Simulated Real-Time Testing of the Prototype Implementation of the SOFIE Model: The 2025 Space Weather Prediction Testbed Exercise

TL;DR

This study reports on the on-site, simulated real-time testing of SOFIE, a physics-based SEP model within the CLEAR framework, during NOAA SWPC's Space Weather Prediction Testbed 2025. SOFIE integrates ambient solar wind modeling (AWSoM-R), CME flux rope generation (EEGGL), and shock-driven SEP acceleration/transport (M-FLAMPA) to produce two-dimensional proton-flux distributions on a Sun-centered sphere and Earth-directed time–intensity profiles. Results from two well-observed historical events (10 Sep 2017 and 4 Nov 2001) show that SOFIE can deliver 4-day SEP forecasts significantly faster than real time (e.g., 4-day predictions in about 5–13 hours on 1,000 CPU cores) while reproducing key CME and SEP features, though grid resolution and early connectivity can affect ESP timing and amplitude. The study demonstrates that physics-based SEP predictions are operationally viable, provided grid-refinement strategies and forecaster feedback are used to balance speed and accuracy, and it outlines a practical dual-setup workflow (coarse early forecasts followed by higher-accuracy runs) to support future human space exploration missions.

Abstract

The CLEAR Space Weather Center of Excellence's solar energetic particle (SEP) prediction model, SOlar wind with FIeld lines and Energetic particles (SOFIE), was run and evaluated on-site during the Space Weather Prediction Testbed (SWPT) exercise at NOAA's Space Weather Prediction Center (SWPC) in May 2025. As a physics-based SEP simulation and prediction model, SOFIE simulates the acceleration and transport of energetic particles in the coronal mass ejection (CME) driven shock in the solar corona and inner heliosphere. It has been validated against historical events. However, questions remain regarding whether a physics-based model, traditionally considered computationally expensive, could meet operational needs. The SWPT exercise offered a valuable opportunity to evaluate SOFIE's performance under simulated real-time conditions. Interactive feedback during the exercise from SWPC forecasters, SRAG console operators, CCMC personnel, and M2M SWAO analysts led to significant strategic improvements in the model setup to meet operational requirements. The resolution of the simulation domain was optimized by combining a coarser background grid with higher-resolution regions along the CME path and facing toward Earth, reducing computational cost without compromising accuracy. In this work, we present the operational performance of SOFIE and its capability to predict SEP fluxes significantly faster than real time. SOFIE was able to complete a 4-day SEP simulation within 5 hours on a supercomputer with 1,000 CPU cores during the SWPT exercise. This marks a critical milestone in demonstrating both the robustness and operational usefulness of SOFIE to support future human space exploration.

Figures (9)

• Figure 1: Schematic diagram of the prototyped SOFIE model suite (middle), as well as its inputs (left) and SEP outputs (right). Items with a marker were used in the SWPT exercise.
• Figure 2: The input photospheric magnetogram and the steady-state solar wind solutions for the 10 September 2017 event. (a) Input GONG magnetogram as of 15:04 UT on 10 September 2017, with the black dashed box marking the parent AR (NOAA AR 12673) for this event. (b) Angular grid resolution ($\Delta\phi$) of the steady-state simulation in the SC ecliptic plane in heliographic rotating (HGR) coordinates. The white solid circle at the center denotes the inner boundary of the SC domain at a heliocentric distance ($r$) of 1.1 $R_\mathrm{s}$. The HCS is plotted as orange lines, and the Sun-to-Earth connection line is plotted as a magenta dashed line. (c) Mesh size ($\Delta x$) of the steady-state simulation in the IH ecliptic plane in Carrington heliographic (HGC) coordinates. The black solid circle at the center denotes the inner boundary of the IH domain at $r = 20$$R_\mathrm{s}$, and the white dashed circle marks the IH shell at $r = 1.7$ au. The HCS is marked as orange lines. Multiple AMR criteria are indicated as: (I) HCS refinement within the IH shell, (II) HCS refinement beyond the IH shell, and (III) Earth-directed cone refinement. (d) Steady-state solar wind speed in the IH ecliptic plane in HGC coordinates. Multiple white curves with arrows indicate magnetic field lines, with the Earth-connected field line highlighted in magenta. The black solid and dashed circles represent heliocentric distances of 20 $R_\mathrm{s}$ and 1 au, respectively.
• Figure 3: CME simulation results during the 10 September 2017 event. Panel (a) shows the initial 3D CME flux rope inserted above the parent AR at the inner boundary of SC ($r=1.1\; R_\mathrm{s}$), viewed from Earth and colored by the radial magnetic field strength ($B_r$). Panels (b)–(c) show the LASCO/C2 observation and SOFIE-synthesized WL images at 16:11 UT on 10 September 2017. The color scale indicates the relative change in WL total brightness, expressed as the ratio of the CME signal to the background solar wind before the eruption. In both panels, the field of view is limited within $r = 6\; R_\mathrm{s}$, with the central black solid circle marking an occultation disk at $r = 2\; R_\mathrm{s}$. Panel (d) shows the flow speed in the IH ecliptic plane in heliographic inertial (HGI) coordinates, 40 hours after the eruption, when the ICME flank approaches Earth. Multiple magnetic field lines are marked as white curves with arrows, and a black dashed circle denotes the heliocentric distance of 1 au. The magenta and purple scatter points denote Earth and its 15$^\circ$ westward position in the ecliptic plane, respectively. Panel (e) shows the modeled solar wind plasma parameters at Earth (magenta) and its 15$^\circ$ westward position in the ecliptic plane (purple), together with ACE observations at Earth (black). Parameters shown from top to bottom are the plasma number density ($N_\mathrm{p}$), solar wind speed ($U$), plasma temperature ($T_\mathrm{p}$), and magnetic field strength ($B$).
• Figure 4: Distribution of energetic protons in the 10 September 2017 event. Panels (a) and (b) show SOFIE-modeled fluxes of $>$10 MeV and $>$100 MeV protons, respectively, on a logarithmic scale on the 1 au sphere, 10 hours after the eruption. In each panel, the $x$- and $y$-axes are Carrington longitude and latitude, respectively. The thin gray vertices correspond to the field line intersections with the 1 au sphere, and the edges indicate the triangulation skeleton constructed via the Delaunay triangulation approach delaunay1934spherelee1980two. The Earth's location is marked as "E" in a magenta circle, and the CME flux rope footpoint on the solar surface is marked as "F" in a green square. Panels (c) and (d) show the $>$10 MeV and $>$100 MeV proton time--intensity profiles at Earth, respectively, 6.27 real-time hours after the eruption, corresponding to the time when SOFIE predicts 10-hour proton fluxes. In each panel, GOES observations are plotted as a black curve, followed by a dark gray circle and a light gray curve indicating past, current, and future fluxes, respectively. SOFIE results are plotted as cyan and red curves with square markers, representing the $>$10 MeV and $>$100 MeV energy channels, respectively. A dashed-dotted vertical line represents the CME eruption time, and a vertical purple band marks the prediction leading time of 3.73 hours, also indicated in the legend. Panels (e) and (f) are similar to panels (c) and (d), but 12.73 real-time hours after the eruption, when SOFIE predicts 4-day proton fluxes, yielding a leading time of 83.27 hours.
• Figure 5: The input photospheric magnetogram, the angular resolution of the grid in SC, the mesh size in IH, and the steady-state solar wind solutions in IH for the 4 November 2001 event, shown in panels (a)–(d), respectively, with the same plot style as Figure \ref{['fig:steady201709']} except for adjusted plotting ranges and ticks.