Evaluation of mixed-layer height from a hectometric-scale NWP model using a dense LiDAR network in Berlin, Germany

Hectometric scale numerical weather prediction (NWP) models are expected to provide improved spatio-temporal weather and climate information in urban regions. Key to understanding the interaction of urban surfaces with the overlying atmosphere is knowledge of the depth of the mixed-layer. Here, we use simulations with the UK Met Office Unified Model with grid-spacing down to 100 m over Berlin, Germany during two days with significantly different weather conditions (18 April 2022 and 4 August 2022) and obtain fields of mixed-layer height from model output (Z_mod) with a new algorithm (termed MMLH). We evaluate MMLH using aerosol attenuated backscatter-derived mixed-layer height from a network of 25 automatic lidar-ceilometers (Z_obs) operated in the city and its surroundings as part of the urbisphere-Berlin campaign, that was designed for observing intra-urban and urban-rural differences in boundary-layer characteristics.  Z_mod compares best to Z_obs in the afternoon and is consistently able to reproduce the vertical extent of the mixed-layer during late afternoon on both case days. MMLH performance is better at 100 m grid-resolution relative to a 300 m configuration. This could be related to higher vertical resolution on the 100 m grid, allowing for tighter vertical gradients in aerosol to be resolved. Both days show a distinct influence of the city on the mixed-layer height , with an urban plume downwind of the city evident in both the observations and model. By nighttime, the model’s urban plume leads to a deeper mixed-layer 10 km downwind of the city compared to upwind (~200 m 18 April; ~500 m 4 August). Both Z_obs and Z_mod show a similar urban plume at night on both days, but on 18 April Z_obs was generally higher by ~100 m over the city. From this study, we identify future avenues to improve hectometric-scale models in urban environments.