Metrics for the evaluation of warm convective cloud fields in a large-eddy simulation with Meteosat images

Abstract

The representation of warm convective clouds in atmospheric models and satellite observations can considerably deviate from each other partly due to different spatial resolutions. This study aims to establish appropriate metrics to evaluate high-resolution simulations of convective clouds by the ICON Large-Eddy Model (ICON-LEM) with observations from Meteosat SEVIRI over Germany. The time series and frequency distributions of convective cloud fraction and liquid water path (LWP) are analyzed. Furthermore, the study focuses on size distributions and decorrelation scales of warm convective cloud fields. The investigated metrics possess a pronounced sensitivity to the apparent spatial resolution. At the fine spatial scale, the simulations show higher occurrence frequencies of large LWP values and a factor of two to four smaller convective cloud fractions. Coarse-graining of simulated fields to the optical resolution of Meteosat essentially removes the differences between the observed and simulated metrics. The distribution of simulated cloud sizes compares well with the observations and can be represented by a power law, with a moderate resolution sensitivity. A lower limit of cloud sizes is identified, which is 8–10 times the native grid resolution of the model. This likely marks the effective model resolution beyond which the scaling behaviour of considered metrics is not reliable, implying that a further increase in spatial resolution would be desirable to better resolve cloud processes below 1 km. It is finally shown that ICON-LEM is consistent with spatio-temporal decorrelation scales observed with Meteosat having values of 30 min and 7 km, if transferred to the true optical satellite resolution. However, the simulated Lagrangian decorrelation times drop to 10 min at 1 km resolution, a scale covered by the upcoming generation of geostationary satellites.