Evaluation of convective cloud microphysics in numerical weather prediction models with dual-wavelength polarimetric radar observations: methods and examples

Abstract

Abstract. The representation of cloud microphysical processes contributes substantially to the uncertainty of numerical weather simulations. In part, this is owed to some fundamental knowledge gaps in the underlying processes due to the difficulty of observing them directly. On the path to closing these gaps, we present a setup for the systematic characterization of differences between numerical weather model and radar observations for convective weather situations. Radar observations are introduced which provide targeted dual-wavelength and polarimetric measurements of convective clouds with the potential to provide more detailed information about hydrometeor shapes and sizes. A convection-permitting regional weather model setup is established using five different microphysics schemes (double-moment, spectral bin (“Fast Spectral Bin Microphysics”, FSBM), and particle property prediction (P3)). Observations are compared to hindcasts which are created with a polarimetric radar forward simulator for all measurement days. A cell-tracking algorithm applied to radar and model data facilitates comparison on a cell object basis. Statistical comparisons of radar observations and numerical weather model runs are presented on a data set of 30 convection days. In general, simulations show too few weak and small-scale convective cells. Contoured frequency by altitude diagrams of radar signatures reveal deviations between the schemes and observations in ice and liquid phase. Apart from the P3 scheme, high reflectivities in the ice phase are simulated too frequently. Dual-wavelength signatures demonstrate issues of most schemes to correctly represent ice particle size distributions, producing too large or too dense graupel particles. Comparison of polarimetric radar signatures reveals issues of all schemes except the FSBM to correctly represent rain particle size distributions.

Influence of cloud microphysics schemes on weather model predictions of heavy precipitation

“…The exact opposite is true for the P3, Thompson 2-mom, and Thompson aerosol-aware simulations. These results are consistent with the findings of Köcher et al (2022), where the simulated differential reflectivity was statistically compared with the observed one. In particular, the SBM scheme did not produce larger differential reflectivity in the lower elevations, suggesting the absence of large drops in the SBM, while the P3 and Wu et al (2021).…”

“…We conclude that it is not the rain mass produced that is the problem, but rather the distribution across droplet sizes: compared to the radar-observed heavy rain events, the P3, Thompson 2-mom, and Thompson Aerosol-aware schemes produce large raindrops too frequently, while the SBM simulations produce too few. The results related to the Thompson 2-mom and Morrison schemes are in conflict with a previous study by Wu et al (2021), but are consistent with Köcher et al (2022) and Putnam et al (2016), highlighting the problem of evaluating microphysical schemes using case studies and demonstrating the importance of statistical evaluation as in this study.…”

“…A total of 30 convective weather days in 2019 and 2020 were used for comparison. A detailed description of the radar observation and model simulation dataset can be found in Köcher et al (2022). It is also summarized below.…”

“…Depending on the grid spacing and type of model used, the available computational power may also simply be insufficient to provide weather simulations in a limited amount of time. In a recent study by Köcher et al (2022), microphysics schemes are statistically evaluated over a dataset of 30 convection days in 2019 and 2020. This dataset consists of weather simulations with 5 different microphysics schemes of varying complexity, as well as simultaneous polarimetric C-band radar observations from DWD, and is therefore well suited to address the aforementioned following problems:…”

“…In Köcher et al (2022), cloud microphysics of varying complexity are assessed by a statistical comparison of the observed radar signals with the simulated radar signals from the model output, i.e., by applying the approach (1) described above. This study builds on the study by Köcher et al (2022) and goes one step further by additionally using the approach (2) described above to evaluate the microphysics, i.e., we obtain hydrometeor information from polarimetric radar observations and compare it to the simulated hydrometeors. The focus is on high-impact weather events, i.e., hail and heavy rain.…”