Evaluation of the Urban Climate Model PALM-4U over Hilly Terrain Using Wind and Turbulence Observations

We used boundary layer observations in the Stuttgart area from two summer episodes for model evaluation of the urban climate model PALM-4U. In summer 2017, radiosondes and Doppler lidars were operated, and the lidar virtual tower technique was applied to provide vertical profiles of wind speed and direction at different sites. In summer 2018, two Doppler lidars were operated in vertical stare mode providing vertical wind speeds and their variances. PALM-4U with grid spacings of 10 m in 2018 and 40 m in 2017 was driven by COSMO analysis data. The PALM-4U output data were evaluated with the observations at the corresponding sites. For 14 to 15 August 2017, the normalised root mean square error (NRMSE) between simulated and measured wind speed time series at 100 m agl is about 0.5, except for the last hours of the investigated period. The RMSE for wind direction is 30–35°. At 700 m agl, the NRMSE for wind speed is 0.2–0.3 and the RMSE for wind direction is about 15°. The greater NRMSE differences found in the night and morning of 15 August can be explained by differences between the observed and simulated height of the low-level jet which caused stronger simulated turbulent downward mixing of momentum in the morning. On 20 June 2018, the daytime convective boundary-layer evolution was well represented in the model. However, as the wind speed was low, 1-h integration times turned out to be not suitable for model evaluation because the uncertainty in variances of the vertical wind is considerable (50%). To overcome the poor statistics due to the low number of eddies dominating, 3-h integration times turned out to be necessary. For these time intervals, the simulated profiles lie within the error bars of the observations. Theoretical considerations provide suggestions for experimental set-ups and synoptic conditions to capture vertical profiles allowing model evaluations under convective conditions based on 1-h intervals.