THE FINDING: Cutting injection volumes measurably reduces how often earthquakes happen within 7–19 km of disposal wells. field manual · methodology

Causal Analysis of
Induced Seismicity
in the Permian Basin

A decision-support tool for regulators and operators. Click an earthquake, see which wells contributed, how much, and what volume threshold keeps risk below the limit. Built on doubly-robust TMLE and Causal Forest methods.

Enter Dashboard Try the Demo (no login) Field Manual
SPE-228051   L. Matthews  |  2025
TMLE Extension   L. Matthews  |  2026

Primer

Correlation Is Not Causation

Two things can happen at the same time without one causing the other. Ice cream sales and drowning deaths both go up in summer — but banning ice cream wouldn't save anyone. The hidden third factor (hot weather) drives both. That hidden factor is called a confounder.

Spurious (Fake)

Ice cream → Drowning

Looks causal. But controlling for temperature makes the link vanish. The arrow was never real.

Real Causation

Smoking → Cancer

After controlling for genetics, pollution, income — the link holds. The arrow is real.

The Test

Control for confounders

If the link survives after adjusting for every confounder, it's causal. If it disappears, it was spurious.

The Permian Basin Confounding Problem

Saltwater disposal wells and earthquakes cluster in the same areas. But the same deep sedimentary basins that make good disposal targets also sit on top of critically stressed faults. Geology is the confounder.

Without Causal Inference
G Geology (faults, depth, rock)

WInjection

SEarthquakes
Is the W–S link real, or is it all geology?
Without controlling for G, we can't tell. Just like ice cream and drowning.

What SPE-228051 Did: Drew the Causal DAG

The paper formalized the physical mechanism as a causal diagram. It identified five geologic and operational confounders, controlled for them, and tested whether the injection–seismicity link survived. It did.

SPE-228051: The Causal Model
G1Fault Dist
G2Fault Count
G3Depth
G4Days Active
G5Neighbor Vol
↓ ↓ ↓ ↓ ↓ (controlled for)
WVolume
PPressure
SSeismicity
After controlling for G1–G5, the WPS link survives. Injection causes seismicity through pressure buildup at distances consistent with pore-pressure diffusion (7–19 km).

The advance: SPE-228051 moved the field from "wells and earthquakes are in the same place" to "changing injection volume causes changes in earthquake magnitude." The causal DAG was right. But the estimator (OLS) had limitations.


Methodology

From OLS to TMLE: A Better Estimator

The original paper used ordinary least squares (OLS), which forces a straight line through the data. The real relationship between injection and seismicity is nonlinear. We replaced OLS with TMLE — a doubly-robust, nonparametric method that lets the data speak for itself.

OLS (Paper)TMLE v3 + Forest (This Work)
EstimatorLinear OLSDoubly robust TMLE (NNLS SuperLearner, 5-fold CV)
Per-well effectsNot availableCausal Forest CATE + honest CI at all 20 radii
Spatial interferenceNot addressedNeighbor well volumes controlled (VIF 1.12)
Dose-responseStraight lineNonparametric curve with extrapolation warnings
Pressure band (7–19 km), regHAL-TMLENot testedψ=+7.7×10-3, p=7.2×10-4 (April 2026 vintage; see scoreboard)
Pressure band (7–19 km), XGBoost-GPUNot testedψ=+6.1×10-4, p=1.3×10-4 (plug-in)
Near-field (1–6 km)Not testedInconclusive (basis-sensitive)
Confidence intervalsAssumes equal varianceBessel-corrected cluster IF SEs
ValidationDoWhy refutersR tlverse (<0.1%) + 6 sensitivity analyses

Three Layers of Causal Analysis

The dashboard stacks three levels, each more specific than the last:

Population

TMLE Shift + Dose-Response

"If all wells reduce volume 10%, how does earthquake risk change?" For area-wide policy.

Per-Well

Causal Forest CATE

"How much did this specific well contribute to this specific earthquake, controlling for what neighboring wells were doing?" For targeted shut-ins.

Threshold

Per-Well Dose-Response

"At what volume does this well's contribution cross the regulatory limit?" For volume caps.

Two Types of Events, Two Playbooks

Concentrated Causation

A few wells dominate

Top 3–5 wells have CATEs with CIs excluding zero. Use the CATE waterfall + threshold curve. Action: targeted shut-in or volume cap.

Distributed Causation

No well stands out

Every well's CI crosses zero. Collective effect is real but individual signal is below noise floor. Action: area-wide volume reduction via TMLE shift.


The Dashboard

How Regulators and Operators Use This

Decision Workflow

Trigger
M ≥ 3.5
Step 1
Click Event
Step 2
See CATEs
Step 3
Click Well
Action
Set Volume Cap
Regulator

After an event

Click the earthquake. Rank wells by CATE. Target wells where the CI excludes zero. Use the threshold curve to set a well-specific volume cap. Other operators continue normally.

Operator

Compliance & planning

Check if your well's CI crosses zero (no significant contribution). Use the threshold curve to find your safe operating volume. Prioritize reductions at highest-CATE wells across your fleet.

Earthquakes (TexNet, live)
Well-days (RRC daily H-10)
Latest catalogued event in model
Causal Forest radii (1–20 km)
~100%
Of the causal effect is event frequency, not magnitude

Try the Dashboard

Click any earthquake on the map. See which wells contributed and how much. Click a well card for its threshold curve.

SEIS Dashboard — live demo on the M4.8 curated event (full version requires access) Open Full Screen →

Ready to explore?

Open the full dashboard, read the FAQ, or dive into the methodology.