Developing the Next Generation Space Weather Modeling Framework: Year 2
Abstract
Supported by the Space Weather with Quantified Uncertainty (SWQU) NSF program, we have been developing the Next Generation Space Weather Modeling Framework at the University of Michigan for two years. The main goal of the project is to provide useful probabilistic forecast of major space weather events about 24 hours before the geospace impact occurs. We are using the first-principles models in the Space Weather Modeling Framework in combination with uncertainty quantification and data assimilation. Using the advanced MaxPro experimental design and fully automated Python scripts, we have performed about a thousand simulations with our solar corona and heliosphere model generating steady state solar wind solutions and coronal mass ejections. Based on these simulations, we have performed the uncertainty quantification analysis using the Bayesian inversion formula and a newly defined distance metric adapted to solar simulations. One important finding is that the physically meaningful range of the background solar wind model parameters depends on the solar cycle. We have identified the three most important parameters that impact the background solar wind model and a fourth parameter that impacts the CME eruption model. The reduced dimensionality of the parameter space enables reducing the size of the ensemble. Data assimilation provides further opportunity to improve the predictions. We are using in-situ observations at L1 prior to the CME and coronal while-light image observations right after the eruption to find the optimal parameters for the ensemble simulations. To make the ensemble simulations feasible, the model should take advantage of GPUs. We have already ported the Geospace model to run efficiently on a GPU. The operational Geospace model can run on a single GPU significantly faster than real time at the same speed as using about 100 CPU cores. We are currently porting the solar corona and heliosphere model to multiple GPUs. The main product of the project, the Michigan Sun-to-Earth Model with Quantified Uncertainty and Data Assimilation (MSTEM-QUDA) is available as an open-source distribution at https://github.com/MSTEM-QUDA to the entire community.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2022
- Bibcode:
- 2022AGUFMSM25C2001T