Quantitative Image-Based Validation Framework for Assessing Global Coronal Magnetic Field Models

Problem Statement
The solar corona's magnetic field is the dominant force governing space weather propagation, yet it remains notoriously difficult to measure directly in the atmospheric layers above the photosphere. Consequently, researchers rely on extrapolations from indirect photospheric measurements and magnetic field models (such as Potential-Field Source Surface and Magnetohydrodynamic models) to understand coronal structure. A critical challenge in space physics is the lack of "ground-truth" precedent to validate the orientation of segmented features in coronal images against the true magnetic field. While algorithms like the Quasi-Radial Feature Tracing Algorithm (QRaFT) can segment quasi-radial features in polarized brightness (pB) images to approximate magnetic field orientation, it is difficult to determine if these segmented features accurately represent the underlying magnetic structure without a rigorous validation framework.

Methodology
The authors present a quantitative framework to validate image-based coronal segmentation methods against magnetic field models, and vice-versa. The framework consists of four primary components:

1. Image Processing: Sequential polarized images from the STEREO COR-1 Coronagraph Instrument are calibrated, background-subtracted, and stacked to generate representative pB images.
2. Synthetic Image Generation: The Magnetohydrodynamic Algorithm outside a Sphere (MAS) model, specifically a high-resolution simulation of the August 21, 2017 solar eclipse, is used as a "ground truth." The FORWARD toolset converts the 3D MAS solution into synthetic pB images co-aligned with real observations from the perspective of the STEREO A satellite.
3. Image Segmentation: The QRaFT algorithm is applied to both the real COR-1 pB observations and the synthetic FORWARD pB images. QRaFT detects quasi-radial features, approximates their geometry using polynomial fits, and calculates their orientation angles in the plane-of-sky (POS).
4. Comparative Analysis: The orientation of the segmented features is compared against the projected MAS magnetic field vector ($\vec{B}$) in the POS. The angular difference ($\Delta\theta$) is calculated for each pixel.

To isolate sources of error, the study compares four distinct data populations against the MAS POS $\vec{B}$:

• MAS ne POS: Central plane-of-sky electron density (1-to-1 model comparison).
• MAS ne LOS: Line-of-sight integrated electron density (adds geometric integration complexity).
• FORWARD pB: Synthetic polarized brightness (adds Thomson scattering physics).
• COR-1 pB: Real observational data (adds instrumental noise and artifacts).

Statistical analysis includes calculating median and mean absolute angular differences ($|\Delta\theta|$), standard deviations, and 95% confidence intervals. Additionally, the authors employ Jensen-Shannon Divergence (JSD) and Kullback-Leibler Divergence (KLD) to compare the shapes of probability distributions, and Tukey's Honestly Significant Difference (HSD) test to evaluate the statistical significance of mean differences between populations.

Key Results

• Accuracy of Segmentation: The QRaFT algorithm successfully identifies the global large-scale orientation of the coronal magnetic field within approximately $\pm 10^\circ$ to $15^\circ$ of the POS-projected MAS magnetic field.
• Baseline Error: The MAS ne POS population (a direct model-to-model comparison) yielded a mean absolute angular difference of $10.479^\circ \pm 0.118^\circ$. The authors interpret this as a baseline estimate of the systematic error inherent in the QRaFT methodology and assumptions.
• Error Propagation:
  • The difference between MAS ne POS and MAS ne LOS was statistically insignificant ($0.025^\circ$), suggesting that off-POS electron density contributions do not significantly distort the orientation analysis.
  • The difference between MAS ne LOS and FORWARD pB was statistically significant ($1.616^\circ$), indicating that the inclusion of Thomson scattering physics introduces measurable error.
  • The COR-1 pB observations showed the highest angular differences (mean $14.554^\circ$), reflecting the cumulative effect of model limitations, scattering physics, and observational artifacts (e.g., background subtraction issues and solar rotation smearing).
• Distribution Similarity: JSD metrics confirmed that the probability distributions of angular errors for the model-based populations (MAS ne POS, MAS ne LOS, FORWARD pB) are highly similar to each other and distinct from a randomized control dataset, validating the consistency of the framework.

Significance and Claims
The paper claims to provide a novel, quantitative framework for cross-validating coronal image segmentation algorithms against MHD models. The primary significance lies in the ability to:

1. Quantify Uncertainty: Move beyond qualitative visual comparisons to provide rigorous statistical metrics (e.g., $|\Delta\theta|$, JSD, HSD) that describe the accuracy and uncertainty of both segmentation algorithms and MHD models.
2. Isolate Error Sources: By comparing different data populations, the framework can distinguish between errors arising from geometric projection, scattering physics, and observational noise.
3. Improve Space Weather Modeling: The authors argue that accurate segmentation of large-scale plasma density features is critical for constraining the boundary conditions of space weather models (such as WSA). Since small uncertainties in the orientation of quasi-radial magnetic field lines can propagate into large uncertainties in solar wind speed predictions, this framework offers a pathway to improve the fidelity of space weather forecasting by validating and refining both the observational segmentation techniques and the underlying physical models.

The study concludes that the developed methodology is universally applicable to various feature tracing algorithms and MHD models, offering a ready-to-use toolset for the space weather community to assess model-data consistency.