A combination of waveform inversion and reverse-time modelling for microseismic event characterization in complex salt structures

The increased emission of greenhouse gases into the atmosphere, causing climate changes, leads to a strong requirement of renewable energy resources. However, they are intermittent and need buffer storage to bridge the time gap between production and public demands. The injection of gas (e.g. compressed air or hydrogen) in sealed underground structures like salt caverns is one approach to solve this problem. Possible risks related to cavern storage are gas leakages from the injection tube into the surrounding sediments, material failure in salt rock surrounding the cavern during irregular operation and in the most extreme case a partial collapse of the cavern. For the early detection of these problems, a geophysical monitoring strategy is required. The objective of this paper was to map possible leakage paths outside of the salt structures and local failures within the cavern walls by the localization of crack-induced microseismic events. Classical methods require arrival time picking and phase identification. An alternative approach is elastic reverse-time modelling (RTMOD), where the recorded microseismic events are numerically backpropagated from the receiver positions into the elastic underground model. The resulting seismic wavefield focuses at the location of the event, which can be subsequently imaged by estimating the maximum of the seismic energy at each underground point. However, the success of this approach highly depends on the used elastic background model. In case of complex salt bodies, the strong velocity contrast between the salt and the surrounding sediments is a major problem. Therefore, we propose a combined monitoring approach, consisting of a seismic full waveform inversion of active source reflection seismic data to accurately image the background velocity model and subsequent RTMOD for the microseismic event localization. Accuracy and sensitivity with respect to the acquisition geometry and random noise will be demonstrated using a complex benchmark model. Furthermore, the localization accuracy is discussed for three different scenarios covering the detection of a partial cavern collapse, a gas leakage and the occurrence of cracks within the cavern wall due to extreme loading conditions during irregular operation.