Michigan Geospace Model With Ensemble Forecasting and Particle Filtering: Modeling the May 2024 Extreme Storm

Abstract This paper introduces an updated version of the University of Michigan Geospace model. The Geospace model has been used for operational forecasting at the Space Weather Modeling Center of the National Oceanic and Atmospheric Agency since 2016 with regular updates. The current operational version, Geospace V2, was launched in 2021 and is driven by solar wind and magnetic field observations near the first Lagrange (L1) point. The model provides global indices and local magnetic perturbation predictions that space weather forecasters use to assess risks posed by geomagnetic storms. The new model, Geospace V3, uses an ensemble of simulations to mitigate the sensitivity of the model to small perturbations, and instead of a single solution can provide a distribution of forecasts. Furthermore, we introduce a particle‐filter‐based version Geospace V3 + Dst, where at each hour‐mark, the real‐time Dst index is used to select the best‐performing ensemble member, whose output is then used to continue the simulation. Lastly, we document a new version Geospace V3E optimized for extreme storms that uses a reduced radius for the inner boundary of the global magnetohydrodynamic model BATS‐R‐US. We validate the new model for the May 2024 extreme geomagnetic storm. We discuss issues of observational and propagation errors of the L1 observations used at the upstream boundary of the Geospace model. We show that the new Geospace model produces significantly improved SYM‐H/SMR and CPCP index predictions compared to the current operational version.