Abstract

Five earthquake swarms occurred from 2007 to 2011 near Berne, New York. Each swarm consisted of four to twenty-four earthquakes ranging from M 1.0 to M 3.1. The network determinations of the focal depths ranged from 6 km to 24 km, 77% of which were ≥14 km. High-precision, relative location analysis showed that the events in the 2009 and 2011 swarms delineate NNE-SSW orientations, collinear with NNE trends established by the distribution of the spatially distinct swarms; the events in the 2010 swarm aligned WNW-ESE. Focal mechanisms determined from the largest event in the swarms include one nodal plane that strikes NNE, collinear with the distribution of the swarms and relative events within the swarms.

Two, possibly related explanations exist for the Berne earthquake swarms. (1) The swarms were caused by reactivations of proposed blind NE- and NW-striking rift structures associated with the NE-trending Scranton gravity high. These rift structures, of uncertain age (Proterozoic or Neoproterozoic/Iapetan opening), have been modeled at depths appropriate for the seismicity. (2) The NNE-trending swarms were caused by reactivations of NNE-striking faults mapped at the surface north-northeast of the earthquake swarms. Both models involve reactivation of rift-related faults, and the development of the NNE-striking surficial faults in the second model probably was guided by the blind rift faults in the first model. The Berne swarms may be evidence that these faults are seismically capable and, if so, could sustain a maximum event on the order of Mw 5.7–6.6, based on fault segment length.

INTRODUCTION

The most accurate seismic hazard assessments that can be made for a region utilize knowledge about the potentially active faults in the region, particularly, the locations of individual active faults, the maximum magnitudes that can occur on those faults, and the recurrence rates of potentially damaging earthquakes on those faults. For active plate-boundary regions like California, active faults are revealed by (1) surface faulting in areas of recent strong earthquakes, (2) geomorphic signatures of geologically recent earthquake activity, and (3) alignments of seismicity on suspected active structures. For example, Petersen et al. (2014) included information on over 2000 active faults in the western United States in their 2014 version of the U.S. National Seismic Hazard Maps. In contrast, for intraplate regions like the central and eastern United States, because information on seismically active faults is generally lacking, the hazard computations are based primarily on the locations of the small and moderate earthquakes in the region (Petersen et al., 2014). For these areas, inferences about the possible locations of future strong earthquakes and the magnitudes of those earthquakes are made from the seismicity data set alone, because direct geological information about active faults is lacking, except at a few localities, such as Mineral, Virginia (e.g., McNamara et al., 2014; Horton et al., 2015a; Shah et al., 2015), and Charlevoix, Quebec, Canada (for a review of this and other sites, see Coppersmith et al., 2012; Horton et al., 2015a). Thus, any new information about possible seismically active geologic structures in an intraplate region like the central and eastern United States is important to improve future seismic hazard assessment.

For an intraplate region like eastern North America, the rate of seismicity is generally much lower than for plate-boundary regions, and strong earthquakes associated with surface ruptures are rare. Furthermore, in most places, the routine, small earthquakes that are detected with modern regional seismic networks so far have been generally insufficient in number to delineate the local geologic structures that are seismically active (e.g., Ebel and Kafka, 1991; Coppersmith et al., 2012). For this reason, earthquake swarms in eastern North America provide a special opportunity to look for alignments of epicenters that might illuminate the local structure or structures on which the earthquakes took place. One such locality in the eastern United States is southwest of Albany, New York, in the Mohawk Valley region near the town of Berne, where five earthquake swarms were recorded by local and regional seismic network stations from 2007 to 2011 (Figs. 1 and 2; Jacobi et al., 2012; Ebel et al., 2014; Kim et al., 2016). Each swarm consisted of four to twenty-four earthquakes ranging in magnitude from 1.0 to 3.1 (Tables 1 and 2). Because so many of these events were scattered across such a small region, the relative location method of Ebel et al. (2008) can be used to compute very precise spatial distributions of the events in these swarms. Precise relative locations of the events in these swarms provide an opportunity to delineate the faults that may have been active in these earthquake swarms and to relate the geology to any lineaments or surfaces defined by the earthquake loci established by the relative location analysis. The 2011 swarm began ∼1 d before the 2011 M 5.8 Mineral, Virginia, earthquake and continued for ∼110 h after the time of the Mineral main shock (e.g., Chapman, 2013; Horton et al., 2015b), which occurred ∼600 km to the south-southwest along the general strike of the Mesozoic Atlantic-opening structures in eastern North America. Although the occurrence of the 2011 Berne swarm at the same time as the Mineral, Virginia, Mw 5.8 main shock may have been coincidental, both the Mineral main shock and the 2011 Berne swarm took place on NE-SW–oriented structures.

The purpose of this study was twofold. The first objective was to compute highly precise relative locations of the events in each local swarm in order to look for alignments of the epicenters. Relative locations of the events in the different swarms near Berne were computed, and the locations of the different swarms relative to each other also were estimated. The second purpose of this study was to compare the locations of the events and the orientations of the hypocentral alignments in the different swarms to the local geology in order to assess the local faults that may have been associated with the seismicity.

The major result of this study is that the hypocenters of the Berne area swarms are consistent in depth and alignment with presumed and modeled Proterozoic or Neoproterozoic/Iapetan-opening rift structures in the region of the northeastern termination of the Scranton gravity high (Benoit et al., 2014). The epicenters of the Berne area swarms also align with the southern extensions of Taconic normal faults, which themselves may have been guided by the older rift structures. The relationship between the Berne seismicity and known geologic structures is inherently speculative due to the midcrustal depths at which the Berne-area earthquake swarms took place.

SEISMOTECTONIC AND GEOLOGIC SETTING

The Berne earthquake swarms that are the subject of this study were centered in eastern New York State at the southern end of a broad swath of scattered seismicity that extends from the Canadian border to south of the Mohawk Valley (Fig. 1). In New York State, this swath of seismicity approximately coincides with the Adirondack dome; to the south, the seismicity band ends in the area where the Cambrian–Ordovician cover sequence above the Precambrian units of the Adirondack dome dips south beneath Silurian–Devonian rocks (Fig. 1).

Two relatively large earthquakes with determined focal mechanisms have occurred in the Adirondack dome region since 1980 (Fig. 1): the 2002 Mw 5.0 Au Sable Forks, New York, event (Seeber et al., 2002) and the 1983 Mw 4.7 Goodnow, New York, event (Seeber and Armbruster, 1986; Nabelek and Suarez, 1989; Seeber et al., 2002). The hypocenter of the Au Sable Forks event was 11 km deep, and the Goodnow event was 8 km deep. Both events had thrust focal mechanisms and occurred in the Precambrian rocks of the Adirondack dome. The rupture surfaces are estimated to dip west at 55° (Goodnow) and 45° (Au Sable; e.g., Seeber et al., 2002). Seeber et al. (2002) suggested that the N-striking rupture plane of the Au Sable Forks event was related to faults similar to the N-striking Iapetan-opening and Taconic faults located near Lake Champlain (thick, black lines west of Lake Champlain in Fig. 1), whereas the rupture surface of the Goodnow, New York, event was related to inferred faults that are marked by prominent NNE-trending topographic lineaments (labeled Taconic “normal” faults in Fig. 1) and by more localized NNW-trending lineaments in the same region. The NNE-trending lineaments likely represent structurally deeper, northerly extensions of the Taconic faults mapped in the Paleozoic rocks in the Mohawk Valley (e.g., Bradley and Kidd, 1991; Valentino et al., 2011, 2012; Jacobi, 2012).

The Berne earthquake swarms are located at the approximate northeastern termination of the Scranton gravity high, a feature in the gravity field that extends northeast from Pennsylvania into eastern New York (Fig. 2). Sources proposed to explain the Scranton gravity high include a Neoproterozoic mafic-magma rift basin/Iapetan opening rift basin (e.g., Diment et al., 1972; Rankin, 1976; Hawman and Phinney, 1992a, 1992b; Jacobi, 2002) or a Proterozoic mafic-magma rift basin (Benoit et al., 2014). A significant positive shear wave velocity anomaly in the mantle from 165 to 210 km depth (Li et al., 2003) is coincident with the Scranton gravity high and suggests that the Scranton gravity high also may be associated with mantle heterogeneities (Jacobi, 2011, 2012).

Based on Euler deconvolution of the Scranton gravity anomaly and assumed steep-sided density bodies, Benoit et al. (2014) calculated that a NE-striking mafic-magma rift structure is located in the Precambrian basement from ∼8 to 22 km deep in northeastern Pennsylvania. The proposed steep dips of the mafic-magma rift basin boundaries are supported by the spatial coincidence of the density-contrast boundaries found for varying depths by Euler deconvolution. In New York State, where the Precambrian units rises to the surface in the Mohawk Valley region, the Euler deconvolution analyses yield somewhat shallower depths for the proposed mafic-magma rift structure—in the range of 5–20 km (Benoit et al., 2014).

Benoit et al. (2014) proposed that the structures that give rise to the Scranton gravity high are truncated by an approximately NW-striking crustal shear zone located ∼0.5 km southwest of the southwestern epicenters of the Berne swarms (in Fig. 2, the shear zone is indicated by the thick, dashed, NW-trending blue line labeled “B”). Shallow, intermediate, and deep Euler deconvolution solutions for a density-contrast boundary occur in the area of the NW-striking crustal shear zone (Fig. 2), suggesting the shear zone dips steeply (Benoit et al., 2014). NW-trending topographic lineaments are observed in the same area (Fig. 2), although the topographic lineament trends are slightly divergent to the strike of the proposed crustal shear zone; one NW-trending lineament passes through the locale of the Berne swarms (Fig. 2).

Inward-facing, NE-striking boundary faults of the modeled rift basin lie along the steeper gradients of the Scranton gravity high. The location of the proposed northwestern boundary fault, based on Euler deconvolution solutions, is partially shown as a NE-trending, thick, dashed blue line labeled “A” west-southwest of the Berne swarm in Figure 2 (Benoit et al., 2014). Coincident shallow- and intermediate-depth Euler deconvolution solutions (Fig. 2) suggest that this boundary fault is steeply dipping (Benoit et al., 2014). This modeled boundary fault is coincident with the NE-trending Susquehanna River/Scranton gravity high fault (S-R/S-G-H, Fig. 2) that Jacobi (2002) proposed based on topographic lineaments, minor faults, and gravity gradients (gray zones with dashed black boundaries labeled as “S-R/S-G-H” in Fig. 2). Jacobi (2002) suggested that these lineaments reflected reactivations of the Scranton gravity high structures. The epicenters of two seismic events are located on the proposed Susquehanna River/Scranton gravity high fault system (Jacobi, 2002).

The epicenters of the Berne earthquake swarms of 2007–2011 were located on strike with, and SSW of, the Saratoga-McGregor fault, which is the easternmost NNE/NE-striking normal fault in the Mohawk Valley region (Fig. 2). The faults in the Mohawk Valley, from the Saratoga-McGregor fault west, generally have been assumed to be Taconic normal faults that accommodate the eastward deepening of the Taconic foreland basin (e.g., Bradley and Kidd, 1991). Assumed westerly dips on faults such as the Sprakers, Dolgeville, and Fonda faults result in local grabens and horsts (e.g., Bradley and Kidd, 1991). However, the actual dips on the faults are poorly constrained, the sense-of-motion on the faults is much more complicated than merely normal fault motion (many may have undergone strike-slip and thrust motion), and the motion history is not just Taconic, but ranges from Iapetan-opening to possibly postglacial times (see Appendix A for details concerning the Mohawk Valley fault systems).

No deep reflection studies have been published for the area around the Berne earthquake swarms, but a COCORP study ∼150 km north of Berne in the Adirondack dome revealed prominent subhorizontal to shallow-dipping reflectors of unknown origin between ∼18 and 26 km depth (the Tahawus complex; Klemperer et al., 1985). Other subhorizontal reflectors at 1.6–3.2 km depth may be related to the Marcy anorthosite, and more steeply dipping intermediate reflectors between the Tahawus and anorthosite reflectors were proposed to represent nappes such as those mapped at the surface (Klemperer et al., 1985). The complexity of the Precambrian structures in the Adirondack dome would suggest that it is not reasonable to extrapolate from these seismic lines 150 km south to the Berne area, although reflectors at ∼18 km depth were also recognized ∼100 km northwest of these seismic lines in the western Adirondack dome and ∼75 km southeast in the Taconic convergence zone (Klemperer et al., 1985).

A significant low-velocity zone is located in the shallow mantle (100–300 km depth) that marks thinned lithosphere beneath southern New England, called the Northern Appalachian anomaly (e.g., Levin et al., 1995, 2000; Li et al., 2003; Skryzalin et al., 2015; Pollitz and Mooney, 2016; Menke et al., 2016, 2018). The approximate, relatively sharp, steeply dipping western boundary of the low-velocity zone lies parallel to, and ∼15–30 km west of, the western thrusts on Figure 1 (and ∼10 km east of the Berne swarm; Menke et al., 2016), or lies up to 75 km farther west (∼45–55 km west of the Berne swarm; Schmandt and Lin, 2014; Pollitz and Mooney, 2016). The low-velocity zone is thought to indicate upwelling asthenosphere that is related to mantle processes functioning at the continental margin (“edge-driven upwelling”; e.g., Menke et al., 2016, 2018). Abbott and Menke (2019) proposed that geothermal springs and seismicity are linked to the edge-driven upwelling. The field of identified geothermal springs extends from southern New England into eastern New York along the Hudson River valley (Menke et al., 2018). Menke et al. (2018) suggested that He+3 anomalies such as those in the spring water at Saratoga Springs (Siegel et al., 2004) indicate that mantle sourcing is occurring along deep fracture systems that extend to the mantle and that are open because of the geodynamics of the edge-driven upwelling. Saratoga Springs is located along the Saratoga-McGregor fault (which is on-strike with the epicenters of the Berne earthquake swarms; Fig. 2).

SEISMIC DATA ANALYSIS

Earthquake Swarm Data Set

Five earthquake swarms took place in the Berne area of the Mohawk Valley from 2007 to 2011 (Table 1), and they were composed of as few as four events to as many as 24 events (Table 1), which were detected and located by regional seismic network monitoring (Table 2). It is quite possible that there were many smaller events associated with these swarms that were not detected by the regional seismic network. As is clear from Tables 1 and 2, the largest earthquake in each swarm varied from ML 2.8 to 3.1, magnitudes that were large enough to be detected widely by the regional seismic stations but well below the threshold at which damage would occur. Nevertheless, the local residents near the swarms felt or heard many of the events, and this raised some concern among the local population.

One unusual aspect of these swarms compared to other earthquake swarms that have been observed in the past in the northeastern United States (such as the 1982–1988 Moodus, Connecticut, swarms; see Ebel, 1989) is that the swarms in different time periods in Table 1 took place at different locations that were separated by a few to several kilometers. Thus, these swarms provide a rather unique data set, because not only can the relative locations of the events within each one of the swarms be computed with high precision, but also the relative locations of one swarm with respect to the other can be determined. Taken together, these two independent sets of analyses can provide a high-precision image of the full spatial distribution of the events.

The focal depths of the events in the swarms as determined from hypocenters computed using the regional seismic network data ranged from 6 km to 24 km; most are listed by the Lamont Cooperative Seismic Network as being 14 km or deeper, with a secondary mode between 8 and 11 km depth (Table 2). The focal depths of the swarm events are not well constrained from the regional seismic network data because of the sparse station coverage, with the closest seismic stations to the Berne area at ∼25 km (HCNY) and 30 km (TRY, Fig. 3). With this station distribution, the depths of the deeper hypocenters are better constrained than those of the shallower hypocenters. Ebel and Kafka (1991) noted that the depths of earthquakes in the accreted terranes along the U.S. east coast rarely exceed ∼12 km, whereas the depths of many earthquakes in the stable craton of eastern Canada can be as deep as 32 km. From the focal depths reported in Table 1, it appears that the focal depths of these earthquake swarms are more like those of cratonic North America earthquakes than those of the accreted terrane earthquakes, an observation that is consistent with the geology, since the westernmost thrusts of the Taconic orogen lie east of Berne (Figs. 1 and 2; see Macdonald et al., 2017).

Relative Location Analysis

The relative location analysis method that was used in this study was described by Ebel et al. (2008). For each swarm, an event that was well recorded by the regional seismic network stations was chosen as a master event, and the hypocenters of the other events in the swarm were located relative to the hypocenter of that master event. Because the smaller events were detected by only a few of the regional seismic network stations, with lower signal-to-noise ratios, reliable relative locations could be computed only for the larger events in each swarm (general larger than about magnitude 1.7 in Table 2). Each relative location was computed using the arrival-time differences of the P and S waves between a second event and the master event at each seismic station where those seismic waves were recorded with signals above the background noise. The arrival-time differences were found by cross-correlating the P and S waveforms of the second event with those of the master event, and only arrival-time differences from normalized cross-correlations with high coefficients (larger than ∼0.5) were used in the relative location computations. A jackknife analysis was employed to estimate the uncertainty in each relative location due to the data that were available. In the jackknife analysis, a set of relative locations was computed. One member of this set of relative locations was the original relative location computed using the data from all of the available seismic stations. Additional solutions were found by taking the original data for P and S relative arrival times, dropping the P and S readings from one of the seismic stations, and then computing a new solution with this reduced data set. The original solution plus the set of solutions computed with data from each one of the stations dropped were then accumulated and used to compute the standard deviations of the relative latitude, longitude, and depth of the master event and the second event. The standard deviations are an estimate of the uncertainties in the relative locations of the two events. Although the uncertainties in the relative locations found with the jackknife analysis varied from event pair to event pair, in general, the two standard deviation values of the uncertainties were less than ∼100 m. Thus, for most of the events, the relative locations are precise to better than 100 m with 95% confidence. The uncertainty of the absolute locations of the relative event patterns is unchanged from the average uncertainty of the events in Table 2, which is ∼2–3 km.

The locations of the seismic stations from which data were obtained for the relative location analysis are shown in Figure 3. Although only two of the stations (HCNY and TRY) are within ∼35 km from the sources, there was a good azimuthal coverage of seismic stations around the events. For this reason, the uncertainties in the event locations are comparable in the north-south and east-west directions.

The results of the relative location analyses for all but the 2007 swarms are shown in Figure 4. There were only four events in the 2007 swarm, and one of the events was too small to give a reliable relative location. On the other hand, the other swarms had a number of events for which precise relative locations could be computed. The four plots in Figure 4 show the locations of events relative to four different master events as a function of epicenter and time. Several points can be noted from these plots. First, the events in the two swarms in 2009 and in the 2011 swarm trend in a NE-SW orientation, whereas the events in the 2010 swarm align in the WNW-ESE orientation. Second, the events in the February 2009 and February 2010 swarms spread bilaterally from the first event in the swarm, whereas the events in the April–December 2010 and the August 2011 swarms spread unilaterally from the location of the first event in the swarm. Third, the events in each swarm spanned a total length of less than 1 km, so each swarm was very localized in spatial extent. It is the very precise relative arrival times from the cross-correlation analyses that allow such fine detail in the spatial patterns of the swarms to be extracted in the determinations of the relative event locations. Fourth, the relative focal depths in each of the four panels of Figure 4 do not yield a reliable indicator of rupture plane dip.

The results in Figure 4 come from relative location analyses of each individual swarm. A separate relative location analysis was carried out using the master events of each of the swarms in an attempt to precisely locate the position of one swarm relative to that of the other swarms. The event on 18 February 2010 at 14:20 was used as the master event for this analysis. The results of this analysis are given in Table 3 and are shown in Figure 5. Approximate latitudes and longitudes are indicated in Figure 5 under the assumption that the master event for the 2010 swarm was at the absolute location for this event given by the Lamont Cooperative Seismic Network in Table 2. It is clear that the 2011 swarm was resolvably to the northeast of the master event, as was the 22 March 2009 09:21 event. The 24 July 2007 12:56 event was northwest of the master event in this analysis, although its location is approximately at the edge of the 95% uncertainty interval in the relative location analysis.

The relative location analyses carried out on these swarms did not provide a direct constraint on the absolute depths for the events, although they did constrain the relative depths. The absolute locations of the swarm events determined by the Lamont Cooperative Seismic Network (http://almaty.ldeo.columbia.edu:8080/data.search.html) show focal depths of 13–18 km for the 2007 swarm, 6–16 km for the two swarms in 2009 (with most between 7 and 10 km depth and another group at 14–16 km depth), 17–22 km (except one at 14 km) for the 2010 swarm, and 18–24 km (with most 20 km or deeper) for the 2011 swarm (Table 2). In contrast, our relative location analysis found that the relative depths of determinable events within each swarm varied by <200 m, and the relative depths between swarms varied by <4.1 km (Figs. 4 and 5). The relative location results in Table 3 indicate that the 2011 swarm was 2–4 km deeper than the events in the other swarms, and this depth difference is well resolved by the relative location analysis. Figure 5 shows the orientations of the distributions of the events in each swarm along with an estimate of the absolute depths of the master events based on the average absolute depth of the events in the 2010 swarm from Table 2 and the relative depths of Figure 5. The approximate epicentral alignment of events within each swarm is similar to the general trends (NE and WNW) described by the loci of swarms (Fig. 5). If the 2009 and 2011 events occurred on the same slip surface, then that surface dips WNW.

The improvement in the event locations based on the relative location analysis (shown in Figs. 4 and 5) compared to the absolute event locations computed by the Lamont Cooperative Seismic Network is illustrated in Figure 6. Figure 6A displays the locations of the swarm earthquakes computed by the Lamont Cooperative Seismic Network, where an independent absolute location was computed for each earthquake. The absolute locations of these events as computed by the Lamont Cooperative Seismic Network are scattered in two general areas, one to the northeast and one to the south of Berne, with the event epicenters in each area suggesting NW-SE trends. Figure 6B shows the locations of the swarm earthquakes from the relative location analysis, where the relocated events from Figure 5 are displayed on the same base map as in Figure 6A. In this case, all of the swarms except for that in 2010 show NE-SW alignments that roughly parallel or align with the major structural features observed at the surface. The WNW-ESE trend of the 2010 swarm also is subparallel to a surface lineament.

Focal Mechanism Analysis

Only one of the swarm events, that of 24 July 2007 with ML 3.1, was recorded sufficiently well by seismic stations in the northeastern United States and nearby Canada that a first-motion focal mechanism could be computed. A range of focal mechanisms was found compatible with the first-motion data. All of the focal mechanism solutions that are consistent with the first-motion data for the 24 July 2007 event have P-axis orientations (i.e., SHmax) between about N65°E and N80°E. Two mechanisms that represent the end members of the range of possible focal mechanisms are shown in Figure 1. This earthquake was an oblique-slip event with some normal faulting component. The left-lateral strike-slip component and P-axis orientation are consistent with the generally ENE-directed SHmax in New York State and Pennsylvania west of the Hudson River (including one ENE- and one NE-directed SHmax ∼45 km south of the Berne events; e.g., Zoback and Zoback, 1989; Hurd and Zoback, 2012).

There are only a few other published focal mechanisms for earthquakes in northeastern New York State. Most of those solutions, e.g., the 2002 Au Sable Forks, New York, event and the 1983 Goodnow, New York, event in the Adirondack dome, have a strong thrust component (SHmax = S1; see Fig. 1; Hurd and Zoback, 2012). For the 24 July 2007 event, one nodal plane strikes WNW/NW and dips steeply, and the other nodal plane strikes NE (approximately N42°E) and has a dip to the NW ranging from ∼40° to 70°, with most solutions dipping at ∼45°. The NE-striking nodal plane is parallel to the possible rupture plane defined by the relative location plots in Figure 4 (e.g., the August 2011 earthquake swarm defines a rupture plane that strikes N42°E). In contrast, the NW-striking nodal planes (at about N48°W) are not parallel to the WNW rupture plane (at N77°W) defined by the February 2010 relative epicentral locations (Fig. 4). The NW strike of the nodal planes is similar, however, to the trend of the NW lineaments. If the NE-striking nodal planes with a westerly dip are the correct solution, then the westerly dipping solution is consistent with the possibility that the 2009 and 2011 swarm events were related to the same NW-dipping zone of slip surfaces, although most nodal plane solutions have steeper dips than the inferred dip from the 2009 and 2011 swarms.

An earthquake of MLg 2.6 on 25 August 2013 occurred near Glen Falls, New York (43.351°N, 73.794°E; Fig. 1), for which a well-constrained focal mechanism was computed for this study due to the operation of the EarthScope Transportable Array seismic stations in the area at the time of the event. That focal mechanism also shows oblique strike-slip motion with a steeply west-dipping, NNE-striking nodal plane; the N28°E strike of the nodal plane is very close to the N30°E strike of a 6-km-long segment of the nearby Saratoga-McGregor fault mapped in detail by Geraghty and Isachsen (1980) and Tice (1993).

RELATIONSHIP WITH LOCAL GEOLOGY

The Berne earthquake swarms are located at the approximate northeastern termination of the NE-trending Scranton gravity high (Fig. 2). In one possible explanation for these earthquakes, the Berne seismicity swarms are related to motion on rift basin boundary faults that are defined by modeled sources for the Scranton gravity high. The depths of the rift structures have been calculated to range from ∼8 to 22 km in northeastern Pennsylvania and from ∼5 to 20 km in the Mohawk Valley region, based on Euler deconvolution analyses of the Bouguer gravity anomaly by Benoit et al. (2014). Significantly, these depths are similar to the range of hypocentral depths of the Berne earthquake swarms: 6 km to 24 km, with a primary mode at 20 km and secondary modes at 10 km and 14–15 km as listed by the Lamont Cooperative Seismic Network (Table 2).

The 2010 Berne swarm is ∼3 km NE of a crustal shear zone proposed by Benoit et al. (2014) that terminates the northeastern extension of the Scranton gravity high. The hypocenters of the 2010 swarm are midcrustal, ranging from 14 km to 22 km, with a mode at 18–19 km (Table 2). These depths are similar to the deeper portions of the modeled rift basin (Benoit et al., 2014). The 2010 swarm defines a rupture surface at depth that is oriented roughly WNW-ESE (Fig. 4) and that is nearly coincident with a WNW-trending topographic lineament (Fig. 6B). The fact that the rupture surface determined from the 2010 earthquake swarm is so close in depth, orientation, and location to the lineament and to the modeled approximate termination of the rift structures suggests that the 2010 earthquake swarm resulted from reactivation of the cross-strike crustal shear zone that marks the northeastern termination of the rift graben proposed by Benoit et al. (2014).

The 2007, 2009, and 2011 Berne earthquake swarms describe surfaces that strike northeast (Figs. 4 and 6B), similar to the strikes and depths of the proposed rift basin boundary faults (Fig. 2). However, the epicenters are not precisely coincident with, or on strike with, the locations of the proposed boundary faults. The swarms are located immediately northeast of the proposed NW-striking, steeply dipping crustal shear zone discussed above (Fig. 2). If isolated rift faults extend NNE beyond the proposed NW/WNW-striking crustal shear zone (or stepover), the seismicity could represent reactivation of these isolated Proterozoic/Neoproterozoic faults.

If the 2009 and 2011 Berne earthquakes swarms developed on the same general slip surface, then that surface dips west (Fig. 5). A westerly dip is consistent with the westerly dipping nodal plane solution of the Berne earthquake (Fig. 1), but the westerly dip inferred from the swarms appears unrealistically low, on the order of 26° to 37°, based on the values in Figures 4 and 5 and Tables 1 and 2. If the proposed west-dipping fault is correct, and the fault initiated in the Proterozoic/Neoproterozoic, then the fault could be related to the southeastern boundary rift fault by a stepover in the location of Benoit et al.’s (2014) crustal shear. However, if the midcrustal seismicity analyzed in this study is aligned with the surface lineaments and is causally related to them, then the structures in this model must be essentially vertical, as also proposed by Benoit et al. (2014).

The seismically active, NW/WNW-striking fault that is suborthogonal to the main NNE-striking faults is similar to those noted by Kafka et al. (1985) and Seeber et al. (2002) for faults in New Jersey, where seismically active faults strike at a high angle to the main Mesozoic-opening faults (Seeber et al., 1998). The fact that the Berne earthquake swarms are localized in a region where NW/WNW-striking cross-faults related to the termination of the Scranton gravity high rift structures are proposed to intersect NNE-striking rift-related boundary faults suggests that crustal weakness, local stress rotations, and local stress concentrations related to the intersecting major fault systems may have promoted and localized the Berne earthquake swarms. Such relationships among structural complexity, local stress deviations, and elevated seismicity have been proposed elsewhere (e.g., Talwani, 1999; Li et al., 2009; Stein et al., 2009), including the Mineral, Virginia, seismicity (e.g., Shah et al., 2015).

In the second alternative model that relates the Berne earthquakes to local geologic structures, the seismicity swarms that have a NNE trend are related to motion on deeper segments of the NNE-striking faults in the Mohawk Valley region northeast of the Scranton gravity high (Fig. 2). The general NE trend of the epicenters in each earthquake swarm (except for the WNW trend of the 2010 swarm) is consistent with the strikes of most of the faults in the Mohawk Valley, and, more specifically, is approximately on-strike with the Saratoga-McGregor fault. If the Saratoga-McGregor fault is extended along strike to the SSW from its proposed (not observed) southerly limit, then the surface expression of this fault would pass within 2 km of the epicenters of the 2009 swarms and within 4 km of the epicenters of the 2007 and 2011 swarms (Fig. 6B). Additionally, the epicenters of the 2011 swarm are also within 2 km of a meter-wide fracture-intensification domain (FID) that outcrops northwest of Altamont, New York; the FID has NE-striking, vertical fractures with minimal throw (Fig. 6B; see Appendix A; Jacobi et al., 2012). However, since the focal depth estimates of the relocated Berne earthquake swarms are estimated in this study to be at depths between 19 km and 24 km, these earthquake swarms occurred well within the Precambrian basement, not the Paleozoic section, because Precambrian basement is only ∼1.2 km (4000 ft) below sea level in the Berne region (e.g., Domrois et al., 2015).

The earthquake swarms could have occurred on the Saratoga-McGregor fault (and associated splays such as that inferred from the fracture-intensification domain) only if the Saratoga-McGregor fault system is essentially vertical. A steep dip can be inferred from the Euler deconvolutions of Bouguer gravity anomalies, which show coincident 6 and 12 km depth solutions along the Saratoga-McGregor fault northeast from the Berne swarm (Fig. 2). A steep dip of the rupture surface(s) also might be inferred from the narrow band of epicenters in individual swarms, given the significant variation in absolute hypocenter depths from the Lamont earthquake catalog. However, our relative location analysis indicates that all but the 2011 swarm took place within ∼2.0 km of the same focal depth.

Southwest of the Berne swarms, three-dimensional (3-D) seismic data indicate that the NNE/NE-striking Mohawk Valley faults generally have very steep dips in the Paleozoic section, but there is no indication of fault dips in the Precambrian basement in the seismic data (Jacobi, 2011, 2012). However, there are some suggestions that the deep “keels” of the Mohawk Valley faults in the Precambrian units do dip steeply. (1) The NNE-trending lineaments in the Adirondack dome that are inferred to be the structurally deeper counterparts of the Mohawk Valley faults in the Paleozoic section are fairly straight across the dome, even though the dome presumably exposes deeper erosion levels in the middle of the dome (Bradley and Kidd, 1991; Valentino et al., 2011, 2012; Jacobi, 2012). (2) A mesoscale fault located on the probable northward extension of the Little Falls fault in the Precambrian section of the Adirondack dome is steeply dipping (82°SE; Valentino et al., 2012).

A steeply dipping fault extending to 24 km depth (about the maximum depth of the Berne earthquakes) would suggest a strike-slip fault, but the core raised on a splay of the Saratoga-McGregor fault (star labeled “Core” in Fig. 2) did not display any strike-slip kinematic indicators, only dip-slip motions (see Appendix A; Hanson et al., 2010). Other northerly striking Mohawk Valley faults do exhibit oblique-slip and strike-slip kinematic indicators, and map patterns suggest that some of the faults in the Mohawk Valley were subjected to strike-slip motion at some time in their past (e.g., Hoffmans fault; see Appendix A; Jacobi et al., 2015).

In contrast, if the Saratoga-McGregor fault and similar faults to the west do have intermediate dips, on the order of 60°, such as the dip of the surface expression of the Hoffmans fault near the Mohawk River (e.g., Jacobi et al., 2016; Mitchell and Jacobi, 2016; see also Appendix A), then the NNE-striking earthquake swarms would have occurred on a fault that outcrops west of the Saratoga-McGregor fault, if the fault dips east (such as the Hoffmans fault or splays of the Saratoga-McGregor fault), or east of the Saratoga-McGregor fault, if the fault dips west. It may be that intermediate dips characterize the Mohawk Valley faults in the Paleozoic section, and steep dips characterize the faults in Precambrian basement; essentially, the fault systems are flower structures. In that case, given the highly inexact location of the Saratoga-McGregor fault trace immediately north-northeast of the Berne swarms (Fig. 2), it is likely that the Saratoga-McGregor fault system is the fault system associated with the Berne earthquake swarms.

The deep earthquakes of the Berne swarms may have occurred near the intersection of the NNE-striking faults with a master detachment fault such as that proposed for Taconic times in the Mohawk Valley by Bradley and Kidd (1991). If the 18-km-deep seismic reflectors recognized under both the Taconics and the Adirondacks represent the same structure (an observation broached by Klemperer et al., 1985), then there is a possibility that this structure is a Mesozoic rift detachment. In that case, the Berne swarms probably occurred on a Mesozoic fault system that consisted in part of the reactivated NNE-striking Iapetan-opening/Taconic fault system. Such a model is consistent with Wheeler’s (1995) suggestion that thinned continental crust from the Mesozoic opening extends west as far as western New York State. The 2011 M 5.8 Mineral, Virginia, earthquake, which occurred during the 2011 Berne swarm, and the Mineral, Virginia, aftershocks, were also proposed to have occurred along several Paleozoic and perhaps Mesozoic reactivated fault systems (e.g., Horton et al., 2015b). The intersection of normal faults and the master detachment has been proposed as the source for seismic activity in active orogenic belts (e.g., Greece; Kapetanidis et al., 2015).

It is probable that the second model of reactivation of NNE-striking faults in the Mohawk Valley (with primarily Taconic throw) is not independent of the first model, which involves reactivation of rift structures associated with the Scranton gravity high. The faults with significant Taconic throw in the Mohawk Valley region were probably guided by Proterozoic/Neoproterozoic rift structures, based on seismic reflection data both to the north and west-southwest of the Berne area, as discussed previously (e.g., Jacobi, 2011; Séjourné et al., 2002). Furthermore, selected faults of both the Scranton rift faults and the faults in the Mohawk Valley may have been reactivated during the present Atlantic-opening cycle. The first model, which postulates motion on rift faults associated with the Scranton gravity high, is attractive because the hypocentral depths closely match the modeled depths of the rift structures and because the swarms appear to have been located at the intersection of rift boundary faults with cross-strike faults that mark either stepovers or a crustal shear zone. It has been proposed that fault intersections can be locales of weakened crust and stress concentrations and therefore can be preferential places where earthquakes take place in intraplate settings (e.g., Talwani, 1999; Li et al., 2009). However, the NNE-striking rupture surfaces of the Berne earthquake swarms were located northeast of the proposed crustal shear zone that is thought to terminate the Scranton gravity high rift structure (Benoit et al., 2014). Either other rift structures extend northeast from the proposed termination of the rift structure associated with the main body of the gravity high (as suggested by Euler deconvolutions of Bouguer gravity anomalies east and northeast of the Berne swarm conducted by Benoit et al. [2014]; Fig. 2), and these rift structures guided later Taconic fault development (such as the Saratoga-McGregor fault), or the proposed location of the termination of the Scranton rift is inexact, as would be expected in gravity modeling. In the case where the Scranton rift structures extend farther northeast than inferred by Benoit et al. (2014), the first and second models that we have proposed to explain the Berne swarms become inextricably linked.

DISCUSSION

The possible association of the Berne earthquake swarms with the Saratoga-McGregor fault and/or parallel normal faults raises the question of whether the Saratoga-McGregor fault and other northerly striking faults in the Mohawk Valley that extend north into the Adirondack dome and the lowlands to the east are associated with modern seismicity. This relationship is explored in Figure 1, where the instrumental seismicity of the Adirondack dome and nearby areas is plotted along with the major mapped faults.

The three published focal mechanism solutions in the Adirondack region all have rupture planes that are parallel to NNE- and N- striking faults that are northward extensions of the faults in the Mohawk Valley. The focal mechanism of the MLg 2.6, 25 August 2013 earthquake near Glen Falls, New York (Fig. 1), has a NNE-striking nodal plane that is within 2° of the strike of a 6-km-long mapped segment of the nearby Saratoga-McGregor fault. The Mw 5.0 Au Sable Forks earthquake of 20 April 2002 (http://www.ldeo.columbia.edu/LCSN/NYQuake_2002/hrv.cmt.txt; Seeber et al., 2002) had a pure thrust focal mechanism with nodal planes that strike just east of N, parallel to nearby faults, the easternmost of which links southward to the Saratoga-McGregor and associated faults (Fig. 1; Seeber et al., 2002). The Mb 5.1 Goodnow, New York, earthquake of 7 October 1983 (Nabelek and Suarez, 1989; Seeber et al., 2002) had a thrust focal mechanism with northerly striking nodal plane orientations (Fig. 1). Although many of the prominent lineaments and proposed faults trend NNE in the Adirondack dome, N-trending lineaments (and assumed faults) occur in the immediate region of the Goodnow event (Seeber et al., 2002). It thus appears that the N- and NNE-striking Mohawk Valley faults that extend north through and east of the Adirondack dome are seismically capable.

Detailed mapping of the Saratoga-McGregor fault and the parallel Hoffmans fault shows that those faults consist of a series of linked, relatively short fault segments as suggested by the fault map patterns (e.g., Fisher, 1980; Geraghty and Isachsen, 1980; Tice, 1993; Jacobi et al., 2016). The seismic capability of the shorter segments of the Saratoga-McGregor fault (on the order of 2.5 km) is about Mw 5.7, i.e., about the same magnitude as the 2011 Mineral, Virginia, event, and the seismic capability of mapped longer segments, on the order of 10 km, is about M 6.3, assuming accepted fault length versus magnitude relationships (e.g., Wells and Coppersmith, 1994). The 12 km distance between the February 2009 swarm and the 2011 swarm is about the same as the longer mapped segments of the Saratoga-McGregor fault, if the two swarms lie on a single fault system.

If a Mohawk Valley fault like the Saratoga-McGregor fault is capable of generating a Mw ≥ 6 seismic event, then this capability has implications for the seismic hazard of the northeastern United States and southeastern Canada. On the other hand, there is no information on the past earthquake history of this fault, nor is there any information on what might be the possible repeat times of strong earthquakes on the fault, if indeed it is seismically active. Tice (1993) carried out an investigation to look for paleoseismic indicators of recent faulting on the Saratoga-McGregor fault, but he found no evidence for Holocene fault movement at the site of his investigation. Further, Abbott and Menke (2019) suggested that aseismic zones may occur within 10 km of thermal springs, such as those at Saratoga Springs.

As argued above, the primary association of the Berne swarms may well be with rift structures related to the Scranton gravity high, rather than to faults exposed in the Mohawk Valley and the Adirondack dome. In that case, we have little knowledge about the length of the faults or their slip potential, because those faults are not exposed at the surface. The NW-striking crustal shear zone, or stepover, that truncates the rift and gravity high in the region of Berne is estimated to be on the order of 90 km long (Fig. 2). If the shear zone is partitioned into smaller fault segments, as is common elsewhere and as can be inferred from jogs in the gravity gradients (Fig. 2), then the estimated length of the fault segments (20 km) would be capable of hosting an Mw 6.6 event, based on the scaling laws of Wells and Coppersmith (1994). Similarly, we can estimate the length of NE-striking rift fault segments from jogs in the gravity gradient along the long axis of the Scranton gravity high and from jogs in the structural trends based on the Euler deconvolution solutions of Benoit et al. (2014, their fig. 6). These lengths are generally ∼20 km long, which suggests that each of these fault segments, if active, may be capable of generating an Mw 6.6 event, again based on the length scaling laws of Wells and Coppersmith (1994).

It may well be that the NNE-striking faults in the Mohawk Valley are related to the rift faults of the Scranton gravity high and that the Berne earthquake swarms are related to the intersection of the faults in the Mohawk Valley with the rift faults and cross-faults of the Scranton gravity high. In that case, the Berne swarms may be localized at the intersections of these different basement structures due to local stress concentrations and weakened crust there, and the faults in the Mohawk Valley to the NE away from intersections of structures may not be associated with modern seismicity represented by the Berne swarms.

The ultimate source of the stresses that caused the fault reactivations of the Berne swarms is beyond the scope of this report, but it may be linked to two alternative models: (1) geodynamics of the asthenospheric edge-driven upwelling centered in southern New England (Menke et al., 2016, 2018), or (2) adjustments of the slightly undercompensated mafic bodies modeled for the Scranton gravity high (Benoit et al., 2014) in response to the present stress field.

In the first alternative model, the imprecise steep, western boundary of the Northern Appalachian anomaly has been proposed to lie ∼10 km east of the Berne swarms, based on mantle tomographic images (Menke et al., 2016, 2018). Geothermal springs along the Hudson River valley and the He+3 anomaly in the waters at Saratoga Springs are consistent with this proposal (Menke et al., 2016, 2018). Since Menke et al. (2018) suggested that the He+3 anomaly indicates that the asthenospheric edge-driven upwelling is promoting open fractures from the surface into the mantle to conduct He+3 to the surface, that fracture system (the Saratoga-McGregor fault) must be responding to stresses set up by the edge-driven upwelling. Countering this argument are the SHmax measurements ∼45 km to the south of the Berne swarm that have ENE/NE-trending SHmax orientations typical for the eastern continental interior of New York and Pennsylvania rather than the dispersed orientations typical of New England and the Northern Appalachian anomaly (Zoback and Zoback, 1989). Nevertheless, if Menke et al. (2018) were correct concerning the significance of the He+3 anomalies, then, to the extent that the seismicity in southern New England is driven by the geodynamics of the proposed asthenospheric edge-driven upwelling (e.g., Abbott and Menke, 2019), fault reactivation in easternmost New York may also be responding to these same dynamics. The second alternative model of slight undercompensation could be an additive effect to the stress field set up by the asthenospheric edge-driven upwelling, or the model could be the driving mechanism.

CONCLUSIONS

Five recent earthquake swarms were recorded by local and regional seismic network stations from 2007 to 2011 near Berne, New York, west-southwest of Albany, New York. Each swarm consisted of four to 24 earthquakes ranging in magnitude from 1.0 to 3.1. The focal depths of the events reported in regional seismic networks ranged from 6 km to 24 km, although most were 14 km or deeper, with a secondary mode between 7 and 10 km depth. The 2011 swarm began ∼1 d before the 2011 M 5.8 Mineral, Virginia, earthquake and continued for ∼110 h after the time of the Mineral main shock.

For each swarm, an event that was well recorded by the regional seismic network stations was chosen as a master event, and the hypocenters of the other events in the swarm were located relative to the hypocenter of that master event. A jackknife analysis was used to estimate the uncertainties in the relative locations, which were found to have precisions better than 100 m with 95% confidence. The events in the two swarms in 2009 and in the 2011 swarm trend NE-SW, whereas the events in the 2010 swarm align WNW-ESE. The events in each swarm span a total length of less than 1 km.

Two broad, possibly related, explanations relate the Berne earthquake swarms to the local geology. The depth range of the earthquakes from the absolute locations computed using the regional network stations (6 km to 24 km with 77% 14 km or deeper) and the orientations of the rupture surfaces (NE- and WNW-striking) are consistent with the depths and strikes of faults of a Proterozoic or Neoproterozoic rift basin that has been proposed to lie beneath the Scranton gravity high. These faults are inferred based on modeling of a Bouguer gravity high that is centered in the Scranton, Pennsylvania, area (Benoit et al., 2014). The NE-striking rift faults are terminated by a NW-striking crustal shear zone or stepover in the area of the Berne earthquake swarm.

The second alternative is that the Berne earthquake swarms are related to NNE-striking faults and WNW-striking cross-faults similar to those that have been mapped in the Mohawk Valley region, northeast of the proposed rift structure of the Scranton gravity high. Although a southwest on-strike extension of the surface expression of the Saratoga-McGregor fault would pass within a few kilometers of the epicenters of the Berne earthquake swarms, the midcrustal hypocenters of the earthquakes in the swarms do not permit a direct connection between the epicenters and the on-strike surface features unless those features are nearly vertical. A focal mechanism and the spatial distribution of the swarms might suggest that much of the Berne seismicity took place on fault surfaces that dip WNW, precluding a direct relationship between the swarms and the presumed E-dipping Saratoga-McGregor fault. The rupture surfaces of the swarms might align with other faults mapped at the surface in the Mohawk Valley.

If the swarms were localized where WNW- and NNE-striking rift system faults of the Scranton gravity high intersect, then it may be that the swarms are not indicators of potential seismicity along the NNE-striking faults that extend NNE from Berne through the Mohawk Valley into the Adirondack dome and Lake Champlain Valley. In contrast, if the seismicity swarms were related to the faults in the Mohawk Valley, then coupled with previous and new focal mechanism studies to the north, the Berne swarms may be evidence that these faults are seismically capable and could sustain events on the order of at least Mw 5.7. If the Berne earthquake swarms are related to the rift structures of the Scranton gravity high, then, based on fault length/magnitude relationships, the rift faults, if active, could be capable of generating about an Mw 6.6 event in localized areas of the rift structures, especially the broken crustal regions where cross-faults intersect the NE-striking rift faults. The ultimate source of the stresses that caused the fault reactivations of the Berne swarms may be linked to geodynamics of the proposed asthenospheric edge-driven upwelling centered in southern New England.

APPENDIX A: MOHAWK VALLEY NORMAL FAULTS

The dominant faults in the Mohawk Valley, from the Saratoga-McGregor fault west (Fig. 2), strike NNE/NE and generally have been assumed to be Taconic normal faults (e.g., Bradley and Kidd, 1991). However, few damage zones and slip surfaces of faults in the Mohawk Valley are actually exposed, precluding definitive knowledge of fault dip and motion. Single-outcrop exceptions for some faults include the Little Falls fault (“L-F” in Fig. 2), which dips east in exposures along the New York State Thruway (Jacobi et al., 2005), the Dolgeville fault (“Do” in Fig. 2), which displays a vertical fault breccia (Jacobi et al., 2005), the Manheim fault (“Man” in Fig. 2), which has a nearly vertical fault breccia as well as a mine adit along the fault that dips ∼77°E (e.g., Bradley and Kidd, 1991), a splay of the Noses fault (“No” in Fig. 2) with a very steep possible fault surface, and the Hoffmans fault (“Ho” in Fig. 2), which has parallel minor vein-fill faults that dip ∼60°E and an erosional notch up a hillside from which an easterly dip can be inferred (Jacobi et al., 2016; Mitchell and Jacobi, 2016).

A splay of the Saratoga-McGregor fault appears to have had primarily dip-slip (normal and reverse) motion on fault surfaces dipping between 35° and 60°, based on kinematic indicators in an unoriented core of the Ordovician Utica Group at the fault splay (core site is the star labeled “Core” in Fig. 2; Hanson et al., 2010). The kinematic indicators include slickenfibers, restraining/releasing bends in calcite veins, and laminae offsets. Estimates of the cumulative down-to-the-east stratigraphic throw of the Saratoga-McGregor fault vary from ∼150 m (Bradley and Kidd, 1991) to more than 730 m (Bosworth and Putman, 1986).

In contrast to the 35°–60° fault dips in the Utica core on the Saratoga-McGregor fault splay, a meter-wide fracture-intensification domain in the Upper Ordovician Schenectady Formation displays vertical fractures that strike NNE, parallel to the assumed Saratoga-McGregor fault strike (outcrop site is labeled “outcrop” in Fig. 6; Jacobi et al., 2012). Some of the fractures in the fracture-intensification domain display very modest offset (<2 cm). The fracture-intensification domain exposure is ∼3 km northwest of the proposed southerly on-strike extension of the Saratoga-McGregor fault (Fig. 6). The fracture-intensification domain is located ∼2 km northeast of the 2011 Berne earthquake swarm. These types of fracture-intensification domains have been found to be associated with fault systems in other parts of the Appalachian basin in New York State (e.g., Jacobi, 2002; O’Hara et al., 2017). It is probable that splays of the Saratoga-McGregor fault are characterized by such fracture-intensification domains in this area and at this stratigraphic interval.

Although the NNE/NE-striking faults did sustain significant normal-fault motion during the Taconic orogeny, oblique and strike-slip Taconic motion has also been proposed (Jacobi, 2011; Valentino et al., 2011, 2012; Jacobi et al., 2015, 2018). Taconic strike-slip motion was inferred from rhombochasms that were active in basal Utica time (Ordovician), based on the timing of basin infill observed in 3-D seismic surveys west-southwest of the Mohawk Valley (Jacobi, 2011; Jacobi et al., 2015, 2018). A Taconic age for a proposed rhombochasm associated with the tail of the Hoffman’s fault (northeast of label “Ho” in Fig. 2) was also suggested based on Trenton-age breccias found in the rhombochasm (Jacobi et al., 2016). Jacobi et al. (2015, 2018) also proposed a rhombochasm origin for small grabens with Cambrian–Ordovician sediments in the Precambrian Adirondack dome; these grabens are associated with lineaments that extend south to Taconic-aged faults in the Mohawk Valley, and the geometry of the lineaments at the grabens implies a component of left-lateral motion. A component of left-lateral strike-slip motion was also inferred from offsets of Precambrian structural patterns along NNE-striking faults in the Adirondack dome and from inferred fault geometries based on magnetic models of the blind Piseco Lake graben in the Adirondack dome (Valentino et al., 2011, 2012), but Valentino et al. (2012) suggested that the strike-slip component of motion was related to Cretaceous uplift of the Adirondack dome. Oblique and horizontally plunging slickenfibers on steeply dipping fault surfaces observed in outcrop along the Hoffmans fault (e.g., Hrywnak et al., 2014; Jacobi et al., 2015) and in core acquired between the Hoffmans and Fonda faults have unknown ages, but they are suspected to record multiple events, based on crosscutting and abutting relationships (Hrywnak, et al., 2014; Jacobi et al., 2015; Schweigel et al., 2018). The faults in the Mohawk Valley are proposed to have undergone reverse slip in the latest Taconic (e.g., Kidd et al., 1995; Jacobi, 2011; Jacobi et al., 2016).

The faults in the Mohawk Valley region were presumed to be of only Taconic age, since the Ordovician-Silurian contact south of the mapped faults does not appear to be significantly offset (e.g., Fisher, 1980; Fig. 2). Additionally, the poor outcrop and imprecise stratigraphy in the Upper Ordovician shales and coarser clastics south of the mapped faults in the Mohawk Valley do not promote fault recognition, even if faults did exist in these units (e.g., Fisher, 1980; Fig. 2). However, recent research reviewed below has shown that the fault-motion history of the Mohawk Valley faults, including the Saratoga-McGregor fault, is considerably more complicated than the original assumption that fault motion was restricted to Taconic times.

The NNE/NE-striking faults in the Mohawk Valley region were probably initially Iapetan-opening faults, based on growth-fault geometries observed in 2-D and 3-D seismic data to the west-southwest and lineament studies that appear to connect those faults observed in the seismic data with the faults in the Mohawk Valley (Jacobi, 2002, 2011, 2012). Similarly, growth-fault geometries of faults observed in 2-D seismic data shot to the north in Quebec indicate that the Taconic normal faults there too initiated as Iapetan-opening faults (e.g., Séjourné et al., 2002). Iapetan-opening motion is also thought to be recorded by mylonitic foliation along the Saratoga-McGregor fault north of Saratoga Springs (Bosworth and Putman, 1986).

The faults in the Mohawk Valley region were reactivated during the Silurian, as inferred from Silurian unit pinchouts at the southern extensions of the faults that were extrapolated along lineaments (Jacobi and Smith, 2000) and from proprietary 3-D seismic images to the southwest that display Silurian growth-fault geometries across the faults (e.g., Jacobi, 2011). Lineaments and proprietary 3-D seismic images also indicated that the faults were reactivated (with small throws) after the Devonian Catskill delta complex was deposited, suggesting Neoacadian and/or Alleghanian motion (e.g., Jacobi and Smith, 2000; Jacobi, 2011). Differences in model time-temperature histories based on apatite fission-track ages for sites across NNE-striking faults in the southeastern Adirondacks suggest Cretaceous and Eocene reactivation of the faults during uplift of the Adirondack dome (Roden-Tice et al., 2000). Recent (postglacial) motion on the Saratoga-McGregor fault has been proposed based on the nonbeveled, nonincised fault scarps (Bosworth and Putman, 1986; Tice, 1993). Earthquake epicenters located near some of the faults in the Mohawk Valley, such as the Herkimer and Sprakers faults, suggest that this system of faults may be active today (e.g., Jacobi, 2002).

The normal faults in the Mohawk Valley region were thought to be related to plate flexure and plate subsidence as the Laurentian plate entered the subduction zone in Taconic eastward subduction zone models (e.g., Jacobi, 1981; Rowley and Kidd, 1981; Bradley and Kidd, 1991). However, in the most recent tectonic models, the faults underwent Late Ordovician motion related to retro-arc foreland basin subsidence (e.g., Macdonald et al., 2014, 2017; Karabinos et al., 2017; Jacobi and Mitchell, 2018), although limited eastward subduction during final continent-arc collision also may have occurred.

ACKNOWLEDGMENTS

We would like to thank our students, who assisted in many facets of this project, including Alex O’Hara, with whom Jacobi worked in the field. The Mohawk Valley geology studies were partially funded by New York State Energy Research and Development Authority (NYSERDA) grants to Jacobi. Jacobi thanks William Kidd, John Martin, Charles Mitchell, Rich Nyahay, and Dave Valentino for the many years of interesting discussions concerning Mohawk Valley geology. We also thank Bill Stephenson and two additional, anonymous reviewers, whose suggestions considerably improved the original manuscript.

Gold Open Access: This paper is published under the terms of the CC-BY-NC license.