Testing for the ‘predictability’ of dynamically triggered earthquakes in The Geysers geothermal field

Abstract

The Geysers geothermal field is well known for being susceptible to dynamic triggering of earthquakes by large distant earthquakes, owing to the introduction of fluids for energy production. Yet, it is unknown if dynamic triggering of earthquakes is 'predictable' or whether dynamic triggering could lead to a potential hazard for energy production. In this paper, our goal is to investigate the characteristics of triggering and the physical conditions that promote triggering to determine whether or not triggering is in anyway foreseeable. We find that, at present, triggering in The Geysers is not easily 'predictable' in terms of when and where based on observable physical conditions. However, triggered earthquake magnitude positively correlates with peak imparted dynamic stress, and larger dynamic stresses tend to trigger sequences similar to mainshock-aftershock sequences. Thus, we may be able to 'predict' what size earthquakes to expect at The Geysers following a large distant earthquake. Highlights ► Template matching event detection enhances statistical triggering test quality. ► Maximum triggered earthquake size positively correlates with dynamic stress. ► However, when and where triggering will occur is not easily 'predictable'.

Going Beyond Rate Changes as the Sole Indicator for Dynamic Triggering of Earthquakes

“…Causality is assumed primarily because local events are either coincident with the passage of the surface waves, as determined by the presence of local earthquakes in the wavefield [11][12][13][14][15] or the triggered events initiate within several hours after the passage of the mainshock seismic waves, leading to the concept of delayed dynamic triggering [16][17][18][19] . The prevalence of identified mainshocks generating remote, small magnitude earthquakes has led to two hypotheses: (1) remote dynamic triggering is ubiquitous 13 and (2) dynamically triggered events are small, often below the magnitude of completeness (Mc) and to identify these events requires catalog enhancement 14,[20][21][22][23][24] .…”

“…This movement argues that we should abandon relying on a given significance value, like a p-value of 0.05 or 95% confidence interval (which were arbitrarily chosen to begin with), and instead move toward a compatibility assessment. In our analysis, we examine statistical significance in context of the spatial characteristics of the triggered events and the temporal behavior of the rate changes (factors also examined by Aiken et al 21 ).…”

“…However, if following this path, then we will also need to define indicators of dynamic triggering beyond rate. These could include the spatial and temporal classes identified in this and other papers 21 . Our findings suggest down-weighting the link to dynamic triggering for bursty sequences and increasing the weight for widespread spatial triggering.…”

“…Our findings suggest down-weighting the link to dynamic triggering for bursty sequences and increasing the weight for widespread spatial triggering. Other factors to consider include peak ground velocity and strain changes 14,17,29,30,35,[47][48][49][50] and the orientation of incoming seismic waves with respect to the local faults 12,21,[32][33][34][35] .…”

Analysis of Microseismicity Framing ML > 2.5 Earthquakes at The Geysers Geothermal Field, California

“…Application of the matched filter technique to a particular data set may result in a different decrease of the magnitude of completeness depending on the scope of each study. For example, Aiken et al () investigated potential dynamic triggering at The Geysers by applying a matched filter analysis to continuous vertical recordings half a day before and one day after 25 distant main shocks. The authors were able to lower the magnitude of completeness by 1 full order of magnitude, down to M L ~0, by utilizing around 3,000 templates from the entire geothermal field.…”