Evaluating Planetary Boundary Layer and Land Surface Models via Dual-Polarization WSR-88D and Flux Tower Observations

Recent work has taken advantage of the existing national network of dual-polarization WSR-88D radars to produce accurate daytime planetary boundary layer (PBL) depth estimates for every radar volume scan (roughly every 10 min or less) at WSR-88D sites across the United States. We expand on this work by comparing hourly forecasts of daytime PBL depth from the Mellor–Yamada–Nakanishi–Niino eddy diffusivity–mass flux (MYNN-EDMF) PBL scheme to estimates obtained from WSR-88D observations for calendar year 2022. Forecasts from the operational Rapid Refresh (RAP) model that uses the Advanced Research version of the Weather Research and Forecasting (WRF-ARW) Model are used to generate this large dataset for analysis. We find that MYNN-EDMF forecasts of PBL depth can differ significantly from WSR-88D observations in both timing of growth and maximum depth of the PBL with consistent biases in PBL depth across regions and seasons. These biases are then evaluated through the comparison of parameterized surface sensible and latent heat fluxes to those observed by flux towers located near selected radar sites. We find that errors in predicted PBL depths are strongly correlated with the over-/underestimation of surface sensible heat flux. Further assessment of modeled soil moisture reveals that the land surface model utilized by RAP is likely contributing to PBL depth biases in summer months through erroneous drying of soil, which increases surface sensible heat flux and therefore PBL depth. SIGNIFICANCE STATEMENT: The boundary layer plays a significant role in many weather phenomena that impact society. Recent work has shown that characteristics of this layer, such as its depth, can be measured by operational weather radars. This study compares these observations to predictions from a weather model to understand that model’s biases. We find that the model consistently overestimates boundary layer depth in summer months and further explore potential causes of this bias. This work will help model developers address model errors and potentially improve weather model performance.