Evaluation of ice and snow content in the global numerical weather prediction model GME with CloudSat

Abstract. The present study evaluates the global numerical weather prediction model GME with respect to the grid-scale parameterization of frozen particles, both ice and snow, focusing on the performance of a diagnostic versus a prognostic precipitation scheme. As a reference, CloudSat Cloud Profiling Radar observations are utilized – the so far only near-globally available data set which vertically resolves clouds. Both the observation-to-model and the model-to-observation approach are applied and compared to each other. For the latter, the radar simulator QuickBeam is utilized. Criteria are applied to further improve the comparability between model and observations. The two model versions are statistically evaluated for a four-month period.

The comparison reveals that the prognostic scheme reproduces the shape of the CloudSat frequency distributions for both ice water content (IWC) and reflectivity factor well, while the diagnostic scheme produces no large IWCs or reflectivity factors because snow falls out instantaneously. However, the prognostic scheme overestimates the occurrence of high ice water paths (IWP), especially in the mid-latitudes. Sensitivity tests show that an increased fall speed of snow successfully reduces IWP. Both evaluation approaches capture the general features, but for details, the two together deliver the largest informational content. In case of limited resources, the model-to-observation approach is recommended. Finally, the results indicate that the lack of IWC in most global circulation models might be attributed to the use of diagnostic precipitation schemes, i.e., the lack of snow aloft.

Based on its good performance the prognostic scheme went into operational mode in February 2010. The adjusted snow fall speed went operational in December 2010. However, continual improvements of the ice microphysics are necessary, which can be assessed by the proposed evaluation technique.