Representation of mesoscale flow characteristics in a convective-scale analysis dataset: validation using Doppler wind lidar measurements from the 'Swabian MOSES 2023' Campaign

Forecasting high-impact convective events, which pose significant risks to people and property, remains a persistent challenge. This difficulty can partly be attributed to biases and errors in analysis data, which serve as initial conditions for numerical weather predictions (NWP), are often used to improve process understanding of such events from the model perspective. A systematic validation of the quality of analysis and forecast data, in particular in complex terrain, is often restricted by the limited availability of detailed observations.

Here we leverage observations from the ‘Swabian MOSES 2023’ field campaign, which took place in the Black Forest region, Southwestern Germany, during the summer 2023. Among others, the field campaign included an extended meso-scale network of 10 Doppler wind lidars, which are used to retrieve vertical profiles of the wind speed and direction.  These observations enable the characterization of the dynamic structure of the lower troposphere on the mesoscale in moderately complex terrain.

The retrieved wind profile observations, which extend from the near-surface to approximately 4 km above ground, are compared to the regional convective-scale analysis dataset. For this purpose, the ICON-D2 analysis data from Deutscher Wetterdienst, available at 2.2 km horizontal resolution and produced using the Icosahedral Nonhydrostatic (ICON) model , are used. The three months of continuous observations provide a comprehensive independent data set for validating the representation of mesoscale wind characteristics in the analysis across the Black Forest region. Overall, the convective-scale analysis captures the vertical profiles of horizontal wind well, with wind speed bias remaining on average below 0.5 m s^-1. Yet, we find differences between different measurement locations, depending to some extent on the local topography, with the largest differences in areas with more complex terrain. Moreover, the analysis tends to overestimate the zonal wind component at lower altitudes, while underestimating the meridional wind component at higher altitudes. Furthermore, the data show that the mean absolute error between the analysis and the observations is larger during rainy than during dry weather conditions, which highlights the added value of the observations, particularly during convective situations. Finally, we demonstrate the value of a high-resolution analysis dataset by comparing the observations against a range of analysis datasets, including ICON-D2, ICON-EU, and ICON-global, with spatial resolutions of 2.2 km, 6.5 km, and up to 13 km, respectively.