North American Ice Sheet build-up during the last glacial cycle, 115–21 kyr

lable at ScienceDirect

Quaternary Science Reviews 29 (2010) 2036e2051

Contents lists avai

Quaternary Science Reviews

journal homepage: www.elsevier .com/locate/quascirev

North American Ice Sheet build-up during the last glacial cycle, 115e21 kyr

Johan Kleman a,*, Krister Jansson a, Hernán De Angelis a, Arjen P. Stroeven a, Clas Hättestrand a,Göran Alm a, Neil Glasser b

aDepartment of Physical Geography and Quaternary Geology, Stockholm University, S-106 91 Stockholm, Swedenb Institute of Geography and Earth Sciences, Aberystwyth University, Aberystwyth, Ceredigion, Wales SY23 3DB, UK

a r t i c l e i n f o

Article history:Received 13 July 2009Received in revised form20 February 2010Accepted 23 April 2010

* Corresponding author. Tel.: þ46 8 164813.E-mail address: [email protected] (J. Kle

0277-3791/$ e see front matter � 2010 Elsevier Ltd.doi:10.1016/j.quascirev.2010.04.021

a b s t r a c t

The last glacial maximum (LGM) outline and subsequent retreat pattern (21e7 kyr) of North Americanice sheets are reasonably well established. However, the evolution of the ice sheets during their build-upphase towards the LGM between 115 and 21 kyr has remained elusive, making it difficult to verifynumerical ice sheet models for this important time interval. In this paper we outline the pre-LGM icesheet evolution of the Laurentide and Cordilleran ice sheets by using glacial geological and geomor-phological records to make a first-order reconstruction of ice sheet extent and flow pattern. We mappedthe entire area covered by the Laurentide and Cordilleran ice sheets in Landsat MSS images andapproximately 40% of this area in higher resolution Landsat ETMþ images. Mapping in aerial photo-graphs added further detail primarily in Quebec-Labrador, the Cordilleran region, and on Baffin Island.Our analysis includes the recognition of approximately 500 relative-age relationships from crosscuttinglineations. Together with previously published striae and till fabric data, these are used as the basis forrelative-age assignments of regional flow patterns. For the reconstruction of the most probable ice sheetevolution sequence we employ a stepwise inversion scheme with a clearly defined strategy for delin-eating coherent landforms swarms (reflecting flow direction and configuration), and linking these topreviously published constraints on relative and absolute chronology. Our results reveal that ice-dispersal centres in Keewatin and Quebec were dynamically independent for most of pre-LGM time andthat a massive Quebec dispersal centre, rivalling the LGM in extent, existed at times when the SW sectorof the ice sheet had not yet developed. The oldest flow system in eastern Quebec-Labrador (Atlanticswarm had an ice divide closer to the Labrador coast than later configurations). A northern Keewatin-Central Arctic Ice Sheet existed prior to the LGM, but is poorly chronologically constrained. There is alsoevidence for older and more easterly Cordilleran Ice Sheet divide locations than those that prevailedduring the Late Wisconsinan. In terms of ice sheet build-up dynamics, it appears that “residual” ice capsafter warming phases may have played an important role. In particular, the location and size of remnantice masses at the end of major interstadials, i.e. OIS 5c and 5a, must have been critical for subsequentbuild-up patterns, because such remnant “uplands” may have fostered much more rapid ice sheetgrowth than what would have occurred on a fully deglaciated terrain. The ice-sheet configuration duringstadials would also be governed largely by the additional topography that such “residual” ice constitutesbecause of inherent mass balanceetopography feedbacks.

� 2010 Elsevier Ltd. All rights reserved.

1. Introduction

The last glacialmaximum (LGM) outlines and subsequent retreatpatterns (21e7 kyr) of the North American ice sheets are reasonablywell established. However, the evolution of these ice sheets duringtheir build-up towards the LGM (115e21 kyr) has remained elusivebecause of the fragmented nature of the older landform record andconflicting age assignments of key stratigraphical sequences. This

man).

All rights reserved.

situation has left numerical ice sheet models unverifiable for thistime interval and hampers General Circulation Model experimentsbecause of the substantial uncertainty in ice sheet configuration andtopography through the long ice-growth phase. Our currentunderstanding of North American ice sheet evolution stems fromtwo distinctly different approaches.

The first approach is grounded in the compilation of geologicalevidence for, and timing of, the extent of the Laurentide andCordilleran ice sheets (LIS, CIS). Dyke and Prest (1987) presenteda reconstruction of the post-LGM evolution of the LIS basedprimarily on radiocarbon constraints for local deglaciation. In their

mailto:[email protected]

www.sciencedirect.com/science/journal/02773791

http://www.elsevier.com/locate/quascirev

J. Kleman et al. / Quaternary Science Reviews 29 (2010) 2036e2051 2037

reconstruction, radiocarbon dates constrain the outline of the icesheet at successive time intervals and the landform recordwas usedto infer changes in ice-flow patterns. Dyke et al. (2002) presentedthe most comprehensive synthesis concerning the LGM configu-ration of the LIS, which portrayed a ridge or high saddle betweenthe Keewatin and Quebec dispersal centres, and also suggested OIS3 outlines. Dredge and Thorleifson (1987) attempted to reconstructthe ice sheet outline during the Middle Wisconsinan and arrived atthree alternative reconstructions, ranging from a coherent icesheet, intact over Hudson Bay, to a scenario with smaller remnantdomes in Keewatin and Quebec. They concluded that the datingevidence was inadequate to distinguish clearly between thesereconstructions, and pointed to the Hudson Bay Lowlands (HBL) asa key area for understanding the evolution of the LIS during thebuild-up phase towards its LGM configuration. Clark et al. (1993)summarized the available stratigraphical information pertainingto the pre-LGM part of the last glacial cycle in an effort to constrainice sheet configurations during that time. Like Dredge andThorleifson (1987), they also presented alternatives for the evolu-tion of ice cover in the HBL. Hence, due to its critical location and itsrich but poorly dated stratigraphical record, HBL is clearly thesingle-most important region for constraining the shape andvolume evolution of the proto-LIS during the build-up phase. Thework of Boulton and Clark (1990) represented a new approach tousing the geomorphological record to reconstruct former icesheets, primarily because the glacial lineation record was consid-ered to consist of an amalgamation of discrete patches of glaciallineations (flow sets) formed at different times during the lastglacial cycle. Using Landsat MSS imagery to map flow sets andaerial photography for relative-age determinations of crosscuttinglineation sets, they portrayed LIS evolution as a stack of events,marked by major shifts in the location of dispersal centres. Thedeglaciation pattern was not traced, and meltwater-generatedlandforms were not used in the analysis.

The second approach is through numerical ice sheet modelling.Reconstructions of the LIS at the LGM based on numerical ice sheetmodelling have ranged froma thickmono-domed ice sheet (Dentonand Hughes, 1981) to a relatively thin multi-domed ice sheet (Clarket al., 1996). More recent ice sheet models (Marshall et al., 2000,2002; Tarasov and Peltier, 2004) seem to converge on the LIS atthe LGM exhibiting (i) a multi-domed shape with a complexconfiguration of ice streams and inter-stream ridges in peripheralareas, (ii) frozen-bed core areas under the main domes, and (iii) icethicknesses substantially smaller than for the early monodomereconstructions. Few numerical modelling exercises focus on, orinclude, the growth phase of, the ice sheet (Huybrechts andT’siobbel,1995; Charbit et al., 2007; Bintanja and van deWal, 2008).

These numerical ice sheet models commonly conflict with thegeological evidence for ice sheet extent by showing; (i) too muchice cover in Alaska and the Mackenzie mountains (NW sector),areas that were largely ice-free at the LGM, (ii) continuous andmassive glaciation of British Columbia instead of the geologicallyindicated ephemeral ice cover (Clague, 1989), and (iii) an inabilityto generate a sufficiently large SE extent of the Quebec dome duringice sheet build-up (Kleman et al., 2002) without overshootingdocumented ice margins in other areas. The problem of validatingnumerical models of the pre-LGM North American ice sheets hasbeen confounded by the scarcity and extremely uneven spatialcoverage of reliable geological control on ice sheet extent betweenthe last interglacial and the LGM (Clark et al., 1993).

In an effort to reconstruct a credible build-up history of NorthAmerican ice sheets, adequate for the validation of numerical ice-sheetmodel predictions,wepresent anewglacial geomorphologicaldata set for the LIS and CIS areas (Figs. 1 and 2), and performa comprehensive integration of this data set with previously

published stratigraphical and chronological evidence. In contrast toprimarily dating-driven reconstructions, we combine the extensivelandform record, which reflects ice-flow organisation, with pub-lished dating evidence, which we, where possible, correlate toidentifiable flow events. We accordingly aim to present a coherenttime-slice reconstruction of the LIS andCIS for the 115e21 kyr build-up phase towards the LGM. We report the results on two levels ofconfidence and generalisation; (i) a robust configuration scheme,which rests primarily on relative-age determinations but is spatiallyincomplete due to the fragmented nature of the landform record,and (ii) a more complete but also more speculative event sequence,where ice sheet outlines are shown for a number of defined timeslices. The results on level (ii) represent the most likely recon-structed pre-LGM evolution of North American ice sheets, inte-grating available stratigraphical, chronological andgeomorphological evidence. For (i) wemake no inference regardingthe absolute age of flow trace swarms. Themore robust but spatiallyand temporally incomplete level (i) results are fully embraced in themore speculative and interpretative level (ii) results.

1.1. The challenge

We note that ice sheet reconstructions for the 115e21 kyr build-up phase are far more difficult to achieve than reconstructions forthe LGM and deglacial time, for the following reasons:

- The time interval is much longer, and can rarely be addressedwith radiocarbon dates.

- All landforms and deposits created in this time interval havebeen subjected to overriding ice during the LGM and laterstages, which in most cases has led to full or partial destructionand reshaping. Furthermore, overriding may have resulted incomplete burial of pre-existing landforms but may equally sohave resulted in a complete absence of new sediments orlandforms due to cold-based conditions.

- The landform record is, therefore, spatially more fragmentedthan for the LGM and deglacial periods (Kleman et al., 2006).

- The number of stratigraphical sections reflecting this timeinterval is much smaller.

These difficulties add up to a formidable methodological chal-lenge, and require a focus on only the first-order patterns in theevolution of ice sheets in this time interval. The reliability andspatial and temporal precisionwhich has been achieved for aspectsof post-LGM ice sheet evolution (Dyke et al., 2003; Ehlers andGibbard, 2003) is unattainable for the 115e21 kyr interval.

2. Materials and methods

2.1. The data set

An important source of information concerning pre-LGM glacialevents is the almost omni-present record of glacial lineations(Figs. 1 and 2), represented in drift as till lineations or in bedrock asglacial striae. Their value for ice-sheet reconstructions lies in thefact that whereas glacier flow is laminar and highly organised overlarge areas, it also shifts over time in response to climatically drivenice-sheet configuration changes and internally driven ice dynam-ical changes. The most important ice dynamical condition is thesubglacial thermal organisation (Kleman and Stroeven, 1997;Kleman and Glasser, 2007).

The landform record used in this study was extracted mainlythrough new mapping. More than 80,000 lineations were mappedin MSS satellite imagery, but, for cartographic reasons, onlyw30,000 are shown in Fig. 1a. We reduced the number of lineations

Fig. 1. a) Glacial landforms in the LIS area. b) Locations of sites with crosscutting till lineations.

J. Kleman et al. / Quaternary Science Reviews 29 (2010) 2036e20512038

using normal cartographic principles, by thinning symbol densitywithout affecting the outline and directional information of thelandform systems. Around 90% of the CIS area, and approximately40% of the terrestrial LIS area, were also mapped in Landsat ETMþimagery. Mapping from aerial photographs was used to add furtherdetail in some key areas. In addition, information on till lineationsfrom the Glacial Map of Canada (GMC; Prest et al., 1968) wasadopted for the Interior Plains (IP), parts of the HBL, the Great Lakesregion, Quebec-Labrador, and Newfoundland. Eskers, terminalmoraines and ribbed moraine were extracted from the GMC, withimportant amendments from satellite and aerial photographinterpretations.

A key feature of our analysis includes the recognition of around500 locations (through an analysis of 160,000 microfilmed aerial

photographs) where relative-age relationships can be establishedfrom crosscutting lineations. These locations, together with previ-ously published striae and till fabric data from key sites, comprisea set of around 700 locations where crosscutting relationships havebeen documented. Thesewere used for the relative-age assignment(i.e. stacking) of flow patterns. Previously published striae and tillfabric observations were added to the information derived from tilllineations in defining flow-directional swarms in cases where theyprovided important additional information. Data from fieldmapping of crosscutting sets of striae and till fabric are includedfrom central Quebec-Labrador, southern British Columbia, and theYukon. In essence, our glacial landform map shows a much greaterresemblance to the map of Prest et al. (1968) than to that presentedby Boulton and Clark (1990). We note that crosscutting lineation

Fig. 2. Main glacial lineation systems and preserved fragments of older lineationsystems (red lines) in the Cordilleran Ice Sheet area. (For interpretation of the referencesto colour in this figure legend, the reader is referred to the web version of this article.)


systems are underrepresented on the GMC, because the carto-graphic principle employed in the construction of that map wasthat in the case of crosscutting lineations, both sets were only puton the map if the landforms in the two sets were of approximatelyequal size (Douglas Grant, personal communication 1998). We havefailed to find evidence for some of the lineation sets mapped byBoulton and Clark (1990). It should also be noted that some of ourmapped swarms are composite, in the sense that they are definedboth by landforms and regional sets of striae.

2.2. Classification and spatial delineation of flow systems (swarms)

Using established palaeoglaciological principles, we linkparticular glacial landforms (Fig. 3) to specific glaciodynamicconditions (Kleman et al., 2006; Stokes et al., 2009): (i) Till linea-tions, which form when the ice sheet bed is thawed and sufficient

subglacial debris is available. (ii) Terminal moraines, which reflectdeposition during standstills of the ice margin or slow recessionduring a quasi-balanced state of the ice-sheet. (iii) Eskers, whichmark melting, mass loss, and extensive subglacial drainage, formtime-transgressively close to the ice margin during ice-sheetretreat. (iv) Ribbed moraine, which is considered to reflect inward-transgressive subglacial thawing (Hättestrand and Kleman, 1999;Kleman and Hättestrand, 1999; Finlayson and Bradwell, 2008;Van Landeghem et al., 2008)

The main components in the inversion procedure are calledswarms (Kleman et al., 2006). These represent glacial landformsystems with internally coherent morphological characteristics.Swarms are defined on the basis of spatial continuity and theconformance to a glaciologically plausible flow pattern, i.e.a minimum-complexity assumption. The swarm types recognisedare: (i) The deglacial envelope, which embraces the entire glaciatedarea, is commonly defined by landform swarms that consists of tilllineations with eskers, glacial lake traces and ribbed moraine.Although till lineations are not deglacial features as such, they areinterpreted to have formed close to the retreating ice margin whenthey are well aligned with deglacial meltwater landforms (Klemanand Borgström, 1996; Kleman et al., 1997, 2006), (ii) “Event”swarms, which are defined by landform systems which compriseabundant flow traces but lack aligned meltwater traces, and, hence,were created well inside the ice margin, and (iii) Ice-streamswarms, which are identified using established criteria (Stokes andClark, 1999), such as well defined lateral margins, high elongationratios of lineations in ice-stream trunks, and strong flow conver-gence in ice-stream heads. There is an important differencebetween our swarms and the flow sets of Boulton and Clark (1990).The flow sets defined by Boulton and Clark (1990) are based solelyon lineations, whereas we also consider the distribution of ribbedmoraine and meltwater traces in defining three types of landformswarms (Kleman et al., 2006), an approach similar to that of Stokeset al. (2009). A crucial uncertainty is whether a similar alignment oflandforms really indicates that the landforms are of a similar age, i.e. can we trust correlations based on directional evidence? Whenthe morphology of the landforms in question is considered inrelation to the average morphology of the host swarms and theoutline of the swarms, robust conclusion regarding the viability ofcorrelation can usually be reached. We therefore regard similarityin orientation (within 10�) in close proximity, as good evidence forsimilarity in age.

2.3. Deciphering procedure

We aggregate the swarms to form glaciologically plausible icesheet flow patterns and place them in the proper relative chro-nology according to the crosscutting relationships in the data set. Inthe final analysis step, these swarms are correlated to publishedchronological constraints, to arrive at the most plausible evolutionsequence of ice sheet configurations.

For the particular application of probing into the proto-LISevolution, the inversion scheme of Kleman et al. (2006) has beenamended with additional considerations. We have done this inresponse to the vastly increased complexity of North American icesheets evolution when compared to analysing only sub-regions oflarge ice sheets, ice sheets of a smaller size, and/or those havinga simpler topographical and climatological context, such as theFennoscandian Ice Sheet (Kleman et al., 1997). We have focused theanalysis on a restricted number of ice-flow patterns of regionalsignificance. Directional differences between these ice-flowpatterns reflect flow-directional changes that are likely related tosignificant changes in ice-sheet outline or dome location. Thisconsideration has led to a restrictive use of “coastal” data, because

Fig. 3. Landsat ETMþ images of: a) Till lineations on Belcher Islands, Hudson Bay, recording Lateglacial ice flow towards southwest. Truly parallel glacial lineations, witha “smeared” appearance, are easily separated from bedrock structural trends. b) Key landforms in the reconstruction of NAIS; till lineations, eskers, and terminal moraines in an arealocated 50 km north of Great Bear Lake, NWT, Canada. c) Crosscutting till lineations south of Baker Lake reveal the relative age of two ice-flow events. Small westward-trendinglineations, forming part of the deglacial Dubawnt ice-stream swarm are superimposed on large northward-trending older lineations.

Fig. 4. The retreat pattern of the Laurentide Ice Sheet. The 22e9 kyr interval is basedon Dyke et al. (2003). The spatial pattern of the final, post-9 kyr, retreat in Keewatin,Foxe Basin and Quebec is based on detailed regional reconstructions (Jansson et al.,2002, 2003; De Angelis and Kleman, 2005, 2007). For the purpose of restricting theanalysis to the pre-LGM stages, we first identified and excluded the landform swarmsthat due to a conformable pattern were judged to have formed sometime during theshown post-LGM retreat sequence. The most important of the landform swarms thatsurvived this initial age-screening are shown in Fig. 6.


coastal or near-coastal flow sequences typically reflect changes incoastal or continental shelf drainage routes but in most cases havelimited explanatory power for the locations of distant inlanddomes. We have in all but two cases, both of them concerning thesouthern part of the Quebec dome, avoided correlation of flowsystems on opposite sides of “apparent” ice divides. In the absenceof absolute age control on tills, it is usually unreliable to correlateopposing flow patterns on the basis of inferred synchronicity. Wehave left swarms of restricted size and with unclear relative andabsolute chronology as residuals, rather than including themspeculatively into sequences. These residuals represent primecandidates for future research.

For the CIS region we used a slightly different appoach, inresponse to the fact that ice-flow pattern in this mountainousregion almost everywhere conforms strongly to topography, whichprompted radically different flow conditions in valleys andlowlands compared to uplands. Ice covers on uplands were thinner,often exhibited divergent flow and were likely cold-based. Wediscern a unified morphological imprint, consisting of five spatiallyseparated swarms. Whether this pattern is of deglacial age orreflects earlier parts of the LateWisconsinan is unclear (see below),and we therefore refrain from labelling it as a composite deglacialswarm. Fragments of older flow systems primarily occur preservedunder LGM ice divides or cold-based/slow-flow zones of the LateWisconsinan.We have emphasized the fairly abrupt limits betweenareas exhibiting few or no Late Wisconsinan flow traces, and thosewhich bear a strong Late Wisconsinan imprint. The lateral limits ofswarms therefore have different implications in the CIS (Fig. 2) andLIS (Fig. 1) areas. In the CIS map they mark the interpreted laterallimits of strong or erosive flow, whereas in the LIS areas the lateralswarm limits simply mark the limit of observations.

The overall post-LGM retreat pattern until 9 kyr B.P, based onthe largely dating-based reconstruction by Dyke et al. (2003) isshown in Fig. 4. For the purpose of restricting the analysis to thepre-LGM (115e21 kyr) time frame, we first identified the landformswarms (the deglacial envelope and ice-stream swarms) thatreasonably fit this documented post-LGM retreat pattern, andexcluded these from the further analysis. The foremost criterion foraccepting a swarm as part of the retreat pattern was a generalconformity of esker and lineation trends to the dating-driven ice-marginal retreat pattern (i.e. generally perpendicular).

In Fig. 5 we display some of the pre-LGM swarms that havea clear morphological manifestation. The ice-sheet wide distribu-tion of pre-LGM swarms remaining after the initial sorting processdescribed above is shown in Fig. 6. In Fig. 7 we show the individual

swarms, together with a scheme giving their relative age and theminimum extent of ice required by each swarm.

2.4. The swarms

The swarms onwhich the analysis of the 115e21 kyr evolution isbased (Fig. 5) arepresented inanascendingageorder. This is becausedeciphering of the glacial geological and geomorphological record is“archaeological” in nature, with younger and more complete infor-mation layers being successively “peeled off” to reveal the morefragmentary older record of preserved or incompletely erasedlandforms and deposits. Thus, with increasing time depth, traces ofglacial events become fewer and more difficult to correlate. Thenomenclature used in the description of ice-flow patterns is asfollows: For those swarms that can be confidently correlated topreviously described glacial sediments, we adopted previouslyaccepted nomenclature in the Laurentide area (Severn, Sachigo,Rocksand, Shagamu, Caledonian; Fig. 6).We assigned newnames toice-flowpatterns that have not beenpreviously described or named,

Fig. 5. Key landform swarms. a) The Aberdeen swarm is defined by lineations reflecting a dispersal centre in northeastern Keewatin. The deglacial flow has left little or no imprint inthe picture area south of the Dubawnt ice-stream swarm. b) The Atlantic swarm occurrs as the oldest morphological sratum at scattered location in south-central Quebec. In thepictured area its direction differ by 60� from the deglacial (generally referred to as the SW-system) flow that is indicated by the direction of eskers. c) The Churchill swarm occurs asa preserved fragment between zones reflecting deglacial flow from the north. d) Heavily till-covered area near Pointe Louis XIV, northeasternmost James Bay. The smaller lineationsreflect deglacial flow, but underlying “ghosts” are compatible with the Sachigo, and possibly also the Rocksand flow directions. The primary evidence for the latter flow directionsare striae and till fabric data from a wider region surrounding James Bay. For location of photos, see Fig. 6.


referring to type-regions where they are particularly prominentlyexpressed (Aberdeen, Garry, Churchill, Atlantic, Fundy; Fig. 6). TheCordilleran swarms have all been given new names, based onmorphological and flow pattern coherence, without attempts toadhere to previous chronological assignments (Fig. 2). We alsoemploy two additional well-established HBL stratigraphical desig-nations that pertain to regionally ice-free conditions. This is some-what inconsistent but necessary, because of the great importance ofthe Prest Sea and Bell Sea sediments in the interpretation of the HBLstratigraphical sequences.

The importance of each data type is different for different timeintervals. For the deglacial pattern (Fig. 4), a coherent pictureemerges from dating of the first occurrence of ice-free conditions(Dyke et al., 2003), from landforms, sediments and striae reflectinglate ice flow, and, independently, from landforms reflecting theinward-transgression of the meltwater system (eskers, channels,glacial lake traces). The extensive database on crosscutting linea-tions primarily aided in separating older lineation sets from Late-glacial or deglacial lineations. The age relationships between older

lineation swarms (those reported in Figs. 7 and 8) were primarilybased on published striae and till fabric data.

For the LGM, the spatial extent is reasonably well constrained bydated end moraines and extra-marginal radiocarbon evidence. Thearrangement of ice sheet domes and divides at the LGM is moreuncertain, and was inferred on the basis of patterns of major icestreams terminating at the continental shelf break and at terrestrialLGM margins. The lack of a dating method for flow traces thatformed well inside the ice sheet margin causes considerableuncertainty regarding themore precise interior flow patterns at theLGM.

For older events, dated evidence of ice margin locations isextremely scarce or totally lacking in most regions, leaving thespatial extent and pattern of older till lineations, till fabrics, andstriae as the only spatial information sources. Correlations based ontill properties are of local importance, but do not permit correla-tions over long distances.

For each swarm we discuss the evidence defining the swarm,the relative-age constraints, and the glaciological implications of

Fig. 6. Landform swarm patterns considered to be older than the LGM. The swarms are described in the text and were used to reconstruct the ice sheet configurations in Fig. 7 andthe event sequence in Fig. 8.


the described ice-flow pattern. We then discuss published ageconstraints and possible correlations for each swarm.

3. Results

3.1. LGM

Fig. 7a shows the location ofmajor ice streams terminating at theLGM margin (Kleman and Glasser, 2007) and Fig. 8a shows ourpreferred reconstruction of the LGM ice sheet based on thisevidence. The form line indicating the approximate interior layout isbased on the three main considerations of realistic size of drainagesfor themajor ice-stream corridors, a northesouth asymmetry in theice-sheet caused by a higher fraction of thawed bed in the moresoutherly areas, and the effects of high topography in the north-eastern sector, particularly Baffin Island and the Torngatmountains.

In the area of confluence between the LIS and the CIS, tilllineations trending NNWand SSE, parallell to the Rocky Mountainfront, occur scattered over a large region. These lineations,together with the oldest ice-stream traces in the McKenziecorridor indicate an ice divide over the Interior Plains, approxi-mately at the latitude of Lake Athabasca. These flow patternsindicate that an ENEeWSW oriented ice divide existed over the IPat some time, as previously suggested by Dyke et al. (1982). Thisflow configuration remains undated, with one of the fewconstraints on its age being the subsequent establishment of theLate Wisconsinan Keewatin Dome. Cosmogenic nuclide exposureages (Jackson et al., 1997) on the foothills erratics train (Stalker,1956) place the age of confluence between Cordilleran and Lau-rentide ice firmly in the Late Wisconsinan. Hence, we havetentatively assigned this pattern to near-LGM time, and suggestthat radial flow from a Keewatin Dome resumed only after theLGM, but acknowledge substantial uncertainty about the age ofthe NNWeSSE flow pattern across the IP.

We have reconstructed a generally EeW trending main icedivide joining the Keewatin and Quebec-Labrador domes. Theelevation of this divide in its central part may have fluctuated in

concert with the quasi-cyclic Heinrich events, provided that thelatter were a consequence of pulses of enhanced flow of the HudsonStrait Ice Stream (Alley and MacAyeal, 1994; Hemming and Hajdas,2003; Hemming, 2004; Hulbe et al., 2004). Because no existingdating method has a direct bearing on deep interior subglacialevents (such as the formation of striae and till lineations), any effortto pinpoint the precise location of the main ice divide at a givenpoint in time (such as the LGM) is futile. Similarly, we know of nogeological or geomorphological data that can allow discriminationbetween a low monodome centre of the ice sheet, a ridge, ora multidome situation with a high saddle over Hudson Bay at theLGM. The old issue of monodome vs. multidome is firmlyentrenched in the research history (Denton and Hughes, 1981;Peltier, 1994) but since monodome vs. multidome is a topologicalclassification which may involve subtle differences in absoluteelevation and flow pattern, we abstain from elaborating on thisquestion. In the absence of reliable constraints, we depict simplya single form line that defines a high ridge over Hudson Bay anda triple junction with the Foxe Dome (Fig. 8a). We acknowledgesubstantial uncertainty regarding the precise location of the ridgeconnecting the Quebec and Keewatin domes.

In the Cordilleran region, most flow traces form part of five largeswarms; Mayo, Whitehorse, Liard, Prince George and Kamloops(Fig. 2). These swarms collectively form a plausible ice-flow patternand have traditionally been associated with Late Wisconsinan flow(Clague, 1989). However, it remains unclear whether this mainCordilleran flow pattern reflects either a LGM (in a broad sense), ora deglacial flow pattern or else a combination of both. This uncer-tainty stems from the current lack of understanding of the degla-ciation pattern, which in turn is a result of the difficulty in tracingice-marginal retreat in high-relief terrain using geomorphology,and the even greater difficulties concerning stratigraphic correla-tion of complex valley fills in the deep valleys dissecting this region.

A broad classification of the LGM dynamics of different sectorsof the LIS is found in Kleman and Glasser (2007), and a goodaccount of what is known about the interior configuration of the LISand CIS at the LGM is given in Dyke et al. (2002).

Fig. 7. aeg) Landform swarm patterns considered to be older than the LGM in a relative chronology framework, which is spatially divided into four main sectors. For a fulldescription of the systems, see text. h) A relative chronological framework for the swarms shown in aeg. The extremely fragmentary nature of pre-LGM systems in the CIS areamake us abstain from suggesting pre-LGM ice sheet outlines. For a discussion on pre-LGM CIS ice divide locations, see text.


Fig. 8. aeh) The reconstructed evolutionary sequence of NAIS. Ice sheet configurations are based on the swarms shown in Figs. 6 and 7, with extrapolations based on topographicconditions and assessment of glaciological plausibility. We have reconstructed the ice-margin at the distal termination of coherent lineation swarms but acknowledge these asminimum-extent configurations. For ice sheet configurations defined by several spatially separated fragments of lineation systems, the most distal swarm fragment was the guidingfeature. The reconstructed configurations are minimum-complexity solutions. i) The preferred absolute ages of the described ice sheet configurations, based on publishedstratigraphical data. Column A gives our preferred interpretation, whereas B represents an alternative interpretation based on assignment of the Rocksand till to OIS stage 5(Thorleifson et al., 1993). Location of Pelly and Nimpo swarms not shown in this figure, see instead Fig. 2.


3.2. Severn (Fig. 7b)

SW-trending lineations (indicating ice flow from NE), striae(Parent et al., 1995), and till fabric (Thorleifson et al., 1993) in theJames Bay region and the HBL indicate ice flow, younger than theSachigo and Rocksand flows (see below), from ENE. This swarm canbe traced with only minor gaps from the James Bay region to thesouthwestern corner of Hudson Bay, with the Churchill swarmrepresenting a possible extension to the west. This Severn ice sheetflowpattern is ourmain tool for correlatingevents in theQuebec andCentral sectorsof the ice sheet. Possible correlation to ice-flowtraces

in the Keewatin sector hinges on the interpretation that the smallChurchill swarm relates to the Severn flow, an inference which ispossible but not indisputable. There is a noticeable absence of eskersin the Severn swarm, also in areas where it is not overprinted byyounger ice-flow features. Because eskers are diagnostic of iceretreat patterns, also for pre-LGM retreat swarms (Lagerbäck andRobertsson, 1988; Kleman et al., 1997), we interpret this swarm asreflecting ice expansion and infilling of Hudson Bay during the finalice-build-up towards the LGM, an age assignment in accord withthat suggested by Thorleifson et al. (1992). In the till stratigraphy ofThorleifson et al. (1992), the Severn flow is firmly sandwiched


between theWinisk till, and the older Sachigo till. The suggestion byBoulton and Clark (1990) that this flow (Severn) reflects a retreatphase (OIS 3) appears at odds with the absence of eskers, and,moreover, the pattern indicates thick ice and diverging ice flowoverHudson Bay, conditions we regard as incompatible with retreat andpossible ice-free conditions during OIS 3 (Andrews et al., 1983).

3.3. Sachigo (Fig. 7c)

The Sachigo flow, which in the HBL reflects ice flow from ENE, iscontinuously traceable from the HBL to the regions immediately Sand E of James Bay. In the HBL it is mostly defined by till fabrics(Thorleifson et al., 1992). In its eastern part it is defined by a wide-spread occurrence of striae (Veillette and Roy, 1995), which formsa glaciologically plausible up-ice extension of the Sachigo flowpattern documented in till fabrics of the HBL.

The ice-flow pattern indicates that ice emanated froma dispersal centre in S or SWQuebec-Labrador. An important aspectof the Sachigo ice-flow pattern is that its layout clearly indicatesthat it formed during a major glacial event, with such a southerlyQuebec Dome location that its extent must have rivalled the LGMmargins in its southern part. At the same time it indicates little orno ice in the SW Laurentide area (Liverman et al., 1989). The HBL tillstratigraphy clearly shows the Sachigo to be older than the Severnand younger than the Rocksand tills (see below). The Sachigo tillpostdates at least two ice-free periods in the HBL stratigraphy, thePrest Sea and Bell Sea sediments (see below) because it containsreworked marine molluscs from both ice-free episodes. Given thatit predates the LGM, we regard OIS 4, or a cold phase during OIS 3,as the most plausible ages for formation of the Sachigo swarm.

3.4. Rocksand (Fig. 7d)

The Rocksand flow, which reflects ice flow from the ESE in theHBL, is based on till fabric and lineation evidence (Thorleifson et al.,1993; Parent et al., 1995; Clark et al., 2000). It formed undera massive Quebec dome with its centre in an even more southerlyposition than during the Sachigo flow event. It is distinguishedfrom the Sachigo flow event on the basis of both the HBL stratig-raphy (Thorleifson et al., 1993) and striae patterns in the James Bayregion (Veillette and Roy, 1995). The Rocksand till containsreworked Bell Sea molluscs only, in contrast to the Sachigo till,which contains both Bell and Prest Sea molluscs (Andrews et al.,1983) Like the Sachigo swarm it reflects a Quebec dispersalcentre with a size that approaches its later LGM extent. An ice-flowpattern from ESE in the HBL indicates a NeS oriented margin westthereof, a situation which we regard as incompatible with ice onthe southern and central IP. The reason is that the combination ofESE-flow in the HBL simultaneously with an ice margin far to thewest on the IP would define a glaciologically implausible configu-ration, with long flowlines to the west and short flowlines to thesouth, a situation that could not be explained by geological andmeteorological differences for the western and southern sectors ofsuch an ice sheet. This argumentation is similar for both theRocksand and Sachigo flow patterns. The major discrepancy in ice-flow configuration compared to the ice-flow pattern during theLGM indicates an age for the Rocksand flow that is considerablyolder. The glaciological implications are broadly similar to those ofthe Sachigo, i.e. an extremely southerly location of the QuebecDome. The Rocksand flow is defined by four spatially separatedswarms (Fig. 6). These swarms together form a meaningful glaci-ological pattern, but in the absence of absolute dating control, weacknowledge uncertainty about whether they represent one ormore events. The anti-clockwise directional shift in ice-flowdirection reflected in the RocksandeSachigoeSevern sequence in

the HBL could be compatible with a shifting flow pattern duringcontinuous ice coverage, but does not in itself prove the absence ofan intervening deglaciation event.

3.5. Atlantic (Fig. 7e)

The Atlantic configuration is defined by the scattered occurrenceof locally oldest flow traces reflecting SW ice flow in central Quebec(cf. flow sets 16 and 17 in Clark et al. (2000), Fan E in Jansson et al.(2002)). In central Quebec, this swarm consists of large SW-trending drumlins indicating a more easterly dispersal centre thanany other ice sheet configuration. In its W sector it is directionallysimilar to LateWisconsinan flow, but in the SE part of its occurrencethe landforms from the Atlantic flow clearly are separable fromyounger landforms in both direction and morphology. There are nounambiguous crosscutting relationships with the Sachigo orRocksand swarms.

The Atlantic swarm is part of a controversy surrounding the ice-flow sequence in Quebec-Labrador, and in particular the age of N-directed flow into Ungava Bay (Kleman et al., 1994; Veillette et al.,1999; Clark et al., 2000; Dyke et al., 2002; Jansson et al., 2002),Veillette et al. (1999) reported evidence that this ice flow intoUngava Bay is younger than a single event of SW-directed flow(interpreted as the main deglacial flow) in central Quebec. Clarket al. (2000) and Jansson et al. (2002), in contrast, identified twoSW-directed ice-flow events, of which the Atlantic swarm is theolder, and a number of individual N-directed ice stream or surgeevents that are younger than the Atlantic swarm. The latterschemes better explain the striae and landform record, bypermitting the differing relative-age relationships that are actuallyobserved between SW and N ice flow at different localities.

The main glaciological implications of the Atlantic configurationare; (1) it probably occurred during a time with small LIS volumesand during which the ice sheet had a NNWeSSE oriented ice dividein eastern Quebec-Labrador; (2) it is compatible with, but does notrequire, ice-free conditions in Hudson Bay and James Bay and; (3) itindicates a more northerly dispersal centre than the Sachigo andRocksand swarms and, unlike these, is compatible with fairlyretracted marginal positions in southern Quebec. Its absolute age isunknown, but it appears likely that it reflects a build-up stage in thelast glacial cycle, or an ice sheet remnant during an interstadial. Ifolder than the Wisconsinan, such restricted ice volumes are onlycompatible with a stage preceding the Illinoian maximum, analternative we consider less probable due to the preservation of theAtlantic morphology.

3.6. Shagamu (Fig. 7f)

The Shagamu swarm is defined primarily by till fabrics reportedby Thorleifson et al. (1992), and striae in Quebec (Bouchard andMartineau, 1985). It includes the oldest traces of glaciation in theHBL and, according to Thorleifson et al. (1992), is of Illinoian orolder age. Our Shagamu swarm is composite and based ongeographically scattered evidence. The inclusion in this swarm oflineation patches reflecting old S flow in central Quebec is morequestionable than those from S and SE of James Bay. However,exclusion of the central Quebec patches would not materially alterthe interpretation that the Shagamu swarm is radically differentfrom all other pre-LGM swarms in that it indicates ice flow into SECanada from a dispersal centre located in the Hudson Bay region.

3.7. Aberdeen and Garry (Fig. 7g)

In central and northern Keewatin, a widespread pattern of ice-flow traces, the Aberdeen swarm, reflecting a dispersal centre in


northern Keewatin, is clearly older than flow from the Late Wis-consinan Keewatin Dome. The Aberdeen swarm is composed ofdiscrete patches of till lineations which together form a divergentpattern (Kleman et al., 2002). Striae that are conformable with thispattern (Lee, 1959) extend its spatial extent to the E. This Aberdeenpattern has previously been interpreted to represent an ancestralNE Keewatin dispersal centre (Boulton and Clark,1990; Kleman andBorgström, 1996; Kleman et al., 2002, 2006). Similar conclusionsregarding the presence of an ancestral Northern Keewatin dispersalcentre were reached by McMartin and Henderson (2004), whorecognised two consecutive flow patterns emanating from thisgeneral area.

Our mapping shows consistent overprinting of an even older setof lineations, Garry, from a slightly more westerly direction. Thesimilarity of the Aberdeen and Garry ice-flow patternsmay indicatethat these two swarms formed close in time. In fact, they may bothrepresent different stages of the same event.

Both the Aberdeen and Garry swarms indicate ice-dispersalcentres located N of the Late Wisconsinan dome centre. Because ofthe reversed relative chronology andmodest differences inmappedpattern they cannot be correlated directly to phases A and B inMcMartin and Henderson (2004), but we infer that both sets ofobservations embrace the same broad glacial event. Although theSWmargin was potentially located further outwards than the mostdistally preserved lineations would indicate, flow-line symmetryarguments can be invoked to suggest that both the Aberdeen andGarry configurations indicate thin or no ice on the IP. A conservativeextrapolation of the outline of the two swarms, using the evidencefor southerly directed flow near Hudson Bay as a key element,indicates that at least the northern half of the Hudson Bay was ice-covered during both events. Of the pre-LGM ice-flow systems, theNorthern Keewatin ice sheet that is indicated by the Aberdeen (andpossibly Garry) swarm, is the only one for which a candidate endmoraine zone, at Thlewiatza River, has been identified (Klemanet al., 2002). Reliable correlations with the Sachigo, Rocksand, orAtlantic swarms are, however, lacking, although for the Aberdeenswarm we regard the Sachigo as the most likely correlativecontemporaneous ice mass in Quebec/Ungava, based on patternalone. A particular difficulty in correlating swarms and events inthe western and central sectors is the apparently long-lastinginterlobate nature of ice in the Churchill region, with evidence forlarge variations in till fabric for individual till units (Dredge andNielsen, 1985; Nielsen, 2002), and ice flow emanating from Hud-son Bay and Keewatin, respectively.

3.8. Caledonian and Fundy swarms (Fig. 7c and d)

The composite Caledonian stage of Stea (2004) is here repre-sented by two ice-flow events, an earlier W flow and a later flowfrom the NW.We apply “ Fundy ” for the older flow and “ Caledonian” for the younger flow from the NW. The Fundy swarm is defined bystriae and till fabrics reflecting Laurentide ice flow over Nova Scotiafrom the W to WNW, a direction indicating a more southerlydispersal centre than later ice-flow directions (Stea, 2004).

The ice-flow pattern represented by the Fundy swarm is gla-ciologically enigmatic because it occurs at 45� to the shelf edge overa wide region. This configuration potentially indicates that the icemarginwas at the shelf edge outside southern Nova Scotia, but thattherewas little or no grounded ice in the Laurentian Channel. Basedon spatial pattern alone, and in particular the uniquely southerlydispersal centre location it indicates, the Rocksand swarm appearsto be the most likely correlative ice-flow pattern. The source areasfor glacial events in Nova Scotia have previously mainly been dis-cussed in terms of local (spreading centres in New Brunswick orNovia Scotia), Appalachian, and Laurentide (invasion from the NW

across the Saint Lawrence valley) ice sources. The Fundy phasecannot easily be assigned to either of these. The broad parallell flowfrom the west is most easily explained as the flank flow of a NeSoriented ice sheet ridge west of the area. In its southern part, sucha ridge would indeed constitute an “Appalachian” ice divide, butthe northern extension that sems to be required to explain the flowpattern cannot be classified as either Appalachian or Laurentide, asit cuts obliquely across the Saint Lawrence depression.

The Caledonian swarm is defined by striae and till fabrics, andindicates ice flow from a more NW direction than during thepreceding Fundy flow. Glaciologically, it is compatible with an icemargin along the shelf edge, or inwards of but parallel to the shelfedge. Its spatial pattern fits well with the Sachigo swarm. Indeed,the northward dispersal centre shift, from an extreme southerlyposition as indicated on the west side of the Quebec dispersalcentre, from Rocksand to Sachigo, and a similar shift evidenced inthe east from Fundy to Caledonian, suggests, but does not prove,that the two pairs of eventsmay be related. Stea (2004) and Seaman(2004) assigned an age of 75e40 kyr for the Caledonian stage,which embraces both our Fundy and Caledonian swarms. Below,we discuss possible correlations within the group of the Rocksand,Sacihigo, Fundy and Caledonian swarms.

3.9. Ungava (Fig. 6)

The N-directed ice flow converging on Ungava Bay, the UngavaSwarm, was initially described as a single coherent flow system(Hughes, 1964; Klassen and Thompson, 1993; Kleman et al., 1994),but has later been shown to consist of up to 5 discrete units (Clarket al., 2000; Jansson et al., 2002, 2003), each one representing anice stream, or a surge event. Kleman et al. (1994), based on cross-cutting lineations in the most proximal zone in central Quebec,suggested the Ungava Swarm was of pre-Late Wisconsinan age. Theflow sequence in central Quebec has subsequently been shown to beexceedingly complex, with one or more of the previously supposedsingle flow-directional events now being interpreted as repeatedflows (N-flow in Veillette et al. (1999), SW-flow in Jansson et al.(2002)), and with the recognition of new ice-flow events (Clarket al., 2000; Clarhäll and Jansson, 2003). Based on these develop-ments, we here explicitly depart from the pre-Late Wisconsinan ageassignment of Kleman et al. (1994). In agreement with others(Veillette et al., 1999; Dyke et al., 2003), we consider the Ungava Bayswarm to be a lateglacial feature, albeit composed of several sub-events, of possible correlative age to the documented northwardadvances ontoMeta Incognita Peninsula, Baffin Island (Manley,1996;Kleman et al., 2001). However, there remains a possibility that thesouthernmost part of the Ungava Swarm reflects a separate olderflow event of unknown absolute age (Veillette et al., 1999).

3.10. Churchill (Fig. 7b)

In northernmost Manitoba there exists a restricted patch ofWSWeENE trending till lineations (Fig. 5), truncated to theWand Eby two lateglacial ice streams. We informally name this restrictedbut potentially important swarm the Churchill swarm. It canneither be reliably correlated to the till stratigraphy in northernManitoba (Nielsen, 2002), nor to other swarms. Based on itsdirection only, it is possible that the Churchill swarm represents anoutlier of the Severn swarm.

3.11. Prest Sea and Bell Sea episodes

The Prest Sea and Bell Sea episodes fall outside the employedswarm concept (as defined by ice-flow indicators), but must never-theless be included in a paleoglaciological analysis because


a meaningful discussion on pre-LGM North American ice sheetevolution has to include consideration of periods with conditions ofminimum ice cover. Again, the evidence from the HBL is crucial. Wehere follow the sequence of events described by Thorleifson et al.(1993, pp. 52e55), who also refer to the sites where the Prest Seaand Bell Sea episodes were originally defined (Skinner, 1973).Although the evidence for the Prest Sea and Bell Sea episodes doesnot yield direct glacial geological evidence such as the configurationsof presumably smaller ice sheets that may have existed in Keewatinor Quebec during those times, due to the central location of the HBL,ice-free conditions there still efficiently constrain ice volume duringthese episodes to less, and probably much less, than approximatelyhalf the LGM volume. The Prest Sea deposits are considered to beyounger than the Bell Sea deposits based on consistently loweramino acid alle/lle ratios (Andrews et al., 1983). The absolute age ofpost-Bell Sea ice-free conditions (Prest Sea) is disputed, with radio-carbon and TL dating giving inconclusive or contradictory results(Dredge and Thorleifson, 1987; Vincent and Prest, 1987; Clark et al.,1993). Thorleifson et al. (1993), in their preferred alternative, assignthe Bell Sea to the Sangamonian (OIS 5e), and the Prest Sea to anEarly Wisconsinan interstadial, probably OIS 5a. However, they alsofind that if TL dates are deemed insufficiently reliable, the two eventsmay instead pertain to OIS 7 and 5e. Despite renewed efforts toradiocarbon date crucial samples, the age of the Prest Sea stillremains uncertain (McNeely and Nielsen, 2000). By nature, non-glacial deposits in the HBL stratigraphy require a reduction in theKeewatin and Quebec dispersal centres. In addition, the evidence formarine conditions is also direct evidence for a lack of confluence ofthe Quebec and Keewatin and/or Foxe-Baffin ice-dispersal centres.Glaciolacustrine conditions in the HBL stratigraphy, on the otherhand, are indicative of the presence of an unbroken eastewest icebarrier to the north of the HBL.

3.12. Pelly (Figs. 2 and 6)

The Pelly swarm (Figs. 1a and 2) is defined by the widespreadoccurrence of degraded drumlins in the Whitehorse e PellyMountains region, which occur as a regional ice-flow system that iscrosscut by the main Late Wisconsinan system in this region.

The Pelly lineations are oriented northesouth, but no indisput-able directional indicator has yet been found. When the differentparts of the Pelly swarm are viewed in a topographic context, itappears most realistic that ice flow emanated from the region east ofPellyMountains and flowed in a SSWdirection towards the coast. Nochronological constraints appear to exist for this flow, except that itis overprinted by the main Late Wisconsinan system.

3.13. Nimpo (Figs. 2 and 6)

A notable group of flow traces, the Nimpo swarm, converge oncols draining towards the Bella Coola inlet on the coast (Fig. 2). Thisolder system indicates drainagewestward through the Coast Range,as opposed to the later northeastward flow out from the CoastRange. We refute both the relative age and direction of flowpreviously reported by (Tipper,1971) for this system. Crag-and-tailsclearly indicate ice flow towards the west. Superimposition of smallNE-trending drumlinoids, show that the underlying larger formspredate, instead of postdate, the main Late Wisconsinan flowsystem. Like for the Pelly and Quesnel swarms, we know of no solidage constraints for the Nimpo swarm.

3.14. Quesnel (Figs. 2 and 6)

The Quesnel swarm (Fig. 2) consists of a small number ofdegraded till lineations and associated striae, apparently preserved

under a regional ice divide (saddle), located between the LateWisconsinan Prince George and Kamloops swarms. Other, possiblyrelated fragments of older flow are preserved in the up-ice parts ofthe Prince George swarm. We have found no indisputable direc-tional indicators, so it is unclear wheteher the Quesnel swarmrepresents ice flow from the Coast Range or the Cordillera.

A particularly enigmatic observation concerns the southern halfof the Prince George swarm.

Here, ice flow from a very large catchment area, in the westreaching up-ice to the ice divide in the Coast Range, and in thesouth to the saddle beween the Prince George and Kamloopsswarms, was funneled into a narrow portion of the RockyMountainTrench (Fig. 2), south of which the flow traces simply peter out. Thesimplest explanation for this pattern is that this part of the trenchwas an area with massive ablation and ice loss when the PrinceGeorge swarmwas formed. On the other hand, based on the up-iceparts of the discrete corridors which together form the Kamloopsswarm, this same part of the trench must have constituted thecatchment area for the flow going to the southern margin of the CIS.Therefore substantial non-synchroneity within the 5-swarmcomposite flow pattern here assigned to the Late Wisconsinan isindicated. The Mayo, Whitehorse, Liard, and Kamloops swarms canall be traced to the vicinity of the Late Wisconsinan maximummargin, but this is not the case for the Prince George swarm. Theevidence for continuity of flow from the Prince George swarm andeastwards across the Rocky Mountains is meagre, at best. Twoalternatives appear to exist; either the Prince George swarm isdeglacial, or very significant local ablation areas existed within theRocky Mountain trench in near-LGM time.

4. Configurations and absolute chronology e synthesis anddiscussion

The synthesis of our work, which integrates the evidence for thespatial extents and relative ages of swarms with absolute chro-nology, is shown in Fig. 8. Here we show inferred ice sheet outlinesand tentative ice surface form lines based on flow directions in theproximal parts of each swarm. Fig. 8a shows the two chronologicalalternatives we discuss in the text.

In agreement with previous authors (Boulton and Clark, 1990;Clark et al., 1993), we emphasize the importance of the HBL stra-tigraphy (Thorleifson et al., 1992, 1993) for understanding the pre-LGM history of the LIS. The HBL stratigraphy is a composite derivedfrom observations at many locations, primarily from river sections,and is probably the most complete record of both glacial and non-glacial events available for any part of the LIS. In addition, its centrallocation implies that it reflects both shape and volume changesmore directly and sensitively than more peripheral sites. Swarmsthat can be tied to the HBL stratigraphy using the direction andpattern of the landform systems, also in some cases extend intoeastern (Quebec) and possibly also western (Keewatin) LIS sectors,adding to the correlation possibilities where chronologicalconstraints are lacking. It is apparent that the absolute age of theRocksand swarm which reflects westward flow from a uniquelysoutherly dispersal centre in Quebec, is a key element for a credibleevent sequence reconstruction. Importantly, the Rocksand swarmcan be traced in both central and eastern sectors of the LIS whichallows us to link the uniquely southerly dispersal centre to theevolution in the entire south-central portion of the ice sheet. Wereason that such a southerly ice-dispersal centre could not bemaintained at times other than during stadials and, because itclearly predates the LGM, an OIS 4, 5d or 5b age appears to be themost realistic. The eastward ice flow of the Fundy swarm in NovaScotia which emanated from a uniquely southerly dispersal centreor northesouth oriented ridge joining Appalachian and Quebec


Domes, was assigned an age in the 40e75 kyr interval by Stea et al.(1998), but the constraints on its age do not appear strong, theprimary one being the presence of striae with this direction ona marine rock bench of assumed Sangamonian age (Rampton et al.,1984). Seaman (2004) suggests that evidence for W to E flow in theupper St John river valleymay be correlative to the early Caledonianphase (the flow we here label as Fundy) of Stea et al. (1998), butagain, the evidence is not conclusive. In summary, we regard anEarly Wisconsinan age for the Fundy flow as likely, but notconclusively shown.

The Rocksand and Fundy swarms may offer an opportunity tocorrelate events on opposing sides of an ice divide on the basis ofextreme but compatible configurations. The Sachigo and Caledo-nian swarms, on thewest and east sides of the Quebec ice-dispersalcentre, respectively, both indicate the presence of a more northerlydispersal centre than during the Rocksand and Fundy swarms.There are no known intermediate flow patterns between theRocksand and Sachigo, and the same holds true for the Fundy andthe Caledonian. In our view, the simplest and most likely expla-nation is that the Rocksand/Sachigo and Fundy/Caledonian pairsreflect the same northward shift of an Appalachian-Quebecdispersal centre, but we also acknowledge the possibility that theRocksand and Fundy flows are unrelated to each other.

Thorleifson et al. (1993) suggested an OIS 5 age for the Rocksandtill, based on the presence of redeposited Bell Sea (by them assignedto stage 5e) shell fragments, and the absence of redeposited younger(Prest Sea) shell fragments. Parent et al. (1995) suggested anOIS 4 agefor NW flow east of Hudson Bay, based on their evidence for long-distance transport and long duration of flow for this stage, whichpreceded westward flow assigned to the LGM. They presented anattractive proposal to correlate this NW flow to the Rocksand in theHBL. In the end they abstained from making this correlation due tothe Thorleifson et al. (1993) OIS 5 assignment for the Rocksand. Wehere adopt the Rocksande Quebec NW flow correlation as originallycontemplated by Parent et al. (1995). This correlation is inferred onthe strength of the argument that it is unlikely that two separateevents would yield swarms that are directionally fully conformable,but spatially complementary with no overlap. This correlation is animportant element in establishing the flow sequence labelled A inFig. 8. The LGM-like southern extent of the Quebec dome during theRocksand indicates that the most realistic age for the Rocksandconfiguration we depict was OIS 4, i.e. the most pronounced coldperiod and sea-level lowstand before the LGM (Mix, 1992; Lambeckand Chappell, 2001). The inference that there was a massiveQuebec dome duringOIS 4 contrasts sharply to the Boulton and Clark(1990) reconstruction,who suggested the ice sheetwas in a retractedstate during this time. Flow sequence A is compatible with resultspresented by Hillaire-Marcel and Causse (1989), who presentedU/Thdating evidence of an 80 kyr (OIS 5/4 transition) major advance ofQuebec ice into the St Lawrence Lowland.

Our alternative, flow sequence B (Fig. 8), follows the Thorleifsonet al. (1993) suggestion of an OIS 5 age for the Rocksand. Followingclimate arguments presented above we regard OIS 5d or 5d as theonly possible substages for major ice-sheet build-up within OIS 5.Of these two, we regard OIS 5d as the less likely alternative, becauseit would imply that a very southerly major dome was build in thesouthern Quebec sector immediately after OIS 5e, when no residualice caps existed in the Laurentide area. The general pattern else-where, such as the Garry and Aberdeen swarms in Keewatin, andthe early stages in Fennoscandia, suggest an insolation drivennortherly build-up directly following the interglacial. TheThorleifson et al. (1993) alternative interpretation of an OIS 6 orolder age for the Rocksand appears appears unlikely because theRocksand configuration is clearly incompatible with the Shagamu,for which OIS 6 appears the most likely alternative.

For the swarms younger than the Rocksand, our reasoning is thefollowing:

As stated above, we propose a correlation of the Caledonian andSachigo swarms. The evolution from the Sachigo/Caledonian to theSevern configuration, whatever any intermediate stages were like,involves an important change from the distinct SE Quebec dispersalcentre to a Hudson Baymonodome or possibly “longWNWeESE icedivide” across the Hudson Bay. The Sachigo/Caledonian configura-tion, like the Rocksand/Fundy, still reflects a pronounced easterlylocation of the ice sheet with little or no ice in the SW, whereas theSevern starts to resemble the LGM configuration. In contrast toThorleifson et al. (1993) we do not favour a post-LGM age for theSevern, because its monodome-like configuration fits poorly withthe distinct Keewatin and Quebec ice-dispersal centres that arethought to have existed from the LGM throughout the deglacialphase (Dyke and Prest, 1987). We consider an age immediatelypredating the LGM to be the most likely alternative, also becausethe Severn flow pattern appears to be associated with long-trans-port paths of erratics towards the SW. Such long-transport isconceptually far more likely to occur during ice sheet build-up thanduring ice sheet decay (Kleman et al., 2008). We have retained anOIS 3 age in both alternatives in Fig. 8. Bio- and litho-stratigraphicalevidence in the Lake Ontario Basin (Eyles and Eyles, 1993), indicatethat the St Lawrence Riverwas repeatedly blocked by ice emanatingfrom the Quebec dispersal centre during the Early and MiddleWisconsinan, but that Wisconsinan pre-LGM ice sheets neverreached south of Toronto. Our reconstruction and age assignmentsof the Sachigo/Caledonian and Rocksand/Fundy configurations aredifficult to reconcile with such a demand for a restricted southwardextent of ice in the eastern Great Lakes region, but we also note thatthere is considerable debate over the interpretation and geneticassignment of key beds in the Scarborough formation in Toronto(Eyles and Eyles, 1993; Godin et al., 2002).

In the Keewatin sector, we acknowledge great uncertaintyregarding the ages of the Aberdeen and Garry swarms, because nochronological constraints seem to exist. Moreover, the Aberdeenand Garry flow traces nowhere geographically overlap with otherpre-LGM swarms. However, we have tentatively correlated theAberdeen swarm with the Sachigo/Caledonian swarms, and alsoregard it as plausible that Garry and Rocksand/Fundy swarms are ofthe same age. If so, the Caledonian/Sachigo/Aberdeen swarmsindicate an early stage of confluence between Keewatin and Quebecsources of ice. The transition from the Caledonian/Sachigo/Garrystage to the Severn stage appears to be a glaciologically plausibleevolution yielding the mass increase that characterized the build-up towards the LGM.

The Atlantic swarm, which is here explicitly defined for the firsttime, is the oldest clearly discernible flow event in central Quebec(Clark et al., 2000; Jansson et al., 2002), although unambiguouscrosscutting evidence for relative age is meagre. In alternative A,the Atlantic swarm is assigned to either OIS 5b or 5d and in alter-native B it is assigned to OIS 5d. Because the Atlantic swarm indi-cates a restricted ice volume, an alternative placement within theOIS 4e3e2 intervals seems unlikely. A pre-Sangamonian (i.e. Illi-noian or older) age also appears rather unlikely. The small icevolume indicated is certainly at odds with it reflecting an Illinoian(OIS 6) ice sheet. An OIS 6 decay age for the swarm also seemsunlikely because of the indication of a southerly located continuousice divide from Labrador to Newfoudland. Such an ice divide did notoccur during the Late Wisconsinan deglaciation, during which iceremnants instead were north-centrally located in Quebec, repre-senting rather symmetrical shrinkage towards terrestrial coreareas. It therefore appears climatologically implausible during theOIS 6 deglaciation. The final alternative, an Illinoian build-up,would involve preservation of striae and well-developed drumlins


through two full ice-sheet cycles, in an area that contains no knownIllinoian maximum or decay stage flow traces. Although notimpossible, we also consider this last alternative to be unlikely.

The Shagamu swarm, which in both alternatives in Fig. 7 isconsidered to predate the Wisconsinan, indicates an ice sheet witha mass distribution radically different from those prevailing duringany other time leading up to the LGM. If it is indeed Illinoian in age(Thorleifson et al., 1992), it indicates a radically different massdistribution within the ice sheet at that time, despite the fact thatthe maximum stage outlines during the Late Wisconsinan andIllinoian show only modest differences. The Shagamu configurationclearly merits further investigation.

The Pelly, Nimpo and Quesnel swarms give morphologicalevidence for older configurations in the Cordilleran region (Fig. 6).The Pelly and Nimpo swarms reflect more easterly ice dividelocations than those dominating during the Late Wisconsinan,whereas the significance of the Quesnel swarm is unclear. It isuncertain whther the Pelly and Nimpo swarms represent the sameevent. Our speculative interpretation is that these two configura-tions may be of OIS 4 age, in line with previous age assignments ofdrift in the northern CIS area (Ward et al., 2007).

5. Wider implications

The presentation of glaciologically plausible ice sheet outlinesand a pre-LGM glacial history on an absolute time scale in NorthAmerica can be tested independently using numerical ice sheetmodels. When comparing our results to existing numerical icesheet model outputs of LIS evolution, the following two challengescan be identified. Firstly, it is striking that our evidence for a longbuild-up phase towards the LGM characterized by the persistenceof dynamically separate ice-dispersal centres in northern Keewatinand Quebec contrasts strongly with many numerical ice sheetmodels which typically feature a rapid growth towards a mono-dome configuration early in the last glacial cycle. Secondly, thepronounced southerly extent of the ice sheet during one or more ofthe major Early/Middle Wisconsinan stadials, with OIS 4 as ourpreferred alternative, represents an important new boundarycondition for future Global Circulation Modelling efforts in that it issignificantly different from the often-modelled LGM configuration.

In terms of the North American ice sheet build-up dynamics, wespeculate that residual inland ice caps may have played an impor-tant role. At the onset of glacial conditions immediately followingfull interglacials, we infer that the ice-build-up pattern was deter-mined by two factors alone; solid-rock topography and the mass-balance pattern. After interstadials that left residual ice caps moreor less intact, the build-up pattern would also be strongly influ-enced by a third factor, the location and extent of these ice caps. Thelocation of such residual ice would, in turn, be a function of inter-stadial climate and mass-balance patterns, and might be quiteunrelated to mass-balance patterns during cooling and build-upphases, and equally unrelated to solid-rock topography. Thedeglacial (8 kyr) Keewatin ice sheet remnant is a splendid exampleof such spurious location that cannot be explained by topography.

In effect, “residual” ice domes surviving interstadials wouldaffect build-up during orbitally controlled cooling phases in twofundamental ways, i) the areal extent of cold high-elevation initialtopography would have been much larger than if same area wasfully ice-free, and ii) the location of “highland” areas would bedifferent and governed by the preceeding glacial evolution.

In particular the ice sheet history of the Quebec sector throughits succesive stages (Atlantic/ Rocksand/ Sachigo/

Severn/ LGM), can be interpreted in this way. In our preferred agealternative (A), the smallest Quebec-centred configuration, repre-sented by the Atlantic swarm, formed directly following the

Sangamonian, when no residual ice would have been present inQuebec. Conversely, the Rocksand/Caledonian/Garry swarm and icesheet configuration probably developed on remnant ice from theAtlantic configuration and consequently attained a much largersize. If both the Sachigo/Fundy/Aberdeen and Severn configurationsare OIS 3 in age, as we suggest, the likely explanation for them notexceeding the preceding Rocksand configuration is the increasedsummer insulation during OIS 3 compared to OIS 4. We also notethat the larger fraction of “interstadial time” during OIS 3, ascompared to OIS 4, would likely help in holding back ice-sheet size.In essence, we anticipate OIS 3 ice to fluctuate around volumessmaller than during OIS 4, but not be eliminated.

The consequence of this reasoning is that the timing of climaticreversals may have been as important as topography in deter-mining glaciation patterns in eastern North America. This behav-iour sets the LIS apart from other, smaller, mid-latitude ice sheets(Kleman et al., 1997) which are likely to have had a more direct andrapid response to climate and a simpler growth-out-of-highlandsorigin (Flint, 1943; Andrews and Barry, 1978). Our results thereforeindicate that the LIS, through a pronounced hysteresis effect,modulated a basically simple climatic pulsebeat into a uniqueevolutionary sequence in terms of ice sheet size and location.

The records of the Keewatin and Arctic sectors do not presentlyallow similar conclusions to be advanced. This is partly because ofthe spatial fragmentation caused by the large marine areas, andprobably also because initial ice sheets in these regions were coldbased for longer periods of time than was the case in the easternand south-central regions of Quebec and HBL.

During the decade-long work preceding this paper, a fairly clearpicture has emerged of what presently constitutes the issueslimiting our understanding of the build-up phase of ice sheets inNorth America. It is clear that the absolute ages of most pre-LGMsystems still remain questionable, and that knowledge about keyissues such as OIS 3 interstadial ice extent is very fragmentary(Dyke et al., 2002). The first coalescence of Quebec and Keewatin/Arctic ice-dispersal centres remains unconstrained, despite itsimportance for the damming-up of large proglacial lakes in theHudson Bay region. The timing of the east-centred CIS configura-tion is likewise poorly known. These issues cause considerableuncertainty concerning key information such as the area andvolume of ice sheets, and proglacial lake extent, and are animpediment to our understanding of ice sheet e climate interac-tions through the last glacial cycle. Themost obvious research need,which has the capacity to resolve two of the issues stated above,would be a renewed full-scale attempt to reliably date key parts ofthe phenomenally rich HBL stratigraphy.

We regard the here presented geomorphologically basedreconstruction of Ice Sheet build-up during the 115e21 kyr asa testable hypothesis. It highlights some geographic regions aspotentially more important than others for stratigraphic investi-gations, because only there will new dating control efficiently linkto older landform systems of regional significance. Only withstronger linkage between stratigraphical and morphological datawill our understanding about older ice sheet extents and configu-rations improve.

We see two fundamental ways to test the solidity of ourreconstructed evolution sequence; i) Dating of potential intersta-dial and interglacial key sites, in particular renewed efforts atdating inter-till sediments in the Hudson Bay Lowlands, and ii) Ice-sheet modelling. The ouput of numerical models is not generallyregarded as a valid test of reconstructed historical events, but weargue that it should be possible to efficiently employmodels in sucha role, because they obey fundamental physical laws that do not inany organised fashion enter into geologically and geo-morphologically based reconstructions. In particular, models are


strong in calculating terrestrial retreat rates (where meltingdominates over calving) and, assuming a reasonable realistic mass-balance scheme, they are also realistically “precipitation-limited” inthe speed with which major ice masses are built. In effect, failure ofrealistic ensemble-modelling to capture first-order evolutionpatterns, such as the one presented here, may indicate eithera flawed model, or alternatively, point to fundamental problems ofgeologically and geomorphologically driven reconstructions. Themore precise nature of mismatches is likely to provide informationabout the most profitable direction of search.

6. Conclusions

We conclude that ice domes in Keewatin and Quebec weredynamically independent for most of pre-LGM time and thata massive Quebec dome, rivalling the LGM in extent, existed attimes when the SW part of the ice sheet had not yet developed. Anorthern Keewatin-Central Arctic Ice Sheet existed prior to theLGM, but is poorly constrained chronologically. The oldest flowsystem in Quebec (Atlantic swarm) had an ice divide closer to theLabrador coast than any later configuration. The subsequent Rock-sand configuration, with possible Fundy and Garry correlatives, hada uniquely southerly dispersal centre and was the most extensivepre-LGM configuration. We tentatively assign the Rocksand to OIS4, with OIS 5b as a possible alternative. Subsequent ice sheetconfigurations during OIS 3, the Sachigo/Caledonian/Aberdeen andthe Severn (leading up to the LGM configuration) were probablyspatially less extensive. In the Cordilleran area, older flow patternsindicate significantly more easterly ice divide locations than thoseprevailing during the Late Wisconsinan. No firm dating constraintsfor these exist, but based on the assumption that the abundance ofpreserved flow traces in general decreases with age, we tentativelyassign them to OIS 4. The location and size of remnant ice masses atthe end of interstadials is an important aspect of the history of icesheet growth in eastern North America because the survival of suchresidual ice caps may have determined the location of subsequentgrowth centres.

Acknowledgements

We thank the late Douglas R. Grant for support and encour-agement in this work. The staff at the National Air Photo Library inOttawa offered generous assistance. Karin Willis and Ola Fredinhelped with map work during the early stages of this work. Thestudy was funded through grants from the Swedish ScienceResearch Council (VR), FORMAS, and the Swedish Space Board. Thisis a contribution from the Bert Bolin Centre for Climate Research atStockholm University. We sincerely thank David Sugden and ananonymous reviewer for their constructive comments.

References

Alley, R.B., MacAyeal, D.R., 1994. Ice-rafted debris associated with binge/purgeoscillations of the Laurentide Ice Sheet. Paleoceanography 9 (4), 503e511.

Andrews, J.T., Barry, R.G., 1978. Glacial inception and disintegration during the lastglaciation. Annual Review of Earth and Planetary Sciences 6, 205e228.

Andrews, J.T., Shilts, W.W., Miller, G.H., 1983. Multiple deglaciations of the HudsonBay lowlands, Canada, since deposition of the Missinaibi (last-interglacial?)formation. Quaternary Research 19, 18e37.

Bintanja, R., van de Wal, R.S.W., 2008. North American ice-sheet dynamics and theonset of 100,000-year glacial cycles. Nature 454, 869e872.

Bouchard, M.A., Martineau, G., 1985. Southeastward ice flow in central Quebec andits paleogeographic significance. Canadian Journal of Earth Sciences 22 (10),1536e1541.

Boulton, G.S., Clark, C.D., 1990. A highly mobile Laurentide ice-sheet revealed bysattelite images of glacial lineations. Nature 346, 813e817.

Charbit, S., Ritz, C., Philippon, G., Peyaud, V., Kageyama, M., 2007. Numericalreconstructions of the Northern Hemisphere ice sheets through the last gla-cialeinterglacial cycle. Climate of the Past 3 (1), 15e37.

Clague, J.J., 1989. Quaternary geology of the Canadian Cordillera. In: Fulton, R.J. (Ed.),Quaternary Geology of Canada and Greenland. Geological Society of America,Boulder, CO, pp. 15e95.

Clarhäll, A., Jansson, K.N., 2003. Time perspectives on glacial landscape formation eglacial flow chronology at Lac aux Goélands, northeastern Québec, Canada.Journal of Quaternary Science 18 (5), 441e452.

Clark, C.D., Knight, J.K., Gray, J.T., 2000. Geomorphological reconstruction of theLabrador sector of the Laurentide Ice Sheet. Quaternary Science Reviews 19(13), 1343e1366.

Clark, P.U., Clague, J.J., Curry, B.B., Dreimanis, A., Hicock, S.R., Miller, G.H., Berger, G.W., Eyles, N., Lamothe, M., Miller, B.B., Mott, R.J., Oldale, R.N., Stea, R.R., Szabo, J.P., Thorleifson, L.H., Vincent, J.-S., 1993. Initiation and development of theLaurentide and Cordilleran ice sheets following the last interglaciation.Quaternary Science Reviews 12, 97e114.

Clark, P.U., Licciardi, J.M., MacAyeal, D.R., Jenson, J.W., 1996. Numerical reconstruc-tion of a soft-bedded Laurentide Ice Sheet during the last glacial maximum.Geology 24 (8), 679e682.

De Angelis, H., Kleman, J., 2005. Paleo-ice streams in the Northern Keewatin sectorof the Laurentide Ice Sheet. Annals of Glaciology 42, 135e144.

De Angelis, H., Kleman, J., 2007. Palaeo-ice streams in the Foxe/Baffin sector of theLaurentide Ice Sheet. Quaternary Science Reviews 26 (9e10), 1313e1331.

Denton, G.H., Hughes, T.J. (Eds.), 1981. The Last Great Ice Sheets. Wiley-Interscience,New York, 484 pp.

Dredge, L.A., Nielsen, E., 1985. Glacial and Interglacial Deposits in the Hudson BayLowlands: a Summary of Sites in Manitoba. Geological Survey of Canada, Paper85-01A, pp. 247e257.

Dredge, L.A., Thorleifson, H., 1987. The Middle Wisconsinan history of the Lauren-tide Ice Sheet. Géographie physique et Quaternaire 41 (2), 215e235.

Dyke, A.S., Prest, K., 1987. Late Wisconsinan and Holocene history of the LaurentideIce Sheet. Géographie physique et Quaternaire 41 (2), 237e263.

Dyke, A.S., Dredge, L.A., Vincent, J.-S., 1982. Configuration and dynamics of theLaurentide ice sheet during the late Wisconsin maximum. Géographie physiqueet Quaternaire 36 (1e2), 5e14.

Dyke, A.S., Andrews, J.T., Clark, P.U., England, J.H., Miller, G.H., Shaw, J., Veillette, J.J.,2002. The Laurentide and Innuitian ice sheets during the Last Glacial Maximum.Quaternary Science Reviews 21 (1e3), 9e31.

Dyke, A.S., Moore, A., Robertson, L., 2003. Deglaciation of North America. GeologicalSurvey of Canada. Open File 1574 (CD ROM).

Ehlers, J., Gibbard, P.L., 2003. Extent and chronology of glaciations. QuaternaryScience Reviews 22 (15), 1561e1568.

Eyles, C.H., Eyles, N., 1993. Sedimentation in a large lake: a reinterpretation ofthe late Pleistocene stratigraphy at Scarborough Bluffs, Ontario, Canada.Geology 11 (3), 146e152.

Finlayson, A.G., Bradwell, T., 2008. Morphological characteristics, formation andglaciological significance of Rogen moraine in northern Scotland. Geomor-phology 101 (4), 607e617.

Flint, R.F., 1943. Growth of North American ice sheet during the Wisconsin age.Bulletin of the Geological Society of America 54 (3), 325e362.

Godin, L., Brown, R.L., Dreimanis, A., Atkinson, G.M., Armstrong, D.K., 2002. Analysisand reinterpretation of deformation features in the Rouge River valley, Scar-borough, Ontario. Canadian Journal of Earth Sciences 39 (9), 1373e1391.

Hättestrand, C., Kleman, J., 1999. Ribbed moraine formation. Quaternary ScienceReviews 18, 43e61.

Hemming, S.R., 2004. Heinrich events: massive late Pleistocene detritus layers ofthe North Atlantic and their global climate imprint. Reviews of Geophysics 42,RG1005.

Hemming, S.R., Hajdas, I., 2003. Ice-rafted detritus evidence from 40Ar/39Ar ages ofindividual hornblende grains for evolution of the eastern margin of the Lau-rentide ice sheet since 43 14C ky. Quaternary International 99e100, 29e43.

Hillaire-Marcel, C., Causse, C., 1989. The Late Pleistocene Laurentide Glacier: Th/UDating of its major fluctuations and d18O range of the ice. Quaternary Research32 (2), 125e138.

Hughes, O.L., 1964. Surficial geology, Nichicun-Kaniapiskau map-area, Quebec.Geological Survey of Canada Bulletin 106, 20 p.

Hulbe, C.L., MacAyeal, D.R., Denton, G.H., Kleman, J., Lowell, T.V., 2004. Catastrophicice shelf breakup as the source of Heinrich Event icebergs. Paleoceanography19, PA1004. doi:10.1029/2003PA000890.

Huybrechts, P., T’siobbel, S., 1995. Thermomechanical modelling of NorthernHemisphere ice sheets with a two-level mass-balance parameterization. Annalsof Glaciology 21, 111e116.

Jackson Jr., L.E., Phillips, F.M., Shimamura, K., Little, E.C., 1997. Cosmogenic 36Cldating of the Foothills erratics train, Alberta, Canada. Geology 25 (3), 195e198.

Jansson, K.N., Kleman, J., Marchant, D.R., 2002. The succession of ice-flow patternsin north-central Quebec-Labrador, Canada. Quaternary Science Reviews 21 (4),503e523.

Jansson, K.N., Stroeven, A.P., Kleman, J., 2003. Configuration and timing of UngavaBay ice streams, Labrador-Ungava, Canada. Boreas 32 (1), 256e262.

Klassen, R.A., Thompson, F.J., 1993. Glacial history, drift composition, andmineral exploration, central Labrador. Geological Survey of Canada Bulletin435, 76 p.

Kleman, J., Borgström, I., 1996. Reconstruction of palaeo-ice sheets: the use ofgeomorphological data. Earth Surface Processes and Landforms 21,893e909.

Kleman, J., Stroeven, A.P., 1997. Preglacial surface remnants and Quaternary glacialregimes in northwestern Sweden. Geomorphology 19 (1e2), 35e54.


Kleman, J., Hättestrand, C., 1999. Frozen-bed Fennoscandian and Laurentide icesheets during the Last Glacial Maximum. Nature 402, 63e66.

Kleman, J., Glasser, N.F., 2007. The subglacial thermal organisation (STO) of icesheets. Quaternary Science Reviews 26 (5e6), 585e597.

Kleman, J., Borgström, I., Hättestrand, C., 1994. Evidence for a relict glacial landscapein Quebec-Labrador. Palaeogeography, Palaeoclimatology, Palaeoecology 111,217e228.

Kleman, J., Hättestrand, C., Borgström, I., Stroeven, A., 1997. Fennoscandian paleo-glaciology reconstructed using a glacial geological inversion model. Journal ofGlaciology 43 (144), 283e299.

Kleman, J., Marchant, D., Borgström, I., 2001. Geomorphic evidence for late Glacialice dynamics on southern Baffin Island and in outer Hudson Strait, Nunavut,Canada. Arctic, Antarctic, and Alpine Research 33 (3), 249e257.

Kleman, J., Fastook, J., Stroeven, A.P., 2002. Geologically and geomorphologicallyconstrained numerical model of Laurentide Ice Sheet inception and build-up.Quaternary International 95e96, 87e98.

Kleman, J., Hättestrand, C., Stroeven, A., Jansson, K.N., De Angelis, H., Borgström, I.,2006. Reconstruction of paleo-ice sheets e inversion of their glacial geomor-phological record. In: Knight, P.G. (Ed.), Glacier Science and EnvironmentalChange. Blackwell Publishing Ltd., London, pp. Chapter 38.

Kleman, J., Stroeven, A.P., Lundqvist, J., 2008. Patterns of Quaternary ice sheeterosion and deposition in Fennoscandia and a theoretical framework forexplanation. Geomorphology 97 (1e2), 73e90.

Lagerbäck, R., Robertsson, A.-M., 1988. Kettle holes e stratigraphical archives forWeichselian geology and palaeoenvironment in northernmost Sweden. Boreas17 (4), 439e468.

Lambeck, K., Chappell, J., 2001. Sea level change through the last glacial cycle.Science 292 (5517), 679e686.

Lee, H.A., 1959. Surficial geology of Southern District of Keewatin and the KeewatinIce Divide, Northwest Territories. Geological Survey of Canada Bulletin 51, 42 p.

Liverman, D.G.E., Catto, N.R., Rutter, N.W., 1989. Laurentide glaciation in west-central Alberta: a single (Late Wisconsinan) event. Canadian Journal of EarthSciences 26 (2), 266e274.

Manley, W.F., 1996. Late-glacial flow patterns, deglaciation, and postglacial emer-gence of south-central Baffin Island and the north-central coast of Hudson Strait,eastern Canadian Arctic. Canadian Journal of Earth Sciences 33 (11), 1499e1510.

Marshall, S.J., Tarasov, L., Clarke, G.K.C., Peltier, W.R., 2000. Glaciological recon-struction of the Laurentide Ice Sheet: physical processes and modelling chal-lenges. Canadian Journal of Earth Sciences 37 (5), 769e793.

Marshall, S.J., James, T.S., Clarke, G.K.C., 2002.NorthAmerican Ice Sheet reconstructionsat the Last Glacial Maximum. Quaternary Science Reviews 21 (1e3), 175e192.

McMartin, I., Henderson, P.J., 2004. Evidence from Keewatin (central Nunavut) forpaleo-ice divide migration. Géographie physique et Quaternaire 58 (2e3),163e186.

McNeely, R., Nielsen, E., 2000. Manitoba Radiocarbon Dates: Geological Radio-carbon Dates (Section 1). Manitoba Industry, Trade and Mines. GeologicalSurvey Open File Report OF2000-1.

Mix, A.C., 1992. The marine oxygen-isotope record: constraints on timing andextent of ice growth events (120e65 ka). Geological Society of America e

Special Paper 270, 19e30.Nielsen, E., 2002. Quaternary Stratigraphy and Ice-flow History along the Lower

Nelson, Hayes, Gods and Pennycutaway Rivers and Implications for DiamondExploration in Northeastern Manitoba. Manitoba Geological Survey, Report ofActivities 2002. Manitoba Industry, Trade and Mines.

Parent, M., Paradis, S.J., Boisvert, É, 1995. Ice-flow patterns and glacial transport inthe eastern Hudson Bay region: implications for the late Quaternary dynamicsof the Laurentide Ice Sheet. Canadian Journal of Earth Sciences 32 (12),2057e2070.

Peltier, W.R., 1994. Ice age paleotopography. Science 265, 195e201.Prest, V.K., Grant, D.R., Rampton, V.N., 1968. Glacial Map of Canada. Geological

Survey of Canada Map 1253A.Rampton, V.N., Gauthier, R.C., Thibault, J., Seaman, A.A., 1984. Quaternary geology of

New Brunswick. Geological Survey of Canada, Memoir 416.Seaman, A.A., 2004. Late Pleistocene history of New Brunswick, Canada. Quaternary

Glaciations - Extent and Chronology, Part II: North America. Elsevier, Amster-dam, pp. 151e167.

Skinner, R.G., 1973. Quaternary stratigraphy of the Moose River Basin, Ontario,Canada. Geological Survey of Canada Bulletin 225, 77 p.

Stalker, A.M., 1956. The erratics train, foothills of Alberta. Geological Survey ofCanada Bulletin 37, 32 p.

Stea, R.R., Piper, D.J.W., Fader, G.B.J., Boyd, R., 1998. Wisconsinan glacial and sea-level history of Maritime Canada and the adjacent continental shelf: a correla-tion of land and sea events. Geological Society of America Bulletin 110 (7),821e845.

Stea, R.R., 2004. The Appalachian glacier complex in maritime Canada. In: Ehlers, J.,Gibbard, P.L. (Eds.), Quaternary Glaciations e Extent and Chronology, Part II:North America. Elsevier, Amsterdam, pp. 213e232.

Stokes, C.R., Clark, C.D., 1999. Geomorphological criteria for identifying Pleistoceneice streams. Annals of Glaciology 28, 67e74.

Stokes, C.R., Clark, C.D., Storrar, R., 2009. Major changes in ice stream dynamicsduring deglaciation of the north-western margin of the Laurentide Ice Sheet.Quaternary Science Reviews 28 (7e8), 721e738.

Tarasov, L., Peltier, W.R., 2004. A geophysically constrained large ensemble analysisof the deglacial history of the North American ice-sheet complex. QuaternaryScience Reviews 23 (3e4), 359e388.

Thorleifson, L.H., Wyatt, P.H., Shilts, W.W., Nielsen, E., 1992. Hudson Baylowland Quaternary stratigraphy: evidence for early Wisconsinan glaciationcentered in Quebec. Geological Society of America Special Paper 270,207e221.

Thorleifson, L.H., Wyatt, P.H., Warman, T.A., 1993. Quaternary stratigraphy of theSevern and Winisk drainage basins, Northern Ontario. Geological Survey ofCanada Bulletin 442, 59 p.

Tipper, H.W., 1971. Glacial geomorphology and Pleistocene history of central BritishColumbia. Geological Survey of Canada Bulletin 196, 89 p.

Van Landeghem, K.J.J., Wheeler, A.J., Mitchell, N.C., 2008. Seafloor evidence forpalaeo-ice streaming and calving of the grounded Irish Sea Ice Stream: impli-cations for the interpretation of its final deglaciation phase. Boreas 38 (1),119e131.

Veillette, J.J., Roy, M., 1995. The Spectacular Cross-striated Outcrops of James Bay,Québec. Geological Survey of Canada, Paper 1995-C, pp. 243e248.

Veillette, J.J., Dyke, A.S., Roy, M., 1999. Ice-flow evolution of the Labrador Sector ofthe Laurentide Ice Sheet: a review, with new evidence from northern Quebec.Quaternary Science Reviews 18 (8e9), 993e1019.

Vincent, J.-S., Prest, V.K., 1987. The Early Wisconsinan history of the Laurentide IceSheet. Géographie physique et Quaternaire 41 (2), 199e213.

Ward, B.C., Bond, J.D., Gosse, J.C., 2007. Evidence for a 55e50 ka (early Wisconsin)glaciation of the Cordilleran ice sheet, Yukon Territory, Canada. QuaternaryResearch 68 (1), 141e150.

North American Ice Sheet build-up during the last glacial cycle, 115–21 kyr

Documents