Parameter Optimization of SEP Acceleration and Transport Models Towards Forecasting: Application to Multi-Spacecraft Observations of the 9 October 2021 Event

Solar energetic particle (SEP) events pose severe risks to manned spaceflight and most space infrastructure, such as satellites, impacting the broader modern technology-based society. Accurate and timely prediction of SEP events is therefore essential. However, SEP propagation to near-Earth involves multiple complex physical processes, while observational constraints remain limited. Consequently, no physical model has yet been established that can reproduce the full chain of processes accurately and efficiently. We aim to provide physically interpretable reproduction and prediction of SEP events from pre-flare solar observations to near-Earth particle fluxes by improving and connecting models for each process. This presentation focuses on optimizing free parameters in the SEP acceleration and transport model. We coupled a CME propagation model (SUSANOO, SUSANOO-CME; Shiota et al., 2014, 2016) with an acceleration and transport model with the diffusive shock acceleration and the focused transport equation (Minoshima et al., in review), and applied it to the SEP event on 9 October 2021. We attempted simultaneous reproduction of proton fluxes across five spacecraft. We introduced spatial refinement by treating key parameters (injection efficiency ε, acceleration efficiency ξ, and transport mean free path λ) independently for each magnetic field line or heliolongitude, enabling a detailed representation of spatial structure. Furthermore, we implemented black-box optimization using an evolution strategy to robustly and efficiently explore a high-dimensional parameter space. As a result, our framework achieved a lower mean absolute error (MAE) than grid search. Parameter importance analysis revealed that injection efficiency (ε) exerts the strongest influence on proton flux, consistent with physical understanding. This framework achieves accuracy and computational efficiency, representing a significant step toward generalizing initial condition settings in SEP prediction models. Future work will apply this approach to multiple events to establish a model suitable for operational forecasting.