2  Executive Summary

2.1 The one-sentence version

Cutting saltwater-disposal injection volumes measurably reduces earthquake frequency — how often quakes happen — within 7–19 km of disposal wells in the Permian Basin, and the evidence for this meets a causal standard that the current regulatory process does not require but should.

2.2 The five-minute version

What we did. We linked every TexNet-catalogued earthquake in the Permian Basin (7,581 events, 2017–2026) to every saltwater-disposal well within 20 km (795 wells, 918,720 well-days of daily Railroad Commission injection records) and asked a counterfactual question: what would seismicity have been if each well had injected 10% less? Answering that question credibly requires controlling for the things that make high-injection wells different from low-injection wells — fault proximity, formation depth, well age, neighborhood injection — and we use targeted learning, the same family of methods used in FDA-grade epidemiology, to do it.

What we found.

  1. The effect is real and positive at pressure-diffusion distances. Pooled across the 7–19 km band — the distances where pore-pressure diffusion operates on a one-year timescale — a 10% volume increase causes a statistically significant increase in expected seismicity (basin-scale combined test: z = 4.25, p ≈ 2 × 10⁻⁵ under the current data vintage; significant in every vintage tested once a documented cross-validation artifact was repaired).

  2. The mechanism is frequency, not magnitude. Essentially all of the causal effect operates through how often events occur, not how large a given event is. A volume cap is therefore an event-rate-reduction tool. This is the single most policy-relevant finding: it means volume management reduces the number of felt-event triggers that today’s reactive curtailment responds to.

  3. Per-well attribution is operational. For any catalogued event, the pipeline ranks every well within radius by its causal contribution, with honest confidence intervals — and for any well, it produces a volume-threshold curve: the injection level above which that specific well’s marginal contribution to expected seismicity exceeds a chosen risk tolerance.

  4. Near-field estimates (1–6 km) are inconclusive and we say so. At very short distances the positive-event counts are too small for the basis-function machinery to behave stably; we document the sensitivity rather than reporting a fragile number.

What it means for policy. The Railroad Commission’s current Seismic Response Area protocol is reactive: events happen, then injection is curtailed nearby. The evidence in this manual supports a complementary proactive instrument: permit conditions tied to causal dose-response thresholds, applied before events occur, targeted at the wells whose attribution and threshold curves identify them as marginal-risk carriers. Model permit-condition language is provided in the appendix.

What it means for operators. The same pipeline that supports regulation also supports defense: an operator whose wells sit below their threshold curves has auditable, causal-standard evidence of proactive risk management — for regulators, insurers, and litigation.

2.3 What this evidence is not

It is not a claim that any specific earthquake was caused by any specific well; per-event attributions are model-based estimates with stated uncertainty. It is not a near-field (under 7 km) claim. And it does not yet model the operator-feedback loop (operators reducing volumes in response to nearby events), which is the program’s next methodological step — documented openly in the Methods Defense.