Sensitivity analysis of PBL schemes by comparing WRF model and experimental data

Abstract. This work discusses the sources of model biases in reconstructing the Planetary Boundary Layer (PBL) height among five commonly used PBL parameterizations. The Weather Research and Forecasting (WRF) Model was applied over the critical area of Northern Italy with 5 km of horizontal resolution, and compared against a wide set of experimental data for February 2008. Three non-local closure PBL schemes (Asymmetrical Convective Model version 2, ACM2; Medium Range Forecast, MRF; Yonsei University, YSU) and two local closure parameterizations (Mellor Yamada Janjic, MYJ; University of Washington Moist Turbulence, UW) were selected for the analysis. Vertical profiles of aerosol number concentrations and Lidar backscatter profiles were collected in the metropolitan area of Milan in order to derive the PBL hourly evolution. Moreover, radio-soundings of Milano Linate airport as well as surface temperature, mixing ratio and wind speed of several meteorological stations were considered too.

Results show that all five parameterizations produce similar performances in terms of temperature, mixing ratio and wind speed in the city of Milan, implying some systematic errors in all simulations. However, UW and ACM2 use the same local closure during nighttime conditions, allowing smaller mean biases (MB) of temperature (ACM2 MB = 0.606 K, UW MB = 0.209 K), and wind speed (ACM2 MB = 0.699 m s⁻¹, UW MB = 0.918 m s⁻¹).

All schemes have the same variations of the diurnal PBL height, since over predictions of temperature and wind speed are found to cause a general overestimation of mixing during its development in winter. In particular, temperature estimates seem to impact the early evolution of the PBL height, while entrainment fluxes parameterizations have major influence on the afternoon development. MRF, MYJ and ACM2 use the same approach in reconstructing the entrainment process, producing the largest overestimations of PBL height (MB ranges from 85.51–179.10 m). On the contrary, the best agreement between model and both Lidar and balloon observations was identified for YSU (MB = −27.54 m and 30.15 m, respectively).