Seismic attenuation is an important physical parameter for characterizing rocks and providing significant constraints on the viscosity, rigidity, temperature, and mineral composition of the Earth's crust and mantle (Jackson & Anderson, 1970;Knopoff, 1964). Additionally, seismic attenuation measurements can provide independent constraints on the interpretation of seismic velocity models (Deen et al., 2006;Faul & Jackson, 2005;Godey et al., 2004;Goes & van der Lee, 2002;Goes et al., 2000;Hwang et al., 2009;Lee, 2003;Schutt & Lesher, 2006;Shapiro & Ritzwoller, 2004;Sobolev et al., 1996). Previous seismological investigations suggest that anelasticity and velocity variations exhibit strong sensitivity to temperature anomalies in the uppermost mantle (Anderson, 1967;Faul & Jackson, 2005;Goes et al., 2000;Jackson et al., 2002;Karato, 1993;Knopoff, 1964;Sato et al., 1989;Wang et al., 2017). Such anelasticity can be estimated by measuring the attenuation of teleseismic body waves, as they provide frequencies intermediate to those of long-period surface waves and regional earthquake body waves (Solomon, 1972;Wang et al., 2017).
Anelastic or intrinsic attenuation refers to the conversion of seismic energy into heat, generally caused by grain boundary friction (Jackson & Anderson, 1970) and the movements of dislocations through the mineral grains (Gorich & Muller, 1987). Anelastic attenuation of seismic waves in a medium is expressed in terms of the seismic quality factor (Q), which is inversely proportional to anelastic attenuation. The relation between Q and energy dissipation is
where E o is the maximum value of elastic energy stored during one cycle of loading, and δE is the energy loss during the cycle (Knopoff, 1964).
In addition to anelasticity, scattering can also lead to the reduction of body-wave amplitude. Akinci et al. (1995) propose that the energy dissipation of body waves due to scattering is more prominent at shorter distances and decreases substantially as the propagation distance increases. The study also suggests that Abstract Seismic attenuation is an important parameter for characterizing subsurface morphology and thermal structure. In this study, we use P-wave amplitude spectra from 588 teleseismic events recorded by 477 broadband seismic stations in the southeastern United States to examine the spatial variations of seismic attenuation in the crust and upper mantle. The resulting seismic attenuation parameter (∆t*) measurements obtained using the spectral ratio technique reveal a zone of relatively low attenuation in the Gulf of Mexico Coastal Plain and the southwestern terminus of the Piedmont province. Spatial coherency analysis of the ∆t* observations suggests that the center of the low attenuation layer is located within the uppermost mantle at about 70 km depth. This low attenuation anomaly lies along the suture zone between Laurentia and Gondwana and approximately coincides with the east-west trending Brunswick magnetic anomaly. The origin of this low attenuation anomaly can be attributed to low attenuation bodies in the form of remnant lithospheric fragments in the deep crust and the uppermost mantle. The contribution of scattering to the observed ∆t* is estimated by calculating the ratio of amplitude on the transverse and vertical components in the P-wave window. Relative to the rest of the study area, the Gulf of Mexico Coastal Plain demonstrates weaker scattering which is suggestive of a more homogenous crustal and uppermost mantle structure.
intrinsic attenuation is dominant at larger epicentral distances, without showing a strong frequency dependence. Laboratory experiments have also reported frequency-independent Q for many solids up to moderately high frequencies (Knopoff, 1964). On the basis of previous laboratory and observational studies (Dziewonski, 1979;Jackson & Anderson, 1970;Knopoff, 1964), frequency-independent Q is assumed in this and numerous previous studies (e.g., Hwang et al., 2009) to estimate the attenuation of teleseismic P-waves for frequencies up to 1 Hz.
The southeastern United States (SEUS) is tectonically more stable than the western U.S. but is more active than most portions of the central U.S.
(Figure 1). Since the late Proterozoic (>1.0 Ga), multiple phases of terrane accretion, orogenies, and continental breakups have taken place in the SEUS (Cook et al., 1979;Hatcher, 2010;Thomas, 2006). The Grenville orogeny is the oldest known Mesoproterozoic mountain-building episode that marks the assembly of the supercontinent Rodinia (∼1.1 Ga; Denison et al., 1984;Thomas, 1985). The Grenville Front separates the SEUS continental margin from the Mazatzal province (Figure 1), which was formed due to the accretion of juvenile volcanic arcs to the older stable core of North America. Following the breakup of Rodinia (∼570 Ma), several episodes of continental accretion and orogenies resulted in the formation of the SEUS terrane. The Alleghanian orogeny represented by the collision of Laurentia and Gondwana at ∼330 Ma resulted in the formation of the supercontinent Pangea (Iverson & Smithson, 1983;Rankin et al., 1991). This collision marked the formation of the Appalachian Mountains and the addition of the Suwannee terrane, which has significantly different tectonic attributes, lithology, and fossil accumulations than Laurentia (Mueller et al., 2014). A regionally extensive swath of lower-than-normal magnetism known as the Brunswick Magnetic Anomaly (BMA; Figure 1) lines up with the Suwannee Suture zone (Higgins & Zietz, 1983;Mueller et al., 2014;Williams & Hatcher, 1983).
The attenuation structure from previous continental-scale studies in North America suggests relatively low attenuation in the eastern and southeastern U.S. in comparison to the tectonically active western U.S. (Der et al., 1982;Der & McElfresh, 1977;Hwang et al., 2009;Lawrence et al., 2006;Solomon & Toksöz, 1970). A recent study (Cafferky & Schmandt, 2015) computes the spatial variation of seismic attenuation across the U.S. using teleseismic P-wave spectra from deep earthquakes for multiple frequency bands between 0.08 and 2 Hz. All frequency bands yield a high attenuation region near the Appalachian margin and low to medium attenuation is reported in the continental interior. Most of the previous seismic attenuation studies are conducted for the entire continental U.S. with a limited spatial resolution in the SEUS. In this study, a comprehensive assessment of seismic attenuation and the effects of scattering beneath the SEUS is conducted using data from the USArray and other portable or permanent deployments listed in the Data Availability Statement section. The results support the existence of remnant lithospheric segments in the crust and upper mantle beneath the Gulf of Mexico (GoM) Coastal Plain.
Seismic data used in the study were recorded by 477 broadband seismic stations and were obtained (Figure 1) from the Incorporated Research Institutions for Seismology (IRIS) Data Management Center (DMC). The stations include 220 USArray Transportable Array (TA) stations which sampled the study area with ∼70 km spacing. The cutoff magnitude (M c ) for data requesting is computed using SHRIVASTAVA ET AL.
10.1029/2021GC009715 2 of 15 -90˚-85˚-802 53 03 54 0˚G r e n v il le G r e n v i l l e A p p a l a c h i a n M a z a tz a l A p p a l a c h i a n P l a t e a u V a l l e y & R i d g e Blue Ridge Mts. P i e d m o n t Coastal Plain S u w a n n e e S u t u r e BMA -3000 -1500 0 1500
M c = 5.2 + (∆-∆ min )/(180.0-∆ min )-D/D max , where ∆ is the epicentral distance (which ranges from 30° to 180°) in degree, D is the focal depth in km, ∆ min = 30°, and D max = 700 km (Liu & Gao, 2010). The events were recorded by both portable and permanent seismic stations in the area of 25°-40° North and 80°-90° West, between March 1993 and January 2019.
To enhance the quality of the measurements, only events recorded by a minimum of 10 stations were kept, and as a result, 588 teleseismic events (Figure 2) were used in the study.
Several techniques have been developed to estimate the amplitude of seismic attenuation. These techniques can be broadly classified as either time-domain methods or frequency domain methods. Wavelet modeling (Jannsen et al., 1985), rise-time (Gladwin & Stacey, 1974), and analytical signal (Taner et al., 1979) methods are some of the major techniques used to compute seismic attenuation in the time domain, whereas methods such as spectral ratio (Teng, 1968), spectral matching (Raikes & White, 1984), coda normalization (Aki, 1980), and spectral modeling (Gao, 1997;Halderman & Davis, 1991) work in the frequency domain.
Attenuation is measured in terms of the attenuation parameter t*, which is defined as the total traveltime of the wave along the raypath divided by the quality factor (Kovach & Anderson, 1964), that is,
where V(r) is the velocity of the waves, and Q(r) is the quality factor.
In the frequency domain, the amplitude spectrum ( ik A f) of an event "k" recorded at station "i" can be expressed as (Teng, 1968)
where k S f is the source spectrum of the source wavelet, ik R f is the spectrum of the near-receiver effects, i I f is the spectrum of the instrument response, and ik G f is the spectrum of Green's function, which can be written as
The spectral ratio method used in the study is the most widely used technique to estimate seismic bodywave attenuation relative to a reference spectrum (e.g., Der & McElfresh, 1976;Hwang et al., 2009;Solomon & Toksöz, 1970;Teng, 1968). One of the benefits of using this method is that for teleseismic events, the source signal and common path effects are removed. The requested vertical component seismograms are re-sampled into a sampling frequency of 20 sps, and a section of the vertical component seismogram with a total length of 51.2 s (i.e., 1024 data samples) starting from 10 s before the theoretical arrival time of the P (or PKP) is selected for computing the spectrum. The instrument response is removed by deconvolving the seismograms with the instrument response function. A 10-s window length before the arrival of P-wave is used to determine the noise amplitude. The signal-to-noise ratio (SNR) between the maximum absolute value of the signal amplitude and mean absolute noise amplitude is computed for every trace, and seismograms with an SNR smaller than 10.0 are not used for the study. The P-wave section of the seismogram is tapered using the customary cosine-sum window, with the form where w(n) is a zero-phase function, N is a positive integer, and the numerical value of 0 a is set as 0.54, which categorizes this tapering function as a Hamming window. A bandpass filter with corner frequencies of 0.1 and 0.5 Hz is applied to the selected seismograms. Amplitude spectra of all the filtered high-quality seismograms are computed using the Fourier transform. To minimize the effects of heterogeneities outside the study area, for each event we use the mean spectrum computed over all the stations that recorded this event as the reference spectrum in the spectral ratio. Additionally, to exclude seismograms with high noise, the minimum correlation coefficient between each spectrum and the mean spectrum is set to 0.9. The relative attenuation factor Δt ik between station "i" and the reference spectrum from event "k" is calculated by fitting the spectral ratio with a straight line using the least-squares method (Der & McElfresh, 1977;Solomon & Toksöz, 1970), that is,
where C is the ratio between the near receiver effects of station "i" and the reference spectrum from event 3) and is assumed to be frequency independent. The automatically computed results are then manually checked to remove measurements with abnormal data or with a nonlinear frequency variation of the spectral ratios. Furthermore, Δt ik measurements with an absolute value greater than 1.0 s, or a standard deviation greater than 0.2 s are excluded. Figure 3 shows an example of the spectral ratio and associated spectra and seismograms.
The resulting 14,702 individual ∆t* measurements (Figure 4a) obtained using the spectral ratio method are used to compute the station-averaged ∆t* measurements if the number of measurements obtained at the station is three or greater (Figures 4b and 5a and Table S1), which vary from -0.62 s ± 0.03 to 0.60 s ± 0.04 s in the study area and demonstrate systematic spatial variations. The station-averaged ∆t* measurements (Figure 5a) are spatially interpolated by averaging the measurements in overlapping 1° by 1° blocks with a moving step of 0.1° (Figure 5b). We experimented with different values of the size of the blocks for smoothing and found 1° to be a balanced value that most clearly demonstrates the spatial variation of station averaged ∆t* measurements. As the block size for smoothing increases, both the spatial resolution and peak-topeak range of the ∆t* measurements decrease, and vice versa. To get a sense of the uncertainties in the ∆t* measurements, we compute the spatial distribution of the standard deviation (SD) of the ∆t* measurements (Figure 6). Areas with the largest SD are in the Florida Peninsula and along the southernmost part of the GoM Coastal Plain. Some previous studies (e.g., Cafferky & Schmandt, 2015;Dong & Menke, 2017) used only events with hypocenters deeper than at least 150 km for attenuation measurements for the purpose of reducing the impact of the reduction in high frequency components by the lithosphere on the source-side. To explore the influence of including the shallow events, we compute ∆t* measurements by only using events with a focal depth ≥150 km. The results (Figure S1) show similar spatial variations with those obtained using all the events (Figure 5), even with a much-reduced number of measurements.
Based on the characteristics of the measurements (Figure 5) and the tectonic setting, we divide the study area into five areas: Mazatzal Province (A), Grenville Province (B), Southern Appalachian Mountain Range and Piedmont Province (C), GoM Coastal Plain (D), and Florida Peninsula (E). Area A is part of the stable central North American craton and is characterized by positive V p anomalies in the upper-most mantle (Figure 5a; Golos et al., 2020). Physiographically, the eastern portion of Area A belongs to the Appalachian Plateau, located adjacent to the southern end of the Appalachian Mountains. The southeastern region of the area displays higher attenuation relative to other regions in the area. The western portion of Area A consists of the Interior Lowlands that include several structural depressions that have filled with sediments mostly eroded from the mountains (Swaby et al., 2016). Attenuation decreases gradually westward as the lithospheric thickness beneath the stable part of North America Craton increases. The northwestern region of Area A consists of the Illinois Basin comprising of a thick layer of Cambrian through Pennsylvanian sedimentary rocks (Swann, 1968) and Proterozoic granites and rhyolite in the basement which date back to around 1.55 Ga (van der Pluijm & Catacosinos, 1996;van Schmus et al., 2007). Negative ∆t* measurements obtained in the Illinois Basin coincide with the strong positive V p anomaly mapped by Golos et al. (2020) that extends up to the depth of 70 km. The southernmost tip of Area A is characterized by negative ∆t* observations belonging to a zone of low attenuation pervasively observed along the northern border of the GoM Coastal Plain (Figure 5). The average ∆t* value for Area A is close to zero (0.01 s ± 0.01 s).
Area B occupies the Grenville Province tectonically and is composed of the Appalachian Plateau except for the southernmost quarter which belongs to the GoM Coastal Plain. The observed ∆t* values show a sharp contrast between the Appalachian Plateau and the GoM Coastal Plain, with mean values of 0.03 s ± 0.01 s for the former, and -0.30 s ± 0.03 s for the latter region. The NE portion of the area, which is found to possess high V P anomalies in the uppermost mantle (Figure 5a), shows reduced ∆t* measurements. The SW boundary of Area B traverses the area with negative ∆t* values, suggesting that the observed spatial variation of the ∆t* measurements is not controlled by Precambrian basement terranes, but by physiographic divisions which are mostly the result of post-Precambrian tectonic activities.
Area C is a physiographical province mostly consisting of the Blue Ridge Mountains and Valley and Ridge of the southern Appalachian Mountains in the west, and a plateau region of the Piedmont Province in the east. Although the western and the eastern regions of Area C are physiographically distinct from each other, both regions share similar crystalline igneous and metamorphic rocks (Swaby et al., 2016). The northernmost part of this area is comprised of the Appalachian Plateau. Similar to Areas D and E, tectonically it is part of the Paleozoic-Cenozoic Appalachian Province. The ∆t* measurements are comparable to Area B, with an areal mean value of 0.16 s ± 0.01 s but are higher than those observed on the GoM Coastal Plain (Figure 5). No obvious change in ∆t* is observed across the boundary between this area and Area B, which is a tectonic boundary.
Area D belongs to the GoM Coastal Plain, which is composed of very young rocks, ranging in age from the Cretaceous to the present. It is characterized by a well-defined E-W zone of low ∆t* measurements, except for the NE and SW corners of the area. The zone of negative ∆t* closely follows the northern boundary of the GoM Coastal Plain and extends to the southernmost portions of Areas A and B. The mean ∆t* value for this area is -0.17 s ± 0.02 s which is the lowest among all the five areas.
Area E, which includes the Florida Peninsula, is characterized by ∆t* values that are intermittent between those observed in Areas A-C and D, with a mean value of -0.03 s ± 0.02 s. The ∆t* values demonstrate a southward gradual increase (Figure 5a), and the same trend is observed for the uncertainty of the measurements (Figure 6).
We estimate the optimal depth of the observed ∆t* anomalies by adapting a procedure that was developed for estimating the depth of the source of anisotropy using shear wave splitting measurements (Liu & Gao, 2011). Spatial coherency of seismic attenuation parameters is used to estimate the depth of seismic attenuation by computing a spatial variation factor ( Δt F ). The geometric distribution of the ray-piercing points is computed at a depth incremental interval of 5 km from 0 to 400 km, based on the IASP91 Earth model (Kennett & Engdahl, 1991). For each depth, the study area is divided into overlapping blocks of 0.2° × 0.2° at a distance of 0.05° between the centers of the neighboring blocks.
F values are then calculated at each depth using
where N is the number of blocks, M i is the number of measurements for the i-th block, Δ ij t is the attenuation parameter in the i-th block, and t i is the average ∆t* over all the measurements in block i. A detailed explanation of this approach along with the FORTRAN program is illustrated in Gao & Liu (2012). The assumption adopted in this approach is that the attenuation of body waves is caused by a single horizontal layer with spatially variable thickness. This means that the resulting optimal depth corresponding to the minimum value of Δt F indicates the center of the layer. Figure 7 shows the calculated Δt F plotted against the assumed depth of attenuation for the SEUS. The resulting Δt F shows that the optimal depth is about 70 km, that is, in the uppermost mantle. Note that the optimal depth can be viewed as the weighted mean depth computed by the magnitude of lateral variations of the observed ∆t* values. For a homogenous layer, it is the depth of the center of the layer. If ∆t* variations decrease with depth, the resulting optimal depth is smaller than the center of layer and vice versa. In all cases, the actual thickness of the layer cannot be determined.
Previous larger-scale body wave (Cafferky & Schmandt, 2015;Hwang et al., 2009;Solomon & Toksöz, 1970) and surface wave (Bao et al., 2016;Baqer & Mitchell, 1999;Pasyanos, 2013) attenuation studies show a common pattern of high attenuation in the Appalachian Mountains and attenuation beneath the GoM Coastal Plain. Cafferky & Schmandt (2015) map the upper mantle ∆t* values across the contiguous U.S. by inverting teleseismic P-wave amplitude spectra for multiple frequency bands ranging between 0.08 and 2 Hz. The ∆t* values obtained for all the frequency bands are consistent with the results obtained in this study. Cafferky & Schmandt (2015) display the ∆t* measurements using a median smoothing radius of 1.75° and 5° that resulted in two different spatial resolutions. The spatially interpolated results from this study (Figure 5b) are obtained by overlapping 1° by 1° blocks and are similar to their results obtained using the median smoothing radius of 1.75°. They report the lowest mean 95% confidence interval of ∆t* measurements (0.09 s) for the 0.08-2 Hz band, and highest (0.21 s) for the 0.08-0.5 Hz band calculated using over 16,000 ∆t* measurements across the entire contiguous U.S. In comparison, the 95% confidence interval of 14,702 ∆t* measurements in our study is 0.01 s for the SEUS. Note that in our study the frequency band is 0.1-0.5 Hz, where the teleseismic P-wave is the strongest (Figure 3b) which could account for the differences in the small confidence interval of our measurements.
A surface wave attenuation study (Gallegos et al., 2014) uses a two-station method to estimate Lg attenuation in the central and eastern U.S. Their results reveal a low crustal attenuation anomaly beneath the GoM Coastal Plain, which correlates with the location of the low attenuation anomaly observed in our study. Lawrence et al. (2006) measure seismic attenuation beneath the North American continent using waveform cluster analysis and further correlate the results with the travel time. The study finds that seismic travel times and attenuation are weakly correlated (R 2 < 0.3). The sparsely populated seismic stations over a large study area and decreased waveform coherence between the stations produced large-scale variations in seismic attenuation, and therefore, the attenuation structure in the SEUS is mapped with a low spatial resolution relative to those using data from the USArray. Other previous studies (Der et al., 1982;Der & McElfresh, 1977;Hwang et al., 2009;Solomon & Toksöz, 1970) calculate the crustal and upper mantle seismic attenuation beneath the U.S., and none of them reveals the low attenuation anomaly observed in our results near the southwestern terminus of Piedmont (Figure 5b). This is possibly due to the limited number of stations used in most of these studies to produce continent-scale attenuation maps, hence unable to resolve detailed features, and only major trends are reported. Our results make a more comprehensive assessment of P-wave attenuation using a large number of waveforms thereby obtaining a more detailed attenuation structure of the SEUS.
We next compare our results with those from previous velocity tomography studies to provide constraints on the geological implications of the attenuation measurements. The shear velocity in the upper mantle beneath the study area is as much as 15%-20% higher than that in the western U.S. as reported in the studies of body wave travel-times (Golos et al., 2018;Grand & Helmberger, 1984;Melbourne & Helmberger, 1998) SHRIVASTAVA ET AL.
10.1029/2021GC009715 8 of 15 -90˚-85˚-802 53 03 54 0˚-90˚-85˚-802 53 03 54 0˚G r e n v il le G r e n v i l l e A p p a l a c h i a n M a z a tz a l A p p a l a c h i a n P l a t e a u V a l l e y & R i d g e Blue Ridge Mts. P i e d m o n t Coastal Plain S u w a n n e e S u t u r e BMA SD of station-averaged t* (s) -90˚-85˚-802 53 03 54 0˚0 2012) obtain high-velocity anomalies in the SEUS using seismic ambient noise data. Another high resolution 3-D shear velocity model of the crust and uppermost mantle beneath Mexico and the southern U.S., constrained by Rayleigh wave group velocity measurements up to 90 s period, reveals higher seismic velocities in the SEUS relative to the southwestern U.S. in the uppermost mantle (Spica et al., 2016).
Scattering is an important factor that can lead to the decay of the amplitude because of the heterogeneity of the Earth's crust and mantle (Shapiro & Kneib, 1993). Most of the rocks and minerals contain heterogeneities in the form of grains, mineral boundaries, pore edges, cracks, etc., and the seismic energy is scattered when it encounters these features. Different modes of scattering are often determined based on the ratio between the scale of heterogeneity of the medium, a, and the wavelength (Wu & Aki, 1985).
A small S r (<<0.01) indicates that the size of the heterogeneities is extremely small relative to the seismic wavelength, leading to insignificant scattering. Scattering from heterogeneities with 0.01 < S r < 0.1 is termed as Rayleigh scattering, and that from heterogeneities with S r in the range of 0.1 and 10 is termed as Mie scattering, which produces strong attenuation and distinguishable scattering in the seismic signal.
Theoretically, there should be zero energy on the transverse component of P-waves in an isotropic medium that is free of heterogeneities capable of producing scattering. Therefore, most of the energy in the P-wave window on the transverse component is the scattered energy due to 3-D heterogeneity. To examine the lateral variation of the strength of scattering, we calculated the ratio of the mean absolute amplitude between the transverse component and that of the vertical component for all the events that we used to estimate the ∆t*. We selected a signal window that is 5 s before and 10 s after, and a noise window that is 5-15 s before, the theoretical P-wave arrival time for both the vertical and transverse components. A bandpass filter with corner frequencies of 0.1 and 0.5 Hz is used, which is identical to the one used in ∆t* calculation. The ratio of the noise normalized absolute mean amplitude between the transverse and vertical components is calculated for each of the event-station pairs using / , / Ts Tn Rtz Zs Zn (9)
where Z s and T s are the mean absolute vertical and transverse amplitudes in the signal window, and, Z n and T n are the mean absolute vertical and transverse amplitudes in the noise window, respectively.
The station-averaged Rtz measurements for the entire study area (Figure 8a) range from 0.086 to 0.424, with a mean value of 0.187 ± 0.040 s. The GoM Coastal Plain, which is an area characterized by low attenuation anomalies (Figure 5b and Rtz measurements. A strong positive correlation between ∆t* and Rtz values would reveal potential scattering artifacts, and zero or negative correlation would indicate that the ∆t* values likely reflect intrinsic attenuation (Cafferky & Schmandt, 2015). High positive XCCs are observed near the Appalachian Mountains and northwest part of the Appalachian Plateau (Figure 8b), which is consistent with the scattering estimates obtained by Cafferky & Schmandt (2015) from T/Z spectral analysis of teleseismic P-waves in the frequency between 0.08 and 0.4 Hz. Using receiver function, Schulte-Pelkum & Mahan (2014) report high scattering in the Appalachian Mountains. A recent study that uses the USArray data to estimate the crustal attenuation of high frequency (1-20 Hz) shear waves reports high scattering in parts of Interior Plains and Appalachian Highlands (Eulenfeld & Wegler, 2017). They also report low scattering in the parts of the GoM Coastal Plain including the Lower Mississippi Region. The relatively low Rtz values and mostly negative XCCs observed along the GoM Coastal Plain in our study suggest a relatively more homogenous crustal and upper structure in this area.
The geographic variation of ∆t* across the SEUS provides insights into the structure and dynamics of the upper mantle. The estimated ∆t* measurements correspond with the P-wave velocity anomalies in the uppermost mantle (Figure 5a). The strongest correspondences include high ∆t* in the Appalachians, where upper mantle velocities are low, and low ∆t* in the GoM Coastal Plain where an E-W strip of high upper mantle velocities are reported in numerous velocity tomography studies (Bensen et al., 2008;Biryol et al., 2016;Gaite et al., 2012;Golos et al., 2018;Spica et al., 2016). Previous seismic studies have revealed mantle upwelling beneath several sections of the Appalachians along the eastern North American margin (Savage et al., 2017;Schmandt & Lin, 2014), and some of which are attributed to edge-driven convection (Menke et al., 2016). A recent study (Byrnes et al., 2019) estimates the upper mantle seismic attenuation beneath the Appalachian Mountains using the tight station spacing of 10-25 km. They interpret the high-attenuating upper mantle as the result of the removal of mantle lithosphere from a 100 km wide region beneath the central Appalachian Mountains.
The low attenuation anomaly observed beneath the GoM Coastal Plain lies within the proposed Suwannee suture zone (Thomas, 2011), and roughly coincides with the east-west trending BMA (Figure 5b) (Higgins & Zietz, 1983;Mueller et al., 2014;Williams & Hatcher, 1983). This magnetic anomaly located within the study area runs from Alabama across southern Georgia up to North Carolina's northern banks in the Atlantic Ocean. The source of the BMA is ambiguous because of its apparent connection with both the Permo-Carboniferous Alleghanian orogeny (330-270 Ma) and the volcanic rocks that caused the emplacement of the Central Atlantic Magmatic Province (∼200 Ma). Lower crustal seismic reflectors overlapping the location of magnetic low in southern Georgia (McBride & Nelson, 1988) and more recent magnetic models (Parker, 2014) suggest that the continental segment of the BMA can be explained by the collision of Laurentia and Gondwana. Seismic data have revealed the remnants of Pangea's breakup in the shallow crust near the BMA, providing evidence for the collision zone in the deep crust (Parker, 2014). Therefore, the low attenuation anomaly along the path of BMA can be explained by the presence of remnant lithospheric fragments in the deep crust or the uppermost mantle.
S-to-P receiver function studies using the USArray data (e.g., Hopper & Fischer, 2018;Liu & Gao, 2018) suggest that in the SEUS, the lithosphere has an average thickness of ∼70 km, which is comparable to the estimated optimal depth of the weighted center of the anomalous attenuation layer (Figure 7). In addition, seismic tomography studies suggest a high velocity band approximately overlaps with the low attenuation zone along the GoM Coastal Plain in the depth range of ∼20-∼200 km relative to the Appalachians (Golos et al., 2020). Using the Qp values in the PREM model (Dziewonski & Anderson, 1981) for the remnant lithosphere (Q L = 1400) and the surrounding asthenosphere (Q A = 195), and Vp values of 8.1 km/s for the lithosphere (V L ) and 8.0 km/s for the asthenosphere (V A ), respectively, the required vertical length of the remnant lithospheric slab (R L ) in order to produce the observed -0.17 s ± 0.02 ∆t* value is as large as 305 ± 35 km which is inconsistent with results from seismic tomography studies. One way to produce a more reasonable R L is to use a smaller Q A value. For instance, when a Q A value of 120 is used, R L would reduce to ∼180 ± 20 km which is more in accordance with tomography results (e.g., Golos et al., 2020).
Additional studies are needed to isolate the intrinsic attenuation from the observed total ∆t* and to more Geochemistry, Geophysics, Geosystems accurately determine the absolute Qp value for the proposed lithospheric segments and that of the ambient asthenosphere, in order to confirm the existence of the lithospheric segments and their spatial distribution.
We have utilized 14,702 teleseismic P-wave amplitude spectra recorded by 477 broadband seismic stations to map the spatial distribution of ∆t* in the SEUS. This large data set of events recorded by the dense array of stations including 220 USArray TA stations provides a better constraint on the crustal and upper mantle P-wave attenuation structure than previous larger-scale attenuation studies. The resulting ∆t* measurements show a systematic contrast between the Appalachian Mountain range and the GoM Coastal Plain exhibiting high and low attenuation, respectively. An east-west strip of low attenuation anomaly is identified beneath most of the GoM Coastal Plain. The weighted center of this anomaly is located at about 70 km depth as estimated using the spatial coherency approach. This anomaly lies along the Suwannee suture zone that separated Laurentia and Gondwana during the Alleghanian orogeny. It also coincides with the path of the Brunswick Magnetic anomaly, providing evidence of low attenuation and fast velocity bodies in the upper-most mantle that are likely remnant lithospheric segments extending from the crust to the middle upper mantle. The ratio between the transverse and vertical amplitudes of the P-wave is calculated to estimate the lateral variation of scattering. Areas of relatively high transverse/vertical ratios are observed in the Appalachian Plateau and the Floridan Peninsula, whereas low ratios observed in the GoM Coastal Plain indicate that this area is relatively less capable of producing scattering.