Modern shelf bathymetry bordering the Gulf of Alaska exhibits shelf-crossing sea valleys that suggest focused pathways for ice flow during glacial conditions. Using an integrated seismic data set between the present Yakutat and Alsek Sea Valleys, we investigate the glacial stratigraphic record in order to improve our understanding of the regional glacial system during maximum glacial conditions. Our investigations reveal four glacial unconformities, of which, the latter two are overlain by sediment packages hundreds of meters thick. We suggest that these unconformities are indicative of ice advance phases during the Little Ice Age (LIA), the Last Glacial Maxima (LGM), and two pre-LGM advances with glacial retreat sequences preserved from the youngest two. The advances were dominated by ice expanse from the Malaspina Glacial system and Alsek River districts rather than the Hubbard Glacial system and only show distinctive morainal bank development near the shelf edge and near the mouth of, or within, modern bays, fjords, or river valleys. This observation strongly supports rapid and continuous retreat from glacial maxima conditions during climatic warming. All, except the inferred LIA sequence, transgress the shelf and exhibit concentrated erosion in overdeepened troughs, analogous to cross-shelf troughs similar to those observed on other high-latitude glaciated shelves. The two Alaskan troughs discussed here may be end-member examples of high-sediment flux systems due to the temperate glacial setting combined with an actively exhuming orogen.
 One important tool in determining glacial advance-retreat history on glaciated margins globally is the preserved erosional surfaces and sedimentary deposits. These features can be observed in high-resolution seismic reflection records and interpreted using glacial sequence stratigraphy [e.g., Powell and Cooper, 2002; Solheim et al., 1998]. One important task is to use the preserved stratigraphy to consider the processes and rates of processes involved in glacial advances onto continental shelves and retreats from those shelves globally.
 Along the southern Alaskan shoreline, eight across-shelf sea valleys have been interpreted to act as conduits for glaciers during expansion (Figure 1) [Carlson et al., 1982]. Common characteristics of these valleys alluding to a glacial origin include the following: U-shaped cross sections; concave longitudinal sections that tend to shoal at the seaward end, possibly indicating grounded ice; till-like sediments collected from the walls and outer shelf deposits adjacent to the valleys; and a pre-Holocene erosional surface incised into lithified strata [Carlson et al., 1982]. Many of the valleys display overdeepened sections, where the depositional surface deepens toward land, common in modern sub-glacial environments [Alley et al., 2006].
 Concurrent with continued orogenesis of the St. Elias Mountains of Alaska, which include the Bering, Malaspina, and Hubbard Glacial systems (Figure 1), dramatic climatic cooling during the late Pliocene (~3 Ma) resulted in extensive glacial cover across the Gulf of Alaska margin [Lagoe et al., 1993] that may have slightly preceded northern hemisphere-wide glaciation. In the Pleistocene, marine-terminating glaciers dominate the delivery of sediment to the margin, as evident from the abundance of ice-rafted debris within coastal sediments and thick deposits on the shelf and in the deep-sea Surveyor Fan [Eyles et al., 1991; Rea and Snoeckx, 1995; Lagoe and Zellers, 1996; Worthington et al., 2008, 2010; Reece et al., 2011]. Glaciations appear to have intensified with the mid-Pleistocene climate transition from 41 to 100 kyr cycles, and a feedback between an increased glacial erosion and tectonic response of the mountain range has been documented [Berger et al., 2008; Worthington et al., 2008, 2010]. The end of the Last Glacial Maximum (LGM) and gradual return to present climatic conditions for the northern Pacific was roughly synchronous with worldwide LGM dates around ~22,000–17,000 B.P. [Mann and Hamilton, 1995]; though various age constraints apply a broader range between 25,000–14,000 B.P. [Heusser, 1995].
 Late Holocene climate conditions over the past ~2000 years are punctuated by multi-century climate fluctuations, e.g., the Little Ice Age (LIA) and the Medieval Warm Period (MWP) [Calkin et al., 2001]. Beginning approximately 1500–1300 B.P. (~600 A.D.), nearly all southern Alaskan glaciers advanced [Calkin et al., 2001; Barclay et al., 2001]. Transition to the MWP occurred around ~900 A.D.; warmer climatic conditions persisted for nearly 300 years, suppressing glacial advance across the margin except for Hubbard Glacier, which exhibited an out-of-phase behavior with climate advance through the MWP and indeed to modern times [Barclay et al., 2001]. At the commencement of the LIA in Alaska around ~1200 A.D., equilibrium line altitudes were depressed between 150 and 200 m [Calkin et al., 2001], signaling the onset of a threefold advance period (13th–15th centuries, 17th, and 19th centuries) for most Alaskan Glaciers that, in total, spanned 1200–1900 A.D. [Calkin et al., 2001] in contrast to the shorter duration Little Ice Age in Europe.
 The Hubbard and Malaspina glacial systems dominate glacial activity in the Yakutat area (Figure 2). Hubbard Glacier appears to have advanced out of phase with most glacial systems in southern Alaska. Evidence for late Pleistocene expansions of Hubbard Glacier is preserved at the southern end of Russell Fiord, where two terminal moraine deposits are thought to represent late Wisconsinan ice limits of the glacier [Barclay et al., 2001]. Radiocarbon and tree ring dating suggest that the latest of the “neoglacial” advances of Hubbard Glacier, which began around ~3100 B.P., brought ice margins to the southern end of Russell Fjord by 1710 B.P.; in Yakutat Bay, researchers suggest that by 1000 A.D. the terminus of Hubbard Glacier had reached the mouth of Yakutat Bay, depositing the bay-mouth wide morainal bank [Calkin et al., 2001; Barclay et al., 2001; Carlson, 1989]. By 1308, retreat of the glacier was underway. The current advance phase for Hubbard Glacier was underway by 1895, resulting in two recent damming events of Russell Fjord in 1986 and 2002 [Motyka and Truffer, 2007; Trabant et al., 2003; Willems et al., 2011].
 Ice originating in the Bagley Ice Field flowing south out of the Seward Throat expands along the shores of the Gulf of Alaska between Icy Bay and Yakutat Bay, producing the Malaspina Glacier system (Figures 1 and 2). Dubbed the largest temperate piedmont glacier in the world, the Malaspina and contributing tributary glaciers (Agassiz, Seward Lobe, Marvine, and Hayden Glaciers) combine to cover an area of 5000 km2, all of which exist in the ablation zone below 600 m elevation [Sauber et al., 2005]. Current morphologic and hydraulic characteristics of the Malaspina Glacier and its foreland are comparable to environments that existed along the southern Laurentide margin during the Pleistocene [Gustavson and Boothroyd, 1987]. The Malaspina Foreland is populated by large fan deltas composed of outwash plain and barrier spit deposits. Various ice-contact deposits including collapsed lake deposits, collapsed eskers, crevasse filling, and stagnant ice provide evidence for the passive to current stagnant nature of the terminus. Due to the proximity to the current shoreline, no terrestrial record exists for the advance-retreat history of Malaspina Glacier into Yakutat Bay or onto the continental shelf.
 This paper focuses on evaluating the preserved glacimarine signature of glaciation in the Yakutat Bay region and Yakutat Shelf along the southeastern coast of Alaska, USA. We present an interpretation of seismic reflection data as a case study for using sub-surface data to determine ice extent and the dynamics of glacial retreat. The preserved glacial sequence stratigraphic signatures of these advance-retreat sequences are comparable to those associated with margins characterized by marine-terminating or shelf-proximal glaciers in Alaska, Greenland, Svalbard, Norway, and Antarctica [e.g., Landvik et al., 2005; Solheim et al., 1998; O'Cofaigh et al., 2005; O'Brien and Harris, 1996; Wellner et al., 2006; Ottesen et al., 2005; Heroy and Anderson, 2005].
 During September 2008 as part of the St. Elias Erosion/Tectonics Project (STEEP), 1250 km of 2-D multi-channel seismic (MCS) reflection and refraction data were collected throughout the Gulf of Alaska basin aboard the R/V Marcus G. Langseth. Interpretations along reflection lines STEEP01a/b and STEEP02, running roughly east-west and north-south, respectively (blue lines in Figure 1), were used in this study and their processing is discussed in Worthington et al. . During the summer of 2004, aboard the R/V Maurice Ewing, 1800 km of high resolution 2-D MCS data was collected in and around fjords and the inner shelf of the Gulf of Alaska (red lines in Figure 1) in an attempt to better understand the mechanics of late Quaternary sediment delivery to the margin. Dual 45/45 in3 GI (generator/injector) air guns were used, achieving vertical resolutions of 5 m or greater. Particular lines significant to this study include goa2001, goa2002, and goa2003, located offshore of the present day Malaspina Glacier, just south of the mouth of Yakutat Bay. The processing of these high-resolution data is discussed in Berger et al. .
 In June 1979, Western Geophysical Company acquired a 30,400 km2 grid of MCS data (yellow lines in Figure 1) aboard the R/V Anne Bravo in the Yakutat-Fairweather shelf district as part of a hydrocarbon exploration survey (Figure 1). The data are available for research purposes through data archives at the USGS. A 650 in3 air gun array at 4500 psi was used during acquisition; vertical resolutions are generally 25 m throughout the data set. Western Geophysical completed the processing of the data through post-stack time migration.
 Our interpretation methods were to identify glacimarine sequences within the upper ~1 s of the available data (e.g., Figures 3 and 4). These identifications were made based on reflector truncations in the case of sequence-bounding basal erosional surfaces and in acoustic facies changes in the case of the top of the retreat phase strata. Structure contour maps were made for the most recent two horizons in the Yakutat Bay, outer Yakutat Sea Valley, and Alsek Sea Valley areas; a time-thickness isopach map was made by subtracting these horizons. We assign acronyms for horizons that identify their general geographic positioning relative to major landforms across the shelf; these acronyms include Yakutat Bay (YB), the Yakutat Sea Valley (YSV), and the Alsek Sea Valley (ASV). Horizons are further classified based on their stratigraphic character and were given -u or -ur suffixes in order to discriminate between erosional and retreat-phase depositional surfaces, respectively.
2.2 Erosional Horizons and Overlying Sediments
 In the Yakutat area, we identify four regional erosional surfaces as well as two surfaces bounding the top of glacial retreat phase sedimentary deposits that overlie the two most recent of these erosion surfaces (Figures 3 and 5). Chronologically (based on superposition), the mapped surfaces from oldest to youngest are the deeper pre-YSVu erosion surface, the shallower pre-YSVu erosion surface, the YSVu erosion surface, the YSVur top of retreat sequence (Figures 3, 5, and 6), the YBu erosion surface, and the YBur retreat surface (Figures 3 and 5). In the Alsek portion of the shelf, we map two surfaces which are ASVu erosion surface and the subsequent ASVur top of the retreat sequence (Figures 4 and 7). Within the upper 1 s of seismic data, there are older surfaces in the Alsek Sea Valley area and offshore Icy Bay, but these were not mappable beyond two of the seismic lines within our existing dataset.
2.3 YBu and YBur Surfaces
 The most recent erosional surface mapped near Yakutat Bay is the YBu horizon. The horizon originates in Yakutat Bay and extends south across the shelf where it terminates mid-shelf at the seafloor (Figures 3 and 8a). To the west (Figure 8a), just south of Icy Bay, the surface does not extend into the Pamplona Zone thrust belt [e.g., Worthington et al., 2010]. Within the high-resolution lines (goa2001 and goa2002) that cross into the northern and eastern flanks of the modern morphological Yakutat Sea Valley, the surface is significantly deeper (Figure 5). A structure contour map of the erosional surface (YBu) shows an arcuate or semicircular geometry relative to Yakutat Bay, with an overdeepened region to the south of the present position of the Malaspina Glacier (Figure 8a).
 The sediments deposited above the YBu erosional surface are capped by horizon YBur (Figures 5 and 8b); these sediments are thin within the Yakutat Sea Valley, thicker east of the Sea Valley, and display a clear thickening trend at the mouth of the Yakutat Bay (Figures 5 and 8c). There are two acoustically distinct packages within the sediments between YBu and YBUr along east-west oriented high-res line goa2002 (Figure 5): a lower unit that extends into the Yakutat Sea Valley and an upper dome-shaped deposit cut by channels that is not regionally extensive (Figure 5). The upper unit thickens landward and its surface expression is very irregular with several hills and troughs in the entrance to Yakutat Bay (Figure 10). Separating these two packages is a horizon exhibiting high amplitude reflections. Both the upper and lower units within the sedimentary deposit above YBu lack consistent internal layering and appear slightly transparent in areas; the capping horizon YBur is very reflective and can be easily distinguished from the overlying and underlying sediments (Figure 10).
 At the mouth of the Yakutat Bay, large bank deposits on the seafloor can be seen in seismic sections and in bathymetry (Figures 2, 3, and 10). These deposits have been previously interpreted as morainal banks formed during the last time Hubbard Glacier expanded to the mouth of the Yakutat Bay [e.g., Carlson, 1989]. Internal stratigraphy is unresolvable within the banks either due to the highly reflective nature of the capping horizon YBur, the short-period multiples of which propagate through the entire structure (Figure 10), or due to the presence of glacial diamict. Sediment thicknesses show a substantial amount of accumulation from between the YBu and YBur surfaces within the Bay, reaching nearly 500 m or > 400 m thick at the western most region of its mouth (Figure 8c).
2.4 YSVu and YSVur Surfaces
 Extending laterally from the inner shelf to the shelf edge and as far west as Icy Bay, the YSVu horizon records a significant and extensive erosional event (Figure 11a). Like YBu, this surface rises toward the seafloor as it progresses to the outer shelf (Figures 3 and 11a). Toward the Yakutat Bay, the YSVu surface truncates against the YBu surface (Figures 3 and 10), thereby limiting the extent of the preserved surface to the middle and outer shelves. Within the axis of the present day YSV, the YSVu surface is overdeepened in this area (Figures 6 and 9) and follows a similar trajectory of the present YSV to the shelf edge (Figure 11a).
 To the west, the intensity of the structural deformation associated with the Pamplona Zone has rendered interpretation across the thrust belt difficult, although the surface appears to remain at depth (Figures 9 and 11a). Lines steep02 and fw-066 illustrate the difficulties of interpreting the YSVu horizon into the Yakutat Bay and westward (Figures 3, 9, and 10); the previously discussed YBu deepens across the mouth of the Yakutat Bay, ultimately truncating the YSVu surface in this area.
 Within the Yakutat Sea Valley, erosion by successive glacial advances has created a significant buried trough and modern bathymetric expression of that trough that extends close to the shelf edge (Figure 11a). The overlying sediment package capped by YSVur mimics the plan view morphology of the YSVu surface (Figures 11b and 11c), although the thickest sediments seem limited to the flank of the buried trough below the modern Sea Valley (Figures 6 and 9). Uncertainty in our interpretations of the sediment package to the west exists due to multiple interferences associated with present YSV's ledges (Figures 6 and 9). Nonetheless, we interpret a large, broken rim of sediments bordering the outer reaches of the YSVu trough and flanks (Figure 11c). The rim shows a nearly constant thickness of ~100 m; however, the eastern section displays two significantly thicker regions reaching 200+ ms. The seismic character of the sediment rim is highly chaotic and transparent, except for distinct reflections that dip toward the axis of the trough (Figures 6 and 9). The inner portions of the trough are devoid of sediments, although a triangular-shaped (in cross-section) axis-parallel reflector may depositionally correlate with the flank sediments bounded by the YSVu and YSVur reflectors; alternatively, this feature may be pre-YSVu sediments that escaped erosion (Figures 6 and 9). Like the YSVu surface, this feature is not imaged towards the Yakutat Bay due to truncation by the YBu horizon. The greatest preservation of the YSVu surface and overlying material beneath the YSVur surface is to the east where the erosion that resulted in the YBu surface was not as deep (e.g., Figure 6).
2.5 ASVu and ASVur Surfaces
 In the easternmost sector of our study area, the Alsek River discharges along the southern Alaskan coast onto the shelf region occupied by the present ASV (Figure 1). Cutting obliquely across the shelf and buried below a package of draped sediment reaching thicknesses of ~250 m is the locally extensive ASVu (Figures 4, 7, and 12a). The surface, which reaches the shelf's edge, displays areas overdeepened to depths of 550+ m. Like the YSVu surface, this horizon displays a trough-like appearance with the majority of scouring occurring along the central axis. An exception is noted in that the YSVu horizon appears to have a westward extension closer to the modern shoreline. Along the main trough axis, the surface is somewhat irregular, exhibiting variations in relief up to 200 m resulting in two very distinct lows or basins along the troughs trajectory. Sediments dip west and north (Figure 4) of a basement highly known locally as Fairweather Ground [e.g., Bruns, 1983; Worthington et al., 2010] and have been eroded by the event that resulted in the ASVu surface as well as locally at the seafloor (Figure 4). Sediment accumulation atop the ASVu surface is similar in character to the previously described deposits to the west. For the most part, accumulation is confined to the outer rim of the shelf-crossing trough (Alsek Sea Valley), with minimal thicknesses occurring within the trough itself (Figure 12c). Also, the capping horizon ASVur converges with ASVu on the outer shelf, confining accumulation between the mid and inner shelf regions (Figures 4 and 12c). Along the westward arm of these surfaces, a larger NW-SE trending ridge of sediment exists atop the limits of the ASVu surface (Figure 7). The sediments confined by the ASVu/ASVur horizons, in general, show no internal structure; the surface expression undulates in areas with peaks and trough showing up in many places (Figures 4 and 7).
2.6 Pre-YSVu Surfaces
 In the Yakutat Bay and south across the shelf, two pre-YSVu surfaces are mapped to the shelf edge (Figure 3). Within and proximal to the Yakutat Bay, it is possible to map these horizons; however, outer shelf areas were not as easily mapped because of navigational inconsistencies between surveys and interference by seafloor multiples. As a result, these surfaces were not gridded, although inferences about their possible locations relative to mapped horizons were made. Along line steep02, the two surfaces are delineated crossing the shelf and approaching the seafloor very close to the shelf's edge (Figure 3). Within the Yakutat Bay, the two horizons are imaged at depth (Figure 10). South of Yakutat Bay, along lines goa2001 and goa2002, the same trend of deepening into the region beneath the modern Yakutat Sea Valley that was seen for YBu and YSVu also occurs with the pre-YSVu horizons (Figure 5).
 In goa2002, thick packages of chaotic, yet well-defined sediments overtop the two events and are cut by channels (Figure 5). The seismic character of the packages is similar to those of the ones bound by YBu/YBur and YSVu/YSVur. Despite our inability to reliably map the pre-YSVu surfaces across the shelf, taking into account the positions of the two horizons relative to YSVu and our interpretations of the surfaces along line steep02, we assume that these pre-YSVu horizons do in fact make it to the shelf edge within the vicinity of the present YSV (Figure 3).
 At the shelf's edge above and below the oldest interpreted pre-YSVu horizon and below the younger pre-YSVu surface, we delineate packages of sediments with internal reflectors dipping towards shore, much like the sediments bound by the YSVu/YSVur events. The dipping reflectors, where present, tend to downlap onto the underlying horizon followed by truncation via the overlying horizon. Other than the distinct, local intervals of dipping reflectors, the sediments show no internal structure and are seismically homogenous.
3.1 Regional Erosion
 Pleistocene stratigraphic records in the Pacific region indicate regional climate variability consistent with a globally cooler climate, resulting in conditions conducive for exponential glacial expanse [i.e., Rea and Snoeckx, 1995]. Therefore, we interpret the numerous erosional events across the shelf as evidence for glacial advances and the direct signatures of the extent of glacial occupation (Figures 3, 8, 11, and 12).
 Our investigation of the recent stratigraphic record along the southern Alaskan shelf reveals four erosional events originating from the Yakutat Bay and a singular event buried beneath the present ASV (Figures 3, 4). Horizon YBu is the shallowest surface that we interpret to be an erosional event and is the last recognizable erosional event seen near the Yakutat Bay (Figures 8 and 10). The surface does not reach the shelf edge, but does show correlation with sediment deposited within and just south of the Bay (Figures 5 and 10). Several researchers attribute the large morainal banks existing at the present seafloor in the Yakutat Bay to Hubbard Glacier expanse and subsequent retreat around 1000 A.D. [Calkin et al., 2001; Barclay et al., 2001; Carlson, 1989]. These dates suggest that the Hubbard Glacier would have been at the mouth of the Yakutat Bay 200 years prior to the accepted onset of the LIA ice expanse in the region. Out-of-phase advances and retreats by tidewater glaciers are not uncommon as their advance and retreat cycles are not primarily governed by climatic oscillations [Motyka and Truffer, 2007]. Nonetheless, our interpretations and recent seafloor imaging suggest a different origin for the large accumulations of sediments at the mouth of the bay.
 The structure contour maps of the YBu and YBur surfaces indicate enhanced erosion and deposition directly offshore of the present position of the Malaspina Glacier (Figure 8). The large, pronged basin extending offshore is considerably deeper than the surrounding areas, as if sediments were sourced directly from the Malaspina Glacier. Within the bay, a locally smaller area eroded to around 575 m also exists adjacent to the shoreline. Second, the grid for surface YBur (Figure 8a) reveals a sediment rim that is concave in the direction of the Malaspina foreland. Comparing these results with present seafloor bathymetry in the Yakutat Bay, we see a similar orientation of what is likely the same sediment rim (Figure 2). Therefore, we interpret the rim not as indication for a 1000 A.D. advance of the Hubbard Glacial system, but alternatively as sediments deposited as a morainal bank during the retreat of the Malaspina Glacier at the end of the LIA and, therefore, the YBu event as the LIA advance/retreat surface. Furthermore, because of the large volume of sediment accumulation, it is likely that the glacier experienced a period of relative stability within the Yakutat Bay before retreating to its present position [e.g., Dowdeswell et al., 2008].
 The Hubbard Glacier may have experienced an out-of-phase advance prior to the LIA, but this advance likely was constrained to the glacial trough linking the overdeepened parts of Disenchantment Bay with the deeper bathymetry to the east side of the Yakutat Bay (Figure 2). There is no evidence for Hubbard Glacial system advances reaching to the shelf independent of the Malaspina Glacier within our available seismic data. Rather, the Disenchantment Bay and the east side of the Yakutat Bay appear to have been occupied during repeated and likely locally asynchronous advances (Figure 2).
 Two erosional surfaces, YSVu and ASVu, are interpreted to represent the most recent glacial advances to reach the shelf edge within our study area (Figures 11 and 12). Though the two surfaces, and thus the glacial advances, could not be correlated directly, their similar geographic expanses suggest that they are coeval. As the last time interval to experience glacial advances to the shelf edge, we interpret the two events as coeval and suggest that they represent the LGM advance during the latest Pleistocene. Comparing the two surfaces, we see striking similarities in the style of erosion (Figures 11a and 12a). The YSVu horizon shows greater lateral expanse to the west, again indicating confluence of ice in the Malaspina District (Figures 1 and 9). However, from the mid-shelf outwards and to the east of the YSVu surface and west of the ASVu surface, no evidence exists for ice occupation over the shelf in this area. Both horizons rise to the seafloor and terminate rather than merge, implying a grounded ice-free embayment during the LGM between the Yakutat and Alsek ice lobes (Figures 11 and 12). Both the YSVu and ASVu surfaces imply concentrated ice advance in large, shelf-crossing troughs toward the outer limits of the shelf. Within the troughs, we see the greatest degree of erosion, suggesting more intense ice flow. Such a phenomenon is likely the result of locally streaming ice in these areas, with faster ice resulting in erosion that is more vigorous as has been suggested for Svalbard and Norwegian shelf glaciers [e.g., Dowdeswell and Siegert, 1999]. In the Bering Trough to the west, high-resolution seismic images show a similar tendency for ice advance surfaces over the last ~1 Ma to concentrate in ice streams within a shelf-crossing trough, instead of large and broad sheets [Berger et al., 2008]. These shelf-crossing troughs appear to be self-sustaining as a locus of erosion during the advances and deposition during the retreats, but this deposition is not significant enough to remove the tendency for the next advance to reoccupy the trough.
 Isopach maps between the two erosional surfaces that record ice advance (YSVu and ASVu) and their respective retreat surfaces (YSVur/ASVur) provide an idea of the degree of sediment accumulation on the shelf during the LGM (Figures 11c and 12c). Above the YSVu surface, the broken rim of sediments is likely a grounding zone wedge (Figure 11c). We do not imply or assume any lengthy time of stability for the ice in this area, as one might expect for a terminal or recessional moraine. Alternatively, we propose that sediments were either bulldozed at the ground line during advance or represent retreat phase sediments accreted to the trough margins. Based on the internal architecture of the material, we favor the latter because a cross-section across the trough shows internal reflections dipping toward the trough axis (Figures 6 and 9). At this orientation, it seems the material was deposited in stages during the retreat, building towards the axis of the trough, possibly analogous to ice stream boundary ridges that have been identified in Antarctica [Anderson et al., 1992]; these ridges have been interpreted as indicating lateral movement of ice streams during retreat which makes sense for retreat in these Alaska cross-shelf troughs as well.
 The small window that we have of the YSVu surface that is not truncated allows us to draw insights about ice stability during the occupation on the outer shelf. The lack of any buttressing topography across the shelf suggests the ice was not well confined. Because of this lack of confinement, the ice could be unstable, especially within the extended region along the trough [Schoof, 2007]. Inspection of the landforms indicative of ice occupation reveals no transverse ridges or recessional deposits along the outer limits of the shelf (Figure 11). This observation also implies that ice occupation was limited at the shelf edge and retreat was rapid, rather than punctuated or episodic [Dowdeswell et al, 2008].
 Comparing the interpreted LGM retreat deposits in the Yakutat versus Alaska systems, several characteristics suggest a different style of retreat for the Alsek following the LGM. First, the ASVu erosional surface is segregated into two distinct lows separated by a higher relief section that transverses the entire trough (Figure 12a). The outer low is devoid of retreat sediments other than a slight mantle over the shoreward sections. Not until the midvalley high do significant deposits of sediment begin to appear: sediment accumulation circumvents the second low and reaches a thickness of 150+ m (Figures 4, 7, and 12c). In addition, to the west of the second smaller low, a ridge oriented obliquely to the main trough cuts perpendicular across the western arm extension running parallel to the shore (Figure 12a). The differences in appearance of the ASVu and YSVu surfaces, combined with the considerably different amounts of sediment accumulation between the two areas, give reason to believe that the retreat of the ice mass occupying the Alsek glacial trough at the LGM was episodic and even stalled for a period of time before retreating completely. The two overdeepened lows on the ASVu surface are evidence for intense subglacial erosion during the respective periods of ice occupation. Like during the formation of the YSVu surface, the outer shelf during formation of the ASVu surface was likely a highly unstable environment, and retreat from this terminal position was rapid, resulting in little to no sediment accumulation. However, it seems that stability was reached again approximately midshelf where the glacier stalled and eroded a second overdeepened basin and possibly remained at this position long enough to deposit recessional ridges to the west and a significant rim of material around the overdeepened low (Figure 12). Internally, the ridges and rim material do not show accretionary structures like the LGM retreat material bordering the Yakutat system but are more chaotic and seismically transparent, similar to the morainal banks described in the Yakutat Bay. Therefore, we interpret the deposits overlying the ASVu surface to represent at least two stages of retreat along the shelf during the LGM before retreating onshore. The initial retreat was apparently rapid and reached close to the modern foreland. There is no evidence of a later advance through the Alsek River corridor by glacial ice during the LIA in contrast to the Yakutat Sea Valley.
 Prior to the LGM event, intense glacial activity during Plio-Pleistocene time continually resulted in shelf-wide glacial expanse. The two pre-YSVu surfaces interpreted across the shelf represent these expansions (Figure 3), although direct age correlation to a particular event is not possible. Based on interpretations, we assume that they were likely just as expansive, if not more expansive than the LGM event. Seismic imaging suggests that these glacial advances may have been more widespread, especially in the middle shelf, as erosional surfaces reached the shelf edge in some locations where the LGM erosional surface terminates at the seafloor landward of the shelf edge (Figure 3). Interpretations at the mouth of the Yakutat Bay show the pre-LGM horizons deepen towards the Malaspina (Figure 5). This observation may be an additional evidence that the mechanics of the Yakutat Bay glacial system are dominated by and dependent on ice flow from the Malaspina complex. Survey lines displaying cross-ties of our interpretations beneath the present Yakutat Bay suggest that isolation of ice flow into highly eroded troughs did not occur and ice coverage was broad and widely distributed (e.g. Figure 6). Along line fw-050, large packages of sediments with internal dipping reflectors represent retreat material during the evacuation of ice from the shelf edge (Figure 13). The package directly above the oldest interpreted pre-YSVu surface (green horizon) shows striking similarities to a grounding zone wedge [Mosola and Anderson, 2006], indicating a somewhat stable regime during ice occupation [Dowdeswell et al., 2008]. Comparing this image with what we see during the LGM and the younger pre-LGM event, the retreat was much slower than later retreats at the shelf edge, allowing for substantial sediment accumulation. It seems that this slower retreat may have been the case for some even earlier events (Figures 13). Based on the available data and correlation of sea valleys to Glacial Interval C (mid-Pleistocene and younger [Berger et al., 2008]), we interpret the two pre-YSVu surfaces as recording late Pleistocene advances, though no specific age can be assigned at this time.
3.2 Glacial Stability and Persistent Sea Valleys
 Many stratigraphic signatures found along previously glaciated margins are direct evidence for ice occupation and delineate ice contact, proximal and distal environments associated with glacial termini position. Many of these signatures occur in predicable lateral and vertical successions, allowing for correlation and interpretation of respective environments relative to one another. Moreover, the presence or lack of assemblages of specific depositional landforms provides evidence for ice behavior and relative stability at terminal and recessional positions [e.g., Anderson et al., 1992; Dowdeswell et al., 2008]. Powell and Cooper  extrapolate such assumptions using a model specific to temperate glacier tidewater margins and discuss the capability of using typical landform assemblages to interpret depositional environments over several glacial cycles. They conclude that like traditional sequence stratigraphic models, the tidewater model is useful in predicting certain depositional systems tracts associated with a complete sea level cycle, and these tracts are bound by erosional surfaces and sequence boundaries, i.e., highstand, lowstand, and transgressive systems tracts. Based on observational data, a prescribed set of depositional structures is assumed to be predictable for each systems tract. Examples that our data support include grounding lines and proglacial wedges during lowstand, onlapping sediments during transgression, and prograding and/or aggrading deltaic deposits during highstand (Figure 5).
 Building on this idea, certain landform assemblages can be used to identify stability during terminal and recessional periods of ice occupation [Dowdeswell et al., 2008]. Specific to our study area, grounding zone wedges are packages of sediment deposited at the glacier or ice sheet toe during the grounding stages [e.g., Dowdeswell and Fugelli, 2012]. The presence of these deposits provides evidence for momentary stability and/or slow retreat of ice [e.g., Dowdeswell et al., 2008]. At the same time, dearth and frequency of these events suggest rapid retreat due to instability or periodic, stable pauses during retreat, respectively. Throughout the imaged area, the grounding line deposits that occur in specific areas along the shelf (namely, within and just south of Yakutat Bay) associated with the LGM event beneath the Alsek Sea Valley and at the shelf's edge (Figures 3, 10-12, and 13). The existence of grounding zone wedges or morainal material only at the shelf edge and proximal to bays and forelands implies that retreats from glacial maxima along this margin are consistently rapid and with only rare pauses (e.g., Alsek).
 Shelf-crossing bathymetric lows (such as the Yakutat Sea Valley, Alsek Sea Valley, and the trough in Disenchantment Bay and eastern Yakutat Bay) overlie eroded depressions that persist through multiple glacial cycles. In the Gulf of Alaska, the shelf is crossed by eight sea valleys that bear morphologic similarities to Yakutat and Alsek Sea Valleys, implying a common glacial process. A self-fulfilling aspect to these glacial systems appears to be in whereby eroded lows are not fully filled during the interglacials, such that the same shelf and bay locations are reoccupied with each subsequent glacial advance. This observation supports the interpretation that these troughs are caused by ice streams as proposed for Bering Trough in Alaska [Berger et al., 2008] and shelf-crossing troughs similar to those imaged here on other Northern and Southern Hemisphere glaciated margins [e.g., Landvik et al., 2005; Solheim et al., 1998; O'Cofaigh et al., 2005; O'Brien and Harris, 1996; Wellner et al., 2006; Ottesen et al., 2005; Heroy and Anderson, 2005]. Some caveats regarding the Yakutat and Alsek Sea Valleys when compared to these other settings are that (1) these troughs cross a wide shelf yet show little evidence of lateral migration, (2) there is only minimal evidence for step-wise retreat facies such successive grounding zone wedges, and (3) despite long-lived successive glaciations in these troughs, there are kilometers of Plio-Pleistocene sediments preserved on the Yakutat Shelf [e.g., Worthington et al., 2008, 2010] and thus no bedrock-related features. We interpret some of these differences to be caused by temperate glaciers eroding an actively deforming orogen resulting in significant sediment flux through the paleo-ice streams that occupied these Alaskan troughs.
 Over the last several million years along the coast of southeastern Alaska, glacial influence and variability in the face of climatic-tectonic interactions have resulted in a substantial offshore record that provides invaluable evidence regarding the dynamic nature of such interactions. Seismic imaging reveals five major erosional events across the region that indicate ice occupation and expanse along the shelf dominated by ice emanating from the Malaspina Foreland and the Alsek River Valley. The Hubbard Glacial system, in contrast, advances in a trough that forms Disenchantment Bay and occupies the eastern part of the Yakutat Bay, but does not reach the shelf in the areas of data coverage. Three glacial retreat surfaces provide evidence for sediment accumulation during events herein named YBu, YSVu, and ASVu. Isopachs between these events reveal several stratigraphic signatures characteristic of the style of retreat. Based on relative timing and knowledge of past glaciations in Alaska, the five events likely represent ice expanse during the Little Ice Age (YBu), the Last Glacial Maxima (YSVu and ASVu), and two Plio-Pleistocene pre-LGM events. For the glacial maxima, the glacial systems appear to transition to ice streams advancing across the shelf in self-sustaining, overdeepened sea valleys or shelf-crossing troughs similar to those observed on other high-latitude glaciated shelves. The two Alaskan troughs discussed here may be end-member examples of high-sediment flux systems due to the temperate glacial setting combined with an actively exhuming orogen. Retreat on the Yakutat Shelf from glacial maxima, however, appears to be rapid with few examples of grounding zone wedges or moraines except for near the shelf edge and proximal to bays or forelands.
 We thank the captain and crew of the R/V Maurice Ewing for the assistance in acquiring the 2004 high-resolution seismic data. The research supported by NSF Ocean Drilling Program award OCE-0351620 and Continental Dynamics award EAR-0408584. Christopher R. Elmore is supported by a UT Institute for Geophysics, Gale White student fellowship. Manuscript benefitted from a review by John Anderson. This is University of Texas Institute for Geophysics Contribution #2558. The authors thank John Jaeger for assistance with the bathymetry of Yakutat Bay.