3 Regulator One-Pager
The case for causal permit conditions, on one page
Written to be lifted directly into a briefing packet. Every claim links to the underlying evidence.
3.1 The problem with the current standard
Texas regulates injection-induced seismicity reactively. Under the Railroad Commission’s Seismic Response Area (SRA) protocol, a felt earthquake triggers curtailment of disposal wells within a radius. This works — but it pays for risk reduction after the seismic event, with a blunt spatial rule that treats every well within the circle identically, regardless of whether that well’s injection actually contributes to seismic hazard.
3.2 What the evidence now supports
A causal analysis of nine years of TexNet earthquake data and daily RRC injection records (7,581 events; 795 disposal wells; 918,720 well-days) establishes:
| Finding | Evidence grade |
|---|---|
| Injection volume causes increased seismicity at 7–19 km | Basin-scale combined test, z = 4.25, p ≈ 2 × 10⁻⁵; direction stable across every data vintage and estimator tested |
| The mechanism is event frequency, not magnitude | ~100% of the pooled causal effect operates through P(event), not E[magnitude] — stable across vintages |
| Per-well causal attribution is feasible today | Honest-forest per-well contributions with confidence intervals, live on the dashboard for every catalogued event |
| Per-well volume thresholds are computable | Dose-response curves per well identify the injection level at which marginal risk exceeds tolerance |
In plain terms: a 10% volume reduction at pressure-diffusion distances reduces how often earthquakes happen, and we can now say which wells carry the marginal risk, instead of curtailing every well in a circle.
3.3 The instrument this enables
A causal permit condition: disposal permits in seismically active areas carry a volume schedule derived from the well’s position on its dose-response curve, reviewed annually against refreshed data. Compared to SRA curtailment it is:
- Proactive — applied before felt events, not after.
- Targeted — calibrated per well, not per circle. Wells with no attributable risk are not penalized.
- Auditable — the dose-response machinery, its assumptions, and its failure modes are public and reproducible (github.com/Project-Geminae/induced-seismicity).
- Defensible for operators — compliance generates causal-standard evidence of proactive management.
Draft regulatory text is provided in Appendix A — Model Permit Language.
3.4 What we are not claiming
- No specific earthquake is attributed to a specific well with certainty; attributions carry stated uncertainty.
- Near-field effects (under 7 km) are statistically inconclusive in this data and are not a basis for the proposed instrument.
- The headline effect size varies across data vintages (the direction and mechanism do not); the instrument is therefore designed around rankings and thresholds, which are stable, rather than point estimates, which move.
3.5 Contact
Lewis Matthews · Project Geminae · lewis@projectgeminae.net Live dashboard access on request.