A04
Precipitation processes
PD. Dr. Silke Trömel
University of Bonn | +49-228-73779 |
Summary
Atmospheric models still do not adequately represent precipitation generating processes and previous studies identify cloud and precipitation processes as one of the largest uncertainties in current weather and climate prediction models. This incomplete representation of precipitation generating processes is assumed to be at least partly responsible for their inability to reproduce observed regional patterns of drying and wetting. We compared different precipitation datasets covering the European domain and then evaluated the simulated precipitation and precipitation trends against them. We then aimed to relate the trends in precipitation with trends in terrestrial water storage (from satellite observations and reanalysis) and found significant discrepancies, suggesting complex dynamics which could be linked to glacier melt, topographical influences and/or human water use (i.e., irrigation), and still require further investigation.
We focus on Germany and Türkiye, two countries with contrasting climate regimes and hydrological trends, which are monitored by modern polarimetric radar networks. Microphysical properties such as mean volume diameter (Dm), total number concentration (Nt), and ice/liquid water content (IWC/LWC) were analyzed using contoured frequency-temperature diagrams. Observations show that Dm increases with temperature due to riming and aggregation, Nt decreases toward the melting layer (ML), and IWC and LWC show site-dependent patterns that may be influenced by processes like evaporation, coalescence, and seasonal sample biases. Simulations differ from observations by showing fewer but larger particles above the ML and differing IWC gradients, indicating differences in aggregation and riming intensities but potentially also modeling deficiencies, e.g. in the ice nucleating scheme that requires further investigation. Contoured frequency-temperature diagrams of radar moments reveal consistent vertical profiles across German radar sites, while Turkish sites exhibit more variability. Germany shows stronger gradients of reflectivity above the ML, indicating more intense riming and aggregation, with notable seasonal changes across all sites. Stratiform events with dominating riming processes—more frequent in summer and more common in Germany—are associated with higher, colder cloud tops, sagging MLs, and distinct radar signatures, patterns observed consistently across both regions. These observational findings are now being further exploited for model evaluation and improvements. Whether the regional differences are influenced by anthropogenic interventions in the land-atmosphere system remains to be investigated also with supporting scenario simulations.
Graphical Summary
Contribution to CRC
The project contributes to the CRC’s key objectives by investigating deficiencies in precipitation generating processes and feedback mechanisms in the modelling platform, and the impact of anthropogenic modifications on precipitation generating processes and microphysical characteristics. Overall, we work on the model simulation scenarios and reanalyses provided by other projects and produce a detailed analysis of precipitation characteristics.
Approach
The project analyzes six years of radar data (2015-2020) from three radar sites in Germany and five in Türkiye. Using the EMVORADO radar forward operator, synthetic radar observations are generated to evaluate model simulations.
We compare model simulations, precipitation products, and polarimetric radar observations across Germany and Türkiye, focusing on trends, microphysical properties, and precipitation-generation signatures. By examining these in relation to terrestrial water storage (TWS) trends from reanalysis data, we aim to uncover if model inaccuracies in precipitation trends contribute to disparities with observed TWS trends.
Microphysical data—like mean volume diameter, water content, and hydrometeor types—are derived from radar data and compared to model simulations to detect potential biases. Further, we explore the impacts of anthropogenic influences by assessing differences in simulated rain rates, water content, and particle types across historic climate scenarios.
Through these comparisons, particularly between observed and simulated radar data, the project seeks to improve our understanding of model limitations in capturing human-driven changes in precipitation patterns.
Main Results in 2022, 2023 and 2024
Data collection and quality control: C-band radar observations from three radar sites in Germany and five radar sites in Türkiye for the period 2015-2020 were acquired during the second half of 2022 and concluded in early 2023. The data went through extensive data quality assurance and preprocessing. Several data quality issues in the Turkish data required special attention. State-of-the-art methods for melting layer (ML) detection, stratiform classification, riming detection and microphysical retrievals were adapted and applied to the data (e.g., Trömel et al., 2023; Blanke et al., 2025).
Trends in observed and modelled precipitation: Originally, the IMERG precipitation product (Huffman et al., 2023) was considered the best reference dataset freely available covering the European domain. However, considerable biases were found between IMERG and other precipitation products derived from rain gauges, satellite observations, ground radars, and reanalysis. This led to an unplanned effort to explore the suitability of IMERG as a reference dataset, with special emphasis on the focus regions Germany and Türkiye. This effort is not only important for the trend analysis but for proper validation of the simulations. Over Germany, a country with a large observational network, all rain-gauge based datasets (HYRAS, E-OBS, GPCC, CPC) perform similarly, with radar-based German products (RADOLAN RW, RADKLIM) coming closely behind, for yearly to daily precipitation accumulations. GSMAP is the best-performing satellite product, while IMERG V07B does not perform better than the ERA5 reanalysis. Over Türkiye, a country with a less dense observational network, the observational uncertainty is larger and IMERG V07B performs similarly to CPC and GSMaP. The full European domain is only covered by a small sample of datasets, also depending on the temporal resolution. Overall IMERG V07B performs slightly worse than E-OBS in yearly and monthly accumulations (reference: GPCC), but similarly to ERA5, EURADCLIM or CPC and better than other satellite products. All in all, there are better-performing datasets that can be used for evaluation over Germany, while IMERG remains among the best satellite-based estimates for other regions where other products are not available. See Hammoudeh et al. (2025) for details on the results for Germany at daily scale. Publication of results including the full European domain and other scales is in preparation.
As a next step, we aimed to relate simulated/observed trends in precipitation to trends in reanalyses of TWS and IMS-simulated TWS to elucidate whether insufficiently reproduced precipitation trends in the simulations explain the differing trends between the reanalysis of TWS and IMS model output. IMERG V07B and GPCC present similar precipitation trends over the whole domain and the D02 baseline simulation reproduces the observed trends to a certain degree (Fig. 1). The south of Germany has a larger simulated positive trend compared to the observations. Most of Türkiye has simulated trends similar to the observed ones, except for some localized differences. A research question from Phase 1 was whether feedback mechanisms exist – i.e. changes in precipitation processes due to drying – which the model does not represent. Thus, we would expect larger differences in the modeled vs observed precipitation in Türkiye compared to Germany due to more anthropogenic modifications in Türkiye, however this is not the case.
TWS from the D02 baseline simulation is unfortunately not yet available, so it remains to be explored whether the model reproduces the observed TWS trends and how it relates to precipitation trends. Fig. 1 also shows trends in TWS from GRACE (provided by D07) and the C01 reanalysis for the common period 2003-2020. The observational TWS from GRACE (Kvas et al. 2019) shows a drying trend over most of Europe. A drying trend covers most of Germany except the north. In Türkiye, drying trends are seen over most of the country except for the center-north which experiences a wetting trend. TWS reanalysis shows an overall good agreement with GRACE except some localized differences. In Türkiye the drying trend observed in GRACE is not clearly reproduced in the reanalysis, with an apparent influence from the complex topography. Referring to observations, the trends in precipitation do not directly relate to trends in TWS. For instance, a positive precipitation trend in the Alps and south of Germany contrasts with a negative trend in TWS, which could be related to melting glaciers. A positive precipitation trend is evident in most of Türkiye except negative trends mainly in southwest and central and north-eastern Türkiye. We continue to work in collaboration with D07 to understand the differences between TWS and precipitation trends.
Fig. 1: 2003-2020 TWS (GRACE, C01 Reanalysis) and precipitation (IMERG V07B, GPCC, D02 Baseline) trends.
Microphysical retrievals: Polarimetric microphysical retrievals of mean volume diameter (Dm), total number concentration (Nt) in logarithmic scale, and ice and liquid water content (IWC/LWC) have been calculated and explored in the form of contoured frequency-temperature diagrams. Dm increases with increasing temperature, consistent with particle growth due to riming and aggregation processes, reaching values between 1-4 mm with variability among radar sites. Dm values below the ML concentrate around 1 mm. Nt diminishes towards the ML, consistent with faster-falling particles and aggregation. IWC shows considerable variability among radar sites. Radar-derived IWC diminishes towards the ML at two of the Turkish sites, as expected assuming constant precipitation flux (Fig. 2), although there is also a slight increase in IWC in the dendritic growth layer (-20 to -10 °C). The other Turkish and the German radar sites, however, either do not show a strong gradient in IWC or show a slight increase towards the ML. LWC typically diminishes towards the surface/warmer temperatures, which could be due to evaporation or coalescence processes but also because the sample size of events in warmer temperatures is biased towards summer events. Simulations (taken from Trömel et al., 2023 due to delayed D02 simulations) show smaller Nt values above the ML and reach larger Dm above the ML compared to observations, i.e. few particles undergo more intense aggregation and riming processes. Simulated IWC profiles show gradients growing towards the ML in the simulations (Fig. 2), which will be explored together with differences in observed and modelled Nt in Phase 2.
Fig. 2: IWC frequency-temperature diagrams for radar stations in Gaziantep, Turkey (left panel); Prötzel, Germany (middle) and simulations from Trömel et al. (2023, right panel) for an X-band radar site in Germany.
Radar process signatures: Contoured frequency-temperature diagrams of the radar moments (horizontal reflectivity DBZH, differential reflectivity ZDR, specific differential phase KDP and cross-correlation coefficient RHOHV) show similar average vertical profiles among the German radar sites, and some differences are found when comparing to the Turkish sites, which show more variability. For instance, the gradient of DBZH above the ML, which indicates the intensity of riming and aggregation processes, is larger in Germany than in Türkiye. This gradient also shows strong seasonal variability among all sites. The strength of the ML, measured by the maximum value of DBZH and its difference to DBZH below the ML, also shows a seasonal variation although the median values are similar among different sites. A riming detection algorithm identifies riming in around 20% of all stratiform events in Germany and is more frequent during summer than in winter. In Türkiye, riming is more variable between regions and has a stronger seasonality. Despite the regional differences in riming frequency, rimed stratiform events show at all sites (German and Turkish) higher and colder cloud tops, a sagging ML, larger gradients and absolute values of DBZH above the ML, lower values of ZDR directly above the ML (Fig. 3), and greater values of DBZH and ZDR below the ML, compared to non-rimed events.
Fig. 3: Distributions of reflectivity gradient (left), reflectivity (middle) and differential reflectivity (right) directly above the melting layer in one site in Germany (Prötzel, first row) and the five sites in Türkiye (second to sixth rows), for all rimed stratiform events (yellow) and non-rimed stratiform events (purple).
To exploit these observational insights for model evaluation and improvement, high-resolution convection-permitting simulations are needed. Preliminary simulations with TSMP2-ICON at ~3 km resolution in the Euro-CORDEX domain with ERA5 initial and boundary conditions were analyzed for a collection of 10 precipitation events. The ML signature tends to be wider and less defined in the simulations compared to the observations. ZDR is considerably lower above and in the ML but larger below the ML. While observed ZDR and KDP tend to increase towards higher altitudes, polarimetric signatures above the ML are mostly missing in the simulations. These discrepancies can be at least partly explained by the deficiencies in the COSMO-ICON modeling framework described above in retrieval space, where too few ice particles produce large aggregates and additional excessive graupel production generates unrealistic ZDR profiles below the ML. Limitations of the T-matrix simulations in EMVORADO to produce appropriate polarimetric signals for fluffy low-density particles like snow are responsible for too small simulated ZDR and KDP values above the ML.
Regional differences and possible anthropogenic impacts: For this task we have so far focused on the Turkish radar sites because of the more contrasting hydrological trends. We find differences in the average profiles of DBZH, ZDR and KDP (not shown) between the radar in Sivas, at the center-north of Türkiye covering an area with a wetting-to-neutral trend, compared to the radars in the center south of Türkiye (Hatay and Gaziantep), covering an area with a drying trend. The relation of these differences to possible anthropogenic impacts on the drying and wetting trends and the deficiencies of the model to reproduce these trends remain to be investigated when the scenario simulations become available.
References
Blanke, A., Gergely, M., & Trömel, S. (2025). A new aggregation and riming discrimination algorithm based on polarimetric weather radars. Atmospheric Chemistry and Physics, 25(7), 4167-4184. https://doi.org/10.5194/acp-25-4167-2025
Kvas, A., Behzadpour, S., Ellmer, M., Klinger, B., Strasser, S., Zehentner, N., & Mayer‐Gürr, T. (2019). ITSG‐Grace2018: Overview and evaluation of a new GRACE‐only gravity field time series. Journal of Geophysical Research: Solid Earth, 124. https://doi.org/10.1029/2019JB017415
Hammoudeh S., Goergen K., Belleflamme A., Giles J. A. , Trömel S. and Kollet S., 2025: Evaluating precipitation products for water resources hydrologic modeling over Germany. Front. Earth Sci. 13:1548557. https://doi.org/10.3389/feart.2025.1548557
Huffman, G.J., E.F. Stocker, D.T. Bolvin, E.J. Nelkin, Jackson Tan, 2023: GPM IMERG Final Precipitation L3 Half Hourly 0.1 degree x 0.1 degree V07, Greenbelt, MD, Goddard Earth Sciences Data and Information Services Center (GES DISC), Accessed: 03 April 2024, 10.5067/GPM/IMERG/3B-HH/07
Trömel, S., U. Blahak, R. Evaristo, J. Mendrok, L. Neef, V. Pejcic, T. Scharbach, P. Shrestha, and C. Simmer, 2023: Fusion of radar polarimetry and atmospheric modeling. In V. N. Bringi, K. V. Mishra, & M. Thurai (Eds.), Advances in Weather Radar, Volume 2: Precipitation science, scattering and processing algorithms (pp. 293–344). IET The Institution of Engineering and Technology, ISBN-13: 978-1-83953-624-3, https://doi.org/10.1049/SBRA557G_ch7
