From Correlation to Causation: A Petrophysics-Informed Framework for Optimizing Hydraulic Fracturing in Tight Gas Reservoirs

Achieving sustainable development in unconventional resources requires a fundamental shift from correlation-based predictions to causal decision-making. Conventional data-driven approaches often mistake spurious correlations for effectiveness, neglecting the interplay between engineering treatments and reservoir physics. This study introduces a petrophysics-informed causal framework to disentangle these effects. Analyzing 646 wells, we first replicated conventional practice using a high-fidelity XGBoost model (<i>R</i> ² = 0.80) interpreted by SHAP, confirming strong positive associations between production and large-scale treatments. However, by subsequently employing a Causal Forest to control for confounders, we revealed a stark contradiction: these same treatments often exert a statistically significant negative causal effect on output, indicating systemic material waste. We resolve this paradox by demonstrating that treatment efficacy is strictly modulated by rock properties; for instance, high injection rates (<i>q</i> _prop) prove causally effective only within a specific resistivity range (<i>R</i> _<i>t</i> ) indicative of rock brittleness. This framework provides validated decision rules to reject "one-size-fits-all" strategies, enabling designs tailored to subsurface conditions for simultaneously enhanced production and efficiency.