Understanding Poroelastic Stressing and Induced Seismicity with a Stochastic/Deterministic Model: an Application to an EGS Stimulation at Paralana, South Australia, 2011

Abstract

Induced seismicity is the occurrence of earthquakes associated with engineering activities such as Enhanced Geothermal System (EGS) operations. The induced microseismicity distribution carries information about both the reservoir and the stimulation process. On the other hand, induced seismicity poses a variety of environmental, economic and public safety risks. Therefore, it is important to investigate the mechanisms of induced seismicity using numerical models. There are at least two mechanisms responsible for injection induced seismicity: pressurization within a fracture and poroelastic stress changes. Fracture pressure increase, by reducing the effective normal stress, allows for the shear failure of existing faults and fractures. Pressure increases also modify the wider reservoir stress field through poroelastic effects. If large enough, poroelastic stress changes can also induce shear failure. The goal of this study is to use a numerical model and its application to the 2011 Paralana EGS stimulation to characterize events triggered by the pressurization mechanism and those triggered by poroelastic effects. We use a model that mixes deterministic and stochastic methods to generate synthetic catalogs of microseismicity, which comprise event locations, times and magnitudes. The model couples: (1) deterministic reservoir simulation of subsurface fluid flow, heat transfer, and stress and permeability evolution with FEHM; and (2) a mixed deterministic-stochastic seismicity model in which events in the reservoir are triggered by Mohr-Coulomb failure, magnitudes are assigned from a Gutenberg-Richter probability distribution. Mohr-Coulomb failure can be caused by either pressurization or poroelastic effects stress changes; thus our model captures the two triggering mechanisms. We have applied our model to the 2011 Paralana-2 EGS stimulation, South Australia, during which about 3 million liters of water were injected over a period of 5 days, inducing 7085 microearthquakes with a maximum observed magnitude of 2.5. We calibrate our model using injection records and a high-resolution microearthquake dataset. We find that poroelastic mechanisms must be accounted for to obtain a realistic spatial distribution of seismicity. In particular, observations of events triggered at large distance from the injection point early on during the stimulation are difficult to account for by pressurization alone. Comparing the data and the model, we explore the relative importance of the two mechanisms in different parts of the reservoir and at different times.