Neogene Development of the Terror Rift, Western Ross Sea, Antarctica

The Terror Rift is a 350 km‐long, 50–70 km‐wide, north‐trending deep basin in the western Ross Sea, adjacent to the Transantarctic Mountains. The Terror Rift lies within the broader Victoria Land Basin, which experienced 100 km of a mid‐Cenozoic extension. New fault and post‐29 Ma seismic stratigraphic interpretations were developed using all the seismic reflection data available from the Antarctic Seismic Data Library System and were correlated to all the scientific coreholes. A new 3D velocity model was used for depth conversions. Depth‐converted seismic profiles are used to image the faulting along the rift margins. Two segments of steep normal‐separation faults, which are connected by a broad anticline, border the eastern margin, whereas steep to sub vertical faults, combined with a stratal dip, produce the relief present at the western margin. The overall geometry of the Terror Rift shows an asymmetric half‐graben structure with a main bounding fault that dips alternatively eastward and westward in the southern and in the central parts, until it becomes a symmetric graben in the north. Both the eastern and western faults were active since at least 29 Ma in the southern and central Terror Rift and at least since 21 Ma in the north. This fault activity appears to have continued within the whole Terror Rift into post‐Miocene time, as suggested by the significant component of post‐Miocene vertical slip. The measurements of large sedimentary rock thicknesses changes over time between the rift shoulders and the deepest part of the basin and agree with the continuous faulting and relative subsidence in the southern Terror Rift that occurred between 29 and 13 Ma. These changes differ from several published papers that proposed that no significant tectonic activity occurred between 23 and 13 Ma. Despite the growth of the Terror Rift basin, the extension after 21 Ma was only ∼2–4 km across offshore mapped faults in the Terror Rift, and this minor extension agrees with a published plate tectonic study that suggests that the Terror Rift was a transtensional dextral transform boundary between 26 and 11 Ma. Therefore, the Terror Rift may have changed from an orthogonal rift into a transtensional transform boundary after 26 Ma, with the faulting after 11 Ma considered to be intraplate deformation.

The timing of the extension differs among basins within the Ross Sea part of the WARS. Extension occurred between 105 Ma and 80 Ma in the eastern Ross Sea (P. G. Fitzgerald & Baldwin, 1997;Lawver & Gahagan, 1994;Luyendyk et al., 1996;Siddoway et al., 2004). Sea floor spreading across what is now a deep bathymetric basin, which is aligned with the sedimentary basin Central Trough, has been proposed for ∼80-55 Ma ; D. S. Wilson & Luyendyk, 2009). This sea floor spreading is linked to the extension and crustal thinning in the Central Trough and in the Northern Basin. Sea floor spreading of about 180 km occurred across the Adare Trough between 43 and 26 Ma; this spreading affected the area that is now filled with sediments in the Northern Basin (Cande et al., 2000;Davey et al., , 2016Granot et al., 2010Granot et al., , 2013. The Victoria Land Basin formed at the same time, and it was stretched to 95-100 km of its total 140-160 km width ; F. J. Davey & De Santis, 2006). The Terror Rift is considered to be as a younger feature that formed during the Neogene inside the already existing Victoria Land Basin (Cooper et al., 1987a(Cooper et al., , 1987bF. J. Davey & Brancolini, 1995;C. R. Fielding et al., 2006C. R. Fielding et al., , 2008Henrys et al., 2007), with an initial E-W orthogonal crustal extension, across N-S faults (Cooper et al., 1987a;Hall et al., 2007).
The timing of the initiation of the focused extension and subsidence of the Terror Rift within the Victoria Land Basin is an open question, with two sets of conflicting interpretations and models. The first set of models proposes that the Terror Rift extension and tectonic subsidence started during the Middle Miocene time after a period of inactivity and passive thermal subsidence between 23 and 13 Ma (R. C. Fielding, 2018;Fielding et al., 2008;Hall et al., 2007;Henrys et al., 2007). These models are based on examinations of the Cape Roberts and ANDRILL cores and on interpretations of seismic reflection data (R. C. Fielding, 2018; SAULI ET AL.  Ryan et al., 2009). Fielding et al., 2008;Henrys et al., 2007). The other model proposes that tectonic motion across the western Ross Sea was continuous between 43 Ma and 11 Ma, with a change in the relative motion direction at 26 Ma (Cande et al., 2000;Granot & Dyment, 2018;Granot et al., 2013). This model is based on plate tectonic reconstructions that used marine magnetic anomalies and fracture zone orientations preserved in the deep ocean, including the Adare Trough. The calculated pole of rotation between East and West Antarctica from 26 to 11 Ma resulted in a total right-lateral displacement of 30 km across Terror Rift, and includes between 0 and 20 km of extension (Granot & Dyment, 2018).
Oblique rifting initiated from 50 to 40 Ma has been proposed for the westernmost Ross Sea and the adjacent Transantarctic Mountains, based on geological field work and incorporating the timing of denudation (Rossetti et al., 2006;Storti et al, 2008;T. J. Wilson, 1995). Alternatively, a mid-Cenozoic (∼30 Ma) change from the Paleogene E-W extension to Neogene oblique rifting, including strike-slip faulting, was interpreted (Hamilton et al., 2001;Salvini & Storti, 1999;Salvini et al., 1997). The Neogene oblique rifting hypotheses are based in part on interpretation of seismic reflection data that were correlated with scientific core holes. The Salvini et al. (1997) model proposed that long NW-SE right-lateral faults to cut from North Victoria Land ( Figure 1) across the Northern Basin and Adare Trough to the central and eastern Ross Sea. This model was disproved by additional newer seismic reflection data, and by the lack of offsets in the N-S oriented marine magnetic anomalies, some of which are continuous between the Adare Trough and Northern Basin (F. J. Davey et al., 2016;Granot et al., 2010).
We present a comprehensive study of the geometry of Terror Rift in the Ross Sea, and its evolution through late Cenozoic time. This study is based on interpretations of faults and of stratigraphy as imaged on all the available seismic reflection data and sampled by scientific core holes. The rift is studied in detail over its full 350 km length north of the ice shelf edge. We outline the Terror Rift geographically as the deep sedimentary basin that formed in the western part of the Victoria Land Basin, and temporally by post-dating the de-activation of the eastern Victoria Land Basin rifting, which occurred sometime before the formation of a 26-30+ Ma unconformity (RSU6, Figure 3).

Data Set and Methods
This study utilized seismic reflection data, bathymetry, and stratigraphic data from deep scientific core drilling ( Figure 2). The seismic reflection data set includes all the multichannel seismic reflection data (MCS) available from the SCAR Antarctic Seismic Data Library System for Cooperative Research (https://sdls.ogs. trieste.it/; Childs et al., 1994) for the study area. In addition, the following single channel seismic reflection profiles (SCS) were integrated: Italian data acquired in 2001/2002(Sauli et al., 2014 and United States of America data collected in 1990 (Anderson & Bartek, 1992;Bartek et al., 1996) and 2003(Bart, 2004. The water is deep enough over much of the study area that the SCS images the mapped faults and the stratigraphic reflectors above the first water bottom multiple. Additionally, where primary reflections dip and the shallow multiple reflections are flat, stratigraphy can often be interpreted through the multiple reflections even in SCS data. Faults in the Terror Rift were interpreted on all the seismic reflection data shown in Figure 2. These faults were correlated between reflection profiles and were digitally 3D viewed ( Figure 5). Faults were drawn on fifteen depth-converted seismic reflection profiles ( Figure S4). How the faults were correlated from one profile to the next is explained in the Supplemental text.
Velocity information for travel time to depth conversions were obtained by converting into interval velocities the stacking velocities of sub-horizontal reflections from processing of MCS data available on ANTOSTRAT (Brancolini et al., 1995a). Stacking velocities were converted into interval velocities using Dix's equation (Dix, 1995). Additional interval velocity information was obtained from sonobuoy refraction seismic data, retrieved from original publications and reports (Cochrane et al., 1992;F. J. Davey et al., 1982;Houtz & Davey, 1973;Sato et al., 1984) (for more information see supplemental data). We created 250 depth-time charts at sonobuoy locations and at selected stacking velocity points, which were located from just east of 180° longitude through the western shore of the Ross Sea ( Figure 2). The depth-time charts were entered into the software IHS Kingdom Suite software. This software calculated the average velocities for each unconformity grid at each depth-time chart. Gridding of the average velocities between control points produced average velocity grids from the sea surface to each digital unconformity grid. These were used to convert the travel time grids into depth. Time and depth grids were paired, including sea surface, sea floor, and subbottom grids, and a 3D velocity model was created in the IHS Markit Kingdom Suite. This 3D velocity model was applied to the seismic reflection profiles shown in Figures 6 and S4, and a depth version of each profile was created in the IHS Markit Kingdom Suite. The same velocity model was used to convert faults into depth. These depth faults were then displayed on the depth seismic profiles.
We analyzed how vertical motion was accommodated along 300 km of the Terror Rift. The components of the relative vertical relief between the western edge of our data and the deepest point of the basin, and between that basin and the highest part of the eastern rift shoulder, were measured on 15 depth-converted profiles, which are presented in the supplemental data ( Figure S4). These measurements were summed for each flank of the rift, for each depth section, for both faulting and dip. SAULI ET AL.

Stratigraphy
Our seismic stratigraphic correlations are part of a broader effort to interpret the Oligocene through middle Miocene reflectors by using all the available seismic reflection data from most of the Ross Sea (the area of Figure 4; C. C. Sorlien et al., 2016Sorlien et al., , 2019. This work continued and includes seismic stratigraphic correlations throughout the Victoria Land basin, which included the Terror Rift. The starting points were the ANTOSTRAT (Brancolini et al., 1995a(Brancolini et al., , 1995b interpretation at Deep Sea Drilling Program sites 270, 272, and 273. The ANTOSTRAT (Brancolini et al., 1995a(Brancolini et al., , 1995b stratigraphy contains eight seismic sequences that overly the Mesozoic acoustic basement and are bounded by seven unconformities (e.g., RSU6 -RSU1) with ages ranging from Late Oligocene (RSU6) to Lower Pleistocene.
The C. Sorlien et al. (2019) RSU5 unconformity is 570 m below the bottom of DSDP site 273 in Central Trough, and at this location, it is nearly the same as the ANTOSTRAT (Brancolini et al., 1995a(Brancolini et al., , 1995b interpretation in two way time but is significantly deeper in depth. A revision of the depth and age of the RSU5 unconformity is ongoing after the recent IODP Expedition 374, which drilled site U1521; this site is located near the western edge of the Eastern Basin (Figures 1 and 4) and that reached sediments dated ca. 18 Ma near the bottom of the hole (McKay et al., 2019). Therefore, we used the names RSU5B_vlb (19 Ma) and RSU5C_vlb (21 Ma) labels in the Victoria Land Basin (extension "_vlb" stands for Victoria Land Basin). The ages of these reflectors are the same as the ages of the RSU4A and RSU5 unconformities from Brancoliniet al. (1995aBrancoliniet al. ( , 1995b. In this paper we correlated different reflectors by using the published seismic stratigraphy nomenclature and ages assigned ( Figure 3): (a) The age of the unconformity RSU6 the age is > 26-30+ Ma on the basis of the sediments ages at DSDP 270  in the eastern Ross Sea, although this unconformity cannot be traced directly across the Ross Sea and its age may be diachronous from east to west. A potentially equivalent reflector labeled with age 29 Ma in the southern Terror Rift is correlated from the Cape Roberts core holes (Cape Roberts Science Team, 1998  (21 Ma) of Brancolini et al. (1995aBrancolini et al. ( , 1995b and renamed them RSU5C_vlb and RSU5B_vlb respectively (d) Reflector Wb (well bottom of DSDP 273 of Sauli et al., 2014) is dated to ∼18 Ma (e) In the southern Terror Rift the reflector at 29 Ma is from Hamilton et al. (2001) and R. C. Fielding (2018).
The reflector dated to ∼21 Ma is from Fielding and Thompson (1999)  Terror Rift for 100 km northward. These unconformities were paired with the RSU unconformities from the eastern flank of the Terror Rift, based on their similar ages. The seismic stratigraphic correlation from the eastern rift shoulder into the deep basin was determined where the vertical component across faults was the least. The unconformities were correlated along profiles that were more parallel to the rift trend in the deep basin and on eastern rift shoulder. This loop-tying (supporting information and Figure S1) procedure allowed our stratigraphic interpretation across faults where the vertical components are greatest.
The Rg (∼13 Ma) and Rh (7.6 Ma) unconformities were correlated from over-ice seismic reflection profiles in the McMurdo Sound area, as discussed in Pekar et al. (2013) and Wenman et al. (2020), northward within the basin, and to the eastern shoulder of the central and northern rift. The Ri unconformity (∼4.6 Ma) could not be directly correlated north from the area near ANDRILL 2A, so we interpreted a regional unconformity farther north as Ri (Figure 3). This interpretation is the same as that of Fielding et al. (2008) for seismic line IT90-75.
The results of the scientific coring provided an age control for the reflectors that were correlated by use of seismic reflection data. In the following paragraphs the regional-scale deformations of the sedimentary rocks from differential subsidence, faulting, and folding are presented. The thicknesses of sedimentary rocks intervals are related to sediment availability and to the accommodation space produced by spatially variable subsidence. The thickness of the middle Miocene through lower Pliocene rocks is limited by the significant erosion on the rift shoulders as well as in the northern rift basin.
The upper Oligocene-lower Miocene section between RSU6 and RSU5C_vlb (29 Ma-21 Ma) is interpreted to be ∼3 km thick in the center of the Terror Rift (extrapolated on profiles in Figures 5, 6b, 6f, and 6h). The late Oligocene and older strata thicken toward the Terror Rift from the Coulman High westward, and from the western shoulder eastward (Figures 6b, 6f, 6h, and 8). These strata clearly dip more steeply than the overlying Miocene sediments on the western rift flank. The reflectors diverge in a fan pattern from the western edge of our interpretations toward the basin center. This fan pattern is also observed between the eastern shoulder of the central rift and the basin, with reflectors diverging westward (Figures 6e and 6f). This pattern suggests a continuous subsidence across the Terror Rift, with the basin center subsiding most rapidly and with subsidence rates decreasing both east and west toward the rift shoulders. In addition to the E-W variation in sedimentary thickness, there is a N-S variation. The late Oligocene and younger section is thickest near Ross Island (  Figure S4l. (h) Profile O-O′ is displayed in depth and is located in Figures 2 and 5. The ridge at the western end of the figure is near the Cape Roberts scientific core holes (Figures 1 and 2). The Terror Rift basin at this latitude is an asymmetric half-graben, and the structural relief is tilted in the west and is formed by a very large newly interpreted fault in the east. The 21 Ma horizon has a vertical separation of nearly 4 km across the eastern fault from the RSU5C_vlb horizon (21 Ma). Rh and Ri are also offset by the fault, which requires post-Miocene activity. The phrase "change in thickness" indicates where the measurements of thickness changes shown in the graph of Figure 11 were made.
A part of the lower and middle Miocene section, namely, RSU5C_vlb-Rg (21 to ∼13 Ma) dips more gently than the sedimentary section below within parts of the Terror Rift (Figures 6b, 6f, and 6h). This interval has a uniform thickness in the VLB east of the Terror Rift (Figure 8). The strata of this section are tilted and eroded on the western rift flank (Figures 6b,  6d, 6f, and 6h) and are tilted and significantly faulted on the eastern side, west of the Lee Arch ( Figure 6f) and across the mid-basin antiform (Figures 6d and 6b).
The middle and upper Miocene sections, which are delimited by the regional unconformities Rg-Rh (∼13-7.6 Ma), are significantly reduced in thickness in the northern depocenter with respect to the southern and central depocenter (Figures 6a, 6b, 6d, and 6h). This interval is thin or completely eroded away at or immediately below the sea floor along the eastern rift shoulder, including at the crest of Lee Arch (Figures 6b, 6d-6f, and 6h). This interval is also mostly or completely eroded away within the northern deep basin, especially on the Mid-Basin antiform (Figures 6a  and 6b). The erosion of the Miocene to Early Pliocene sedimentary section at the northern edge of the Terror Rift, might have been caused by repeated ice-sheet oscillations over the past 13 Ma that were reported at the AND-2A (McKay et al., 2009;Naish et al., 2009).
In the northern Terror Rift the inferred Ri unconformity (∼4.6 Ma) is interpreted as a subhorizontal erosional unconformity within the basin and on the eastern rift shoulder where it is not eroded away. The vertical component of the offset of Ri by faults is between 500 m and 1 km (Figures 6a-6c). Many faults offset Ri throughout the Terror Rift (Figure 6), and it is also involved in tilting (Figures 6e and 6h).

Faulting
On the eastern flank of the Terror Rift we interpreted different sets of faults with NNW-SSE and N-S striking directions. The NNW-SSE striking faults (i.e., the green fault set at the northern and southern rift ends in Figure 5 and segments 1, 3, and 4 in Figure 9) consists of down-tobasin normal-separation faults. The N-S striking faults (green and pink fault sets in Figure 5, segment 2 in Figure 9) are normally separated with both down-to-basin and up-to-basin stratigraphic separation, dissect the SAULI ET AL.  On the western flank of the Terror Rift, there is an NNW-SSE striking set (i.e., the red fault set in Figures 5 and 9) of steep to sub vertical faults, which locally exhibit reverse separation (Figure 6g). The largest of these down-to-basin faults delimits the deep rift basin center and dissect the mid-basin antiform in the north (Figures 6b and 6d The overall geometry of the Terror Rift graben varies spatially from south to north. The rift displays a half graben structure in the south with the main bounding fault located on the east side (Figure 6h). Part of the central rift includes a half graben with the main bounding fault located on the west side ( Figure 6f). Northward the graben becomes symmetric with a broad flatter basin center (segments 3 and 4 in Figures 6b, 6d and 9).
Some faults exhibit reverse separation of strata, while others are subvertical (Figures 6b, 6d-6h). Most nonvertical faults exhibit normal separation for all pre-Pliocene stratigraphic horizons, with larger separations for older horizons than for younger horizons. There are local fault strands with reverse separation that strike parallel to those with normal separation (Figure 6g).
The depth-converted MCS profiles that cross the Terror Rift are presented from north to south (Figures 6a-6h). These profiles show how the vertical relief was accommodated on the rift flanks. On the northeastern border, a change in vertical relief accommodation from tilting to faulting occurs within a 30 km distance (Figure 6a). Farther south, along the eastern rift flank, the accommodation is dominated by a series of seven faults ( Figure 6b) and changes southward, where it is almost entirely dominated by tilting, with most of the faults being up-to-basin (Figures 6e and  6f). On the southeastern border, nearly all of the relief occurs across one narrow fault zone that delimits a half graben (Figure 6h).
The vertical relief between the Terror Rift margin and its axial basin is taken up by a combination of tilting and faulting along its western flank and by a spatial alternation between faulting and tilting along strike on its eastern flank, which forms discrete segments (Figure 9). Where the vertical relief between the rift shoulder and basin is mainly due to faulting, the stratigraphic separation, as expected, is down-to-basin (profile D-D′, Figure 6b). However, on those segments where the down-to-basin relief is due to tilting, there are zones of up-to-basin faults that dip away from the middle of the basin and toward the rift shoulder (profiles F-F′, G-G′, J-J′, Figures 6d-6f). These faults mostly strike north and link the northern and southern eastern rift border faults (Figures 5, 7, and 9). Unconformities can be correlated in the seismic reflection data across the tilted boundaries and then correlated on both sides of the faults to demonstrate the vertical separation across the fault boundaries.

Discussion
In the following paragraphs we discuss the results of our comprehensive interpretation with their geological implications and in light of previous published models. This discussion includes geometry, stratigraphic architecture, and kinematics over time for all the 350 km of the portion of the Terror Rift north of the Ross Ice SAULI ET AL.

10.1029/2020GC009076
12 of 20 The rift trends N-S for most of its length but trends NNW-SSE in the south and dies out on the north near the NW-SE faults, which were interpreted as right-lateral by Rossetti et al. (2006). The "AZ" symbols indicate the eastern parts of the accommodation zones of. The orange 1, 2, 3, and 4 numbers refer to the segmentation of the eastern rift boundary (see the text for more explanation). The yellow fault is the Transantarctic Mountains (TAM) Border Fault modified from Olivetti et al. (2018). It is only inferred along most of its length, but is imaged in small areas on the south of the figure.
Shelf edge. We briefly consider the evidence for Plio-Quaternary tectonic activity in the study area and present our results for measured post-21 Ma and post-13 Ma crustal extension in the Terror Rift.

Geometry of the Terror Rift
How the vertical relief of reflectors is accommodated is different in the west versus the east, and it varies in four segments along the eastern flank. The vertical relief across the western flank of the Terror rift is due to both a stratal dip and faulting, without distinct segmentation along the trend (Figures 9 and 10). The vertical relief due to faulting ranges between 1 and 2 km, as measured for the 21 Ma horizon (Figure 10, solid red line), while the vertical relief due to dip increases from near 2 km in the north to 6 km in the south (Figure 10, solid blue line). This does not include any possible additional relief that is present west of the available offshore seismic reflection data along the front between the Transantarctic Mountains and offshore basin (Figures 7 and 9). It can be hypothesized that the Victoria Land Basin and the deeper inner Terror Rift share their western rift boundary near the current coastline along both interpreted and inferred faults (Ferraccioli et al., 2009;Hamilton et al., 2001;Pekar et al., 2013).
The eastern flank of the Terror Rift is partitioned into four structural domains, which alternate between accommodating the vertical relief into the basin by either tilting or faulting (Figures 9 and 10). The fault-controlled segments (numbered 1 and 3 in Figure 9) are linked by a segment where the vertical relief is due to dip (Figures 5, 9, and 10). This dip segment (numbered 2 in Figure 9) is cut by N-striking faults with a vertical relief up-to-basin. This segment also includes southward diverging down-to-basin faults east of the crest of the Lee Arch (Figures 6d-6f). The segment is separated from the northern fault segment (numbered 3) by a discontinuity interpreted as a WNW-striking accommodation zone by Hall et al. (2007) (Figure 9). In their interpretation of the northern accommodation zone, which coincides with ours, Hall et al. (2007) included an ∼23 km right step or offset of fault zones, anticline axes, and the deep basin center across this zone. Instead of interpreting this lateral offset across the accommodation zone, we map a double right bend of ∼7 km (Figures 5 and 9). This broad double bend is dominated by down-to-the east faults that we map as continuous across the Hall et al., (2007) accommodation zone. There are no data to constrain the deep basin at the accommodation zone. Simply gridding the interpreted 21 Ma horizon produces no offset of the deepest basin there (Figures 5 and 9).
The anticline zone (Lee Arch), on the eastern border of the Terror Rift, is dissected by N-S-striking upto-basin and down-to-basin steep to subvertical faults (Figures 6d-6f). This kind of geometry can form during an orthogonal extension, such as a rollover fold in an extensional basin (Withjack et al., 1995;Xiao & Suppe, 1992). Nevertheless this geometry is also similar to a negative flower structure, where faults with opposite dips converge downward (e.g., Sylvester, 1988), and are often interpreted to be diagnostic of strikeslip motion (Lowell, 1972;Ramsay & Huber, 1987;Sylvester 1988). The relation of faulting to the northern part of Lee Arch anticline is seen on F-F′ profile (Figure 6d). The western limb of the anticline can be explained as due to the more rapid subsidence of the deepest part of the Terror Rift that resulted from crustal thinning, thermal contraction, and sediment loading. The eastern limb exhibits back tilting into the imaged normal-separation faults. These faults may or may not be bookshelf faults, rotating about a horizontal axis, above an inferred gently-dipping detachment below (Mukherjee & Khonsari, 2018). Such a detachment is not imaged, but gently-dipping faults are expected in a location that extended for 100 km (Fossen et al., 2000). The same mechanism could explain the similar anticline as imaged on G-G′ (Figure 6e), where the back-tilting into west dipping faults is less obvious.
The west dipping fault on the southernmost O-O′ profile shows nearly 4 km of normal vertical separation of the ∼21 Ma unconformity in the half-graben basin, to the RSU5C_vlb reflector (21 Ma) on the east footwall of that fault (Figure 6h). We advocate that the down-to-east tilting could partly be back tilting into this SAULI ET AL.

10.1029/2020GC009076
13 of 20 Figure 10. Components of Terror Rift relief due to faulting and due to dip for the eastern margin (dashed) and western margin (solid). The positive dip on the eastern side is to the west, while the positive dip on the western side is to the east. The eastern margin has four segments, where the relief alternates between faulting and dip. The western margin relief is due to both faulting and to dip, except in the south, where it is mostly due to dip. Relief on the western margin does not include any proposed fault system between the Transantarctic Mountains and the Ross Sea. major fault (Figure 12). Normal slip on a fault interpreted as listric by R. C. Fielding (2018) is expected to produce such back tilts. In the southern Terror Rift, many profiles, such as O-O′, end near this fault, and our interpretation of the stratigraphy across the fault is limited. Thus, we do not directly know the timing of the middle Miocene and younger activity across this fault or of the pre-21 Ma activity across it. However, the continuous back tilting started before RSU6 (29 Ma) and continued after Rh (7.6 Ma), which suggests continuous activity on the fault. We tentatively interpret ∼1 km of vertical separation for Rh (7.6 Ma) across this fault. Similarly, Ri is truncated by this fault, and there is a minimum of a half kilometer of vertical separation across it. The fault reaches close to the sea floor and may have even been active during the Quaternary. The large vertical component of offset component and its association with the tilting of the half-graben does not preclude a similar or larger right-lateral component of motion across this fault.
Vertical faults, and steep faults with reverse separation are imaged in the northern and southern parts of the Terror Rift (west side of Figures 6b-6d, 6f, and 6g). Such a geometry suggests that they are strike-slip faults. The most conclusive evidence that supports the strike-slip motion comes from the quantitative analysis of marine magnetic anomalies by Granot and Dyment (2018), which results in an estimate of 30 km of transtensional right-lateral motion, and between 0 and 20 km of extension between 26 and 11 Ma across the Victoria Land Basin (Granot & Dyment, 2018). We have not interpreted significant Miocene activity or tilting across the eastern Victoria Land Basin (Figure 8), so the displacement was probably accommodated within the Terror Rift.
Some global strike-slip fault zones have vertical components that do not increase consistently downwards into older sedimentary rocks. These strike-slip faults can even vary within a short distance from normal to reverse separation with switching of the side that is downthrown, such as for the Palos Verdes fault near Los Angeles (Brankman & Shaw, 2009; C. C. Sorlien et al., 2013), and the Newport-Inglewood fault also near Los Angeles (C. C. Sorlien et al., 2015). In contrast, we interpret that even the subvertical fault strands have the same sense of vertical stratigraphic separation along their full lengths, and this vertical separation increases with depth and thus ages at any one location on a fault strand. While these observations do not support strike-slip motion, they also do not rule it out.

Stratigraphic Architecture and Passive 23-13 Ma Thermal Subsidence versus Ongoing Faulting and Tectonic Subsidence
Several publications have proposed that in the southern Victoria Land Basin, the sedimentary fill recorded passive thermal subsidence between 23 and 13 Ma, followed by renewed rifting activity after 13 Ma (R. C. Fielding 2018;Fielding et al., 2008;Henrys et al., 2007). Our results do not support the concept of an early Miocene interval of rift inactivity. We infer a continuous tilting over time between ∼29 Ma and present in the southern Terror Rift and between 21 and 17 Ma in the central Terror Rift (Figure 11). This continuous tilting activity exhibits a segmented distribution on the eastern rift flank, and a spatial variation when moving northward on the west rift flank (Figures 9 and 10).
We cannot be confident of the possible differential vertical motion between 29 and 21 Ma across the central and northern Terror Rift because RSU6 (29 Ma) is poorly imaged to nonimaged in the deep basin and is mostly extrapolated below the younger horizons. However, the tilting between RSU6 (29 Ma) and RSU5C_ vlb (21 Ma) or the 21 Ma horizons on the western part of the D-D′ and J-J′ profiles (Figures 6b and 6f) implies that differential vertical motion by tilting and probably by faulting also occurred on this flank of the central and northern Terror Rift. The thickness between RSU5C_vlb (21 Ma) and Wb (18 Ma) is much greater in the basin than on the rift shoulders on both the eastern and western margins, as displayed in the B-B′ and D-D′ profiles (Figures 6a and 6b). This dramatic decrease in thickness across the eastern rift SAULI ET AL. border faults between 21 Ma and 18 Ma implies fault activity during this time interval (Figures 6a and 6b). Similarly, the changes in thickness due to tilting for this same age interval are evident in the F-F′, G-G′, and J-J′ profiles (Figures 6d-6f) in the northern and central Terror Rift. Additionally, the thickness between ∼18 Ma or ∼17 Ma and Rg (13 Ma) is greater in the basin than on both rift shoulders in the northern and southern regions, as observed in the F-F′ and O-O′ profiles (Figures 6d and 6h). This thickness increase is much less dramatic than the thickness change for the interval below, in part because of the removal of part of this interval by later erosion (Profiles B-B′, D-D′ in Figures 6a and 6b).
In conclusion, due to the evidence of tilting and fault activity in the southern Terror Rift between 29 and 21 Ma, in the northern Terror Rift between 21 Ma and 18 Ma, and in southwestern and central Terror Rift continuously since 29 Ma, it seems likely that the entire rift was active during these time intervals.
The graph in Figure 11 shows the differences in the current compacted sedimentary rock thicknesses between locations on the rift shoulder and in the deepest part of the sedimentary basin over time. The measurements at the eastern and western boundaries of the central and northern rift were made on profiles J-J′ and D-D′ (Figures 6f and 6b), where the location of the measurements are indicated). The thickness changes of sedimentary rocks in the southern rift were measured only on the western flank of the profile O-O′ (Figure 6h) with the reported location of the measurements). Sedimentary rocks for a given age interval were compacted more by deeper burial in the deepest basin than on the flanks of the rift. Therefore, our measures of the thickness changes of these sedimentary rocks underestimate the differences in the rates of sediment accumulation between the deep basin and the rift flanks. These thickness changes across faults provide information on the rate of the vertical slip components throughout the analyzed time intervals and can be considered in light of the interplay between slip rates, the regional subsidence and the sedimentation rate (Doglioni et al., 1997).
The resulting graph shows a high rock thickness change rate in the southern Terror Rift between 29 and 13 Ma, with the slowest rate being between 13 and 7.5 Ma. The central rift has a large change in rock thickness between 21 and 17 Ma, a low rate between 17 and 13 Ma, and a higher rate between 13 Ma and the present. This is true for both the eastern flank and part of the western flank and include a large fault. Thickness changes in the northern rift are high between 21 and 18 Ma and then are very small until 4.6 Ma. The rate SAULI ET AL.  Figure 6h, on the right. The changes in sediment thickness across the fault show that there was growth on the fault and a faster vertical relief from 29 to 13 Ma than from 13 to 0 Ma (see also Figure 11). increases after 4.6 Ma on the eastern flank. In summary, thickness changes decreases after 17 or 18 Ma in the central and northern rift, while thickness changes decrease between 13 and 7.5 Ma in the southern rift.
These thickness changes are related to both the vertical motions of faults and to tilting and hence to the tectonic activity that affected the Terror Rift mostly between 29 and 13 Ma or between 21 and 18 Ma in the southern and northern sectors. Decreased tectonic activity appeared to occur, at least for the southern Terror Rift, when the renewed Terror rifting phase had been proposed by many authors (e.g., Fielding et al., 2008). The two hiatuses in thickness change for the periods of 18-4.6 Ma and 17-13 Ma and, in particular, the hiatus in the northern Terror Rift may have resulted from the influence on sedimentation of glacial discharge/ erosion during the periodic Miocene oscillations of the David Glacier, which streamed preferentially within the northern part of the Terror Rift. Moreover, the greater thickness change rate after 4.6 Ma may in part be related to postglacial isostatic rebound at the rift flanks.
The depositional history of the southern Terror Rift after 29 Ma to the present is reconstructed during continuous rifting along the depth profile O-O′ (Figure 6h), which is comparable, for the location, to the R. C. Fielding (2018) diagram ( Figure 12). His reconstruction was simplified, and no decompaction of the sediments was considered. In our depth profile, the tilt rate is constant, with no major changes after 29 Ma. The figure clearly displays the continuity of subsidence and faulting, which implies continuous synsedimentary deposition, with sediment strata that thicken basinward and with no significant major changes in the geometry of the basin infill over time.
Some portions of the subsidence that are associated with tilting result from differential thermal subsidence and sediment loading. However, without full basin modeling, it is not possible to guess the percentage of tilting due to this process versus that due to active extension. Such modeling would include the removal of both the sedimentary and water loads (backstripping), thermal subsidence related to the cessation of the rapid 43-26 Ma extension, and isostatic correction using a realistic elastic thickness of the Terror Rift crust if flexure is corrected for (Allen & Allen, 1990;Steckler & Watts, 1978;Watts & Ryan, 1976). Alternatively, Airy (local) isostatic correction would be performed. If the subsidence between, for example, 23 and 13 Ma can be fully explained by such modeling, then extension and crustal thinning are not required.

Evidence for 11 Ma to Present Tectonic Activity
Fault and tilt activity continued after 11 Ma, which is the age that Granot and Dyment (2018) proposed that East and West Antarctica became one plate. Published seismic reflection data have imaged late Miocene and younger sedimentary rocks being offset and deformed by faults that possibly reach the seafloor (Fielding et al., 2008;Hall et al., 2007;Hamilton et al., 2001;Henrys et al., 2007).
We present profiles E-E′ and D-D′ (Figures 6c and 6b, respectively), which show 800 m of vertical separation for unconformity Ri (∼4.6 Ma) across two faults that have sea floor scarps or fault-line scarps. There are several hundred meters of additional vertical separation across the adjacent four large faults of Ri (Figure 6b). These profiles are located in the northern Terror Rift and we acknowledge that some of the fault movement may be related to glacial isostatic adjustment after late Pleistocene deglaciation (Hampel et al., 2010). The grounded David Glacier extended in the northeastern Terror Rift, and the stresses released since the Last Glacial Maximum during the postglacial unloading and rebound processes may be responsible for some of the fault movement. However, there have been numerous advances and retreats of this glacier and of the ice sheet during and after Middle Miocene time. These faults exhibit an increasing vertical offset component of the older rocks, which cannot be easily explained by glacial isostatic adjustment.
Our overall interpretation suggests the presence of Plio-Quaternary and possibly Holocene fault activity and is consistent with other geological evidence, such as volcanism within the Terror Rift and on its flanks (Giggenbach et al., 1973;P. R. Kyle 1990; P. R. Kyle et al., 1992;LeMasurier, 1990;Rilling et al., 2007;Rowe et al., 2000), and the high heat flow measurements (Blackman et al., 1987;Della Vedova et al., 1992.

Extension Across Terror Rift Since 21 Ma
To evaluate the crustal extension across the northern and southern Terror Rift since 21 Ma, the horizontal offset components of the faulted 21 Ma horizon were measured along the seismic depth profiles across the rift. These measurements were carried out by summing the heaves across the faults along the profiles. The profiles are close to being perpendicular to the strikes of most of the faults, so the apparent extension due to faulting is sufficiently close to the true extension.
The measured extension of the RSU5C_vlb (21 Ma) horizon across the D-D′ profile in the northern Terror Rift (Figures 6b and S5b) indicates 2.0 km of extension, while the post-21 Ma extension across L-L′ in the southern Terror Rift ( Figure S5a) is 3.6 km; additionally, an unknown amount of extension was accommodated by diking beneath a volcanic feature located east of L-L′ and by an unknown amount of extension across the TAM Border fault(s).
Our interpretation indicates only a few km of post-21 Ma extension and indicates a significant change in the Terror Rift dynamics from the earlier nearly orthogonal extension to an oblique extension or transtensional motion (Granot & Dyment, 2018;Granot et al., 2013).

Summary and Conclusions
We present the geometry of the entire 400 km length of the Terror Rift, which is located north of the ice shelf edge. The stratigraphy dated at core holes, which was correlated through grids of seismic reflection profiles, allowed the sedimentation and relative vertical motion over the last 29 million years to be studied. Published plate tectonic studies combined with our interpretation of subvertical faults suggest that the Terror Rift was spatially focused within an older broader basin at ∼26 Ma, as oblique rifting and transtensional strike-slip faulting replaced earlier nearly orthogonal rifting. The vertical relief was accommodated differently on the two opposite rift flanks and was spatially different among the four segments along the eastern flank. The entire western side of the Terror Rift, which is along the Transantarctic Mountain front, was formed by a combination of both faulting and tilting, with tilting forming most of the vertical relief in the south. The eastern side is segmented into four segments that exhibit alternate faulting and tilting. A broad anticline in the southern tilting segment is dissected by steep to subvertical faults that converge downward and form a strike-slip flower structure geometry. The geometry of the Terror Rift varies spatially from a southern and central half-graben structure, which dips down to the east or down to the west toward the northern symmetric graben with a broad flatter basin center. In contrast to the 23-13 Ma tectonic quiescent phase that was proposed in several publications, we conclude that rifting activity, with ongoing faulting and subsidence, was continuous from at least the late Oligocene through the Miocene and Pliocene and possibly to the Quaternary-present. The new comprehensive geometry and newly proposed temporal continuity of the rifting activity fit with a new model for Terror Rift emplacement in which an early east-west extensional episode successively evolved into dextral transtensional motion or oblique rifting at ∼26 Ma (Granot & Dyment, 2018). The results of this work will also influence broader studies on the evolution of the development of the ice sheet and coastal glacial systems across the TAM during the Cenozoic. Improved knowledge of the tectonic framework will help to define the boundary conditions for glacioisostatic modeling and paleosea floor depth reconstructions, which are crucial for estimating the sensitivity of the ice sheet to past and future climate changes.

Data Availability Statement
The seismic stratigraphic interpretation for much of Ross Sea has been provided to the US Antarctic Program Data Center at http://www.usap-dc.org/view/dataset/601098. Information about the grids is given at: https://www.usap-dc.org/readme/601098.