Effects of source model variations on Coulomb stress analyses of a multi-fault intraplate earthquake sequence

Highlights

• Coulomb stress modelling has been used to investigate the rupture behaviour of the 1987 to 1988 Tennant Creek earthquake sequence in Australia.

• Coulomb stress triggering explains the sequence evolution for most of the fault sequence models.

• Increases in knowledge of source models may increase confidence in attributing spatiotemporal patterns of earthquakes to stress changes.

Abstract

Fault models are quickly produced and iteratively improved over weeks to years following a major earthquake, to characterise the dynamics of rupture, evaluate the role of stress transfer, and contribute to earthquake forecasting. We model Coulomb stress transfer (ΔCFS) between the largest foreshock (Mw 5.4; 1 year prior to first mainshock) and three Mw 6.1 to 6.5 earthquakes that occurred in a 12-hour period on January 22, 1988 in central Australia (Tennant Creek earthquake sequence) to investigate the role of static stress transfer in earthquake triggering relative to progressive source model development. The effects of fault model variance are studied using ΔCFS modelling of five different fault source model sequences (27 total models) using different inputs from seismic and geospatial data. Some initial models do not yield positive ΔCFS changes proximal to hypocentres but in all models, preceding earthquakes generate positive ΔCFS (≥0.1 bar) on ≥10 to 30% of the forthcoming receiver fault rupture areas. The most refined and data-integrative model reveals ΔCFS ≥ +0.7 to +13 bars within 2 km of impending hypocentres and large (≥30 to 99%) areas of positive ΔCFS. When compared to global compilations of threshold ΔCFS prior to impending ruptures (average = 3.71 bar, median = 1 bar), this suggests that Coulomb stress change theory adequately explains the Tennant Creek rupture sequence. In the most-refined model, earthquake inter-event times decrease as ΔCFS increases, suggesting that higher stress magnitudes may have more rapidly (within hours) triggered successive events, thus accounting for some temporal aspects of this sequence. ΔCFS analyses provide a useful framework for understanding the spatiotemporal aspects of some intraplate earthquakes. The progressive refinement of source models using emergent data may reduce epistemic uncertainties in the role of stress transfer that result from different model inputs, approaches, and results.

Introduction

Multi-fault earthquakes and earthquake sequences are common in many tectonic settings on Earth (e.g., Sieh et al., 1993; Fletcher et al., 2016; Wei et al., 2011; Beavan et al., 2012; Elliott et al., 2012; Hamling et al., 2017; Quigley et al., 2019; Clark et al., 2012). Distinct fault ruptures may be triggered as part of a quasi-continuous cascading seismic moment release, or be triggered seconds, minutes, hours, days, or even decades following preceding earthquakes (Belardinelli et al., 2003a, Belardinelli et al., 2003b; Freed, 2005; Nissen et al., 2016; Stein, 1999). Stress changes due to moderate-to-large earthquakes may affect the location of subsequent events by processes including static stress transfer, dynamic (i.e., coseismic) stress changes and viscoelastic stress change (e.g., King et al., 1994; Stein, 1999; Nostro et al., 1997; Reasenberg and Simpson, 1992; Stein et al., 1997; Lin and Stein, 2004; Steacy et al., 2013). The relative importance of these different types of stress changes in earthquake sequences may be challenging to evaluate as they all may provide scientifically valid explanations for aspects of the same sequence. Static stress changes estimated from Coulomb stress calculations (e.g., Harris and Simpson, 1998) have been shown to provide an important explanation for some instances of earthquake triggering and clustering (e.g., King et al., 1994; Harris, 1998; Stein, 1999; Harris et al., 1995; Freed and Lin, 2001; Freed, 2005; Steacy et al., 2005; Mohammadi and Bayrak, 2015; Mohammadi et al., 2017) by suppressing or encouraging rupture on receiver faults. Coulomb stress change analysis may be a potentially powerful forecasting tool if it can be reliably applied to rapidly-developed fault models (Steacy et al., 2014).

Although the calculation of Coulomb stress changes is well-established, the early stages of many earthquake sequences see the emergence of multiple diverse models for fault ruptures that vary in fault geometries and co-seismic slip due to different modelling approaches and utility of different datasets (e.g., seismological data, geologic data, optical data, InSAR, geodetic data). Model variations impart significant epistemic uncertainty to interpretation of the relationships between stress transfer and earthquake sequencing (Wang et al., 2014; Zhan et al., 2011). Changing parameters of the Coulomb stress change model, including the stress change tensor (related to changing geometries and slip distributions of the source fault), receiver fault geometry, and the friction coefficient and Skempton's coefficient, may impact on how well static stress change analyses explain observation of spatiotemporal patterns of earthquake sequences (Lin and Stein, 2004; Zhan et al., 2011; Wang et al., 2014; Mildon et al., 2016). Fault model uncertainties are rarely applied to stress-triggering studies; although some studies have examined uncertainties in receiver fault geometries (e.g., Harris and Simpson, 2002; Steacy et al., 2005; Lasocki et al., 2009). Woessner et al. (2012) conducted a thorough analysis of the effects of fault model uncertainties on Coulomb stress models for the moment magnitude (Mw) = 5.9 June 2000 Kleifarvatn earthquake in southwest Iceland.

Another limitation of many Coulomb stress change studies is the ambiguity with which estimated stress changes on receiver faults are interpreted to have been enough to trigger rupture. A Coulomb stress increase of 0.01 MPa (0.1 bar) is commonly proposed to be the threshold for potential earthquake triggering (Harris, 1998; Reasenberg and Simpson, 1992; Freed, 2005; King et al., 1994; Stein, 1999). However, the stress threshold to trigger instantaneous rupture on receiver faults concurrent with the hypocentral source fault rupture may be significantly higher. For example, Coulomb stress changes of >0.1 MPa (Zhan et al., 2011) and 1 to 1.5 MPa (Walters et al., 2018) were insufficient to generate spontaneous rupture during the 2010–2011 Canterbury earthquake sequence and 2016 Central Italy seismic sequence, respectively. Instead subsequent receiver fault ruptures occurred days (Walters et al., 2018) to months (Zhan et al., 2011) after initial stress loading from prior mainshocks.

In this paper, we apply Coulomb stress modelling to investigate the rupture behaviour of the 1987 to 1988 Tennant Creek earthquake sequence in Australia. We aim to investigate (1) whether static stress changes on receiver faults induced by preceding earthquakes provide an explanation for the observed spatiotemporal patterns of this sequence, including the hypocentral locations and inter-event timing, and (2) whether differences in fault geometry and rupture kinematics (associated with different rupture models), influence static stress changes significantly enough to cast uncertainty over whether Coulomb stress models adequately explain this sequence. We also consider maximum calculated stress change increases on receiver faults in the context of stress triggering thresholds and time lags between source and receiver fault ruptures for this and other earthquake sequences globally.