Estimating crustal thickness using SsPmp in regions covered by low-velocity sediments: Imaging the Moho beneath the Southeastern Suture of the Appalachian Margin Experiment (SESAME) array, SE Atlantic Coastal Plain
Abstract
Deconvolved waveforms for two earthquakes (Mw: 6.0 and 5.8) show clear postcritical SsPmp arrivals for broadband stations deployed across the coastal plain of Georgia, allowing mapping of crustal thickness in spite of strong reverberations generated by low-velocity sediments. Precritical SsPmp arrivals are also identified. For a basement in which velocity increases linearly with depth, a bootstrapped grid search suggests an average basement velocity of 6.5 ± 0.1 km/s and basement thickness of 29.8 ± 2.0 km. Corresponding normal-incidence Moho two-way times (including sediments) are 10.6 ± 0.6 s, consistent with times for events interpreted as Moho reflections on coincident active-source reflection profiles. Modeling of an underplated mafic layer (Vp = 7.2–7.4 km/s) using travel time constraints from SsPmp data and vertical-incidence Moho reflection times yields a total basement thickness of 30–35 km and average basement velocity of 6.35–6.65 km/s for an underplate thickness of 0–15 km.
Key Points
- Unstacked broadband waveforms show clear postcritical SsPmp arrivals over low-velocity sediments of the Georgia coastal plain
- Precritical SsPmp arrivals are also visible
- Average basement Vp (6.5–6.6 km/s) allows for, but does not require, high Vp in the lower crust associated with Mesozoic underplating
1 Introduction
The southeastern United States is an ideal location to study the influence of tectonic inheritance on rift dynamics. Inboard of the Atlantic volcanic passive margin, onshore Triassic rift basins overlie a Late Paleozoic crustal-scale suture zone, the Suwannee suture, buried beneath the Atlantic Coastal Plain (ACP). Seismic refraction profiling suggests that magmatic underplating was localized along the Atlantic rifted margin [Holbrook and Kelemen, 1993], but the relationship between dike intrusion and underplating along the trace of the suture and beneath onshore rift basins remains uncertain [Heatherington and Mueller, 2003]. Fundamental constraints on crustal structure are required to evaluate the extent of crustal modification associated with emplacement of the Central Atlantic Magmatic Province (CAMP) [Hames et al., 2000].
The Southeastern Suture of the Appalachian Margin Experiment (SESAME) array consists of 85 broadband stations along three transects crossing parts of the ACP and southern Appalachians [Parker et al., 2013] (Figure 1). A primary objective of SESAME is to determine the relationship between the Late Paleozoic Suwannee-Wiggins suture (SWS) and the Triassic-Jurassic South Georgia basin (SGB). The SWS trends roughly E-W across southern Georgia and has been interpreted to mark the Late Paleozoic collision of a Gondwanan crustal fragment, the Florida block, with Grenville-age North American crust during the formation of Pangea [McBride et al., 2005]. The SGB, an offset rift basin located ~250 km inboard of the continental margin [Chowns and Williams, 1983; McBride and Nelson, 1991; Chenin and Beaumont, 2013], cross cuts the trend of the suture and occupies a transitional position between other onshore Atlantic and Gulf of Mexico basins [Clendenin, 2013]. Reactivation of the suture zone may have focused crustal thinning and mafic magmatism along the trend of the suture [McBride and Nelson, 1988; Lizarralde et al., 1994]. Alternatively, onshore Mesozoic rifting along the Atlantic margin may have been controlled by low-angle detachment faults that sole out at midcrustal levels [Ussami et al., 1986; Gouiza et al., 2010], leaving inboard basins laterally offset from localized crustal thinning and magmatic underplating along the Atlantic margin [Lister et al., 1986; McBride and Nelson, 1991].
Unfortunately, deep structure beneath the ACP has been difficult to image because of a thick column of very low velocity, unconsolidated sediments and poorly consolidated sedimentary rocks. Conversions and multiples generated within sedimentary columns may completely overwhelm Ps conversions from the lower crust and mantle [Zelt and Ellis, 1999; Yeck et al., 2013], rendering standard approaches for analyzing receiver functions for deep structure ineffective [Langston, 2011]. Generally, the problem is indicated by an apparent delay in the direct P arrival, Pp, due to interference with strong Ps conversions from the base of the sediments [Zelt and Ellis, 1999]. This effect is observed in receiver functions generated for SESAME stations in the ACP.
In this paper we investigate an alternative approach for imaging deep structure using SsPmp, the wide-angle P wave reflection from the Moho generated by conversion of the direct S wave, Ss, at the free surface [Zandt and Randall, 1985]. The method takes advantage of large reflection coefficients at postcritical angles of incidence to generate high-amplitude P wave reflections from the Moho. An important advantage of this approach is that the arrivals are large enough to identify and model without stacking waveforms for large numbers of earthquakes [Tseng et al., 2009]. In principle, given laterally homogeneous structure and events covering a range in ray parameter, SsPmp-Ss differential times can also be used to provide independent estimates of the average P wave velocity of the crust.
Here we investigate the utility of the method for imaging the Moho beneath low-velocity sediments of the ACP.
2 The Southeastern U.S. Atlantic Coastal Plain
The ACP in Georgia and Florida is characterized by poorly consolidated Cretaceous and Cenozoic carbonates and siliciclastics up to 2000 m thick [Chowns and Williams, 1983]. Drilling and seismic data indicate that several Triassic rift basins containing sedimentary rocks correlative with the Newark Group underlie the ACP sediments [Chowns and Williams, 1983], further complicating the velocity structure. In this paper, we focus on the E line, where Triassic rift sediments are significantly thinner than along the W line. Several drill holes along that line bottom out at depths of 1280–1360 m in Paleozoic tuffs, granitic plutons, and Paleozoic sandstone/shale, rather than rift-related sedimentary rocks [Chowns and Williams, 1983]. Along the E line, Coastal Plain sediments increase in thickness from zero at the Fall Line to roughly 1.5 km beneath the southern end of the profile.
3 Methods: Waveform Processing
To avoid interference with the depth phases pS and sS, it has been customary in SsPmp studies to use earthquakes restricted to depths greater than 80 km [Yu et al., 2012]. More recently, Yu et al. [2013] have described a method for deconvolving vertical and radial components using a “pseudo-S” trace consisting of pure-SV motion. This trace contains the source-time function, S-type multiples generated on the receiver side and, importantly, depth phases generated on the source side, allowing the use of earthquakes at all source depths. Details of the method are described in the supporting information. The procedure is illustrated in Figure S1 in the supporting information with synthetic waveforms, in Figure 2a for Event 1 using a single station, and in Figure S2 for the same event using all stations from Line E.
We also computed the envelope function [Bracewell, 1978]. This removes postcritical phase shifts, thus facilitating consistent measurement of SsPmp-Ss differential times for traces with different ray parameters [Phinney et al., 1981]. In practice, SsPmp-Ss differential times were picked from the radial component deconvolved by the pseudo-S phase (R/Q) and corrected for postcritical phase shifts using synthetic envelope functions; details are presented in the supporting information. Phase shifts may be avoided by recording in the precritical range, where SsPmp waveforms are smaller relative to Ss but may still be prominent and well resolved (Figure S5). For Line E, all Events 1 and 2 SsPmp paths are postcritical; precritical SsPmp from other events will be incorporated in future work.
4 Observed SsPmp Phases
Results for Events 1 and 2 (Table S1) for Line E (Figures 3, S8a, and S8b) show clear Ss and SsPmp waveforms that are continuous for over 350 km. Ss arrivals show little evidence of delays caused by interference with shallow conversions and multiples. Average differential SsPmp-Ss times are 5.7–6.4 s (p = 0.128–0.131 s/km) for Event 2 and slightly smaller (5.2–6.3 s), as expected, for the larger ray parameters sampled for Event 1 (p = 0.132–0.135). To the northwest, across the Inner Piedmont and Blue Ridge, SsPmp waveforms become broader and more complex (Figures S8c and S8d). Possible implications for the structure of the lower crust and crust/mantle transition are considered in the supporting information.
5 Modeling SsPmp-Ss Differential Times for Crustal Structure
SsPmp waveforms are also of interest because they can be used to constrain P wave velocity structure without conducting an active-source experiment. We first derived a sediment model based on well data and COCORP (Consortium for Continental Reflection Profiling) data. With this fixed sediment model, we then compared observed and predicted SsPmp-Ss differential times to determine best fitting basement velocity and basement thickness parameters. These steps are described in the next three subsections.
5.1 Effects of Low-Velocity Sediments
To investigate the effect of low-velocity sediments on SsPmp waveforms and travel times, we used the reflectivity method [Randall, 1989] to generate synthetic data for a series of one-dimensional models consisting of a sedimentary layer overlying a 10-layer, roughly 30 km crystalline crust (“Series 1” model in Tables S2 and S3). Average velocities within the sedimentary column (Vp = 2.4 km/s; Vs = 0.75 km/s) were based on active-source reflection data from the SE Georgia Coastal Plain [Iverson and Smithson, 1983; Barnes and Reston, 1992] and borehole measurements of unconsolidated sediments in the Mississippi Embayment [Langston, 2011]. These velocities are considerably smaller than values used for sedimentary layers in previous SsPmp modeling studies [Yu et al., 2012, 2013]. Detailed results are presented in the supporting information. In brief, replacing the single layer with a 10-layer velocity gradient (with the same average velocity) to more realistically incorporate the effects of compaction minimizes reverberations, yielding a reasonably close fit to the observed waveforms (Figure S6).
5.2 Modeling Basement Structure
We used a grid search for finding a range of one-dimensional crustal models that minimize the misfit between observed and computed time picks. Single-layer models generally employed for the analysis of steeply incident Ps arrivals are less useful for SsPmp because of significant refraction effects observed at wide angles of incidence. Instead, we parameterized the crust as a single sedimentary layer (with a thickness that varied from station to station, based on well and seismic reflection data), underlain by a basement with velocities increasing linearly with depth. We fixed the velocity at the top of the basement and carried out the grid search over a range of velocity gradients (corresponding to a range of average basement velocities) and total basement thicknesses. The procedure was repeated by bootstrapping (200 runs) to include randomly picked subsets of the picks [Efron and Tibshirani, 1991]. Details of the procedure are described in the supporting information.
5.3 Modeling Results
Ideally, a separate model would be derived for each station using a number of earthquakes providing a range of ray parameters to constrain variations in average basement velocity beneath the profile. For the preliminary analysis considered here, we used two earthquakes (Events 1 and 2) and combined the data from multiple stations.
Modeling of 37 picks for the two earthquakes (Table S2: Series 1; picks from stations with high S/N waveforms) yielded estimates of 29.8 ± 2.0 km for basement thickness and 6.5 ± 0.1 km/s for average basement velocity, where the uncertainties represent 1 standard deviation from the best fit values found for the individual bootstrap runs. The results are summarized in Figure S7 and discussed in detail in the supporting information.
To get some idea of variations in basement thickness and basement velocity along the profile, we repeated the modeling separately for 10 individual stations (two picks for each station, i.e., one pick for each earthquake) that showed the highest signal levels (Table S2: “Series 2”). Basement thicknesses ranged from 28.3 to 34.9 km (mean for all stations: 31.0, with a standard deviation of 2.3 km) and average basement velocities ranged from 6.41 to 6.72 km/s (mean: 6.6 ± 0.1 km/s). Average sediment thicknesses for these stations were 1.1–1.5 km. Corresponding normal-incidence two-way times to the Moho were 9.8–11.5 s (mean: 10.5 ± 0.6 s). These times are consistent with times for events interpreted as Moho reflections on COCORP Lines 16 and 17 [McBride and Nelson, 1988, 1991; Barnes and Reston, 1992].
Next (Table S2: “Series 3”), we computed basement thicknesses for the two earthquakes separately for individual stations (one pick for each station) assuming a uniform average basement velocity of 6.6 km/s, the mean value derived in Series 2. Resulting basement thicknesses range from 29.2 to 33.1 km (mean for all stations: 31.3 ± 1.1 km) for Event 1 (21 stations), compared to a range of 28.1–32.5 km (mean: 30.6 ± 1.4 km) for Event 2 (18 stations). With the exception of a few outliers, Moho depths plotted at the corresponding reflection points for the two earthquakes are in close agreement (Figure 4).
Corresponding Moho two-way times computed using the appropriate values of sediment thickness for each station are 10.5 ± 0.3 s and 10.3 ± 0.4 s, respectively. Again, these values compare favorably with the times observed on nearby COCORP sections.
To test the sensitivity of our results to the choice of fixed model parameters, we repeated the modeling for Series 1, perturbing the values for each of the parameters one at a time. The results are summarized in Table S4 and discussed in the supporting information.
As an additional check, we computed basement thicknesses using the same model parameterization for several stations along the D line, for comparison with results from previous receiver-function and active-source wide-angle results [Parker et al., 2013; Hawman et al., 2012]. Thicknesses for stations in the Carolina Terrane are in close agreement, but for stations in the Inner Piedmont and Blue Ridge, the SsPmp estimates are 4–11 km smaller than estimates based on previous studies. Factors possibly contributing to the disparities are examined in the supporting information.
Lastly, to investigate the possibility of underplating, we extended the model parameterization to include a high-velocity layer with fixed velocity (7.2–7.4 km/s) and thickness at the base of the crust (see supporting information for details). For mafic layer thicknesses up to 10 km, there is little change in average basement velocity and total basement thickness (Figures S11 and S12). Mafic layer thicknesses greater than 10 km indicate two-way times for normal-incidence Moho reflections that are greater than times (10–11 s) interpreted for coincident COCORP reflection profiles [Barnes and Reston, 1992]. For a mafic layer thickness between 0 and 15 km, the total basement thickness and average basement velocity range from 30 to 35 km and 6.35 to 6.65 km/s, respectively.
For Line E, our estimates of total crustal thickness (basement + sediments: 29–35 km, with a few outliers up to 38 km) are slightly smaller than thicknesses (35–40+ km) recently found based on combined analysis of Rayleigh waves and Ps receiver functions for six roughly coincident stations of the Transportable Array (TA) [Schmandt et al., 2015]. The differences are greatest for stations in northern Florida, where Coastal Plain sediments are thickest (1.5–1.6 km), suggesting that at least part of the disparity may be due to differences in average velocity assigned to the upper crust. However, those differences are relatively small (average P wave velocity = 5.0 km/s in the uppermost 11 km for our study compared with roughly 5.2–5.5 km/s for the TA stations in north Florida), as are the differences in average P wave velocity for the whole crust (6.0 km/s for our study compared with roughly 6.2 km/s for the same TA stations). These differences in average crustal velocity account for only a 1 km difference in total crustal thickness.
Our crustal thicknesses for Line E are in better agreement with the results of Shen and Ritzwoller [2016], who found thicknesses decreasing from roughly 39 km at the Fall Line to 34–36 km in northern Florida, using data from the same TA stations.
6 Discussion
Although the picks included in the foregoing analysis cover only a very narrow range in ray parameter, the preliminary estimates for average basement velocity beneath the Eastern Line are similar to estimates for average crustal velocity beneath the Carolina Terrane (6.5–6.6 km/s) and the Blue Ridge (6.5–6.6 km/s) [Prodehl et al., 1984; Hawman et al., 2012]. Average crustal Vp/Vs ratios of 1.69–1.72 suggest a felsic-to-intermediate average composition for the Inner Piedmont and Carolina terrane [Parker et al., 2013]. However, the crustal structure of the ACP was modified by dike intrusion during emplacement of the CAMP at ~201 Ma. The volume of underplated magma inboard of the margin is poorly constrained, but our modeling results and previously interpreted Moho two-way travel times are consistent with up to 10 km of underplating. East of the SESAME E line, offshore seismic refraction data (Figure 1) indicate a 3–4 km thick lens of high-velocity (7.2 km/s) material at the base of the crust [Lizarralde et al., 1994]. These constraints suggest that the addition of ~5 km of underplated material to the base of the crust beneath the SESAME E line may be more reasonable (Figure S12a).
The new results support earlier interpretations of COCORP lines over the ACP that show a thin crust (29–38 km) and relatively flat Moho (Figure 4). This is in marked contrast with crustal thickness variations across the neighboring crystalline terranes [French et al., 2009; Parker et al., 2013, 2015]; there the crust gradually thickens to the northwest, with a significant root that is fairly localized beneath the high elevations of the Blue Ridge Mountains. Active-source wide-angle soundings [Hawman et al., 2012] and receiver functions for additional broadband stations [Parker et al., 2013] indicate that crustal thicknesses in the range 51–56 km extend 200 km to the northeast of Line D in a 30–50 km wide band along strike beneath the mountains.
Figure 4 highlights two unanswered questions regarding the southern Appalachians: (1) the timing and mechanisms responsible for observed crustal thickness variations and (2) the preservation and/or generation of high elevations and significant relief. The second question can be explained at least in part by enhanced erosion in valleys compared with mountain summits. This has been documented in the Blue Ridge Mountains [McKeon et al., 2014] and can generate an increase in absolute elevation via isostatic rebound [England and Molnar, 1990]. Low Pn velocities (7.9 km/s) beneath the Blue Ridge suggest that structures within the uppermost mantle contribute to the support of topography [MacDougall et al., 2015]. Alternative mechanisms include dynamic processes such as thermal uplift and rebound associated with delamination of a portion of the lithosphere [Wagner et al., 2012].
Regarding the first question, we consider two models for the present configuration of the Moho. One possibility is that the basic features of the crustal architecture were in place by the Jurassic, with the Blue Ridge root either a remnant of a much broader zone of crustal thickening formed during Alleghanian collision and thinned by Mesozoic extension, or an earlier structure associated with Taconic or Grenville collision [Parker et al., 2013]. An Alleghanian age for the root would be consistent with models involving head-on collision [Hatcher, 2010] accompanied by significant crustal thickening, as opposed to transpression [Mueller et al., 2014], during the final stages of the orogeny.
An alternative model involves the northwestward retreat of the Blue Ridge Escarpment [Spotila et al., 2004], the steep (300–500 m of local relief) southeastern flank of the Blue Ridge province that is interpreted to be a “great” escarpment [Ollier, 1985] initially formed in response to Triassic thermal uplift and rifting [Pazzaglia and Gardner, 2000]. According to this model, the crust southeast of the escarpment rebounded and thinned as erosion progressed, leaving thicker crust to the northwest [Spotila et al., 2004]. However, this would not explain thinner crust observed northwest of the Blue Ridge Mountains nor would it explain the relief along the northwestern flank of the range in North Carolina/Tennessee and Georgia, which is just as steep as the southeastern flank but clearly unrelated to escarpment development.
It is possible that both mechanisms have been at work, with escarpment retreat modifying crustal thickness variations initially developed in response to collision and subsequent rifting. To be viable, any model must explain the disparity in geometry between the root and the relatively flat Alleghanian detachment, which dips gently to the southeast. Significant post-Jurassic isostatic rebound of the crust concentrated mainly beneath the terranes southeast of the escarpment is not consistent with the present geometry.
If the Blue Ridge root does (at least in part) predate the escarpment, then it may be slowing the escarpment's northwestward retreat because of the sizable reservoir of crustal material feeding isostatic uplift. The importance of this effect would depend on the present flexural rigidity of the continental lithosphere, a subject of ongoing debate [Bechtel et al., 1990; McKenzie and Fairhead, 1997; Braun et al., 2014].
7 Conclusions
This study provides new Moho depth coverage beneath a ~400 km long profile crossing the Suwannee suture zone and South Georgia rift basin. Deconvolved waveforms for two small-magnitude earthquakes show clear, consistent SsPmp arrivals for broadband stations deployed across the SE Atlantic Coastal Plain, allowing mapping of the crust-mantle transition in spite of strong reverberations generated by low-velocity sediments. As noted by previous workers, amplitudes for postcritical SsPmp are strong enough to allow determination of Moho depths without stacking waveforms for multiple events. Waveforms for precritical SsPmp arrivals, although much smaller in amplitude, have also been identified.
Assuming a basement in which velocity increases linearly with depth, a bootstrapped grid search using differential SsPmp-Ss times suggests an average basement velocity of 6.5 ± 0.1 km/s (basement thickness: 29.8 ± 2.0 km). Estimates for picks from individual stations are 6.6 ± 0.1 km/s (basement thickness: 31.7 ± 3.0 km). Corresponding predicted normal-incidence two-way times to the Moho (including the effects of Coastal Plain sediments) are 10.6 ± 0.6 s, which is consistent with times for events interpreted as Moho reflections on coincident COCORP lines. Models incorporating a 5–15 km thick high-velocity (7.2–7.4 km/s) layer at the base of the crust yield similar fits to the SsPmp-Ss data, but mafic layer thicknesses less than 10 km yield better fits to the COCORP data. For mafic layer thicknesses up to 15 km, the total basement thickness and average basement velocity range from 30 to 35 km and 6.35 to 6.65 km/s, respectively.
Thus, the SsPmp data are consistent with, but do not require, the addition of significant amounts of mafic material to the crust during Mesozoic rifting. Increasing the coverage in ray parameter and azimuth would improve estimates of average crustal properties and would allow better tracking of lateral variations, but determination of velocities within the crust, including more definitive resolution of high-velocity layers, will require more detailed refraction/wide-angle reflection profiling [Shillington et al., 2015].
Acknowledgments
We thank our station hosts for their generous support, the staff at the PASSCAL Instrument Center (especially N. Barstow, P. Miller, and G. Slad) for their assistance in the field, and the members of our field crew. Comments by two anonymous reviewers significantly improved the manuscript. Waveform data were obtained from the IRIS Data Management Center. Data for all the SESAME stations will become available in 2016. This work was supported by NSF grants EAR-0844154 (R.B.H.), EAR-0844276 (K.M.F.), and EAR-0844186 (L.S.W.).
