- 1National Space Institute, Technical University of Denmark, Kongens Lyngby, Denmark (eljens@dtu.dk)
- 2Institute of Physics, University of Belgrade, Belgrade, Serbia
- 3DTU Compute, Technical University of Denmark, Kongens Lyngby, Denmark
- 4Danish Meteorological Institute, Copenhagen, Denmark
While thunderstorm processes, such as the acceleration of electrons to relativistic energies, are widely studied, the computational challenges involved have made definitive proofs difficult to acquire. High precision in electric discharge simulations is achieved by resolving particles individually, via for example Monte Carlo methods, rather than by applying a fluid approximation; however, this is computationally expensive, and the multiscale nature of thunderstorm processes incurs additional difficulties. To address these challenges, we have developed the Atmospheric STReamer And Relativistic Particle Engine (ASTRAPÉ), a fully 3D GPU-based Monte Carlo particle-in-cell code capable of tracing approximately 109 computational particles, modeling all relevant electron-molecule collisions and solving the Poisson equation to include space charge effects. We will present the particulars of the GPU implementation, along with benchmarking against existing data and performance metrics. Additionally, we will discuss code optimization for LUMI (Large Unified Modern Infrastructure), Europe’s first pre-exascale supercomputer, which allows for exceptionally fast streamer simulations. Finally, we will discuss how ASTRAPÉ can be used to study the generation of relativistic electrons in thunderclouds.
How to cite: Fangel-Lloyd, E., Gourbin, P., Dujko, S., Gammelmark, M., Karlsson, S., Jara Jimenez, A. R., and Köhn, C.: ASTRAPÉ: Atmospheric STReamer And relativistic Particle Engine – a GPU-based particle code for pre-exascale supercomputing , EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-8563, https://doi.org/10.5194/egusphere-egu25-8563, 2025.