Causal Inference for the Characterization of Microseismic Events Induced by Hydraulic Fracturing

Abstract

Abstract This study proposes a workflow that employs causal inference techniques on microseismic data acquired during hydraulic fracturing operations on 2 horizonal wells in Marcellus Shale. The study quantifies the causal relationships between a new microseismic event and the prior "spatiotemporally proximal" microseismic events, while taking into account the confounders that influence both the causes and effects. In doing so, we explain the magnitude, location, and occurrence of a new microseismic event produced during hydraulic fracturing as a consequence of the prior "spatiotemporally proximal" microseismic events. The causal relations quantified in this study are beyond statistical correlation/association tests. The study provides new insights into the microseismic-source mechanisms, such as: 1) Magnitude of a new microseismic event does not depend on the number and the spatial and temporal concentrations of the spatiotemporally proximal, prior microseismic events; 2) Regions with high microseismic magnitude events produce a new microseismic event earlier in time; and 3) Microseismically active regions produce a new microseismic event much earlier in time. Selection of true confounders is crucial for obtaining accurate causal estimates. Failure to properly select confounders can result in significant overestimation or underestimation of the causal estimates, as high as +/- 100%. Certain treatment-outcome pair exhibit large differences between the causal estimates and correlation coefficients that confirm the independence of causation and correlation. A causal analysis with true confounders reveals the true causal relationship that cannot be quantified using correlation/association methods.