Combining radar POlarimetry, weather forecast MOdel outputs and DOppler radar observations for RIming analysis (POMODORI)


Joint project between
Ludwig-Maximilians-University of Munich (LMU) and Deutscher Wetterdienst (DWD), Phase 2

LMU: Paul Ockenfuß (PhD), Stefan Kneifel (PI)
DWD: Mathias Gergely (Scientist), Michael Frech (PI)

Abstract

The idea behind PROM-POMODORI is to gain a better quantitative understanding of the riming process, which is one of the main growth mechanisms of frozen precipitation and yet is difficult to identify in remote sensing observations and to faithfully represent in atmospheric models. PROM-POMODORI aims to overcome current challenges in investigating riming by leveraging a combination of polarimetric measurements from the German C-band radar network, vertically pointing Doppler radar measurements from Ka-band cloud radars and C-band weather radars, and outputs from atmospheric models over Germany. The focus is estimating the degree of riming both in an atmospheric column above the radars locally and on assessing the spatiotemporal dimensions and the variability of riming events.


pomodori_prom_abstract.jpg

Figure 1: Correlate riming retrievals from vertically pointing cloud radars with measurements of DWD C-band radar polarimetric variables and atmospheric (thermo)dynamics from ICON-D2 model outputs. First, analyzing riming above radar sites and then throughout the radar scan volume.

Status 2025

Contribution of Ludwig-Maximilians-University of Munich

In the beginning of the POMODORI project, we have focused on the extension of the existing method for riming detection by Kneifel and Moisseev, 2020. By clustering riming into distinct events, we were able to perform a comprehensive characterization of riming events at two cloud radar sites in Germany. This includes the frequency, duration, spatial extent and temperature characteristics of strong riming in stratiform clouds. We find those events to be occuring mainly between $0°\text{C}$ and $-15°\text{C}$, with average durations of around $20\,\text{min}$. We also searched for systematic differences between riming and non-riming situations in other datasets, e.g. radiosonde profiles. Here, we see significant shifts towards higher liquid water path for the soundings during riming events. These results have been published in Ockenfuß et al., 2025. Since then, we have investigated if a similar analysis would also be possible based on the operational birdbath scan, performed by the 17 C-band radars operated by the German Weather Service. For this task, the operational birdbath data from 2021 to 2025 was clutter filtered and transferred into an analysis-ready dataset. Based on these datasets, we were able to assess the general potential of the operational birdbath scan. We compared the effective sensitivity (i.e. after clutter removal) of the birdbath scan with long-term cloud radar observations and we presented two case studies of a stratiform and convective event, where we are able to directly compare the reflectivity, mean Doppler velocity and Doppler spectra of the operational radars with a research cloud radar. In the stratiform case, we focus especially on riming, while the convective case exhibits strong hail signatures. In close cooperation between DWD and LMU, those results were submitted for publication to the Bulletin of the American Meteorological Society and are currently under review.

In order to further explore the potential for quantitative retrievals from the operational birdbath data, the DWD investigated the information contained in convective hail Doppler spectra, while at the LMU, we are focusing on riming in stratiform clouds.

The previously developed riming detection, which relies on the increase in detected mean Doppler velocity, serves as our benchmark. To validate the transfer of the retrieval to the operational birdbath scan, we first test it on a “mockup” birdbath dataset, constructed by downsampling cloud radar data to match the resolution of the operational birdbath scan. Since in general there is no additional sensor data at the operational radar sites, a melting layer detection algorithm for the operational birdbath scan had to be developed first. Finding good agreement between mockup and benchmark datasets, we apply the new retrieval to radar data recorded during the winters of 2021 to 2024. This results in the first nationwide distribution of riming in wintertime clouds. There is a north-south gradient in the distribution, which can be linked to Germany's precipitation climatology. Notably, we show that the occurrence of riming events correlates more strongly with precipitation intensity than with the total number of precipitation hours across sites. The temperature distribution associated with riming is consistently between $-15°\text{C}$ and $0°\text{C}$ at all sites, except at the Feldberg site, which hints at a possible mountain effect. These results are currently prepared for publication.

In summary, we were able to demonstrate the potential of the operational birdbath scan as a valuable tool for cloud and microphysical research. Our findings also highlight general challenges and solutions involved in transferring retrieval algorithms from research-grade cloud radars to the operational birdbath scan.


Figure 2: Spatial distribution of riming event frequency (a) and total wintertime precipitation (b) in Germany.


Contribution of Deutscher Wetterdienst

The work at DWD focused on analyzing the vertically pointing birdbath scans that were implemented across the German C-band radar network in PROM phase 1: HydroColumn (Gergely et al., 2022) and on collecting and providing radar data for further analysis by our POMODORI project partners at LMU and by collaborators at University of Bonn and at DLR.
In PROM-POMODORI, we extended the signal processing chain of the Doppler spectra recorded during the operational DWD birdbath scans to also allow postprocessing and analysis of the Doppler spectra collected in severe convective storms (Gergely et al., 2024). All spectral postprocessing routines are summarized in the PyBathSpectra Python toolkit that is publicly available on GitHub (Gergely 2024). This approach to spectral postprocessing of weather radar Doppler spectra allows for the first time a detailed retrieval of the hail size distribution N(D) and an estimation of the vertical air motion of convective storms (Gergely et al., 2025). Three examples of retrieved N(D) are shown in Fig. 3. The full retrieval algorithm is published in the PyBathHail package (Gergely 2025) and could also be tested for Doppler spectra recorded with other weather radars.


Figure 3: Summary of profiles of hail size distributions N(D) retrieved from C-band radar birdbath scans for 3 hailstorms (with consistent scaling of x- and y-axes for all storms): storms A and B were recorded by the MHP radar in southern Germany, storm C by the FLD radar in central Germany. Adapted from Gergely et al. (2025).


Together with our project partners at LMU, we have explored how the operational DWD birdbath scans can be analyzed similarly to shorter-wavelength cloud research radar data (Frech et al., 2024). Based on long-term time-series of operational weather and cloud research radar data since spring 2021, we found a surprisingly good agreement between weather features detected by the German C-band radars and nearby Ka-band cloud radars over a wide range of precipitation conditions (see Fig. 1). But the weather radars also allow an analysis of severe weather with intense precipitation (like hailstorms) in contrast to cloud radars that are strongly affected by atmospheric attenuation in these weather conditions. Conversely, C-band radars can miss weak radar echos of very thin clouds or near the cloud top. These findings are very promising for transferring existing quantitative riming retrievals for cloud radars (Kneifel and Moisseev, 2020) to the national C-band radar network and determining long-term spatial and temporal statistics of riming events over larger regions (Ockenfuß et al., 2025).

For our collaboration with the PROM-POLICE project at the University of Bonn, we provided DWD birdbath scan data and postprocessing routines for the identification of significant riming in the radar data and for transferring these results to quasi-vertical profiles of polarimetric C-band radar scans via machine learning (Blanke et al., 2025). While a detailed quantification of riming in the 3D atmospheric volume covered by all German C-band radars still remains out of reach for now, the results show that the occurrence of significant riming is often closely correlated with polarimetric variables (e.g. the depolarization ratio) that can be measured at a high spatial and temporal resolution across the entire lower atmosphere above Germany, offering a promising avenue to further research efforts.

We also provided DWD radar data from operational sites as well as from the dedicated test and research radar in Hohenpeißenberg to collaborators of the PROM-IcePolCKa project at DLR/LMU. They found benefits for a synergistic utilization of weather radar polarimetry and high-resolution vertically pointing cloud radars, depending on weather radar coverage and vertical resolution (Heske et al., 2025).

References

  • Blanke, A., M. Gergely, and S. Trömel, 2025: A new aggregation and riming discrimination algorithm based on polarimetric weather radars, Atmos. Chem. Phys. 25, 4167-4184, https://doi.org/10.5194/acp-25-4167-2025.
  • Frech, M., S. Kneifel, P. Ockenfuß, and M. Gergely, 2024: Exploring the untapped potential of operational weather radars for vertical profiling of precipitation and clouds, Bull. Amer. Meteor. Soc., accepted with minor revisions.
  • Gergely, M., M. Schaper, M. Toussaint, and M. Frech, 2022: Doppler spectra from DWD's operational C-band radar birdbath scan: sampling strategy, spectral postprocessing, and multimodal analysis for the retrieval of precipitation processes, Atmos. Meas. Tech. 15, doi: 10.5194/amt-15-7315-2022, 2022.
  • Gergely, M., P. Ockenfuß, S. Kneifel, and M. Frech, 2024: Postprocessing methods to characterize multimodal precipitation in Doppler spectra from DWD's C-band radar birdbath scan, 12th European conference on RADar in meteorology and hydrology (ERAD 2024) – Book of Extended Abstracts, L. Baldini, M. Montopoli, E. Adirosi, A. Bracci, F. Napolitano, and N. Pierdicca, Eds., Citta Universitaria Sapienza University of Rome, Rome, Italy, https://doi.org/10.5281/zenodo.14170845.
  • Gergely, M., 2024: PyBathSpectra Python package (following Gergely et al. 2022, 2024), MIT license, https://github.com/birdbathDWD/PyBathSpectra.
  • Gergely, M., P. Ockenfuß, S. Kneifel, and M. Frech, 2025: Retrieval of the hail size distribution and vertical air motion from weather radar Doppler spectra, J. Atmos. Oceanic Technol., submitted.
  • Gergely, M., 2025: PyBathHail Python package (following Gergely et al. 2025), MIT license, https://github.com/birdbathDWD/PyBathHail.
  • Heske, C., F. Ewald, and S. Groß, 2020: Augmenting the German weather radar network with vertically pointing cloud radars, 2025: implications of resolution and attenuation, EGUsphere [preprint], https://doi.org/10.5194/egusphere-2025-691.
  • Kneifel, S., and D. Moisseev, 2020: Long-term statistics of riming in nonconvective clouds derived from ground-based Doppler cloud radar observations, J. Atmos. Sci. 77, doi: 10.1175/JAS-D-20-0007.1.
  • Ockenfuß, P., S. Kneifel, M. Gergely, and M. Frech, 2025: Spatial and temporal scales of riming events in nonconvective clouds derived from long-term cloud radar observations in Germany, J. Geophys. Res. Atmos. 130, e2024JD042180, https://doi.org/10.1029/2024JD042180.
  • Ockenfuß, P., 2023: Exploring spatial scales and favorable conditions of riming using long-term cloud radar observations, 4th International Summer Snowfall Workshop, 11 - 13 September 2023, Leipzig, Germany, oral presentation, https://monsun.meteo.uni-leipzig.de/~drops/issw4/.
  • Ockenfuß, P., S. Kneifel, M. Gergely, M. Frech, 2024: Spatial and Temporal Scales of Riming Events Derived from Long-term Cloud Radar Observations in Germany, International Conference on Clouds and Precipitation, 14 – 19 July 2024, Jeju, South Korea.
  • Ockenfuß, P., S. Kneifel, M. Gergely, and M. Frech, 2024: Potential for Profiling of Snow and Precipitation with Operational C-Band Birdbath Scans. In 12th European conference on RADar in meteorology and hydrology. Book of Abstracts, https://doi.org/10.5281/ZENODO.14099624.
  • Ockenfuß, P., M. Gergely, S. Kneifel, and M. Frech, 2025: Melting Layer and Riming Detection from Vertically Pointing C-band Radar Observations, PrePEP - Conference: Precipitation Processes – Estimation and Prediction, 16 - 21 March 2025, University of Bonn, Germany, oral presentation, Book of Abstracts, https://doi.org/10.5281/ZENODO.15735778.
  • Ockenfuß, P., G. Köcher, and S. Kneifel, 2025: Visualizing the Spatial Structure of Strong Riming Events using Scanning Cloud Radars, EGU General Assembly 2025, 27 April - 2 May 2025, Vienna, Austria, oral presentation.

Status 2024

For the first year of PROM-POMODORI at LMU, we focused mainly on data from the vertically pointing Ka-band cloud radars located in Jülich and Lindenberg. The main objective was to create a database of riming events, serving as a training dataset and benchmark for future riming retrievals based on the C-band radars of the German Meteorological Service (DWD). To achieve this, we re-implemented the riming retrieval from Kneifel et al. (2020), which is based on the higher fall velocity of rimed particles compared to pristine crystals. The algorithm was improved and subsequently applied to 14 years of data (2010-2024). From the algorithm output, we created a database of continuous riming events. Four examples of riming events are shown in Figure 1.

Combining this database with thermodynamic profiles from atmospheric models, we were able to derive a high number of characteristics for each event, e.g. duration, height, vertical extent, temperature, etc. Using wind information from the NWP models, we investigated different approaches to estimate the horizontal extent of the detected events. This analysis has provided insight into the typical temporal and spatial scales and the associated variability of riming, which forms the groundwork to implement a polarimetric C-band retrieval, but is also interesting from a microphysical perspective.

For the Lindenberg site, we also investigated possible correlations between riming and radiosonde measurements. Particularly, we derived estimates of the liquid water content from the radiosonde and showed that riming events have a significantly higher probability to coincide with high liquid water content than similar ice clouds where no riming is observed.
Currently, we are moving toward C-band radar data and analyze the potential for identifying and estimating riming from these weather radar measurements at longer wavelengths.




Figure 1: Selected examples of riming events. The upper part of each panel shows Dopplervelocity, the lower part the retrieved rime mass fraction. Hatched areas are excluded due to a high convection index. Events are indicated by a green overlay. The number of samples per event is depicted by the green number.


For riming events identified by the LMU project partners based on Ka-band cloud radars, the DWD researchers have analyzed concurrent polarimetric measurements from the operational scanning C-band weather radars to see whether a typical riming signature can also be observed in the C-band polarimetric data recorded above the cloud radars. Figure 2 shows an example where such a riming signature is indeed evident in the C-band radar volume-scan data: elevated radar reflectivity, differential reflectivity close to 0, and a very high correlation coefficient approaching a value of 1. However, other riming events identified in Ka-band cloud radar data do not show any clear signature in the corresponding polarimetric C-band radar measurements.

To gain a better quantitative understanding of riming across Germany, DWD has provided long-term C-band radar data from the operational (vertically pointing) birdbath scans (Frech et al. 2017) of all 17 C-band radars from 2021 through 2023 to the project partners at LMU for transferring the riming retrieval algorithm which has been extensively tested for vertically pointing cloud radars to the C-band radar birdbath scan. Additionally, DWD continues to explore the new birdbath scan (developed as part of PROM phase 1 and implemented in DWD's operational radar scanning cycle; see Gergely et al. 2022) in the context of analyzing convective storms and, particularly, for estimating hail characteristics relevant to assessing the damage potential of these thunderstorms.




Figure 2: PPI of Essen C-band radar at 2.5° elevation in October 2021, recorded when a significant riming event was identified based on the vertically pointing Ka-band cloud radar at Jülich. Here, a characteristic riming signature can also be observed in the polarimetric radar data above the cloud radar site.


PROM-POMODORI partners have presented these results at the 2022 North American Hail Workshop, at the AMS Conference on Radar Meteorology 2023, and at the International Summer Snowfall Workshop 2023.

Looking ahead, the next tasks of the POMODORI project include publishing the results obtained in the first year of the project in peer-reviewed journals. Estimating the degree of riming from the long-term dataset of operational C-band radar birdbath scans will also provide a first estimate of the variability of riming across larger regions.

References