Valley morphology and Quaternary seismic stratigraphy of the Manicouagan impact crater lake (Eastern Canada)

Lakes in formerly glaciated regions can provide valuable paleoclimate archives. Lake Manicouagan (Manikuakan, according to Innu toponymy), formed in the basin of the ~214‐Ma Manicouagan impact crater of eastern Québec, is a key area for reconstructing long‐term environmental change, as it was directly affected by the Pleistocene glaciations and the waxing and waning of the Laurentide Ice Sheet. Here, we present high‐resolution seismic data revealing an overdeepened bedrock valley filled with a sedimentary sequence. We assess its potential to serve as a paleoclimate archive. The varying shape of the overdeepened valley indicates complex erosional processes. A lower narrow V‐shaped gorge is indicative of either pressurized subglacial meltwater or pre‐Quaternary fluvial erosion, or a combination of both. Three scenarios are discussed regarding deposition of the sedimentary sequence: (i) deposition only during and after retreat of the last glacial episode; (ii) deposition during multiple glacial–interglacial cycles; and (iii) deposition mainly during a subglacial lake stage. We suggest subglacial followed by proglacial sedimentation as the most probable scenario for deposition of the sedimentary succession. We recommend considering the sediments of Lake Manicouagan as a paleoclimate archive reaching back at least to 7.5 ka, but the lake also probably contains sediments deposited before the last deglacial period.


Introduction
Deep and large lakes form ideal physiographic settings for climate science investigations, as they potentially contain a thick and undisturbed record of sedimentation that reflects changes in depositional regimes controlled by climatic variabilities and events over long timescales (Melles et al., 2012, Zolitschka et al., 2013, Litt et al., 2014, Wagner et al., 2019).Lakes of middle and high latitudes containing such sedimentary sequences are valuable paleoclimate archives that can be used to reconstruct environmental changes during periods of deglaciation and phases of glacial readvances (Mullins and Hinchey, 1989, Eyles et al., 1991, Mullins and Halfman, 2001, Charlet et al., 2008, Christofferson et al., 2008, Waldmann et al., 2010, Greenwood et al., 2015, Gagnon-Poiré et al., 2018, Trottier et al., 2020, Trottier et al., 2021).These paleoclimate archives provide important constraints for climate modeling, in particular for the reconstruction of the behavior of ice sheets during a warming climate.However, the potential of long-term sedimentary archives in lakes is often constrained by numerous factors such as low sedimentation rates or non-deposition (Desiage et al., 2015, Lajeunesse et al., 2017), or simply because most lakes worldwide only formed following the end of the last glacial, hence being younger than 20 ka (Kelts 1988).In addition, missing accommodation space, drystand periods (De Cort et al., 2021) or absent sediment preservation in a lake result in a sedimentary hiatus.Direct glacial erosion or mass movements can remobilize sediments and thus be the reason for the lack of sediment preservation (Gagnon-Poiré et al., 2018, Trottier et al., 2020).Despite these many limiting factors, seismic and coring investigations conducted in deep lake basins of formerly glaciated regions have allowed the observation and interpretation of thick sedimentary sequences of several hundreds of meters (Mullins and Hinchey, 1989, Eyles et al., 1991, Christofferson et al., 2008).These thick sedimentary sequences suggest high sedimentation rates that offer the potential of high-resolution paleoclimate archives.In particular, lakes situated in impact craters provide large accommodation space for sediments and are therefore prone to host paleoclimate archives of long-term environmental change (Melles et al., 2012, Zolitschka et al., 2013).
With regard to the origin of overdeepened lake basins in glacial landscapes, there is an ongoing debate as to whether they were excavated solely by direct glacial erosion in bedrock (Mullins andHinchey, 1989, Eyles et al., 1991) or whether these incisions could also be the result of fluvial erosion and subsequent overprinting during glacial periods (Montgomery and Korup, 2011, Lajeunesse, 2014, Cooper et al., 2016).Complicating matters further, such lakes may be subglacial (Livingstone et al., 2012) or proglacial (Carrivick and Tweed, 2013), and the discrimination of the resulting sedimentary deposits is ambiguous (Livingstone et al., 2012).For North America, the existence of subglacial lakes beneath the Laurentide Ice Sheet (LIS) has been proposed by both local field studies (Christoffersen et al., 2008, Guyard et al., 2011) and regional modeling (Livingstone et al., 2013).The regional modeling predicts likelihoods for the occurrence of paleosubglacial lakes during the last glacial episode (Livingstone et al., 2013).A likelihood of 40% or higher is estimated by the regional model for the Manicouagan area.These modeled predictions for the occurrence of paleo-subglacial lakes have, however, a significant uncertainty.Therefore, Livingstone et al. (2013) suggest carrying out detailed field investigations based on their likelihood predictions.Local observations from excavations during the construction of the Daniel-Johnson dam in the south of the Manicouagan Reservoir revealed narrow gorges in the lower part of the valley (Lajeunesse, 2014) that have been interpreted as preserved base segments of preglacial fluvial valleys.In contrast, Jansen et al. (2014) proposed that inner gorges are the result of pressurized subglacial meltwater in most instances, hence pointing at a glacial formation process.These contrary interpretations by Lajeunesse (2014) and Jansen et al. (2014) show that the formation processes and evolution of deep lake basins in formerly glaciated terrain, such as Lake Manicouagan (Manikuakan, according to Innu toponymy), remain controversial and that their formation is likely to be polygenetic.Local field investigations, as carried out in this study, help to constrain and refine conceptual models for erosion processes.To prevent a bias and to test the most likely models for formation, this study aimed to obtain both perpendicular and longitudinal seismic profiles.
In general, ultra-high-frequency chirp sonar systems and high-frequency seismic systems are widely chosen tools to image lake sedimentary sequences.Due to their high frequencies (1-10 kHz) and high resolution, chirp data are well suited for investigating the Holocene development of lakes, such as sediments deposited as a result of mass movements related to seismicity or delta deposits associated with freshwater inflows (e.g.Lajeunesse et al., 2017, Trottier et al., 2020, Trottier et al., 2021).Multi-channel reflection seismic systems (MCS) in the sub-kilohertz range have the capability to resolve the entire sedimentary basin infill of lakes and the bedrock morphology due to increased penetration.In this study, we rely on high-resolution MCS data to examine the lake valley morphology and develop a Quaternary seismic stratigraphy for Lake Manicouagan.In addition, we use Chirp data for analysis of the shallow subsurface in areas where no MCS data are available.In essence, we make use of geophysical imaging methods to (i) assess Lake Manicouagan's sedimentary succession including its potential to hold a longterm paleoclimate archive, and (ii) evaluate the possible processes responsible for the erosion of Lake Manicouagan's thalweg and its resulting complex basement morphology.

Study area and settings
The Manicouagan Reservoir is located in northeastern Canada in the subarctic province of Québec ~540 km north-east of Québec City (Figure 1A).The reservoir is easily identified in satellite photographs due to its annular shape and a diameter of about 65 km (Dence, 1977, Phinney et al., 1978, Flamini et al., 2019).The Manicouagan Reservoir is between 2 and 10 km wide and it completely surrounds the Île-René-Levasseur (Figure 1B).The area was initially shaped by a meteorite that struck the Grenville Province during the Triassic (~214 Ma) and deformed Precambrian crystalline metamorphic and igneous rocks (Phinney et al., 1978, Hodych and Dunning, 1992, Spray et al., 2010).The resulting Manicouagan impact structure is among the largest of its kind, spanning roughly 90 km (Earth Impact Database, 2021).It is a complex impact crater characterized by a central uplift peak with rings (Spray and Thompson, 2008).The entire structure is well preserved but has been reshaped by glaciations since the late Triassic (Spray and Thompson, 2008).The central uplift peak is preserved as the reservoir central island (Île-René-Levasseur, Figure 1B).It is characterized by a horseshoe shape, Mont de Babel, with a height of 952 m a.s.l.(Biren and Spray, 2011), and probably impact-related high-angle basement faults (Spray and Thompson, 2008).Further, the impact resulted in a melt sheet located in the center of the reservoir reaching ~1500 m in thickness (Spray and Thompson, 2008).The melt rocks cooled in two stages and are texturally inhomogeneous but chemically homogeneous (Phinney et al., 1978).The geology of the Manicouagan area is dominated by shock-metamorphic overprinted rocks due to the meteorite impact (Murtaugh, 1976).Mesozoic igneous rocks and uplifted mafic rocks are present on Île-René-Levasseur, whereas the surrounding areas mainly consist of Paleoproterozoic mafic and undivided gneiss as well as Paleoto Mesoproterozoic undivided gneiss, mafic intrusive-diorite, gabbro and anorthosite (Murtaugh, 1976, Phinney et al., 1978).Quaternary deposits such as tills are deposited on top of the crystalline bedrock (Veillette, 2004).
Large regions of northern North America were covered or directly affected by ice sheets such as the LIS during the cold periods of the Late Pleistocene.The Manicouagan area was covered by the Québec-Labrador Ice Dome of the LIS at least since 60 ka (Marshall et al., 2000).A maximum thickness of 2000-3000 m is estimated for this ice dome for the last glacial maximum at 20 ka (Marshall et al., 2000).Reconstructions of the deglaciation by Dalton et al. (2020) indicate that the Manicouagan area was fully glaciated until 8.5 ka and that the ice had completely retreated at 7.3 ka.Veillette (2004) reconstructed three generations of ice movements for the Manicouagan area (Figure 1B).One generation is backed by traces of a relict ice movement east of the Manicouagan Reservoir, which are interpreted as movement towards 30-55°.This direction of ice movement has been proposed to correlate with the northeastward direction of the oldest striated surfaces in the southern Labrador region and is linked to the ice dispersal centers located in the Québec Highlands (Veillette, 2004).Another ice movement in an eastward direction of 105-110°is reconstructed based on striated grooves (Veillette, 2004).These grooves were subsequently filled with till, which protected them from erosion by later ice movements.A later ice flow direction of 160-170°is reconstructed based on striations (Veillette, 2004).The chronological order of these two movements has been deduced from two cross-striated outcrops.The chronological order of the ice movement towards 30-55°could not be reconstructed; it is therefore outlined with a question mark in Figure 1.
Until the 1960s, today's Manicouagan Reservoir consisted of two crescent-shaped lakes, namely Manicouagan (Manikuakan, according to Innu toponymy) in the east and Mushalagan (Mushalakan, according to Innu toponymy) in the west (Figure 1B).Construction of the artificial Daniel-Johnson dam in the south of the reservoir between 1962 and 1965 closed the outflow of the two lakes into the Manicouagan River (Manikuakanishtiku, according to Innu toponymy) (Figure 1B).As a result, the lake level rose by ∼140 m and the two lakes were connected, giving the Manicouagan Reservoir an annular shape (Flamini et al., 2019).In this paper, we focus on the eastern crescent-shaped former lake of the reservoir, Lake Manicouagan.

Data acquisition
A high-resolution MCS dataset was acquired in 2016 in the eastern portion of Manicouagan Reservoir from an 8.2-m launch (R/V Louis-Edmond-Hamelin).The survey focused on  (Dressler and Reimold, 2004).The directions of former ice flows reconstructed by Veillette (2004)  the deepest parts of the eastern sector of the reservoir, which corresponds to the limits of Lake Manicouagan prior to construction of the Daniel-Johnson dam.The acquired dataset consists of 111 km of seismic profiles (Figure 1C).For the generation of acoustic signals, a Sercel Mini-G Gun with a chamber volume of 200 cm 3 /0.2L was used.The source was triggered every 6 s and the gun pressure varied between 70 and 130 bar due to limited availability of compressed air.The survey speed was ~2 m s −1 , which results in an average shotpoint distance of 12 m.Recording of the acoustic response was realized by using a Geometrics MicroEel analog solid-state streamer with a group interval of 6.25 m.Every group consisted of three hydrophones.The streamer contained 24 channels, resulting in an active length of 150 m.To minimize noise from vibrations of the vessel, a lead-in of 50 m was deployed additionally to the active section of the streamer.The signal was digitized and recorded using a Geometrics Geode and a portable computer.The recording time was set to 3 s.Positioning was realized with a Hama GPS-Receiver.The vertical resolution of the seismic dataset is in the meter range in the upper subsurface and decreases with increasing depth.
In addition to the seismic profiles, 293 km of chirp subbottom profiles and a bathymetry grid covering 212 km 2 were collected during surveys in 2014 and 2016.A Knudsen Chirp 3212 was used as the sub-bottom profiler (3.5 kHz) for imaging the shallow subsurface.The vertical resolution of the chirp sub-bottom profiler data is in the decimeter range.In this study, we only use this dataset to fill gaps in the MCS dataset.High-resolution bathymetry data were acquired using a Reson Seabat 8101 and Kongsberg EM 2040 multibeam echosounders in 2014 and 2016, respectively.The grid cell size of the bathymetry grid presented here is 100 × 100 m.

Seismic data processing
All seismic profiles were processed using the commercial VISTA Desktop Seismic Data Processing Software (Schlumberger, version 2018).Processing included a 20/40-Hz low-cut filter, the subtraction of harmonic noise of 60, 120 and 240 Hz, both despiking and predictive deconvolution for signal enhancement, compensation of spherical divergence, common-midpoint binning and normal-move-out correction.The common-midpoint bin size was set to 3.125 m, which resulted in an average fold of five.Quality checks were carried out after each major processing step for all profiles.Only the first 10 channels were used to produce the final commonmidpoint stack due to an insufficient signal-to-noise-ratio in the rear part of the streamer.After stacking, a finite differences time migration was applied to the data.A data-driven velocity model derived from an iterative velocity analysis was used for the normal-move-out correction and migration.The model consists of three different units to which different sound velocities have been assigned.The velocity of the water column was set to 1460 m s −1 , according to the measured average water sound velocity from conductivity, temperature, depth sensor casts during hydrographic surveys.For the sedimentary sequence an average sound velocity of 1600 m s −1 was determined.The velocity of the bedrock was iteratively set to 6000 m s −1 .The limited offset range of the streamer did not allow for a precise velocity analysis.

Data analysis and interpretation
All available data were loaded into the software IHS Markit ® Kingdom version 2018 for analysis and interpretation.Boundary reflections of the seismic units as well as key reflections within the units were picked using IHS Markit Kingdom.The picked upper boundary reflections of unit 1 were used to calculate the surface of the bedrock valley using the Flex Gridding algorithm provided by IHS Markit Kingdom.Further, the picked surfaces of units 2, 3 and 4 were used to calculate the unit boundary surfaces in the same way.The thicknesses of the units in two-way-traveltime (TWT) were calculated by subtracting their top and base surface using IHS Markit Kingdom.The calculated grids were exported as ASCII XYZ data.Final geospatial analysis and visualization were carried out using QGIS version 3.18.0-Zürich.The imported ASCII XYZ data were gridded for visualization using the spatial interpolation algorithm 'Inverse Distance Weighting' with nearest neighbor search radius of 50 m provided by QGIS.

Seismic unit description
Four different seismic units are classified in the seismic dataset based on their acoustic characteristics.Three representative profiles are plotted in Figures 2-4.Further profiles are provided in the Supporting Information (Figs S1 -S4).Depths are given as both TWT and as meters converted with a constant sound velocity of 1600 m s −1 .This constant sound velocity is used due to the uncertainties in the velocity model.The converted depths are therefore not accurate and should be interpreted with care.To ease interpretation, depths in meters are used for analysis of the data.The lake level is defined as reference for the depth scalar as 0 m, or a TWT of 0 s, respectively.
Unit 1 is characterized by chaotic discontinuous to subparallel reflections (Figures 2-4).It shows high amplitudes near the top of the unit, but amplitudes quickly fade out with depth.The unit is exposed at the lake floor only in shallow areas of the lake, whereas it is covered by units 2-4 in the deeper parts of the lake.The boundary between unit 1 and the overlying unit is defined by high-amplitude reflections, which are disrupted in places (Figures 2-4).The disruptions of these reflections are mainly observed in steep sections and are probably artefacts caused by steep inclination and the short streamer (Figures 2-4).The base of unit 1 is not imaged in this dataset due to the limited penetration of our acoustic signal.Profiles oriented perpendicular to the longitudinal axis of the lake show that unit 1 has steep flanks and resembles an eroded valley, which is filled by the other three units (Figures 2 and 3).The longitudinal profile shows an undulating surface at the upper boundary of unit 1 (Figure 4).Unit 1 is present in the entire study area.
Units 2-4 generally show horizontal to subhorizontal reflections (Figures 2-4).These units are a valley fill and onlap on unit 1, where the boundary between the units is steeply inclined.Unit 2 consists of chaotic to low-continuity reflections with generally low amplitudes (Figure 3).This unit fills the lowest part of the eroded valley formed by unit 1 and is overlain by unit 3. The boundary between unit 2 and unit 3 is an unconformity formed by the deepest continuously traceable reflections (Figure 4).Unit 2 is present in the northern and central parts of the lake and only occasionally occurs in the southern part.The maximal thickness of unit 2 is 180 m (Figure 5A).Unit 3 shows continuous wavy reflections which are parallel to subparallel (Figures 2-4).In its upper part, this unit is characterized by medium to high amplitudes, which decrease with depth.Unit 3 is underlain either by unit 2 or by unit 1 in areas where unit 2 is absent (Figures 2-4).Unit 3 is exposed and hence forms the lake floor in some areas.In other areas unit 3 is overlain by unit 4 (Figures 2 and 4).Units 3 and 4 are separated by an unconformity.Unit 3 is present in the entire working area and has a thickness of up to 270 m (Figure 5B).Unit 4 is a mostly transparent unit, bound by continuous parallel reflections with only minor dips (Figures 2  and 4).This unit forms the lake floor in the northern part of the working area where it is a thin drape of up to 20 m on unit 3 (Figure 5C).Unit 4 is absent in the central and southern part of the lake.

Geomorphological analysis of the eroded valley
For a comparison of the shape of the cross-valley profiles (see Figure 6A for location), the boundary between units 1 and 2 is plotted from six seismic transverse profiles in a perspective view (Figure 6B) and from all seismic cross-profiles (Figure 6C).For better comparison, the profiles have been centered around the deepest point of the thalweg.The selected profiles show significant variations in incision depths.The thalweg depths range between 400 and 690 m (Figure 6C) and a decreasing trend of the thalweg towards the south is visible (Figure 6B).Additionally, the shape of the valley also varies.Three different types of valley shapes were distinguished visually (Figure 6).Eight profiles illustrate a U-shaped characteristic, which is wide at its lower part.In contrast, five other profiles exhibit a V-shaped lower part.The remaining five profiles are characterized by an even narrower lowermost V-shaped part that is superimposed by an upper U-shaped part (Figure 6).This narrow lowest part is incised with a depth of 50-80 m and a width of only 250-290 m (Figure 6C).From deep to shallow, it widens to a U-shape in the upper part associated with terraces at least on one side.The distribution of the valley shape types is shown in Figure 6A.The U-shaped valley type is distributed throughout the working area and is absent only at the southern end.The V-shaped type and superimposed type profiles are distributed throughout the entire lake.All three shape types alternate repeatedly and no clusters can be identified.Valley flanks oriented towards Île-René-Levasseur show similar characteristics in morphology (Figure 6C) as their slope angle (on average 35°), slope shape and height of the valley flanks (between 95 and 200 m) are similar.In contrast, the opposite flank shows more variety in the depth of the valley shoulder (between 65 and 385 m) and slope angle (between 10°and 35°) (Figure 6C).
The surface of the eroded valley forming the base of Lake Manicouagan, represented by the top of unit 1, was gridded for geomorphological analysis (Figure 7).All seismic and the chirp sub-bottom profiles were used for reconstruction of this surface.Due to the limit of penetration, the shallow chirp sub-bottom profiles could only be used in areas where unit 1 is shallow and units 2-4 are thin enough or absent.
The valley thalweg, reflecting the present shape of the lake (Figure 1), has one major direction change from NW-SE to N-S (Figure 7) and meanders with a low sinuosity coefficient of 1-1.3.The depth of the valley ranges between 10 and 690 m (Figure 7).A depth of 100 m is observed at the northernmost end of the grid.It increases towards the deeper basins in the northern part of the lake where a depth of up to 640 m is observed (Figure 7B).The basins are separated by local highs, here referred to as ridges.Some of them are shown in the data example (Figure 4) and their locations are plotted in Figure 7 (labeled Rn).The full extent and geometry of these ridges cannot be determined with the limited data coverage.Therefore, it is unclear whether the ridges extend across the entire valley.Five ridges (R1-R5) are located in the northern part of the   1 for location of the profile.The x-axis shows the distance along profile in meters and the y-axis the two-way-traveltime (TWT) in seconds (left side) and the converted depth in meters (right side).A constant seismic velocity of 1600 m s −1 is used for the conversion.Vertical exaggeration (V.E.) is ~17.The interpretation is plotted next to the profile.Four units are identified in the dataset.Unit 1 is interpreted as bedrock.The profile shows that the surface of unit 1 is undulating with local highs of the bedrock, referred to as ridges.These ridges are numbered in the entire dataset from north to south (R1-R16).Units 2, 3 and 4 are interpreted as sedimentary infill of the valley formed by unit 1.The zoom shows the separation of the different units in more detail.[Color figure can be viewed at wileyonlinelibrary.com] lake.A height difference of up to 200 m is observed between the basins and the ridges (Figure 7B).The depth of the central part of the eroded valley varies between 340 m (ridges) and 690 m (basins) and shows a decreasing trend towards the south (Figure 7B).Seven additional ridges (R6-R12) are observed in this part, which further separate the basins (Figure 7).The deepest basin of the lake (between R6 and R7) is located in the central part and reaches a depth of 690 m.The maximum observed height difference between ridge and basin is 350 m.The trend of decreasing depth of the thalweg continues towards the south (Figure 7B).Depth varies between 160 and 590 m in the southern part of the lake.Here, four ridges (R13-R16) are observed (Figure 7B).The longitudinal shape of the eroded valley is not well imaged by the surface grid in the southern part of the lake, due to missing longitudinal seismic profiles in this area (Figure 7).Additionally, local basins and ridges cannot be determined in the entire southern area of the lake.

Analysis of the valley infill
Thickness grids of the valley infill were calculated using all available profiles (Figure 5).The thickness of the valley infill correlates strongly with the shape of the eroded valley.The infill is thickest in the local basins and thin above the ridges.The greatest thickness of 280 m is observed in the northern part of the lake between ridge R2 and R3.A thickness of only 60 m is observed above ridge R2 in this area.An overall decreasing trend towards the south is visible in the thickness of the valley infill, which correlates with the overall trend of a decreasing incision depth of the valley.A maximum thickness of 230 m is observed in the central area of the lake between ridges R6 and R7.In the southern part of the lake, the maximum thickness of the valley infill is 205 m.Only isolated spots of a thick infill are visible in this part (Figure 5D), which probably does not reflect the actual shape of the valley infill due to artefacts caused by the sparse availability of data in this region.Instead, the valley infill is probably continuous in this southern part of the lake, comparable to the other parts of the lake.Overall, with its varying thickness, the valley infill levels the morphology of local basins and ridges.The recent bathymetry (Figure 1) shows only one deep longitudinal basin in the northern and central part of the lake, a shallower basin at the northern end of the southern part and a shallower lake floor in the south of the lake.The recent lake floor shape highlights that the undulating morphology of the eroded valley is leveled by the infill.

Formation of the eroded valley
The subsurface of lakes and basins in North America is often characterized by an acoustic basement with transparent seismic facies as a result of weak or no signal penetration.This lowermost unit is usually interpreted as bedrock or glacial till (Eyles et al., 1991, Eyles et al., 2003, Duchesne et al., 2010, Normandeau et al., 2013, Lajeunesse et al., 2017).In analogy, unit 1 of our dataset, showing weak signal penetration, represents the acoustic basement of Lake Manicouagan.It is interpreted as metamorphic bedrock rather than till deposits based on geological information available from the shore (Murtaugh, 1976) and from excavation of the valley in the south (Lajeunesse, 2014).The bedrock was probably eroded by multiple processes that formed the present valley.These erosive processes probably followed pre-existing morphology that resulted from the meteorite impact into the Greenville units ~214 Ma (Hodych and Dunning, 1992) or are related to even older structural valleys induced by the postulated St. Lawrence rift system (Kumarapeli and Saull, 1966).These theories cannot be verified by this study due to missing penetration of the acoustic signals into the deeper subsurface.In general, direct glacial, subaerial preglacial fluvial and subglacial meltwater erosion processes have the ability to erode bedrock, each resulting in different valley shape characteristics.

Direct glacial erosion
Direct glacial erosion by grounded ice is widely associated with the formation of U-shaped valleys (e.g.Harbor, 1992, Cook and Swift, 2012, Patton et al., 2016).These eroded glacial troughs are oriented parallel to the forward direction of ice flow.For the Manicouagan area, at least three different ice flow directions have been reconstructed for the last glacial episode (Veillette, 2004), outlining the potential of different directions and phases of glacial erosion.Numerous other ice flow events probably occurred in the Manicouagan area, which cannot be reconstructed from erosional bedforms because they were only short-lived, were not erosive or their evidence has been eroded by subsequent glacial events.U-shaped profiles are observed in the dataset from Lake Manicouagan, suggesting that direct glacial erosion shaped a large portion of the existing valley, including its upper sidewalls.The incision depth of erosive ice is crucial for its ability to excavate or deepen a valley and to excavate its potential pre-glacial infill.In our case study, the incision depth of the observed U-shaped erosion varies greatly along the lake valley, ranging from 400 to 690 m, causing several localized glacial overdeepenings.Locally, five seismic cross-profiles reveal a lowest narrow base of the eroded valley with a V-shape incised between 50 and 80 m deeper than the base of the overlying U-shape.This superimposed valley shape type is observed in the northern, central and southern parts of the lake.Therefore, this observation implies that direct glacial erosion did not excavate to the lowest base along the entire lake valley.

Subaerial fluvial erosion
Subaerial fluvial erosion is commonly linked to a V-shaped incision into the strata (Harbor, 1992, Montgomery, 2002).This type of erosion can result in valleys that resemble narrow gorges.Detailed imaging of the base of these narrow gorges is often not feasible using a seismic system as their lateral extent is only in the meter range, which is close to the resolution limit of the seismic system used.Despite constraints in resolution, V-shaped narrow gorges with an incision depth between 50 and 80 m are observed in the seismic dataset from Lake Manicouagan.The width of the top of these narrow gorges ranges between 250 and 290 m.Here, these narrow gorges are superimposed by a large U-shaped depression.This upper Ushaped part of the valley found throughout the lake is similar to theoretical transverse profiles of glaciated valleys derived by Davis (1906) and refined by Martini et al. (2001).The authors propose that a glacier overriding a V-shaped fluvial valley induces the major forces against the flanks of the valley, thereby forming frozen plugs at the top of the flanks and the bottom of the valley.After glacial erosion, the valley then has a lower narrow V-shaped base superimposed by a U-shaped depression and the valley bottom may consist of a combination of glacial and fluvial deposits (Martini et al., 2001).Such narrow V-shaped valleys superimposed by a U-shape have been reported by field studies from several lakes and valleys in North America (Pugin et al., 2019).Lajeunesse et al. (2017) reported a V-shaped incision at the thalweg of a U-shaped valley from Lake Témiscouata.This incision is interpreted as a fluvial channel eroded during a phase of sea level lowstand before the Quaternary glaciations.For the Manicouagan Reservoir, a study based on excavation at the Daniel-Johnson dam construction site and data from the Manicouagan River report a buried valleywithin-valley morphology with a sharp V-shaped base superimposed by a U-shaped valley (Lajeunesse, 2014).Archive photographs show flutes along the well-polished sidewalls of the valley as well as a well-polished bedrock bench and a pothole (Lajeunesse, 2014).The author concludes that the V-shaped valley and the observed bedforms are typical features of a fluvial bedrock channel.These observations from the south of the Manicouagan Reservoir point to subaerial fluvial erosion of the narrow V-shaped valley prior to the last glacial episode.This interpretation fits with the overall conclusion of Lajeunesse (2014) who argues that many gorges and the bottom of V-shaped valleys of eastern Canada were not eroded by the LIS but instead are preserved preglacial fluvial networks.However, our seismic data show that the valley thalweg reaches a depth of 400-690 m below the present lake level, which is 360 m above present sea level.Considering a total glacio-isostatic uplift of about 200 m for Lake Manicouagan and 120 m for the modern shelf (Andrews, 1974), the thalweg was 120-410 m below sea level during the Last Glacial Maximum.This rules out subaerial fluvial erosion in the Quaternary because the fluvial base level is below sea level.Prior to the Quaternary, the thalweg would have had to be higher relative to sea level to allow such fluvial erosion.Further, the seismic data of this study reveal a decreasing trend in depth of the thalweg from north to south, suggesting a slope gradient opposite to the modern fluvial drainage pathway.From this perspective, an interpretation of a subaerial fluvial erosion of the V-shaped valley implies that river flow must have changed direction in the past due to migration of the fluvial catchment (Bishop, 1995) or a former river system must have drained through an adjacent valley.The Toulnustouc River valley in the east of Lake Manicouagan may have been a possible pathway for a fluvial system to flow through the northern and central parts of the working area and then follow the adjacent river valley eastward instead of flowing south to the Manicouagan River.However, our seismic dataset in the souther area has gaps, so this idea remains speculative and cannot be explored further in this study.A reversal of the river flow direction seems possible if the Manicouagan area was part of the Eocene to Pliocene Bell mega-river system that drained North America from the Cordillera to the Hudson Bay (Bell, 1895).However, it is unclear whether the Manicouagan area was part of that river system because reconstruction of the drainage boundary of the Eocene to Pliocene Bell mega-river system has uncertainites as all sedimentary evidence of it was eroded by the LIS (Bell, 1895, Wilson, 1903, Corradino et al., 2021).

Subglacial fluvial erosion by pressurized meltwater
Another explanation for a lower narrow V-shape superimposed by a U-shaped valley is meltwater erosion at the base of an ice sheet under high hydrostatic pressures (Jansen et al., 2014).Such an erosion by meltwater deepens the thalweg of the Ushaped valley and results in a superimposed U-shape above a V-shape (Kehew et al., 2012).In this case, the resulting valley is often termed a 'tunnel valley' (e.g.Ó Cofaigh, 1996, Kehew et al., 2012, van der Vegt et al., 2012).In general, valleys eroded by subglacial meltwater are characterized by an undulating, convex-upward or adverse base (Kehew et al., 2012).In North America, narrow gorge-shaped valleys formed by pressurized meltwater have been interpreted at the base of the New York Finger Lakes (Mullins and Hinchey, 1989).A Vshaped bedrock valley is also reported from Lake McDonald and interpreted as the result of subglacial meltwater flows under high hydrostatic pressure (Mullins et al., 1991).A third example is Lake Okanagan, which is characterized by a V-shaped valley and interpreted to have been filled with subglacial meltwater deposits as well as deglacial and postglacial sediments (Eyles et al., 1991).The observed thalweg of Lake Manicouagan is undulating and shows a decreasing incision depth from north to south, suggesting a slope gradient opposite to the modern fluvial drainage pathway.Both observations are strong indications for subglacial meltwater erosion and they fit the interpretation of overdeepening by glacial erosion.
Alternatively, the valley morphology of Lake Manicouagan could represent an assemblage of erosional remnants from pre-Quaternary fluvial incision, localized overdeepenings and subglacial processes.The alternating distribution of the three classified valley shape types suggests such an assemblage of erosional remnants.With the existing dataset, it is not possible to confirm or refute one or a combination of formation mechanisms for the observed lower narrow V-shaped gorges.It cannot be fully clarified whether they were eroded by a subaerial fluvial system prior to Quaternary glaciations, or by pressured subglacial meltwater during the last glaciation.To address this origin problem, dating the valley infill would allow to narrow down the time of its formation and help distinguish between possible formation mechanisms.

Lake Manicouagan, an overdeepened basin
The seismic dataset acquired at Lake Manicouagan reveals an eroded bedrock valley reaching a depth of 400-690 m.Today, the lake surface is 360 m above present sea level.Consequently, the thalweg is 40-330 m below present sea level, making the lake basin an overdeepened basin.These numbers, however, do not consider glacio-isostatic uplift of ~200 m for the lake basin and only 120 m for the modern shelf area in postglacial times (Andrews, 1974).Even today, the Manicouagan area is uplifted by 5-10 mm yr −1 , whereas the modern coast is no longer uplifting (Peltier et al., 2015).Thus, the thalweg was ~80 m deeper in relation to sea level during the glacial period.The great cumulative depth between 120 and 410 m of this overdeepened basin has probably been predetermined by the meteorite impact and pre-Quaternary erosion processes and is not only the result of glacial erosion.Such overdeepened basins provide accommodation space for sedimentary deposits and are therefore well suited to preserve sediments during repeated glacial cycles (Cook and Swift, 2012).Considering the complex fill of the Lake Manicouagan valley, the sediment sequences shown hold the potential to shed light on the evolution of Lake Manicouagan during cycles of glaciations up to its present state.

Valley sedimentary infill
A sample-based sedimentological analysis of the valley infill is not possible because the sedimentary units of Lake Manicouagan have not yet been drilled.Nevertheless, new insights from the seismic stratigraphy of units 2-4 can be used for interpretation of the sedimentary infill by incorporating the variety of available studies from other lakes.

Deposition of unit 2
Unit 2 is characterized by chaotic reflections with low amplitudes and fills the lowest part of the eroded valley.Similar units have been observed in other lakes and estuaries in North America, South America and Europe (Table S1).For instance, units characterized as mostly transparent to chaotic and discontinuous are reported from studies in Lake St-Joseph (Normandeau et al., 2013), Lake Melville (Syvitski and Lee, 1997), Lago Puyehue (Heirman et al., 2011) and 17 perialpine Swiss lakes (Finckh et al., 1984).Often these units fill the lowermost parts of bedrock troughs and are interpreted as sediments deposited prior to the last glacial episode that were preserved and compacted during ice transgression.Deep drilling in Lake Zurich confirms this interpretation based on the seismic data for the Swiss lakes (Lister 1984b).Such preserved sediments are separated from the overlying deposits by an erosional unconformity (e.g.Burschil et al., 2019).We observe an unconformity as a boundary at the top of unit 2, but no clear evidence for erosion can be imaged, such as crosscutting.Therefore, this unconformity could be the result of an erosion process or a change in depositional conditions.In contrast to the preserved ice-loaded sediment interpretation, comparable units from Lake Okanagan (Eyles et al., 1991), Shuswap Lake (Eyles and Mullins, 1997), Kalamalka Lake (Mullins et al., 1990), the New York Finger Lakes (Mullins and Hinchey, 1989), Lake Annecy (Van Rensbergen et al., 1998, Beck et al., 2001), Lake Le Bourget (van Rensbergen et al., 1999, Chapron et al., 2004) and Lake Neuchâtel (Ndiaye et al., 2014) are interpreted as ice-proximal subaqueous outwash sediments deposited at the end of glaciation as part of the deglaciation sequence.Further, the lowest seismic unit from the St. Lawrence Estuary is described as onlap fill of bedrock troughs with seismically transparent facies containing some low-amplitude reflections (Duchesne et al., 2010).This unit occurs as discontinuous patches in some areas of the St. Lawrence Estuary.St-Onge et al. (2008) hypothesized that this unit is either part of a thick deglacial sequence or sediment of pre-late-Wisconsin age.Therefore, this hypothesis supports both outlined theories.In comparison, these studies suggest that unit 2 from Lake Manicouagan consists of either subaqueous ice-proximal outwash sediments or older pre-Wisconsin sediments deformed by overriding ice.One important constraint for the interpretation of unit 2 is the erosion of the bedrock valley and the excavation depth of the overriding ice sheet during the last glacial period, as discussed above.Ultimately, sediment samples and age dating are needed to answer which of the two interpretations is correct.

Deposition of unit 3
Unit 3 is characterized by parallel to subparallel, continuous wavy reflections.Comparable seismic units characterized by high-amplitude parallel reflections are interpreted as glaciolacustrine deposits in Lake Okanagan (Eyles et al., 1991), the New York Finger Lakes (Mullins and Hinchey, 1989), Lake Témiscouata (Shilts et al., 1992, Lajeunesse et al., 2017), Mazinaw Lake (Eyles et al., 2003), Lake Neuchâtel (Ndiaye et al., 2014) and 17 perialpine Swiss lakes (Finckh et al., 1984).A seismic unit reported from Lake St-Joseph, characterized by parallel, high-amplitude reflections that drape underlying units, is interpreted as sediments deposited in the late phase of the Champlain Sea in a low-energy setting (Normandeau et al., 2013).Another example is a seismic unit from the St. Lawrence Estuary described as a series of parallel high-amplitude reflections, which changes its appearance from flat to wavy to discontinuous depending on the distance to the mouth of the Saguenay River (Duchesne et al., 2010).The seismic unit from the St. Lawrence Estuary is interpreted as a deglacial unit with a change from a glaciomarine iceproximal to an ice-distal setting (St-Onge et al., 2008, Duchesne et al., 2010).Eyles and Mullins (1997) describe a continuous, high-frequency seismic unit with decreasing amplitude from Shuswap Lake and interpret this unit as rhythmically deposited silts and sands transported by underflows.A seismically transparent unit is interpreted as massive silt deposited rapidly during deglaciation from Kalamalka Lake (Mullins et al. 1990).Further, seismically transparent units with laterally continuous reflections are reported from the Great Slave Lake, interpreted as a subglacially deposited unit (Christoffersen et al., 2008), and from Lago Puyehue, interpreted as sub-or proglacial glaciolacustrine sediments (Heirman et al., 2011).Based on the similarities of the units with the descriptions above, unit 3 is interpreted as glaciolacustrine sediments deposited either in a subglacial lake or during ice retreat.No internal unconformities are observed within unit 3, whether truly absent or unresolvable with the acoustic characteristics used.Therefore, we interpret the entire unit, with a thickness of up to 270 m, to reflect a single glacial cycle.

Deposition of unit 4
The uppermost sedimentary unit 4 is a mostly transparent unit, bound by continuous parallel reflections.This uppermost unit drapes the underlying units.Similar units are observed in other lakes and described as units with a draping geometry, some faint parallel reflections and a semi-transparent appearance.These units from Lake Témiscouata (Shilts et al., 1992, Lajeunesse et al., 2017), Lake St-Joseph (Normandeau et al., 2013), Great Slave Lake (Christoffersen et al., 2008), Lake Melville (Syvitski and Lee, 1997), the New York State Finger Lakes (Mullins and Hinchey, 1989) and Mazinaw Lake (Eyles et al.,2003) were interpreted as postglacial sediment.Postglacial sediments with different acoustic facies, however, are described from Lake Okanagan (Eyles et al., 1991), Shuswap Lake (Eyles and Mullins, 1997) and Kalamalka Lake (Mullins et al. 1990).These units are described as well-defined parallel laminations which drape the underlying hummocky topography.The thicknesses of all these postglacial units are reported to be in a range of a few meters up to 22 m (Syvitski and Lee, 1997, Eyles et al., 2003, Christoffersen et al., 2008, Normandeau et al., 2013, Lajeunesse et al., 2017), excluding Lake Okanagan (Eyles et al., 1991) and Shuswap Lake (Eyles and Mullins, 1997) with maximum postglacial thicknesses of 60 and 68 m, respectively.The shallowest unit 4 in Lake Manicouagan has a thickness of up to 20 m, which fits well in the range of reported thicknesses of postglacial units.Therefore, we interpret unit 4 as postglacial sediments.Unit 4 is only imaged in the northern part of the lake, whereas it is not observed in the central and southern part of the lake.This suggests a varying postglacial sedimentation pattern throughout the lake, probably due to modern-day fluvial tributaries in the north of the lake.The expected sediment production in the lake is probably low in the central and southern area because the resulting deposits are below the limit of seismic resolution in the meter range.

Potential sedimentation rates and implications for the deposition age
The overdeepened basin of Lake Manicouagan provides a large accommodation space for sedimentary deposits.Local basins of the bedrock valley contain the thickest sediment deposits, whereas the sediment thickness decreases above the inter-basin ridges.Sedimentation rates from different lakes in a proglacial setting or during glacial retreat are used to discuss the potential timing of sediment deposition.In general, a wide range of deglacial sedimentation rates is reported from different North American Lakes (Table 1), with a minimum sedimentation rate of 0.5 mm a −1 and a maximum rate of 6.3 mm a −1 .Lakes from other geographic settings are also included to show the possible range of deglacial sedimentation rates with a maximum of 110 mm a −1 (Table 1).For Lake Manicouagan, an average sedimentation rate of ~37 mm a −1 would be necessary to deposit the entire sedimentary sequence of up to 280 m in the 7500 years that spans the local deglaciation and postglacial times.Moreover, these rates must be much higher during deglaciation when considering a much lower postglacial sedimentation rate of ~3 mm a −1 or an even lower rate of <1 mm a −1 as reported elsewhere in the southeastern Canadian Shield (Normandeau et al., 2013, Trottier et al., 2019).Considering these reported sedimentation rates from North America, it is rather unlikely that the entire sequence of Lake Manicouagan was deposited during and after the last retreat of the LIS.Sedimentation rates five to six times higher than deglacial sedimentation rates reported would be needed to explain the deposition of the sedimentary sequence during a single phase of ice retreat.However, the sedimentation rate of ~37 mm a −1 for Lake Manicouagan is exactly within the range of rates reported from other geological settings, including modern proglacial lakes, where sedimentation rates can be studied by considering the exact depositional conditions (e.g.Lister, 1984a, Piret et al., 2021).These geological settings in the Alps or Andes are characterized by high relief.Yet, such high deglacial sedimentation rates have not been reported from the Canadian Shield (Table 1) where Quaternary sediment supply seems to be low (Christoffersen et al., 2008).Therefore, it is likely that parts of the sedimentary sequence were deposited prior to the last deglaciation episode.To be able to clarify the discrepancy in sedimentation rates, sediment samples and age dating are indispensable.

Conceptual model for the development of the lake
The analysis and interpretation of the sedimentary units and of the geomorphology of the bedrock basement of Lake Manicouagan allow us to derive a conceptual model consisting of three stages (Figure 8).Prior to the onset of the Quaternary glaciations, the Manicouagan area was shaped by a meteorite impact ~214 Ma, resulting in an impact structure characterized by a circular shape and a central uplift (Hodych and Dunning, 1992).The landscape of the Manicouagan area was probably further modified by pre-Quaternary erosional processes, which presumably predetermined the later development and present-day shape of the lake valley.

Stage I -Erosion of the valley
Two hypotheses are proposed here for the erosion of the valley in the Quaternary.(i) Glacial erosion excavated most of the present-day lake valley.In the process, the ice overrode pre-Quaternary deposits.These deposits could persist under direct glacial erosion in narrow gorges.Such gorges are reported by Lajeunesse (2014) from the south of the Manicouagan Reservoir.An analogy for such preservation is provided by narrow gorges reported from the Alps (Montgomery and Korup 2011).(ii) The entire present-day lake valley was eroded by direct glacial erosion and the lower narrow base was incised by pressurized meltwater drainage during the Wisconsinan or perhaps earlier glaciations.Furthermore, a combination of pre-Quaternary, Steffen Fjord, South America*/modern proglacial lake 21-40 Christoffersen et al., 2008 Great Slave Lake, North America/former subglacial lake 2 Kuhn et al., 2017 Pine Island Bay, Antartica/former subglacial lake 0.057-0.19Hodgson et al., 2009 Lake Hodgson, Antartica/subglacial lake 0.041 Siegfried et al., 2023 Mercer Subglacial Lake, Antartica/subglacial lake 0.49-2.3*Included to show the possible range of sedimentation rate, although these studies investigate a different geological setting.
SEISMIC STRATIGRAPHY OF LAKE MANICOUAGAN glacial and subglacial erosion is also possible, forming an assemblage of erosional remanants.Waxing and waning of the ice sheet are assumed for every glacial cycle, but not outlined in the model.

Stage II -Deposition of the sedimentary sequence
Three different scenarios are discussed below with regard to the deposition of the sedimentary sequence inside the eroded valley (units 2-4).The first scenario suggests that most of the sedimentary sequence was deposited during the last retreat of the LIS.Following this scenario, the sediments forming unit 2 represent subaqueous ice-proximal outwash sediments deposited beneath an ice sheet or during its retreat.Alternatively, unit 2 consists of preserved pre-Quaternary deposits.Unit 3 is deposited during ice retreat in this scenario.Analogies for this scenario include the New York Finger Lakes (Mullins and Hinchey 1989) and the Lakes in British Columbia (Eyles et al. 1991, Eyles and Mullins 1997, Mullins et al. 1990).However, high average deglacial sedimentation rates for North America would be necessary during Lateglacial and postglacial times for this scenario.Therefore, it is possible that parts of the sedimentary succession were deposited prior to the retreat of the last glacial episode, which leads our considerations to the other two scenarios of sediment deposition.
The second scenario proposes that the sediment deposits are the result of multiple glacial-interglacial cycles.Glaciolacustrine sediments forming unit 3 were deposited during glacial retreat, followed by interglacial lacustrine sediments.The deposited sediments were then either locally eroded by a warm-based ice sheet during the next ice advance, or were preserved beneath an overriding cold-based ice sheet.In general, thermal conditions at the glacial bed determine the ability for glacial erosion (Bierman et al., 2014).Cold-based ice is frozen to the ground and therefore is predominantly nonerosive, while warm-based ice can be highly erosive.For example, sediments have persisted under the Greenland Ice Sheet in a cold-based environment for millions of years (Bierman et al., 2014).A lake analog for the preservation of sediment beneath a cold-based ice sheet is Lake CF8 on the Clyde foreland, Baffin Bay (Briner et al., 2007).According to scenario 2, this depositional cycle probably repeatedly occurred several times during the Wisconsin and during earlier glaciations and led to successive deposition of the sedimentary sequence in Lake Manicouagan.Consequently, this scenario implies that Lake Manicouagan may contain sediments older than the last glacial episode.However, the repeated glacial-interglacial cycles should have led to interglacial units and to erosional unconformities within unit 3, if the ice sheet was warm-based; neither are observed in our seismic data.It is possible that these interglacial units were eroded by a subsequent advance of an ice sheet, but this would probably have resulted in erosional unconformities.Therefore, it of debate whether these unconformities and interglacial units are missing in Lake Manicouagan, or whether they are simply not resolvable with the seismic system used.Briner et al. (2007) reported a preserved glacial-interglacial sequence from Lake CF8 on the Clyde foreland (Baffin Bay) with a thickness of 3.2 m, spanning three glacial-interglacial cycles.The individual units have thicknesses between 0.1 and 1.2 m.Based on the acoustic characteristics of the seismic system used in this study, such thin units would not be detectable as individual layers in our data due to the limited vertical resolution in the meter range.In contrast, postglacial unit 4 observed in our dataset has a thickness of up to 20 m and is well imaged with the seismic system used.It is unclear why the older interglacial units should have a significantly lower thickness than the latest interglacial unit.Therefore, scenario 2 is conceivable but rather unlikely.
The third scenario suggests that Lake Manicouagan was a subglacial lake during the last glacial episode.The glaciolacustrine sedimentary unit 3 may have been deposited during the subglacial lake stage.This scenario could explain missing erosional surfaces of repeated glaciations and missing interglacial units.Similar observations of missing unconformities and interglacial units were made in the Great Slave Lake and interpreted as suggesting subglacial lake deposits (Christoffersen et al., 2008), which strongly supports this scenario.However, reported subglacial sedimentation rates of 0.041-2.3mm a −1 are comparatively low (Table 1).This means that Lake Manicouagan must have existed as a subglacial lake for at least 117 000 years asumming that the entire unit 3 (270 m) was deposited with a sedimentation rate of 2.3 mm a −1 during this time.This minimum estimated period, however, is much longer than the time span during which the area was covered by the Québec-Labrador Ice Dome (after 60 ka; Marshall et al., 2000).Therefore, purely subglacial sedimentation for unit 3 is unlikely.The most likely sequence of formation is a combination of subglacial followed by proglacial sedimentation during the last deglaciation period.This sequence would explain the thickness of up to 270 m and represents a combination of scenarios 1 and 3. Sediment sampling and age dating are essential to clarify the deposition history of unit 3.

Stage III -Postglacial deposition
The final stage in our conceptual model is the deposition of postglacial sediment since the last deglaciation 7.5 ka.This Holocene drape forms the shallowest unit 4 representing the modern sedimentary processes in an open-water lake.

Paleoclimate archive potential
Lake Manicouagan is an overdeepened basin directly affected by Quaternary glaciations.The lake holds a thick sedimentary sequence of up to 280 m, and its thickest unit 3 shows no visible internal unconformities in the seismic data.This indicates strongly that Lake Manicouagan holds a mostly undisturbed record of long-term climate variations, including local climate fluctuations since at least the late Pleistocene.In general, proglacial (Carrivick and Tweed, 2013) as well as subglacial (Livingstone et al., 2012) lake deposits have a high potential to reflect short-to long-term Quaternary climate change.Therefore, the sediment infill of Lake Manicouagan has a significant potential to be a paleoclimate archive, assuming that scenario 3 (subglacial lake) or a combination of scenarios 1 and 3 (subglacial followed by proglacial sedimentation) most closely reflects the evolution of the lake basin.However, if the entire sedimentary sequence was deposited during deglaciation only (scenario 1), Lake Manicouagan could be envisaged as a sink for rapidly deposited Lateglacial sediments.In any other scenario, the lake sediments contain information from sedimentation prior to the last deglaciation.Assuming scenario 2 to be correct, the infill of the lake provides an archive of multiple glacial-interglacial cycles albeit with some hiatuses related to erosional periods.Assuming scenario 3 to be correct, Lake Manicouagan is filled with a sedimentary sequence that contains subglacial sediments.Lake Manicouagan therefore has a high potential to serve as a paleoclimate archive.Nevertheless, samples from long sediment cores are needed to validate and refine the suggested scenarios that are currently based solely on geophysical data.

Conclusions
In this paper we report our results on a seismic investigation of the Quaternary stratigraphy of the large and deep Manicouagan impact crater lake, which allows us to discuss the processes responsible for the erosion of its enclosing bedrock valley and assess its potential as a paleoclimate archive.The key results of this investigation are: • The shape of the valley indicates a complex sequence of multiple separate or combined erosional processes.• The observed lower narrow V-shaped gorges were either eroded by pressurized meltwater during the last glacial episode, or are a remnant of pre-Quaternary erosion, or are a combination of both.• The origin of the valley remains to be confirmed by determining the age of the lowermost sedimentary infill.• Lake Manicouagan is an overdeepened basin, thus providing significant accommodation space for a mostly undisturbed sedimentary sequence that reaches 280 m in thickness.
• Three scenarios are proposed regarding deposition of the sedimentary sequence: 1.The entire sedimentary sequence was deposited during the last retreat of the LIS and in a postglacial environment during a single phase.Note that this scenario is only valid for deglacial sedimentation rates five times higher than reported for North America.Such rates have been reported from other geological settings such as the Alps or Andes with a much steeper relief.2. The sedimentary sequence is a result of multiple glacial-interglacial cycles.This scenario would need thin glacial-interglacial layers and erosional unconformities that are beyond the resolution of our seismic system and hence not imaged in our data to hold true. 3. Lake Manicouagan was a subglacial lake during the last glacial episode.It is unlikely that unit 3 was completely deposited in a subglacial regime as the necessary sedimentation rates are far beyond those published.
• We suggest that unit 3 was deposited initially in a subglacial lake regime followed seamlessly by proglacial sedimentation, which would explain its thickness of up to 270 m with reasonable sedimentation rates.The entire deposition of the sedimentary infill would hence have occurred under a combination of scenarios 1 and 3.
Our results allow us to evaluate the significance of subarctic Lake Manicouagan as a paleoclimate archive.Subarctic lakes, in general, have a high potential to serve as climate archives for (pre)glacial times.However, it is important to investigate whether sediments were continuously deposited over time or whether hiatuses in the sedimentary sequence exist, such that deposits do not reflect a continuous paleoclimate archive.Preglacial and interglacial sedimentary units also need to be sheltered from direct glacial erosion, for example in a narrow gorge at the bottom of the lake.Lakes fulfilling these criteria are potential sedimentary archives of short-to long-term Quaternary climate variations, especially if they form subglacial waterbodies during glacial periods.Therefore, we consider Lake Manicouagan to be a significant paleoclimate archive if scenario 3 or a combination of scenarios 1 and 3 is correct.Age dating of samples from long sediment cores is essential to validate the scenarios and to gain detailed insight into valley erosion and sediment deposition at Lake Manicouagan during the Quaternary.

Figure 1 .
Figure 1.(A) Location of the Manicouagan Reservoir in eastern Canada.The GEBCO 2020 Grid (GEBCO Compilation Group 2020) is plotted as background topography.The red box indicates the zoom shown in B. (B) Map of the Manicouagan area showing the present and former shape of Lake Manicouagan (Manikuakan, according to Innu toponymy) and Lake Mushalagan (Mushalakan, according to Innu toponymy) (Dressler and Reimold, 2004).The directions of former ice flows reconstructed by Veillette (2004) are plotted as arrows and the reconstructed ice retreat during the last deglaciation (Dalton et al. 2020) is plotted as dashed lines.The red box indicates the extent of the detailed map shown in C. (C) Map of the working area with plotted seismic profiles, chirp sub-bottom profiles and bathymetry grid.The lake is divided into a northern, central and southern area.Seismic profiles shown in other figures are plotted in bold and labeled.In B and C, the Canadian Digital Elevation Model (DEM), Edition 1.1 (Natural Resources Canada, 2016), is plotted as background topography.[Color figure can be viewed at wileyonlinelibrary.com] Figure 1.(A) Location of the Manicouagan Reservoir in eastern Canada.The GEBCO 2020 Grid (GEBCO Compilation Group 2020) is plotted as background topography.The red box indicates the zoom shown in B. (B) Map of the Manicouagan area showing the present and former shape of Lake Manicouagan (Manikuakan, according to Innu toponymy) and Lake Mushalagan (Mushalakan, according to Innu toponymy) (Dressler and Reimold, 2004).The directions of former ice flows reconstructed by Veillette (2004) are plotted as arrows and the reconstructed ice retreat during the last deglaciation (Dalton et al. 2020) is plotted as dashed lines.The red box indicates the extent of the detailed map shown in C. (C) Map of the working area with plotted seismic profiles, chirp sub-bottom profiles and bathymetry grid.The lake is divided into a northern, central and southern area.Seismic profiles shown in other figures are plotted in bold and labeled.In B and C, the Canadian Digital Elevation Model (DEM), Edition 1.1 (Natural Resources Canada, 2016), is plotted as background topography.[Color figure can be viewed at wileyonlinelibrary.com]

Figure 2 .
Figure 2. Seismic cross-profile AWI-20168150 from the northern part of the lake.The location of the profiles is plotted in Figure 1.The x-axis shows the distance along profile in meters and the y-axis the two-way-traveltime (TWT) in seconds (left side) and the converted depth in meters (right side).A constant seismic velocity of 1600 m s −1 is used for the conversion.Vertical exaggeration (V.E.) is ~3.The interpretation is plotted next to the profile.Three units are imaged on this profile: Unit 1 is interpreted as bedrock, and units 3 and 4 are interpreted as sedimentary infill of the valley formed by unit 1.The zoom shows the separation of units 1 and 3 in more detail.[Color figure can be viewed at wileyonlinelibrary.com]

Figure 3 .
Figure 3. Seismic cross-profile AWI-20168260 from the southern part of the lake.The location of the profile is plotted in Figure 1.The x-axis shows the distance along profile in meters and the y-axis the two-way-traveltime (TWT) in seconds (left side) and the converted depth in meters (right side).A constant seismic velocity of 1600 m s −1 is used for the conversion.Vertical exaggeration (V.E.) is ~3.The interpretation is plotted next to the profile.Three units are shown on this profile: Unit 1 is interpreted as bedrock, and units 2 and 3 are interpreted as sedimentary infill of the valley formed by unit 1.The zoom shows the separation of the different units in more detail.[Color figure can be viewed at wileyonlinelibrary.com]

Figure 4 .
Figure4.Seismic along profile AWI-20168250 from the northern part of the lake.See Figure1for location of the profile.The x-axis shows the distance along profile in meters and the y-axis the two-way-traveltime (TWT) in seconds (left side) and the converted depth in meters (right side).A constant seismic velocity of 1600 m s −1 is used for the conversion.Vertical exaggeration (V.E.) is ~17.The interpretation is plotted next to the profile.Four units are identified in the dataset.Unit 1 is interpreted as bedrock.The profile shows that the surface of unit 1 is undulating with local highs of the bedrock, referred to as ridges.These ridges are numbered in the entire dataset from north to south (R1-R16).Units 2, 3 and 4 are interpreted as sedimentary infill of the valley formed by unit 1.The zoom shows the separation of the different units in more detail.[Color figure can be viewed at wileyonlinelibrary.com]

Figure 5 .
Figure 5. Isopach maps of the valley infill units (A-C) and the cumulative valley infill (D).The Canadian Digital Elevation Model (DEM), Edition 1.1 (Natural Resources Canada, 2016), is plotted as background topography.Unit 2 (A) has a thickness of up to 180 m and is present in the northern and central part of the working area and occasionally occurs in the southern area.Unit 3 (B) has a thickness of up to 270 m and is present in the entire working area.Unit 4 (C) is present only in the northern part of the working area and has a thickness of up to 20 m.The cumulative valley infill (D) has a maximum thickness of 280 m.In general, the sediment thickness decreases towards the south, which corelates with the decreasing depth of the bedrock valley and is thereby linked to the decreasing accommodation space.[Color figure can be viewed at wileyonlinelibrary.com]

Figure 6 .
Figure 6.(A) Map showing the distribution of the visually classified different types of valley shapes.The cross-profiles are colored from north (dark blue) to south (light green) to allow a separation in the other subplots (B, C).The Canadian Digital Elevation Model (DEM), Edition 1.1 (Natural Resources Canada, 2016), is plotted as background topography (B) Picked surface of the eroded valley in a perspective view from six seismic crossprofiles.Location of the profiles is shown in A. The deepest point of every profile is assigned to 0 m distance along the profile.The varying shape of the bedrock valley is outlined.Further, the upslope direction of the thalweg is indicated by the black arrows.(C) Picked surface of the eroded valley from all seismic cross-profiles.The x-axis shows the distance along the profile in meters and the y-axis the converted depth in meters.The presentday lake level is used as reference of zero for depth estimation.The deepest point of every profile is assigned to 0 m distance along the profile.The different profiles are colored to allow separation.The plot illustrates the varying shape of the bedrock valley and the different shapes of the flanks.[Color figure can be viewed at wileyonlinelibrary.com]

Figure 7 .
Figure 7. (A) Plot of the picked surface of the eroded valley (unit 1).Note the data gaps in the southern part of the working area.The depth of the eroded valley is plotted in meters and varies between 10 and 690 m.The present-day lake level is used as reference of zero for depth estimation.The deepest incisions are visible in the northern part of the lake.The depth of the valley decreases towards the south.Picked ridges (local highs) are plotted as an overlay in white and labeled from north to south (R1-R16).The Canadian Digital Elevation Model (DEM), Edition 1.1 (Natural Resources Canada, 2016), is plotted as background topography.(B) Thalweg profiles picked through the eroded valley (white dashed line in A).The profiles show the decreasing trend of the eroded valley from north to south and the ridges.[Color figure can be viewed at wileyonlinelibrary.com]

Figure 8 .
Figure 8. Conceptual model for the evolution of Lake Manicouagan consisting of three stages.Three different scenarios for deposition of the sedimentary sequence are illustrated in stage II.Scenario 1: the entire sedimentary sequence is deposited during the last retreat of the LIS and during a postglacial environment.Scenario 2: the sediment deposits are the result of multiple glacial-interglacial cycles.Scenario 3: most of the sediments were deposited during a subglacial lake stage.[Color figure can be viewed at wileyonlinelibrary.com]

Table 1 .
Sedimentation rates from different lakes in a proglacial or subglacial setting or during glacial retreat for a comparison with the estimated sedimentation rate from Lake Manicouagan.