northerly valley flow was only predicted by the runs initialized at 0000 UTC 25 August, with a model lead time of two hours that was practically useless to the operational forecasters. Hypothetically, with a more powerful supercomputer allowing faster NWP production, these high-resolution models could have provided a useful ultrashort-term forecast for this windstorm. It is expected that the forecast from the SBC-1.0km would be better than that from the HRDPS-2.5km, as verified by observations. The potential advantages of a wildfire-coupled NWP model are also discussed in this study.