An Abstract of the Dissertation Of

Home , Magnetic pulsations, Magnetotellurics

AN ABSTRACT OF THE DISSERTATION OF

Benjamin S. Murphy for the degree of Doctor of Philosophy in Geology presented on May 28, 2019.

Title: Magnetotelluric Constraints on Lithospheric Properties in the Southeastern United States

Abstract approved:

______Gary D. Egbert

By inverting EarthScope long-period magnetotelluric (MT) data from the southeastern United States (SEUS), we obtain electrical conductivity images that provides key insights into the geodynamics of this region. Significantly, we resolve a highly electrically resistive block that extends to mantle depths beneath the modern Piedmont and Coastal Plain physiographic provinces. As high resistivity values in mantle minerals require cold mantle temperatures, the MT data indicate that the sub- Piedmont thermal lithosphere must extend to greater than 200 km depth. This firm bound appears to conflict with conclusions from seismic results; tomography shows that velocities in this region are generally slightly slow with respect to references models. This observation has led to a seismically-informed view of relatively thin (<150 km), eroded thermal lithosphere beneath the SEUS. However, resolution tests demonstrate that our MT constraints are robust. Furthermore, narrow-band biases in MT transfer functions from the SEUS due to geomagnetic pulsations associated with field-line resonances support the presence of bulk resistive lithosphere in this region. We show that, by considering anelastic prediction of seismic observables, MT and seismic (tomography, attenuation, receiver function) results are in fact consistent with thick (~200 km), coherent thermal lithosphere in this region. Our results demonstrate the danger of interpreting seismic results purely in terms of reference models and the importance of integrating different geophysical techniques when formulating geodynamic interpretations.

Magnetotelluric Constraints on Lithospheric Properties in the Southeastern United States

Benjamin S. Murphy

A DISSERTATION

submitted to

Oregon State University

in partial fulfillment of the requirements for the degree of

Doctor of Philosophy

Presented May 28, 2019 Commencement June 2020

Doctor of Philosophy dissertation of Benjamin S. Murphy presented on May 28, 2019

APPROVED:

Major Professor, representing Geology

Dean of the College of Earth, Ocean, and Atmospheric Sciences

Dean of the Graduate School

I understand that my dissertation will become part of the permanent collection of Oregon State University libraries. My signature below authorizes release of my dissertation to any reader upon request.

Benjamin S. Murphy, Author

ACKNOWLEDGEMENTS

The research presented here was supported by the NSF Graduate Research Fellowship Program under Grant 1314109-DGE to Oregon State University. Thanks to Anne Fulton and Bug, Moo, Peanut, and Cinder Fluffton for all their support.

Chapter 1: We thank Emily Hart, Kyle Jones, Allen Hooper, Kyle McDonald, Taryn Bye, Eleanor Kester, Lana Erofeeva, Brady Fry, Lu Pellerin, and Adam Schultz for their work in collecting and processing these data. Bo Yang and Naser Meqbel assisted greatly with use of ModEM in this project. Conversations with and input from Nikki Moore, Phil Wannamaker, Dave Graham, Gene Humphreys, Steve Constable, Paul Bedrosian, and Eric Kirby greatly benefited this work. We thank Rob Harris and Brandon Schmandt for valuable feedback on a preliminary version of this paper. We also thank Kate Selway, Ian Ferguson, and Peter Shearer for very helpful reviews of the manuscript. This research was supported by NSF Grant EAR1053628 and by the NSF Graduate Research Fellowship Program under Grant 1314109-DGE. (Copyright for this chapter is held by Elsevier B.V.; it is reproduced here under the Author’s Right to Personal Use.) Chapter 2: We are grateful to Robert McPherron for helping us to understand the physics of geomagnetic pulsations and field-line resonances. We also thank Maxim Smirnov for his help with the robust array processing program. We thank two anonymous reviewers and editor Juanjo Ledo for a helpful review of the manuscript. This research was supported by the NSF Graduate Research Fellowship Program under Grant 1314109-DGE and by NSF Grant EAR1447109. Chapter 3: We thank Brandon Schmandt and Derek Schutt for their valuable insights that inspired and greatly benefited this work. We also thank Uli Faul, Emily Hopper, Jeff Park, Shun Karato, and Tolu Olugboji, who all provided helpful feedback on these calculations and interpretations. Jeff Park inspired the idea that the Piedmont Resistor may possibly represent a Mesozoic example of craton formation. We acknowledge Lara Wagner, Fred Pollitz, and Brandon Schmandt for graciously providing their surface wave tomography models to us for these analyses. We thank Stephan Thiel, an anonymous reviewer, and editor Uli Faul for a helpful and constructive review of this manuscript. This work was supported by the NSF Graduate Research Fellowship Program under Grant 1314109-DGE and by NSF Grant EAR- 1820688.

TABLE OF CONTENTS Page Introduction ...... 1 Chapter 1: Electrical Conductivity Structure of Southeastern North America: Implications for Lithospheric Architecture and Appalachian Topographic Rejuvenation ...... 3 1.1 Abstract ...... 4 1.2 Introduction ...... 4 1.3 Data and Methods ...... 6 1.4 Results ...... 7 1.5 Discussion ...... 10 1.5.1 Well Resolved Shallow Structures ...... 10 1.5.2 Poorly Resolved Structures ...... 13 1.5.3 Deep Lithospheric Conductivity Contrast and the Piedmont Resistor ...... 14 1.5.3.1 Constraints from the Magnetotelluric Data: Thick Sub-Piedmont Lithosphere14 1.5.3.2 Comparison to Seismic Images: Minor Agreement and Major Conflict...... 16 1.5.3.3 Possible Explanation: A Metasomatized Lithospheric Root that Regrew after Delamination ...... 18 1.5.3.4 Implications for Appalachian Topographic Rejuvenation ...... 20 1.6 Conclusions ...... 21 1.7 References ...... 22 Chapter 2: Source Biases in Mid-Latitude Magnetotelluric Transfer Functions due to Pc3-4 Geomagnetic Pulsations ...... 29 2.1 Abstract ...... 30 2.2 Introduction ...... 30 2.2.1 The Quasi-Uniform Source Assumption in the Magnetotelluric Method ...... 32 2.2.2 Geomagnetic Pulsations and Field-Line Resonances ...... 32 2.3 Data and Methods ...... 34 2.4 Results ...... 35 2.5 Discussion ...... 38 2.5.1 Pc’s and the 3D Earth ...... 39 2.5.2 Implications for MT Data Inversions ...... 39 2.5.3 Implications for Earth Structure ...... 40 2.6 Conclusions ...... 40 2.7 References ...... 41

TABLE OF CONTENTS (cont.) Page Chapter 3: Synthesizing Seemingly Contradictory Seismic and Magnetotelluric Observations in the Southeastern United States to Image Physical Properties of the Lithosphere ...... 45 3.1 Abstract ...... 46 3.2 Introduction ...... 46 3.2.1 Geologic setting ...... 49 3.2.2 Previous seismic results ...... 51 3.2.3 Previous MT results ...... 52 3.2.4 Anelasticity, grain size, and seismic velocity ...... 56 3.3 Methods ...... 57 3.3.1 Resistivity refinement and mapping to temperature ...... 57 3.3.2 Seismic velocity and attenuation calculations...... 59 3.4 Results ...... 59 3.5 Discussion ...... 64 3.5.1 Role of anelasticity in resolving thick thermal lithosphere ...... 64 3.5.2 Role of oxygen fugacity ...... 65 3.5.3 Body wave tomography and lithospheric discontinuities ...... 66 3.5.4 Origin of the Piedmont Resistor...... 68 3.5.5 Modern geodynamics of the southeastern United States ...... 70 3.5.6 Comparison to cratonic regions ...... 71 3.5.7 Implications for geophysical interpretation and geodynamic investigations ...... 73 3.6 Conclusions ...... 74 3.7 References ...... 75 Conclusions ...... 84 Bibliography ...... 85 Appendices ...... 101

LIST OF FIGURES

Figure Page

Figure 1.1: Overview map of the study region...... 5

Figure 1.2: Depth slices and cross sections through the preferred inverse solution...... 8

Figure 1.3: Vertical magnetic field transfer function data plotted as induction vectors...... 9

Figure 1.4: Isostatic gravity map for the southeastern United States...... 12

Figure 1.5: Resistivity as a function of temperature for olivine, orthopyroxene, and clinopyroxene...... 15

Figure 1.6: Possible model for the formation of the sub-Piedmont lithospheric resistor...... 21

Figure 2.1: Map showing distribution of EarthScope MT sites identified as clearly showing source bias, plotted over vertically integrated Earth conductance (15-150 km)...... 31

Figure 2.2: Transfer functions and cross-phase analyses from sites shown in Figure 2.1...... 36

Figure 2.3: Results from frequency domain robust principal components analysis of electric and magnetic fields from the array shown in Figure 2.1...... 37

Figure 3.1: Resistivity as a function of temperature for a range of compositions...... 47

Figure 3.2: Map of the region considered in this study...... 50

Figure 3.3: Comparison between MT and seismic surface-wave results...... 53

Figure 3.4: Map views of average Vs over 75-150 km depth from five surface wave tomography models and geometric mean resistivity over the same depth range...... 54

Figure 3.5: Depth profiles, obtained from the points shown in Figure 3.4, from the five surface wave models and the electrical resistivity inverse solutions...... 55

Figure 3.6: Final refined resistivity solutions and the temperature profiles used to generate the vertical maximum resistivity distribution...... 62

Figure 3.7: Comparison between Vs, Q-1, and MT-derived resistivity values...... 63

Figure 3.8: Comparison of predicted resistivity-Vs relationships that are computed with and without depth-dependent fO2 offset with respect to the FMQ buffer...... 66

Figure 3.9: Predicted body-wave velocities as a function of depth...... 68

Figure 3.10: Comparison between the geodynamic interpretation from seismic results alone and the geodynamic framework illuminated by jointly interpreting seismic and MT results. 70

LIST OF APPENDICES

Page Appendix 1: Supplemental Materials to Accompany Electrical Conductivity Structure of Southeastern North America: Implications for Lithospheric Architecture and Appalachian Topographic Rejuvenation ...... 102 A1.1 Preferred inverse model data fit ...... 103 A1.2 Constraints on maximum depth of the Piedmont resistor ...... 105 A1.3 Constraints on the horizontal extent of the Piedmont resistor ...... 108 A1.4 Constrains on sub-Appalachian lithospheric resistivity ...... 111 A1.5 Robustness of other features described in the main text...... 114 A1.6 Additional supporting information for the Appalachian Highland-Piedmont contrast ...... 118 A1.7 Consideration of tipper data ...... 121 Appendix 2: Supplemental Materials to Accompany Source Biases in Mid-Latitude Magnetotelluric Transfer Functions due to Pc3-4 Geomagnetic Pulsations ...... 123 A2.1 Cross-Phase Analysis ...... 124 Appendix 3: Supplemental Materials to Accompany Synthesizing Seemingly Contradictory Seismic and Magnetotelluric Observations in the Southeastern United States to Image Physical Properties of the Lithosphere...... 132 A3.1 Seismic surface-wave inverse solutions ...... 133 A3.2 Seismic velocity calculations ...... 134 A3.3 Modified resistivity inverse solutions ...... 136 A3.4 Geotherm calculations ...... 147 A3.5 Further resolution tests ...... 152

LIST OF APPENDIX FIGURES

Figure Page

Figure A1.1: Distribution of site-by-site nRMSE...... 103

Figure A1.2: Representative data misfits for the preferred inverse model and the forward models used to test the depth constraint on the Piedmont resistor...... 104

Figure A1.3: Spatial distribution of site misfits for the preferred inverse model and the forward models used to evaluate the depth extent of the Piedmont resistor...... 105

Figure A1.4: Cross sections through forward models that were used to constrain the depth extent of the Piedmont resistor...... 106

Figure A1.5: Plots of an inversion in which a region of the sub-Piedmont lithosphere was forced to remain conductive and the associate misfit...... 107

Figure A1.6: Resolution tests to constrain the northwestern limit of R1...... 109

Figure A1.7: Resolution tests to constrain the southeastern limit of R1...... 110

Figure A1.8: Results from a forward test in which cells in the preferred inverse solution between ~100 km and ~250 km were set to be at least 1000 Ωm...... 112

Figure A1.9: Results from a test in which all model cells beneath ~100 km were set to be at least 300 Ωm...... 113

Figure A1.10: Results from an inversion that includes resistive ocean lithosphere...... 115

Figure A1.11: Results from an inversion that includes resistive ocean lithosphere as well as conductive shelf sediments...... 116

Figure A1.12: Results from an inversion that started with a 30 Ωm half-space and that included the Atlantic Ocean and the Gulf of Mexico...... 117

Figure A1.13: Distributions of apparent resistivities calculated for off-diagonal components of the impedance tensor...... 119

Figure A1.14: Off-diagonal components of the impedance tensor at 1092 s plotted as apparent resistivity and phase...... 120

Figure A1.15: Observed vertical magnetic field transfer function data plotted as induction vectors and the corresponding forward-modeled induction vectors for various possible structures that are presented in the main text...... 122

Figure A2.1: Conceptual plots to illustrate the physics of geomagnetic field line resonances...... 125

Figure A2.2: Transfer functions and cross-phase analyses for a pair of sites located in the conductive Snake River Plain...... 126

LIST OF APPENDIX FIGURES (cont.)

Figure Page

Figure A2.3: Transfer functions for the two sites used in the cross-phase analyses shown in Figure 2.2...... 127

Figure A2.4: Plots of each component of the remote-reference estimate for the full impedance tensor of site SCV57...... 128

Figure A2.5: Different TF estimates for site SCV57...... 129

Figure A2.6: Comparison between robust remote reference transfer function estimates that excluded record times with clearly apparent Pc activity and the original robust remote reference transfer function estimates...... 130

Figure A2.7: Raw electric and magnetic field time series from EarthScope MT site IAN37...... 131

Figure A3.3.1: Inverse solutions obtained through the iterative resistivity redistribution process described in the main text...... 137

Figure A3.3.2: Map views of the iterative re-inversion process...... 138

Figure A3.3.3: Site-by-site misfits for each step in the iterative redistribution...... 139

Figure A3.3.4: Map of site-by-site misfits for the refined resistivity distributions...... 140

Figure A3.3.5: Depth slices at 120 km through the refined resistivity solutions...... 141

Figure A3.3.6: Locations of the sites used in the following data plots...... 142

Figure A3.3.7: Data fit from site SCW55 for the three refined inverse solutions...... 143

Figure A3.3.8: Data fit from site SCW56 for the three refined inverse solutions...... 144

Figure A3.3.9: Data fit from site SCV56 for the three refined inverse solutions...... 145

Figure A3.3.10: Data fit from site REV55 for the three refined inverse solutions...... 146

Figure A3.4.1: Mantle reduced heatflow calculation...... 148

Figure A3.4.2: Locations of the heatflow measurements used in the mantle reduced heatflow calculation...... 148

Figure A3.5.1: Resolution test that shows the effect of extending thick (~200 km) thermal lithosphere beneath the Coastal Plain...... 152

LIST OF APPENDIX TABLES

Table Page

Table A3.1.1: Information about the five seismic surface-wave models that we compare to our MT inverse solutions...... 133

Table A3.4.1: Heatflow measurements used for the reduced heatflow calculation...... 149

Table A3.4.2: Heatflow measurements used to calculate median surface thermal gradient over the Piedmont...... 150

Table A3.4.3: Temperature constraint points used to construct the quadratic spline for our third geotherm...... 151

DEDICATION

This work is dedicated to my mentor, colleague, and friend Alfred Kwok, who passed away on the side of Deerhorn Mountain in the Sierra Nevada in September, 2016. Ever since my first semester at Pomona College, Kwok believed I was capable of great things. Throughout my time at Pomona, he challenged me to excel in my intellectual pursuits. His support, encouragement, and friendship have had a profound impact on my life.

Introduction Although often overshadowed by the more geologically active west, the geologically “dead” eastern U.S. actually presents many perplexing and enigmatic geoscientific problems, including the questions of topographic rejuvenation and passive margin seismicity. Collection of long-period magnetotelluric (MT) measurements in the southeastern U.S. (SEUS) as part of the EarthScope program in 2015-2016 provided the opportunity to gain new insights into the geodynamics of this region via geophysical imaging of electrical conductivity. Our initial analysis of the MT data from the SEUS (Murphy & Egbert, 2017; Chapter 1) revealed a surprising, highly electrically resistive lithospheric block beneath the Piedmont and Coastal Plain physiographic provinces in this region. The only plausible interpretation of this geoelectric structure is that it represents cold, thick thermal lithosphere beneath the Piedmont and Coastal Plain (thermal lithosphere-asthenosphere boundary at >200 km depth). This conclusion appeared to be in strong disagreement with results from seismic tomography of the region; seismic images have shown slightly slow velocities beneath the southeastern US that are interpreted as being due to thin, warm, eroded lithosphere. Abundant resolution tests, however, demonstrate that the MT data require high resistivities (>300 Ωm) to at least 200 km depth in this region. Any changes to our inverse solutions to shrink or remove this highly resistive block significantly degrade our data fit to unacceptable levels. Consequently, as a given resistivity value provides an upper bound on temperature, these MT data require thick, cold lithosphere beneath the Piedmont. We initially attempted to explain the apparent contradiction between seismic and MT results by invoking the effects of unusual mantle compositions. Although both seismic velocity and electrical conductivity are fairly insensitive to bulk composition in the upper mantle, the former is slightly more sensitive to mineralogy and iron content (expressed as magnesium number, or similarly bulk melt fertility) than the latter. Therefore, we attempted to explain the relatively slow seismic velocities (with respect to references models) in the context of the thick, cold lithosphere required by our MT data by hypothesizing that the mantle lithosphere in the SEUS is fertile (in the bulk compositional sense) and rich in pyroxene minerals. Such an uppermost mantle composition could theoretically produce slow seismic velocities (relative to reference models) at the cold temperatures required by MT. However, very special geodynamic circumstances would be required to create such a mantle composition. We invoked formation of the Central Atlantic Magmatic Province (CAMP) as a means to produce such unusual conditions, although the scenario was somewhat contrived.

Despite the apparent contradiction between seismic and MT results, continued analysis (Murphy & Egbert, 2018; Chapter 2) of MT data from the SEUS has provided further support for highly resistive lithosphere in this region. We have demonstrated that short spatial scale geomagnetic pulsations (Pc’s) associated with field-line resonances (FLRs) can introduce subtle biases in MT transfer function data at spatial locations where the Earth is in bulk resistive. The pervasive presence of these biases in our SEUS MT data provides further support for the presence of thick, highly resistive lithosphere in this region. As the apparent strong contradiction between our MT findings and seismic images for the SEUS remained perplexing, we further explored the match-up between seismic and MT results by comparing these results together to laboratory experiments for seismic observables. We found that, in fact, there is no disagreement between the different geophysical techniques when using anelastic predictions of seismic observables as a function of temperature and pressure and when also taking into account subtle changes in mapping between electrical conductivity and temperature due to depth-dependent mantle oxygen fugacity (Murphy & Egbert, 2019; Chapter 3). Observed seismic velocities, as well as attenuation and receiver function observations, match predictions based upon the lithospheric thermal field required by our MT data very well when using models of anelastic controls on seismic parameters. There only seemed to be a discrepancy between geophysical observations because the seismic results had only been interpreted in comparison to global seismic reference models. Those reference models do not map to a meaningful physicochemical reference state, so comparison to them does not provide a meaningful or unique inference of lithospheric properties. Together, the available seismic and MT data support the view of thick, coherent thermal lithosphere beneath the SEUS. Much seismic research has proposed an interpretation of edge convection and associated lithospheric erosion along the passive margin of eastern North America; however, we have demonstrated that these geophysical datasets together preclude such processes in the SEUS. We interpret the present lithospheric state to be due to formation of CAMP, which comprises one of the largest large igneous provinces in the world. Eruption of such a large volume of material would have thoroughly depleted the mantle and would have consequently produced a large block of buoyant, rheologically strong material that, resistant to convection, would have conductively cooled to lithospheric temperatures. Our synthesis of seismic and MT observation requires no special mantle compositions to have formed with eruption of CAMP, so the anelastic calculations provide a more natural way of integrating the geophysical results.

Chapter 1

Electrical Conductivity Structure of Southeastern North America: Implications for Lithospheric Architecture and Appalachian Topographic Rejuvenation

Benjamin S. Murphy & Gary D. Egbert

Earth and Planetary Science Letters Volume 462, 15 March 2017, Pages 66-75

1.1 Abstract We present the first three-dimensional view of the lithospheric electrical conductivity structure beneath southeastern North America. By inverting EarthScope long-period magnetotelluric (MT) data, we obtain an electrical conductivity image that provides new insights into both the architecture of the Appalachian Orogen and the cryptic post-rifting geodynamic history of the southeastern United States. Our inverse solutions reveal several elongate electrically conductive features that we interpret as major terrane sutures within the Appalachian Orogen. Most significantly, we resolve a highly electrically resistive layer that extends to mantle depths beneath the modern Piedmont and Coastal Plain physiographic provinces. As high resistivity values in mantle minerals require cold mantle temperatures, the MT data indicate that the sub-Piedmont thermal lithosphere must extend to greater than 200 km depth. This firm bound conflicts with conclusions from seismic results. The boundary between the anomalously thick, resistive sub-Piedmont lithosphere and the relatively thin, moderately conductive sub-Appalachian lithosphere corresponds within resolution to the modern Appalachian topographic escarpment. This newly recognized contrast in lithospheric properties likely has important implications for Appalachian topographic rejuvenation.

1.2 Introduction Beneath the dense vegetation and the thick accumulations of saprolite, the crust and mantle lithosphere of the southeastern United States (Fig. 1.1) record a complicated history of accretionary and collisional orogenies, continental rifting, and passive-margin evolution. Although often scientifically overshadowed by the more active west, the eastern margin of North America nevertheless provides a natural laboratory for studying cryptic geodynamic and tectonic problems. The recent acquisition of EarthScope seismic and magnetotelluric (MT) data on the east coast of the United States now permits unprecedented three- dimensional views of the lithosphere beneath this enigmatic region. The ancestral Appalachian mountains were the product of a full Wilson cycle that spanned the late Mesoproterozoic to the late Paleozoic (Hatcher, 2010). The Grenville Orogeny at ~1.1 Ga marked the assembly of Rodinia, which persisted in various forms into the Neoproterozoic. Following Iapetan rifting of this supercontinent in the latest Neoproterozoic to early Cambrian, a series of Paleozoic terrane-suturing events in eastern Laurentia culminated in the Alleghenian Orogeny, which marked the assembly of Pangea and the closure of the Rheic Ocean. In the southern Appalachian Orogen, two pre-Pangean accretionary events are generally recognized: the mid-Ordovician Taconic Orogeny and the

5 late-Ordovician to Silurian Cherokee Orogeny, in which the Carolina Superterrane (Carolina Zone) docked onto Laurentia (Hibbard et al., 2010; Fig. 1.1). In the mid-Mesozoic, Pangea was rifted apart with the opening of the Atlantic Ocean. During this time, at ~200 Ma, the Central Atlantic Magmatic Province (CAMP; e.g., Whalen et al., 2015) formed as the largest continental large igneous province so far recognized in the geologic record. Geochemical evidence indicates that a mantle plume did not generate CAMP; rather, these flood basalts are likely due to passive continental rifting associated with orogenic collapse (Whalen et al., 2015; Frizon de Lamotte et al., 2015). With growth of the Atlantic Ocean, eastern Laurentia became a passive margin. The region has experienced no major tectonism since the mid Mesozoic.

Figure 1.1: Overview map of the study region. Orange inverted triangles indicate EarthScope MT sites that constitute our dataset. Black solid lines delineate major terrane boundaries as determined from geologic datasets (adapted from Hatcher, 2010). Black dotted line indicates the recently reinterpreted surface trace of the Suwannee suture (Boote and Knapp, 2016; Hopper et al., 2016). White dashed lines divide the major physiographic provinces referenced in the text (adapted from USGS physiographic province dataset). Basemap shows modern Southern Appalachian topography. Abbreviations for US states that are shown on the map are as follows: VA, Virginia; WV, West Virginia; KY, Kentucky; TN, Tennessee; NC, North Carolina; SC, South Carolina; GA, Georgia; AL, Alabama.

Multiple lines of evidence suggest that the modern Appalachians Mountains are the result of topographic rejuvenation that occurred after the ancestral Alleghenian highlands had been eroded away. Geomorphic evidence indicates that the topography of the Southern and Central Appalachians is not in equilibrium (Gallen et al., 2013; Miller et al., 2013; Prince and Spotila,

2013). Apatite (U-Th)/He thermochronology suggests that topographic relief increased in the Southern Appalachians in the late Cretaceous (McKeon et al., 2014). Most importantly, the Cenozoic stratigraphic record on the Atlantic shelf shows major pulses of sedimentation in the Cretaceous and the Miocene that likely represent periods of topographic rejuvenation (Poag and Sevon, 1989). Multiple mantle-driven mechanisms for such low-magnitude topographic disequilibrium and rejuvenation have been proposed, including both dynamic subsidence and dynamic uplift due to large-scale mantle flow (Moucha et al., 2008; Spasojević et al., 2008; Liu, 2014) and hydration of the mantle beneath eastern Laurentia due to dewatering of the deep Farallon slab (van der Lee et al., 2008). Surficial processes, such as stream capture and basin reorganization (e.g., Willet et al., 2014), have also been suggested as an explanation for topographic disequilibrium. Despite much work on the history and development of modern Appalachian topography, no single explanation has yet to be widely accepted. Here, we use recently collected EarthScope long-period MT data from the southeastern United States (Fig. 1.1) to provide a new geophysical perspective on the geologic architecture and geodynamic evolution of this region. Our work expands significantly on previous two- dimensional MT studies of the Appalachian Orogen (Ogawa et al., 1996; Wannamaker, 2005), which imaged several of the features we observe here but which did not thoroughly explore the implications of those features due to the limitations of their data. With the EarthScope MT dataset, we are now able to better constrain and interpret lithospheric electrical structures.

1.3 Data and Methods Our dataset comprises full MT impedance tensors and vertical magnetic field transfer functions in the period range of ~10 s to ~10,000 s from 124 Earthscope Transportable Array (TA) long-period MT sites in the southeastern United States (Fig. 1.1). These data are available for community use through IRIS (http://ds.iris.edu/spud/emtf). We prescribed error floors of 5% for impedance tensor components (treating the two rows separately) and 0.03 for the real and imaginary parts of the vertical magnetic field transfer function components. We inverted the data with ModEM, a three-dimensional isotropic electromagnetic inversion code (Kelbert et al., 2014). Nominal horizontal grid-cell size for our model domain is 12 km. Vertical grid-cell size telescopes logarithmically to a maximum model depth of 900 km. Cell thickness is ~200 m at the surface; vertical grid-cell size is ~50 km at 300 km depth. Because the typical site spacing is ~70 km and the shortest period in our dataset is ~10 s, our model

7 parameterization is sufficiently fine to model crust- and lithosphere-scale structures as well as the effects of distortion. We use a nested inversion approach to account for possible effects of large-scale electrical structures beyond our data footprint. The larger model within which our inversions were embedded was derived from a coarse-scale inversion of all available EarthScope MT data in the eastern half of the United States. This large model included both the Atlantic Ocean and the Gulf of Mexico. We assess gross model resolution through multiple inversion runs that utilized a range of prior resistivity models and that had various constraints imposed. Note that, in the ModEM regularization approach, deviations from a prior model are penalized (Kelbert et al., 2014); therefore, only features that appear for a range of prior models are considered required by the data. Results from an illustrative selection of inversion runs are shown in the Appendix 1. For the preferred solution shown here, a ModEM smoothing setting of 0.2 with two passes was utilized. In the inversion code, model smoothing is part of the inversion regularization. For our model discretization and covariance settings, features are smoothed over a length scale of approximately 30 km. This spreads bulk resistivity values into adjacent cells to the extent that regional-scale features are easily identified while also providing reasonable localization of those features. The preferred inversion began with a prior half-space resistivity of 180 Ωm that included ocean cells frozen to 0.3 Ωm. A normalized root-mean-square misfit of 1.75 was achieved after 350 iterations of the ModEM inversion. Further discussion of the data fit as well as resolution tests are presented in Appendix 1.

1.4 Results Our preferred inverse solution is shown in Figure 1.2. We identify five well resolved electrically conductive structures, denoted C1 through C5; three poorly resolved but significant electrically conductive features, identified EC1 through EC3; and two zones of high resistivity, denoted R1 and R2 (Fig. 1.2). The features presented here are those that we currently consider robust in at least some capacity based upon multiple inversions and resolution tests (see Appendix 1). Structures C1, C2, and C3 are major Appalachian-parallel, mid- to lower-crustal conductors. These three structures follow the northeast-southwest, slightly sinuous grain of the Appalachian Orogen. The average vertically integrated conductances (~25-50 km depth) are 3300 S for C1, 2800 S for C2, and 1500 S for C3. (Average conductances are calculated for the spatial footprint of each conductor as defined in Figure 1.2.) The signature of highly conductive structure C1 is apparent in the vertical magnetic field transfer functions. These

Figure 1.2: Depth slices and cross sections through the preferred inverse solution. Each cross section is paired with the corresponding topographic profile. Labels that begin with C denote resolved conductive structures that are described in the text; labels that begin with EC denote unresolved conductive features mentioned in the text; and labels that begin with R denote resistive structures discussed in the text. The off-shore southeastern limit of R1 is poorly resolved at present; see Figure A1.7 and the accompanying text for a discussion of resolution. data are plotted in Figure 1.3 as induction vectors (e.g., González-Castillo et al., 2015), which (with the Parkinson sign convention used here) point to anomalous conductors. C1 was

9 observed in a previous two-dimensional MT study of the Appalachian Orogen (Wannamaker, 2005). C4 is a weak, east-west trending conductive feature in the mid to lower crust of central Georgia. Its average vertically integrated conductance (~25-50 km depth) is 210 S; this is marginally greater than a conductance of 190 S for the same region of the prior model. Although the conductivity contrast with the surrounding model domain is minimal, this structure appears in all inverse solutions and is therefore considered to represent a real feature. Structure C5 is a robust conductor in the mid to lower crust of central South Carolina that, at least at mid-crustal depths, weakly connects southwestward to C4 and the southernmost extent of C3.

Figure 1.3: Vertical magnetic field transfer function data (tippers) at 1092 s plotted as induction vectors with the Parkinson sign convention. When plotted in this manner, these arrows point towards anomalous conductivity contrasts. These data delineate a feature along which the induction vectors change direction by 180° in the northwestern portion of our data coverage. This linear feature corresponds to the trace of the New York-Alabama (NY-AL) Lineament. The induction vectors from the vast majority of the southern sites in our dataset point towards a major conductivity contrast beyond our spatial footprint to the southwest, in southern Alabama or the northern Gulf of Mexico.

Anomalies EC1, EC2, and EC3 appear consistently in inversions (Figs. A1.10-12) beyond the footprint of our dataset. Although the exact positions and geometries of these structures are not resolved due to present limits on data coverage, their persistence in inversions suggests that they do represent real electrically conductive features outside of the current array. EC1 appears offshore of the study region, beneath the continental shelf; EC2 is imaged in southeastern Alabama along the border with Georgia; and EC3 lies to the north of

10 the array along the West Virginia-Virginia border. Induction vectors (Fig. 1.3) point to EC2 and EC3 and are therefore consistent with the high conductivities observed in the inverse solutions in these areas. Perhaps the most enigmatic feature that appears in all inverse solutions is the large, thick block of resistive material beneath the Piedmont and Coastal Plain physiographic regions (Figs. 1.1, 1.2). We refer to this structure, which was observed in a previous two-dimensional MT study of the region (Wannamaker, 2005), as R1. In the preferred inverse solution (Fig. 1.2), high resistivity values associated with this structure extend to depths greater than 300 km. Resolution tests (see Appendix 1) indicate that this feature must maintain bulk resistivity values of at least 1000 Ωm to a depth of at least 200 km. Strikingly, the northwestern limit of structure R1 coincides with the modern Appalachian topographic escarpment (Figs. 1.1, 1.2; see also Figs. A1.6-7 for resolution tests). At ~200 km depth, the region to the northwest of the escarpment displays resistivity values on the order of 100 Ωm, whereas the region to the southeast displays resistivity values higher than 1000 Ωm in association with structure R1 (see Figs. A1.8-9 for resolution tests). This sharp boundary in deep resistivity is clearly evident in plots of apparent resistivity and phase in an Appalachian-parallel coordinate system (Figs. A1.13-14). Therefore, both the data themselves and multiple inversions indicate the presence of a major conductivity contrast between the mantle beneath the modern Appalachian highlands and the mantle beneath the Piedmont and Coastal Plain. Finally, structure R2 is a small Appalachian-parallel resistive feature that appears in the lowermost crust and uppermost mantle beneath western North Carolina. This structure exhibits resistivity values up to 1000 Ωm and is separated from structure R1 by a zone of moderate electrical conductivity that extends downward from structure C3.

1.5 Discussion 1.5.1 Well Resolved Shallow Structures Structures C1, C2, C3, C4, and C5 are spatially coincident with terrane sutures inferred from studies of surface geology. As sutures have frequently been associated with elongate crustal conductors (e.g., Boerner et al., 1996; Jones et al., 2005; Wannamaker, 2005; Yang et al., 2015), we interpret these conductive structures as representing the major terrane sutures of the Southern Appalachian Orogen. The high conductivities associated with these structures are likely due to graphitic or authigenic-sulfide-rich sediments that became trapped in these zones of crustal convergence (e.g., Boerner et al., 1996; Wannamaker, 2005). We interpret

C1 to represent the main Grenville suture; C2, the Taconic suture; and C3, the main Carolina Superterrane suture. We interpret C4 and C5 to be associated with the Suwannee terrane. Structure C1 coincides with the New York-Alabama (NY-AL) Lineament (Fig. 1.3), a continent-scale aeromagnetic anomaly (e.g., Steltenpohl et al., 2010) that, based upon regional contrasts in radiogenic isotope affinities, has been interpreted to represent the main Grenville suture (Fisher et al., 2010). Our electrical resistivity images therefore provide strong geophysical support for this geologically motivated interpretation. Both the induction vectors (Fig. 1.3) and inverse solutions indicate that structure C1 is highly conductive, more so than any other feature in the model. Repeated reactivation of the corresponding fault system to accommodate both orogen-perpendicular and orogen-parallel deformation throughout the latest Neoproterozoic and the Paleozoic (Steltenpohl et al., 2010) likely helped to concentrate and connect conductive phases and thereby increase bulk conductivity over time (e.g., Wannamaker, 2005). Recent research indicates that the Suwannee suture likely lies to the north of its previously defined trend (Fig. 1.1; Boote and Knapp, 2016; Hopper et al., 2016; cf. Hatcher, 2010). Therefore, C5 may be part of the reinterpreted Suwannee suture and C4 may be a suture within the larger Suwannee terrane. (The downward continuation of C5 into the mantle shown in the preferred inverse solution is not robust; cf. Figs. A1.10-12.) However, whereas seismic receiver function images indicate that the Suwannee suture dips shallowly through the crust, C5 appears at mid- to lower-crustal depths almost directly beneath the inferred surface trace of the redefined Suwannee suture. Resolution tests indicate that depth is poorly constrained for at least the weak southwestern extension of structure C5 (Figs. A1.10-12), so this discrepancy may be due to the poor upper-crustal resolution of these long-period MT data. Curiously, the footprint of C5 is spatially coincident with an isolated relative isostatic gravity low (Figs. 1.2, 1.4). Structure C5 may therefore hold additional significance beyond representing a terrane suture. Further analysis is needed to explore the nature of this structure, especially as its relationship with C3 and C4 is complex. Incorporation of broadband MT data would greatly clarify these crustal-scale structures. These conductive features (at least C1, C2, and C3) define lateral limits on the extent of rocks of different terrane affinities in the mid to lower crust. In particular, our results indicate that Grenvillian rocks do not extend southeastward in the lower crust beneath the Piedmont and Coastal Plain, as has been suggested previously (e.g., Cook and Vasudevan, 2006). Furthermore, the interpreted mid- to lower-crustal electrical expressions of the Grenville, Taconic, and Carolina sutures lie almost directly beneath their surface expressions. Therefore,

12 based on the MT images presented here, these sutures appear to extend essentially continuously through the entire crust in the Southern Appalachian orogen. This structural model stands in contrast to that from other work (e.g., Cook and Vasudevan, 2006; Hibbard et al., 2002) that has interpreted the upper-crustal expression of these sutures to be significantly offset (~400 km) to the northwest from their continuation at depth. Our conclusions here stand at least for the Appalachian orogen north of the (redefined) Suwannee suture; because compelling seismic images indicate that the Suwannee suture is a low-angle detachment (Hopper et al., 2016), crustal architecture is likely far more complicated at the southernmost end of the orogen. Further detailed exploration of the geologic implications of these structures is left for future study.

Figure 1.4: Isostatic gravity map for the southeastern United States (from Phillips et al., 1993; data available at http://mrdata.usgs.gov/geophysics/gravity.html). Note the relative isostatic gravity low that spatially corresponds to electrically conductive structure C5. The major Appalachian-parallel isostatic gravity high coincides with electrically resistive structure R1.

Resistive structure R2 appears approximately coincident with a seismically detected westward-dipping mid-lithospheric discontinuity (Wagner et al., 2012). Previous studies have interpreted this seismic structure as a dangling remnant of either the Iapetan slab from accretion of the Carolina Superterrane (Wagner et al., 2012) or the Rheic slab after closure of the ocean between Laurentia and Gondwana (Whalen et al., 2015). However, because Carolinian subduction is believed to have been directed westward (Hibbard et al. 2010), neither of these previous explanations seems geometrically consistent with our resistivity

13 images, which show structure R2 to the northwest of the downward continuation of what we interpret to be the Carolina Superterrane suture (C3). Alternatively, this feature could be a dangling remnant of the Iapetan slab that was trapped during the Taconic Orogeny. However, this interpretation is also geometrically problematic, as Taconic subduction is believed to have been directed to the east (e.g., Abbott and Greenwood, 2001) and structure R2 appears to the southeast of the conductive signature of the Taconic suture (C2). This structure therefore remains enigmatic.

1.5.2 Poorly Resolved Structures The high bulk conductivity values associated with EC1 in the preferred inverse solution (Fig. 1.2) do not appear in all solutions (cf. Figs. A1.10-12). This feature is likely an edge effect due to conductive shelf sediments or some other unresolved offshore, crustal-scale conductive structure. Such a conductor could correspond to the Alleghenian suture, which would mark the Laurentian boundary along which Pangea was assembled. However, EC1 appears in inversions southeast of the Suwannee suture, which is the Alleghenian suture in the southernmost Appalachian orogen (e.g., Boote and Knapp, 2016). This unresolved conductor could alternatively represent a structure within the Suwannee terrane. The coherent, large amplitude, southwest-pointing induction vectors that cover most of Georgia and South Carolina (Fig. 1.3) require the presence of a major electrical conductivity contrast in the crust or upper mantle beneath southern Alabama or the northernmost Gulf of Mexico. Such a structure is represented in inverse solutions as EC2, although the characteristics of this feature are unresolved. Recent surface wave tomography (Pollitz and Mooney, 2016) shows an isolated slow seismic anomaly in southern Alabama that may be due to lithospheric modification associated with passage of the Bermuda hotspot track underneath southern Laurentia at the end of the Cretaceous (Cox and Van Arsdale, 2002). Plume impingement could increase lithospheric conductivity by introducing water or even mantle sulfides into the constituent mantle rocks. However, slow shear-wave velocities in southern Georgia (Pollitz and Mooney, 2016) that may also be associated with this hot-spot track spatially coincide with resistive structure R1 in our inverse solutions, which possibly conflicts with what would be expected if hot-spot refertilization had significantly altered the lithosphere in this region. Alternatively, the thick sediment package in the Gulf of Mexico or the edge of resistive structure R1 could be responsible for the pattern in the induction vectors and for the appearance of EC2 in inverse solutions. Forward-modeling tests suggest that either enhanced mantle conductivity or a sharp southwestern edge on R1 may explain the

14 induction vectors (Fig. A1.15); however, further expansion of the EarthScope MT dataset into Alabama will be necessary to resolve the lithospheric electrical conductivity structure of this area. Structure EC3 coincides with a region of Eocene volcanism that hosts the youngest volcanic rocks found on the eastern margin of the United States (e.g., Mazza et al., 2014). This volcanism has been ascribed to asthenospheric upwelling associated with a delamination event (Mazza et al., 2014), although hotspot activity is an equally compelling explanation (Chu et al., 2013). As EC3 lies at the edge of our current data footprint, at present our data cannot clearly distinguish between these two competing hypotheses.

1.5.3 Deep Lithospheric Conductivity Contrast and the Piedmont Resistor We interpret resistive structure R1 as representing thick sub-Piedmont lithosphere. In the discussion that follows, we rely on the thermal definition of the lithosphere: the thickness of mantle and crustal material that lies above the ~1300°C isotherm.

1.5.3.1 Constraints from the Magnetotelluric Data: Thick Sub-Piedmont Lithosphere In our preferred inverse solution, the lithosphere beneath the modern Appalachian highlands displays resistivity values of ~100-300 Ωm, which can be explained by moderate hydration (~10-100 ppm H2O, Fig. 1.5; Selway, 2014; Gardés et al., 2014) at lower- lithosphere temperatures of ~1000° – 1200° C. The moderately conductive lithosphere beneath the modern Appalachian highlands only extends to depths of ~150-200 km, as typical asthenospheric conductivity values for moderately hydrated mantle minerals (~30 Ωm, Fig. 1.5; Constable, 2006; Fullea et al., 2011) appear in the inverse solution by a depth of 200 km in this region. Although mantle minerals could display resistivity values of ~30 Ωm at lithospheric temperatures with more than ~500 ppm incorporated H2O (Gardés et al., 2014), such water contents would be at or above the solubility limit at pressures corresponding to 200 km depth (~6 GPa; Ardia et al., 2012). Additionally, if Appalachian lithosphere were to extend to ~250-300 km depth, temperatures at ~100 km would be ~600° C for a reasonable conductive geotherm and the uppermost mantle lithosphere would then need to be hydrated with >1000 ppm H2O in order to produce resistivity values of ~100 Ωm (Fig. 1.5; Yoshino et al., 2009; Jones et al., 2012; Dai and Karato, 2014; Gardés et al., 2014). Such levels of hydration are unrealistic (Keppler and Bolfan-Casanova, 2006). Based on these considerations, the MT data suggest that the modern Appalachian thermal lithosphere is no more than ~150-200 km thick.

Figure 1.5: Resistivity as a function of temperature for olivine (ol), orthopyroxene (opx), and clinopyroxene (cpx). As these three minerals are the dominant phases in the upper mantle, they will control mantle resistivity values (Fullea et al. 2011; Selway 2014). Black lines are for dry olivine and demonstrate the minimal effect that compositional variations have on electrical conductivity (compositional effects are similarly small for pyroxenes). Red and green lines are representative of dry orthopyroxene and clinopyroxene resistivity, respectively. Blue lines demonstrate the effects of adding small amounts of water to mantle minerals. The blue-shaded region denotes the resistivity values that are observed for structure R1. The green-shaded region denotes resistivity values that are observed for the region beneath the modern Appalachian highlands. The curves plotted here are representative of mantle mineral resistivities as a function of temperature. Other calibrations differ slightly but are generally in agreement with those shown here. These calibrations are from Fullea et al. (2011), Jones et al. (2012), and Gardés et al. (2014).

In contrast, the lithosphere beneath the Piedmont and Coastal Plain (R1) must extend to depths greater than 200 km and must be minimally hydrated. For all volumetrically significant mantle minerals (those that will dominantly control electrical conductivity), the maximum resistivity that could possibly be observed at 1300° C (the approximate temperature at the LAB) under dry conditions is ~300 Ωm (Fig. 1.5; Fullea et al., 2011; Jones et al., 2012; Selway, 2014). Even small amounts of incorporated water (~10 ppm H2O) will lower the maximum possible temperature (Fig. 1.5). Therefore, the resistivity of dry mantle

16 minerals provides an upper limit on temperature as well as a lower limit on sub-Piedmont thermal lithospheric thickness. As the data require resistivity values of ~1000 Ωm to a depth of at least 200 km, temperatures at that depth must be ≤1200° C and the LAB beneath the Piedmont must therefore lie deeper than 200 km. As electrical conductivity of dry mantle minerals is controlled primarily by temperature alone (Constable, 2006; Fullea et al., 2011), this conclusion is robust. At lithospheric temperatures and in the absence of interstitial water, thermally activated small-polaron hopping, which amounts to valence swapping between Fe2+ and Fe3+ within the crystal lattice (Elliot, 1998), is the dominant conduction mechanism in olivine, orthopyroxene, and clinopyroxene (e.g., Selway, 2014). Although this mechanism is dependent on iron content and oxygen fugacity to a limited extent (these control the number of charge carriers), temperature (which controls carrier mobility) is the overwhelmingly dominant factor that determines electrical conductivity (Fig. 1.5; Constable, 2006; Fullea et al., 2011). Since dry mantle minerals display minimal electrical anisotropy (e.g., Du Frane et al., 2005), this conclusion does not depend on mantle fabric. As electrical conduction in dry mantle silicates is well studied and understood in terms of both laboratory results (e.g., Constable 2006) and solid-state physics theory in general (e.g., Elliot, 1998), our bound on sub-Piedmont lithospheric temperature and, by extension, thermal lithospheric thickness rests on a solid foundation. Our data thus strongly suggest the presence of a step in the LAB between the modern Appalachian highlands and the Piedmont. This coastward increase in LAB depth juxtaposes lithospheric resistivity values that would be expected for Phanerozoic lithosphere and anomalously high lithospheric resistivity values.

1.5.3.2 Comparison to Seismic Images: Minor Agreement and Major Conflict Our observation of a slightly hydrated lithosphere beneath the modern Southern Appalachian topographic highs is compatible with results from a study of local Pn travel times that indicate uppermost-mantle hydration in this region (MacDougall et al., 2015). Our depth range for the LAB beneath the Appalachian highlands also agrees with estimates of local lithospheric thickness from body- and surface-wave tomography (Biryol et al., 2016; Pollitz and Mooney, 2016). However, the thick sub-Piedmont lithosphere that we observe with the MT data stands in sharp conflict with images derived from body- and surface-wave tomography (Schmandt and Lin, 2014; Pollitz and Mooney, 2016; Biryol et al., 2016). Mantle lithosphere that extends beyond 200 km depth would follow an anomalously cold conductive

17 geotherm and would therefore be expected to be clearly visible in body-wave travel times as a well-defined fast anomaly. However, no such anomaly is observed in seismic travel-time datasets or in body-wave tomography inversions (Schmandt and Lin, 2014; Biryol et al., 2016). Continent-scale body-wave tomography (Schmandt and Lin, 2014) shows that mantle velocities in this region deviate very little from the reference model. Local compressional body wave tomography (Biryol et al., 2016) shows heterogeneous mantle compressional velocities that generally exhibit small deviations (<1%) from the reference model, although more significant slow-velocity anomalies (up to ~3%) are also detected beneath the Piedmont and Coastal Plain. While local surface-wave inversions (Pollitz and Mooney, 2016) do not image an anomalous feature that could correspond to this large lithospheric resistor, more recent continent-scale surface-wave work (Shen and Ritzwoller, 2016) does appear to show relatively high shear wave velocities beneath the Piedmont and Coastal Plain and lower shear wave velocities beneath the adjacent Appalachians. Shear-wave splitting measurements show a contrast in behavior between the two lithospheric domains that we image here, although this pattern could be the result of overprinting by shallow lithosphere structures, such as sutures (Long et al., 2016). Measurements of local Pn travel times do indicate higher seismic velocities beneath the Piedmont than beneath the modern Appalachians (MacDougall et al., 2015), but this contrast could be characteristic of only the shallowest mantle (Biryol et al., 2016). It is difficult, but not impossible, to reconcile the seismic and the MT results. While temperature is the first-order control on seismic velocity in the upper mantle, the effects of lithology or composition may be more important in defining anomalies than is widely appreciated. For example, an increase in the iron content of mantle rocks and a decrease in ambient temperature have opposite effects on seismic velocities (Cammarano et al., 2003; Lee, 2003). That is, a decrease in magnesium number by 4-5 units can offset effects of decreasing temperature by 100-200° C on seismic velocities so that the net deviation from a reference model is roughly zero. Changes in mineralogy can also affect seismic velocities.

For example, a decrease in olivine content from 100% to 60% will decrease VP by ~1% (Lee,

2003), and enrichment in orthopyroxene (up to ~50%) in particular can further lower VP by up to 1% (Schutt and Lesher, 2010). Furthermore, mantle pyroxenite can be invisible to body- wave tomography, as under certain circumstances cold pyroxenite may exhibit similar seismic velocities to fertile asthenospheric (i.e., hot) peridotite (Erdman et al., 2016, their Fig. 3). As the MT data lead to the seemingly inescapable conclusion that sub-Piedmont mantle

18 temperatures are colder than would be expected, careful consideration of possible compositional effects on seismic velocity is warranted.

1.5.3.3 Possible Explanation: A Metasomatized Lithospheric Root that Regrew after Delamination A major orogen-parallel, post-Alleghenian delamination event has been invoked to explain the generation of CAMP lavas (Whalen et al., 2015). If the resulting void in the mantle lithosphere were filled by material that formed from asthenospheric consumption of the delaminant or that trapped significant volumes of partial melt, then the regrown, now-cold lithospheric root could plausibly be electrically resistive but not particularly seismically anomalous. The model that we propose here is shown graphically in Figure 1.6. Foundering of the subducted Rheic slab may have led to the initiation of Pangean rifting in this region prior to 200 Ma, but a large, pan-Appalachian delamination at 200 Ma spurred CAMP magmatism and accelerated the rifting of Pangea (Whalen et al., 2015). This delamination would have occurred beneath the ancestral orogenic hinterland, which lies beneath the modern Piedmont and Coastal Plain, and would have removed the lowermost, likely eclogitized crust. Fertile asthenospheric peridotite would then upwell to fill the void left by this delamination. If the regrown sub-Piedmont lithosphere were highly fertile, so that the constituent rocks have very low magnesium numbers; rich in pyroxenes, so that any constituent lherzolites could almost be classified as olivine websterites; and/or rich in mantle pyroxenite, then the structure that we observe here could be seismically invisible (Cammarano et al., 2003; Lee, 2003; Schutt and Lesher, 2010; Erdman et al., 2016). If the step in the LAB between these two lithospheric domains were ~50 km, then for reasonable conductive geotherms the temperature contrast would be small enough for compositional differences to offset the temperature difference. Such unusual mantle compositions could reconcile the MT and the seismic results as long as the regrown lithosphere were minimally hydrated. Excess mantle water could have been removed during CAMP magmatism. The heterogeneity observed in local body-wave tomography (Biryol et al., 2016) may reflect compositional variations in this regrown lithosphere root. Such atypical mantle lithologies could form if the regrown lithosphere trapped abundant asthenospheric melts or if the upwelling asthenosphere consumed significant amounts of the delaminated lithosphere. The Alleghenian lithosphere was likely hydrated to some extent; lithospheric absorption of asthenospheric melts that were generated during dehydration of the

19 delaminant (e.g., Elkins-Tanton, 2005) could result in fertilization of the regrown sub- Piedmont lithosphere. If the Alleghenian lithosphere were already warm, as would be expected following orogenesis, and if it detached slowly, then melting of the delamination itself would be expected (Wang and Currie, 2015). Indeed, the geochemistry of CAMP lavas indicates a shallow-mantle pyroxenite source (Whalen et al., 2015), so delaminated material must have melted to feed this continental flood basalt. By assimilating this melt, the upwelling asthenosphere could have become enriched in a pyroxene component or effectively metasomatized. Global geochemical studies indicate that assimilation of eclogitic material can indeed produce abundant secondary mantle pyroxenite (e.g., Sobolev et al., 2007). With the transition from an active rift zone to a passive margin, the eastern margin of North America would have cooled and subsided. Especially because particularly fertile lithosphere would be relatively dense (Cammarano et al., 2003; Lee, 2003), subsidence of the regrown root could have resulted in a depressed LAB that is still intact today (Fig. 1.6). Stability of this thick root could be maintained if the constituent mantle rocks remain dry, as the MT data indicate they are, and thereby strong. Further work will be necessary to rigorously test the geodynamic and geochemical requirements and consequences of this possible model. Seismic studies suggest that the crust beneath the modern Appalachian highlands is thicker than the crust beneath the Piedmont lowlands (Hawman et al., 2012; Schmandt et al., 2015; Shen and Ritzwoller, 2016). The model proposed here could provide an explanation for thinner crust in the ancestral orogenic hinterland compared to the ancestral foreland, as the lower Piedmont crust would have been removed during the delamination event. An isostatic gravity high that spatially coincides with the highly resistive structure imaged here (R1; Figs. 1.2, 1.4; Simpson et al., 1986) also supports the existence of exceptionally fertile, dense lithosphere beneath the Piedmont. Although this Appalachian- parallel isostatic high has been interpreted as due in part to mafic lithologies such as CAMP lavas, a single coherent explanation for its origin has been lacking (Simpson et al., 1986). The model that we propose here provides such an explanation for this gravity signal. Thermal measurements from the southeastern United States show that heat flow is generally higher in the Piedmont than in the adjacent Appalachians (Smith et al., 1981), seemingly in contrast to what would be expected for the thick sub-Piedmont thermal lithosphere that we image here. However, differences in heat production between the rocks of the Appalachian crust and the rocks of the Piedmont crust could mask the deeper mantle heatflow signature. Alternatively, higher heat production within the thick, fertile sub-Piedmont

20 lithosphere could yield a higher heat flow than from the adjacent sub-Appalachian lithosphere. In this scenario, the temperature difference between the sub-Piedmont mantle and the adjacent sub-Appalachian mantle could be diminished by high internal heat production in the fertile sub-Piedmont lithospheric root. This low thermal contrast would then certainly allow compositional effects to exert significant control on seismic wave speeds. Recent seismic body-wave tomography has been used to argue for continuing piecemeal delamination of the lithosphere beneath the southeastern United States (Biryol et al., 2016). Based on our results here, particularly our observation of the coherently resistive nature of the sub-Piedmont lithosphere, we cannot support the interpretation that delamination events occurred throughout the Cenozoic and are still occurring at present.

1.5.3.4 Implications for Appalachian Topographic Rejuvenation Regardless of the exact nature and history of this anomalously thick sub-Piedmont lithosphere, its existence does indicate a major, previously unappreciated contrast in the upper mantle beneath the southeastern United States. The spatial correspondence between this lithospheric boundary and the modern Appalachian topographic scarp (Fig. 1.2) is likely not a coincidence; rather, this contrast in lithospheric properties may hold important consequences for Appalachian topographic rejuvenation. Edge-driven, small-scale convection at the boundary between relatively thin Appalachian lithosphere and relatively thick Piedmont lithosphere could have driven the episodes of topographic rejuvenation that are suggested by the geomorphic, thermochronologic, and Atlantic sedimentary records (Fig. 1.6). Shear-induced flow associated with overall mantle motion beneath eastern North America could drive such a small convection cell (Fig. 1.6; Kaislaniemi and van Hunen, 2014). Vertical mantle flow associated with edge convection in this region may be responsible for elongating the edge of this resistive anomaly with depth (Kaislaniemi and van Hunen, 2014), so that it may in fact extend to depths significantly greater than 200 km. Hydration of the sub-Appalachian lithospheric mantle by dewatering of the Farallon slab (van der Lee et al., 2008) could have played an important role in enabling this small-scale convection.

Figure 1.6: Possible model for the formation of the sub-Piedmont lithospheric resistor. Delamination of pyroxenite-rich mantle and eclogitized lower crust from the Alleghenian Orogen leads to CAMP magmatism as well as development of a void in the Appalachian lithosphere that is filled by upwelling asthenosphere. This asthenospheric mantle assimilates material from the delaminant and thereby becomes highly fertilized or metasomatized. As the eastern Laurentian margin cools after opening of the Atlantic, this region subsides. Lithospheric cooling and the high density of the fertile regrown root allow for development of a step in the LAB, which now supports edge convection that can drive topographic disequilibrium.

1.6 Conclusions Inversion of MT data from the southeastern United States provides new insights into the architecture of the Appalachian Orogen and also reveals the presence of an anomalous lithospheric block beneath the Piedmont and Coastal Plain. Based on these MT data, the lithosphere beneath the Piedmont must extend to >200 km depth. In contrast, the lithosphere beneath the modern Appalachians only extends to 150-200 km depth. Although such relatively thick Piedmont lithosphere has not been observed by seismic studies, the MT data

22 require low temperatures (~1100° C) to depths of ~200 km. This lithospheric step is therefore robust. The origin of the thick sub-Piedmont lithosphere and the cause of its apparent long-term geodynamic stability are unclear at present. However, the contrast between highly resistive lithosphere and moderately conductive lithosphere corresponds to the topographic escarpment of the modern Appalachian Mountains. This structure therefore likely holds important implications for Appalachian topographic rejuvenation. We propose that it has enabled edge- driven convection that has provided dynamic support for topography in the eastern United States. Further detailed study of this region is clearly warranted. Future collaboration with seismologists, geodynamicists, and geochemists is necessary in order to better characterize the composition and physical state of the Piedmont resistor and to explore the implications of this structure for the past and present geodynamics of the southeastern United States. The eastern boundary of this structure is poorly resolved at present due to the current limits of our dataset. A seafloor MT survey offshore of North Carolina, South Carolina, and Georgia would be tremendously beneficial in better understanding this region.

1.7 References Abbott, R.N., Greenwood, J.P., 2001. Retrograde metamorphism of eclogite in the southern Appalachian Mountains, U.S.A.—A case involving seamount subduction? Journal of Metamorphic Geology 19, 433-443. http://dx.doi.org/10.1046/j.0263-4929.2001.00321.x.

Ardia, P., Hirschmann, M.M., Withers, A.C., Tenner, T.J., 2012. H2O storage capacity of olivine at 5-8 GPa and consequences for dehydration partial melting of the upper mantle. Earth and Planetary Science Letters 345-348, 104-116. http://dx.doi.org/10.1016/j.epsl.2012.05.038. Biryol, C.B., Wagner, L.S., Fischer, K.M., Hawman, R.B., 2016. Relationship between observed upper mantle structures and recent tectonic activity across the Southeastern United States. Journal of Geophysical Research: Solid Earth 121, 3393-3414. http://dx.doi.org/10.1002/2015JB012698. Boerner, D.E., Kurtz, R.D., Craven, J.A., 1996. Electrical conductivity and Paleo-Proterozoic foredeeps. Journal of Geophysical Research 101, 13775-13791. http://dx.doi.org/10.1029/96JB00171.

Boote, S.K., Knapp, J.H., 2016. Offshore extent of Gondwanan Paleozoic strata in the southeastern United States: The Suwannee suture zone revisited. Gondwana Research 40, 199-210. http://dx.doi.org/10.1016/j.gr.2016.08.011. Cammarano, F., Goes, S., Vacher, P., Giardini, D., 2003. Inferring upper-mantle temperatures from seismic velocities. Physics of the Earth and Planetary Interiors 138, 197-222. http://dx.doi.org/10.1016/S0031-9201(03)00156-0. Chu, R., Leng, W., Helmberger, D.V., Gurnis, M., 2013. Hidden hotspot track beneath the eastern United States. Nature Geoscience 6, 963-966. http://dx.doi.org/10.1038/ngeo1949. Constable, S., 2006. SEO3: A new model of olivine electrical conductivity. Geophysical Journal International 166, 435-437. http://dx.doi.org/10.1111/j.1365-246X.2006.03041.x. Cook, F.A., Vasudevan, K., 2006. Reprocessing and enhanced interpretation of the initial COCORP Southern Appalachians traverse. Tectonophysics 420, 161-174. http://dx.doi.org/10.1016/j.tecto.2006.01.022. Cox, R.T., Van Arsdale, R.B., 2002. The Mississippi Embayment, North America: a first order continental structure generated by the Cretaceous superplume mantle event. Journal of Geodynamics 34, 163-176. http://dx.doi.org/10.1016/S0264-3707(02)00019-4. Dai, L., Karato, S., 2014. High and highly anisotropic electrical conductivity of the asthenosphere due to hydrogen diffusion in olivine. Earth and Planetary Science Letters 408, 79-86. http://dx.doi.org/10.1016/j.epsl.2014.10.003. Du Frane, W.L., Roberts, J.J., Toffelmier, D.A., Tyburczy, J.A., 2005. Anisotropy of electrical conductivity in dry olivine. Geophysical Research Letters 32, L24315. http://dx.doi.org/10.1029/2005GL023879. Elkins-Tanton, L.T., 2005. Continental magmatism caused by lithospheric delamination. In Foulger, G.R., Natland, J.H., Presnall, D.C., Anderson, D.L., eds. Plates, plumes, and paradigms. Geological Society of America Special Paper 388, 449-461. http://dx.doi.org/10.1130/0-8137-2388-4.449. Elliott, S., 1998. The Physics and Chemistry of Solids, Wiley, England. Erdman, M.E., Lee, C.-T. A., Levander, A., Jiang, H., 2016. Role of arc magmatism and lower crustal foundering in controlling elevation history of the Nevadaplano and Colorado Plateau: A case study of pyroxenitic lower crust from central Arizona, USA. Earth and Planetary Science Letters 439, 48-57. http://dx.doi.org/10.1016/j.epsl.2016.01.032.

Fisher, C.M., Loewy, S.L., Miller, C.F., Berquist, P., Van Schmus, W.R., Hatcher, R.D., Jr., Wooden, J.L., Fullagar, P.D., 2010. Whole-rock Pb and Sm-Nd isotopic constraints on the growth of southeastern Laurentia during Grenvillian orogenesis. Geological Society of America Bulletin 122, 1646-1659. http://dx.doi.org/10.1130/B30116.1. Frizon de Lamotte, D., Fourdan, B., Leleu, S., Leparmentier, F., de Clarens, P., 2015. Style of rifting and the stages of Pangea breakup. Tectonics 34, 1009-1029. http://dx.doi.org/10.1002/2014TC003760. Fullea, J., Muller, M.R., Jones, A.G., 2011. Electrical conductivity of continental lithospheric mantle from integrated geophysical and petrological modeling: Application to the Kaapvaal Craton and Rehoboth Terrane, southern Africa. Journal of Geophysical Research 116, B10202. http://dx.doi.org/10.1029/2011JB008544. Gallen, S.F., Wegmann, K.W., Bohnenstiehl, D.R., 2013. Miocene rejuvenation of topographic relief in the southern Appalachians. GSA Today 23, 4-10. http://dx.doi.org/10.1130/GSATG163A.1. Gardés, E., Gaillard, F., Tarits, P., 2014. Toward a unified hydrous olivine electrical conductivity law. Geochemistry, Geophysics, Geosystems 15, 4984-5000. http://dx.doi.org/10.1002/2014GC005496. González-Castillo, L., Galindo-Zaldívar, J., Junge, A., Martínez-Moreno, F.J., Löwer, A., Sanz de Galdeano, C., Pedrera, A., López-Garrido, A.C., Ruiz-Constán, A., Ruano, P., Martínez-Martos, M., 2015. Evidence of a large deep conductive body within the basement of Guadalquivir foreland Basin (Betic Cordillera, S-Spain) from tipper vector modelling: Tectonic implications. Tectonophysics 663, 354-363. http://dx.doi.org/10.1016/j.tecto.2015.08.013. Hatcher, R.D., Jr., 2010. The Appalachian orogen: A brief summary, in Tollo, R.P., Bartholomew, M.J., Hibbard J.P., and Karabinos, P.M., eds., From Rodinia to Pangea: The Lithotectonic Record of the Appalachian Region. Geological Society of America Memoir 206, 1-19. http://dx.doi.org/10.1130/2010.1206(01). Hawman, R.B., Khalifa, M.O., Baker, M.S., 2012. Isostatic compensation for a portion of the Southern Appalachians: Evidence from a reconnaissance study using wide-angle, three- component seismic soundings. Geological Society of America Bulletin 124, 291-317. http://dx.doi.org/10.1130/B30464.1. Hibbard, J.P., van Staal, C.R., Rankin, C.W., 2010. Comparative analysis of the geological evolution of the northern and southern Appalachian orogen: Late Ordovician-Permian. In Tollo, R.P., Bartholomew, M.J., Hibbard, J.P., Karabinos, P.M., eds., From Rodinia to

Pangea: The Lithotectonic Record of the Appalachian Region. Geological Society of America Memoir 206, 51-69. http://dx.doi.org/10.1130/2010.1206(03). Hopper, E., Fischer, K.M., Wagner, L.S., Hawman, R.B., 2017. Reconstructing the end of the Appalachian orogen. Geology 45, 15-18. http://dx.doi.org/10.1130/G38453.1. Jones, A.G., Ledo, J., Ferguson, I.J., 2005. Electromagnetic images of the Trans-Hudson orogen: the North American Central Plains anomaly revealed. Canadian Journal of Earth Sciences 32, 457-478. http://dx.doi.org/10.1139/e05-018. Jones, A.G., Fullea, J., Evans, R.L., Muller, M.R., 2012. Water in cratonic lithosphere: Calibrating laboratory-determined models of electrical conductivity of mantle minerals using geophysical and petrological observations. Geochemistry, Geophysics, Geosystems 13, 27. http://dx.doi.org/10.1029/2012GC004055. Kaislaniemi, L., van Hunen, J., 2014. Dynamics of lithospheric thinning and mantle melting by edge-driven convection: Application to Moroccan Atlas mountains. Geochemistry, Geophysics, Geosystems 15, 3175-3189. http://dx.doi.org/10.1002/2014GC005414. Kelbert, A., Meqbel, N., Egbert, G.D., Tandon, K., 2014. ModEM: A modular system for inversion of electromagnetic geophysical data. Computers & Geosciences 66, 40-53. http://dx.doi.org/10.1016/j.cageo.2014.01.010. Keppler, H., Bolfan-Casanova, N., 2006. Thermodynamics of Water Solubility and Partitioning. In Keppler, H., Smyth, J.R., eds. Water in Nominally Anhydrous Minerals. Reviews in Mineralogy and Geochemistry 62, 193-230. http://dx.doi.org/10.2138/rmg.2006.62.9. Lee, C.-T. A., 2003. Compositional variation of density and seismic velocities in natural peridotites at STP conditions: Implications for seismic imaging of compositional heterogeneities in the upper mantle. Journal of Geophysical Research 108, B92441. http://dx.doi.org/10.1029/2003JB002413. Liu, L., 2014. Rejuvenation of Appalachian topography caused by subsidence-induced differential erosion. Nature Geoscience 7, 518-523. http://dx.doi.org/10.1038/ngeo2187. Long, M.D., Jackson, K.G., McNamara, J.F., 2016. SKS splitting beneath Transportable Array stations in eastern North America and the signature of past lithospheric deformation. Geochemistry, Geophysics, Geosystems 17, 2-15. http://dx.doi.org/10.1002/2015GC006088. MacDougall, J.G., Fischer, K.M., Forsyth, D.W., Hawman, R.B., Wagner, L.S., 2015. Shallow mantle velocities beneath the southern Appalachians from Pn phases. Geophysical Research Letters 42, 339-345. http://dx.doi.org/10.1002/2014GL062714.

Mazza, S.E., Gazel, E., Johnson, E.A., Kunk, M.J., McAleer, R., Spotila, J.A., Bizimis, M., Coleman, D.S., 2014. Volcanoes of the passive margin: The youngest magmatic event in eastern North America. Geology 42, 483-486. http://dx.doi.org/10.1130/G35407.1. McKeon, R.E., Zeitler, P.K., Pazzaglia, F.J., Idleman, B.D., Enkelmann, E., 2014. Decay of an old orogen: Inferences about Appalachian landscape evolution from low-temperature thermochronology. Geological Society of America Bulletin 126, 31-46. http://dx.doi.org/10.1130/B30808.1. Miller, S.R., Sak, P.B., Kirby, E., Bierman, P.R., 2013. Neogene rejuvenation of central Appalachian topography: Evidence for differential rock uplift from stream profiles and erosion rates. Earth and Planetary Science Letters 369-370, 1-12. http://dx.doi.org/10.1016/j.epsl.2013.04.007. Moucha, R., Forte, A.M., Mitrovica, J.X., Rowley, D.B., Quéré, S., Simmons, N.A., Grand, S.P., 2008. Dynamic topography and long-term sea-level variations: There is no such thing as a stable continental platform. Earth and Planetary Science Letters 271, 101-108. http://dx.doi.org/10.1016/j.epsl.2008.03.056. Ogawa, Y., Jones, A.G., Unsworth, M.J., Booker, J.R., Lu, X., Craven, J., Roberts, B., Parmelee, J., Farquharson, C., 1996. Deep electrical conductivity structures of the Appalachian Orogen in the southeastern U.S. Geophysical Research Letters 23, 1597- 1600. http://dx.doi.org/10.1029/95GL03601. Phillips, J.D., Duval, J.S., Ambroziak, R.A., 1993. National geophysical data grids: gamma- ray, gravity, magnetic, and topographic data for the conterminous United States. United States Geological Survey Digital Data Series 9. http://mrdata.usgs.gov/geophysics/gravity.html (accessed Oct. 1, 2016). Poag, C.W., Sevon, W.D., 1989. A record of Appalachian denudation in postrift Mesozoic and Cenozoic sedimentary deposits of the U.S. Middle Atlantic continental margin. Geomorphology 2, 119-157. http://dx.doi.org/10.1016/0169-555X(89)90009-3. Pollitz, F.F., Mooney, W.D., 2016. Seismic velocity structure of the crust and shallow mantle of the Central and Eastern United States by seismic surface wave imaging. Geophysical Research Letters 43, 118-126. http://dx.doi.org/10.1002/2015GL066637. Prince, P.S., Spotila, J.A., 2013. Evidence of transient topographic disequilibrium in the landward passive margin river system: knickpoints and paleo-landscapes of the New River basin, southern Appalachians. Earth Surface Processes and Landforms 38, 1685- 1699. http://dx.doi.org/10.1002/esp.3406.

Schmandt, B., Lin, F.-C., 2014. P and S wave tomography of the mantle beneath the United States. Geophysical Research Letters 41, 6342-6349. http://dx.doi.org/10.1002/2014GL061231. Schmandt, B., Lin, F.-C., Karlstrom, K.E., 2015. Distinct crustal isostasy trends east and west of the Rocky Mountain Front. Geophysical Research Letters 42, 10290-10298. http://dx.doi.org/10.1002/2015GL066593. Schutt, D.L., Lesher, C.E., 2010. Compositional trends among Kaapvaal Craton garnet peridotite xenoliths and their effects on seismic velocity and density. Earth and Planetary Science Letters 300, 367-373. http://dx.doi.org/10.1016/j.epsl.2010.10.018. Selway, K., 2014. On the Causes of Electrical Conductivity Anomalies in Tectonically Stable Lithosphere. Surveys in Geophysics 35, 219-257. http://dx.doi.org/10.1007/s10712-013- 9235-1. Shen, W., Ritzwoller, M.H., 2016. Crustal and uppermost mantle structure beneath the United States. Journal of Geophysical Research: Solid Earth 121, 37. http://dx.doi.org/10.1002/2016JB012887. Simpson, R.W., Jachens, R.C., Blakely, R.J., Saltus, R.W., 1986. A New Isostatic Residual Gravity Map of the Conterminous United States With a Discussion on the Significance of Isostatic Residual Anomalies. Journal of Geophysical Research 91, 8348-8372. http://dx.doi.org/10.1029/JB091iB08p08348. Smith, D.L., Gregory, R.G., Emhof, J.W., 1981. Geothermal Measurements in the Southern Appalachian Mountains and Southeastern Coastal Plains. American Journal of Science 281, 282-298. http://dx.doi.org/10.2475/ajs.281.3.282. Sobolev, A.V., Hofmann, A.W., Kuzmin, D.V., Yaxley, G.M., Arndt, N.T., Chung, S.-L., Danyushevsky, L.V., Elliott, T., Frey, F.A., Garcia, M.O., Gurenko, A.A., Kamenetsky, V.S., Kerr, A.C., Krivolutskaya, N.A., Matvienkov, V.V., Nikogosian, I.K., Rocholl, A., Sigurdsson, I.A., Sushchevskaya, N.M., Teklay, M., 2007. The Amount of Recycled Crust in Sources of Mantle-Derived Melts. Science 316, 412-417. http://dx.doi.org/10.1126/science. 1138113. Spasojević, S., Liu, L., Gurnis, M., Müller, R.D., 2008. The case for dynamic subsidence of the U.S. east coast since the Eocene. Geophysical Research Letters 35, L08305. http://dx.doi.org/10.1029/2008GL033511. Steltenpohl, M.G., Zietz, I., Horton, J.W., Jr., Daniels, D.L., 2010. New York-Alabama lineament: A buried right-slip fault bordering the Appalachians and mid-continent North America. Geology 38, 571-574. http://dx.doi.org/10.1130/G30978.1.

28 van der Lee, S., Regenauer-Lieb, K., Yuen, D.A., 2008. The role of water in connecting past and future episodes of subduction. Earth and Planetary Science Letters 273, 15-27. http://dx.doi.org/10.1016/j.epsl.2008.04.041. Wagner, L.S., Stewart, K., Metcalf, K., 2012. Crustal-scale shortening structures beneath the Blue Ridge Mountains, North Carolina, USA. Lithosphere 4, 242-256. http://dx.doi.org/10.1130/L184.1. Wang, H., Currie, C.A., 2015. Magmatic expressions of continental lithosphere removal. Journal of Geophysical Research: Solid Earth 120, 7239-7260. http://dx.doi.org/10.1002/2015JB012112. Wannamaker, P.E., 2005. Anisotropy versus Heterogeneity in Continental Solid Earth Electromagnetic Studies: Fundamental Response Characteristics and Implications for Physicochemical State. Surveys in Geophysics 26, 733-765. http://dx.doi.org/10.1007/s10712-005-1832-1. Whalen, L., Gazel, E., Vidito, C., Puffer, J., Bizimis, M., Henika, W., Caddick, M.J., 2015. Supercontinental inheritance and its influence on supercontinental breakup: The Central Atlantic Magmatic Province and the breakup of Pangea. Geochemistry, Geophysics, Geosystems 16, 3532-3554. http://dx.doi.org/10.1002/2015GC005885. Willett, S.D., McCoy, S.W., Perron, J.T., Goren, L., Chen, C.-Y., 2014. Dynamic Reorganization of River Basins. Science 343, 9. http://dx.doi.org/10.1126/science.1248765. Yang, B., Egbert, G.D., Kelbert, A., Meqbel, N.M., 2015. Three-dimensional electrical resistivity of the north-central USA from EarthScope long period magnetotelluric data. Earth and Planetary Science Letters 422, 87-93. http://dx.doi.org/10.1016/j.epsl.2015.04.006. Yoshino, T., Matsuzaki, T., Shatskiy, A., Katsura, T., 2009. The effect of water on the electrical conductivity of olivine aggregates and its implications for the electrical structure of the upper mantle. Earth and Planetary Science Letters 288, 291-300. http://dx.doi.org/10.1016/j.epsl.2009.09.032.

Chapter 2

Source Biases in Mid-Latitude Magnetotelluric Transfer Functions due to Pc3-4 Geomagnetic Pulsations

Benjamin S. Murphy & Gary D. Egbert

Earth, Planets and Space Volume 70, Article 12, 22 January 2018

2.1 Abstract The magnetotelluric (MT) method for imaging the electrical conductivity structure of the Earth is based on the assumption that source magnetic fields can be considered quasi- uniform, such that the spatial scale of the inducing source is much larger than the intrinsic length scale of the electromagnetic induction process (the skin depth). Here, we show using EarthScope MT data that short spatial scale source magnetic fields from geomagnetic pulsations (Pc’s) can violate this fundamental assumption. Over resistive regions of the Earth, the skin depth can be comparable to the short meridional range of Pc3-4 disturbances that are generated by geomagnetic field-line resonances (FLRs). In such cases, Pc’s can introduce narrow-band bias in MT transfer function estimates at FLR eigenperiods (~10-100 s). Although it appears unlikely that these biases will be a significant problem for data inversions, further study is necessary to understand the conditions under which they may distort inverse solutions.

2.2 Introduction The fundamental assumption underlying the magnetotelluric (MT) method is that external source fields have wavelengths that are large compared to the fundamental length scale characterizing electromagnetic (EM) induction in the Earth—that is, the skin depth (e.g., Weidelt and Chave, 2012). With the assumption of a quasi-uniform source, the impedance (transfer function, TF, relating electric to magnetic fields at a point on the Earth’s surface) is well defined and independent of the actual (large-scale) spatial structure of the source fields. Because EM induction in a conductor is a diffusive process, the impedance and other uniform-source TFs used in MT (e.g., vertical magnetic field TFs) should be very smooth functions of frequency (Weidelt, 1972). Indeed, excessively rapid variations in TF curves estimated from field data are often taken as evidence of source complications, most often due to relatively short-wavelength anthropogenic EM signals such as electric trains (e.g., Larsen et al., 1996; Egbert, 1997; Egbert et al., 2000). Possible biases in MT TFs due to complications in naturally occurring external sources are also occasionally considered, but generally only at long periods (>1000 s) and near the auroral or equatorial electrojets (e.g., Viljanen, 2012, and references therein). As part of the EarthScope USArray project, long-period (~101-104 s) MT data have been acquired since 2006 in a series of short (20-30 day) deployments on a quasi-uniform 70 km grid. As of this writing, approximately 1000 sites across about half of the continental United States (US) have been occupied (Fig. 2.1). A significant fraction of the impedances from this

31 dataset display unphysical “humps” that significantly deviate from the smooth trend of the TF over a narrow range of periods, typically much less than half a decade in width, between 10 and 100 s (Fig. 2.2). A conservative marking of sites where impedances clearly exhibit these unphysical features is shown in Figure 2.1. A larger fraction of sites, perhaps as great as 15%, show at least some evidence for similar biases of smaller amplitude. The affected sites are most common in, or on the edges of, areas where the Earth is relatively resistive (Fig. 2.1). As increasing resistivity increases the skin depth and makes the quasi-uniform assumption easier to violate, this observation suggests that these humps reflect source bias. In this paper, we provide further evidence for this conclusion and demonstrate that short- wavelength natural sources associated with geomagnetic pulsations (Pc’s) can explain the observed biases.

Figure 2.1: Map showing distribution of EarthScope MT sites identified as clearly showing source bias (gray filled circles), plotted over vertically integrated Earth conductance (15-150 km) calculated from the inverse solutions of Meqbel et al. (2014) (northwestern US), Yang et al. (2015) (north-central US), and Murphy and Egbert (2017) (southeastern US). For reference, the conductance of constant 100 Ωm Earth between 15 km and 150 km would be 1350 S. Markers with purple centers are sites used in the array analysis shown in Figure 2.3. Markers with black dots in the center are sites used in the cross-phase analyses shown in Figure 2.2. Hollow circles show all available EarthScope MT sites as of this writing. Key geologic features in the northwestern US that are discussed in the text are shown on the map as well. MHB: Medicine Hat Block; Wy: Wyoming Craton; SRP: Snake River Plain; CP: Colorado Plateau.

The possibility that Pc’s, which in the ~10-100 s period band usually result from geomagnetic field line resonance (and hence exhibit rapid variations meridionally), might violate the quasi-uniform source assumption has been discussed previously from a theoretical perspective by Pilipenko and Fedorov (1993, 1994). Our results demonstrate with a large collection of MT impedance estimates that such source effects can indeed cause problems for the MT method, at least in resistive areas.

2.2.1 The Quasi-Uniform Source Assumption in the Magnetotelluric Method Mathematically, the external magnetic field satisfies the quasi-uniform source (“plane- wave”) assumption when the skin depth δ for induced currents is much less than the spatial scale of the source, i.e.,

√휌푇 휆(푇) ≫ 훿 = , (1) 2 where T is the period (in s) of the exciting source harmonic component; λ is the spatial wavelength (in km) of the source; and ρ is the electrical resistivity (in Ωm) of the region of the Earth in question. For a one-dimensional Earth, a rigorous mathematical treatment is possible. Dmitriev and Berdichevsky (1979) showed that if the source magnetic field varies linearly over a distance of at least three times the skin depth, the local relationship between electric and magnetic fields will be nearly identical to that obtained with an exactly uniform source. (See also Wait, 1954, 1982.) When there are lateral variations in resistivity, a rigorous analysis is not possible, and there is some evidence that source effects may be stronger (e.g., Egbert and Booker, 1989). Nonetheless, the fundamental physical intuition remains the same: the inducing external magnetic field should not vary significantly over length scales that are several times the skin depth in the Earth.

2.2.2 Geomagnetic Pulsations and Field-Line Resonances In the ~10-100 s period band, geomagnetic pulsations are a dominant component of the geomagnetic power spectrum measured on the Earth’s surface (e.g., McPherron, 2005). These signals originate beyond the magnetosphere (in the foreshock, bowshock, and/or magnetopause) as ULF hydromagnetic waves that propagate into the magnetosphere through a variety of mechanisms (Hughes, 1994; Clausen et al., 2009; Menk, 2011). Within the magnetosphere, these hydromagnetic disturbances can excite standing Alfvén waves along geomagnetic field lines (Hughes, 1994). A single field line can be pictured as a forced,

33 damped harmonic oscillator (Waters et al., 1991), with the amplitude of the response greatly enhanced at the natural resonant frequency of the field line. (See Fig. A2.1 for a graphical summary.) Although this model of field-line resonance (FLR) is very simple, it explains many observed characteristics of Pc’s (Hughes, 1994). FLRs drive electrical currents in the ionosphere at the ends of the field lines (McPherron, 2005). Ionospheric Pedersen currents (parallel to driving electric field) cancel out the magnetic field of the FLRs, thereby screening their direct signature at the Earth’s surface, but east-west ionospheric Hall currents (perpendicular to driving electric and magnetic fields) produce electromagnetic radiation that is observed on the ground as dominantly north-south magnetic oscillations (McPherron, 2005). The FLR signature at the Earth’s surface is therefore the result of ionospheric screening processes (Hughes and Southwood, 1976; McPherron, 2005). Because geomagnetic field lines have distinct resonant frequencies, the resulting magnetic disturbances observed on the ground are nearly monochromatic and highly sinusoidal; they are therefore identified as a type of continuous pulsation (Pc). Pulsations associated with FLRs are generally in the Pc3-4 band (~10-150 s) (McPherron, 2005). The resonant period of geomagnetic field lines depends on the length of the field line, the magnetic field strength along the field line, and the plasma density along the field line (Hughes, 1994; Waters et al., 2006; Menk, 2011). Field-line length increases with geomagnetic latitude, so the fundamental FLR period generally increases geomagnetically poleward. Because magnetic field strength and plasma density are highly dynamic, the FLR period fluctuates in time; for example, daily variations in plasma density cause daily variations in FLR period (Poulter et al., 1988). On the Earth’s surface, the width of a given FLR is of order 100 km (Hughes, 1994; Menk, 2011; Chi et al., 2013). As can be seen from the simple harmonic oscillator model of an FLR as well as from more complex theoretical and empirical treatments, amplitudes will be enhanced over a limited range of latitudes centered on the main resonant field line (several hundred kilometers), and there will be a 180° phase shift moving geomagnetically north- south through this zone (Waters et al., 1991; Hughes, 1994; Chi and Russell, 1998; Kawano et al., 2002; see also Fig. A2.1). Although large-scale magnetospheric perturbations (fast- mode compressional waves) are also observed on the ground in this same period range (Kawano et al., 2002; Chi et al., 2013), when FLRs are excited, their effects typically dominate the total incident field at the Earth’s surface (Ádám et al., 2005). Therefore, near the (latitude-dependent) resonant frequency, the total source magnetic field can vary rapidly over only a few hundred kilometers in the geomagnetic north-south direction. Where the

Earth is in bulk moderately conductive (~100 Ωm), from Eqn. 1 the skin depth at an FLR period of 30 s will be ~30 km, still relatively small compared to Pc source wavelengths. However, if the Earth is in bulk relatively resistive (~1000 Ωm or more), the skin depth at 30 s will be ~90 km or more, comparable to source wavelengths expected for Pc’s. More resistive Earth or a longer-period FLR will make the skin depth an even larger fraction of the spatial scale of the Pc’s. Clearly, over resistive Earth the quasi-uniform source assumption of the MT method can be violated in the Pc3-4 band.

2.3 Data and Methods In order to investigate the link between field-line resonances and biases in MT TFs, we utilized two different analyses to explore the spectral and spatial characteristics of the source horizontal magnetic fields. We used EarthScope MT time series and robust remote-reference TF estimates (Kelbert et al., 2011), which are archived for community use with IRIS (http://ds.iris.edu/spud/emtf). To detect the presence and period localization of FLRs in the EarthScope MT time series, we use the cross-phase technique (e.g., Waters et al., 1991; Chi and Russell, 1998; Kawano et al., 2002; Menk et al., 2004; Waters et al., 2006; Chi et al., 2013; see also Fig. A2.1 and supplemental text in Appendix 2), a well-established analytical method in ULF space physics research. Because the phase of the FLR changes significantly when moving meridionally through a resonance, the phase difference as a function of period (cross-phase) between two sites spaced at the length scale of an FLR should show a sharp peak at the resonance period (Waters et al., 1991; Chi and Russell, 1998; Kawano et al., 2002). We selected pairs of sites in which the two sites lie roughly on the same geomagnetic meridian and are separated in latitude by ~70-200 km. With the N-S magnetic field time series from these pairs of sites (the X geomagnetic component), we used the wavelet-based cross-phase calculation approach documented by Waters et al. (2006). However, because the MT method uses a time-averaged frequency-domain analysis, we then averaged the cross-phase estimates in time. The resulting cross-phase curve thus provides an estimate of the time-averaged FLR period appropriate for comparison to the MT TF estimates. Note that, due to induced currents in the Earth, there may be spatially localized anomalous horizontal magnetic fields that also cause variations in the cross-phase spectrum. Because these induced internal fields result from a diffusive induction process, they will be smooth and broadband, in contrast to the sharp (localized in period) phase signature of FLRs.

To further characterize the spatial patterns of EM fields at the analyzed MT sites, we applied the array processing approach of Egbert and Booker (1989; see also Egbert, 1997), to small (~6-10 site) arrays of simultaneously occupied EarthScope sites. We specifically used the processing scheme extended by Smirnov and Egbert (2012); this technique is essentially a robust frequency-domain principal components analysis (PCA) to allow for missing data and only partial overlap of site occupation intervals. The PCA scheme decomposes the frequency domain data from the array (including all five channels from each MT site, so 30-50 channels total) into a small number (typically 2-5) of coherent spatial modes. The analysis provides estimates of incoherent noise levels in each channel and signal levels in each spatial mode. If the uniform source assumption held exactly, only two modes (corresponding to uniform source polarization parameters) would have power above the incoherent noise level; additional modes are generally indicative of source complications (either natural or anthropogenic). The spatial modes combine information about both source spatial structure and the response of the Earth.

2.4 Results Analyses of station pairs across the US reveal sharp peaks in the cross-phase spectra in the Pc3-4 range, thereby confirming the signature of FLRs in the EarthScope magnetic field time series. Two examples of these analyses are shown in Figure 2.2, in both cases for sites located over resistive Earth (Fig. 1) where skin depths will be large. Note that the peak in the cross-phase spectrum shifts to longer periods at higher geomagnetic latitude. At both pairs of sites, the peak period is consistent with expectations for FLRs (e.g., Poulter et al., 1988). These peaks imply a phase shift of ~15° (on average) over the distance between the stations in each pair (~150 km) in the same period range as the biases in the corresponding MT TFs, which are also shown in Figure 2.2. This correlation is strong evidence that the short spatial scale of the FLRs is contaminating the TF estimates where the skin depth is large. A similar analysis of sites in the relatively conductive Snake River Plain still shows the presence of the FLR cross-phase peak, but the TFs do not show any clear biases (see Fig. A2.2). Our analysis of a large number of station pairs shows that FLR signatures are always present in the EarthScope magnetic field data (and consequently always induce telluric currents), but apparently only where skin depths become large do they bias the TF estimates.

Figure 2.2: Transfer functions and cross-phase analyses from sites shown in Figure 2.1. The upper four plots in each column display the robustly estimated remote-reference TFs for the indicated site. TF estimates were calculated with EMTF (Egbert and Booker, 1986). In each of these plots, blue is used to denote TFs that involve the N-S magnetic field (X geomagnetic component: YX impedance element and Tx vertical field TF component) and red is used to denote TFs that involve the E-W magnetic field (Y geomagnetic component: XY impedance element and Ty vertical field TF component). The bottom plot in each column shows the calculated time-averaged cross-phase spectra for the specified pair of sites for the N-S (X) component of the magnetic field (estimated using the wavelet technique outlined in Waters et al., 2006). (See Fig. A2.3 for TFs from NCT57 and IAL37; see Fig. A2.4 for plots of the full impedance tensor from SCV57.)

Figure 2.3: Results from frequency domain robust principal components analysis of electric and magnetic fields from the array shown in Figure 2.1, which includes the two sites from the southeastern US used in Figure 2.2. (a) Power spectrum of 6 dominant modes, given in signal-to-noise units expressed in decibels (dB). Signal power above 0 dB indicates that the mode is above the incoherent noise level and therefore significant. If the uniform-source assumption that underlies the MT method held exactly there would only be two significant modes (shown in red). In the FLR period band for this array (~20-40 s), the curve for the third mode becomes significant. Additional significant modes at longer periods are also associated with source effects (primarily in the vertical magnetic fields) due to increasing skin depths with respect to large-scale source structure (e.g., Jones and Spratt, 2002). (b-d) Linear combinations of the first three spatial modes at a period of 26 s, displayed as complex vectors plotted on a map of array sites. The first two panels show linear combinations, defined following the approach in Egbert (1997), that best approximate uniform magnetic sources, linearly polarized geomagnetically N-S and E-W. Panel (d) represents the remaining signal in the first three modes. This mode exhibits a clear north-south spatial gradient and very large vertical magnetic components, consistent with expectations for a spatially localized FLR signature. In these plots, horizontal magnetic (electric) fields are shown with blue (red) arrows, with real parts solid and imaginary parts dashed. The vertical magnetic field is shown with a blue circle, scaled as for the horizontal components to indicate amplitude, with the position of the drawn radius denoting phase. To scale the lengths of electric and magnetic field vectors in a physically meaningful way, a reference Earth resistivity (given in each panel) is assumed.

The robust PCA approach provides further information about source spatial structure associated with FLRs and the observed TF biases. Results for a small array located over the highly resistive Piedmont Resistor of Murphy and Egbert (2017), where all six sites show TF biases at ~30 s (analyses from two of these sites are shown in Fig. 2.2), are shown in Figure 2.3. In Figure 2.3a, power in signal-to-noise ratio units is plotted as a function of frequency for the first 6 modes. Over the same band of periods where the cross-phase analysis identifies the FLR (and where the TF biases appear), a third mode rises significantly above the estimated background incoherent noise level (defined to have an SNR of 0 on this plot). The

38 three estimated PCA modes in this band likely represent a mixture of nearly uniform magnetic field sources and N-S gradients associated with FLR. Following the approach of Egbert (2002), we seek linear combinations of the three estimated modes that most closely approximate uniform magnetic sources, polarized linearly in the N-S and E-W (geomagnetic) directions (Figs. 2.3b, c). The dominant signal remaining after removing these quasi-uniform components (Fig. 2.3d) exhibits strong meridional gradients in both the horizontal magnetic and electric field components. The vertical component of magnetic field (shown by the circles) is also very large (comparable to horizontal components). Note that the fine-scale heterogeneity in the spatial modes reflects complexities in the Earth rather than source complications or noise. This is particularly true of the electric field vectors, which can be strongly distorted by even small-scale superficial heterogeneity. The general pattern illustrated in Fig. 2.3 holds in subarrays across the continental US. Again, it is clear that short-wavelength gradients in horizontal magnetic fields associated with Pc’s are present in all the EarthScope MT data, but only where skin depths become large do their effects become noticeable in the MT TF estimates.

2.5 Discussion As the sharp peaks in the cross-phase spectra of Figure 2.2 indicate, there are significant phase shifts in source magnetic fields between sites that are meridionally separated on the order of 100 km. As these phase shifts occur in the Pc3-4 band and generally move to longer periods at higher geomagnetic latitudes, they are associated with localized, narrow-band FLRs. In resistive areas where the skin depth can become comparable to the spatial scale of FLRs (~100 km), the quasi-uniform-source assumption of the MT method will be violated. Because FLRs essentially cause a meridionally localized amplification of magnetic field variability in the Pc3-4 band, the signal associated with them will be spatially coherent (both zonally and meridionally). Therefore, these biases cannot be removed by remote-reference processing. (Note the TF estimates shown in Figs. 2.2 and A2.2 are robust remote-reference estimates; Fig. A2.5 shows that using different remote sites does not significantly change the biases.) Exclusion of time series segments with significant Pc activity can mitigate the resulting biases in TF estimates somewhat (see Fig. A2.6). However, we have not been able to completely remove these biases with simple data selection procedures. Furthermore, the Pc’s associated with FLRs appear to be a major source of signal for the MT method in the Pc3-4 band, at least at mid-latitudes. During times without clear Pc activity there is little MT

39 signal, which can result in noisy and unreliable TF estimates (Ádám et al., 2005). It is thus unclear if the source biases we have identified in the Pc3-4 band can be completely avoided.

2.5.1 Pc’s and the 3D Earth The spatial distribution of EarthScope MT sites that display source biases in impedance estimates (Fig. 2.1) suggests that to first order this problem will be most severe directly over resistive Earth. This pattern is clearest in the central US, over resistive Archean- to Paleoproterozoic-aged cratonic lithosphere that was imaged by Yang et al. (2015), and in the southeastern US, over the highly resistive Piedmont Resistor of Murphy and Egbert (2017). However, there are many sites over resistive areas where biases in the Pc3-4 band are not clear. Furthermore, in the western US, stations that appear most affected are often on the edges of, rather than directly above, the resistive Wyoming Craton, Medicine Hat Block, and Colorado Plateau (imaged by Meqbel et al., 2014). The exact details of how the short spatial scale of Pc’s interact with three-dimensional conductivity structures is likely more complex than the simple skin depth argument suggests. Large conductivity contrasts as well as high resistivity may be important determinants of the impact of short-wavelength sources on MT TF bias. Further numerical modeling work is necessary to explore this possibility.

2.5.2 Implications for MT Data Inversions As Figure 2.2 indicates, the effect of Pc-related source biases on MT impedances is clearly noticeable with good quality data. However, the effect is still relatively small compared to the uncertainties and inaccuracies inherent in MT inversions. We therefore consider it unlikely that these small biases in impedances will lead to substantial biases in inversion results, especially if appropriate error floors are set. However, there may be circumstances in which the impacts on inverse solutions will be more severe, so this issue deserves more careful consideration. As also indicated by Figure 2.2, in comparison to the effect on impedance data, the effect on the vertical magnetic field TFs (tippers) generally appears to be more significant. This is perhaps to be expected, as tippers are more dependent on source characteristics than impedances (e.g., Jones and Spratt, 2002). Biases in the tippers may sometimes be large enough that these data should be excluded from an inversion. In general, care should be taken in working with MT data in the FLR band, especially in places where the Earth is relatively resistive. Our observations of Pc-related biases in MT TFs also demonstrate the necessity of using error floors in MT data inversions. In the case documented here, the biases associated with

40 violation of the assumptions of the MT method are clearly noticeable; however, it is possible that measured data could include other far more subtle biases that could go unnoticed and unappreciated. Therefore, error floors are necessary to reflect the reality of working with MT data: intrinsic complexities and necessary approximations in the MT method, from data acquisition to data inversion, make it impossible to perfectly model real data.

2.5.3 Implications for Earth Structure Finally, we note that the widespread appearance of biases in the Pc3-4 band in MT TFs can be a useful first-order indicator of resistive Earth. For example, the pervasive presence of biases in the TF estimates from the Piedmont and Coastal Plain regions of the southeastern US is an additional strong line of evidence for the highly resistive, highly anomalous Piedmont Resistor documented by Murphy and Egbert (2017).

2.6 Conclusions Despite abundant empirical and theoretical study of FLRs in the space physics literature, that they can bias MT TFs at relatively short periods (~10-100 s) and at relatively low latitudes (<60°) has not been widely appreciated by the MT community. The potential for biases in MT TFs posed by Pc’s has been acknowledged theoretically before (e.g., Pilipenko and Federov 1993, 1994), and it was even pointed out at the very inception of the MT method. In a 1953 letter to Louis Cagniard, one of the pioneers of MT, James R. Wait raised the problem of geomagnetic pulsations (at the time called ‘micropulsations’):

The current system in the upper atmosphere giving rise to short period (less than 60 sec) variations of the magnetic [field] such as “micro pulsations” are confined to a limited region as observed by several investigators. In this case the infinite [current] sheet representation is not permissible.

Cagniard replied to Wait’s concern by flatly dismissing the potential for a violation of the plane-wave approximation:

I am entirely in accord with you except regarding the question of the micropulsations being limited to a very restricted region. Only yesterday, M. Migaux, director of the Compagnie Générale de Geophysique, the man who “sees” the most micropulsations,

told me that they do not exist except where there are industrial disturbances. Along with St. Thomas, we have to see to believe.

(From letters published in Vozoff, 1986, pp. 41, 43.) Now, using real data, we have shown that at mid latitudes (30-50°) geomagnetic pulsations in the Pc3-4 band (~10-100 s) associated with field-line resonances can and do violate a fundamental assumption of the MT method. They therefore can bias MT TF estimates where skin depths are large. Clearly, the infinite current sheet representation in the MT technique may at times be impermissible. Our observations herein stress the importance of understanding the magnetospheric/ionospheric sources that the MT technique exploits, of appreciating when the source fields may violate the assumptions of the MT method, and of carefully considering possible biases when interpreting and inverting estimated MT responses.

2.7 References Ádám A, Verõ J, Szendrõi J (2005) Solar eclipse effect on geomagnetic induction parameters. Ann Geophys 23:3487-3494. doi:10.5194/angeo-23-3487-2005 Chave AD, Jones AG (2012) Introduction to the magnetotelluric method. In: Chave AD, Jones AG (eds) The Magnetotelluric Method: Theory and Practice. Cambridge University Press, pp. 1-18. doi:10.1017/CBO9781139020138.002 Chi PJ, Engebretson MJ, Moldwin MB, Russell CT, Mann IR, Hairston MR, Reno M, Goldstein J, Winkler LI, Cruz-Abeyro JL, Lee D-H, Yumoto K, Dalrymple R, Chen B, Gibson JP (2013) Sounding of the plasmasphere by Mid-continent MAgnetoseismic Chain (McMAC) magnetometers. J Geophys Res-Space 118:3077-3086. doi:10.1002/jgra.50274 Chi PJ, Russell CT (1998) An interpretation of the cross-phase spectrum of geomagnetic pulsations by the field line resonance theory. Geophys Res Lett 25(24):4445-4448. doi:10.1029/1998GL900211 Clausen LBN, Yeoman TK, Fear RC, Behlke R, Lucek EA, Engebretson MJ (2009) First simultaneous measurements of waves generated at the bow shock in the solar wind, the magnetosphere and on the ground. Ann Geophys 27:357-371. doi:10.5194/angeo-27-357- 2009 Dmitriev VI, Berdichevsky MN (1979) The Fundamental Model of Magnetotelluric Sounding. Proc IEEE 67(7):1034-1044. doi:10.1109/PROC.1979.11386

Egbert GD (1997) Robust multiple-station magnetotelluric data processing. Geophys J Int 130:475-496. doi:10.1111/j.1365-246X.1997.tb05663.x Egbert GD (2002) Processing and interpretation of electromagnetic induction array data. Surv Geophys 23:207-249. doi:10.1023/A:1015012821040 Egbert GD, Booker JR (1986) Robust estimation of geomagnetic transfer functions. Geophys J Int 87:173-194. doi:10.1111/j.1365-246X.1986.tb04552.x Egbert GD, Booker JR (1989) Multivariate Analysis of Geomagnetic Array Data: 1. The Response Space. J Geophys Res 94(B10):14227-14247. doi:10.1029/JB094iB10p14227 Egbert GD, Eisel M, Boyd OS, Morrison HF (2000) DC trains and Pc3s: Source effects in mid-latitude geomagnetic transfer functions. Geophys Res Lett 27(1): 25-28. doi:10.1029/1999GL008369 Hughes, JW (1994) Magnetospheric ULF Waves: A Tutorial With a Historical Perspective. In: Engebretson MJ, Takahashi K, Scholer M (eds) Solar Wind Sources of Magnetospheric Ultra-Low-Frequency Waves. American Geophysical Union, Washington, pp. 1-11. doi:10.1029/GM081p0001 Hughes WJ, Southwood DJ (1976) The Screening of Micropulsation Signals by the Atmosphere and Ionosphere. J Geophys Res 81(19):3234-3240. doi:10.1029/JA081i019p03234 Jones AG, Spratt J (2002) A simple method for deriving the uniform field MT responses in auroral zones. Earth Planets Space 54:443-450. doi.org/10.1186/BF03353035 Kawano H, Yumoto K, Pilipenko VA, Tanaka Y-M, Takasaki S, Iizima M, Seto M (2002) Using two ground stations to identify magnetospheric field line eigenfrequency as a continuous function of ground latitude. J Geophys Res 107(A8). doi:10.1029/2001JA000274 Kelbert A, Egbert GD, Schultz A (2011) IRIS DMC Data Services Products: EMTF, The Magnetotelluric Transfer Functions. doi:10.17611/DP/EMTF Larsen JC, Mackie RL, Manzella A, Fiordelisi A, Rieven S (1996) Robust smooth magnetotelluric transfer functions. Geophys J Int 124:801-819. doi:10.1111/j.1365- 246X.1996.tb05639.x McPherron RL (2005) Magnetic Pulsations: Their Sources and Relation to Solar Wind and Geomagnetic Activity. Surv Geophys 26:545-592. doi:10.1007/s10712-005-1758-7 Menk FW, Mann IR, Smith AJ, Waters CL, Clilverd MA, Milling DK (2004) Monitoring the plasmapause using geomagnetic field line resonances. J Geophys Res 109:A04216. doi:10.1029/2003JA010097

Menk FW (2011) Magnetospheric ULF Waves: A Review. In: Liu W, Fujimoto M (eds) The Dynamic Magnetosphere. Springer, pp. 223-256. doi:10.1007/978-94-007-0501-2_13 Meqbel NM, Egbert GD, Wannamaker PE, Kelbert A, Schultz A (2014) Deep electrical resistivity structure of the northwestern U.S. derived from 3-D inversion of USArray magnetotelluric data. Earth Planet Sci Lett 402:290-304. doi:10.1016/j.epsl.2013.12.026 Murphy BS, Egbert GD (2017) Electrical conductivity structure of southeastern North America: Implications for lithospheric architecture and Appalachian topographic rejuvenation. Earth Planet Sci Lett 462:66-75. doi:10.1016/j.epsl.2017.01.009 Pilipenk VA, Federov EN (1993) Magnetotelluric sounding of the crust and hydromagnetic monitoring of the magnetosphere with the use of ULF waves. Ann Geofis 36(5-6):19-32. doi:10.4401/ag-4243 Pilipenko VA, Federov EN (1994) Magnetotelluric Sounding of the Crust and Hydromagnetic Monitoring of the Magnetosphere with the Use of ULF Waves. In: Engebretson MJ, Takahashi K, Scholer M (eds) Solar Wind Sources of Magnetospheric Ultra-Low-Frequency Waves. American Geophysical Union, Washington, pp. 283-292. doi:10.1029/GM081p0283 Poulter EM, Allan W, Bailey GJ (1988) ULF Pulsation Eigenperiods within the Plasmasphere. Planet Space Sci 36(2):185-196. doi:10.1016/0032-0633(88)90054-2 Smirnov MYu, Egbert GD (2012) Robust principal component analysis of electromagnetic arrays with missing data. Geophys J Int 190:1423-1438. doi:10.1111/j.1365- 246X.2012.05569.x Wait JR (1954) On the relation between telluric currents and the Earth’s magnetic field. Geophysics 19(2):281-289. doi:10.1190/1.1437994 Wait JR (1982) Geo-Electromagnetism. Academic Press, 268 p. Waters CL, Menk FW, Fraser BJ (1991) The Resonance Structure of Low Latitude Pc3 Geomagnetic Pulsations. Geophys Res Lett 18(12):2293-2296. doi:10.1029/91GL02550 Waters CL, Menk FW, Thomsen MF, Foster C, Fenrich FR (2006) Remote Sensing the Magnetosphere Using Ground-Based Observations of ULF Waves. In: Takahashi K, Chi PJ, Denton RE, Lysak RL (eds) Magnetospheric ULF Waves: Synthesis and New Directions. American Geophysical Union, Washington, pp. 319-340. doi:10.1029/169GM21 Weidelt P (1972) The Inverse Problem of Geomagnetic Induction. Zeitschrift für Geophysik 38:257-289.

Weidelt P, Chave AD (2012) The magnetotelluric response function. In: Chave AD, Jones AG (eds) The Magnetotelluric Method: Theory and Practice. Cambridge University Press, pp. 120-164. doi:10.1017/CBO9781139020138.006 Viljanen A (2012) Description of the magnetospheric/ionospheric sources. In: Chave AD, Jones AG (eds) The Magnetotelluric Method: Theory and Practice. Cambridge University Press, pp. 96-121. doi:10.1017/CBO9781139020138.005 Vozoff K (ed) (1986) Magnetotelluric Methods. Society of Exploration Geophysicists, 763 p. Yang B, Egbert GD, Kelbert A, Meqbel NM (2015) Three-dimensional electrical resistivity of the north-central USA from EarthScope long period magnetotelluric data. Earth Planet Sci Lett 422:87-93. doi:10.1016/j.epsl.2015.04.006

Chapter 3

Synthesizing Seemingly Contradictory Seismic and Magnetotelluric Observations in the Southeastern United States to Image Physical Properties of the Lithosphere

Benjamin S. Murphy & Gary D. Egbert

Geochemistry, Geophysics, Geosystems Volume 20, Issue 6, June 2019, Pages 2606-2625

3.1 Abstract Although seismic velocity and electrical conductivity are both sensitive to temperature, thermal lithosphere properties are derived almost exclusively from seismic data because conductivity is often too strongly affected by minor highly conductive phases to be a reliable indicator of temperature. However, in certain circumstances, electrical observations can provide strong constraints on mantle temperatures. In the southeastern United States (SEUS), magnetotelluric (MT) data require high resistivity values (>300 Ωm) to at least 200 km depth. As dry mantle mineral conduction laws provide an upper bound on temperature for an observed resistivity value, the only interpretation is that lithospheric temperatures (<1330°C) persist to 200 km. However, seismic tomography shows that velocities in this region are generally slightly slow with respect to references models; this observation has led to a view of relatively thin (<150 km), eroded thermal lithosphere beneath the SEUS. We show that MT and seismic (tomography, attenuation, receiver function) results are consistent with thick (~200 km), coherent thermal lithosphere in this region. Reduced seismic velocities (relative to reference models) can be explained by considering the effect of finite grain size (anelasticity). Calculated velocity as a function of temperature is overall slower when including anelastic effects, even at reasonable grain sizes of 1 mm – 1 cm; this permits mantle temperatures that are colder than would typically be inferred. We argue for a geodynamic scenario in which the present thermal lithosphere in the SEUS formed in association with the Central Atlantic Magmatic Province (CAMP) and has subsequently survived intact for ~200 Ma.

3.2 Introduction Many geologic and geodynamic problems require insight into the temperature distribution in some area of interest. Although multiple geophysical observables are sensitive to the thermal field in the mantle, seismic velocity is most commonly used to infer upper mantle temperature because seismic data coverage is extensive and because this geophysical observable is largely controlled by temperature (e.g., Cammarano et al., 2003). In the upper mantle, electrical conductivity (imaged by inversion of long-period magnetotelluric measurements) is also sensitive to temperature: increasing temperature causes an increase in conductivity, or equivalently a decrease in electrical resistivity (Fig. 3.1). However, minor highly conductive phases, such as water (e.g., Gardés et al., 2014), melt (e.g., Pommier & Garnero, 2014), mantle graphite (e.g., Mareschal et al., 1995; Jones et al., 2003), or even mantle sulfides (e.g., Ducea & Park, 2000), can drastically increase bulk electrical

47 conductivity (Selway, 2014) and thereby mask the comparatively moderate sensitivity to temperature (Fig. 3.1).

Figure 3.1: Resistivity as a function of temperature for a range of compositions. The gray shaded region in the lower right of the plot indicates resistivity-temperature combinations that are physically impossible in the real Earth (based on numerous laboratory mineral physics experiments). Dry mantle mineral conduction laws therefore provide an upper bound on the possible observable resistivity for a given temperature, or equivalently a maximum temperature for an obtained (i.e., inverted) resistivity value. The SEO3 model (Constable, 2006), which depends upon pressure via the oxygen fugacity dependence, is evaluated at 3 GPa. Other dry mantle mineral calibrations are from Jones et al. (2012). Hydrous olivine calibration is from Gardés et al. (2014). Partial melt conductivity is from Pommier & Garnero (2014); note that, because even small amounts of partial melt will wet grain boundaries and thereby become interconnected, the conductivity of the partial melt dominates the bulk average for the system.

It is crucial to note, however, that such minor conductive phases always decrease electrical resistivity (or, equivalently, increase electrical conductivity). Therefore, in the upper mantle, maximum bulk resistivity cannot exceed the resistivity of dry silicate minerals

(Fig. 3.1; see also Murphy & Egbert, 2017). For all major mantle minerals at lithosphere temperatures, electrical conduction is enabled by small polaron diffusion, which can be viewed as electron valency swapping (or hole migration) between Fe2+ and Fe3+ (Elliot, 1998; Selway, 2014). (Ferric iron substitution for ferrous iron introduces a hole defect state into the valence band, thereby enabling extrinsic semiconduction.) This conduction mechanism is only weakly dependent upon oxygen fugacity, iron content (magnesium number), and bulk mineralogy (less than half an order of magnitude for fO2 and less than a quarter of an order of magnitude for composition). In contrast, the effect of temperature (more than three orders of magnitude over the range of reasonable mantle lithosphere temperatures; Fig. 3.1) is much greater. Similarly, the effect of anisotropy in dry mantle mineral electrical conductivity is less than a factor of two at lithospheric temperatures (Du Frane et al., 2005), and grain size only matters when mineral grains become very small (<10 μm; Pommier et al., 2018; Selway, 2014). Consequently, in regions where the Earth is in bulk resistive, so that the significant influence of minor ‘contaminant’ phases can be ruled out, MT can provide a strong constraint on temperature. Translating the observed resistivities to temperature using empirical conduction laws derived from laboratory measurements, such as the SEO3 dry olivine conduction model of Constable (2006), provides a firm upper bound on mantle temperature. As Figure 3.1 shows, details in bulk composition will have only a minor effect on this bound, and allowing for even low concentrations of water or other conductive phases would require even lower temperatures. In practice, the conductivity of mantle lithosphere obtained from MT data is often significantly higher than would be expected for dry mantle minerals under a reasonable geotherm. One reason for this may involve the limitations of the MT technique; the mantle lithosphere is often bounded between highly conductive layers in the lower crust and the asthenosphere, so the actual resistivity is difficult or impossible to resolve. Regardless, well- resolved resistivity values are often much lower than would be expected under dry conditions, so trace conductive phases and water appear to be ubiquitous (e.g., Eaton et al., 2009; Jones et al., 2013; Selway, 2014; cf. Jones et al., 2002). Consequently, regions where MT can provide a firm constraint on temperature are relatively rare. Purely thermal interpretations of MT images have generally only been invoked in well-studied cratonic regions, such as the Kaapvaal Craton (e.g., Muller et al., 2009). The southeastern US (SEUS; Fig. 3.2) is one notable location where well-resolved electrical resistivity values are high enough that electrical conduction mechanisms other than dry temperature-dependent conduction can be ruled out (Murphy & Egbert, 2017). MT data therefore provide a firm

49 upper bound on temperature and, consequently, thermal lithosphere thickness in the SEUS. However, this MT-derived thermal constraint seems at odds with seismic observations, which appear to provide a fundamentally different view of basic lithospheric properties in this area (Fig. 3.3). Here, we review the apparent discrepancy between seismic- and MT-derived views of the lithospheric thermal fields, manifested most clearly in differing estimates of thermal lithosphere thickness, and demonstrate that recovered absolute Vs values in the SEUS are consistent with the thick (~200 km), coherent thermal lithosphere required by MT. We primarily focus on comparison to surface wave tomography, as such results are the most appropriate for mantle-lithosphere-scale structure and provide absolute shear wave velocity rather than perturbations from an assumed (and often physicochemically meaningless; Cammarano et al., 2003; Cobden et al., 2008) reference model, although we do also consider body-wave and converted-wave seismic observations. A key conclusion from this analysis is that anelastic controls on seismic velocity must be taken into account when formulating geodynamic interpretations for the deep lithosphere, at least in tectonically quiet regions. We also discuss how formation of the Central Atlantic Magmatic Province (CAMP) is most likely responsible for the lithospheric conditions that we observe today beneath the SEUS.

3.2.1 Geologic setting Although a passive margin at present, the geologic history of the SEUS comprises a complicated sequence of tectonic processes. Continental lithosphere in this region is the product of a full Wilson cycle (e.g., Hatcher, 2010) that began with the assembly of the supercontinent Rodinia in the Meso- to Neoproterozoic. Following late Neoproterozoic rifting, the modern southeastern margin of the Laurentian continent became an accretionary margin along which a series of exotic terranes were sutured through the Paleozoic. Convergent orogenies ended in the late Paleozoic with the collision of Gondwana along the modern eastern margin of Laurentia to form the supercontinent Pangea. Geodynamic analyses have indicated that this continental collision formed an Andean-scale orogenic belt (e.g., Slingerland & Furlong, 1989). In the early Mesozoic, passive rifting of Pangea led to opening of the Atlantic Ocean (Frizon de Lamotte et al., 2015). During this time, at ~200 Ma, the Central Atlantic Magmatic Province (CAMP; Fig. 3.2) erupted across the Atlantic rift zone; voluminous lava flows and dike swarms that comprise this large igneous province are found along the passive margins of all Atlantic-bounded continents (e.g., McHone, 2000). Geochemical evidence from the SEUS

50 indicates, at least in this region, that CAMP erupted in association with orogenic collapse and slab foundering and that constituent magmas were formed from a subduction-modified mantle (Whalen et al., 2015). Following the formation of CAMP and the waning of rift-flank tectonism in the Jurassic, the SEUS became a passive margin.

Figure 3.2: Map of the region considered in this study. Red lines show the distribution of Central Atlantic Magmatic Province (CAMP) dikes (from Ragland et al., 1983) in the southeastern US. Note that the Coastal Plain (tan masked region, the portion of the margin influenced by Cenozoic sea level fluctuations and therefore covered in young sedimentary rocks) obscures dike distribution towards the coast. The Piedmont physiographic region comprises the low-relief region between the Coastal Plain and the modern Appalachian Mountains (high-relief, high-elevation region shown on basemap). The orange triangle along the West Virginia-Virginia border shows the approximate location of spatially limited post-CAMP volcanic rocks (e.g., Mazza et al., 2014; Mazza et al., 2017). The white line shows the cross-section location for subsequent figures. State name abbreviations: GA—Georgia, NC—North Carolina, SC—South Carolina, TN—Tennessee, VA—Virginia, WV—West Virginia.

There have been no further margin-wide tectonic or magmatic events since rifting and formation of CAMP, although there is geologic, geochemical, and geophysical evidence for localized post-CAMP modification of the SEUS lithosphere. Spatially limited outcrops of late-Jurassic- and Eocene-aged volcanic rocks along the border between Virginia and West Virginia (Mazza et al., 2014; Meyer & van Wijk, 2015; Mazza et al., 2017; see Fig. 3.2) that are coincident with both a slow seismic anomaly (e.g., Wagner et al., 2018) and a conductive geoelectric anomaly (Evans et al., 2016) in the mantle lithosphere provide evidence for post-

51 rifting lithospheric modification, although more consideration is needed to determine whether such modification is due to hotspot impingement (e.g., Chu et al., 2013), to a delamination event (e.g., Mazza et al., 2014; Schmandt & Lin, 2014; Meyer & van Wijk, 2015), to localized thermal erosion (e.g., Evans et al., 2016), to a combination of such processes, or to some completely other process. Regardless, such post-200 Ma lithospheric evolution is spatially localized. Eruption of CAMP was the last major margin-wide event to affect the mantle lithosphere beneath the SEUS.

3.2.2 Previous seismic results The lithosphere in the SEUS has been imaged by numerous surface-wave tomography studies that used a variety of datasets and inversion approaches (e.g., Schmandt et al., 2015; Porter et al., 2016; Shen & Ritzwoller, 2016; Pollitz & Mooney, 2016; Savage et al., 2017; Wagner et al., 2018). These surface-wave imaging results generally show heterogeneous, small-amplitude deviations from reference models in the Piedmont and Coastal Plain physiographic provinces (Figs. 3.2, 3.4, 3.5). Although these seismic results strictly only provide insight into properties of the seismic lithosphere-asthenosphere boundary (LAB), which in itself does not necessarily denote a physicochemical boundary or map in a simple way to the thermal-rheological LAB (e.g., Eaton et al., 2009), these images have nevertheless been interpreted to show relatively thin (<150 km) thermal lithosphere beneath the SEUS (e.g., Fig. 3.3). They have further been used to argue that the mantle lithosphere in this region has been modified since opening of the Atlantic by passive continental margin processes such as delamination and lithospheric erosion by edge convection (e.g., Wagner et al., 2018). Figures 3.4 and 3.5 compare five of these surface-wave tomography models to each other and to the MT results. (See also Table A3.1 for a summary of the datasets and inversion approaches used to derive each solution.) Body-wave tomography studies similarly show mottled small-amplitude deviations from standard velocity reference models (Biryol et al., 2016; Schmandt & Lin, 2014). With velocity anomalies generally skewed toward relatively slightly slow, these results have also been used to argue for piecemeal delamination beneath the SEUS (Biryol et al., 2016). Specifically, this interpretation envisions lithospheric loss occurring incrementally across the passive margin over the last ~200 Ma. Converted-wave imaging results (using Sp phases at ~0.1 Hz) generally reveal a negative velocity gradient at approximately 70 km in this region (Hopper & Fischer, 2018; Liu & Gao, 2018).

3.2.3 Previous MT results Multiple MT imaging studies, which all used drastically differing inversion approaches, have consistently revealed high resistivity values beneath the SEUS (Fig. 3.3; Wannamaker, 2005; Evans et al., 2016; Murphy & Egbert, 2017; Gribenko & Zhdanov, 2017). Extensive resolution tests indicate that resistivity values exceeding 300 Ωm must persist to at least 200 km depth (Murphy & Egbert, 2017). As dry mantle mineral electrical conduction laws provide an upper bound on temperature for a given observed resistivity (Fig. 3.1), the MT data require lithospheric temperatures (<1330°C) to at least 200 km depth. Note that the MT data also essentially require that the lithosphere in this region be completely dry. Even a small amount of incorporated water would require colder temperatures, and therefore a deeper thermal lithosphere-asthenosphere boundary, to explain such high resistivity values (Fig. 3.1). The MT inverse solution of Murphy & Egbert (2017) is compared to surface-wave tomography results in Figures 3.4 and 3.5. This highly resistive structure (herein referred to as the Piedmont Resistor) is apparent in the raw MT data for this region (Murphy & Egbert, 2017), and other observations from various electromagnetic datasets support the presence of a large, resistive body beneath the SEUS. For example, source biases due to Pc3-4 geomagnetic pulsations, indicative of bulk resistive Earth, are ubiquitous in MT transfer functions in the SEUS (Murphy & Egbert, 2018). Resistive upper mantle beneath the SEUS also appears in the global geomagnetic depth sounding (GDS) results of Sun et al. (2015). Although lithospheric structure in those GDS results is poorly resolved due to the period range of the dataset, this observation provides conditional but independent (as GDS relies on a different set of assumptions than MT) support for the presence of highly resistive lithosphere beneath the SEUS.

Figure 3.3: Comparison between MT (top: Murphy & Egbert, 2017) and seismic surface-wave (bottom: Wagner et al., 2018) results. Solid lines indicate thermal lithosphere-asthenosphere boundary (LAB) estimates for each geophysical technique independently, without taking into account the effects we consider here. Based on the usual interpretation approaches, these images provide significantly different views of thermal lithosphere thickness. The white line in the upper plot is drawn following the interpretation of Murphy & Egbert (2017), which honors MT resolution tests that indicate the MT data require high resistivities (>300 Ωm), and hence lithospheric (<1330°C) temperatures, to 200 km depth. The black line in the lower plot is drawn to follow the 0% Vs anomaly contour (reference value is 4.5 km/s, as used by Wagner et al. [2018] in their imaging study). Cross sections follow the transect shown in Figure 3.2. Horizontal axis tick interval is 100 km.

Figure 3.4: Map views of average Vs over 75-150 km depth from the five surface wave tomography models considered here (Schmandt et al., 2015; Porter et al., 2016; Shen & Ritzwoller, 2016; Pollitz & Mooney, 2016; Wagner et al., 2018) and geometric mean resistivity (Murphy & Egbert, 2017) over the same depth range. Surface wave models are quite variable. Models are plotted roughly at native resolution. The white dots on the resistivity map show the points at which model depth profiles are extracted for subsequent calculations.

Figure 3.5: Depth profiles, obtained from the points shown in Figure 3.4, from the five surface wave models (Schmandt et al., 2015; Porter et al., 2016; Shen & Ritzwoller, 2016; Pollitz & Mooney, 2016; Wagner et al., 2018) and the electrical resistivity inverse solutions. The solid line is the median value of all profiles, and the shaded regions indicate the range of extracted profiles (darker region is the 25th-75th percentile; lighter region is the full range). Note that there is considerable variability between models. Also plotted are velocities from the AK135 and PREM reference models. The lower right panel shows the median resistivity profiles from the inverse solution of Murphy and Egbert (2017) and the refined resistivity solutions shown in Figure 3.6.

3.2.4 Anelasticity, grain size, and seismic velocity Seismic velocities are controlled by two separate effects: anharmonicity and anelasticity (e.g., Karato, 1993). Anharmonicity describes how seismic velocities change (via the constituent elastic parameters) as a function of temperature, pressure, and composition, but it ignores the fact that rocks are composed of discrete crystal grains. Anelasticity describes the effect that grain boundaries have on seismic properties; this component is dependent upon temperature, pressure, and seismic wave frequency. At low enough frequencies, grain boundaries do not behave elastically; instead, they behave viscously. Theories of anelasticity are parameterized in terms of a critical frequency at which a continuum of viscoelastic/viscous grain-boundary processes begins to exert significant control on seismic observables (e.g., Jackson & Faul, 2010; Olugboji et al., 2013; Karato et al., 2015). Around this critical frequency, the value of which depends on temperature, pressure, and grain size, elastically accommodated grain-boundary sliding (EAGBS) causes a peak in attenuation and a major reduction in shear modulus, which consequently decreases both Vp and Vs. (For example, at 1000°C and 4 GPa, for a grain size of 5 mm, the critical frequency is ~5 Hz, or ~0.2 s.) At lower frequencies (longer periods, seconds to hundreds of seconds for the specified conditions), diffusionally accommodated grain-boundary sliding (also termed ‘absorption band behavior’) leads to a slow but continual increase in attenuation and decrease in shear modulus. At very low frequencies (very long periods, hundreds of seconds for the specified conditions), steady state creep further significantly increases attenuation and decreases the shear modulus. (See the Supplemental Information of Karato et al. [2015] for a graphical summary of this continuum of anelastic behavior.) Anelasticity has long been known to be an important control on seismic velocities (e.g., Karato, 1993), but only recently have rigorous attempts been made to quantify the magnitude of this effect (e.g., Jackson & Faul, 2010; Karato et al., 2015) and to integrate its consequences into geophysical interpretation (e.g., Goes et al., 2012). Though much of the current research on anelasticity has been specifically targeted at understanding the cause of mid-lithospheric discontinuities (MLDs; e.g., Karato et al., 2015, Selway et al., 2015) and at characterizing the nature of the lithosphere-asthenosphere boundary (LAB; e.g., Olugboji et al., 2013; Olugboji et al., 2016), the proposed models of anelastic controls on seismic observables can, and indeed should, be applied more broadly when formulating geodynamic interpretations of observed seismic anomalies. Some work has taken anelasticity into account when determining seismic constraints on temperature (e.g., Goes et al., 2005; Yang & Forsyth, 2008; Armitage et al., 2015; Eeken et al., 2018). However, all too frequently seismic

57 velocity anomalies, expressed as percent deviations from an assumed reference model, are only qualitatively interpreted in terms of relative temperatures, with slow (fast) anomalies being translated directly into hot (cold) lithosphere or even asthenosphere. Different models of anelasticity invoke different dependences upon grain size (e.g., Jackson & Faul, 2010; Karato et al., 2015; see also section A3.2). However, regardless of the exact model chosen for the anelastic contribution to seismic observables, grain size is an important parameter. In the upper mantle, mineral grain size is controlled by the balance between static grain growth and bulk strain (e.g., Dannberg et al., 2017). At elevated temperatures, crystal grains tend to grow in order to minimize the Gibbs free energy of their surface interfaces (Evans et al., 2001). Strain tends to reduce grain size via dislocation creep, with higher strain rates causing a greater reduction in grain size (Austin & Evans, 2007). Calculations leveraging semi-empirical models of these effects predict an equilibrium grain size of 1 mm – 1 cm for a strain rate of ~10-17-10-16 s-1, the upper bound on lithospheric strain rate in the central and eastern United States (e.g., Calais et al., 2016; Kreemer et al., 2018). Petrographic analyses of peridotite xenoliths suggest that grain sizes of ~1mm – 1 cm are typical for the mantle lithosphere (e.g., Armienti & Tarquini, 2002; Ave Lallemant et al., 1980).

3.3 Methods 3.3.1 Resistivity refinement and mapping to temperature Although the results of Murphy & Egbert (2017) constrain the depth extent of the highly resistive lithosphere (the Piedmont Resistor), the resistivity-depth profiles in the most resistive portion of that structure cannot be interpreted in terms of a simple resistivity- temperature relationship, such as a dry olivine conduction law. As suggested by Figure 3.5, translating the average resistivity profile to temperature (Fig. 3.1) would lead to an unphysical geotherm with essentially constant temperature between 60 km and 200 km depth. Therefore, in order to obtain a more physically meaningful electrical resistivity solution, we redistribute the high resistivity values from the Murphy & Egbert (2017) inverse solution to be consistent with resistivity profiles calculated from the SEO3 dry olivine conduction law (Constable, 2006) using several assumed geotherms. We also take into account empirical observations of decreasing oxygen fugacity (fO2) with increasing pressure in the mantle lithosphere (Frost & McCammon, 2008). We perform this calculation iteratively in order to find the solution with the thinnest thermal lithosphere allowed by the MT data. We summarize the procedure below, and provide further details in section A3.3.

First, we take the Murphy & Egbert (2017) inverse solution and limit high resistivity values according to a dry resistivity profile calculated from the SEO3 model with a linear geotherm that is constructed between the surface and an assumed thermal lithosphere- asthenosphere boundary (LAB) depth, initially set to 120 km. We also impose this resistivity- depth profile in the off-shore (oceanic) lithosphere and impose a connection between that lithospheric domain and the Piedmont Resistor in order to better account for a possible large galvanic coast effect (Ranganayaki and Madden, 1980; Evans et al., 2002). (Essentially, coast-perpendicular current channeling within a highly conductive oceanic layer underlain by highly resistive ocean lithosphere could possibly distort the lithospheric Earth electromagnetic response; by forcing a connection between the Piedmont Resistor and resistive oceanic lithosphere, we are attempting to better take into account this possibility.) We use this modified resistivity model as the starting point for a new data inversion. Once the inversion has converged, we modify regions in the resulting solution that have become more resistive at depths exceeding the previously assumed LAB depth to limit those resistivities according to a new geotherm with the LAB at a greater depth. Then, we again restart the inversion. We repeat this process until the inversion stops increasing resistivity values beneath the specified LAB depth. Finally, we perform a final refinement using three different temperature profiles: (1) a uniformly linear geotherm; (2) a nonlinear (parabolic) geotherm that enforces a Moho heatflow value derived from a mantle reduced heatflow calculation (in which measured heatflow is plotted against measured heat production in order to back out the crustal basal heatflow); and (3) a nonlinear geotherm that uses surface geothermal gradient observations to constrain the near-surface temperature profile and that attempts to incorporate mid-lithospheric discontinuity (MLD) observations by using the EAGBS model of Karato et al. (2015) to inform mantle temperature. (See section A3.4. Heatflow data were obtained from the SMU Geothermal Laboratory database; http://geothermal.smu.edu/.) Our inversion techniques are similar to those described by Murphy & Egbert (2017), although we use a slightly different dataset in this study (see section A3.3). The SEO3 dry olivine conduction law (Constable, 2006) plays a key role in deriving these modified inverse solutions and in converting resistivity uniquely back to temperature. It is important to note that SEO3 allows for variations in oxygen fugacity (fO2). Petrologic observations indicate that fO2 generally decreases by several log units from the FMQ buffer with increasing pressure at lithospheric depths (2 – 7 GPa; Frost & McCammon, 2008). We include this depth-dependent fO2 decrease in converting between resistivity and temperature at a given depth.

3.3.2 Seismic velocity and attenuation calculations We use the preferred, modified Burgers model of Jackson & Faul (2010) to calculate seismic velocities (Vs, Vp) and quality factor (Q) as a function of temperature, depth, and grain size. For these calculations, we use pressure values that are appropriate for the specified depth. For calculations pertaining to surface-wave observations, we use a wave period that is representative of the surface wave period range that would sense the specified depth (Forsyth, 1992). For calculations pertaining to body-wave observables, we use a wave period of 10 s (0.1 Hz) and a grain size of 5 mm. Although the model of anelasticity proposed by Karato et al. (2015) requires further experimental work to constrain the numerical values of parameters, we also perform calculations with their framework to demonstrate that our conclusions depend little on the exact model for anelastic effects. (See section A3.2.) Unlike the model of Jackson & Faul (2010), in which the magnitude of anelastic velocity reduction depends upon grain size, the model of Karato et al. (2015) predicts that grain size only affects the frequency dependence of anelasticity, not the magnitude of the velocity reduction. Therefore, this model is only evaluated for a grain size of 5 mm.

3.4 Results The final refined resistivity solutions, based on the three geotherms described in the Methods, are shown in Figure 3.6. These are the solutions with the thinnest resistive block and, consequently, the thinnest thermal lithosphere allowed by the MT data. The geotherms used to construct these final inverse solutions have a thermal LAB (1330°C isotherm) at 200 km, although coarse model discretization at this depth results in some ambiguity in depth to the electrical LAB (the ~300 Ωm contour) in these images. The final refined inverse solutions all fit our dataset equally well (see section A3.3), despite clear differences in resistivity distributions (cf. Figs. 3.5, 3.6). Due to the physics of electromagnetic induction in the Earth, the MT technique is insensitive to even order-of-magnitude variations in highly resistivity structures. Long-period electromagnetic fields will diffuse rapidly with little attenuation through any highly resistive layer, such that, for example, there will be little difference in Earth response between 104 and 105 Ωm material. In our inversions, this insensitivity means that we can significantly adjust high resistivity values over a large range with very little impact on the predicted data; that is, the high resistivity values of the Piedmont Resistor are so high that even order of magnitude increases are in the null space of the MT inverse problem. Therefore, as a range of high resistivity-depth profiles could fit our data, a wide range of geotherms is also allowed. Thus, although we constrain the thermal lithosphere to be

60 at least 200 km thick beneath the Piedmont, since resistivity values >300 Ωm must persist at least to that depth, we poorly constrain the detailed temperature distribution within that lithospheric column. Through our iterative inversion process, we find that high resistivity values are only required by our data to ~200 km beneath the Piedmont; high resistivities in our solutions terminate at a shallower depth beneath the Coastal Plain (Fig. 3.6; see also section A3.3). However, resolution tests (see section A3.5) indicate that our data would allow high resistivity values to persist to ~200 km beneath that region, as they do further inland, so the depth extent of the resistive block is not well resolved by onshore MT data alone. This finding contrasts with the conclusion of Murphy & Egbert (2017) and apparently reflects our efforts to better account for the galvanic coast effect. Previous resolution tests to examine the effect of resistive offshore lithosphere on recovered on-land structure (Figs. S10, S11 of Murphy & Egbert, 2017) suggested that a thick resistor was also required under the Coastal Plain. Those tests included a moderately conductive (~100 Ωm) boundary region between the resistive oceanic lithosphere and the continental lithosphere, beneath the continental shelf. We have no a priori reason to believe that highly resistive lithosphere extends uninterrupted from the continent into the ocean basin (as assumed in our present solutions), but in light of this uncertainty we conclude that our data permit greater leeway in thermal lithospheric thickness directly along the coastline, in contrast to the requirement for a thick (200 km) cold layer slightly further inland. It is interesting to note that our reanalysis of the Murphy & Egbert (2017) inverse solution places particularly high resistivity values at greater depth under western North Carolina (see section A3.3). However, this feature may be an artifact of imposing a connection between the Piedmont Resistor and resistive oceanic lithosphere. Again, as we have no reason to believe that highly resistive lithosphere does in fact extend continuously from the continent into the ocean basin, we cannot draw any meaningful conclusions about this result. We refrain from analyzing this feature until seafloor MT data are available to constrain the electrical characteristics of the lithosphere along the continental margin. In Figure 3.7, we compare calculated surface wave velocity and attenuation (as a function of temperature, depth, and grain size) to seismic observations from the SEUS. Temperatures required for the computations are derived (via SEO3, accounting for depth variations in fO2, as discussed in the Methods) from resistivity profiles extracted from the solutions shown in Figure 3.6 (at the points shown in Fig. 3.4). As this comparison demonstrates, by invoking finite grain size, we can explain the combination of relatively low observed Vs values and

61 high resistivity values. There is considerable variability between seismic models, but overall grain sizes of ~1 mm – 1 cm (reasonable for mantle lithosphere, as discussed in Section 3.2.4) reduce Vs sufficiently at the temperatures the MT observations require. Some models (Schmandt et al., 2015; Shen & Ritzwoller, 2016) are relatively fast through the lithospheric column and do not necessarily require invoking grain size effects. The model of Pollitz & Mooney (2016) is somewhat of an outlier, but velocities at 120 km and 150 km in their model are still consistent with colder temperatures when taking into account reasonable grain sizes. The model of Wagner et al. (2018) is consistent at all depths when taking into account grain sizes of 1 mm – 10 cm. The model of Porter et al. (2016) generally matches our MT-based temperatures inferences with grain sizes of ~10 cm; although slightly larger than typically assumed for the mantle, we nevertheless take this as evidence that anelasticity can indeed reconcile the seemingly disparate MT and seismic results. Because our MT data do not constrain the maximum bulk resistivity values within this structure, we can increase resistivity and thereby decrease temperature at a given depth within the Piedmont Resistor and still fit the data. In Figure 3.7, this adjustment would have the effect of shifting the observation points to the right in each plot, such that observation pairs from the models of Schmandt et al. (2015), Shen & Ritzwoller (2016), and Porter et al. (2016) would fall along the predicted curves for realistic grain sizes of ~1 mm – 1 cm. Furthermore, additional slight adjustments would be possible by considering more realistic compositions (that is, a combination of olivine and pyroxene rather than just olivine, as we consider here). Therefore, given the limitations of the MT data and of our analysis here, the fact that the seismic observations almost always plot on or above the predicted velocity-temperature (Vs-T) curves for 1 mm – 1 cm grain size in Figure 3.7 demonstrates that, by considering the anelastic contribution to seismic velocity, we can reconcile seemingly disparate geophysical observables within a common lithospheric physical state. Observed attenuation values (Bao et al., 2018) are similarly generally consistent with grain sizes of ~1 mm – 1 cm.

Figure 3.6: Final refined resistivity solutions (right column) and the temperature profiles (left column) used to generate the vertical maximum resistivity distribution. Top row shows the refined inverse solution for a linear geotherm with a thermal lithosphere-asthenosphere boundary (LAB), taken to be the 1330°C isotherm, at 200 km (geotherm 1). The middle row shows the refined solution for a parabolic geotherm constructed from a mantle reduced heatflow calculation with a thermal LAB at 200 km (geotherm 2). The bottom row shows the refined solution for a nonlinear geotherm informed by mid-lithospheric discontinuity (MLD) observations with a thermal LAB at 200 km (geotherm 3). The dotted black/white lines in these bottom plots show the approximate depth of observed MLDs in the SEUS (~70 km for ~10 s wave periods; Hopper & Fischer, 2018; Liu & Gao, 2018). See section A3.4 for more information about derivation of these geotherms. Resolution tests (see section A3.5) indicate that high resistivity values are permitted, but not required (as they are further inland), by our data to ~200 km beneath the Coastal Plain. High near-surface conductivities in the Coastal Plain region may limit our sensitivity at great depth here. Cross sections follow the transect shown in Figure 3.2. Horizontal axis tick interval is 100 km.

Figure 3.7: Comparison between Vs (top row), obtained from the five different surface-wave inversions shown in Figures 3.4 and 3.5; Q-1 (bottom row), obtained from the work of Bao et al. (2016); and MT-derived resistivity values from our work here. We map resistivity directly to temperature using the SEO3 dry olivine conduction law with a depth-dependent fO2 offset, as explained in the main text. This comparison is provided at three different depths (corresponding to the three different columns). The solid colored lines in these plots show temperature (and equivalently resistivity) versus predicted Vs and 1/Q, with different colors corresponding to different grain sizes, calculated from the modified Burgers model of Jackson & Faul (2010). The solid black line shows the predicted temperature/resistivity-Vs relationship from the model of Karato et al. (2015). The black dashed line shows the calculated temperature/resistivity-Vs relationship using only the anharmonic pressure and temperature derivatives of the bulk and shear moduli. Symbols in the upper row show the median resistivity-Vs pair extracted from the MT and seismic models; sample points are shown in Figure 3.4 (and are the same as those used to make Fig. 3.5). The bars on these symbols show the full range of extracted Vs values. The different colors of these symbols denote different surface wave models (SR16 = Shen & Ritzwoller, 2016; PLH16 = Porter et al., 2016; PM16 = Pollitz & Mooney, 2016; W18 = Wagner et al., 2018; SLK15 = Schmandt et al., 2015). The three different plotted resistivity values for each seismic model correspond to the three different geotherms shown in Figure 3.6: diamonds for geotherm 1 (linear; upper panel); squares for geotherm 2 (parabolic; middle panel); circles for geotherm 3 (bottom panel). The points in the bottom row similarly show the median observed resistivity-Q-1 pair when comparing attenuation maps (Bao et al., 2016) to our MT images (only available for the 80 s case here). In almost all cases, the relatively slow velocities (compared to reference models) at high resistivities (and therefore cold temperatures) can be explained by reasonable grain sizes of ~1 mm – 1 cm. In some cases, even the predicted purely anharmonic velocity can explain observations.

The predicted Vs values from the model of Karato et al. (2015) over the range of temperatures used to fit the MT data are in fact slightly slower than almost all seismic observations, even when assuming a relatively small velocity reduction as we do here (see section A3.2). Although the model of Karato et al. (2015) requires further laboratory experiments to constrain numerical parameters, this model also can explain the relatively slow seismic velocities with the effects of anelasticity. Again, because our observation points in Figure 3.7 almost always plot above the predicted Vs-T curves for this model (that is, at higher S-wave velocities), we can conclude that this alternative model for anelasticity can also explain both seismic and MT observations.

3.5 Discussion 3.5.1 Role of anelasticity in resolving thick thermal lithosphere By taking into account anelastic controls on seismic observables, we find that both MT and seismic results are consistent with ~200 km-thick thermal lithosphere beneath the SEUS. In almost all cases, with reasonable mantle grain sizes of ~1 mm – 1 cm or greater, calculations for the cold (lithospheric) temperatures required by the MT data provide a reasonable match to seismic observations. Given the significant scatter in surface-wave velocity estimates, the poor sensitivity of the MT method to exactly how resistive the Piedmont Resistor truly is, and the fact that we ignore possible compositional effects in our analysis, the exact lithospheric temperature at a given depth is not well constrained here, even when jointly interpreting both datasets. Regardless, we are able to unify the seemingly contradictory seismic and MT results into a common picture of thick, coherent thermal lithosphere beneath the SEUS. As our analysis demonstrates, this result does not depend on the assumed model of anelasticity. Previous attempts to reconcile the seemingly disparate seismic and MT results (Murphy & Egbert, 2017) invoked the effects of unusual mantle compositions to provide relatively slow seismic velocities at the cold temperature inferred from the MT observations. In contrast to this somewhat contrived interpretation, anelastic effects provide a much simpler and more natural explanation of the slightly slow (with respect to reference models) seismic velocities observed in this area. As grain-size effects operate universally in the mantle, relying on these anelastic calculations does not require invoking any special or unusual geodynamic or geologic circumstances. Certainly, compositional variations that we ignore here may also exert control on the observed seismic properties. However, our results demonstrate that the

65 seismic and MT observations can be easily synthesized via a simple physical mechanism without needing to specify constituent mineralogy.

3.5.2 Role of oxygen fugacity In our calculations, we have included an empirical depth-dependent offset in oxygen fugacity (as discussed in the Methods). Though overall a small effect, this component is in fact important in synthesizing our MT results with seismic results, as shown in Figure 3.8. In accounting for the decrease in fO2 (with respect to the FMQ buffer) with increasing pressure when mapping between resistivity and temperature, we obtain a slightly warmer mantle than would otherwise be inferred for a given resistivity value. These marginally warmer temperatures improve the match between observed seismic velocities and electrical resistivities at grain sizes reasonable for typical (‘normal’) mantle lithosphere (Fig. 3.8). As demonstrated in Figure 3.8, without the depth-dependent fO2 offset (with respect to the FMQ buffer), we would require smaller grain sizes (<1 mm) to match the observations of Wagner et al. (2018), and the results of Pollitz & Mooney (2016) would become more extreme outliers. As such small grain sizes are less reasonable for typical mantle lithosphere, the depth-dependent fO2 offset is an important effect that helps integrate this set of geophysical observables. Recent experimental work (Cline et al., 2018) suggests that seismic velocities may also depend on oxygen fugacity via anelasticity. Specifically, Cline et al. (2018) find a relationship between Q and fO2 such that oxidizing conditions enhance attenuation. As we use observations that mantle lithosphere fO2 decreases (becomes more reducing with respect to the FMQ buffer) with depth, these recent experimental observations would suggest that Vs (via its relationship to Q) should be slightly increased (fast) at more reducing conditions. In Figure 3.7, this would have the effect of shifting predicted curves upwards slightly at greater depths and would potentially provide a better match between these predictions and some of the faster surface-wave-imaged velocities. At more reducing conditions, the Cline et al. (2018) observations would also suggest that attenuation should decrease (Q should become larger). The predicted attenuation curves in the lower plots of Figure 3.7 would therefore shift downwards and perhaps better match the observations of Bao et al. (2016).

Figure 3.8: Comparison of predicted resistivity-Vs relationships (at 180 km) that are computed with (top) and without (bottom) depth-dependent fO2 offset with respect to the FMQ buffer. The resistivity- temperature mapping in the top plot uses an fO2 value approximately 4 log units below the FMQ buffer for the SEO3 calculation, per petrologic observations (Frost & McCammon, 2008). The mapping in the bottom plot uses an fO2 value given by the FMQ buffer. Plot symbology is as in Figure 3.7. By including the fO2 offset (with respect to the FMQ buffer) with depth, resistivity values are mapped to slightly warmer temperatures. These warmer temperatures facilitate a better match to seismic observations.

3.5.3 Body wave tomography and lithospheric discontinuities Although we have mainly considered surface-wave imaging here, our arguments for the importance of anelastic effects in interpreting seismic observations extend to other forms of seismic imaging as well. Anelasticity affects shear modulus, upon which both Vs and Vp depend. As shown in Figure 3.9, this effect is stronger for Vs, but it still does affect Vp. For

67 our assumed geotherms, predicted Vp is in fact slightly slow with respect to the AK135 reference model at depths of ~150-200 km. Therefore, although body wave images of the SEUS show slightly slow velocity anomalies for Vp, which have been used to argue for lithospheric destruction along the continental margin (e.g., Biryol et al., 2016), our conclusions regarding anelastic effects can also reconcile those observations with thick thermal lithosphere beneath the SEUS. Karato et al. (2015) argue that anelasticity can explain observations of mid-lithospheric discontinuities, which for the SEUS are observed (at wave periods of ~10 s) at depths of ~70 km (Hopper & Fischer, 2018; Liu & Gao, 2018). Based upon this model, the observed lithospheric discontinuities could be explained as abrupt negative velocity gradients (NVGs) associated with the onset of anelastic effects within the 200 km thick thermal lithosphere of the SEUS, as shown in Figure 3.9. Note that, in the model proposed by Karato et al. (2015), the location of these NVGs depends crucially on the assumed geotherm. Adjustments to the geotherm that we use here (permitted by the MT data, as discussed in the Results) could vertically adjust the NVGs in Figure 3.9 to better match observations. Of course, as we specifically constructed our nonlinear geotherm, the third in Figure 3.6, by taking into account MLD observations, we are able to reproduce the sharp drop in velocity at roughly the correct depth range. However, with our parabolic geotherm, the second in Figure 3.6, we could make shallow mantle temperatures slightly warmer, as permitted by our MT data, in order to move the NVG upwards and thereby better match observations.

Figure 3.9: Predicted body-wave velocities (calculated at 0.1 Hz) as a function of depth using the models of Jackson & Faul (2010; JF10) and Karato et al. (2015; KOP15). For these calculations we have used a grain size of 5 mm. The upper two plots are calculated using the parabolic geotherm based on our mantle reduced heatflow calculation (geotherm 2; middle row in Fig. 3.6). The lower two plots are calculated using the nonlinear geotherm constrained by mid-lithospheric discontinuity and surface geothermal gradient observations (geotherm 3; bottom row in Fig. 3.6). Symbology is generally similar to that used in Figure 3.7.

3.5.4 Origin of the Piedmont Resistor We interpret the formation of this highly resistive geoelectric structure to be associated with the eruption of CAMP, the last major geologic event in this region. The highest resolved resistivity values are found beneath the Piedmont, spatially associated with CAMP (cf. Figs. 3.2, 3.4). High resistivity values require cold temperatures and also strongly suggest that the mantle lithosphere is completely dry (no dissolved water in nominally anhydrous minerals). Because water is incompatible in the mantle during melting (e.g., Kohn & Grant, 2006), CAMP magmatism would have removed all dissolved water in nominally anhydrous minerals such as olivine and would have thereby left the mantle lithosphere completely dry.

Melt depletion (by increasing Mg#) can cause up to a 1.7% increase in Vs (Lee, 2003; Schutt & Lesher, 2006; Schutt & Lesher, 2010; Afonso & Schutt, 2012). In Figure 3.7, this would have the effect of shifting all predicted Vs-temperature curves upwards by ~0.05 km/s (and would in fact better match the Shen & Ritzwoller [2016] and Schmandt et al. [2015] results). This is small in comparison to the anelastic velocity changes and especially in comparison to the scatter in model values (cf. Figs. 3.4, 3.5, 3.7), so the role of melt depletion is secondary to the role of anelasticity in explaining the observed combinations of seismic velocities and electrical conductivities. The small influence of compositional effects of course meshes with usual upper-mantle tomography interpretation approaches that only invoke temperature perturbations (e.g., Cammarano et al., 2003). Here, it also reinforces our conclusion that anelasticity is the most important effect in synthesizing all available geophysical observables in this area. We stress, however, that our analysis does not preclude the possibility of compositional variability within the sub-SEUS lithospheric column. Rather, our results demonstrate that the seismic and MT results can be integrated without needing to assume any specific composition or mineralogy. We interpret the low-amplitude, heterogeneous nature of recovered seismic velocities in this region to reflect spatial variability in melt depletion and possibly minor spatial variability in temperature and grain-size distributions. The piecemeal-delamination interpretation based on seismic results alone is geologically problematic, as there is a distinct lack of any clear surface expressions of this process, such as post-CAMP volcanic rocks beyond the limited Virginia-West Virginia border region, localized heatflow anomalies outside that border zone, or localized paleosurface deflections that can uniquely be associated with a delamination event (cf. Elkins-Tanton, 2007). The seismic interpretation is further precluded by our MT results. Therefore, the seismic observations must be explained by heterogeneity in composition, grain size, and/or lithospheric temperatures left after widespread but perhaps incongruous mantle melting. The highly electrically resistive structure beneath the Piedmont has alternatively been interpreted as the result of a Carboniferous-Permian (i.e., pre-CAMP) delamination, mantle- melting, and lithosphere-regrowth event (Wannamaker, 2005). Structural data have been used to argue for a large-scale delamination event during the climactic collisional (Alleghanian) orogeny that formed Pangaea (e.g., Sacks & Secor, 1990; Nelson, 1992). Similarly, geochemical data indicate that Alleghanian-aged granitic rocks within the Appalachian orogen formed solely as the result of crustal heating and anatexis (Samson et al., 1995), which could be the result either of a lithospheric delamination event or of significant crustal

70 thickening. A possible alternative explanation for the Piedmont Resistor therefore could be that such a late-Paleozoic delamination event, during construction of the Andean-scale Appalachian orogen, led to removal of the sub-Appalachian lithospheric mantle, melting of the upper mantle that would have purged volatiles, and subsequent subcretion of dry mantle material (Wannamaker, 2005). We consider this model unlikely, as there is no evidence for input of mantle melts, either as a geochemical component of anatectic rocks or as discrete igneous units, into the Alleghanian crust (Samson et al., 1995). Furthermore, it is unclear how such a coherent block of mantle lithosphere would survive intact through the conclusion of supercontinent formation and the subsequent orogenic collapse, rifting, and large-igneous- province eruption. We consider the link to CAMP to be a better explanation for the Piedmont Resistor.

Figure 3.10: Comparison between the geodynamic interpretation from seismic results alone (left) and the geodynamic framework illuminated by jointly interpreting seismic and MT results (right). As discussed in the main text, our MT data preclude the seismically derived interpretation of lithospheric erosion along the passive margin. Rather, our MT data indicate that the thermal lithosphere beneath the SEUS is thick (~200 km) and coherent. Seismic results are consistent with this interpretation. Cross sections follow the transect shown in Figure 3.2. Surface elevations above sea level are vertically exaggerated by a factor of 20. Elevation from ETOPO (Amante & Eakins, 2009); Moho depth from Crust1.0 (Laske et al., 2013).

3.5.5 Modern geodynamics of the southeastern United States The observation of heterogeneous, slightly slow seismic velocity anomalies with respect to reference models has led many previous investigators to the interpretation that continental lithosphere along the eastern North American margin has been eroded over time by edge convection and has experienced piecemeal delamination since the rift-to-drift transition (e.g., Biryol et al., 2016; Savage et al., 2017; Wagner et al., 2018). However, taken together, the

MT and seismic observations support a modern geodynamic picture of relatively thick, coherent thermal lithosphere beneath the SEUS. Furthermore, as the MT data require a large, coherent resistive block beneath the Piedmont and Coastal Plain, they also preclude the possibility of margin-wide small-scale moderately conductive features that would be expected if there were in fact lithospheric replacement by hot, asthenospheric material. In contrast to its insensitivity to the absolute values within highly resistive structures, MT is highly sensitive to conductive (<300 Ωm) structures. Our MT dataset would easily resolve any small-scale (~10s of km), moderately conductive features at lithospheric depths within the Piedmont Resistor. Indeed, our iterative inversion process reveals two highly localized conductive anomalies (see Fig. A3.3.5), although neither is clearly associated with a relatively slow seismic anomaly (cf. Fig. 3.4). Our data do not reveal any other conductive structures of similar or larger scale. Therefore, when taken together, the seismic and MT data are not consistent with the view that the lithosphere along the passive eastern margin of North America has been destroyed over time. Modest heterogeneities in recovered seismic velocities in this area thus must represent variations in composition or physical state within a broadly coherent block of thermal lithosphere. Figure 3.10 schematically illustrates the modern geodynamic configuration illuminated by jointly interpreting these datasets. It is also interesting to speculate as to whether the Piedmont Resistor may even represent a Mesozoic example of craton formation. As discussed above, the high resistivity values of this structure effectively require that the constituent mantle lithosphere be completely dry. Under such dehydrated conditions, and with lithospheric temperatures extending to at least 200 km depth, the mantle lithosphere in this region is most likely highly viscous and rheologically strong (cf., Karato, 2010). As also discussed above, CAMP likely depleted the SEUS lithosphere to some extent, so that the constituent mantle is likely buoyant. Therefore, by increasing lithospheric strength and buoyancy, the eruption of CAMP may have produced the conditions necessary to ensure the stability of this continental lithosphere for >1 Ga (cf. Wang et al., 2014). Given the intense debates pertaining to the formation of cratons (e.g., Griffin et al., 2009; Lee & Chin, 2014), the SEUS clearly warrants further study.

3.5.6 Comparison to cratonic regions Our results here beg a major question regarding the relationship between seismic and MT results: Why do we image moderate seismic velocities and very high resistivities in the SEUS, but fast seismic velocities and moderately high resistivities in cratonic regions? The answer

72 to this question likely involves the order of magnitude difference in age between these domains. Cratonic regions have had ~2 Ga to cool, thereby promoting high seismic velocities, but they have also had that long to become metasomatized. Over this long time frame, electrically conductive phases such as water have been reintroduced to make cratonic lithosphere more conductive than the lithosphere we image here (cf. Kelbert et al., 2012; Meqbel et al., 2014; Yang et al., 2015). In contrast, CAMP formed at ~200 Ma, so there has been much less time for SEUS lithosphere to cool after tectonism and to become metasomatized. As cratonic regions have not been significantly deformed for ~2 Ga, grains have had a very long time to equilibrate to large sizes; this trend towards grain growth likely reduces the importance of grain boundary processes as a control on physical characteristics of the mantle lithosphere. In contrast, mineral grains in the sub-SEUS lithosphere have had an order of magnitude less time to re-equilibrate. Furthermore, the modern lithospheric strain rates in cratonic interiors are likely smaller than those on a passive continental margin, so the current equilibrium grain size in cratonic regions is likely larger than along the edges of the continent. In expanding this comparison to include tectonically active areas, which have deformational ages that are an order of magnitude younger than CAMP and which generally show low seismic velocities and high electrical conductivities, the uppermost mantle beneath the SEUS may represent a class of “middle-aged” lithosphere with geophysical properties distinct from the typical cratonic and orogenic end-members. We also stress that seismic reference models do not translate to a meaningful or realistic mantle physicochemical state (e.g., Cammarano et al., 2003; Cobden et al., 2008). Interpreting seismic velocity anomalies with respect to some value or reference model will obviously depend on what is taken as ‘normal.’ Commonly used reference models do not provide a physicochemically meaningful definition of ‘normal,’ so that interpreting ‘fast’ and ‘slow’ anomalies in terms of ‘cold’ and ‘hot’ regions effectively lacks a baseline with which to define absolute temperature. For example, an average Vs value of 4.5 km/s is commonly taken as a cutoff between lithospheric temperatures and asthenospheric temperatures in the upper mantle. (For an example from the SEUS, see Wagner et al., 2018; in that work, 4.5 km/s is implicitly assumed to be the cutoff between high velocities representative of stable lithosphere and low velocities representative of lithospheric modification and removal.) However, our calculations here (Fig. 3.7) demonstrate that 4.5 km/s can represent upper mantle material at lithospheric temperatures. If a Vs value of 4.4 km/s were used as a reference value for the lithospheric-asthenospheric temperature cutoff, then the SEUS would

73 in fact appear slightly ‘fast’ and ‘cold,’ and cratonic regions would appear even faster and perhaps more in line with their true temperatures. Indeed, experimental work demonstrates that shear velocities could even be less than 4.3 km/s at high lithospheric temperatures of 1250-1300°C (Takei, 2017). Although methodologically useful, reference models and reference values do not necessarily provide direct physicochemical insight or illuminate physical constraints when interpreting seismic results.

3.5.7 Implications for geophysical interpretation and geodynamic investigations It is well established that temperature exerts the dominant control on seismic velocity in the upper mantle (e.g., Cammarano et al., 2003), but there are nevertheless other important factors, including composition (e.g., Lee, 2003; Schutt & Lesher 2006, 2010; Afonso & Schutt, 2012) and grain size (e.g., Jackson & Faul, 2010), that cannot be ignored when formulating geodynamic interpretations of seismic images. In the tectonically active western US, lateral gradients in seismic velocities are often so high that significant variations in temperature must surely be invoked as a first order explanation. However, in a passive setting, such as the eastern US, this is not so. Temperature gradients are small enough that other effects must be considered. Although the importance of anelastic contributions to seismic observables has long been recognized (e.g., Karato, 1993), such effects are not always integrated into geophysical interpretation. Our results underscore the importance of doing so. Furthermore, our results highlight the danger of making geodynamic and geophysical interpretations from seismic observations alone when other deep-sensing observations, such as long-period MT measurements, are available. The interpretation of relatively thin, eroded thermal lithosphere beneath the SEUS may be consistent with seismic data. However, this interpretation is not consistent with the available MT data, which require a thick (~200 km), coherent thermal lithosphere. More careful consideration of absolute seismic velocities in the SEUS demonstrates that the seismic data are also consistent with this model of thick lithosphere. Of course, all geophysical techniques can provide useful constraints, but models and interpretations are never unique. Obviously, integrating constraints from multiple geophysical techniques (as well as other geodynamic information, such as surface topography) can reduce non-uniqueness and improve reliability of interpretations. Upper mantle seismic anomalies are all too often interpreted in terms of relative temperatures (red equals hot, blue equals cold) without sufficient consideration of reference models and other factors that can influence absolute seismic velocity. MT data in the SEUS require high

74 resistivities to great depth and, as we have argued, thus provide an absolute constraint on temperature. Combining this constraint with the observed seismic velocity anomalies has the potential to yield a more meaningful understanding of the current physical state and geologic evolution of the region. It is worth noting that, although some geodynamic studies have acknowledged the importance of anelasticity in assimilating seismic velocity observations into numerical models (e.g., Dannberg et al., 2017), the intrinsic ambiguity in mapping seismic velocity anomalies to temperature fields appears to be underappreciated and too often ignored in geodynamic modeling. Our results here demonstrate the need for care in mapping seismic results to temperature. Indeed, grain size is an important factor in material viscosity, so relatively small seismic anomalies may even be mapped to the wrong physical parameter (that is, temperature instead of viscosity) unless such effects are considered.

3.6 Conclusions By taking into account anelastic controls on seismic velocity and depth-dependent oxygen fugacity, we are able to synthesize seemingly contradictory MT and seismic results from the SEUS in order to arrive at a consistent picture of thick (~200 km), coherent thermal lithosphere beneath this region. Our results underscore the importance of accounting for anelasticity when interpreting small-amplitude seismic velocity anomalies in the mantle lithosphere; clearly demonstrate the value of combining disparate datasets in geophysical and geodynamic interpretations; and highlight the danger of interpreting seismic results purely in terms of reference models. An intriguing test of the ideas presented here will be to compare seismic and MT results from other areas affected by recent voluminous, mantle-derived volcanism, such as the Deccan Traps, other regions that host CAMP magmatism, and even oceanic Large Igneous Provinces such as the Ontong Java Plateau. Current MT results are insufficient to make a clear comparison between MT-derived and seismic-derived estimates of thermal lithosphere properties in those regions. Although much MT work has been completed in the areas of Brazil and India that were affected by these recent flood-basalt events (e.g., Patro & Egbert, 2011; Padilha et al., 2015; Danda et al., 2017; Maurya et al., 2018), all studies so far have focused on crustal and uppermost mantle structures (even though resolvable resistive structures in some cases do extend deeper, beyond the putative depth resolution limit), and there are few three-dimensional inversion results from these areas. Further work in these regions, as well as in other regions affected by CAMP magmatism (specifically the margins

75 of Europe, Africa, and northeastern North America) is clearly warranted to see if similar apparent discrepancies between seismic and MT observations are found. Indeed, it is possible that a careful reexamination of existing datasets would already reveal such discrepancies— given the relative unfamiliarity of the broader geophysics community with MT, and the tendency for MT interpretation to mimic existing interpretations from seismic observations, seemingly anomalous MT results have often been disregarded in the past. However, as we have shown here, MT (especially with recent advances in methodology and laboratory constraints on interpretation) can often provide reliable, and sometimes quite firm, constraints on composition or physical state of the interior. Clearly, if our goal is a meaningful and correct picture of the Earth, constraints from MT must be taken into account, along with the constraints provided by seismology, heat flow, topography, gravity, and magnetics. Only then may we begin to understand the mysterious inner workings of our planet.

3.7 References Afonso, J.C., & Schutt, D.L. (2012). The effects of polybaric partial melting on density and seismic velocities of mantle restites. Lithos, 134-135, 289-303. https://doi.org/10.1016/j.lithos.2012.01.009 Amante, C., & Eakins, B.W. (2009). ETOPO1 1 Arc-Minute Global Relief Model: Procedures, Data Sources and Analysis. NOAA Technical Memorandum NESDIS NGDC- 24, National Geophysical Data Center, NOAA. https://doi.org/10.7289/V5C8276M Armienti, P., & Tarquini, S. (2002). Power law olivine crystal size distributions in lithospheric mantle xenoliths. Lithos, 65(3-4), 273-285. https://doi.org/10.1016/S0024- 4937(02)00195-0 Armitage, J.J., Ferguson, D.J., Goes, S., Hammond, J.O.S., Calais, E., Rychert, C.A., & Harmon, N. (2015). Upper mantle temperature and the onset of extension and break-up in Afar, Africa. Earth and Planetary Science Letters, 418, 78-90. https://doi.org/10.1016/j.epsl.2015.02.039 Austin, N.J., & Evans, B. (2007). Paleowattmeters: A scaling relation for dynamically recrystallized grain size. Geology, 35(4), 343-346. https://doi.org/10.1130/G23244A.1 Ave Lallemant, H.G., Mercier, J-C.C., Carter, N.L., & Ross, J.V. (1980). Rheology of the upper mantle: Inferences from peridotite xenoliths. Tectonophysics, 70(1-2), 85-113. https://doi.org/10.1016/0040-1951(80)90022-0

Bao, X., Dalton, C.A., Jin, G., Gaherty, J.B., & Shen, Y. (2016). Imaging Rayleigh wave attenuation with USArray. Geophysical Journal International, 206, 241–259. https://doi.org/10.1093/gji/ggw151 Biryol, C.B., Wagner, L.S., Fischer, K.M., & Hawman, R.B. (2016). Relationship between observed upper mantle structures and recent tectonic activity across the Southeastern United States. Journal of Geophysical Research: Solid Earth, 121, 3393–3414. https://doi.org/10.1002/2015JB012698 Calais, E., Camelbeeck, T., Stein, S., Liu, M., & Craig, T.J. (2016). A new paradigm for large earthquakes in stable continental plate interiors. Geophysical Research Letters, 43, 10 621–10 637. https://doi.org/10.1002/2016GL070815 Cammarano, F., Goes, S., Vacher, P., & Giardini, D. (2003). Inferring upper-mantle temperatures from seismic velocities. Physics of the Earth and Planetary Interiors, 138, 197-222. https://doi.org/10.1016/S0031-9201(03)00156-0 Chu, R., Leng, W., Helmberger, D.V., & Gurnis, M. (2013). Hidden hotspot track beneath the eastern United States. Nature Geoscience, 6, 963–966. https://doi.org/10.1038/ngeo1949 Cline, C.J., Faul, U.H., David, E.C., Berry, A.J., & Jackson, I. (2018). Redox-influenced seismic properties of upper-mantle olivine. Nature, 555, 355-358. https://doi.org/10.1038/nature25764 Cobden, L., Goes, S., Cammarano, F., & Connolly, J.A.D. (2008). Thermochemical interpretation of one-dimensional seismic reference models for the upper mantle: Evidence for bias due to heterogeneity. Geophysical Journal International, 175(2), 627- 648. https://doi.org/10.1111/j.1365-246X.2008.03903.x Constable, S. (2006). SEO3: A new model of olivine electrical conductivity. Geophysical Journal International, 166, 435–437. https://doi.org/10.1111/j.1365-246X.2006.03041.x Danda, N., Rao, C.K., & Kumar, A. (2017). Geoelectric structure of northern Cambay rift basin from magnetotelluric data. Earth, Planets and Space, 69(140). https://doi.org/10.1186/s40623-017-0725-0 Dannberg, J., Eilon, Z., Faul, U., Gassmöller, R., Moulik, P., & Myhill, R. (2017). The importance of grain size to mantle dynamics and seismological observations. Geochemistry, Geophysics, Geosystems, 18, 3034–3061. https://doi.org/10.1002/2017GC006944 Du Frane, W.L., Roberts, J.J., Toffelmier, D.A., & Tyburczy, J.A. (2005). Anisotropy of electrical conductivity in dry olivine: Geophysical Research Letters, 32, L24315. https://doi.org/10.1029/2005GL023879

Ducea, M.N., & Park, S.K. (2000). Enhanced mantle conductivity from sulfide minerals, southern Sierra Nevada, California. Geophysical Research Letters, 27(16), 2405-2408. https://doi.org/10.1029/2000GL011565 Eaton, D.W., Darbyshire, F., Evans, R.L., Grütter, H., Jones, A.G., & Yuan, X. (2009). The elusive lithosphere–asthenosphere boundary (LAB) beneath cratons. Lithos, 109(1-2), 1- 22. https://doi.org/10.1016/j.lithos.2008.05.009 Eeken, T., Goes, S., Pedersen, H.A., Arndt, N.T., & Bouilhol, P. (2018). Seismic evidence for depth-dependent metasomatism in cratons. Earth and Planetary Science Letters, 491, 148-159. https://doi.org/10.1016/j.epsl.2018.03.018 Elkins-Tanton, L.T. (2007). Continental magmatism, volatile recycling, and a heterogeneous mantle caused by lithospheric gravitational instabilities. Journal of Geophysical Research: Solid Earth, 112, B03405. https://doi.org/10.1029/2005JB004072 Elliott, S. (1998). The Physics and Chemistry of Solids. Wiley, England. Evans, B., Renner, J., & Hirth, G. (2001). A few remarks on the kinetics of static grain growth in rocks. International Journal of Earth Sciences, 90, 88-103. https://doi.org/10.1007/s005310000150 Evans, R.L., Chave, A.D., & Booker, J.R. (2002). On the importance of offshore data for magnetotelluric studies of ocean-continent subduction systems. Geophysical Research Letters, 29(9), 1302. https://doi.org/10.1029/2001GL013960 Evans, R.L., Benoit, M.H., Long, M.D., Elsenbeck, J., & Ford, H.A. (2016). Evidence for Lithospheric Loss Beneath the Appalachians Along the MAGIC Line. AGU Fall Meeting, T42D-07. Forsyth, D.W. (1992). Geophysical Constraints on Mantle Flow and Melt Generation Beneath Mid‐Ocean Ridges. Geophysical Monograph Series, 71, 1-65. https://doi.org/10.1029/GM071p0001 Frizon de Lamotte, D., Fourdan, B., Leleu, S., Leparmentier, F., & de Clarens, P. (2015). Style of rifting and the stages of Pangea breakup. Tectonics, 34, 1009–1029. https://doi.org/10.1002/2014TC003760 Frost, D.J., & McCammon, C.A. (2008). The Redox State of Earth’s Mantle. Annual Review of Earth and Planetary Sciences, 36, 389–420. https://doi.org/10.1146/annurev.earth.36.031207.124322 Gardés, E., Gaillard, F., & Tarits, P. (2014). Toward a unified hydrous olivine electrical conductivity law. Geochemistry, Geophysics, Geosystems, 15, 4984-5000. https://doi.org/10.1002/2014GC005496

Goes, S., Simons, F.J., & Yoshizawa, K. (2005). Seismic constraints on temperature of the Australian uppermost mantle. Earth and Planetary Science Letters, 236(1-2), 227-237. https://doi.org/10.1016/j.epsl.2005.05.001 Goes, S., Armitage, J., Harmon, N., Smith, H., & Huismans, R. (2012). Low seismic velocities below mid-ocean ridges: Attenuation versus melt retention. Journal of Geophysical Research: Solid Earth, 117, B12403. https://doi.org/10.1029/2012JB009637 Gribenko, A.V., & Zhdanov, M.S. (2017). 3-D inversion of the MT EarthScope data, collected over the East Central United States. Geophysical Research Letters, 44, 11800- 11807. https://doi.org/10.1002/2017GL075000 Griffin, W.L., O’Reilly, S.Y., Afonso, J.C., Begg, G.C. (2009). The Composition and Evolution of Lithospheric Mantle: a Re-evaluation and its Tectonic Implications. Journal of Petrology, 50(7), 1185–1204. https://doi.org/10.1093/petrology/egn033 Hatcher Jr., R.D. (2010). The Appalachian orogen: a brief summary. In: Tollo, R.P., Bartholomew, M.J., Hibbard, J.P., Karabinos, P.M. (Eds.), From Rodinia to Pangea: The Lithotectonic Record of the Appalachian Region: Mem. Geol. Soc. Amer. 206, pp. 1–19. Hopper, E., & Fischer, K.M. (2018). The changing face of the lithosphere-asthenosphere boundary: Imaging continental scale patterns in upper mantle structure across the contiguous U.S. with Sp converted waves. Geochemistry, Geophysics, Geosystems, 19. https://doi.org/10.1029/2018GC007476 Jackson, I., & Faul, U.H. (2010). Grainsize-sensitive viscoelastic relaxation in olivine: Towards a robust laboratory-based model for seismological application. Physics of the Earth and Planetary Interiors, 183, 151–163. https://doi.org/10.1016/j.pepi.2010.09.005 Jones, A.G., Snyder, D., Hanmer, S., Asudeh, I., White, D., Eaton, D., & Clarke, G. (2002). Magnetotelluric and teleseismic study across the Snowbird Tectonic Zone, Canadian Shield: A Neoarchean mantle suture? Geophysical Research Letters, 29(17), 1829. https://doi.org/10.1029/2002GL015359 Jones, A.G., Lezaeta, P., Ferguson, I.J., Chave, A.D., Evans, R.L., Garcia, X., & Spratt, J. (2003). The electrical structure of the Slave craton. Lithos, 71(2-4), 505-527. https://doi.org/10.1016/j.lithos.2003.08.001 Jones, A.G., Fullea, J., Evans, R.L., & Muller, M.R. (2012). Water in cratonic lithosphere: calibrating laboratory-determined models of electrical conductivity of mantle minerals using geophysical and petrological observations. Geochemistry, Geophysics, Geosystems, 13, 27. https://doi.org/10.1029/2012GC004055

Jones, A.G., Fishwick, S., Evans, R.L., Muller, M.R., & Fullea, J. (2013). Velocity- conductivity relations for cratonic lithosphere and their application: Example of Southern Africa. Geochemistry, Geophysics, Geosystems, 14, 806-827. https://doi.org/10.1002/ggge.20075 Karato, S.-i. (1993). Importance of anelasticity in the interpretation of seismic tomography. Geophysical Research Letters, 20, 1623–1626. https://doi.org/10.1029/93GL01767 Karato, S.-i. (2010). Rheology of the deep upper mantle and its implications for the preservation of the continental roots: A review. Tectonophysics, 481, 82-98. https://doi.org/10.1016/j.tecto.2009.04.011 Kelbert, A., Egbert, G.D., & deGroot-Hedlin, C. (2012). Crust and upper mantle electrical conductivity beneath the Yellowstone Hotspot Track. Geology, 40, 447–450. https://doi.org/10.1130/G32655.1 Kelbert, A., Egbert, G.D., & Schultz, A. (2011). IRIS DMC Data Services Products: EMTF, The Magnetotelluric Transfer Functions. https://doi.org/10.17611/DP/EMTF.1 Kohn, S.C., & Grant, K.J. (2006). The Partitioning of Water Between Nominally Anhydrous Minerals and Silicate Melts. In J.J. Rosso, ed., Water in Nominally Anhydrous Minerals: Reviews in Mineralogy & Geochemistry 62, 231-241. https://doi.org/10.2138/rmg.2006.62.10 Kreemer, C., Hammond, W.C., & Blewitt, G. (2018). A robust estimation of the 3-D intraplate deformation of the North American plate from GPS. Journal of Geophysical Research: Solid Earth, 123, 4388–4412. https://doi.org/10.1029/2017JB015257 Laske, G., Masters, G., Ma, Z., & Pasyanos, M. (2013). Update on CRUST1.0—A 1-degree Global Model of Earth’s Crust. Geophysical Research Abstracts, 15, EGU2013-2658. Lee, C.-T. A. (2003). Compositional variation of density and seismic velocities in natural peridotites at STP conditions: Implications for seismic imaging of compositional heterogeneities in the upper mantle. Journal of Geophysical Research: Solid Earth, 108. https://doi.org/10.1029/2003JB002413 Lee, C.-T. A., & Chin, E.J. (2014). Calculating melting temperatures and pressures of peridotite protoliths: Implications for the origin of cratonic mantle. Earth and Planetary Science Letters, 403, 273-286. https://doi.org/10.1016/j.epsl.2014.06.048 Liu, L., & Gao, S.S. (2018). Lithospheric layering beneath the contiguous United States constrained by S-to-P receiver functions. Earth and Planetary Science Letters, 495, 79- 86. https://doi.org/10.1016/j.epsl.2018.05.012

Mareschal, M., Kellett, R.L., Kurtz, R.D., Ludden, J.N., Ji, S., & Bailey, R.C. (1995). Archaean cratonic roots, mantle shear zones and deep electrical anisotropy. Nature, 375, 134-137. https://doi.org/10.1038/375134a0 Maurya, V.P., Meju, M.A., Fontes, S.L., Padilha, A.L., La Terra, E.F., & Miquelutti, L.G. (2018). Deep resistivity structure of basalt-covered central part of Paraná Basin, Brazil, from joint 3-D MT and GDS data imaging. Geochemistry, Geophysics, Geosystems, 19, 1994-2013. https://doi.org/10.1029/2017GC007314 Mazza, S.E., Gazel, E., Johnson, E.A., Kunk, M.J., McAleer, R., Spotila, J.A., Bizimis, M., & Coleman, D.S. (2014). Volcanoes of the passive margin: the youngest magmatic event in eastern North America. Geology, 42, 483–486. https://doi.org/10.1130/G35407.1 Mazza, S.E., Gazel, E., Johnson, E.A., Bizimis, M., McAleer, R., & Biryol, C.B. (2017). Post-rift magmatic evolution of the eastern North American “passive-aggressive” margin. Geochemistry, Geophysics, Geosystems, 18, 3–22. https://doi.org/10.1002/2016GC006646 McHone, J.G. (2000). Non-plume magmatism and rifting during the opening of the central Atlantic Ocean. Tectonophysics, 316, 287-296. https://doi.org/10.1016/S0040- 1951(99)00260-7 Meqbel, N.M., Egbert, G.D., Wannamaker, P.E., Kelbert, A., & Schultz, A. (2014). Deep electrical resistivity structure of the northwestern U.S. derived from 3-D inversion of USArray magnetutelluric data. Earth and Planetary Science Letters, 402, 290–304. https://doi.org/10.1016/j.epsl.2013.12.026 Meyer, R., & van Wijk, J. (2015). Post-breakup lithosphere recycling below the U.S. East Coast: Evidence from adakitic rocks. In The Interdisciplinary Earth: A Volume in Honor of Don L. Anderson, eds. G.R. Foulger, M. Lustrino, S.D. King, Geological Society of America Special Papers, 514. https://doi.org/10.1130/2015.2514(06) Murphy, B.S., & Egbert, G.D. (2017). Electrical conductivity structure of southeastern North America: Implications for lithospheric architecture and Appalachian topographic rejuvenation. Earth and Planetary Science Letters, 462, 66–75. https://doi.org/10.1016/j.epsl.2017.01.009 Murphy, B.S., & Egbert, G.D. (2018). Source biases in midlatitude magnetotelluric transfer functions due to Pc3-4 geomagnetic pulsations. Earth, Planets and Space, 70(12). https://doi.org/10.1186/s40623-018-0781-0

Nelson, K.D. (1992). Are crustal thickness variations in old mountain belts like the Appalachians a consequence of lithospheric delamination? Geology, 20(6), 498-502. https://doi.org/10.1130/0091-7613(1992)020<0498:ACTVIO>2.3.CO;2 Olugboji, T.M., Karato, S.-i., & Park, J. (2013). Structures of the oceanic lithosphere- asthenosphere boundary: Mineral-physics modeling and seismological signatures. Geochemistry, Geophysics, Geosystems, 14, 880–901. https://doi.org/10.1002/ggge.20086 Olugboji, T.M., Park, J., Karato, S.-.i, & Shinohara, M. (2016). Nature of the seismic lithosphere-asthenosphere boundary within normal oceanic mantle from high-resolution receiver functions, Geochemistry, Geophysics, Geosystems, 17, 1265–1282. https://doi.org/10.1002/2015GC006214 Padilha, A.L., Vitorello, Í., Antunes, C.E., & Pádua, M.B. (2015). Imaging three‐dimensional crustal conductivity structures reflecting continental flood basalt effects hidden beneath thick intracratonic sedimentary basin. Journal of Geophysical Research: Solid Earth, 120(7), 4702-4719. https://doi.org/10.1002/2014JB011657 Patro, P.K., & Egbert, G.D. (2011). Application of 3D inversion to magnetotelluric profile data from the Deccan Volcanic Province of Western India. Physics of the Earth and Planetary Interiors, 187(1-2), 33-46. https://doi.org/10.1016/j.pepi.2011.04.005 Pollitz, F.F., & Mooney, W.D. (2016). Seismic velocity structure of the crust and shallow mantle of the Central and Eastern United States by seismic surface wave imaging. Geophysical Research Letters, 43, 118–126. https://doi.org/10.1002/2015GL066637 Pommier, A., & Garnero, E.J. (2014). Petrology-based modeling of mantle melt electrical conductivity and joint interpretation of electromagnetic and seismic results. Journal of Geophysical Research: Solid Earth, 119, 4001-4016. https://doi.org/10.1002/2013JB010449 Pommier, A., Kohlstedt, D.L., Hansen, L.N., Mackwell, S., Tasaka, M., Heidelbach, F., & Leinenweber, K. (2018). Transport properties of olivine grain boundaries from electrical conductivity experiments. Contributions to Mineralogy and Petrology, 173(41). https://doi.org/10.1007/s00410-018-1468-z Porter, R., Liu, Y., & Holt, W.E. (2016). Lithospheric records of orogeny within the continental U.S. Geophysical Research Letters, 43, 144–153. https://doi.org/10.1002/2015GL066950 Ragland, P.C., Hatcher, R.D., Jr., & Whittington, D. (1983). Juxtaposed Mesozoic diabase dike sets from the Carolinas: A preliminary assessment. Geology, 11, 394–399.

Ranganayaki, R.P., & Madden, T.R. (1980). Generalized thin sheet analysis in magnetotellurics: an extension of Price's analysis. Geophysical Journal of the Royal Astronomical Society, 60(3), 445-457. https://doi.org/10.1111/j.1365- 246X.1980.tb04820.x Sacks, P.E., & Secor, D.T., Jr. (1990). Delamination in collisional orogens. Geology, 18(10), 999-1002. https://doi.org/10.1130/0091-7613(1990)018<0999:DICO>2.3.CO;2 Samson, S.D., Coler, D.G., & Speer, J.A. (1995). Geochemical and Nd-Sr-Pb isotopic composition of Alleghanian granites of the southern Appalachians: Origin, tectonic setting, and source characterization. Earth and Planetary Science Letters, 134(3-4), 359- 376. https://doi.org/10.1016/0012-821X(95)00124-U Savage, B., Covellone, B.M., & Shen, Y. (2017). Wave speed structure of the eastern North American margin. Earth and Planetary Science Letters, 459, 394-405. https://doi.org/10.1016/j.epsl.2016.11.028 Schmandt, B., & Lin, F.-C. (2014). P and S wave tomography of the mantle beneath the United States. Geophysical Research Letters, 41, 6342–6349. https://doi.org/10.1002/2014GL061231 Schmandt, B., Lin, F.-C., & Karlstrom, K.E. (2015). Distinct crustal isostasy trends east and west of the Rocky Mountain Front. Geophysical Research Letters, 42, 10290-10298. https://doi.org/10.1002/2015GL066593 Schutt, D.L., & Lesher, C.E. (2006). Effects of melt depletion on the density and seismic velocity of garnet and spinel lherzolite. Journal of Geophysical Research: Solid Earth, 111. https://doi.org/10.1029/2003JB002950 Schutt, D.L., & Lesher, C.E. (2010). Compositional trends among Kaapvaal Craton garnet peridotite xenoliths and their effects on seismic velocity and density. Earth and Planetary Science Letters, 300, 367-373. https://doi.org/10.1016/j.epsl.2010.10.018 Selway, K. (2014). On the causes of electrical conductivity anomalies in tectonically stable lithosphere. Surveys in Geophysics, 35, 219–257. https://doi.org/10.1007/s10712-013- 9235-1 Selway, K., Ford, H., & Kelemen, P. (2015). The seismic mid-lithosphere discontinuity. Earth and Planetary Science Letters, 414, 45-57. https://doi.org/10.1016/j.epsl.2014.12.029 Shen, W., & Ritzwoller, M.H. (2016). Crustal and uppermost mantle structure beneath the United States. Journal of Geophysical Research: Solid Earth, 121, 4306–4342. https://doi.org/10.1002/2016JB012887

Slingerland, R., & Furlong, K.P. (1989). Geodynamic and geomorphic evolution of the Permo-Triassic Appalachian Mountains. Geomorphology, 2, 23-37. https://doi.org/10.1016/0169-555X(89)90004-4 Sun, J., Kelbert, A., & Egbert, G.D. (2015). Ionospheric current source modeling and global geomagnetic induction using ground geomagnetic observatory data. Journal of Geophysical Research: Solid Earth, 120(10), 6771–6796. https://doi.org/10.1002/2015JB012063 Takei, Y. (2017). Effects of Partial Melting on Seismic Velocity and Attenuation: A New Insight from Experiments. Annual Review of Earth and Planetary Sciences, 45, 447-470. https://doi.org/10.1146/annurev-earth-063016-015820 Wagner, L.S., Fischer, K.M., Hawman, R., Hopper, E., & Howell, D. (2018). The relative roles of inheritance and long-term passive margin lithospheric evolution on the modern structure and tectonic activity in the southeastern United States. Geosphere, 14, 1–26. https://doi.org/10.1130/GES01593.1 Wang, H., van Hunen, J., Pearson, D.G., & Allen, M.B. (2014). Craton stability and longevity: The roles of composition-dependent rheology and buoyancy. Earth and Planetary Science Letters, 391, 224-233. https://doi.org/10.1016/j.epsl.2014.01.038 Wannamaker, P.E. (2005). Anisotropy Versus Heterogeneity in Continental Solid Earth Electromagnetic Studies: Fundamental Response Characteristics and Implications for Physicochemical State. Surveys in Geophysics, 26(6), 733-765. https://doi.org/10.1007/s10712-005-1832-1 Whalen, L., Gazel, E., Vidito, C., Puffer, J., Bizimis, M., Henika, W., & Caddick, M.J. (2015). Supercontinental inheritance and its influence on supercontinental breakup: the Central Atlantic Magmatic Province and the breakup of Pangea. Geochemistry, Geophysics, Geosystems, 16, 3532–3554. https://doi.org/10.1002/2015GC005885 Yang, B., Egbert, G.D., Kelbert, A., & Meqbel, N.M. (2015). Three-dimensional electrical resistivity of the north-central USA from EarthScope long period magnetotelluric data. Earth and Planetary Science Letters, 422, 87–93. https://doi.org/10.1016/j.epsl.2015.04.006 Yang, Y., & Forsyth, D.W. (2008). Attenuation in the upper mantle beneath Southern California: Physical state of the lithosphere and asthenosphere. Journal of Geophysical Research, 113, B03308. https://doi.org/10.1029/2007JB005118

Conclusions We have shown that MT data from the southeastern United States (SEUS) require high resistivity values (>300 Ωm) to depths of at least 200 km. The inescapable interpretation of these high resistivity values is that thick (~200 km) thermal lithosphere underlies the Piedmont and Coastal Plain physiographic provinces of the SEUS. Although previous seismic imaging results have been used to argue for relatively thin thermal lithosphere that has experienced erosion and piecemeal delamination along the SEUS passive margin, we have demonstrated that, taken together, both seismic and MT results are consistent with uniformly thick (~200 km), coherent thermal lithosphere beneath this region. We interpret this lithospheric block to be associated with the formation of the Central Atlantic Magmatic Province (CAMP) at 200 Ma. Our analysis indicates that the lithosphere in this region must have survived intact since its formation at that time. Our results pertaining to this coherent lithospheric block hold several important and intriguing geodynamic implications. First, our analysis indicates that the SEUS margin has not experienced lithospheric modification (such as delamination) through the Cenozoic, in contrast to previous interpretations from seismic results alone. Rather, our results indicate that the SEUS lithosphere has been stable through this time. Such active mantle processes therefore have not contributed and are not currently contributing to modern geologic processes observed at the surface, such as topographic rejuvenation and transience. (Note, however, our results do not bear on the role of asthenospheric flow in affecting surface characteristics via dynamic topography.) Second, the peculiar nature of this lithospheric block suggests that it may represent an example of Mesozoic craton formation. Our MT results strongly suggest that the constituent lithosphere is completely dry, so this lithospheric block is likely rheologically very strong. Furthermore, if this lithospheric block is the result of CAMP formation, it is also likely compositionally depleted (in the bulk petrological sense) and therefore buoyant with respect to the surrounding undepleted mantle. Such a strong, buoyant block will be difficult to destroy, even over long geologic timescales. Overall, this work has highlighted the importance of integrating constraints from different geophysical techniques in order to gain insights into lithospheric dynamics. Our results also demonstrate the value of MT in lithospheric imaging studies. Without our MT constraints, the geodynamic conclusions described above would be lost in the intrinsic uncertainty of the seismic images and interpretations. By combining these valuable MT constraints with seismic results, we have arrived at valuable insights about the solid Earth.

Bibliography

Abbott RN, Greenwood JP (2001) Retrograde metamorphism of eclogite in the southern Appalachian Mountains, U.S.A.—A case involving seamount subduction? Journal of Metamorphic Geology 19:433-443. https://doi.org/10.1046/j.0263-4929.2001.00321.x Ádám A, Verõ J, Szendrõi J (2005) Solar eclipse effect on geomagnetic induction parameters. Annales Geophysicae 23:3487-3494. https://doi.org/10.5194/angeo-23-3487-2005 Afonso JC, Schutt DL (2012) The effects of polybaric partial melting on density and seismic velocities of mantle restites. Lithos 134-135:289-303. https://doi.org/10.1016/j.lithos.2012.01.009 Amante C, Eakins BW (2009) ETOPO1 1 Arc-Minute Global Relief Model: Procedures, Data Sources and Analysis. NOAA Technical Memorandum NESDIS NGDC-24, National Geophysical Data Center, NOAA. https://doi.org/10.7289/V5C8276M

Ardia P, Hirschmann MM, Withers AC, Tenner TJ (2012) H2O storage capacity of olivine at 5-8 GPa and consequences for dehydration partial melting of the upper mantle. Earth and Planetary Science Letters 345-348:104-116. https://doi.org/10.1016/j.epsl.2012.05.038 Armienti P, Tarquini S (2002) Power law olivine crystal size distributions in lithospheric mantle xenoliths. Lithos 65(3-4):273-285. https://doi.org/10.1016/S0024-4937(02)00195- 0 Armitage JJ, Ferguson DJ, Goes S, Hammond JOS, Calais E, Rychert CA, Harmon N (2015) Upper mantle temperature and the onset of extension and break-up in Afar, Africa. Earth and Planetary Science Letters 418:78-90. https://doi.org/10.1016/j.epsl.2015.02.039 Austin NJ, Evans, B (2007) Paleowattmeters: A scaling relation for dynamically recrystallized grain size. Geology 35(4):343-346. https://doi.org/10.1130/G23244A.1 Ave Lallemant HG, Mercier J-CC, Carter NL, Ross JV (1980) Rheology of the upper mantle: Inferences from peridotite xenoliths. Tectonophysics 70(1-2):85-113. https://doi.org/10.1016/0040-1951(80)90022-0 Bao X, Dalton CA, Jin G, Gaherty JB, Shen Y (2016) Imaging Rayleigh wave attenuation with USArray. Geophysical Journal International 206:241–259. https://doi.org/10.1093/gji/ggw151 Biryol CB, Wagner LS, Fischer KM, Hawman RB (2016) Relationship between observed upper mantle structures and recent tectonic activity across the Southeastern United States. Journal of Geophysical Research: Solid Earth 121:3393–3414. https://doi.org/10.1002/2015JB012698

Boerner DE, Kurtz RD, Craven JA (1996) Electrical conductivity and Paleo-Proterozoic foredeeps. Journal of Geophysical Research 101:13775-13791. https://doi.org/10.1029/96JB00171 Boote SK, Knapp JH (2016) Offshore extent of Gondwanan Paleozoic strata in the southeastern United States: The Suwannee suture zone revisited. Gondwana Research 40:199-210. https://doi.org/10.1016/j.gr.2016.08.011 Calais E, Camelbeeck T, Stein S, Liu M, Craig TJ (2016) A new paradigm for large earthquakes in stable continental plate interiors. Geophysical Research Letters 43:10621– 10637. https://doi.org/10.1002/2016GL070815 Cammarano F, Goes S, Vacher P, Giardini D (2003) Inferring upper-mantle temperatures from seismic velocities. Physics of the Earth and Planetary Interiors 138:197-222. https://doi.org/10.1016/S0031-9201(03)00156-0 Chave AD, Jones AG (2012) Introduction to the magnetotelluric method. In Chave AD, Jones AG (eds) The Magnetotelluric Method: Theory and Practice. Cambridge University Press, 1-18. https://doi.org/10.1017/CBO9781139020138.002 Chi PJ, Engebretson MJ, Moldwin MB, Russell CT, Mann IR, Hairston MR, Reno M, Goldstein J, Winkler LI, Cruz-Abeyro JL, Lee D-H, Yumoto K, Dalrymple R, Chen B, Gibson JP (2013) Sounding of the plasmasphere by Mid-continent MAgnetoseismic Chain (McMAC) magnetometers. Journal of Geophysical Research: Space Physics 118:3077-3086. https://doi.org/10.1002/jgra.50274 Chi PJ, Russell CT (1998) An interpretation of the cross-phase spectrum of geomagnetic pulsations by the field line resonance theory. Geophysical Research Letters 25(24):4445- 4448. https://doi.org/10.1029/1998GL900211 Chu R, Leng W, Helmberger DV, Gurnis M (2013) Hidden hotspot track beneath the eastern United States. Nature Geoscience 6:963–966. https://doi.org/10.1038/ngeo1949 Clausen LBN, Yeoman TK, Fear RC, Behlke R, Lucek EA, Engebretson MJ (2009) First simultaneous measurements of waves generated at the bow shock in the solar wind, the magnetosphere and on the ground. Annales Geophysicae 27:357-371. https://doi.org/10.5194/angeo-27-357-2009 Cline CJ, Faul UH, David EC, Berry AJ, Jackson I (2018) Redox-influenced seismic properties of upper-mantle olivine. Nature 555:355-358. https://doi.org/10.1038/nature25764 Cobden L, Goes S, Cammarano F, Connolly JAD (2008) Thermochemical interpretation of one-dimensional seismic reference models for the upper mantle: Evidence for bias due to

heterogeneity. Geophysical Journal International 175(2):627-648. https://doi.org/10.1111/j.1365-246X.2008.03903.x Constable S (2006) SEO3: A new model of olivine electrical conductivity. Geophysical Journal International 166:435–437. https://doi.org/10.1111/j.1365-246X.2006.03041.x Cook FA, Vasudevan K (2006) Reprocessing and enhanced interpretation of the initial COCORP Southern Appalachians traverse. Tectonophysics 420:161-174. https://doi.org/10.1016/j.tecto.2006.01.022 Cox RT, Van Arsdale RB (2002) The Mississippi Embayment, North America: a first order continental structure generated by the Cretaceous superplume mantle event. Journal of Geodynamics 34:163-176. https://doi.org/10.1016/S0264-3707(02)00019-4 Dai L, Karato S (2014) High and highly anisotropic electrical conductivity of the asthenosphere due to hydrogen diffusion in olivine. Earth and Planetary Science Letters 408:79-86. https://doi.org/10.1016/j.epsl.2014.10.003 Danda N, Rao CK, Kumar A (2017) Geoelectric structure of northern Cambay rift basin from magnetotelluric data. Earth, Planets and Space 69(140). https://doi.org/10.1186/s40623- 017-0725-0 Dannberg J, Eilon Z, Faul U, Gassmöller R, Moulik P, Myhill R (2017) The importance of grain size to mantle dynamics and seismological observations. Geochemistry, Geophysics, Geosystems 18:3034–3061. https://doi.org/10.1002/2017GC006944 Dmitriev VI, Berdichevsky MN (1979) The Fundamental Model of Magnetotelluric Sounding. Proceedings of the IEEE 67(7):1034-1044. https://doi.org/10.1109/PROC.1979.11386 Du Frane WL, Roberts JJ, Toffelmier DA, Tyburczy JA (2005) Anisotropy of electrical conductivity in dry olivine. Geophysical Research Letters 32:L24315. https://doi.org/10.1029/2005GL023879 Ducea MN, Park SK (2000) Enhanced mantle conductivity from sulfide minerals, southern Sierra Nevada, California. Geophysical Research Letters 27(16):2405-2408. https://doi.org/10.1029/2000GL011565 Eaton DW, Darbyshire F, Evans RL, Grütter H, Jones AG, Yuan X (2009) The elusive lithosphere–asthenosphere boundary (LAB) beneath cratons. Lithos 109(1-2):1-22. https://doi.org/10.1016/j.lithos.2008.05.009 Eeken T, Goes S, Pedersen HA, Arndt NT, Bouilhol P (2018) Seismic evidence for depth- dependent metasomatism in cratons. Earth and Planetary Science Letters 491:148-159. https://doi.org/10.1016/j.epsl.2018.03.018

Egbert GD (1997) Robust multiple-station magnetotelluric data processing. Geophysical Journal International 130:475-496. https://doi.org/10.1111/j.1365-246X.1997.tb05663.x Egbert GD (2002) Processing and interpretation of electromagnetic induction array data. Surveys in Geophysics 23(2-3):207-249. https://doi.org/10.1023/A:1015012821040 Egbert GD, Booker JR (1986) Robust estimation of geomagnetic transfer functions. Geophysical Journal International 87:173-194. https://doi.org/10.1111/j.1365- 246X.1986.tb04552.x Egbert GD, Booker JR (1989) Multivariate Analysis of Geomagnetic Array Data: 1. The Response Space. Journal of Geophysical Research 94(B10):14227-14247. https://doi.org/10.1029/JB094iB10p14227 Egbert GD, Eisel M, Boyd OS, Morrison HF (2000) DC trains and Pc3s: Source effects in mid-latitude geomagnetic transfer functions. Geophysical Research Letters 27(1):25-28. https://doi.org/10.1029/1999GL008369 Elkins-Tanton LT (2005) Continental magmatism caused by lithospheric delamination. In Foulger GR, Natland JH, Presnall DC, Anderson DL (eds) Plates, plumes, and paradigms. Geological Society of America Special Paper 388, 449-461. https://doi.org/10.1130/0- 8137-2388-4.449 Elkins-Tanton LT (2007) Continental magmatism, volatile recycling, and a heterogeneous mantle caused by lithospheric gravitational instabilities. Journal of Geophysical Research: Solid Earth 112:B03405. https://doi.org/10.1029/2005JB004072 Elliott S (1998) The Physics and Chemistry of Solids. Wiley, England. Erdman ME, Lee C-TA, Levander A, Jiang H (2016) Role of arc magmatism and lower crustal foundering in controlling elevation history of the Nevadaplano and Colorado Plateau: A case study of pyroxenitic lower crust from central Arizona, USA. Earth and Planetary Science Letters 439:48-57. https://doi.org/10.1016/j.epsl.2016.01.032. Evans B, Renner J, Hirth G (2001) A few remarks on the kinetics of static grain growth in rocks. International Journal of Earth Sciences 90:88-103. https://doi.org/10.1007/s005310000150 Evans RL, Chave AD, Booker JR (2002) On the importance of offshore data for magnetotelluric studies of ocean-continent subduction systems. Geophysical Research Letters 29(9):1302. https://doi.org/10.1029/2001GL013960 Evans RL, Benoit MH, Long MD, Elsenbeck J, Ford HA (2016) Evidence for Lithospheric Loss Beneath the Appalachians Along the MAGIC Line. AGU Fall Meeting T42D-07.

Fisher CM, Loewy SL, Miller CF, Berquist P, Van Schmus WR, Hatcher RD Jr, Wooden JL, Fullagar PD (2010) Whole-rock Pb and Sm-Nd isotopic constraints on the growth of southeastern Laurentia during Grenvillian orogenesis. Geological Society of America Bulletin 122:1646-1659. http://dx.doi.org/10.1130/B30116.1 Forsyth DW (1992) Geophysical Constraints on Mantle Flow and Melt Generation Beneath Mid‐Ocean Ridges. Geophysical Monograph Series 71:1-65. https://doi.org/10.1029/GM071p0001 Frizon de Lamotte D, Fourdan B, Leleu S, Leparmentier F, de Clarens P (2015) Style of rifting and the stages of Pangea breakup. Tectonics 34:1009–1029. https://doi.org/10.1002/2014TC003760 Frost DJ, McCammon CA (2008) The Redox State of Earth’s Mantle. Annual Review of Earth and Planetary Sciences 36:389–420. https://doi.org/10.1146/annurev.earth.36.031207.124322 Fullea J, Muller MR, Jones AG (2011) Electrical conductivity of continental lithospheric mantle from integrated geophysical and petrological modeling: Application to the Kaapvaal Craton and Rehoboth Terrane, southern Africa. Journal of Geophysical Research 116:B10202. https://doi.org/10.1029/2011JB008544 Gallen SF, Wegmann KW, Bohnenstiehl DR (2013) Miocene rejuvenation of topographic relief in the southern Appalachians. GSA Today 23:4-10. https://doi.org/10.1130/GSATG163A.1 Gardés E, Gaillard F, Tarits P (2014) Toward a unified hydrous olivine electrical conductivity law. Geochemistry, Geophysics, Geosystems 15:4984-5000. https://doi.org/10.1002/2014GC005496 Goes S, Simons FJ, Yoshizawa K (2005) Seismic constraints on temperature of the Australian uppermost mantle. Earth and Planetary Science Letters 236(1-2):227-237. https://doi.org/10.1016/j.epsl.2005.05.001 Goes S, Armitage J, Harmon N, Smith H, Huismans R (2012) Low seismic velocities below mid-ocean ridges: Attenuation versus melt retention. Journal of Geophysical Research: Solid Earth 117:B12403. https://doi.org/10.1029/2012JB009637 González-Castillo L, Galindo-Zaldívar J, Junge A, Martínez-Moreno FJ, Löwer A, Sanz de Galdeano C, Pedrera A, López-Garrido AC, Ruiz-Constán A, Ruano P, Martínez-Martos M, (2015) Evidence of a large deep conductive body within the basement of Guadalquivir foreland Basin (Betic Cordillera, S-Spain) from tipper vector modelling: Tectonic implications. Tectonophysics 663:354-363. https://doi.org/10.1016/j.tecto.2015.08.013

Gribenko AV, Zhdanov MS (2017) 3-D inversion of the MT EarthScope data, collected over the East Central United States. Geophysical Research Letters 44:11800-11807. https://doi.org/10.1002/2017GL075000 Griffin WL, O’Reilly SY, Afonso JC, Begg GC (2009) The Composition and Evolution of Lithospheric Mantle: A Re-evaluation and its Tectonic Implications. Journal of Petrology 50(7):1185–1204. https://doi.org/10.1093/petrology/egn033 Hatcher RD Jr (2010) The Appalachian orogen: A brief summary. In Tollo RP, Bartholomew MJ, Hibbard JP, Karabinos PM (eds) From Rodinia to Pangea: The Lithotectonic Record of the Appalachian Region. Geological Society of America Memoir 206, 1-19. https://doi.org/10.1130/2010.1206(01) Hawman RB, Khalifa MO, Baker MS (2012) Isostatic compensation for a portion of the Southern Appalachians: Evidence from a reconnaissance study using wide-angle, three- component seismic soundings. Geological Society of America Bulletin 124:291-317. https://doi.org/10.1130/B30464.1 Hibbard JP, van Staal CR, Rankin CW (2010) Comparative analysis of the geological evolution of the northern and southern Appalachian orogen: Late Ordovician-Permian. In Tollo RP, Bartholomew MJ, Hibbard JP, Karabinos PM (eds) From Rodinia to Pangea: The Lithotectonic Record of the Appalachian Region. Geological Society of America Memoir 206, 51-69. https://doi.org/10.1130/2010.1206(03) Hopper E, Fischer KM, Wagner LS, Hawman RB (2017) Reconstructing the end of the Appalachian orogen. Geology 45:15-18. https://doi.org/10.1130/G38453.1 Hopper E, Fischer KM (2018) The changing face of the lithosphere-asthenosphere boundary: Imaging continental scale patterns in upper mantle structure across the contiguous U.S. with Sp converted waves. Geochemistry, Geophysics, Geosystems 19(8):2593-2614. https://doi.org/10.1029/2018GC007476 Hughes, JW (1994) Magnetospheric ULF Waves: A Tutorial with a Historical Perspective. In Engebretson MJ, Takahashi K, Scholer M (eds) Solar Wind Sources of Magnetospheric Ultra-Low-Frequency Waves. American Geophysical Union, Washington, pp. 1-11. https://doi.org/10.1029/GM081p0001 Hughes WJ, Southwood DJ (1976) The Screening of Micropulsation Signals by the Atmosphere and Ionosphere. Journal of Geophysical Research 81(19):3234-3240. https://doi.org/10.1029/JA081i019p03234

Jackson I, Faul UH (2010) Grainsize-sensitive viscoelastic relaxation in olivine: Towards a robust laboratory-based model for seismological application. Physics of the Earth and Planetary Interiors 183:151–163. https://doi.org/10.1016/j.pepi.2010.09.005 Jones AG, Fishwick S, Evans RL, Muller MR, Fullea J (2013) Velocity-conductivity relations for cratonic lithosphere and their application: Example of Southern Africa. Geochemistry, Geophysics, Geosystems 14:806-827. https://doi.org/10.1002/ggge.20075 Jones AG, Fullea J, Evans RL, Muller MR (2012) Water in cratonic lithosphere: calibrating laboratory-determined models of electrical conductivity of mantle minerals using geophysical and petrological observations. Geochemistry, Geophysics, Geosystems 13:27. https://doi.org/10.1029/2012GC004055 Jones AG, Ledo J, Ferguson IJ (2005) Electromagnetic images of the Trans-Hudson orogen: The North American Central Plains anomaly revealed. Canadian Journal of Earth Sciences 32:457-478. https://doi.org/10.1139/e05-018 Jones AG, Lezaeta P, Ferguson IJ, Chave AD, Evans RL, Garcia X, Spratt J (2003) The electrical structure of the Slave craton. Lithos 71(2-4):505-527. https://doi.org/10.1016/j.lithos.2003.08.001 Jones AG, Snyder D, Hanmer S, Asudeh I, White D, Eaton D, Clarke G (2002) Magnetotelluric and teleseismic study across the Snowbird Tectonic Zone, Canadian Shield: A Neoarchean mantle suture? Geophysical Research Letters 29(17):1829. https://doi.org/10.1029/2002GL015359 Jones AG, Spratt J (2002) A simple method for deriving the uniform field MT responses in auroral zones. Earth, Planets and Space 54:443-450. https://doi.org/10.1186/BF03353035 Kaislaniemi L, van Hunen J (2014) Dynamics of lithospheric thinning and mantle melting by edge-driven convection: Application to Moroccan Atlas Mountains. Geochemistry, Geophysics, Geosystems 15:3175-3189. https://doi.org/10.1002/2014GC005414 Karato S-i (1993) Importance of anelasticity in the interpretation of seismic tomography. Geophysical Research Letters 20:1623–1626. https://doi.org/10.1029/93GL01767 Karato S-i (2010) Rheology of the deep upper mantle and its implications for the preservation of the continental roots: A review. Tectonophysics 481:82-98. https://doi.org/10.1016/j.tecto.2009.04.011 Kawano H, Yumoto K, Pilipenko VA, Tanaka Y-M, Takasaki S, Iizima M, Seto M (2002) Using two ground stations to identify magnetospheric field line eigenfrequency as a

continuous function of ground latitude. Journal of Geophysical Research 107(A8). https://doi.org/10.1029/2001JA000274 Kelbert A, Egbert GD, Schultz A (2011) IRIS DMC Data Services Products: EMTF, The Magnetotelluric Transfer Functions. https://doi.org/10.17611/DP/EMTF.1 Kelbert A, Egbert GD, deGroot-Hedlin C (2012) Crust and upper mantle electrical conductivity beneath the Yellowstone Hotspot Track. Geology 40:447–450. https://doi.org/10.1130/G32655.1 Kelbert A, Meqbel N, Egbert GD, Tandon K (2014) ModEM: A modular system for inversion of electromagnetic geophysical data. Computers & Geosciences 66:40-53. https://doi.org/10.1016/j.cageo.2014.01.010 Keppler H, Bolfan-Casanova N (2006) Thermodynamics of Water Solubility and Partitioning. In Keppler H, Smyth JR (eds) Water in Nominally Anhydrous Minerals. Reviews in Mineralogy and Geochemistry 62, 193-230. https://doi.org/10.2138/rmg.2006.62.9. Kohn SC, Grant KJ (2006) The Partitioning of Water Between Nominally Anhydrous Minerals and Silicate Melts. In J.J. Rosso (ed) Water in Nominally Anhydrous Minerals. Reviews in Mineralogy & Geochemistry 62, 231-241. https://doi.org/10.2138/rmg.2006.62.10 Kreemer C, Hammond WC, & Blewitt G (2018) A robust estimation of the 3-D intraplate deformation of the North American plate from GPS. Journal of Geophysical Research: Solid Earth 123:4388–4412. https://doi.org/10.1029/2017JB015257 Larsen JC, Mackie RL, Manzella A, Fiordelisi A, Rieven S (1996) Robust smooth magnetotelluric transfer functions. Geophysical Journal International 124:801-819. https://doi.org/10.1111/j.1365-246X.1996.tb05639.x Laske G, Masters G, Ma Z, Pasyanos M (2013) Update on CRUST1.0—A 1-degree Global Model of Earth’s Crust. Geophysical Research Abstracts 15:EGU2013-2658. Lee C-TA (2003) Compositional variation of density and seismic velocities in natural peridotites at STP conditions: Implications for seismic imaging of compositional heterogeneities in the upper mantle. Journal of Geophysical Research: Solid Earth 108. https://doi.org/10.1029/2003JB002413 Lee C-TA, Chin EJ (2014) Calculating melting temperatures and pressures of peridotite protoliths: Implications for the origin of cratonic mantle. Earth and Planetary Science Letters 403:273-286. https://doi.org/10.1016/j.epsl.2014.06.048

Liu L, Gao SS (2018) Lithospheric layering beneath the contiguous United States constrained by S-to-P receiver functions. Earth and Planetary Science Letters 495:79-86. https://doi.org/10.1016/j.epsl.2018.05.012 Liu L (2014) Rejuvenation of Appalachian topography caused by subsidence-induced differential erosion. Nature Geoscience 7:518-523. https://doi.org/10.1038/ngeo2187 Long MD, Jackson KG, McNamara JF (2016) SKS splitting beneath Transportable Array stations in eastern North America and the signature of past lithospheric deformation. Geochemistry, Geophysics, Geosystems 17:2-15. https://doi.org/10.1002/2015GC006088 MacDougall JG, Fischer KM, Forsyth DW, Hawman RB, Wagner LS (2015) Shallow mantle velocities beneath the southern Appalachians from Pn phases. Geophysical Research Letters 42:339-345. https://doi.org/10.1002/2014GL062714 Mareschal M, Kellett RL, Kurtz RD, Ludden JN, Ji S, Bailey RC (1995) Archaean cratonic roots, mantle shear zones and deep electrical anisotropy. Nature 375:134-137. https://doi.org/10.1038/375134a0 Maurya VP, Meju MA, Fontes SL, Padilha AL, La Terra EF, Miquelutti LG (2018) Deep resistivity structure of basalt-covered central part of Paraná Basin, Brazil, from joint 3-D MT and GDS data imaging. Geochemistry, Geophysics, Geosystems 19:1994-2013. https://doi.org/10.1029/2017GC007314 Mazza SE, Gazel E, Johnson EA, Kunk MJ, McAleer R, Spotila JA, Bizimis M, Coleman DS (2014) Volcanoes of the passive margin: the youngest magmatic event in eastern North America. Geology 42:483–486. https://doi.org/10.1130/G35407.1 Mazza SE, Gazel E, Johnson EA, Bizimis M, McAleer R, Biryol CB (2017) Post-rift magmatic evolution of the eastern North American “passive-aggressive” margin. Geochemistry, Geophysics, Geosystems 18:3–22. https://doi.org/10.1002/2016GC006646 McHone JG (2000) Non-plume magmatism and rifting during the opening of the central Atlantic Ocean. Tectonophysics 316:287-296. https://doi.org/10.1016/S0040- 1951(99)00260-7 McKeon RE, Zeitler PK, Pazzaglia FJ, Idleman BD, Enkelmann E (2014) Decay of an old orogen: Inferences about Appalachian landscape evolution from low-temperature thermochronology. Geological Society of America Bulletin 126:31-46. https://doi.org/10.1130/B30808.1 McPherron RL (2005) Magnetic Pulsations: Their Sources and Relation to Solar Wind and Geomagnetic Activity. Surveys in Geophysics 26:545-592. https://doi.org/10.1007/s10712-005-1758-7

Menk FW, Mann IR, Smith AJ, Waters CL, Clilverd MA, Milling DK (2004) Monitoring the plasmapause using geomagnetic field line resonances. Journal of Geophysical Research 109:A04216. https://doi.org/10.1029/2003JA010097 Menk FW (2011) Magnetospheric ULF Waves: A Review. In Liu W, Fujimoto M (eds) The Dynamic Magnetosphere. Springer, 223-256. https://doi.org/10.1007/978-94-007-0501- 2_13 Meqbel NM, Egbert GD, Wannamaker PE, Kelbert A, Schultz A (2014) Deep electrical resistivity structure of the northwestern U.S. derived from 3-D inversion of USArray magnetotelluric data. Earth and Planetary Science Letters 402:290-304. https://doi.org/10.1016/j.epsl.2013.12.026 Meyer R, van Wijk J (2015) Post-breakup lithosphere recycling below the U.S. East Coast: Evidence from adakitic rocks. In Foulger GR, Lustrino M, King SD (eds) The Interdisciplinary Earth: A Volume in Honor of Don L. Anderson, Geological Society of America Special Papers, 514. https://doi.org/10.1130/2015.2514(06) Miller SR, Sak PB, Kirby E, Bierman PR (2013) Neogene rejuvenation of central Appalachian topography: Evidence for differential rock uplift from stream profiles and erosion rates. Earth and Planetary Science Letters 369-370:1-12. https://doi.org/10.1016/j.epsl.2013.04.007 Moucha R, Forte AM, Mitrovica JX, Rowley DB, Quéré S, Simmons NA, Grand SP (2008) Dynamic topography and long-term sea-level variations: There is no such thing as a stable continental platform. Earth and Planetary Science Letters 271:101-108. https://doi.org/10.1016/j.epsl.2008.03.056 Murphy BS, Egbert GD (2017) Electrical conductivity structure of southeastern North America: Implications for lithospheric architecture and Appalachian topographic rejuvenation. Earth and Planetary Science Letters 462:66–75. https://doi.org/10.1016/j.epsl.2017.01.009 Murphy BS, Egbert GD (2018) Source biases in midlatitude magnetotelluric transfer functions due to Pc3-4 geomagnetic pulsations. Earth, Planets and Space 70:12. https://doi.org/10.1186/s40623-018-0781-0 Murphy BS, Egbert GD (2019) Synthesizing Seemingly Contradictory Seismic and Magnetotelluric Observations in the Southeastern United States to Image Physical Properties of the Lithosphere. Geochemistry, Geophysics, Geosystems 20(6): 2606-2625. https://doi.org/10.1029/2019GC008279

Nelson KD (1992) Are crustal thickness variations in old mountain belts like the Appalachians a consequence of lithospheric delamination? Geology 20(6):498-502. https://doi.org/10.1130/0091-7613(1992)020<0498:ACTVIO>2.3.CO;2 Ogawa Y, Jones AG, Unsworth MJ, Booker JR, Lu X, Craven J, Roberts B, Parmelee J, Farquharson C (1996) Deep electrical conductivity structures of the Appalachian Orogen in the southeastern U.S. Geophysical Research Letters 23:1597-1600. https://doi.org/10.1029/95GL03601 Olugboji TM, Karato S-i, Park J (2013) Structures of the oceanic lithosphere-asthenosphere boundary: Mineral-physics modeling and seismological signatures. Geochemistry, Geophysics, Geosystems 14:880–901. https://doi.org/10.1002/ggge.20086 Olugboji TM, Park J, Karato S-i, Shinohara M (2016) Nature of the seismic lithosphere- asthenosphere boundary within normal oceanic mantle from high-resolution receiver functions, Geochemistry, Geophysics, Geosystems 17:1265–1282. https://doi.org/10.1002/2015GC006214 Padilha AL, Vitorello Í, Antunes CE, Pádua MB (2015) Imaging three‐dimensional crustal conductivity structures reflecting continental flood basalt effects hidden beneath thick intracratonic sedimentary basin. Journal of Geophysical Research: Solid Earth 120(7):4702-4719. https://doi.org/10.1002/2014JB011657 Patro PK, Egbert GD (2011) Application of 3D inversion to magnetotelluric profile data from the Deccan Volcanic Province of Western India. Physics of the Earth and Planetary Interiors 187(1-2):33-46. https://doi.org/10.1016/j.pepi.2011.04.005 Phillips JD, Duval JS, Ambroziak RA (1993) National geophysical data grids: gamma-ray, gravity, magnetic, and topographic data for the conterminous United States. United States Geological Survey Digital Data Series 9. http://mrdata.usgs.gov/geophysics/gravity.html (accessed Oct. 1, 2016). Pilipenk VA, Federov EN (1993) Magnetotelluric sounding of the crust and hydromagnetic monitoring of the magnetosphere with the use of ULF waves. Annals of Geophysics 36(5- 6):19-32. https://doi.org/10.4401/ag-4243 Pilipenko VA, Federov EN (1994) Magnetotelluric Sounding of the Crust and Hydromagnetic Monitoring of the Magnetosphere with the Use of ULF Waves. In Engebretson MJ, Takahashi K, Scholer M (eds) Solar Wind Sources of Magnetospheric Ultra-Low-Frequency Waves. American Geophysical Union, Washington, 283-292. https://doi.org/10.1029/GM081p0283

Poag CW, Sevon WD (1989) A record of Appalachian denudation in postrift Mesozoic and Cenozoic sedimentary deposits of the U.S. Middle Atlantic continental margin. Geomorphology 2:119-157. https://doi.org/10.1016/0169-555X(89)90009-3 Pollitz FF, Mooney WD (2016) Seismic velocity structure of the crust and shallow mantle of the Central and Eastern United States by seismic surface wave imaging. Geophysical Research Letters 43:118–126. https://doi.org/10.1002/2015GL066637 Pommier A, Garnero EJ (2014) Petrology-based modeling of mantle melt electrical conductivity and joint interpretation of electromagnetic and seismic results. Journal of Geophysical Research: Solid Earth 119:4001-4016. https://doi.org/10.1002/2013JB010449 Pommier A, Kohlstedt DL, Hansen LN, Mackwell S, Tasaka M, Heidelbach F, Leinenweber K (2018) Transport properties of olivine grain boundaries from electrical conductivity experiments. Contributions to Mineralogy and Petrology 173(41). https://doi.org/10.1007/s00410-018-1468-z Porter R, Liu Y, Holt WE (2016) Lithospheric records of orogeny within the continental U.S. Geophysical Research Letters 43:144–153. https://doi.org/10.1002/2015GL066950 Poulter EM, Allan W, Bailey GJ (1988) ULF Pulsation Eigenperiods within the Plasmasphere. Planetary and Space Science 36(2):185-196. https://doi.org/10.1016/0032- 0633(88)90054-2 Prince PS, Spotila JA (2013) Evidence of transient topographic disequilibrium in the landward passive margin river system: knickpoints and paleo-landscapes of the New River basin, southern Appalachians. Earth Surface Processes and Landforms 38:1685- 1699. https://doi.org/10.1002/esp.3406 Ragland PC, Hatcher RD Jr, Whittington D (1983) Juxtaposed Mesozoic diabase dike sets from the Carolinas: A preliminary assessment. Geology 11:394–399. https://doi.org/10.1130/0091-7613(1983)11<394:JMDDSF>2.0.CO;2 Ranganayaki RP, Madden TR (1980) Generalized thin sheet analysis in magnetotellurics: an extension of Price’s analysis. Geophysical Journal of the Royal Astronomical Society 60(3):445-457. https://doi.org/10.1111/j.1365-246X.1980.tb04820.x Sacks PE, Secor DT Jr (1990) Delamination in collisional orogens. Geology 18(10):999- 1002. https://doi.org/10.1130/0091-7613(1990)018<0999:DICO>2.3.CO;2 Samson SD, Coler DG, Speer JA (1995) Geochemical and Nd-Sr-Pb isotopic composition of Alleghanian granites of the southern Appalachians: Origin, tectonic setting, and source

characterization. Earth and Planetary Science Letters 134(3-4):359-376. https://doi.org/10.1016/0012-821X(95)00124-U Savage B, Covellone BM, Shen Y (2017) Wave speed structure of the eastern North American margin. Earth and Planetary Science Letters 459:394-405. https://doi.org/10.1016/j.epsl.2016.11.028 Schmandt B, Lin F-C (2014) P and S wave tomography of the mantle beneath the United States. Geophysical Research Letters 41:6342–6349. https://doi.org/10.1002/2014GL061231 Schmandt B, Lin F-C, Karlstrom KE (2015) Distinct crustal isostasy trends east and west of the Rocky Mountain Front. Geophysical Research Letters 42:10290-10298. https://doi.org/10.1002/2015GL066593 Schutt DL, Lesher CE (2006) Effects of melt depletion on the density and seismic velocity of garnet and spinel lherzolite. Journal of Geophysical Research: Solid Earth 111. https://doi.org/10.1029/2003JB002950 Schutt DL, Lesher CE (2010) Compositional trends among Kaapvaal Craton garnet peridotite xenoliths and their effects on seismic velocity and density. Earth and Planetary Science Letters 300:367-373. https://doi.org/10.1016/j.epsl.2010.10.018 Selway K (2014) On the causes of electrical conductivity anomalies in tectonically stable lithosphere. Surveys in Geophysics 35:219–257. https://doi.org/10.1007/s10712-013- 9235-1 Selway K, Ford H, Kelemen P (2015) The seismic mid-lithosphere discontinuity. Earth and Planetary Science Letters 414:45-57. https://doi.org/10.1016/j.epsl.2014.12.029 Shen W, Ritzwoller MH (2016) Crustal and uppermost mantle structure beneath the United States. Journal of Geophysical Research: Solid Earth 121:4306–4342. https://doi.org/10.1002/2016JB012887 Simpson RW, Jachens RC, Blakely RJ, Saltus RW (1986) A New Isostatic Residual Gravity Map of the Conterminous United States with a Discussion on the Significance of Isostatic Residual Anomalies. Journal of Geophysical Research 91:8348-8372. https://doi.org/10.1029/JB091iB08p08348 Slingerland R, Furlong KP (1989) Geodynamic and geomorphic evolution of the Permo- Triassic Appalachian Mountains. Geomorphology 2:23-37. https://doi.org/10.1016/0169- 555X(89)90004-4

Smith DL, Gregory RG, Emhof JW (1981) Geothermal Measurements in the Southern Appalachian Mountains and Southeastern Coastal Plains. American Journal of Science 281:282-298. https://doi.org/10.2475/ajs.281.3.282 Smirnov MYu, Egbert GD (2012) Robust principal component analysis of electromagnetic arrays with missing data. Geophysical Journal International 190:1423-1438. https://doi.org/10.1111/j.1365-246X.2012.05569.x Sobolev AV, Hofmann AW, Kuzmin DV, Yaxley GM, Arndt NT, Chung S-L, Danyushevsky LV, Elliott T, Frey FA, Garcia MO, Gurenko AA, Kamenetsky VS, Kerr AC, Krivolutskaya NA, Matvienkov VV, Nikogosian IK, Rocholl A, Sigurdsson IA, Sushchevskaya NM, Teklay M (2007) The Amount of Recycled Crust in Sources of Mantle-Derived Melts. Science 316:412-417. https://doi.org/10.1126/science. 1138113 Spasojević S, Liu L, Gurnis M, Müller RD (2008) The case for dynamic subsidence of the U.S. east coast since the Eocene. Geophysical Research Letters 35:L08305. https://doi.org/10.1029/2008GL033511 Steltenpohl MG, Zietz I, Horton JW Jr, Daniels DL (2010) New York-Alabama lineament: A buried right-slip fault bordering the Appalachians and mid-continent North America. Geology 38:571-574. https://doi.org/10.1130/G30978.1 Sun J, Kelbert A, Egbert GD (2015) Ionospheric current source modeling and global geomagnetic induction using ground geomagnetic observatory data. Journal of Geophysical Research: Solid Earth 120(10):6771–6796. https://doi.org/10.1002/2015JB012063 Takei Y (2017) Effects of Partial Melting on Seismic Velocity and Attenuation: A New Insight from Experiments. Annual Review of Earth and Planetary Sciences 45:447-470. https://doi.org/10.1146/annurev-earth-063016-015820 van der Lee S, Regenauer-Lieb K, Yuen DA (2008) The role of water in connecting past and future episodes of subduction. Earth and Planetary Science Letters 273:15-27. https://doi.org/10.1016/j.epsl.2008.04.041 Viljanen A (2012) Description of the magnetospheric/ionospheric sources. In Chave AD, Jones AG (eds) The Magnetotelluric Method: Theory and Practice. Cambridge University Press, 96-121. https://doi.org/10.1017/CBO9781139020138.005 Vozoff K (ed) (1986) Magnetotelluric Methods. Society of Exploration Geophysicists, 763 p. Wagner LS, Stewart K, Metcalf K (2012) Crustal-scale shortening structures beneath the Blue Ridge Mountains, North Carolina, USA. Lithosphere 4:242-256. https://doi.org/10.1130/L184.1

Wagner LS, Fischer KM, Hawman R, Hopper E, Howell D (2018) The relative roles of inheritance and long-term passive margin lithospheric evolution on the modern structure and tectonic activity in the southeastern United States. Geosphere 14:1–26. https://doi.org/10.1130/GES01593.1 Wait JR (1954) On the relation between telluric currents and the Earth’s magnetic field. Geophysics 19(2):281-289. https://doi.org/10.1190/1.1437994 Wait JR (1982) Geo-Electromagnetism. Academic Press, 268 p. Wang H, Currie CA (2015) Magmatic expressions of continental lithosphere removal. Journal of Geophysical Research: Solid Earth 120:7239-7260. https://doi.org/10.1002/2015JB012112 Wang H, van Hunen J, Pearson DG, Allen MB (2014) Craton stability and longevity: The roles of composition-dependent rheology and buoyancy. Earth and Planetary Science Letters 391:224-233. https://doi.org/10.1016/j.epsl.2014.01.038 Wannamaker PE (2005) Anisotropy Versus Heterogeneity in Continental Solid Earth Electromagnetic Studies: Fundamental Response Characteristics and Implications for Physicochemical State. Surveys in Geophysics 26(6):733-765. https://doi.org/10.1007/s10712-005-1832-1 Waters CL, Menk FW, Fraser BJ (1991) The Resonance Structure of Low Latitude Pc3 Geomagnetic Pulsations. Geophysical Research Letters 18(12):2293-2296. https://doi.org/10.1029/91GL02550 Waters CL, Menk FW, Thomsen MF, Foster C, Fenrich FR (2006) Remote Sensing the Magnetosphere Using Ground-Based Observations of ULF Waves. In Takahashi K, Chi PJ, Denton RE, Lysak RL (eds) Magnetospheric ULF Waves: Synthesis and New Directions. American Geophysical Union, Washington, 319-340. https://doi.org/10.1029/169GM21 Weidelt P (1972) The Inverse Problem of Geomagnetic Induction. Zeitschrift für Geophysik 38:257-289. Weidelt P, Chave AD (2012) The magnetotelluric response function. In Chave AD, Jones AG (eds) The Magnetotelluric Method: Theory and Practice. Cambridge University Press, 120-164. https://doi.org/10.1017/CBO9781139020138.006 Whalen L, Gazel E, Vidito C, Puffer J, Bizimis M, Henika W, Caddick MJ (2015) Supercontinental inheritance and its influence on supercontinental breakup: The Central Atlantic Magmatic Province and the breakup of Pangea. Geochemistry, Geophysics, Geosystems 16:3532–3554. https://doi.org/10.1002/2015GC005885

100

Willett SD, McCoy SW, Perron JT, Goren L, Chen C-Y (2014) Dynamic Reorganization of River Basins. Science 343:9. https://doi.org/10.1126/science.1248765 Yang B, Egbert GD, Kelbert A, Meqbel NM (2015) Three-dimensional electrical resistivity of the north-central USA from EarthScope long period magnetotelluric data. Earth and Planetary Science Letters 422:87–93. https://doi.org/10.1016/j.epsl.2015.04.006 Yang Y, Forsyth DW (2008) Attenuation in the upper mantle beneath Southern California: Physical state of the lithosphere and asthenosphere. Journal of Geophysical Research 113:B03308. https://doi.org/10.1029/2007JB005118 Yoshino T, Matsuzaki T, Shatskiy A, Katsura T (2009) The effect of water on the electrical conductivity of olivine aggregates and its implications for the electrical structure of the upper mantle. Earth and Planetary Science Letters 288:291-300. https://doi.org/10.1016/j.epsl.2009.09.032

101

Appendices

102

Appendix 1

Supplemental Materials to Accompany

Electrical Conductivity Structure of Southeastern North America: Implications for Lithospheric Architecture and Appalachian Topographic Rejuvenation

103

A1.1 Preferred inverse model data fit The preferred inverse model presented in the text fits to an overall normalized root-mean- square error (nRMSE) of 1.75, although this number is only meaningful in the context of the resolution tests presented below. The distribution of site-by-site nRMSE values is shown in Figure A1.1, representative data fits are shown in Figure A1.2, and the spatial distribution of model misfit is shown in the left panel of Figure A1.3.

Figure A1.1: Distribution of site-by-site nRMSE.

104

Figure A1.2: Representative data misfits for the preferred inverse model presented in the text and the forward models used to test the depth constraint on the Piedmont resistor. Points are measured data; curves are modeled responses (solid curves are the XY component, dashed curves are the YX component). Each column contains plots of apparent resistivity and phase for a single site. The title for each column gives the site misfit for the preferred inverse model. Site TNT51 lies in eastern Tennessee, in the Appalachian Highlands. Site SCW55 is in the South Carolina Piedmont. Site VAR57 is in the Virginia Piedmont. The apparent resistivity plots for sites SCW55 and VAR57 clearly show that the data fit becomes unacceptable when structure R1 is truncated at 200 km depth or shallower.

105

Figure A1.3: Spatial distribution of site misfits for the preferred inverse model (left large panel) and the forward models used to evaluate the depth extent of the Piedmont resistor (four right smaller panels). The left plot shows actual nRMSE for the preferred model. The right four plots are given in percent change in nRMSE from the preferred inverse model. Negative percent changes (blue) indicate a better fit; positive percent changes (red) indicate a worse fit. These plots show that the data misfit becomes unacceptable if the Piedmont resistor is truncated at depths of 200 km or shallower. The site-by-site misfits used here were calculated for the entire period range. Using only long periods (>100 s or >1000 s) to calculate site-by-site misfits does not change our conclusions about the requirements that the data place upon the depth extent of structure R1.

A1.2 Constraints on maximum depth of the Piedmont resistor In order to evaluate the requirements that the data place upon the depth extent of structure R1 (the Piedmont resistor), we ran several forward models with the structure truncated at progressively shallower depths in order to see how the data misfit changes. In these forward models, the entire model domain below the truncation depth was set to a resistivity of 100 Ωm. The outcomes of these tests are shown in Figures A1.2 and A1.3. Representative cross sections through these forward models are shown in Figure A1.4. In order to test to robustness of structure R1, we ran an inversion with a region beneath the Piedmont frozen to 100 Ωm. The results of this test are shown in Figure A1.5. The inversion made the region around the frozen conductive body highly resistive, thereby isolating it electrically from the rest of the model. This indicates that the data require resistive material beneath the Piedmont

106 to disrupt currents from readily flowing through that region. While this test indicates that isolated conductive material could be embedded within structure R1, it must overall be highly resistive in order to electrically isolate any such conductors. Note also that bulk resistivity values of >1000 Ωm still persist to >200 km in the inverse solution shown in Figure A1.5. The presence of small, localized conductive pockets within R1 therefore does not change our conclusions about the depth extent of the sub-Piedmont thermal lithosphere.

Figure A1.4: Cross sections through forward models that were used to constrain the depth extent of the Piedmont resistor. These cross sections follow transect B-B’, shown in the main text.

107

Figure A1.5: Plots of an inversion in which a region of the sub-Piedmont lithosphere was forced to remain conductive and the associate misfit. This model fits the data slightly worse than the preferred model does (nRMSE of 2.03). This test demonstrates that the Piedmont must be overall electrically resistive in order to impede long-range current flow.

108

A1.3 Constraints on the horizontal extent of the Piedmont resistor Two sets of tests were conducted in order to constraint the limits on the lateral extent of the Piedmont resistor. The results of these tests are shown in Figures A1.6 and A1.7. Figure A1.6 shows that the data misfit increases markedly to unacceptable levels if the northwestern extent of R1 is forced to the southeast, into the Coastal Plain. If the northwestern extent of R1 is forced to the northwest, as in the top two panels in Figure A1.6, the misfit generally increases somewhat, with some sites showing a significant increase in misfit. Although the overall misfit increase in extending resistive material into the sub- Appalachian lithosphere is perhaps not as clear as with the previous case, consideration of Figures A1.8 and A1.9 as well indicates that a conductivity gradient at the location that one appears in the preferred inverse solution is required by the data. In both Figures A1.6 and A1.8, misfit at sites that lie along the conductivity contrast shown in Figure 1.2 increases when resistive material is extended to the northwest. If the magnitude of the conductivity contrast is decreased, as in Figure A1.9, then the misfit at those sites also increases. Together, these three tests indicate that the northwestern extent of R1 is reasonably well located at the modern Appalachian topographic escarpment. These tests also show that the data require a conductivity contrast beneath the Blue Ridge Front between ~1000 Ωm material to the southeast and ~100 Ωm material to the northwest. Figure A1.7 shows that the southeastern extent of R1 cannot lie beneath the Coastal Plain, as the data misfit increases significantly if conductive material is placed beneath the coast of North Carolina and South Carolina. Figure A1.7 also shows that the misfit at certain sites along the coast increases if resistive material is extended far off shore. However, as Figure A1.11 shows, a resistivity model that includes resistive off-shore lithosphere can be found that still fits the data to a level comparable to that of the preferred inverse solution. The fact that the misfit at certain sites increases when resistive material is extended unimpeded beyond the coast therefore suggests that an off-shore conductive feature is required by the data. Structure EC1 therefore likely represents a real feature, as discussed in the main text. However, with no constraints on the geometry of an off-shore conductor, we cannot constrain the extent of R1 beyond the coast.

109

Figure A1.6: Resolution tests to constrain the northwestern limit of R1. The plots on the left show the synthetic models (adapted from the preferred inverse solution shown in Fig. 1.2) that were used. The plots on the right show the change in misfit (calculated from the forward response of each model) with respect to the misfit for the preferred inverse solution. The right plots also show a depth slice at ~150 km through the respective synthetic forward model. In the right plots, red indicates an increase in site misfit, and blue indicates a decrease in site misfit.

110

Figure A1.7: Resolution tests to constrain the southeastern limit of R1. As with Figure A1.6, the plots on the left show the synthetic models (adapted from the preferred inverse solution) that were used. The plots on the right show the change in misfit (calculated from the forward response of each model) with respect to the misfit for the preferred inverse solution. The right plots also show depth slices through the synthetic test models. In the plots on the right, red indicates an increase in site misfit, and blue indicates a decrease in site misfit.

111

A1.4 Constrains on sub-Appalachian lithospheric resistivity Figures A1.8 and A1.9 show tests that were performed to constrain the resistivity of the lithosphere beneath the modern Appalachian topographic highs. As discussed in the previous section above, Figure A1.8 suggests that the data require a conductivity contrast roughly coincident with the modern Appalachian topographic escarpment, as the misfit increases at sites along the northwestern edge of R1 (in the preferred inverse solution) when Appalachian lithosphere is set to 1000 Ωm (comparable to the resistivity of R1). In Figure A1.9, the data misfit at sites along the northwestern edge of R1 increases when Appalachian lithospheric resistivity is increased from ~100 Ωm to ~300 Ωm. The sites directly overlying the sub- Appalachian lithosphere do not show an increase in misfit with these changes, which may be due to screening from crustal conductors. However, the nearby sites located over the conductivity gradient do provide information about and useful constraints on the sub- Appalachian lithosphere. These tests indicate that the data require the resistivity of sub- Appalachian lithosphere to be ~100 Ωm in order to produce the necessary conductivity contrast with R1.

112

Figure A1.8: Results from a forward test in which cells in the preferred inverse solution between ~100 km and ~250 km were set to be at least 1000 Ωm. Upper left plot shows the change in data misfit with respect to the misfit of the preferred solution; red indicates an increased misfit, blue indicates a decreased misfit. The upper right plot shows a cross-section through the synthetic test model. The lower two plots show the absolute normalized misfit for the synthetic forward model (left) and the preferred inverse solution (right).

113

Figure A1.9: Results from a test in which all model cells beneath ~100 km were set to be at least 300 Ωm. As with Figure A1.8, the upper left plot shows the change in data misfit with respect to the preferred inverse solution for this synthetic model (red indicates increased site misfit, blue indicates decreased site misfit); the upper right plot shows a cross section through the synthetic test model; and the lower two plots show the absolute normalized misfits for this test model (left) and the preferred inverse solution (right).

114

A1.5 Robustness of other features described in the main text To evaluate the robustness of other structures described in the main text, we exploit the covariance scheme implemented in ModEM. The code penalizes deviations from a prior model, so we consider structures that appear in multiple inversions that started from different prior models to be required by the data. Figures A1.10, A1.11, and A1.12 present alternative inverse models that fit the data to an overall nRMSE similar to that of the preferred inverse model presented in the text. The inversion shown in Figure A1.10 used a prior model that included thick, resistive oceanic lithosphere as well as the Atlantic Ocean and the Gulf of Mexico. The resulting inverse solution fits the data to an nRMSE of 1.67, which is essentially the same as that of the preferred inverse solution shown in the main text. The inversion shown in Figure A1.11 used a prior model that included resistive oceanic lithosphere and ~5 km of 30 Ωm sediments on the Atlantic shelf. The resulting solution fits the data to an nRMSE of 1.83, again essentially the same as that of the preferred inverse solution shown in the text. The inversion shown in Figure A1.12 used a 30 Ωm half-space that included the Atlantic Ocean and the Gulf of Mexico as a prior model. The resulting inverse model fits the data to an nRMSE of 1.68, once again essentially the same as that of the preferred solution. This prior model is unrealistically conductive, but it is useful to evaluate the robustness of features that appear in the inversions. Note that structure R1 still extends to more than 200 km depth in this inverse solution.

115

Figure A1.10: Results from an inversion that includes resistive ocean lithosphere (note high resistivity values in lower right of the depth slices for 35 km, 50 km, and 100 km depth).

116

Figure A1.11: Results from an inversion that includes resistive ocean lithosphere as well as conductive shelf sediments.

117

Figure A1.12: Results from an inversion that started with a 30 Ωm half-space and that included the Atlantic Ocean and the Gulf of Mexico. Note that structure R2 is blurred in this inversion.

118

A1.6 Additional supporting information for the Appalachian Highland-Piedmont contrast In rotating the impedance tensors to a coordinate system that is aligned with the trend of the modern Appalachian Mountains, we find that the off-diagonal apparent resistivities at longest periods calculated for currents that flow perpendicular to the topographic escarpment are systematically higher in the lowlands (Piedmont and Coastal Plain) than in the Appalachian highlands. This observation is shown in Figure A1.13. Such a relationship is weak for currents that flow parallel to the topographic escarpment because the resistive Piedmont lithosphere is effectively in a parallel circuit with the conductive Appalachian lithosphere. Current will therefore flow preferentially through the Appalachian lithosphere. For currents that flow perpendicular to the trend of the topographic escarpment, the Piedmont lithosphere and the Appalachian highland lithosphere are effectively in a series circuit. Such currents must traverse both domains and therefore are sensitive to both lithospheric regions. Assuming that the apparent resistivity components corresponding to currents flowing perpendicular to the Appalachian trend at the longest periods are normally distributed in each lithospheric domain (this scatter is due to surface distortion), the difference between the mean apparent resistivity value at lowland sites and the mean apparent resistivity value at highland sites (shown in Fig. A1.13) is statistically significant at the 99% confidence level. (This analysis utilizes a one-sided Student’s t-test that assumes non-equal variances for the two distributions.) Figure A1.14 shows the spatial distribution of apparent resistivity and phase values in map view. The high apparent resistivities observed for the Piedmont sites are not an artifact of surface distortion, as both off-diagonal modes display high apparent resistivities (although they do so to differing degrees, as discussed above). The off-diagonal components do not display a large split that would be diagnostic of extreme surface distortion.

119

Figure A1.13: Distributions of apparent resistivities calculated for off-diagonal components of the impedance tensor (rotated to be aligned with the trend of the coastline). The difference in apparent resistivities between sites in the Appalachian Highlands and sites in the Piedmont is clearly visible.

120

Figure A1.14: Off-diagonal components of the impedance tensor at 1092 s plotted as apparent resistivity and phase. The contrast in the data between Piedmont sites and Appalachian highland sites is visible in the apparent resistivity values for currents oriented 45° S of E. The phase for currents oriented 45° S of E along the coast are sensing the transition from conductive seawater to resistive continental rocks.

121

A1.7 Consideration of tipper data Several forward models were evaluated in order to see which of the hypothesized structures to the southwest of our current data footprint could produce the pattern observed in the tipper (vertical magnetic field transfer function) data from Georgia and South Carolina. The results of these tests, plotted as induction vectors with the Parkinson sign convention, are shown in Figure A1.15. A thick package of conductive sediments in the Gulf of Mexico seems to be an unlikely explanation for the observed large-magnitude, spatially coherent induction vectors. Even when the sheet of conductive material is thick (~10 km) and extends slightly on land, the tippers in most of Georgia are not deflected in the manner observed in the actual data. When a cylinder of highly conductive material is placed in the mantle lithosphere (from ~50 km to ~200 km depth), the induction vectors are deflected and elongated through much of Georgia, although they remain undeflected in South Carolina. The tipper response from a large resistive block with similar geometry to structure R1 displays the deflection observed in the data, but it does not display the increased tipper magnitude. Based on these tests, either a conductor beneath the northernmost Gulf of Mexico or the edge of structure R1 seem to be better explanations than the thick sediment package in the Gulf of Mexico, although neither seems to fully explain the induction vector data. A combination of these possible explanations may be necessary to fully reproduce the observed induction vectors.

122

Figure A1.15: Observed vertical magnetic field transfer function data plotted as induction vectors (upper left) and the corresponding forward-modeled induction vectors for various possible structures that are presented in the main text. These data are for a period of ~1000 s. The induction vectors have the same scale in all panels. All forward models included the ocean. The upper right forward model has a cylindrical conductor of 1 Ωm embedded in a uniform 180 Ωm halfspace between ~50 km and ~200 km depth. The lower left model has a thick (~10 km) tabular conductor set to 30 Ωm. The lower right model has a resistive block of similar geometry to structure R1 (described in the main text) set to 1000 Ωm within a 180 Ωm halfspace.

123

Appendix 2

Supplemental Materials to Accompany

Source Biases in Mid-Latitude Magnetotelluric Transfer Functions due to Pc3-4 Geomagnetic Pulsations

124

A2.1 Cross-Phase Analysis For two (real-valued) time series x(t) and y(t), let X(f) and Y(f) be the complex-valued Fourier frequency-domain representations of the time series. The cross-spectrum of these time series is given by the product X*(f)Y(f), and their phase spectrum (or cross-phase) is given by the phase angle of the quantity X*(f)Y(f), where * denotes complex conjugation. By writing X(f) and Y(f) in complex exponential (Euler) form, one can see that the cross- phase is simply the phase difference between the signal (at a specific frequency) in the two given time series:

∗ −푖휑푋(푓) 푖휑푌(푓) 푖(휑푌(푓)−휑푋(푓)) 푋 (푓)푌(푓) = 퐴푋(푓)푒 ∙ 퐴푌(푓)푒 = 퐴푋(푓) ∙ 퐴푌(푓)푒 = 퐵(푓)푒푖∆휑(푓) ∗ Phase[푋 (푓)푌(푓)] = ∆휑(푓) = 휑푌(푓) − 휑푋(푓) The cross-phase can be localized in time as all as frequency by using either a windowed short-time Fourier transform or a wavelet transform. In either case, the cross-phase value for a specific frequency f at a given time segment t is still X*(f, t)Y(f, t), where the complex- valued frequency domain coefficients are obtained by either transform. The cross-phase analysis used in this study utilized the wavelet-based technique documented by Waters et al. (2006). For horizontal magnetic field time series, the N-S (X) geomagnetic component cross- phase between two stations therefore expresses the phase difference (that is, signal lead/lag) between the signals observed at the two different stations as a function of frequency (or period). For a quasi-uniform horizontal magnetic field (the idealized MT source field) over a 1D Earth, the cross-phase between two arbitrarily spaced stations should be zero. In a non-1D Earth, induced currents will generate anomalous horizontal magnetic fields that will cause the cross-phase to be non-zero. However, due to the diffusive nature of solid Earth electromagnetic induction, the cross-phases associated with the anomalous horizontal magnetic fields will show broad-band deviations from zero phase that will vary smoothly and slowly in period. Sharp and rapid (i.e., narrow-band) deviations from zero in the cross-phase spectrum therefore indicate the horizontal magnetic fields are significantly out of phase due to some non-inductive process, which in the case considered here would be geomagnetic field-line resonance in the source fields. Non-zero cross-phases in source magnetic fields violate the idealized plane-wave source assumption of the MT method.

125

Figure A2.1: Conceptual plots to illustrate the physics of geomagnetic field line resonances (FLRs). Part (a) demonstrates resonance due to an external forcing (after Hughes, 1994). The upper plot shows a monochromatic disturbance and the variation in FLR eigenperiod with latitude. Where the period of the incident forcing matches the FLR eigenperiod, the amplitude response due to that forcing will be amplified, as shown in the lower plot. This resonance is superposed over a larger scale overall magnetospheric response to the forcing (this response consists of fast-mode compressional waves and strengthens at larger L-shells, as those field lines extend farther into the magnetosphere; Kawano et al., 2002). In real life, the external forcing will be broadband, so a meridional range of FLRs (as well as harmonics) may be excited. Part (b) shows the fundamental ideas behind FLR detection techniques as laid out by Waters et al. (1991). The two plots here show amplitude and phase as a function of period as would be expected for damped, driven harmonic oscillators. For two meridionally spaced stations, which will be located at field lines with two distinct resonant periods, the difference in their observed amplitude and phase shows the eigenperiod of the FLR between the two stations. The cross- phase technique uses specifically the phase difference. Part (c) displays an alternative interpretation of the cross phase (Chi and Russell, 1998). For a given monochromatic external forcing, a range of field lines will respond, although the amplitude response will be greatest at the resonant field line and the phase response will lead or lag the forcing on either side of the resonant field line. (Because field-line eigenperiod varies with latitude, this is analogous to forcing a harmonic oscillator over a range of periods above and below resonance.) Consequently, there is a 180° phase shift when moving geomagnetically N-S through an FLR, so two meridionally spaced stations will record different phases for that discrete period.

126

Figure A2.2: Transfer functions and cross-phase analyses for a pair of sites located in the conductive Snake River Plain (see Fig. 2.1). Layout and symbology are the same as in Figure 2.2. The cross- phase spectra show clear FLR peaks, but the TFs are not clearly biased in the corresponding period bands. The Tx component of the tipper may show a slight bias, although this is subtle. The higher conductivity of the Earth in the Snake River Plain presumably decreases the skin depth of source harmonic components in the Pc3-4 period band so that the plane-wave assumption is not violated despite the short spatial scale of the FLR.

127

Figure A2.3: Transfer functions for the two sites (NCT57 and IAL57) used in the cross-phase analyses shown in Figure 2.2. Layout and symbology are the same as in Figure 2.2.

128

Figure A2.4: Plots of each component of the remote-reference estimate for the full impedance tensor of site SCV57. Each component is plotted as its real and imaginary part multiplied by the square root of period. Pc-band bias is apparent in all four components.

129

Figure A2.5: Different TF estimates for site SCV57 (accepted EarthScope TF estimates for this site shown on the left in Figure 2.2). For the single-station TF estimate (leftmost column), Pc-band bias is apparent in the vertical magnetic field transfer functions but is obscured in the apparent resistivity and phase by incoherent (spatially localized) noise at short periods. The right two columns show robust remote-reference TFs estimates using remote sites different than that used for the TFs in Figure 2.2. Site TTW52 is ~350 km west of SCV57; site NCT57 (the site used in the cross-phase analysis) is ~150 km north of SCV57. Clearly, using different remote reference sites does not mitigate the biases from FLRs. The remote site for the TFs in Figure 2.2 is NCW59, which is ~150 km east-northeast of SCV57.

130

Figure A2.6: Comparison between robust remote reference transfer function estimates that excluded record times with clearly apparent Pc activity (right) and the original robust remote reference transfer function estimates (left) from the IRIS SPUD EMTF database. The original TF estimates on the left used the entire record length and clearly show biases due to Pc’s associated with FLRs (these are the same TFs shown in Fig. 2). For the TFs on the right, the time series were visually inspected for Pc signatures, and the record segments that clearly showed Pc activity were excluded from TF estimation. Pc’s are apparent in raw magnetic field time series as highly sinusoidal variations with periods of ~10-100 s and peak-to-peak amplitudes of ~1 nT (see Figure A2.7). For this specific station, ~30% of the time series data contained clear Pc activity. While the FLR biases are less severe in these TF estimates, they are still apparent, especially in the vertical magnetic field TFs.

131

Figure A2.7: Raw electric and magnetic field time series from EarthScope MT site IAN37. An arbitrary constant value has been subtracted from each record shown here for simplicity in plotting. FLR activity is clearly apparent as highly sinusoidal, nearly monochromatic Pc3-4 oscillations in all field components.

132

Appendix 3

Supplemental Materials to Accompany

Synthesizing Seemingly Contradictory Seismic and Magnetotelluric Observations in the Southeastern United States to Image Physical Properties of the Lithosphere

133

A3.1 Seismic surface-wave inverse solutions

Table A3.1.1: Information about the five seismic surface-wave models that we compare to our MT inverse solutions. (Thanks to Emily Hopper for starting this table.)

Model Data Type Data Set Model Inversion Approach Parameter Wagner et al. Rayleigh wave phase USArray TA + Vs, azimuthal Crustal thickness 2018 velocity (33 – 143 s) local arrays anisotropy constraints from receiver (SESAME, functions, wavefield PHRACCS) migration, CRUST1.0; starting model parameterized with sediments, crust, and (almost) constant 4.5 km/s mantle Shen & Rayleigh wave phase USArray TA Vs, crustal Joint Bayesian Monte Ritzwoller velocity (8-90 s) and group thickness Carlo inversion 2016 velocity (8-40 s), receiver functions, Rayleigh wave ellipticity Pollitz & Rayleigh wave phase USArray TA Vs Reference crustal Mooney velocity (18-125 s) thickness from 2016 CRUST1.0; reference seismic structure from PREM with appended crustal structure Porter, Liu, Rayleigh wave phase USArray TA Vs Crustal thickness from Holt 2015 velocity (from ambient noise EARS, sediment thickness tomography 8-40 s, wave from CRUST; starting gradiometry 20-150s) model parameterized in terms of sediment, crust, and constant 4.5 km/s mantle Schmandt, Receiver functions for crustal USArray + Vs, crustal Lin, thickness; Rayleigh wave regional thickness Karlstrom phase velocity and ellipticity observatories 2015 (24-100 s)

134

A3.2 Seismic velocity calculations We rely on two different models for anelasticity in our seismic velocity calculations: the extended Burgers model of Jackson & Faul (2010) and the model proposed by Karato et al. (2015). For the model of Jackson & Faul (2010), we use Matlab scripts that are made publicly available by Uli Faul (http://web.mit.edu/hufaul/www/Anelasticity.html). The model presented by Jackson & Faul (2010) includes an explicit grain-size dependence in the calculations of attenuation and seismic velocity; in this model, grain size in part controls the magnitude of velocity reduction. However, Karato et al. (2015) have taken exception with this parameterization, and instead propose a model with implicit grain-size dependence based on theoretical considerations. In their model, grain size controls the characteristic frequency of anelasticity, but it does not affect the magnitude of velocity reduction. Here, we summarize our implementation of the Karato et al. (2015) model. Karato et al. (2015) argue that the magnitude of velocity reduction due to anelasticity is independent of grain size. For their calculations, they assume a constant arbitrary value of several percent, as they currently lack laboratory data to constrain the magnitude of velocity change. Here, to work within their model framework, we assume a velocity reduction of 3%. They further argue that grain size controls the characteristic frequency for the transition between unrelaxed (anelastic effects are insignificant) and relaxed (anelastic effects are important) shear moduli: 푟 퐴 −(퐸+푃푉)⁄푅푇 퐶푤 휔0 = 푒 (1 + ( ) ), (A3.2.1) 푑 퐶0 where A is a constant (2.3x1014), d is grain size (5 mm), E is the activation energy (350 kJ/mol), V is the activation volume (10 cm3/mol), P is the pressure, T is the temperature, R is the gas constant (8.314 J/mol-K), Cw is the water content, C0 is the reference water content, and r is a constant. We assume completely dry conditions (as discussed in the main text), so the term involving water content vanishes. Karato et al. (2015) distinguish between several anelastic effects that affect seismic velocity and attenuation; the two most important in these calculations are Elastically Accommodated Grain-Boundary Sliding (EAGBS) and Diffusionally Accommodated Grain- Boundary Sliding (also called absorption band behavior, AB). AB occurs at lower frequencies and greater depths (higher temperatures) than EAGBS, and both contribute to seismic velocity reduction (see also the main text for more information). The contribution of each anelastic effect to the total velocity reduction can be separated,

훿푉 훿푉 훿푉 = ( ) + ( ) , (A3.2.2) 푉 푉 퐸퐴퐺퐵푆 푉 퐴퐵

135 where V is specifically S-wave velocity. The first term (EAGBS) is given by

훿푉 ∆ ( ) = 2훽, (A3.2.3) 푉 퐸퐴퐺퐵푆 1+(휔⁄휔0) where Δ is the total (maximum) magnitude of velocity reduction (0.03, as discussed above),

ω is the seismic wave frequency, ω0 is the characteristic frequency (Eq. A3.2.1), and β is a constant (0.4) that characterizes the sharpness of the relaxation peak. The second term (AB) is inversely proportional to attenuation,

훿푉 1 휋훼 ( ) = cot 푄−1, (A3.2.4) 푉 퐴퐵 2 2 where α is a constant (0.3) and Q is the quality factor (Q-1 is attenuation). We use a Q value of 170 based on the work of Bao et al. (2016). To get the total anelastic velocity reduction, for a given temperature and pressure (which corresponds to a specific depth based on the assumed geotherm), we first calculate the characteristic frequency of the unrelaxed-relaxed transition. If the seismic frequency (different for surface or body waves; we use 10 s, 0.1 Hz, for body waves) is less than the critical/characteristic frequency, then we add both the EAGBS and the AB term to get the total fractional velocity reduction. If the seismic frequency is greater than the critical frequency, then the AB term is set to zero so that only the EAGBS term contributes to the total fractional velocity reduction. This piecewise addition effectively takes into account the “turning-on” of AB below the characteristic frequency. The formulation of the EAGBS contribution to velocity reduction (Eq. A3.2.3) in effect “ramps up” when the characteristic frequency is close to the seismic frequency, so the “turning-on” effect is included in this function. The calculated total fractional velocity reduction is then used to determine the relaxed S-wave velocity against the value calculated using only the anharmonic temperature and pressure derivatives. To get the P-wave velocity reduction, we scale the S-wave fractional velocity reduction by a factor of (4/3)2 (~1.778), which is an approximate scaling factor for a material with a Poisson ratio of ~0.25 (commonly assumed for the mantle). The Karato et al. (2015) model currently assumes many fixed parameters, as the experimental database to fully constrain this model is yet undeveloped. Further lab work to verify and develop this model is still in progress, but for our purposes here the model as presented suffices as a means of comparison.

136

A3.3 Modified resistivity inverse solutions

Dataset Our dataset here is roughly the same as that used to derive our original inverse solution (Murphy & Egbert, 2017). However, we have omitted several sites along the northern and northeastern edges of our area of interest. Those sites are most sensitive to structures slightly beyond the study area, so they were poorly fit by our original inverse solution. We have also included several sites in the northern portion of our study area that were not available when we developed our original inverse solution.

Error floors We generally still adopt data error floors of 5% of the off-diagonal components for the impedance tensor (treating each row separately) and 0.03 for the vertical magnetic field transfer functions, as in our original work (Murphy & Egbert, 2017). However, because Pc3-4 source biases are common in data from the SEUS (e.g., Murphy & Egbert, 2018), we apply data error floors of 7.5% of the off-diagonal components for the impedance tensor (again, treating each row separately) and 0.075 for the vertical magnetic field transfer functions in the 10-100 s period band.

137

Figure A3.3.1: Inverse solutions obtained through the iterative resistivity redistribution process described in the main text. The left column contains the modified resistivity models; the right column shows the resulting inverse solutions based on those models. The top row shows the results from redistributing resistivity from Murphy & Egbert (2017) to a 120 km-thick thermal lithosphere (linear geotherm). The second row then shows the results from taking that resulting solution and limiting the resistivity values to a 150 km-thick thermal lithosphere (linear geotherm). Note that the inversion still made cells at ~200 km depth more resistive. The third row shows the results from redistributing the resulting resistivity solution to a 180 km-thick thermal lithosphere (linear geotherm). This inversion resulted in only subtle changes from the starting model. The last row shows the results of limiting resistivity values along a linear geotherm with a 200 km thick thermal lithosphere. These model slices follow the same section as that shown in the main text, although these cross sections are extended further into the ocean (see the upper-left panel of Figure A3.3.2 for the cross-section line). Each tick along the x-axis is 100 km.

138

Figure A3.3.2: Map views of the iterative re-inversion process. As in Figure A3.3.1, the left column contains the modified resistivity models and the right column shows the resulting inverse solutions based on those models. The top row shows the results from redistributing resistivity values from Murphy & Egbert (2017) to a 120 km-thick thermal lithosphere (linear geotherm); the second row, the results from taking that resulting solution and limiting resistivity values to a 150 km-thick thermal lithosphere (linear geotherm); the third row, the results from redistributing resistivity to a 180 km- thick thermal lithosphere (linear geotherm); and the last row, the results of limiting resistivity values along a linear geotherm with a 200 km thick thermal lithosphere. In each case, the depth slice is shown at the depth of the thermal LAB for that iteration. Note that the inversion consistently increases resistivity values in the inner (westernmost) Piedmont. In the final iteration, the increase in resistivity values at ~180 km is generally small across the Piedmont. Due to the coarse model discretization at 200 km, resistivity values at that depth are not representative of the value that would be observed at 1330°C (300 Ωm). In this case, the center of the model layer that goes through 200 km depth is actually at ~190 km, so the resistivity values imposed at that depth is higher than 300 Ωm.

139

Figure A3.3.3: Site-by-site misfits for each step in the iterative redistribution. The order of plots is as in Figures A3.3.1 and A3.3.2. The final iteration here uses a linear geotherm.

140

Figure A3.3.4: Map of site-by-site misfits for the refined resistivity distributions. The data misfits for the three refined resistivity inverse solutions are roughly the same and are all generally better than the original solution of Murphy & Egbert (2017). Parabolic geotherm refers to the temperature profile constructed from the mantle reduced heatflow calculation (second row in Fig. 3.6). MLD-informed refers to the nonlinear temperature profile informed by mantle discontinuity observations and near- surface thermal gradient observations (third row in Fig. 3.6).

141

Figure A3.3.5: Depth slices at 120 km through the refined resistivity solutions.

142

Figure A3.3.6: Locations of the sites used in the following data plots (Figs. A3.3.7-A3.3.10). These sites are generally the most sensitive to the structure of the Piedmont Resistor.

143

Figure A3.3.7: Data fit from site SCW55 for the three refined inverse solutions. Parabolic refers to the geotherm constructed with the mantle reduced heatflow calculations (second row in Fig. 3.6 in the main text). MLD refers to the geotherm informed by mantle discontinuity and near-surface geothermal gradient observations (third row in Fig. 3.6 in the main text).

144

Figure A3.3.8: Data fit from site SCW56 for the three refined inverse solutions. Parabolic refers to the geotherm constructed with the mantle reduced heatflow calculations (second row in Fig. 3.6 in the main text). MLD refers to the geotherm informed by mantle discontinuity and near-surface geothermal gradient observations (third row in Fig. 3.6 in the main text).

145

Figure A3.3.9: Data fit from site SCV56 for the three refined inverse solutions. Parabolic refers to the geotherm constructed with the mantle reduced heatflow calculations (second row in Fig. 3.6 in the main text). MLD refers to the geotherm informed by mantle discontinuity and near-surface geothermal gradient observations (third row in Fig. 3.6 in the main text).

146

Figure A3.3.10: Data fit from site REV55 for the three refined inverse solutions. Parabolic refers to the geotherm constructed with the mantle reduced heatflow calculations (second row in Fig. 3.6 in the main text). MLD refers to the geotherm informed by mantle discontinuity and near-surface geothermal gradient observations (third row in Fig. 3.6 in the main text).

147

A3.4 Geotherm calculations We use three different hypothetical geotherms to obtain our final resistivity solutions; our methods for deriving these geotherms are outlined below.

Linear Geotherm (1) For the simplest case, we take a geotherm with a constant temperature gradient from the surface to the lithosphere-asthenosphere boundary (LAB, 1330ºC isotherm). This temperature profile is shown in the first row of Fig. 3.6 in the main text. zLAB = 200 km TLAB = 1330ºC T0 = 0ºC dT/dz = (1330ºC - 0ºC) / (200 km - 0 km) = 6.65ºC/km (k = 3 W/m*K)

T(z) = 6.65z

Nonlinear geotherm informed by mantle reduced heatflow calculations (2) We use the traditional technique of plotting measured heatflow against measured heat production to derive a “reduced” heatflow from the mantle (Fig. A3.4.1). For this calculation, we use data from the SMU Geothermal Lab database (http://geothermal.smu.edu/; Fig. A3.4.2, Table A3.4.1). We then construct a parabolic geotherm with the 1330ºC isotherm at 200 km and the temperature gradient at the Moho set by the calculated value of reduced heatflow. This temperature profile is shown in the second row of Fig. 3.6 in the main text.

zmoho = 37.5 km (average for Piedmont) 2 qmoho = 28.7 mW/m dT/dz = qmoho / k = 9.5667ºC/km (k = 3 W/m*K) zLAB = 200 km TLAB = 1330ºC

T(z) = -0.02333z2 + 11.31667z

148

Figure A3.4.1: Mantle reduced heatflow calculation (left) yields a value of 28.7 mW/m2. The linear fit is derived using robust least squares with Huber weights. We use this value to construct our parabolic geotherm.

Figure A3.4.2: Locations of the heatflow measurements used in the mantle reduced heatflow calculation.

149

Thermal Thermal Heat Depth Heatflow Well Name State Lat Lon Conductivity Gradient Production (m) (mW/m2) (W/m*K) (°C/km) (μW/m3) GA-00012 GA 33.47815 -83.03408 205.5 3.33 19.3 64 4.4 GA-00011 GA 33.54215 -83.11108 199 3.38 19.1 64 4.7 NC-00019 NC 35.95139 -78.32893 192.2 2.96 16.9 50 2.8 NC-00006 NC 36.07088 -78.12052 209.7 3.15 19.3 61 2.3 NC-00025 NC 36.31178 -78.83304 209.2 3.36 12 40 1.4 NC-00022 NC 36.38678 -78.96645 246.8 3.81 10.8 41 1.7 NC-00023 NC 36.42528 -79.03105 209.3 3.67 11.1 41 1.7 NC-00024 NC 36.42758 -78.89474 194.9 3.48 10.3 36 1 NC-00017 NC 35.79109 -78.41743 204.7 3.01 17.8 54 2.49 NC-00020 NC 35.85479 -78.48143 209.8 2.9 17 49 2.3 SC-00012 SC 33.43685 -81.23652 346.9 3.29 21.3 70 2.7 SC-00011 SC 34.70062 -80.46410 187.5 3.1 12.9 40 3.1 SC-00008 SC 33.91984 -82.11915 285 3.84 17.6 68 5.2 SC-00013 SC 34.31343 -81.14482 316.7 3.38 18.3 62 4.2 SC-00010 SC 34.53892 -80.74731 384.3 3.08 14.9 46 2.3 SC-00009 SC 33.91683 -81.16652 170 3.44 13.5 46 2.96 SC-00007 SC 34.16683 -81.03312 245 2.72 16.7 46 2.96 VA-00021 VA 37.09556 -77.59320 210.1 3.28 17.3 57 3.3 VA-00022 VA 37.75305 -77.55020 201 3.19 18.4 59 3.7

Table A3.4.1: Heatflow measurements used for the reduced heatflow calculation. Data are from SMU Geothermal Lab database (http://geothermal.smu.edu/).

Nonlinear geotherm informed by lithospheric discontinuity measurements (3) In order to match MLD observations using the model of Karato et al. (2015), at the discontinuity depth we need temperature and pressure conditions that would cause the anelastic characteristic frequency (see section A3.2 above) to be roughly equivalent to the frequency of the converted body wave used to image the MLD. When those two frequencies are equal, the rate of velocity decrease with depth (and therefore with pressure/temperature) is roughly at its maximum and the specific depth associated with the given temperature/pressure conditions corresponds roughly to the MLD conversion depth. Utilizing Eq. A3.2.1, and assuming a seismic frequency of 0.1 Hz and a grain size of 5 mm, we find that the pressure-temperature conditions must satisfy 8.314 * ln[(2.3x1014)/(2π * 0.1 * 0.005)] = (350x103 + P*10-5)/T, where P is pressure and T is temperature. Receiver function images from the SEUS show an MLD in this region at ~70 km depth (Hopper & Fischer, 2018; Liu & Gao, 2018). The pressure at this depth would be ~2 GPa, so the temperature at this depth would need to be ~870ºC (~1140K) in order to produce an abrupt velocity reduction according to the model of Karato et al. (2015). We use the median value of surface geothermal gradient observations (Table A3.4.2, data from the SMU Geothermal Lab database) to constrain the behavior of the geotherm in the near surface. We assume a linear temperature profile in the mantle. We construct the

150 complete geotherm (shown in the third row in Fig. 3.6 in the main text) using a quadratic spline with input points given in Table A3.4.3.

Thermal Thermal Depth Heatflow Well Name State Lat Lon Conductivity Gradient (m) (mW/m2) (W/m*K) (°C/km) GA-00012 GA 33.47815 -83.03408 205.5 3.33 19.3 64 GA-00011 GA 33.54215 -83.11108 199 3.38 19.1 64 GA-BHT-000019 GA 32.47280 -82.74900 776.8 2.12 11.5 24 GA-00008 GA 33.00016 -83.99991 300 2.93 14.6 43 GA-00005 GA 33.00016 -83.99991 300 2.93 13.9 41 GA-BHT-000010 GA 32.29540 -83.45810 1838.4 2.11 22 46 GA-08001 GA 33.22820 -84.27230 213.4 15.5 GA-BHT-000016 GA 32.03470 -82.81020 1221 2.12 26.5 56 GA-BHT-000021 GA 32.45230 -82.58620 987.8 2.12 24.2 51 GA-00006 GA 32.71687 -83.24988 350 2.47 15.5 38 NC-00021 NC 36.03608 -77.75351 126.7 3.11 19.2 59 NC-00019 NC 35.95139 -78.32893 192.2 2.96 16.9 50 NC-00006 NC 36.07088 -78.12052 209.7 3.15 19.3 61 NC-00018 NC 35.72679 -78.32893 129.9 3.29 13.8 46 NC-00018 NC 35.72679 -78.32893 129.9 3.35 13.8 46 NC-00018 NC 35.72679 -78.32893 129.9 3.44 13.6 47 NC-00007 NC 35.28340 -80.88311 239 2.15 6.5 14 NC-00025 NC 36.31178 -78.83304 209.2 3.36 12 40 NC-00022 NC 36.38678 -78.96645 246.8 3.82 10.7 41 NC-00022 NC 36.38678 -78.96645 246.8 3.8 10.8 41 NC-00022 NC 36.38678 -78.96645 246.8 3.81 10.8 41 NC-00023 NC 36.42528 -79.03105 209.3 3.67 11.1 41 NC-00024 NC 36.42758 -78.89474 194.9 3.48 10.3 36 NC-00014 NC 34.57012 -78.93355 327.1 2.77 22.6 63 NC-00014 NC 34.57012 -78.93355 327.1 2.94 19.9 59 NC-00014 NC 34.57012 -78.93355 327.1 3.27 21.2 69 NC-00014 NC 34.57012 -78.93355 327.1 3.01 21.1 64 NC-00015 NC 35.68339 -78.93305 251 1.68 15.9 27 NC-00017 NC 35.79109 -78.41743 204.7 3.01 17.8 54 NC-00017 NC 35.79109 -78.41743 204.7 3.08 16.8 52 NC-00017 NC 35.79109 -78.41743 204.7 2.97 18.8 56 NC-00020 NC 35.85479 -78.48143 209.8 2.9 17 49 NC-00012 NC 36.41638 -77.91601 207.8 2.74 18.4 51 SC-00005 SC 33.28345 -81.66654 604 2.89 15 44 SC-00012 SC 33.43685 -81.23652 346.9 3.29 21.3 70 SC-00011 SC 34.70062 -80.46410 187.5 3.1 12.9 40 SC-00008 SC 33.91984 -82.11915 285 3.84 17.6 68 SC-00013 SC 34.31343 -81.14482 316.7 3.38 18.3 62 SC-00010 SC 34.53892 -80.74731 384.3 3.08 14.9 46 SC-00009 SC 33.91683 -81.16652 170 3.44 13.5 46 SC-00007 SC 34.16683 -81.03312 245 2.72 16.7 46 VA-00001 VA 36.86927 -77.90221 315 3.26 18 59 VA-00021 VA 37.09556 -77.59320 210.1 3.28 17.3 57 MEDIAN 3.09 16.8 48

Table A3.4.2: Heatflow measurements used to calculate median surface thermal gradient over the Piedmont. Data are from SMU Geothermal Lab database (http://geothermal.smu.edu/).

151

Depth (km) Temperature (°C) 0 0 1 17 5 85 70 870 85 923 100 976 150 153 200 1330

Table A3.4.3: Temperature constraint points used to construct the quadratic spline for our third geotherm. We use the median surface thermal gradient (~17°C/km) to constrain the linear upper- crustal portion of this geotherm. We assume a linear geotherm in the mantle lithosphere.

152

A3.5 Further resolution tests

Figure A3.5.1: Resolution test that shows the effect of extending thick (~200 km) thermal lithosphere beneath the Coastal Plain. In the plot on the upper left, red indicates increased data misfit; blue indicates decreased data misfit. As this plot shows, the increase in data misfit when extending deep high resistivity values beneath the Coastal Plain is very small. Therefore, although our data do not require thick thermal lithosphere beneath the Coastal Plain, they also do not preclude that possibility.