GIC--Related Observations During the May 2024 Geomagnetic Storm in the United States

This paper analyzes a unique dataset of measurements and model outputs from the severe May 2024 geomagnetic storm to evaluate the accuracy of GIC predictions by power operators and global models, while establishing empirical relationships between geomagnetic parameters and GIC magnitudes across the contiguous United States.

The Gist

Imagine the Earth is a giant, invisible electrical circuit. Usually, this circuit is quiet. But every now and then, the Sun throws a massive tantrum, blasting a cloud of charged particles toward us. When this "solar wind" hits Earth's magnetic shield, it creates a storm—like a giant, invisible wave crashing against a dam.

This paper is a report card on how well we can predict what happens when that wave hits the most important part of our modern life: the power grid.

Here is the story of the May 2024 "Mother's Day" storm, told in simple terms.

1. The Big Event: A Solar Tsunami

In May 2024, the Sun unleashed one of the most violent storms in 20 years. It was so strong that the Northern Lights (aurora) were visible as far south as Mexico and even parts of Africa.

When this storm hit, it didn't just make pretty lights; it induced Geomagnetically Induced Currents (GICs). Think of GICs as "ghost currents." The storm shakes Earth's magnetic field, which acts like a giant generator, pushing unwanted electricity into our power lines, pipelines, and cables. If these ghost currents get too strong, they can overheat and destroy massive power transformers, potentially causing blackouts that last for weeks.

2. The Mission: Testing the Crystal Balls

The scientists in this paper wanted to answer a simple question: "How good are our weather forecasts for these electrical ghosts?"

They gathered a massive amount of data from the May 2024 storm:

• Real Measurements: They looked at actual "ghost currents" measured at 47 different power stations across the US.
• The "Pro" Model: They checked the calculations made by the Tennessee Valley Authority (TVA), the actual power company operators who know their specific power lines inside and out.
• The "Student" Model: They used a generic "Reference Model" that tries to guess the currents without knowing the specific details of the power lines (like a student taking a test without the textbook).
• The Space Models: They compared three different super-computer simulations (MAGE, SWMF, OpenGGCM) that try to predict how the magnetic field changes in space before it hits Earth.

3. The Results: Who Got the Grade?

The Power Company (TVA) vs. Reality

Grade: A-
The TVA's model was very good. It predicted the actual ghost currents with high accuracy (about 80% correlation).

• The Analogy: Imagine the TVA is a local chef who knows exactly how their kitchen stove works. When they predicted how much heat the stove would get, they were spot on.
• The Catch: Even the best model had some errors. Sometimes it guessed the current was too high; other times, too low. This is because every power grid is unique, like a fingerprint.

The Generic Model (Reference) vs. Reality

Grade: C
The generic model was okay at seeing the pattern of the storm, but it was terrible at guessing the exact amount of current.

• The Analogy: This is like trying to guess how much water will flow through a specific pipe in your house just by looking at the main water tower. You know water is coming, but you don't know if your specific pipe is clogged or wide open.
• Why? Because this model didn't know the specific layout or resistance of the wires. It's a "one-size-fits-all" approach, and the Earth's power grid doesn't fit into one size.

The Space Models (MAGE, SWMF, OpenGGCM) vs. Reality

Grade: D+
The models that tried to predict the magnetic field in space (which drives the currents on the ground) struggled significantly. They often predicted the storm was either way too weak or way too strong.

• The Analogy: Imagine trying to predict how hard it will rain in your backyard by looking at a weather satellite image of the whole continent. You can see the storm cloud, but you can't see the tiny details of the wind and terrain that determine exactly how much rain hits your specific roof.
• The Problem: These models are great at seeing the big picture, but they miss the local details of Earth's crust (the ground conductivity) which acts like a sponge, soaking up or amplifying the electricity differently in different places.

4. The Big Discovery: A Simple Rule of Thumb

Despite the models struggling with exact numbers, the scientists found a clever shortcut. They discovered that the maximum amount of ghost current at any location could be estimated using a simple formula involving two factors:

1. Where you are (Latitude): The closer you are to the magnetic poles (where the aurora lives), the stronger the storm.
2. What's under your feet (Conductivity): Is the ground rocky and dry (like a sponge that doesn't soak up water) or wet and sandy (like a sponge that soaks up everything)?

The Analogy:
Think of the storm as a giant sprinkler spraying water (electricity).

• Latitude determines how close you are to the sprinkler head.
• Conductivity determines if you are standing on a trampoline (rocky ground, which bounces the current up) or in a mud pit (wet ground, which absorbs it).

The paper found that if you multiply your "distance from the sprinkler" by your "mud factor," you can get a surprisingly good estimate of how wet you'll get, even without knowing the exact layout of the sprinkler pipes.

5. Why This Matters

This research is like a safety manual for the future.

• For Power Companies: It tells them that while their specific models are good, they need to be careful about relying on generic space weather forecasts. They need to know their own "ground conductivity" (what's under their feet).
• For the Future: As we face bigger solar storms, we need better ways to predict these "ghost currents." This paper gives us a new, simpler way to estimate the danger using just location and ground type, which is a huge step toward keeping our lights on when the Sun gets angry.

In short: We can't perfectly predict every spark of electricity from a solar storm yet, but we've found a reliable rule of thumb to know where the danger zones are, helping us protect our power grids from the next big solar tantrum.

Technical Summary

1. Problem Statement

The May 2024 geomagnetic storm (the "Gannon" or "Mother's Day" storm) was the most severe event in over two decades, reaching a G5 rating and causing widespread auroral visibility at mid-latitudes. While such storms pose significant risks to critical infrastructure—particularly electrical power grids via Geomagnetically Induced Currents (GICs)—there is a lack of comprehensive, ground-truth data validating models during extreme events.

• The Challenge: Accurately predicting GICs requires precise knowledge of the surface magnetic field ($\Delta B$), local ground conductivity, and specific power grid network configurations.
• The Gap: Global Magnetohydrodynamics (MHD) models often struggle to accurately predict ground-level magnetic field perturbations. Furthermore, operational power grid models are often proprietary, making it difficult to validate GIC predictions against measurements across a broad geographic area. This study addresses the need to validate existing models and identify empirical relationships for GIC estimation using a unique, large-scale dataset from this specific storm.

2. Methodology

The authors assembled a unique dataset comprising measurements and model outputs from the contiguous United States (CONUS) during the May 10–12, 2024 storm.

• Data Sources:
  • Measurements: GIC data from 47 sites (15 from Tennessee Valley Authority [TVA] and 32 from the North American Electric Reliability Corporation [NERC]) and magnetic field ($\Delta B$) data from 17 magnetometer sites.
  • Model Predictions:
    • GIC Models: TVA's internal operational model (using real-time grid topology) and a "Reference Model" (using estimated grid parameters and standard conductivity models).
    • $\Delta B$ Models: Three global geospace MHD models: MAGE (Multiscale Atmosphere-Geospace Environment Model), SWMF (Space Weather Modeling Framework), and OpenGGCM (Open Geospace General Circulation Model).
• Analysis Techniques:
  • Model Validation: Comparison of modeled vs. measured GIC and $\Delta B$ using Pearson correlation ($r$), coefficient of determination ($r^2$), and Prediction Efficiency ($pe$).
  • Empirical Regression: Analysis of relationships between maximum GIC magnitude ($|GIC|_{max}$) and scaling factors defined in the NERC TPL-007 standard:
    • $\alpha$: Geomagnetic latitude scaling factor.
    • $\beta$: Ground conductivity scaling factor (derived from 3-D magnetotelluric data).
  • Inter-site Correlation: Analysis of how GIC timeseries correlations between site pairs relate to separation distance, latitude differences, conductivity differences ($\Delta \beta$), and transmission line voltage differences.

3. Key Contributions

• Unprecedented Dataset: The paper synthesizes the largest and most diverse set of GMD-related measurements for a single storm event in the US, combining proprietary utility data (TVA) with public regulatory data (NERC) and global model outputs.
• Model Performance Benchmarking: It provides a rigorous, side-by-side comparison of operational utility models, reference models, and global space weather models against ground truth during an extreme G5 storm.
• Empirical Scaling Validation: The study validates the use of NERC's $\alpha$ and $\beta$ scaling factors for estimating maximum GIC magnitudes, offering a simplified method for hazard assessment when detailed network data is unavailable.

4. Key Results

A. GIC Model Performance

• TVA Model: The TVA operational model demonstrated high skill, achieving a correlation coefficient $r > 0.8$ and prediction efficiency ($pe$) between 0.4 and 0.7 at four validation sites. However, it showed systematic biases (over-prediction at some sites, under-prediction at others).
• Reference Model: The Reference model performed significantly worse, with $r^2$ ranging from 0.01 to 0.48 and $pe$ values often negative. This highlights the critical importance of accurate, real-time power grid topology and impedance data in GIC modeling.
• Comparison: The difference between TVA and Reference model outputs serves as a proxy for uncertainty when only generic models are available.

B. Global MHD Model Performance ($\Delta B$)

• The three global models (MAGE, SWMF, OpenGGCM) showed weak to moderate correlations with measured $\Delta B$ ($r$ range: 0.21 to 0.65).
• Prediction Efficiency: All models exhibited negative prediction efficiency (mean $pe \approx -0.68$ for SWMF, $-3.5$ for MAGE), indicating they fail to capture the variance of the observed magnetic field perturbations.
• Systematic Bias: MAGE consistently over-predicted $\Delta B$, while SWMF and OpenGGCM consistently under-predicted it.
• Implication: Using model-predicted $\Delta B$ to drive GIC calculations would result in significantly lower accuracy than using measured $\Delta B$.

C. Empirical Relationships

• Inter-site Correlation: GIC timeseries correlations between sites were weakly dependent on distance or conductivity differences. Sites separated by <10 km could have correlations ranging from 0.0 to 1.0, confirming that local power grid configuration is a dominant factor in GIC variability, often outweighing local geoelectric field similarities.
• Maximum GIC Estimation: A strong linear relationship was found between the maximum GIC magnitude and the product of the scaling factors $\alpha \times \beta$.
  • The regression model $|GIC|_{max} \approx a(\alpha\beta) + b$ yielded a correlation of $r \approx 0.76$ and an RMSE of 11.6 A.
  • Models using the product $\alpha\beta$ (Model 3) or combinations of $\alpha$, $\beta$, and $\alpha\beta$ (Models 4–7) had the lowest Akaike Information Criterion (AIC), indicating they are the most robust predictors.
  • This validates the NERC TPL-007 assumption that the peak geoelectric field can be approximated by the product of latitude and conductivity scaling factors.

5. Significance

• Infrastructure Resilience: The study confirms that while global space weather models are improving, they are not yet sufficient for precise, site-specific GIC hazard assessment without ground truth data. Operational grid models remain superior but require accurate network topology.
• Hazard Assessment: The validation of the $\alpha\beta$ linear relationship provides a practical, data-driven tool for power system operators to estimate worst-case GIC scenarios during extreme storms without needing real-time magnetometer data or detailed network schematics.
• Future Modeling: The systematic over/under-prediction biases in MHD models suggest that scaling factors could be applied to improve operational forecasts. The study also underscores the necessity of high-resolution 3-D ground conductivity models, as 1-D models are insufficient for capturing local GIC variations.

In conclusion, this paper bridges the gap between space weather modeling and power grid engineering, providing critical validation data for the May 2024 storm and establishing empirical rules for estimating GIC risks in the contiguous United States.