Forward and Inverse Modelling of Induced Seismicity Using TOUGH2 Coupled with an Earthquake Simulator

Abstract

Reservoir numerical modelling is undertaken to approximate the current physical state of a reservoir to predict its future response. To improve the match between the model and the real system, calibration is undertaken by adjusting reservoir parameters, producing a more reliable model. Microseismicity presents a means to calibrate reservoir parameters, particularly the permeability of a fault that serves as a primary fluid pathway. Fluid injection causes pressure build-up in the reservoir, which decreases rock yield strength and promotes shear failure: a small earthquake. The location, migration, and number of events provide information about the nature of fluid flow through the reservoir. The study aims to integrate microearthquake (MEQ) data into a reservoir model development and calibration workflow to estimate the permeability of formations and faults. This study considers an induced seismicity model representing an area where fluid is injected. A forward run using TOUGH2 simulator is conducted to estimate pressure changes due to injection into a single well for specified reservoir and fault parameters. Pressure change on the fault is used to compute its seismicity rate as well as the spatiotemporal evolution of the events. The relationship between reservoir pressure evolution and synthetic microseismicity provides the physical link necessary to use field MEQ data for calibration. This study uses seismicity migration rate to estimate permeability of the reservoir and the fault. The result suggests that identifying reservoir parameters is possible given a large MEQ dataset. The confidence in estimation depends on the number of MEQ data used in calibration and is influenced by the accurate estimate of the other reservoir parameters (e.g., porosity) which the inversion relies upon. The methods developed are then applied to set of earthquakes triggered by wastewater injection wells in the Guy-Greenbrier area of Arkansas. This case study involves development of a simple model to estimate the seismicity rate for the area using the fluid pressure evolution, calibration to match the simulated seismicity rate with the actual MEQ data, and inverse modelling to estimate the permeability of the Ozark Aquifer where the fluid is injected and the Guy-Greenbrier fault where most of the microseismic events occur.