Space Weather Modeling Framework (SWMF)

• Space Weather Modeling Framework (SWMF) is a modular coupled system that integrates multi-physics models to simulate the space weather environment from the Sun to Earth.
• Its core component, BATS-R-US, leverages advanced MHD techniques and adaptive mesh refinement to model diverse regimes including ideal, resistive, and Hall MHD.
• SWMF is well validated against multi-instrument data, supports community access via CCMC, and has transitioned from research to operational forecasting.

The Space Weather Modeling Framework (SWMF) is a modular coupled-system framework developed at the University of Michigan by the Center of Space Environment Modeling (CSEM), with the BATS-R-US extended MHD code as its core element. It was built to simulate the space-weather environment “from the upper solar chromosphere to the Earth’s upper atmosphere and/or the outer heliosphere,” and its development over roughly a quarter of a century produced a capability suitable both for research use by the space physics community through the Community Coordinated Modeling Center (CCMC) and for operational use by the NOAA Space Weather Prediction Center (SWPC) (Gombosi et al., 2021). In the literature, SWMF is presented not as a single physics model, but as a high-performance modular framework that couples specialized models across a multi-scale, multi-domain, multi-physics system.

1. Historical formation and institutional setting

SWMF emerged in an environment in which numerical models used in forecast were developed in the framework of individual research projects and the most promising models were later selected for additional testing at SWPC. In the United States, this broader system involved NSF and NASA for scientific research, NOAA/SWPC for forecast operations, and CCMC for increasing the application of models in research and education. CCMC was created in 1998 at NASA’s Goddard Space Flight Center to broaden the user base for numerical modeling of solar and heliospheric phenomena, promote developing interfaces between different models, employ scientific models for teaching purposes, and provide opportunity collaboration between modelers and additional testing of models (Pevtsov, 2016).

Within that setting, CSEM at the University of Michigan sustained the development and maintenance of SWMF and BATS-R-US. The framework’s trajectory is described as unusual for a university-based effort because it required long-term stability, interdisciplinary collaboration, code integration, verification and validation, user support, HPC adaptation, and operational transition. The published account quantifies that investment as more than \$50M and about 200 person-years, and identifies CSEM as the institutional home that organized collaboration among space physicists, applied mathematicians, computational scientists, software engineers, and later data scientists and operations users (Gombosi et al., 2021).

2. Architecture and principal components

Architecturally, SWMF is explicitly a modular coupled-system framework. By 2021 it comprised about a dozen physics domains and models and more than 1 million lines of Fortran 2008 and C++ plus supporting scripts and tools. Its core solver, BATS-R-US—“Block Adaptive-Tree Solar-wind Roe-type Upwind Scheme”—provides the extended MHD/XMHD capability used in many component domains. The code can be configured for ideal and resistive MHD, semi-relativistic MHD, anisotropic MHD, Hall MHD, multispecies, multi-fluid, and non-neutral multifluid plasmas. Its adaptive mesh refinement, shock-capturing Godunov/Roe-type finite-volume heritage, and scalability on parallel machines are identified as foundational engineering achievements (Gombosi et al., 2021).

Two configurations are especially central. The first is AWSoM/AWSoM-R, which couples the solar corona (SC) to the inner heliosphere (IH) and can extend toward the outer heliosphere. The second is the SWMF/Geospace Model, which couples the global magnetosphere (GM), inner magnetosphere (IM), and ionospheric electrodynamics (IE). In the standard geospace setup, BATS-R-US serves as GM, the Ridley Ionosphere Model (RIM) supplies ionospheric electrodynamics, and the Rice Convection Model (RCM) represents the ring current or inner magnetosphere; the Radiation Belt Environment (RBE) model can also be added. The coupling logic is concrete: BATS-R-US sends near-Earth field-aligned currents to RIM, RIM solves for electric potential and returns boundary electric fields and flows, and RCM exchanges pressure and density with the MHD domain inside closed field lines. Additional coupled or couplable elements include M-FLAMPA for solar energetic particles, PWOM for polar wind or ionospheric outflow, radiation belt models, and embedded PIC regions through MHD-EPIC and MHD-AEPIC (Gombosi et al., 2021).

3. Physical formulations and numerical methods

For solar applications, a central concept in SWMF is that the gradient of Alfvén-wave pressure drives the solar wind and the dissipation of Alfvén-wave turbulence heats the corona. In AWSoM, the outward Poynting flux at the solar surface is assumed proportional to magnetic field strength, $S_A \propto B$, with $S_A/B$ treated as a tunable parameter. Published boundary values for current coronal and CME models include $T = 5\times 10^4\,\mathrm{K}$ and $n = 2\times 10^{17}\,\mathrm{m}^{-3}$ at the upper chromosphere or low boundary. In the 2019 AWSoM validation study, the inner boundary is placed at the base of the transition region with $N_e = N_i = 2\times 10^{17}\,\mathrm{m^{-3}}$, $T_e = T_i = T_{i\parallel} = 50{,}000\,\mathrm{K}$, $$(S_A/B)_\odot = 1.0\times 10^6\,\mathrm{W\,m^{-2}\,T^{-1}}$$, and $$L_\perp \sqrt{B} = 1.5\times 10^5\,\mathrm{m}\,\sqrt{\mathrm{T}}$$; no ad hoc coronal heating term is added (Gombosi et al., 2021, Sachdeva et al., 2019).

The numerical formulations span fluid, electrodynamic, and kinetic regimes. AWSoM uses BATS-R-US for the MHD solution; in the low corona, the 2019 validation study used a fifth-order MP5 scheme inside $1.5\,R_\odot$ and standard second-order shock-capturing elsewhere. In geospace applications, SWMF can compute ground magnetic perturbations using Biot–Savart integrals. In kinetic extensions, the framework uses Hall MHD for stability at the interface between MHD and PIC regions and employs divergence control to prevent accumulation of $\nabla\cdot\mathbf{B}$ errors. A quantitative scaling result reported for embedded PIC is that if ion or electron mass-to-charge scales are modified by a factor $S_A/B$0, the PIC cost can drop by roughly $S_A/B$1; for $S_A/B$2, the cost reduction is approximately $S_A/B$3 (Gombosi et al., 2021).

Methodological development within SWMF has continued beyond the framework papers. A 2024 solar energetic particle study presents a numerical scheme that conserves the number of particles based on integral relations for Poisson brackets, implemented within SWMF, together with a new shock-capturing tool for a coronal mass ejection-driven shock originating from the low solar corona. That study uses synthetic observables including extreme ultraviolet and white-light images, proton time–intensity profiles, and energy spectra, and emphasizes the capability of extracting the complex shock surface and examining how that surface affects the particle acceleration process (Liu et al., 2024).

4. Configurations and scientific applications

SWMF has been used across a broad range of heliophysics and planetary problems, including quiet solar wind, CME initiation and propagation, Sun-to-Earth storms, SEP acceleration and transport, geomagnetic indices, mesoscale magnetopause and tail structures, ionospheric outflow, radiation belts, Mercury, Mars, Jupiter, Saturn, comets, the outer heliosphere, and astrospheres (Gombosi et al., 2021). Within this range, AWSoM functions as a flagship SC+IH implementation for ambient solar wind modeling from the low corona to 1 AU. In the 2019 validation study, the SC domain extended from $S_A/B$4 to $S_A/B$5 and the IH domain was a Cartesian cube surrounding the SC sphere, coupled through an overlapping buffer region. Validation used EUV images from STEREO-A/EUVI and SDO/AIA, tomographic reconstructions based on AIA and SOHO/LASCO, InterPlanetary Scintillation reconstructions, and OMNI data near Earth. The study reported good quantitative agreement with observations between the inner corona and 1 AU and stated that the model now reproduces the fast solar wind speed in the polar regions (Sachdeva et al., 2019).

A further example is SOFIE, a data-driven and self-consistent SEP model built upon SWMF. SOFIE couples AWSoM-R for the background corona and solar wind, EEGGL for CME initiation by insertion of a Gibson–Low flux rope, and M-FLAMPA for energetic proton acceleration and transport along time-dependent magnetic field lines. In the reported implementation, AWSoM-R and M-FLAMPA are dynamically coupled every 120 s, 648 magnetic field lines are followed, and nine SHINE challenge or campaign events are modeled. The reported runtime is about real-time speed with 2000 CPU cores during CME propagation through the solar corona and faster than real time after the CME leaves the coronal domain (Zhao et al., 2023).

Planetary applications extend the framework beyond Earth. In support of Mars space-weather studies, a 2022 paper states explicitly that the SWMF that contains the BATS-R-US codes used in the study is publicly available. The specific Mars configuration uses a 3D BATS-R-US Mars multi-species MHD code extended to a Conducting-Core-Surface-to-Interplanetary-Space MS-MHD model, and an extreme Carrington-type CME scenario produced a significant induced surface field of nearly 3000 nT on the dayside (Green et al., 2022).

5. Validation, forecasting, and operational performance

Validation is a recurrent theme in the SWMF literature. AWSoM/AWSoM-R was validated from the low corona to 1 AU using EUV, coronagraph, tomography, interplanetary scintillation, and in-situ solar wind data, while SWMF/Geospace was validated for Dst/SYM-H, Kp, AL, cross-polar-cap potential, and local ground magnetic perturbations. In CCMC-led challenges, SWMF/Geospace consistently ranked among the better global models, and after comparative evaluation it was selected as a top-performing physics-based model for predicting ground magnetic perturbations and adopted for routine operations at SWPC, where it has run continuously since 2016; an upgraded version 2 was deployed in 2020 (Gombosi et al., 2021).

The ionospheric closure problem in SWMF/Geospace has also been examined in detail. The Conductance Model for Extreme Events (CMEE) was incorporated within RIM for usage in space weather simulations and compared against the existing conductance model through six events. The reported result was overall improvement in ionospheric feedback to ground-based space weather forecasts, with substantial improvements in $S_A/B$6 predictive skill. In the Hi-Res SWPC configuration, the Heidke Skill Score at the $S_A/B$7 threshold increased from $S_A/B$8 with the Ridley Legacy Model to $S_A/B$9 with CMEE, and to $T = 5\times 10^4\,\mathrm{K}$0 with CMEE plus oval adjustment; the same study also states that the oval adjustment increased false alarms, so the gain in event detection came with a discrimination tradeoff (Mukhopadhyay et al., 2020).

Forecast-oriented applications are not limited to geospace. The SOFIE study validates its three modules against steady-state background solar wind properties, the white-light image of the CME, and the flux of solar energetic protons at energies of $T = 5\times 10^4\,\mathrm{K}$1, using ACE, SOHO/LASCO C2, and GOES. Its reported strengths lie in reproducing large-scale solar wind context, CME morphology, and SEP timing or connectivity trends, while its main limitations are absolute SEP normalization, omission of perpendicular diffusion, simplified mean free path treatment, and incomplete treatment of prior or multiple eruptions (Zhao et al., 2023).

6. Community access, open science, and long-term challenges

SWMF has long had a dual identity as a research framework and a community model. The full SWMF was made available under a user license, CCMC’s runs-on-request system made it usable even by non-expert researchers without local HPC resources, EEGGL-driven CME simulations were made available through CCMC in 2016, and a substantial core of SWMF was released on GitHub under a noncommercial open-source license in 2020 (Gombosi et al., 2021). A later community survey reports SWMF as fully open source under Apache 2 in 2024, with more than 50 developers over 22 years, about 1M lines of code, computational requirements from 1 to 1M CPU hours and from 1 GB to 1 TB RAM, storage from 1 GB to 10 TB, and output formats including HDF5, FITS, text, binary, Tecplot, and VTK; the same survey lists SWMF as available via GitHub and CCMC (Corti et al., 28 May 2026).

The open-science literature treats SWMF as a model of a broader class of large, coupled, multiphysics, computationally expensive, community-facing frameworks. In that discussion, models are described as tools for numerical experiments, and open science for models is distinguished from open science for observational datasets or ordinary data-analysis software. The recommendations emphasize open use, open validation, open development, and open collaboration, together with preserved run provenance, benchmark problems, configuration files, metadata, user training, and sustained support for simulation service providers such as CCMC (Corti et al., 28 May 2026).

Community validation infrastructures are being organized around that same logic. The Ambient Solar Wind Validation Team embedded in COSPAR ISWAT proposed an open online platform based on CCMC’s CAMEL framework and a seven-component metadata architecture covering observational input data, data preprocessing, model description, model settings, model output, model chain, and model solution. SWMF is explicitly included among the state-of-the-art coronal and heliospheric frameworks for which unified metadata and unified metrics are intended (Reiss et al., 2022).

A persistent issue, however, is sustainability. The SWMF retrospective emphasizes the advantage of the “free spirit” approach of academia for rapid creation and testing of new ideas and methods, while also stressing the difficulty of maintaining a model of this scale under university funding and staffing patterns. This suggests that SWMF is best understood not merely as software, but as a long-duration scientific infrastructure achievement whose continued value depends on documentation, verification and validation, user access, interoperability, and long-term institutional support (Gombosi et al., 2021).