Rare geomagnetic storms expose regional weaknesses in the U.S. power grid

The study, published in AGU Advances on September 4, 2026, addresses a persistent gap in space-weather risk assessment. Scientists can model geomagnetic storms, power engineers can estimate how geomagnetically induced currents affect transformers, and economists can estimate losses from blackouts, but those parts of the problem have often been analyzed separately.

Oughton et al. (2026) describe their work as the first nationwide assessment to connect that full chain while carrying quantified uncertainty through the calculation.

The modeled impacts increase across the four return-period scenarios. A 100-year storm affects about 3.45 million people and 91 000 businesses, with direct losses of about USD 0.66 billion per day and total losses of USD 1.22 billion per day after inter-industry effects are included. Total modeled losses rise to USD 1.46 billion per day in the 150-year case, USD 1.65 billion in the 200-year case and USD 1.81 billion in the 250-year case. Direct losses in the most severe scenario reach about USD 0.98 billion per day.

A 250-year return period does not mean such a storm is expected every 250 years, nor does the study forecast that one is approaching. It is a statistical severity benchmark derived from extreme-value analysis. The return-period maps are also not reconstructions of a single instant during a real storm: the researchers combined estimated peak values at individual locations to build scenarios for probabilistic risk assessment.

The physical pathway begins with rapid changes in Earth’s magnetic field during a geomagnetic storm.

Those variations generate electric fields at Earth’s surface. Long grounded transmission networks then provide paths through which quasi-direct geomagnetically induced currents (GICs) can flow. When GICs enter high-voltage transformers, they can push magnetic cores outside their normal operating range, increasing heating and creating abnormal electrical conditions that can interfere with grid operation. NOAA identifies the same physical process as a recognized space-weather threat to electric transmission systems.

Ground conditions matter because the same magnetic disturbance does not generate the same electric field everywhere. Electrical conductivity varies with geology, and transmission-line length, orientation, and network configuration determine how the resulting field translates into voltage across the grid. In the study, modeled peak geoelectric fields reach about 25 V/km for the 100-year scenario, 28 V/km for the 150-year scenario, and 32 V/km for the 250-year scenario. Induced voltages on individual transmission lines reach approximately 1.1 kV in the strongest modeled cases.

The highest modeled exposure concentrates across northern parts of the United States, particularly around the Great Lakes and into the Northeast. Transformer-vulnerability hotspots appear in Wisconsin, Minnesota, the Upper Midwest, and northeastern coastal areas, where ground conductivity and network topology combine with the modeled geoelectric field. The study’s “failure” metric requires careful interpretation: it is a probabilistic proxy for transformer malfunction or protection-system misoperation at substation level, not a prediction that every affected transformer would be physically destroyed.

That distinction matters because the model is designed to estimate risk distributions rather than predict the fate of individual substations. The researchers reconstructed a contiguous U.S. transmission network covering equipment rated at 161 kV or higher, comprising 10 464 substations and 16 256 transmission lines.

Because detailed transformer configuration, grounding, and other asset information is incomplete or unavailable nationally, Monte Carlo simulations sample plausible transformer types, unit counts, grounding resistance, equipment age, and related parameters. In physical terms, the model repeatedly recalculates the system under different plausible grid configurations instead of treating uncertain infrastructure details as known exactly.

The socioeconomic component then assigns population and economic activity to modeled substation service areas and estimates the consequences of interrupted electricity supply. Direct losses represent economic activity affected inside the outage areas, while an input-output model estimates additional losses transmitted through relationships between industries. Manufacturing, finance and real estate, and education and entertainment carry the largest modeled sectoral losses in the most exposed regions.

That economic layer explains why the USD 1.81 billion figure is larger than the estimated direct loss of USD 0.98 billion in the 250-year scenario. A factory, office, or service business can lose activity because its own electricity supply is interrupted, but businesses elsewhere can also be affected when suppliers or customers inside the outage area cannot operate normally. The study finds that this indirect component scales broadly with the direct disruption as storm severity increases.

The researchers tested the engineering portion of the framework against utility measurements from the Tennessee Valley Authority during the May 2024 Gannon geomagnetic storm. Correlations between simulated and measured GICs varied widely across monitoring sites, from (|r| = 0.02) to (|r| = 0.70), while prediction efficiencies ranged from (−15.86) to (0.48). Agreement was strongest below 0.004 Hz, where slower magnetic variations are especially relevant to sustained GIC flow and transformer heating, while the model underestimated shorter-period peaks.

Physically, that validation result means the framework captured some of the slow variations most important for transformer thermal exposure better than it reproduced every short-lived fluctuation at every site. Performance was uneven, so the national results should be read as probabilistic risk estimates rather than precise forecasts for particular substations. The authors identify greater utility-sector participation and access to detailed infrastructure information as important steps for more refined future assessments.

The operational relevance is already established independently of the new economic estimates. North American Electric Reliability Corporation standard EOP-010-1 requires applicable grid operators to maintain operating plans, processes, and procedures intended to mitigate geomagnetic-disturbance effects. The U.S. Department of Energy also supports transformer thermal-response research and GMD monitoring, and in February 2026 released software intended to make mandatory utility-run geomagnetic-disturbance assessments more efficient.

The USD 1.81 billion estimate should not be interpreted as a complete estimate of everything an extreme storm could cost. The model excludes voltage collapse, cascading grid failures, restoration costs, frequency disturbances, and disruption propagating through other dependent infrastructure. The authors therefore describe the calculated losses as conservative lower bounds. At the same time, the quoted 95% confidence intervals represent uncertainty in selected model parameters and do not capture every uncertainty in the extreme-storm scenarios, future grid configuration, or future population and economic exposure.

The remaining scientific challenge is therefore not simply to run stronger storms through the model. Better estimates require better knowledge of the grid itself: transformer configuration, grounding, mitigation equipment, operational re-sequencing and measured GIC behavior during major storms. More utility measurements from future events would provide the strongest test of whether the framework can narrow those uncertainties enough to support increasingly detailed national and regional space-weather risk assessments.

References:

¹ Oughton, E. J., Bor, D. K., Weigel, R. S., Gaunt, C. T., Dogan, R., Huang, L., Love, J. J., & Wiltberger, M. (2026). Major space weather risks identified via coupled physics-engineering-economic modeling. AGU Advances, 7(6), e2026AV002367. https://doi.org/10.1029/2026AV002367