The temporal evolution of induced seismicity sequences generated by low-pressure, long-term fluid injection

Abstract

Traffic light schemes (TLSs) are commonly used to mitigate induced seismicity caused by subsurface fluid injection. Verdon and Bommer (J Seismol 25:301–326, 2021a) investigated the validity of the implicit assumptions that underpin the successful functioning of TLSs. In particular, they examined the extent to which magnitude jumps (sudden increases in event magnitudes from any preceding seismicity) and trailing events (continued increases in seismicity levels after the end of injection) took place in hydraulic fracturing (HF) induced seismicity sequences. Other technologies such as carbon capture and storage (CCS), wastewater disposal (WWD) and natural gas storage (NGS) involve the gradual but long-term injection of large fluid volumes at low pressure. Hence, we might expect to see a different spatial and temporal evolution of magnitudes for seismicity induced by low-pressure, long-term (LPLT) injections compared to HF. In this study, we compile cases of LPLT injection-induced seismicity in order to examine their temporal evolution. We examine the occurrence of magnitude jumps, trailing events and onset times for seismicity after the initiation of injection. We find that few LPLT injections have produced trailing events, and that magnitude jumps are typically below 1.5 magnitude units. The timescale of event occurrence (relative to the onset of injection) is highly variable, likely reflecting site-specific conditions. For long-term injection, we observe a trend for the largest events to occur within the earlier part of each sequence, with magnitudes then stabilising, or even reducing, as injection continues. Finally, we evaluate the performance of the next record breaking event (NRBE) model as a method for forecasting induced event magnitudes, finding that this method performs reasonably well in most cases, but that in some cases the largest event significantly exceeds this model.

Growth and stabilization of induced seismicity rates during long-term, low-pressure fluid injection

“…Watkins et al [25] made a similar observation, finding that the largest earthquakes typically occurred within the first 20%-40% of the overall observed sequence (fig. 5 of [25]).…”

“…These time series form the basis of our subsequent analysis. We take M MIN and G-R b values for each earthquake catalogue from Watkins et al [25].…”

“…This observation was in stark contrast to the observations made by Verdon and Bommer [26] for HF-induced seismicity, where the largest events were found to be systematically towards the ends of the observed sequences. Watkins et al [25] did not compile any injection data, and so they were not able to rule out the possibility that the changes in the levels of seismicity they observed were driven solely by changes in injection rates. The objective of this study is to examine how the scaling between seismicity and injection volume, as characterized by the S I and S E parameters, evolves during subsurface injection operations.…”

“…Changes in S I and S E that are in fact generated by a spatial change in injection position could be misinterpreted as a temporal change within the same perturbed volume [13,20]. -The long-term, low rate, but ultimately high volume, nature of WWD provides a useful analogue to anticipated future activities, such as CCS, NGS and hydrogen storage, that are thought necessary to meet energy sustainability and energy security objectives [25,[28][29][30].…”

“…This acceleration usually occurs within 1000 days of the onset of the seismicity sequence (note that this is the time from the first observed seismicity at a site, not the start of injection, which in some cases may have been ongoing for many years before the onset of any observed seismicity). After this See electronic supplementary materials and Watkins et al [25] for further details for each site.…”