Speaker
Description
Understanding the dynamics of protons within the Earth's magnetosphere is crucial for assessing the long-term impact of space weather on satellite infrastructure. Accurate simulation of these dynamics often relies on solving a Fokker-Planck diffusion equation, with precise modeling of proton sources and losses as a central requirement. In this work, we present a significant advancement in modeling solar energetic protons (MeV and above) access to the Earth’s magnetosphere, tailored for use in long-term space climate models.
Our model uses a time-reversed Bulirsch-Stoer proton trajectory integrator to calculate, with high fidelity, the fraction of proton flux from interplanetary space that reaches a target location in the magnetosphere (e.g., the transmissivity and the cut-off energy).
To maximize performance and facilitate integration into modern space weather modeling pipelines, we have reengineered a legacy MPI-Fortran codebase into a Python-based, GPU-accelerated framework utilizing CuPy and Jinja. Core IRBEM routines relevant to magnetic field and coordinate transformations are being rewritten in CUDA C++ for efficient GPU execution. The particle pusher’s high level of optimization enables the rapid generation of dense precomputed tables that map proton access probabilities across relevant phase-space parameters (energy, equatorial pitch angle, L*, MLT) and environnement conditions (Kp). These tables can be directly employed by diffusion models such as Salammbo 3D, providing both greater statistical accuracy and computational efficiency than previous approaches.
Moreover, to ensure the robustness of our approach, we investigated different numerical scenarios for proton trapping and boundary conditions within the diffusion equation solver, including a « porous » boundary that enables modeling the trapping mechanism.
We will present results from extensive long-term simulations, benchmarking our model’s predictions against both legacy shielding models and in-situ satellite measurements, thus assessing the value of our approach for future operational space weather and space climate modeling.
Finally, by combining advanced physical modeling capabilities and modern GPU acceleration, our approach not only improves the resolution and adaptability of proton access simulations, but also opens up new possibilities for GPU-based computation throughout the space climate modeling pipeline. In particular, it paves the way for broader GPU utilization within the IRBEM library (e.g., for IGRF and Tsyganenko magnetic field calculations, or for geographic-to-magnetic coordinates conversion, including L* calculation).
| Do you plan to attend in-person or online? | In-person |
|---|
