Abstract This study introduces a new generalisable methodology for evaluating the representation of boundary‐layer turbulence in sub‐kilometre numerical weather simulations through direct comparison with high‐resolution Doppler lidar and sonic anemometer measurements. We derive key boundary‐layer parameters, such as aerosol height, cloud‐base height, turbulent mixing height, sensible heat flux, and boundary‐layer classification, using an established observational algorithm and similar diagnostics from newly available high temporal resolution outputs from UKV (1.5 km) and WMV (300 m) simulations of the Unified Model. For the case study considered here, there is agreement in the temporal evolution and magnitude of bulk boundary‐layer characteristics for both simulations, including the timing of the morning transition from stable to unstable surface conditions. However, significant differences are found in both the shape and magnitude of vertical velocity variance profiles. In the UKV simulations, peak vertical velocity variance is underestimated by up to a factor of 5 compared with observations, while differences in the WMV simulations are smaller (up to a factor of 3). In the UKV simulations, turbulence is predominately unresolved (parametrized). By contrast, in the WMV over 90% of the vertical velocity variance is resolved explicitly in the middle of the boundary layer once the convective regime is established. These results demonstrate that, for this case study, parametrized turbulence is underestimated in both numerical weather prediction simulations. More broadly, this work highlights the value of high‐resolution observations in diagnosing model performance and provides a transferable evaluation framework applicable at any site with high‐frequency simulation data and Doppler lidar observations.
Harvey et al. (2026) studied this question.