Evaluation and intercomparison of meteorological predictions by five MM5-PBL parameterizations in combination with three land-surface models

In this study, MM5 predictions with five PBL parameterizations in combination with three land-surface models (LSMs) are intercompared and evaluated by using a wide variety of observations derived from WMO routine surface weather stations, TRACE-P aircraft experiments, intense radiosonde soundings and satellite measurements. Six scenarios with various PBL schemes and LSMs are designed to investigate the similarities and differences in model predictions. For near-surface variables, all scenarios yield good correlation between prediction and observation for 2 m-temperature (T2) and 2 m-water vapor mixing ratio (Q2), and relatively poor ones for wind fields. On average, T2 was consistently underpredicted by all scenarios, whereas Q2 was overpredicted by five of the six scenarios. It is found that the application of Noah land-surface model instead of the five-layer soil model is able to enhance the prediction accuracy of Q2. For 10 m-wind speed, the GSE scenario (Gayno–Seaman scheme with the five-layer soil model) produces somewhat smaller correlation, but better consistency in magnitude than those of the other scenarios. Model predictions are more consistent for upper air as a result of using FDDA reanalysis nudging and the reducing influence of underlying surface with altitude. All scenarios show the tendencies to underpredict temperature and to overpredict wind speed at altitudes <1 km and to underpredict wind speed at altitudes >3 km. The correlations for water vapor mixing ratio are much smaller at altitudes >3 km than that in the boundary layer. The differences in predicted PBL height among scenarios are large. GSE scenario performs best for phase (correlation), whereas PCX scenario (Pleim–Chang scheme with Pleim–Xiu LSM) produces the best statistics for magnitude of PBL height. Diurnal variation of PBL height over the western Pacific region during the study period is characterized by the typical day and night cycling superimposed by occasional expansion associated with cold front passage. The scenarios similarly predict precipitation spatial variability, but the difference in absolute magnitude is comparatively large among them, even for the scenarios using the same LSM, suggesting the important impact of PBL turbulence parameterization on precipitation. All scenarios predict larger precipitation amount averaged over the entire domain with the exception of GSE scenario. The statistics for satellite estimates are better than those for model predictions, but satellite data tends to underpredict observation by a similar range to that of GSE scenario. In the study period, latent heat fluxes are about twice the sensible heat fluxes over the western Pacific region, with maximum being 300–600 and 100–300 W m−2, respectively. The differences in surface fluxes among scenarios are mainly determined by various LSMs rather than different PBL schemes applied.