"We want to make the LRZ better known in the seismology and geophysics community"
Satellite data illustrate the tectonical plate shift nera Ordin Turkey. In the green zone the plates meet together. Graphics: Sar4Tec/DLR
Just under 10 percent of the research projects for which computing is done at the Leibniz Supercomputing Centre (LRZ) belong to geophysics and seismology. But many more are to come: That is why the LRZ is sponsoring the annual meeting of the German Geophysical Society (DGG) from March 7 to 10, 2022, where geophysicists will exchange experiences and findings, discuss developments, and also new technologies. They can also get to know the computing centre based in Garching, its services as well as simulations and visualizations online. The focus is on the mentoring program for working groups that want to work with the High Performance Computers (HPC). "In the CXS Lab, researchers find experts from different scientific domains," explains André Kurzmann, who holds a doctorate in geophysics. "Generally, questions about software, codes and large data sets come up in about half of all HPC projects. Each project gets its own mentor, and we take care of technical problems, such as when implementing applications or when memory requirements increase." The LRZ is still supporting geophysics primarily with modeling and simulation, but artificial intelligence methods are also likely to be in demand soon, Kurzmann says in an interview
Why does geophysics need supercomputing? What does it compute on the HPC resources of the LRZ? Dr. André Kurzmann: Geophysics and seismology are classically among the pioneers in supercomputing, because for decades they have been processing a wide variety of data, for example on the history of the Earth, the composition of the Earth's crust, earthquakes and tsunamis, or from satellite data, and using them to calculate ever larger, more accurate models. In the meantime, the simulations are spatial, and the necessary calculations take days even on supercomputers. Prof. Dr. Hans-Peter Bunge from Ludwig-Maximilians-Universität and his team have simulated the development of the continents for many years, and today the model is three-dimensional and calculated in the highest resolution. With the help of SuperMUC-NG, Prof. Dr. Alice Gabriel from LMU and her research group were able to clarify why the tsunami that followed the earthquake in Palu, Indonesia, was so exceedingly strong. And last but not least, the LRZ built terrabyte together with the German Aerospace Center, via this high-performance platform for data analysis, remote sensing data are to be immediately analyzed on our supercomputers.
Does geophysics already rely on artificial intelligence methods, which also require computing power? Kurzmann: The focus of geophysics projects at the LRZ is still on calculations and simulation, but machine and deep learning or pattern recognition are interesting in the field of tomography, for example, when geological layers are examined and visualized with radio waves. Artificial intelligence is also advancing the study of earthquake waves or remote sensing data.
The CXS Lab at LRZ supports researchers in geophysics with a mentoring program. Why? Kurzmann: Computing time and supercomputing are valuable, we want to use them efficiently and therefore we support research teams. In the CXS Lab, they find experts from various scientific disciplines. In general, about half of all HPC projects involve questions about software, codes and large data sets. Each project is assigned a contact person. Mentors take care of technical problems, for example when implementing applications or when memory requirements increase. And they often push the computing work by asking questions about the current status of a project. The program is very well received by researchers.
Where do geophysicists particularly often need help? Kurzmann: Geophysicists are do-it-yourself researchers who develop their own software and algorithms. They usually need help with specific problems, such as porting their code to the supercomputer or optimizing their algorithms when the workload of thousands of computational cores doesn't work.
You're a geophysicist yourself - which projects do you particularly enjoy as a mentor? Kurzmann: The projects where I can provide advice and practical support and maybe even take a project further. Sometimes projects get into one-way streets, so when I can help steer research questions in a new direction, it makes me happy. At the interface between medium-sized and large-scale projects, there is also often a need for advice on applications or technology. Here, too, I can contribute my experience and push projects forward.
The LRZ is presenting itself at the annual meeting of the German Geophysical Society: What are you presenting and why? Kurzmann: We want to make the LRZ better known in the seismology and geophysics community. To this end, we are presenting the SuperMUC-NG and the HPC resources, our services and successful work online on a special website. We also show how researchers can present their projects in the Mozilla Hubs. By the way, the simulation of Professor Bunge mentioned at the beginning can already be seen there. (interview: vs)
Dr. Andrè Kurzmann, Geophysist and mentor at LRZ CXS-Lab