On the use of multiple planetary boundary layer parameterization schemes forecasting temperature and precipitation in complex terrain Chui, Timothy Chun-Yiu
Turbulent exchanges of heat and moisture between the Earth and atmosphere are approximated in modern numerical weather prediction (NWP) models by planetary boundary layer (PBL) parameterizations. These parameterizations were originally formulated in flat terrain, even though they are used in weather models that can be applied to anywhere in the world. A comparison of the deterministic forecast performance of different parameterizations has not been conducted in the complex, mountainous terrain of western Canada. The focus of this research was to evaluate eight commonly used PBL parameterizations by verifying their point forecasts of hourly temperature and daily accumulated precipitation against surface observations at 31 stations in British Columbia and Alberta for one year. The verification showed that choosing the optimum parameterizations depends mostly on the forecast variable and on the station, not on the complexity of their formulation. For the raw forecasts, the Mellor-Yamada-Janjić scheme (MYJ) had the best verification statistics on average for temperature, while the Medium-Range Forecast scheme (MRF) had the best verification on average for precipitation. The forecasts were also combined using a variety of linear methods. An ensemble of PBL schemes inversely weighted by their root mean square error (RMSE) produced hourly temperature forecasts that were just as good as the best individual bias-corrected member at each station, while an equally weighted ensemble of linearly regressed precipitation forecasts had only a slightly larger error on average than the best individual precipitation forecast. Daily forecasts of temperature and precipitation, which are variables of interest to both the general public and to industry, can thus be improved over the raw model output by the careful selection and combination of PBL parameterizations.
Item Citations and Data
Attribution-NonCommercial-NoDerivatives 4.0 International