The report presents the second iteration of a geological susceptibility model for estimating the potential for induced seismicity in the Duvernay Formation of central Alberta. It documents the changes made to the model to more faithfully represent the induced seismicity situation in central Alberta. This second version includes updates to the observed seismogenic clusters, including two in the newly emerging East Shale Basin of the Duvernay. As well, a new input feature (proximity to the Leduc Formation) is included in the analysis. Slight methodology tweaks were made to implement a mixed L1/L2 (lasso/ridge) regularization, and to better capture the geometry of horizontal wells. Overall, the new model recovers many attributes of the previous version and performs slightly better.
We previously suggested using geological data as input features, and assignment of wells as seismogenic/aseismic, to train a machine learning algorithm to evaluate tectonic, geomechanical, and hydrological proxies suspected of controlling induced seismicity. However, the initial model was limited by a lack of subsurface geological information, a restricted number of developed wells, and a paucity of induced earthquake clusters.
Schultz, R. and Pawley, S.M. (2019): Induced earthquake geological susceptibility model for the Duvernay Formation, central Alberta – version 2; Alberta Energy Regulator / Alberta Geological Survey, AER/AGS Open File Report 2019-02, 8 p.