Assimilation of 3D Polarimetric Microphysical Retrievals with an Ensemble Kalman Filter in Germany

Abstract

Accurate quantitative precipitation forecasts (QPF) by numerical weather prediction (NWP) models are and remain of high societal interest, especially in times of global warming, which causes an increasing frequency and intensity of heavy precipitation events across the earth. The assimilation of weather radar observations in convective-scale NWP systems has been proven highly beneficial for improving short-term QPF, but research has so far mostly focused on the assimilation of non-polarimetric radar quantities such as horizontal radar reflectivity ZH and radar radial wind observations. Polarimetric radar data carry more independent information about cloud-precipitation microphysics than ZH observations do and thus have the potential to further improve short-term QPF. However, assimilation of polarimetric data is still in its infancy. <br /> In this study, polarimetric information from the C-band radar network operated by the German meteorological service (DWD, Deutscher Wetterdienst) are assimilated in DWD’s operational convective-scale ensemble-based NWP system for the first time. The polarimetric observations are assimilated using 3D microphysical retrievals of liquid and ice water content (LWC and IWC) below and above the melting layer, respectively, and the impact on short-term QPF compared to the assimilation of ZH observations alone is investigated. For this purpose, this thesis develops an LWC estimator based on a large German disdrometer data set and C-band T-matrix scattering calculations. It is designed to exploit and mitigate the respective advantages and shortcomings of the different polarimetric radar moments known for different precipitation characteristics in a hybrid way. When applied to German C band radar observations of four stratiform and five convective warm-season events, the adapted hybrid LWC estimator yields an encouraging close-to-zero bias and better correlations than all tested non-hybrid new and existing estimators from the scientific literature. <br /> With optimized data assimilation settings, the assimilation of the new hybrid LWC estimator below the melting layer mostly improves deterministic and ensemble first-guess QPF over the assimilation of ZH observations alone for two intense stratiform cases in the summers of 2017 and 2021 and an intense convective case in the summer of 2021. Assimilation of polarimetric data above the melting layer using a hybrid state-of-the-art IWC retrieval from the scientific literature mostly degrades the first guess, especially for convective precipitation, likely because of a lower quality of the estimator in these situations. However, first-guess QPF quality is improved notably for the 2021 stratiform case, for which the estimation of specific differential phase profits from a higher radial radar resolution compared to the other cases. Overall, the best first guess is achieved when ZH, LWC, and IWC are assimilated together. <br /> The assimilation of 3D LWC or IWC estimates on average slightly improves nine hour QPF for most forecast hours compared to the assimilation of ZH observations alone, in particular when LWC estimates are assimilated for the 2017 convective case and when IWC estimates are assimilated for the 2021 stratiform case. Nonetheless, the IWC assimilation degrades deterministic nine-hour QPF again for convective precipitation, which suggests the need for the development of convection-adjusted IWC retrievals for future assimilation studies with polarimetric ice microphysical retrievals. Overall, the best QPF over the first six forecast hours is yielded when ZH, LWC, and IWC are assimilated together.