- © 2007 by the Seismological Society of America
A worldwide compilation of well-constrained fault ruptures and focal depths of earthquakes reveals that the Earth's crust in many stable continental regions (SCRs) is characterized by a bimodal depth distribution with a very shallow upper crustal component. The distributions can vary in a) depth of the modes and b) strength of bimodality, probably due to intracrustal boundaries, differences in frictional and rheological properties, heat-flow densities, strain-rates, and tectonic forces or forces stemming from the surface. Overall, SCR ruptures and SCR earthquakes are confined within the upper third (0–10 km) and/or the lower third of the crust (20–35 km), while the midcrust (10–20 km) tends to be aseismic. Historical data indicate that some SCRs show very well-developed bimodal distributions of focal depths (e.g., North Alpine foreland basin in Europe, Kachchh basin in India), while others show weak to no developed bimodal distributions (e.g., Charlevoix seismic zone, New Madrid seismic zone in the central United States). Moreover, many large SCR earthquakes (Mw 4.5–8.0) nucleate on reverse faults and close to the surface (< 5 km). Almost 80% of the seismic moment density of shallow SCR ruptures is released in the uppermost 7 km of the crust. However, focal depths of instrumentally recorded major SCR earthquakes (3.5 < mb < 6.2) and their aftershocks in the northeastern United States and adjacent Canada, for example, suggest systematic overestimates of hypocentral depths of 88 ± 30% (standard mean ± standard mean error), probably due to sparse instrumental coverage. If error estimates for shallow SCR earthquakes, in particular, are of systematic and not of statistical origin, preconceived assumptions of focal depths within the midcrust might have region-specific implications for understanding SCR seismogenesis and for earthquake hazard estimations (e.g., ground motion).
Stable continental regions (SCRs) can be distinguished from tectonically unstable or active continental regions (ACRs) due to their low seismic activity (Johnston 1996) and their extremely low strain rates of about 10–12 per year (Anderson 1986). These low strain rates are manifested by undeformed geologic markers, such as horizontal Paleozoic strata with no significant faulting. For example, the New Madrid seismic zone in the central United States and the Charlevoix seismic zone in southeast Canada can be considered as SCR regions of higher seismic activity, whereas strain rates in both regions are very low if compared with ACRs such as California, Japan, or Turkey. ACRs can exhibit strain rates in the range 10–8–10–7 per year.
But earthquakes do occur in SCRs and they suggest coherent stress deformation over large areas of continents (Sbar and Sykes 1973) and long time periods. Differential stress in the upper SCR crust can measure very high (Brown and Windsor 1990; Townend and Zoback 2000) and pervasively close to failure. This is consistent with observations of trigger-sensitive seismicity in certain SCR regions (Hough et al. 2003) including many instances where Mohr-Coulomb shear stress changes of only 0.01 MPa (0.1 bar) are enough to trigger earthquakes (Evans 1989; Klose 2007a; Seeber et al. 1998; Talwani 1976).
The seismological spotlight has been generally on ACRs, and the shallow part of SCR seismogenesis, in particular, was given relatively little scientific and public attention. Damaging or potentially damaging SCR earthquakes abound and pose a significant threat to lives and property worldwide (Johnston and Kanter 1990), despite the low average level of seismicity. The 2001 M 7.6 Bhuj earthquake in India, for example, was devastating, and the 1811–12 M 7 earthquake sequence in the central United States would have been devastating had it occurred one or two centuries later. Even medium-size seismic events can be very damaging, such as the 1989 M 5.6 Newcastle earthquake in Australia.
Earthquakes are symptomatic of strain and mechanical conditions of the crust (Scholz 2002). For example, the lower depth limit of cratonic seismicity has been studied to determine whether brittle behavior can reach into the upper mantle (Chen and Molnar 1983) or is limited to the lower crust (Maggi, Jackson, McKenzie et al. 2000; Maggi, Jackson, Priestley et al. 2000; Mitra et al. 2005).
Relatively little is known about very shallow seismogenesis. This is particularly true in SCRs, where seismic stations are sparse and recordings of shallow earthquakes tend to be exclusively via ray paths of narrow-degree angles between source and receiver. Nevertheless, reliable depth determinations have been accumulating rapidly since the 1970s.
This article presents quantitative evidence of shallow seismicity within several well-monitored SCRs worldwide. Different data sets of SCR fault extensions and SCR earthquakes were used to show a) similarities and dissimilarities between different SCRs and b) inaccuracies of focal depth determinations of shallow SCR earthquakes.
EARTHQUAKE RUPTURES AND SEISMIC MOMENTS IN SCRs
Worldwide earthquakes in SCRs are summarized in several catalogs (Schulte and Mooney 2005; Johnston 1996), but the number of seismic events is much smaller (21 earthquakes) when delineating earthquakes with well-determined fault geometries (table 1):
the earthquake has a reliable moment magnitude Mw > 4.5, seismic moment Mo, and hypocentral depth;
the fault rupture has a reliable upper and lower boundary (rupture depth)—obtained from local aftershock monitoring or teleseismic waveform modeling.
A vertical profile of the seismic moment density Mo,i / Wi (in Nm per km depth) was determined for each fault i = 0,1,2,..., N with rupture width W (table 1). Mo was assumed to be equally distributed along the width of each rupture to keep the analysis as simple as possible. Overall, the vertical profile of the seismic moment density M̃o (in % per km depth) for shallow (deep) SCR faults was obtained by averaging Mo,i / Wi over N shallow (deep) ruptures as follows: The depth profile of the seismic moment density in the crust that is averaged over 21 SCR faults worldwide shows a bimodal distribution (figure 1). Taking the crustal thickness (Mooney et al. 1998) into account (table 1), SCR faults tend to rupture either in the upper third or lower third of the crust (figure 2). Most ruptures are reverse, indicating high horizontal and low vertical principal stress components in the SCR crust. Shallow faults release most of their seismic moment in the upper 7–10 km of the crust (figure 1). Moreover, most shallow earthquakes nucleate at very shallow depths (< 7 km), while often breaking the surface. Both shallow and deep zones of faults can rupture during a single mainshock, while the midcrust tends to remain aseismic. Both zones can be active during aftershock sequences as well (figure 3C), as aftershocks of the M 7.6 Bhuj earthquake in India exemplify (Antolik and Dreger 2003; Bodin and Horton 2004).
REGION-SPECIFIC FOCAL DEPTHS IN SCRs
The following earthquake catalogs of different SCRs were used to generate seismic profiles (number of events per km depth):
Al Kalabsha, Egypt, North Africa (Awad et al. 2005)
North Alpine foreland basin, Switzerland, West Europe (Earthquake Catalog of Switzerland, ECOS, http://www.seismo.ethz.ch)
Bhuj earthquake aftershocks, India, South Asia (Bodin and Horton 2004)
Charlevoix seismic zone, Canada, North America; relocated data (Richards et al. 2006)
Eastern Tennessee seismic zone, United States, North America (SEUSS catalog)
The earthquake catalogs were compiled from data collected from regional seismic networks (station spacing > 10 km) and local seismic networks (station spacing < 10 km). Magnitude threshold values for the completeness of each catalog were determined by deviations from the linear Gutenberg-Richter distribution to avoid a possible bias and to ensure the comparability among all SCR earthquake depth profiles. Thus, earthquakes were considered when their magnitudes were larger than a network-specific threshold value (figure 3).
Hypocentral depth values of the analyzed SCRs indicate shallow seismogenic components within the upper part of the continental crust. Many of these regions are also characterized by bimodal crustal depth distributions of seismicity with an additional deeper component in lower parts of the crust (figure 3). Earthquakes can nucleate shallower than 10 ± 2 km and are also confined below 20 ± 2 km. This phenomenon is in contrast to generally well-developed unimodal seismicity distributions in ACRs (Sibson 1982; Scholz 2002). Both bimodality of a distribution and depth locations of a mode can vary widely, reflecting different tectonophysical conditions (rock strength, geology, geothermal profile, strain rates, tectonic forces, or forces from the surface of the Earth's crust). A decreasing strength of bimodality (figures 3A–F) might be a result of a reduction in surface heat-flow density, changing strain rates or different geological/rheological properties at upper, middle, and lower crust boundaries. Thus, depth values of dominant modes represent more unstable parts or boundaries in the crust.
The Al Kalabsha region in Egypt (Simpson et al. 1990; Awad et al. 2005), for instance, shows well-developed bimodal seismicity with a very shallow depth mode at 5 km and a deep mode in the mid/lower crust at 19 km (figure 3A). The two depth modes of seismicity in the North Alpine foreland basin reflect intracrustal boundaries (figure 3B), which were imaged, e.g., by reflection/refraction seismic profiling (Schmid and Kissling 2000; Schmid et al. 2004). Shallow peaks at 7 km and 11 km show the upper and lower boundaries of the Mesozoic sedimentary cover above the upper crustal basement. It can be anticipated that the peak at 11 km is biased due to fixed depth values in the velocity models. The deep peak at 25 km may represent the boundary between the upper crustal basement and the lower crust.
Aftershocks of the 2001 M 7.6 Bhuj earthquake in the Kachchh basin in India exemplify a well-developed bimodal depth distribution (figure 3C). The bimodality is even more distinguished when looking at the depth distribution of the seismic moments. The shapes of both modes may explain changes in the rock strength with increasing depth toward the layer boundaries at 11 km and 27 km depth. The Charlevoix seismic zone in southeast Canada shows a weak bimodal seismicity profile (figure 3D). Although the strength of the deeper mode at 23 km is weaker compared with that of other regions (e.g., Al Kalabsha region), seismicity occurs at this depth in the vicinity as well, as exemplified in the Saguenay earthquake sequence (North et al. 1989). Both the eastern Tennessee seismic zone (figure 3E) and the New Madrid seismic zone (figure 3F) show shallow seismicities and no developed bimodal distributions. A more detailed analysis of the seismogenic profile of the New Madrid seismic zone reveals a) a much shallower mode and b) a weak deep mode in its vicinity (figure 4). Although one might suspect that these deep seismic events at 18–22 km depth in the vicinity of the New Madrid seismic zone are not well-determined, these depth values could be very precisely constrained (Kim 2003) and in even deeper depth, ≥ 25 km (Gordon et al. 1970).
ACCURACY OF SHALLOW FOCAL DEPTH DETERMINATIONS
A data set was selected consisting of 25 medium-size SCR earthquakes (3.1 < mb < 6.2) that had nucleated in the North American northeast since 1968 and for which source depths were well-constrained by aftershock monitoring (figure 5). Depth values of mainshocks, which were recorded by regional seismic networks (station spacing > 10 km), came from the national U.S. Geological Survey catalog (http://neic.usgs.gov/neis/epic/epic.html). Aftershocks were precisely located using portable seismographs as local seismic networks (station spacing < 10 km). For these events, a comparison was made between hypocentral depths obtained from regional versus local seismic networks.
Aftershock sequences, tightly clustered around ruptures of 25 analyzed earthquakes in northeast North America, clearly constrain the mainshock locations, because the magnitudes of these earthquake ruptures are of small to medium size (3.1 < mb < 6.2). The data indicate systematic differences between more accurate hypocentral depth values of local seismic networks and values that were determined by regional networks (figure 5). Moreover, the data suggest that in particular, regional networks overestimate shallow earthquake hypocenters by 88 ± 30% on average (standard mean ± standard mean error). This is because trial depths tend to remain unchanged in the iterative location procedure and constraints on focal depths from sparse regional seismic networks are very weak. Preconceived notions on depth distributions, probably derived from ACR seismicities, may come into play in setting trial depths. These depth values should be revised for each SCR when earthquakes nucleate in very shallow but also deep parts of the SCR crust in contrast to ACRs (Sibson 1982).
CONCLUSION AND OUTLOOK
Depth distributions of fault ruptures and earthquakes in well-monitored SCRs worldwide indicate the following features:
Many earthquakes and fault ruptures are bimodally distributed with depth modes in the upper third of the crust (< 10 km) and the lower third of the crust (> 20 km).
Depth and strength of contrasting modes vary regionally and indicate less-stable intracrustal boundaries that can be also located by seismic profiling.
Many medium-to-large earthquakes (4.5 < Mw < 8.0) tend to nucleate at very shallow depths (< 7 km), while the ruptures often reach the surface.
Almost 80% of the seismic moment density for shallow ruptures is released in the uppermost 7 km of the crust.
Hypocentral depths of earthquakes are on average over-estimated by 88 ± 30% (standard mean ± standard mean error).
Overall, depth profiles of hypocenters, fault ruptures, or seismic moment releases within many SCRs worldwide show significant and systematic distinctions when compared with ACRs. These discrepancies have been partially masked by a tendency to overestimate depths of shallow SCR earthquakes, probably due to sparse regional seismic networks in SCRs and ACR-similar crust models for routine hypocenter determinations.
Such differences between ACR and SCR seismogeneses, but also within SCRs, are likely to stem from fundamental dissimilarities in geomechanical properties that could derive from differences in both heat flow and fracture density in the upper crust. Other tectonophysical mechanisms could also affect the depth distribution of seismogenesis, such as strain/stress accumulations, tectonic forces, or forces from the surface of Earth's crust.
Moreover, major SCR earthquakes recorded in northeastern North America since 1968 have only medium-sized magnitudes (3.1 < mb < 6.2). Although spatially more restricted, many more low-magnitude earthquakes in SCRs can reach damage thresholds, because the number of earthquakes increases exponentially with lower magnitude (Gutenberg-Richter relation). Consequently, even moderately sized earthquakes can be very devastating. For example, almost 10,000 people died as a result of the 1993 M 6.4 Killari earthquake (India). The 1989 M 5.6 Newcastle earthquake in southeastern Australia is believed to be the first earthquake in Australian history that caused fatalities, with 13 deaths. It is probably the most costly of any M 5 earthquake, with damage estimated at more than $3.5 billion (U.S., 1989 value) (Klose forthcoming).
Overall, depth distributions of SCR seismicity with a very shallow crustal component have scientific implications, although the presented results need to be further tested and refined since this analysis is based on relatively few reliable data. The occurrence of strong earthquakes over a very wide range of crustal depths, including very shallow events, may also have important practical implications for earthquake hazard estimations. Algorithms that predict ground motion assuming an “effective” SCR source depth (e.g., Atkinson and Wald 2007) are likely to underestimate the maximum ground shaking from shallow earthquakes. Such estimations may erroneously exclude shallow earthquakes from hazard concerns, despite their potential to cause local damage at very short hypocentral distances (Kövesligethy 1907; Yilmaztürk and Burton 1999; Abrahamson and Silva 1997). Our results strongly suggest that region-specific mapping and monitoring of earthquake hypocenters in SCRs will improve the understanding of SCR seismogeneses and the tectonophysical differences between ACRs and SCRs. Furthermore, ground-motion regressions need to be carried out keeping depth as a free parameter.
The authors are grateful to H. Awad for providing the earthquake catalog for the Al Kalabsha region and to K. H. Jacob for reviewing the manuscript. C. D. Klose expresses his gratitude to the German Science Foundation (DFG), which funded this study as part of an Emmy Noether research grant (KL 1833/1).
Lamont-Doherty Earth Observatory, Columbia University