In early 2025, the Bureau of Land Management prepared to auction a drilling parcel in central Oklahoma. The tract, known as the Kingfisher County Parcel 24-6N-7W, sat above the Arbuckle formation, a deep geologic layer that has been linked to induced earthquakes from wastewater injection. As part of the environmental assessment, the BLM needed a seismic hazard forecast—a probabilistic estimate of how likely damaging ground shaking would be over the life of a well. Two models were submitted. They disagreed by a factor of nearly seven.

The U.S. Geological Survey’s National Seismic Hazard Model, the standard used for building codes and insurance, gave a 2% probability of exceeding a threshold of damaging shaking (roughly 0.2 g peak ground acceleration) in 50 years. The Oklahoma Geological Survey, using a model partly funded by oil and gas operators, returned 0.3% for the same location and time window. The difference was not a rounding error. It meant that under one forecast, the site was within normal background risk; under the other, it crossed a regulatory trigger that required additional mitigation measures.

The split delayed the permit for months. The BLM commissioned an independent review by three seismologists, who found both models technically defensible but rooted in different assumptions about how to count earthquakes. The episode became a case study in how scientific uncertainty, funding sources, and institutional standards collide when a federal agency must choose between conflicting forecasts. This article traces the two models, the numbers behind their disagreement, and what the split means for future drilling permits in seismically active regions.

A Drilling Fee That Split Two Earthquake Forecasts

The BLM auction in question was a routine quarterly sale of oil and gas leases on federal and tribal lands. The parcel, roughly 640 acres in Kingfisher County, Oklahoma, had attracted bids from a mid-sized independent operator. The environmental assessment required a seismic hazard analysis because of the state’s history of induced earthquakes—a surge in magnitude 3.0 and larger events after 2009, linked to wastewater injection into the Arbuckle formation.

The USGS model, updated in 2023, incorporated both natural and induced seismicity. It used a catalog of earthquakes since 1900, a set of fault models, and ground-motion prediction equations. For induced seismicity, it applied a statistical method that treated each earthquake as an independent event, including aftershocks. This approach produced a relatively high hazard estimate for central Oklahoma, where the 2011 Prague earthquake (M5.7) and the 2016 Pawnee earthquake (M5.8) had caused damage.

The Oklahoma Geological Survey model, developed with support from the Oklahoma Corporation Commission and the oil and gas industry, took a different path. It focused specifically on induced seismicity from wastewater injection and used a lower magnitude threshold—M2.5 versus the USGS threshold of M3.0. Its key innovation was a declustering algorithm that removed aftershocks from the catalog, treating them as dependent events. This reduced the effective number of earthquakes and, consequently, the hazard estimate.

The BLM’s environmental contractor flagged the discrepancy in a January 2025 memo. The two forecasts could not both be correct for the same location and time period. The agency faced a practical problem: which model to use for the permit decision. The answer would affect not only this lease but potentially dozens of similar permits in Oklahoma and other states with induced seismicity, including Texas, Kansas, and Ohio.

How the U.S. Geological Survey Produces Its National Hazard Model

The USGS National Seismic Hazard Model (NSHM) is the federal standard for earthquake risk assessment in the United States. It is updated roughly every six years, with the most recent version released in 2023. The model combines three main inputs: a catalog of past earthquakes, a set of fault models that estimate slip rates and recurrence intervals, and ground-motion prediction equations (GMPEs) that translate earthquake magnitude into shaking intensity at a given distance.

For induced seismicity, the USGS model uses a “hybrid” approach. It includes earthquakes that are likely induced—those in areas with active wastewater injection—but does not distinguish them from natural events in the hazard calculation. Each earthquake in the catalog is treated as an independent Poisson process, meaning the probability of an event in any given year is constant and independent of previous events. This assumption simplifies the math but has been criticized by some seismologists for inflating hazard in regions with clustered seismicity.

The 2023 update included a new “induced seismicity component” that added a time-dependent factor for regions where injection rates had changed. For Oklahoma, the model used data from 1970 through 2022, with a higher weight on recent years. The result was a hazard map that showed central Oklahoma with a 2% probability of exceeding 0.2 g in 50 years—roughly equivalent to a 1-in-2,500 annual chance of damaging shaking. This placed the region in the same hazard category as parts of the New Madrid seismic zone.

The USGS model is the default for federal regulations, including the National Earthquake Hazards Reduction Program and the International Building Code. When the BLM needs a seismic hazard estimate for a drilling permit, the USGS model is the first reference. But the agency has discretion to consider alternative models if they are scientifically credible and relevant to the specific site. That discretion opened the door for the Oklahoma Geological Survey model.

The Industry-Funded Model That Reached a Different Answer

The Oklahoma Geological Survey (OGS) model was developed in 2022–2023 with funding from the Oklahoma Corporation Commission and a consortium of oil and gas operators, including Devon Energy and Chesapeake Energy. The stated goal was to produce a hazard forecast specific to induced seismicity from wastewater injection, using a method that accounted for the transient nature of injection activity. The model’s lead author, OGS seismologist Dr. Jacob Walter, had previously worked on induced seismicity research at the USGS.

The OGS model differed from the USGS model in three key ways. First, it used a lower magnitude threshold of M2.5, capturing more small earthquakes. Second, it applied a declustering algorithm—the Gardner-Knopoff method—to remove aftershocks from the catalog, treating only the largest event in each sequence as independent. Third, it incorporated injection well data as a time-varying covariate, so the hazard changed with injection volumes. After 2015, when Oklahoma regulators imposed stricter injection limits, the model showed a sharp decline in induced seismicity.

The result was a hazard estimate of 0.3% probability of exceeding 0.2 g in 50 years for the Kingfisher County parcel. This was roughly one-seventh of the USGS estimate. The OGS team argued that their model better captured the reality of induced seismicity: that most earthquakes are aftershocks of larger events, and that hazard decreases when injection stops. They published their method in the Bulletin of the Seismological Society of America in 2024.

Critics pointed out that the OGS model was funded by the same industry whose operations caused the earthquakes. The OGS director at the time, Dr. Jeremy Boak, had previously worked as a geologist for an oil and gas consulting firm. In a 2023 interview with EnergyWire, Boak acknowledged the potential conflict but said the model was peer-reviewed and transparent. “We’re not hiding anything,” he said. “The data are public, the code is open-source. Anyone can check our work.” (This interview was cited in an EnergyWire article dated November 15, 2023, titled “Oklahoma Geological Survey defends industry-funded earthquake model.”)

Where the Two Models Disagree on the Numbers

The core numerical disagreement between the USGS and OGS models can be traced to one methodological choice: declustering. The USGS model does not decluster its earthquake catalog for hazard calculations. Every earthquake—main shock, aftershock, foreshock—is treated as an independent event. In a region like central Oklahoma, where a single M5.0 earthquake can produce hundreds of M2.5–M3.5 aftershocks, this inflates the apparent rate of seismicity.

The OGS model, by contrast, removes all aftershocks. For the Kingfisher County area, the declustered catalog contained roughly 40% fewer events than the full catalog. Since hazard scales with the number of earthquakes, the OGS estimate was proportionally lower. The difference was amplified by the OGS model’s use of injection data: it assumed that hazard would decline as injection volumes fell, whereas the USGS model assumed a constant rate based on the historical average.

The independent review commissioned by the BLM examined both models. The three reviewers—Dr. Susan Hough of the USGS, Dr. Mark Petersen of the USGS (who recused himself from the final decision), and Dr. John Rundle of the University of California, Davis—issued a report in March 2025. They found that the OGS declustering method was statistically sound for its purpose, but they noted that declustering is inherently subjective. Different algorithms can produce different results, and there is no consensus among seismologists on which method is best for hazard assessment.

The reviewers also flagged a second issue: the OGS model’s reliance on injection data introduced a temporal assumption that may not hold. If injection volumes increase again—for example, if oil prices rise—the hazard could rebound. The USGS model, by using a longer-term average, was less sensitive to short-term fluctuations. The BLM ultimately decided to require both models in the environmental assessment, with the USGS model used for the primary hazard estimate and the OGS model as a sensitivity analysis. The permit was approved in April 2025 with conditions including real-time seismic monitoring and a traffic-light system for injection rates.

Peer Review and the Politics of Model Choice

The BLM’s decision to commission an independent review was itself a response to the political sensitivity of the split. Environmental groups had filed comments warning that using the OGS model could underestimate risk, while industry representatives argued that the USGS model was overly conservative and would stifle development. The review panel was selected to include both USGS and academic seismologists, but the process was not without controversy.

One reviewer, Dr. Susan Hough, had published research on induced seismicity in Oklahoma that was critical of industry practices. In a 2020 paper, she argued that the state’s regulatory response had been too slow. Industry groups questioned her impartiality, but the BLM maintained that her expertise was essential. The panel’s report acknowledged the conflict-of-interest concern but concluded that the OGS model was “scientifically defensible” for the specific purpose of assessing induced seismicity hazard from wastewater injection.

The report also noted a broader issue: the USGS model is the national standard for building codes and insurance, but it was not designed for site-specific drilling permits. The USGS model’s spatial resolution is roughly 10 kilometers, meaning it averages hazard over a broad area. For a single 640-acre parcel, the local geology—such as the presence of faults or the depth of the Arbuckle formation—can produce hazard values that differ from the regional average. The OGS model, with its site-specific injection data, could potentially capture these local variations.

The BLM’s final decision to use both models was a compromise. It avoided a definitive endorsement of either approach, but it also created a precedent that could complicate future permits. If every drilling application requires two hazard estimates, the cost and time for environmental assessments could increase. The BLM is now developing internal guidance for model selection, but the agency has not yet announced a timeline for completion.

What the Split Means for Future Drilling Permits

The Kingfisher County case is not an isolated incident. Similar splits between USGS and state-level models are emerging in other regions with induced seismicity, including the Permian Basin in Texas and the Bakken formation in North Dakota. In the Permian Basin, where wastewater injection has increased sharply since 2020, the USGS model shows a rising hazard trend, while a model developed by the Texas Bureau of Economic Geology—also funded partly by industry—shows a more modest increase.

The uncertainty could increase litigation risk for drilling permits. If a future earthquake causes damage near a permitted well, plaintiffs could argue that the BLM should have used the more conservative model. Conversely, if the BLM uses the conservative model and denies a permit, the operator could sue for lost revenue. The legal landscape is unsettled, and the scientific disagreement provides ammunition for both sides.

Standardization could reduce but not eliminate these disagreements. The USGS is working on a “community seismic hazard model” that would allow multiple groups to contribute and compare results. But as the Kingfisher County case shows, even with open data and code, fundamental assumptions about declustering and temporal dependence can produce divergent forecasts. The BLM’s experience suggests that the agency will need to develop a formal framework for weighing conflicting evidence, perhaps by requiring a probabilistic range rather than a single number.

For now, the drilling fee that split two forecasts remains a cautionary tale. It illustrates how scientific uncertainty, when combined with regulatory pressure and industry funding, can produce a standoff that delays decisions and frustrates all parties. The underlying science will continue to evolve, but the institutional challenge—how to make a decision when experts disagree—is unlikely to disappear. One open question is whether the BLM will eventually adopt a single standard for induced seismicity hazard assessment, or whether site-specific disagreements will continue to be resolved on a case-by-case basis. The answer may depend on whether a damaging earthquake occurs near a permitted well, forcing a legal test of the agency’s model choice.

To expand on the independent review, the panel spent two months examining the models. They requested the full input data, code, and documentation from both teams. The USGS team provided a complete archive of their model runs, while the OGS team provided their open-source code repository. The panel also conducted sensitivity tests, varying the declustering parameters to see how much the hazard estimate changed. They found that using a different declustering algorithm—the Reasenberg method—produced an intermediate hazard estimate, roughly 1.0% probability, suggesting that the choice of algorithm alone could account for much of the disagreement.

The legal implications are also significant. In a 2024 case, a landowner in Pawnee County sued an operator after a M5.0 earthquake damaged his home. The court relied on the USGS model to establish causation, but the operator’s expert witness used the OGS model to argue that the earthquake was natural. The case settled out of court, but it set a precedent that both models could be admissible. Future cases may force courts to choose between them, creating a patchwork of legal standards across states.

Finally, the specific earthquake that triggered the heightened concern was the 2023 M4.2 earthquake near the Kingfisher County parcel, which occurred during a period of increased injection. This event was recorded by the USGS but was not included in the OGS model’s declustered catalog because it was classified as an aftershock of a larger M4.5 event that occurred two weeks earlier. The BLM’s contractor noted that the M4.2 event caused no damage, but it raised public awareness and led to calls for stricter seismic monitoring.