ENVIRONMENT AND ENERGY
Creating Virtual Laboratories at High Resolutions for Supporting Meteorological Field Campaigns
Principal Investigator:
Prof. Dr. Roel A. J. Neggers
Affiliation:
University of Cologne, Germany
Local Project ID:
VIRTUALLAB
HPC Platform used:
JUWELS GPU and CPU at JSC
Date published:
Introduction
The VIRTUALLAB project made use of the JUWELS cluster to perform high-resolution atmospheric simulations based on measurements from various meteorological sites and field campaigns. The resolution of these numerical experiments is high enough to resolve small-scale phenomena such as turbulence, convection and associated clouds. These simulations create a "virtual laboratory", allowing scientists to fill existing data gaps to increase our insight into these phenomena and further improve their representation in models used for weather- and climate prediction. Various climate regimes were simulated including the marine subtropics, mid-latitude continental areas, and the high Arctic. A second goal of VIRTUALLAB was to explore new ways of evaluating the simulations against hemispheric camera images, making use of ray-tracing techniques as applied in high-end gaming.
Figure 1: Ray-tracing rendering of a shallow convective cumulus cloud field in a Large-Eddy Simulation (LES) performed on the JUWELS cluster, as part of the VIRTUALLAB project. The meteorological conditions are those observed over Jülich, Germany during the SOCLES campaign in the summer of 2021.
Report
Convective cloud fields play an important role in Earth's climate system by vertically mixing air, generating surface precipitation, changing the circulation in the troposphere, and by affecting its energy and water budgets. While convection has been intensely researched in the past decades, its representation in weather and climate models is still less than perfect. This in particular applies to convective cloud fields that are strongly organized, for example in the form of squall lines, mesoscale complexes, and cloud clusters over oceans, among others. This problem is at least partially caused by a lack of understanding of this phenomenon, which in turn results from existing data gaps in available meteorological measurements.
The overarching goal of the VIRTUALLAB project was to perform high-resolution simulations of organized convection on supercomputers based on observations made during recent meteorological field campaigns and at fixed sites. Large-Eddy Simulations (LES) can provide a best estimate of the natural small-scale variability in the atmosphere that directly surrounds a measurement platform. This allows scientists to fill existing data gaps on processes that cannot yet be fully or adequately observed. As a result, point measurements made at sites of interest can be better interpreted, while measurement strategies can be virtually tested before actual deployment in the field. Supercomputing is needed for this effort, due to the high spatial and temporal resolutions required to fully resolve organized convection across a range of scales, at turbulence-resolving resolutions in mesoscale-permitting domains. A second goal of the VIRTUALLAB project is to use ray-tracing methods as used in modern gaming software to realistically render simulated clouds fields, and use these to test the realism of the clouds in the simulation against images from hemispheric camera systems on the ground. An example of such cloud rendering is shown in Fig. 1.
Use was made of the JUWELS cluster. Both CPUs and GPUs where employed, the former for the cloud simulations, and the latter for the ray-tracing renderings. The simulations typically cover a grid of about 500 x 500 x 300 points, horizontally spaced at about 25-100 m. To perform such large simulations at small timesteps (on the order of seconds), massive parallel computing is required, for which the JUWELS cluster is well suited. Simulations were performed for observed cloud fields at three locations: i) the JOYCE meteorological site in Jülich (geomet.uni-koeln.de/forschung/actris/joyce), ii) the EUREC4A international field campaign that took place in the Carribean sector of the Atlantic (eurec4a.eu), and iii) the MOSAiC international drift experiment in the high Arctic (mosaic-expedition.org), during which the Polarstern Research Vessel drifted with the sea ice for almost a full year.
The scientific work performed during the VIRTUALLAB project yielded various important findings and insights. The simulations for the JOYCE site, and their subsequent evaluation using ray-tracing rendering, provided proof-of-concept for using this technique for evaluating cloud fields in LES experiments (Burchart et al., 2024). In addition, new insights were obtained into how well LES can reproduce the observed cloud fields (Burchart et al., 2025). The simulations for the EUREC4A campaign yielded results that contribute to an international effort to evaluate LES for organized cloud transitions in the Trade Wind flow (Raghunathan et al., 2025). For the MOSAiC campaign the project helped in performing daily simulations covering the full year (Schnierstein et al., 2024), providing a rich virtual database on turbulence and clouds in the Arctic that supports ongoing research on this topic (Neggers et al, 2025).
The VIRTUALLAB project was thus successful in achieving its goals. From a science perspective, various new insights were obtained on clouds and their role in Earth's climate, but also on how well LES simulates these phenomena, and how such simulations can best be configured based on observational data. This knowledge and expertise advance our knowledge of convective clouds, and inform future applications of LES by the scientific community. The work also provided virtual datasets on convective clouds to accompany observational data, which will be useful for improving the representation of this cloud type in weather and climate models. This effort will thus benefit society as a whole, in providing more accurate weather forecasts and in reducting uncertainty in scenarios of future climate change.
References
Burchart, Y., C. Beekmans and R. A. J. Neggers, 2024: A Stereo Camera Network simulator for Large-Eddy Simulations of continental Shallow Cumulus clouds based on three-dimensional Path-Tracing. J. Atmos. Modeling Earth Syst., doi.org/10.1029/2023MS003797
Burchart. Y., B. Pospichal and R. A. J. Neggers, 2025: Confronting Large-Eddy Simulations with Stereo Camera Data by means of reconstructed hemispheric Cloud Size Distributions. Under review for publication in J. Atmos. Modeling Earth Syst., April 2025
Neggers, R. A. J., J. Chylik and N. Schnierstein, 2025: The entrainment efficiency of persistent Arctic mixed-phase clouds as inferred from daily large-eddy simulations during the MOSAiC drift. Accepted for publication in JAS. [preprint on ESSOAr], doi.org/10.22541/essoar.172710718.86866606/v1
Raghunathan, G. N., P. Blossey, S. Boeing, L. Denby, S. Ghazayel, T. Heus, J. Kazil, and R. A. J. Neggers, 2025: Flower-type organized trade-wind cumulus: A multi-day Lagrangian Large Eddy Simulation intercomparison study. Under review for publication in J. Atmos. Modeling Earth Syst., April 2025
Schnierstein, N., J. Chylik, M. D. Shupe and R. A. J. Neggers, 2024: Standardized daily high-resolution large-eddy simulations of the Arctic boundary layer and clouds during the complete MOSAiC drift. J. Atmos. Modeling Earth Syst., doi.org/10.1029/2024MS004296