Analysis of Geological Susceptibility to Induced Seismicity in the Montney Formation Using Supervised Machine Learning

Summary This project aims to determine the most important geological factors influencing the susceptibility to induced seismicity in the Montney Formation geological and geomechanical characteristics including pressure gradient, distance to the Cordilleran foreland thrust and fold belt and known lineaments, proximity to the Precambrian basement and Debolt formation, variation of maximum horizontal stress direction and depth factor were investigated. Supervised machine learning methods including four different Tree-based methods (Decision Tree, Bagging, Random Forest and Gradient Boosting) were used to calculate the feature importance. Geological susceptibility analysis was performed using Logistic Regression, commonly used for the probability estimation. The analysis of the Tree-based algorithms suggests three types of characteristics having the biggest impact on the geological susceptibility to induced seismicity in the Montney Formation: (1) variance of the SHmax direction from the regional trend, (2) vertical distance to Precambrian basement and (3) depth of the injection relative to the Montney top. Pore pressure gradient and distance to the Debolt Formation were interpreted as least influencing the geological susceptibility distribution. The highest discrepancy in geological susceptibility levels was observed in the northern part of the formation. Moreover, the Lower Montney was determined as most susceptible to induced seismic activity of all units.