The papers in this volume illustrate a number of approaches that are becoming increasingly common and offer the prospect of making significant advances in the broad related topics of the science, hazard, and policy issues of large continental intraplate earthquakes. Plate tectonics offers little direct insight into the earthquakes beyond the fact that they are consequences of slow deformation within plates and, hence, relatively rare. To alleviate these problems, we use space geodesy to define the slowly deforming interiors of plates away from their boundaries, quantify the associated deformation, and assess its possible causes. For eastern North America, by far the strongest signal is vertical motion due to ice-mass unloading following the last glaciation. Surprisingly, the expected intraplate deformation due to regional stresses from plate driving forces or local stresses are not obvious in the data. Several approaches address difficulties arising from the short history of instrumental seismology compared to the time between major earthquakes, which can bias our views of seismic hazard and earthquake recurrence by focusing attention on presently active features. Comparisons of earthquakes from different areas illustrate cases where earthquakes occur in similar tectonic environments, increasing the data available. Integration of geodetic, seismological, historical, paleoseismic, and other geologic data provides insight into earthquake recurrence and the difficult question of why the earthquakes are where they are. Although most earthquakes can be related to structural features, this explanation alone has little predictive value because continents contain many such features, of which a few are the most active. It appears that continental intra-plate earthquakes are episodic, clustered, and migrate. Thus on short time scales seismicity continues on structures that are active at present, perhaps in part because many events are aftershocks of larger past events. However after periods of activity these structures may become inactive for a long time, so the locus of at least some of the seismicity migrates to other structures. Analysis of the thermomechanical structure of the seismic zones gives insight into their mechanics: whether there is something special about them that results in long-lived weak zones on which intraplate strain release concentrates, or as seems more likely, that they are not that unusual, so seismicity migrates. Accepting our lack of understanding of the underlying causes of the earthquakes, the limitations of the short instrumental record, and the possibility of migrating seismicity helps us to recognize the uncertainties in estimates of seismic hazards. Fortunately, even our limited knowledge can help society develop strategies to mitigate earthquake hazards while balancing resources applied to this goal with those applied to other needs.

A common view of continental intraplate seismicity is that large earthquakes occur in areas where peculiar local conditions favor lower lithospheric strength and/or higher stress concentration compared to typical intraplate settings. Although there are numerous explanations for these local strength reduction and stress increase effects, their application to seismic hazard assessment is limited to the few specific regions for which these explanations were developed. In this paper, I present four general models that can be used to define seismic hazards based on the associated geodynamic frameworks and their implications for earthquake locations, sizes, and recurrence rates. The four models are defined by the relationships among lithospheric strength contrasts, strain distribution, and earthquake characteristics, and they may apply to different intraplate regions. (1) The random model, defined by the lack of significant lithospheric structure and the spatial and temporal randomness of seismicity, may be applicable to Precambrian cratons and shields. (2) In the plate-boundary model, earthquakes concentrate along lithospheric-scale tectonic structures under low intraplate strain rates, which may apply to eastern North America. (3) The localized weak zone model postulates that large earthquakes are limited to small areas of crustal weakness and high strain concentration (e.g., New Madrid seismic zone in the central United States). (4) The large-scale weak zone model is characterized by high crustal strain concentration in major paleotectonic structures, along which large earthquakes are spatially confined but susceptible to migration with time. This last model may apply to Paleozoic and Mesozoic rift and basin regions, such as the St. Lawrence valley in eastern Canada. Because all four models are built on the relationship between lithospheric strength, strain distribution, and earthquake characteristics, they can be used as a framework for experiments designed to test their validity. I discuss two lines of studies that address the relationship among strength, strain, and earthquakes. The first type deals with strength of the crust and upper mantle using rock rheology, thermal profiles, and average strain rates in intraplate seismic regions. The second type is based on geodetic measurements of intraplate strain rate patterns and amplitudes.

The spatial distribution of seismicity is often used as one of the indicators of zones where future large earthquakes are likely to occur. This is particularly true for intraplate regions such as the central and eastern United States, where geology is markedly enigmatic for delineating seismically active areas. Although using past seismicity for this purpose may be intuitively appealing, it is only scientifically justified if the tendency for past seismicity to delineate potential locations of future large earthquakes is well-established as a real, measurable, physical phenomenon as opposed to an untested conceptual model. This paper attempts to cast this problem in the form of scientifically testable hypotheses and to test those hypotheses. Ideally, thousands (or even millions) of years of data would be necessary to solve this problem. Lacking such a long-term record of seismicity, I make the “logical leap” of using data from other regions as a proxy for repeated samples of seismicity in intraplate regions. Three decades of global data from the National Earthquake Information Center are used to explore how the tendency for past seismicity to delineate locations of future large earthquakes varies for regions with different tectonic environments. This exploration helps to elucidate this phenomenon for intraplate environments. Applying the results of this exercise to the central and eastern United States, I estimate that future earthquakes in the central and eastern United States (including large and damaging earthquakes) have ∼86% probability of occurring within 36 km of past earthquakes, and ∼60% probability of occurring within 14 km of past earthquakes.

Attempts to study earthquake recurrence in space and time are limited by the short history of instrumental seismology compared to the long and variable recurrence time of large earthquakes. As a result, apparent concentrations and gaps in seismicity and hence seismic hazard within a seismic zone, especially where deformation rates are slow (<10 mm/yr), are likely to simply reflect the short earthquake record. Simple numerical simulations indicate that if seismicity were uniform within a tectonically similar seismic zone, such as the Atlantic coast of Canada, St. Lawrence valley, or the coast of North Africa, thousands of years of record would be needed before apparent concentrations and gaps of seismicity and hazard did not arise. Hence, treating sites of recent seismicity as more hazardous for future large earthquakes is likely to be inappropriate, and it would be preferable to regard the hazard as comparable throughout the seismic zone.

We examine the question of a possible difference in the frequency-size statistics of intraplate earthquakes, as opposed to their more numerous interplate counterparts. We use both the Harvard Centroid Moment Tensor catalogue and the data set of the National Earthquake Information Center. In the former case, we quantify earthquakes through their seismic moment and describe their population distribution through the β-value introduced by Molnar. In the latter case, we use traditional b -values computed from both body-wave magnitudes ( m b ) and surface-wave magnitudes ( M s ). We conclude that both β- and b -values for true intraplate earthquakes (i.e., not occurring in areas of broad tectonic deformation) are essentially equivalent to those of interplate earthquakes in similar ranges of moments or magnitudes. This is consistent with a fractal dimension of two for the intraplate seismogenic zones, suggesting that, like along plate boundaries, they consist of two-dimensional faults and not of volumes with greater dimensions. The distribution of earthquakes in deformed regions, principally the Mediterranean-Tethyan belt, follows that of worldwide inter-plate earthquakes but with a greater value for the critical moment expressing the saturation with depth of the width of the fault at the brittle-ductile transition, suggesting that the latter would take place at greater depths under large-scale orogens.

Since 1992, remotely triggered earthquakes have been identified following large (M > 7) earthquakes in California as well as in other regions. These events, which occur at much greater distances than classic aftershocks, occur predominantly in active geothermal or volcanic regions, leading to theories that the earthquakes are triggered when passing seismic waves cause disruptions in magmatic or other fluid systems. In this paper, I focus on observations of remotely triggered earthquakes following moderate main shocks in diverse tectonic settings. I summarize evidence that remotely triggered earthquakes occur commonly in mid-continent and collisional zones. This evidence is derived from analysis of both historic earthquake sequences and from instrumentally recorded M5–6 earthquakes in eastern Canada. The latter analysis suggests that, while remotely triggered earthquakes do not occur pervasively following moderate earthquakes in eastern North America, a low level of triggering often does occur at distances beyond conventional aftershock zones. The inferred triggered events occur at the distances at which SmS waves are known to significantly increase ground motions. A similar result was found for 28 recent M5.3–7.1 earthquakes in California. In California, seismicity is found to increase on average to a distance of at least 200 km following moderate main shocks. This supports the conclusion that, even at distances of ∼100 km, dynamic stress changes control the occurrence of triggered events. There are two explanations that can account for the occurrence of remotely triggered earthquakes in intraplate settings: (1) they occur at local zones of weakness, or (2) they occur in zones of local stress concentration.

We undertook a parametric study, using a two-dimensional distinct element method, to investigate if there is a preferred geometry of intersecting faults that may favor the occurrence of intraplate earthquakes. This model subjects two and three vertical, intersecting faults within a block to a horizontal force across them, representing the maximum horizontal compression (S Hmax ). The main fault is oriented at an angle α with respect to S Hmax , and β is the interior angle between the main fault and the intersecting fault. The third fault is oriented parallel to the main fault and is half its length. The distribution of shear stresses is examined along the faults for different values of α and β, and varying lengths of the main and intersecting faults. In all cases, maximum shear stresses are generated at the fault intersections. The modeling results reveal that the magnitudes of the shear stresses depend on the values of α and β, with an optimum range for α between 30° and 60°. In the case where the sign of the shear stress on the intersecting fault is opposite that on the main fault, the largest stresses at the fault intersections are obtained when β is between 65° and 125°. When the stresses on these two faults are of the same sign, the largest stress values at the intersections are obtained when 145° ≤ β ≤ 170°. The results of the modeling are consistent with the observed geometry of faults in the New Madrid and Middleton Place Summerville seismic zones.

The origin of earthquakes within stable continental regions has been the subject of debate over the past thirty years. Here, we examine the correlation of North American stable continental region earthquakes using five geologic and geophysical data sets: (1) a newly compiled age-province map; (2) Bouguer gravity data; (3) aeromagnetic anomalies; (4) the tectonic stress field; and (5) crustal structure as revealed by deep seismic-reflection profiles. We find that: (1) Archean-age (3.8–2.5 Ga) North American crust is essentially aseismic, whereas post-Archean (less than 2.5 Ga) crust shows no clear correlation of crustal age and earthquake frequency or moment release; (2) seismicity is correlated with continental paleorifts; and (3) seismicity is correlated with the NE-SW structural grain of the crust of eastern North America, which in turn reflects the opening and closing of the proto– and modern Atlantic Ocean. This structural grain can be discerned as clear NE-SW lineaments in the Bouguer gravity and aeromagnetic anomaly maps. Stable continental region seismicity either: (1) follows the NE-SW lineaments; (2) is aligned at right angles to these lineaments; or (3) forms clusters at what have been termed stress concentrators (e.g., igneous intrusions and intersecting faults). Seismicity levels are very low to the west of the Grenville Front (i.e., in the Archean Superior craton). The correlation of seismicity with NE-SW–oriented lineaments implies that some stable continental region seismicity is related to the accretion and rifting processes that have formed the North American continental crust during the past 2 b.y. We further evaluate this hypothesis by correlating stable continental region seismicity with recently obtained deep seismicreflection images of the Appalachian and Grenville crust of southern Canada. These images show numerous faults that penetrate deep (40 km) into the crust. An analysis of hypocentral depths for stable continental region earthquakes shows that the frequency and moment magnitude of events are nearly uniform for the entire 0–35 km depths over which crustal earthquakes extend. This is in contradiction with the hypothesis that larger events have deeper focal depths. We conclude that the deep structure of the crust, in particular the existence of deeply penetrating faults, is the controlling parameter, rather than lateral variations in temperature, rheology, or high pore pressure. The distribution of stable continental region earthquakes in eastern North America is consistent with the existence of deeply penetrating crustal faults that have been reactivated in the present stress field. We infer that future earthquakes may occur anywhere along the geophysical lineations that we have identified. This implies that seismic hazard is more widespread in central and eastern North America than indicated by the limited known historical distribution of seismicity.

At postglacial rebound time scales, the intraplate continental lithosphere typically behaves as an elastic solid. However, under exceptional conditions, the effective viscosity of the lower crust and lithospheric mantle may be as low as ∼10 20 Pa s, leading to ductile behavior at postglacial rebound time scales. We studied the effects of a lithospheric ductile zone on postglacial rebound–induced seismicity and deformation in eastern Canada and the northeastern United States using three types of models: (1) a reference model with no lithospheric ductile layer; (2) a model with a uniform, 25-km-thick, ductile layer embedded in the middle of the lithospheric column; and (3) a model with a dike-like vertical ductile zone, extending from mid-crust level down to the bottom of the lithosphere, along the Precambrian rift structure of the St. Lawrence Valley. Based on geothermal and rock physics data, the viscosity of the ductile zone is set to either 10 20 or 10 21 Pa s. We found that a narrow ductile zone cutting vertically through the lithosphere has larger effects than the uniformly thick horizontal ductile layer. Effects of a lithospheric weak zone on uplift rates may be large enough to be detected by global positioning system (GPS) measurements, especially for low viscosities. While the effect on fault stability is also large, the impact on the onset time of instability is small for sites within the ice margin. The impact on the onset time is more significant for sites outside the ice margin. Effects of a lithospheric weak zone are also significant on present-day horizontal velocities and strain rates and are at the limit of resolution for GPS measurements.

A field of stress-release bedrock structural features occurs on the floor of western Lake Ontario south of Toronto, Canada. These features were investigated using side-scan and multibeam sonars, high-resolution seismic profiling, and submersible dive observations. The study region was mostly stripped of its glacial drift in late glacial time, and the region has since accumulated only a relatively thin, discontinuous cover (1–2 m) of lacustrine sediment. The stress-release features affect the flat to gently dipping interbedded shales and calcareous siltstones of the Upper Ordovician Georgian Bay Formation. The features consist of sub-lakefloor buckles, about 50–100 m wide with structural relief of 5+ m, and surface bedrock popups, 10–15 m wide with a general relief of 1–2 m. Deeper bedrock faults are possibly associated with some of the sub-lakefloor buckles. Trends of the popups and buckles can be grouped into six modes from 7.5° to 347.5°. Abutting and sediment onlap relationships suggest that the pop-ups formed throughout late and postglacial time following the Last Glacial Maximum ∼20,000 yr ago. The earliest set of popups is estimated to have formed before 9500 B.P.; they trend WNW, collinear with isobases of glacial rebound, and do not parallel major geophysical or structural linear zones in the region. These and other factors suggest that this set developed in response to glacial rebound-induced stress. Later popups form an irregular pattern with several orientations of axes, suggesting that the horizontal principal stress vectors were of similar magnitude. The decrease of rebound strain with time and clockwise rotation of modern contours of basin tilting relative to glacial lake isobases suggest that popups today are likely a response to reduced glacial stress combined with far-field tectonic stress.

Although the central and eastern United States is in the interior of the presumably stable North American plate, seismicity there is widespread, and its causes remain uncertain. Here, we explore the evolution of stress and strain energy in intraplate seismic zones and contrast it with that in interplate seismic zones using simple viscoelastic finite-element models. We find that large intraplate earthquakes can significantly increase Coulomb stress and strain energy in the surrounding crust. The inherited strain energy may dominate the local strain energy budget for thousands of years following main shocks, in contrast to interplate seismic zones, where strain energy is dominated by tectonic loading. We show that strain energy buildup from the 1811–1812 large events in the New Madrid seismic zone may explain some of the moderate-sized earthquakes in this region since 1812 and that the inherited strain energy is capable of producing some damaging earthquakes (M >6) today in southern Illinois and eastern Arkansas, even in the absence of local loading. Without local loading, however, the New Madrid seismic zone would have remained in a stress shadow where stress has not been fully restored from the 1811–1812 events. We also derived a Pn velocity map of the central and eastern United States using available seismic data; the results do not support the New Madrid seismic zone being a zone of thermal weakening. We simulated the long-term Coulomb stress in the central and eastern United States. The predicted high Coulomb stress concentrates near the margins of the North American tectosphere, correlating spatially with most seismicity in the central and eastern United States.

A fundamental question about continental intraplate earthquakes is why they are where they are. For example, why are the New Madrid seismic zone earthquakes concentrated on the Reelfoot rift when the continent contains many fossil structures that would seem equally likely candidates for concentrated seismicity? A key to answering this question is understanding of the thermal-mechanical structure of the seismic zone. If it is hotter and thus weaker than surrounding regions, it is likely to be a long-lived weak zone on which intraplate strain release concentrates. Alternatively, if it is not sig-nificantly hotter and weaker than its surroundings, the seismicity is likely to be a transient phenomenon that migrates among many similar fossil weak zones. These different models have important implications for the mechanics of the seismic zone, stress evolution after and between large earthquakes, and seismic hazard assessment. The sparse heat-flow data in the New Madrid area can be interpreted as supporting either hypothesis. There is a possible small elevation of heat flow in the area compared to its surroundings, depending on the New Madrid and regional averages chosen. The inferred high heat flow has been interpreted as indicating that the crust and upper mantle are significantly hotter and thus significantly weaker than surrounding areas of the central and eastern United States. In this model, the weak lower crust and mantle concentrate stress and seismicity in the upper crust. However, reanalysis of the heat flow indicates that the anomaly is either absent or much smaller (3 ± 23 rather than mW m −2 ) than assumed in the previous analyses, leading to much smaller (∼90%) temperature anomalies and essentially the same lithospheric strength. Moreover, if a small heat-flow anomaly exists, it may result from ground-water flow in the rift's fractured upper crust, rather than higher temperatures. The latter interpretation seems more consistent with studies that find low seismic velocities only in parts of the seismic zone and at shallow depths. Hence, although the question cannot be resolved without additional heat-flow data, we find no compelling case for assuming that the New Madrid seismic zone is significantly hotter and weaker than its surroundings. This result is consistent with migrating seismicity and the further possibility that the New Madrid seismic zone is shutting down, which is suggested by the small or zero motion observed geodetically. If so, the present seismicity are aftershocks of the large earthquakes of 1811 and 1812, and such large earthquakes will not recur there for a very long time.

Approximately 8000 lignite exploration cores, each 91 m (300 ft) deep, were used to map the gravel facies of the Upland Complex (Lafayette gravel) preserved on drainage divides in western Kentucky and Tennessee and on Crowley's Ridge in southeastern Missouri and eastern Arkansas. The Upland Complex is interpreted to be the remnant of a high-level terrace of the ancestral Mississippi-Ohio River system. The longitudinal profile of the Upland Complex and its projection on sea-level curves suggest that this alluvial deposit is early Pliocene in age (5.5–4.5 Ma). Sea level during the early Pliocene was +100 m, and the Upland Complex is interpreted to have been an ∼100-m-thick floodplain when initially deposited. Sea-level decline to −20 m at 4 Ma resulted in incision through the Pliocene floodplain, which formed the high-level terrace. Incision through the floodplain occurred in the Western and Eastern Lowlands of eastern Arkansas and their tributary valleys. The upper silt and sand facies of the terrace (∼60 m) were eroded, leaving the basal gravel-rich Upland Complex preserved on drainage divides. The New Madrid seismic zone lies beneath the Eastern Lowlands. There has been up to 100 m of denudation above the seismic zone in the past 4 m.y., and the most recent denudation occurred in the Holocene due to the confluence of the Mississippi and Ohio Rivers stepping north to Thebes Gap, Missouri. The late Wisconsin and Holocene denudation may have perturbed the local stress field and reactivated the New Madrid seismic zone.