A Sensitivity Study of the Hub-Height Wind Speed Forecast to the PBL Scheme and Initial Condition Selection in the WRF Model in Complex Terrain

Daily hub-height wind speed forecasts using the Weather Research and Forecasting (WRF) model have been produced for a one-year long period from 01 June 2014 through 31 May 2015 in the complex terrain of British Columbia. Sixteen runs were produced daily to evaluate the performance of eight planetary boundary layer (PBL) schemes available in the WRF model by using two initial condition sources: the Global Forecast System and North American Mesoscale model. Since none of the PBL schemes were explicitly designed for use in complex terrain, their applicability may be limited. Further, evaluation of hub-height winds is critical because small errors in wind-speed forecasts are compounded when converted to wind power forecasts.

The results presented here evaluate the wind speed forecasts at four wind farms in complex terrain and quantify the relative importance of selecting the correct PBL scheme versus initial condition. It was found that the model grid spacing and PBL scheme selection are the most important factors to consider when forecasting the hub-height wind in complex terrain, while the initial condition source contributed only a small amount to differences in forecast verification statistics. The dominant factor was also location dependent. At some stations, the PBL scheme choice was found to contribute up to 90% of the variance in forecast verification statistics. The contribution to that variance was also found to vary by time of day and season, with the PBL scheme choice having the largest influence in the summer during the daytime hours. Forecast mean absolute error (MAE) can be reduced by 9-15% be choosing a well performing scheme over a poor performing scheme, an important cost saving for power planners. Changing the initial condition resulted in only a 2% improvement in MAE.