Classification

The mapping of the landscapes of glacial erosion employs a simple morphological classification used by the writer in Greenland and elsewhere {Reference Sugden, Waters and BrownSugden, 1974; Reference SugdenSugden and John, 1976). The main characteristics of the classification are described below.

A landscape of areal scouring comprises an irregular rock surface which has everywhere been dominantly shaped by the action of ice (Fig. 1). This landscape type has been described by Linton (1963) as knock-and-lochan topography. A good example of the landscape type occurs around the head of Frobisher Bay (Fig. 2). Here the most obvious feature is the widespread occurrence of glacially abraded rock surfaces, occasionally with striations and crescentic gouges still preserved. At a scale of hundreds or thousands of metres, these rock surfaces form part of smooth and elongated hills streamlined in the direction of ice flow. In between the hills are structurally controlled depressions which often contain irregular lakes. Sample measurements reveal a 1 000 m long hill to be commonly 60 m high and a 200 m long hill to be 10-15 ^m high. At this scale the profile of the hills is not broken up by irregularities induced by plucked or other lee-side effects. At a scale of 10-100 m the streamlined hills can be seen to comprise a series of individual rock knobs and depressions. Typically, the knobs are roches moutonnées with ice-moulded up-stream flanks and plucked and craggy lee sides. Boulders bounded by clear joint faces are common and include some which are still in place as well as those which have been moved horizontally some tens of metres from their original starting place.

Fig. 1. Vertical air photograph of a landscape of areal scouring, near Nettilling Lake, Baffin Island.

A landscape of selective linear erosion consists of glacially excavated troughs separated by upland plateaux with little or no sign of glacial erosion (Fig. 3). An example of this landscape type in the Coronation Fiord area of Baffin Island was examined in detail in the field. The fjords are troughs with steep cliffs rising precipitously above the fjord water to altitudes in excess of 950 m. The upland areas on either side of the troughs are gently sloping and marked by broad and open river valleys. The surface supports regolith derived in the main by weathering from the underlying bedrock but also from till containing distinctive glacial erratics. For example at an altitude of 910 m on the plateau just north of Coronation Fiord glacially faceted boulders of grey fine-grained gneiss and red coarse-grained pegmatite occur on bedrock consisting of coarse-grained garnetiferous biotite-gneiss. Bedrock tors are characterise of the upper surfaces and may be surrounded by erratics. One of the most striking features of this landscape type is the clarity of the break between the cliff top of the glacial trough and the gentle upper slopes. Other examples of this landscape type with seemingly identical characteristics have been described in the Torngat Mountains of Labrador and the Inugsuin Fiord area of eastern Baffin Island by Ives (1957, 1975).

Fig. 2. Distribution of landscapes of glacial erosion in the eastern Canadian Arctic, compiled from LAMDSAT-1 imagery and conventional air photographs, backed up by field work in selected sites.

A landscape of little or no sign of glacial erosion consists of an extensive area where the dominant landforms are river valleys and where long gentle slopes are swathed in regolith derived in the main from the underlying bedrock. Often tors occur on interfluves Such landscapes are common in the northern Canadian Arctic and are easily distinguished on air photographs The landscape type is characteristic of much of Somerset Island and has been described in detail by Reference DykeDyke (1976) (Fig. 4). Although there is no obvious sign of glacial erosion a former ice cover is frequently indicated by the occurrence of glacial erratics, for example, on Somerset Island (Reference DykeDyke, 1976) and elsewhere in the Arctic Archipelago (Blake, 1964, 1975).

Fig. 3. A landscape of selective linear erosion in the vicinity of Saglek Fiord, Labrador. (Photograph by J. D. Ives.)

It has been suggested by a referee that the lack of evidence of glacial erosion associated with the latter landscape type and the plateau areas of landscapes of selective linear erosion could represent areas where mass wasting has masked any landforms of glacial erosion since déglaciation. Whereas this may have occurred in a few localized places, the possibility of it explaining the basic classification given above can be excluded for two main reasons. In the first place, the amount of mass wasting in Arctic Canada since deglaciation appears inadequate to obliterate relief features associated with glacial scouring, at least on hard rocks. Whereas the glacial landforms have a characteristic amplitude of several tens of metres, the effect of mass wasting is measured in metres or less. This can be illustrated from Broughton Island, Baffin Island, by the preservation of such delicate features as lateral moraines 2-10 m high, which have been dated as being at least c. 50 000 years old and possibly more than 120 000 years old (Reference England and AndrewsEngland and Andrews, 1973). Also, the preservation of glacial erratics on hill crests and in the vicinity of tors on hill summits, for example on Broughton Island (Sugden and Watts, in press), implies relatively little mass wasting since déglaciation; if there had been significant landscape modification Ly mass wasting, erratics would tend to have been swept down into valleys or other depressions. In the second place, areas of little or no erosion represent relict surfaces bearing old regolith. One example of this in Baffin Island is illustrated by the discovery of thin Tertiary deposits on such a surface just north-west of the Barnes Ice Cap (Reference Andrews, Andrews, Guennel, Wray and IvesAndrews and others, 1972). Furthermore, studies of weathering pits and clay-mineral assemblages associated with the regolith of the upland component of a landscape of selective linear erosion near Coronation Fiord point to subaerial weathering over long time periods, probably measured in hundreds of thousands of years (Reference Boyer and PheasantBoyer and Pheasant, 1974; Isherwood, unpublished). Field work in the same area revealed glacial erratics in the regolith fresh enough to retain their glacial polish {Sugden and Watts, in press) and thus it seems clear that the formation of the regolith must pre-date at least the last full glaciation of the area. In such a case, the hypothesis of glacial protection implied by the classification given above seems fully compatible with the field evidence.

Fig. 4. Photograph from approximately 50 m above the ground of a landscape of little or no sign of erosion in Somerset Island. Erratics occur around the tors. (Photograph by A. S. Dyke; GSC-203014-D)

The final landscape type is an alpine landscape which consists of massifs shaped essentially by the action of local valley glaciers constrained in their flow by the topography of each individual mountain block. A coarsely dendritic trough pattern, horns and arêtes are characteristic. A spectacular example of the landscape type, much in demand for the making of Canadian whisky advertisements and Secret Service movies, occurs in the Mount Asgard area of Cumberland Peninsula, Baffin Island !

Distribution of the landscape types

The mapping of erosional landscapes relied heavily on the interpretation of LANDSAT-1 images, backed up by the scrutiny of several thousand oblique and vertical air photographs as well as field work in selected locations on Baffin Island. In practice it was found that LANDSAT images (especially bands 5 and 6) were easily interpreted in terms of the classification given above. Landscapes of areal scouring can be picked out by the irregular lake pattern, presence of clear structural lineations revealing the outcropping of bedrock at the surface and the absence of dendritic river patterns (Fig. 5). Areas of little or no sign of erosion are characterized by regular dendritic river patterns, the absence of structurally controlled lakes and a texture indicating the presence of regolith. An intermediate category (light areal scouring) was introduced to cover those areas where the river pattern is clearly preserved but where there are also a few lakes and some structural lineations. Landscapes of selective linear erosion are distinguished by troughs with clearly marked edges and intervening plateau areas covered in regolith and devoid of lakes. In those cases where the intervening upland is ice scoured, the whole landscape type was classified as one of areal scouring. Alpine landscapes are marked by a complex, coarsely dendritic trough pattern and the presence of pointed peaks.

Figure 6 depicts the broad distribution of main landscape types for the whole of the area covered by the Laurentide ice sheet at its maximum. In addition to the erosional categories discussed above, a depositional zone is included and describes those areas where bedrock is largely obscured by drift. The map, covering as it does such a large area, is obviously highly generalized and is intended to be no more than a guide to the broad pattern of landscape variation. The main feature is a broad area of areal scouring extending over the central area of the former ice sheet. Along the south-eastern maritime margin, the zone extends at least as far as the coast, while along the southern and western land margins the zone gives way to a peripheral zone of deposition. In the north the areal scouring gives way to landscapes with little or no sign of glacial erosion. This transition takes place mainly beneath the waters of Viscount Melville Sound but in several places the transition is abrupt and takes place on land. For example, on Somerset Island, Reference Netterville, Netterville, Dyke, Thomas and DrabinskyNetterville and others (1976) have described the change between the characteristic ice-scoured depressions and rock knobs south of the boundary and the fluvial topography, regolith and tors of the area immediately to the north. Landscapes with little or no sign of glacial erosion include most low-lying islands in the Queen Elizabeth Islands. Although there is some discussion about the status of these islands during the Wisconsin glaciation (Reference BlakeBlake, 1975; Reference EnglandEngland, 1976), there is general agreement that they were covered by Laurentide ice at some stage during the Cenozoic, a conclusion apparently confirmed by the existence of Laurentide shield erratics on the Palaeozoic rocks of the islands. The topography beneath the straits is obviously not well known. However, there are indications that it includes both areal scouring and areas with little or no sign of erosion. For example, Reference BornholdBornhold and others (1976) described well-preserved river patterns as well as ice-scoured topography from the floor of Barrow Strait immediately north of Somerset Island. Landscapes of selective linear erosion are, with one exception on Melville Island, restricted to the eastern upland periphery of the former Laurentide ice sheet. The landscape type is most common in the north and most restricted in the south-east. Probably the extent of selective linear erosion is underestimated in the north-west. Here, troughs appear to occur in some straits (Pelletier, 1966) and these could justify classification of more of the Queen Elizabeth Islands as selective linear landscapes, albeit well submerged.

Fig. 5. The appearance of different landscape types on LANDSAT-1 imagery. The image is of an area south of Eclipse Sound in Baffin Island and shows areal scouring in the south, selective linear landscapes in the north-east and little or no sign of glacial erosion in the intermediate area in the east. North is at the top of the image. (E- 1746-16434-6.)

Fig. 6. Main glacial landscape zone, Laurentide ice sheet, compiled form examination of LANDSAT-1 imagery. Topographic map were used in the far north where imagery is unavailble.

Alpine scenery within the confines of the former Laurentide ice sheet occurs patchily along the north-eastern periphery and in part of Labrador. In all cases, alpine areas are adjacent to or surrounded by selective linear landscapes.

Figure 2 shows the erosional landscapes of the eastern Canadian Arctic in detail and relies mainly on the examination of several thousand oblique air photographs and field testing of crucial areas. Clearly, in any such map there will be areas of doubt but the map is offered as a representation of the main pattern and provides a basis for discussion. Landscapes of areal scouring cover most of the area other than the northern and eastern periphery. Both flanks of Hudson Bay and Foxe Basin and southern Baffin Island are heavily ice-scoured. One clear pattern is that uplands are often less intensely ice-scoured than surrounding adjacent lowlands. Many good examples of this relationship can be seen along the axes of peninsulas in southern Baffin Island, Boothia Peninsula and on Southampton Island. Landscapes with little or no sign of erosion form isolated pockets over the northern part of the area. It is clear that one category of this latter landscape type is associated with uplands. In the north the bulk of the central high parts of Devon and Cornwallis Islands fall into this category. Farther south the occurrences are less extensive but the same relationship is apparent. For example, in several places on Baffin Island, Southampton Island and Boothia Peninsula, areas of little or no sign of erosion include the culminating points of uplands which are increasingly affected by areal scouring down their flanks. Another type of landscape with little or no sign of glacial erosion includes the low-lying forelands of central and northern Baffin Island. Landscapes of selective linear erosion are extensive in the north in eastern Devon Island and Elles mere Island and in northern and central Baffin Island, but are restricted to small upland areas in southern Baffin Island and the Torngat Mountains of Labrador.

Intensity of glacial erosion

Although Figures 2 and 6 are helpful in pinpointing the location of different types of glacial erosion, they are not able to provide quantitative information about the intensity of erosion from place to place. The information is potentially important and several attempts were made to derive it. Simple tests using such indicators as degree of stream derangement, streamlining as reflected by lakes, and preferred orientation of depressions proved either unwieldy or unhelpful. However, density of lake basins proved more helpful. Lake basins have long been regarded as diagnostic of glacial erosion and indeed Hobbs (1945) used their distribution as a means of determining the former extent of the Laurentide ice sheet. For the purpose of this paper, lake-basin density was measured for that part of Canada formerly covered by the Laurentide ice sheet. Each Canadian topographic map at a scale of 1 : 500 000 (latest editions available in 1975) was divided into four equal parts. Within each sector a quadrat representing an area 20 km by 20 km in size was randomly located and lake basins of 0.5-2.0 km in diameter were counted. Areas of sloping topography were avoided where possible as well as areas of large lakes. It is impossible to know how many of the lakes sampled were actually excavated in bedrock. However, structural control of the lake borders was obvious in about 90% of cases and it is unlikely that the inclusion of some non-bedrock lakes will seriously affect the pattern. One obvious area of difficulty was in the Mackenzie River delta, where there are many "periglacial" lakes, and as a result this area was left blank.

The resulting map (Fig. 7) reveals a clear pattern which is in full agreement with the maps of landscape type. The boundary between little or no sign of erosion in the Queen Elizabeth Islands and areal scouring to the south is borne out by the contrast in lake density. To the north there are less than ten lake basins per $$400 km², whereas immediately to the south there are between 10 and 50 lake basins per 400 km$$¹. This latter density soon increases southward to densities of 50-100 lake basins per 400 km$$^s or more. Also of interest is a zone with a low density of lakes along the eastern seaboard between Labrador and Lancaster Sound. Within this zone locally high densities occur in the vicinity of Hudson Strait, Cumberland Sound, Home Bay, and the Cambridge Fiord area.

Perhaps the most intriguing pattern is the zone of high lake-basin density (> 100 per 400 km$$$³) situated approximately midway between the shores of Hudson Bay and the former ice-sheet margin. In the west the zone is continuous between the high western plains and the Arctic Archipelago. In the east it is less continuous but a recognizable zone swings round from the Great Lakes through the Ungava Peninsula and is continued in two places in Baffin Island. If lake-basin density is accepted as a measure of the intensity of glacial erosion, one can suggest that there is a zone of maximum erosion situated between the former ice-sheet centre and the periphery. The reality of this pattern is borne out by two independent maps which are reproduced in the National atlas of Canada (1974). One (p. 6-7), showing the area of fresh water as a percentage of 10 000$$$ km grid units brings out the same zone as one with as much as 10-20% of its surface area comprising water. The other (p. 37-38), showing surface materials, portrays areas where bedrock comprises most, or half of the surface materials. In addition to steep mountain areas along the east coast, there is an arcuate zone running from the Great Lakes to Victoria Island which is coincident with the postulated zone of maximum erosion. On either side, most of the surface material consists of unconsolidated drift. Such a patterns would be expected if there were a zone of maximum glacical erosion approximately midway between Hudson Bay and the western ice-sheet margin.

Fig. 7. The density of lake basins 0.5-2.0 km in diameter for unit areas of 400 km² in canade. The data were obtained by sampling from topographic maps at a scale of 1:500 000. Lake-basin density at this scale is thought to be a measure of the intensity of glacial erosion.

At present there seems no evidence to contradict this view of a zone of maximum glacial erosion midway between the ice-sheet centre and periphery. For long there has been a view that an arc of exhumation between shield and Palaeozoic rocks may be the result of glacial erosion (Reference DavisDavis, 1920; Reference WhiteWhite, 1972), although as Gravenor (1975) has pointed out, there is often good local evidence that this existed as a depression in pre-glacial times and has merely been deepened by erosion in places. It is notable that the greatest deepening of this arc occurs in the vicinity of the Great Lakes and coincides with the intermediate zone of maximum erosion postulated in this paper. In 1972, White suggested a contrary view that the zone of maximum glacial erosion occurred beneath the centre of the former Laurentide ice sheet, namely over Hudson Bay. However, this view was put forward as a speculative hypothesis and is contradicted by a great deal of field evidence (Gravenor, 1975; Sugden, 1976).

Theoretical considerations

Of all glaciological variables, there are grounds for believing that basal thermal regime is the most critical affecting the nature of erosion beneath ice sheets. Following from the work of Weertman (1961, 1966), Boulton (1972) suggested that the basal thermal regime controls the occurrence or otherwise of basal slip between ice and rock, the amount of melt water at the base, the relative importance of the mechanisms of abrasion and plucking, and the processes of entrainment of rock particles. Some of the possible relationships have been discussed by Sugden and John ( 1976), and are illustrated in Figure 8 which shows an idealized sequence of zones and primary erosional processes associated with each. From this, it can be suggested that cold-based ice is essentially protective unless it contains basal debris, inherited perhaps from a zone of basal regelation up-glacier. In this latter situation, slight convexities on the bed may cause ice to diverge round the obstacle and bring debris into contact with the bed so that it may accomplish erosion, as suggested by Rôthlisberger ( 1968) ; also some entrainment of debris may take place by plastic flow of ice round already loosened debris. Zones of warm melting (ice at the pressure-melting point and subjected to melting) are associated with erosion by abrasion, fracture, and melt water, while zones of warm freezing (ice at the pressure-melting point but with successive freezing on of melt water) are associated in addition with entrainment by regelation.

The different types of erosional landscape may reflect differing basal thermal regimes. Areal scouring reflects extensive abrasion and plucking, and would be most favoured by basal ice at the pressure-melting point. Probably, areal scouring can also result from the passage of cold-based ice when it is carrying a basal debris load. Landscapes of little or no sign of glacial erosion would be likely to occur whenever the basal ice is cold-based and carrying little debris ; in such situations most forward movement of the ice would result from deformation within the lower layers of the ice, rather than between ice and bedrock. Landscapes of selective linear erosion are intermediate in that they may reflect a situation where ice is warm-based over the sites of troughs and cold-based over the intervening plateaux.

Fig. 8. Idealized model of the relationship between processes of glacial erosion and the basal thermal regime of an ice sheet

Topography has an influence on landscapes of glacial erosion in two main ways. It can have a direct effect by determining whether or not a particular massif will be submerged beneath the ice sheet. Thus it can be suggested that topography will be the dominant factor m determining whether or not a particular massif is eroded by ice-sheet ice or local valley glaciers. Topography can also play an indirect role by influencing the basal thermal regime beneath the ice sheet. This may be through its influence on the thickness of the overlying ice Temperatures beneath thick ice tend to be higher than those beneath thin ice because with the exception of the surface layers, temperatures tend to rise with depth. Topography also affects basal temperatures through its influence on converging or diverging flow. In an area where the bedrock topography favours convergent ice flow, ice velocities will be increased and this will produce more heat by internal deformation and/or basal sliding. On the other hand topography which causes ice divergence will tend to reduce overall velocities and thus the amount of frictional heat generated.

The geology of the ice-sheet bed can be expected to have important implications on the pattern of glacial erosion at a broad and local scale. The influence of joint patterns fracture characteristics, and hardness are all variables which have yet to be considered from a detailed theoretical standpoint and little can be said about them at present. However, permeability is thought to be important in so far as it affects the amount of water present beneath warm-based ice and thus affects the rate of basal slip (Weertman, 1966) and the ice pressure applied to the glacier bed (Boulton, $$[«1974], [«1975]). Beneath the thick ice of the Laurentide ice Sheet, it is reasonable to suppose that the main effect of permeable beds, especially if they are extensive, would be to allow basal melt water to flow through the beds rather than at the ice/ rock interface. This in turn is likely to reduce the rate of erosion. One reason for this is that the rate of basal slip will be reduced and thus less glacier sole and abrasive passes a given point on the bed. Another reason is that the loss of water at the ice/rock interface may increase ice pressures to such an extent that debris particles are arrested by friction with the bed and are deposited as lodgement till {Boulton, $$$[^c1975]).

Testing the relationship between landscape type and basal thermal regime

The relationship between landscapes of glacial erosion and basal thermal regime can be examined by comparing the landscape distribution and character with the distribution of the differing basal thermal regimes beneath the former Laurentide ice sheet. Simulating the basal thermal regime of a former ice sheet is obviously an exercise fraught with difficulty and m order to give adequate treatment of its limitations I have felt obliged to complete the simulation in a separate paper (Sugdcn, 1977}. The main features of the simulation are the use of morphological and climatic input data obtained by analogy with the existing ice sheets of East Antarctica and Greenland, and the use of the steady-state ice-temperature model developed by Budd and others (1971). Figure 9 shows the result of this reconstruction of the thermal regime beneath the Laurentide ice sheet at its Cenozoic maximum. The main feature is a zone of warm melting beneath the central area of the ice sheet. Surrounding this central zone is a ring of warm freezing followed by a ring of cold-based ice and finally, in most places, a peripheral zone of warm melting. It is likely that the width of the warm-freezing zone is underestimated to some extent because the model used did not allow for the horizontal advection of heat by melt water flowing beneath the ice sheet or the complex problem of the amount of heat produced by the freezing of melt water in the cold-based zone. The pattern was subjected to a sensitivity test by changing the input data and it was concluded that the sequence of zones was relatively persistent but that small changes in the input data could shift zone boundaries and widths by hundreds of kilometres. This was especially true of the western plains where the pattern was least stable. The cold-based ice over the Queen Elizabeth Islands was found to be especially stable and persisted m spite of all reasonable changes in input variables. It is notable that the main feature of the simulation—the central zone of warm melting beneath the ice sheet at its greatest maximum—was discovered previously by Hughes (1973) using independently derived input data and the same ice-temperature model. The spatial extent of the zone agrees well in both simulations.

Fig. 9. Zones of contrasting basal, thermal regime beneath the Laurentide ice sheet at its maximum. (Reproduced with the permission of the University of Colorado from Sugden 1977).

The simulation of the ice sheet was made for maximum steady-state conditions on the assumption that most erosion was achieved during maximum ice conditions. There are pieces of evidence which seem to suggest this assumption is realistic. One is simply that such ice-sheet landforms as the larger fjords of eastern Baffin Island and northern Labrador which cut through the main mountain crest occur at the peripheries of the ice sheet and can hardly have been occupied by ice unless the ice sheet was very close to its maximum. Another line of evidence is that erosional forms seem to relate to maximum ice-flow conditions and are relatively unaffected by subsequent changes, as for example occur during déglaciation. An example of this situation is provided by the Ottawa Islands in eastern Hudson Bay Here Andrews and Falconer (1969) noted that the streamlined outline of the islands reflects maximum. ice-flow conditions out of Hudson Bay, and yet overlapping striations of completely different orientations tell of the phase of déglaciation during which the ice centre migrated to Labrador-Ungava. The deglaciation phases were capable only of carving some striations into the previously ice-moulded surfaces. Other similar associations occur west of the Barnes Ice Lap in Baffin Island where a complete reversal of ice direction during déglaciation has had little impact on the main ice-scoured forms (Ives and Andrews, 1963). Even if the view that most erosion was achieved during maximum conditions must remain an assumption at this stage, there is evidence that the less extensive Northern Hemisphere late-Wisconsin ice sheets accomplished little erosion and that therefore a reconstruction of earlier ice sheets is most relevant to a study of glacial erosion. This view of limited late-Wisconsin erosion is based on the juxtaposition of erosional forms, for example in Europe (Ahlmann, 1919), and on the $$$$A??·» ^{f mter}SJ^{acial de}P°^{sits on}976) glaciated surfaces, for example in West Greenland (Sugden and Miller, $$$$$xgft _m eastern Baffin Island (Feyling-Hanssen, i₉₇6) and around southern Hudson Bay (McDonald, 1969; Prest, 1970).

Comparison of the map of basal thermal regime with those of landscape variation poses several suggestive correlations. One is that the area of little or no erosion over the Queen Elizabeth Islands coincides with an area where the ice sheet was cold-based. In the absence of any obvious correlation with either topography or geology, it is reasonable to suggest that here may be a causal link between the former occurrence of cold-based ice and the areas of little or no erosion. Devoid of basal debris, it is to be expected that such cold-based ice would be essentially protective. The only exceptions occur in the area of the deeper straits where calculations suggest that the basal ice would have been at the pressure melting point It is notable that the deeper straits often contain glacial troughs (Pelletier, 1966) and also that zones of light area! scouring are sometimes associated with the sides of straits, for example between Cornwallis and Devon Islands (Fig. 2).

Another interesting relationship is the apparent coincidence of the zone of most intensive glacial erosion with the zone of warm freezing. The warm-freezing zone provides an efficient means of debris evacuation because successive regelation of melt water and associated debris allows the build-up of a considerable thickness of debris-bearing ice. Sample calculations for the Laurentide ice sheet suggest that the layer would have been of the order of 20-50 m thick it is tempting to suggest that the zone of most intense erosion is associated with this zone of efficient debirs evacuation. There is not a perfect spatial correlation between the zone of intensive erosion and the warm-freezing zone but neither should this be expected. In the first place, the width and position of the zone as portrayed in Figure 9 is highly sensitive to the amount of melt water produced beneath the centre of the ice sheet. It may be that climatic conditions were different to those used in the reconstruction with the result that the position and width of the zone would have been different. In the second place, it is highly probable that the position of the zone fluctuated during the course of a glaciation as maximum equilibrium conditions took time to be established. In the third place, it is likely that the simulation underestimates the width of the warm-freezing zone.

The remarkable feature is that the correlation is as good as it is. One might point out that the correlation agrees over such details as the relative width of the zone which is wider in the west than in the east. It might be suggested that geological variations affect the position of the zone of intense erosion. For example, the western edge of the zone on the western plains coincides with the edge of the shield. Whereas such geological conditions may partly influence the position of the zone in this area, it is sufficient at this stage to note that the pattern also occurs in Labrador-Ungava where there is a more uniform spread of shield rocks both within and on either side of the zone of intense erosion.

Fig. 10. The relationship of alpine landscapes to the former maximum Laurentide ice-sheet surface profile. Alpine landscapes are confined to massifs sufficiently high to have escaped inundation by the ice sheet.

The central area of areal scouring around Hudson Bay coincides with an area where the basal ice was at the pressure melting point. This relationship is as expected. However, the areal scouring that occurs between the zone of intense erosion and the former ice-sheet periphery coincides with both cold-based and warm-melting zones. There are several possible explanations and a full answer must await more detailed field study than has been attempted in this paper. One possibility is that the zone has been eroded beneath cold-based ice which was carrying a considerable thickness of debris inherited from the warm-freezing zone immediately up-stream. Another possibility is that some areas, especially nearer the ice-sheet periphery, have been subjected to erosion beneath the outer zone of warm-based ice. This zone was found to be the least stable of all and relatively modest changes in the climatic input variables could dramatically increase the width of the zone. Yet another possibility is that the areal scouring in this intermediate zone relates to conditions before or after the steady-state pattern of Figure 9 was established.

Fig. 11. Profile parallel to the eastern margin of the Laurentide ice sheet and 130 km in from the margin, showing relationships between landscape type, topography and ice-sheet thickness. Differential isostatic depression between the margin and the line of the profile has been allowed for.

There is no obvious direct relationship between landscapes of selective linear erosion and broad variations in basal thermal regime. In Greenland there is a tendency for the landscape type to occur in environments where the maximum ice sheet was nourished under a continental climatic regime rather than a maritime climate (Sugden, 1974). Such a tendency may also be reflected in eastern North America where it can be seen that the landscape type occurs more frequently in the north than in the south.

Testing the relationship between landscape type and topography

At the scale of the Laurentide ice sheet as a whole, the role of topography seems subdued and relatively unimportant. Much of the bed of the former ice sheet is relatively flat and gently sloping. The main exception is in the east, where the land rises to form an uplifted, westward sloping plateau. It is in these eastern and northern areas that topography has played an important role in influencing glacial erosion.

Alpine landscapes occur along the western periphery of the Laurentide ice sheet and, as might be expected, appear to coincide with those massifs which rose above the maximum ice-sheet surface profile (Fig. 10). The ice-sheet profile in Figure 10 is based on the assumption that the ice was grounded as far as the 200 m submarine contour. Inland of this margin, the surface profile was obtained by analogy with existing stable ice sheets and has the form$$$ y = a+bx+cx¹, where a = 0.607, * = 0.005 ^{and c} = —0.2 X io"*$$$ (Sugden, 1977). The mountain altitudes were obtained by taking the mean of the three highest peaks in eachalpine area and these were then plotted at the appropriate distance inland from the ice margin. In addition, the closest adjacent massif unaffected by alpine glaciation was also plotted on the diagram. Following Walcott (1970), an allowance was made for isostatic depression of the ground following inundation by the ice sheet which amounts to 100-200 m at the ice margin and 200-400 m at a distance of 200 km from the ice margin. These figures give differential isostatic depression of $$$c. 200 m between the ice margin and a point 200 km inland and this has been built in to Figure 10. Bearing in mind the assumptions involved in constructing such a diagram, the apparent correlation in Figure 10 between alpine scenery and those massifs which rose above the ice-sheet profile seems too good to be fortuitious. This is especially so when one realizes that the summits of an alpine area may be only a few hundred metres higher than the summits of a non-alpine plateau only a few tens of kilometres away. A good example of this latter relationship is provided by the contrast between the Mount Asgard alpine area of Cumberland Peninsula and the adjacent slightly lower plateau upland around the Penny Ice Cap (Fig. 11).

In view of this apparent correlation, it seems fair to suggest that the contrast between plateau and well-developed alpine scenery reflects the contrast between those areas once submerged beneath the maximum Laurentide ice sheet and those massifs which were sufficiently lofty to escape inundation. As a result, these upstanding massifs were subjected to periglacial and local glacier action throughout the glacial age. As would be expected, the local glaciers were constrained by the shape of the massif on which they formed and thus exploited pre-existing river valleys to form the dendritic trough pattern of today. At lower levels some of these massifs are traversed by troughs whose orientations clearly reflect the action of ice-sheet drainage.

Topography has an important influence on the location of landscapes of selective linear erosion. All occurrences of the landscape type are associated with cither the upland plateau which forms the eastern upland rim of North America or with uplands in the Arctic Archipelago (Fig. 11). The reason for this relationship may be that topography influences the basal thermal regime of the ice sheet so that the ice is warm-based over the troughs and cold-based over the intervening plateaux. The map of basal thermal regime in Figure 9 suggests that most of the upland rim of eastern Canada was covered with cold-based ice. The main exceptions are parts of Baffin Island where there were locally high ice velocities (owing to the thinness of the ice cover) which created sufficient frictional heat near the base for the ice to be warm-based. However, the calculations in Figure 9 were carried out for an average ice thickness over a 100 km wide "stream" of ice. This means that any channelling effect of valleys was ignored. In such a case, it is probably more realistic to imagine most ice being discharged down the valleys and that ice velocities and thus frictional heat was also concentrated in the valleys at the expense of the intervening uplands. Sample calculations assuming that 50-75% of the ice was discharged through the valleys confirmed that the ice would have been cold-based over the plateaux and warm-based in the valleys. Moreover, the calculations suggest that this topographic effect causing convergence or divergence was about ten times more important in creating this pattern of basal thermal regime than the other contributing factor of varying ice thickness.

The topographic relationships just described may help explain the pattern once the glacial troughs were excavated to approximately their present size. However, they do not necessarily explain the origin of the landscape type. To do so, it would seem necessary to postulate the existence of pre-glacial valleys incised into the upland edge of Canada. Any such irregularities would channel ice flow as soon as the ice sheet built up. It is notable that the existence of such pre-glacial river valleys has been postulated on various grounds by others (Mercer, 1956; Ives', 1957; Sim, 1964; Bird, 1967). Moreover, the plan view of many fjord systems would seem to support this hypothesis. Many fjords are sinuous and when viewed as a whole they form roughly dendritic patterns (with varying degrees of tectonic control)—just as would be expected if they followed pre-existing river valleys (Fig. 12). A particularly good example of a dendritic fjord system following pre-existing river valleys occurs in the vicinity of Nansen Sound between Axel Heiberg and Ellesmere Islands (Hattersley-Smith, 1961).

Topography clearly influences the distribution of landscapes of areal scouring at a local scale. A common pattern around Hudson Bay, Foxe Basin, and southern Baffin Island is for uplands to be lightly scoured and to be surrounded by lower, heavily scoured landscapes.

Fig. 12. Dendritic fjord pattern in the Cambridge Fiord area, north-east Baffin Island, as seen in LANDSAT-1 image.

In some situations the upland has escaped all visible signs of glacial erosion. In the central areas of the ice sheet, where thicknesses were of the order of 3 000-4 000$$$⁰⁰° ^m= there would seem to be two main reasons for this, namely the reduced ice thickness over the upland and the tendency for ice to diverge round the upland. Both characteristics would cause lower than average temperatures over the upland. One could speculate that areas with little or no sign of glacial erosion, such as parts of Southampton Island and Hall Peninsula, may have been overlain by cold-based ice and that most debris-carrying ice was diverted round them. Calculations confirmed that the relatively minor modifications to the ice-temperature profiles accompanying such topographic differences would be sufficient to cause such a pattern in the basal temperatures.

Farther north, the topographic relationships of landscapes of areal scouring are different. As areal scouring becomes less common towards the north it is increasingly restricted to areas of ice convergence. Two excellent examples occur in Baffin Island in the vicinity of Home Bay and Cambridge Fiord. Figure 11 shows how these two areas are topographic lows across the uplands of Baffin Island. Calculations of ice discharge along the coast of Baffin Island, made on the basis of the form and flow lines of the ice sheet, suggest that these two zones were evacuating between six and nine times as much ice per unit length of coast than the intervening area. Clearly, such a contrast can be expected to lead to important contrasts in the creation of frictional heat. Other fine examples of the association of areal scouring with zones of convergence can be seen at the head of Eclipse Sound, on either side of Somerset Island and on southern Devon Island (Fig. 2). In all these cases it can be suggested that much of the ice sheet was cold-based and that frictional heat was able to produce enough heat to raise the basal ice to the pressure melting point only where there was convergent ice flow.

The relationships between topography and areal scouring may be conceived as different ends of a continuum in that they are local variations superimposed on an overall tendency for basal ice temperatures to fall from the ice-sheet centre towards the north. Areal scouring becomes progressively less common towards the north where it is increasingly confined to especially favourable topographic locations. The converse argument also applies to landscapes with little or no sign of glacial erosion.

Testing the relationship between landscape type and geology

Geological variations seem most important at a local scale, though their importance at a macro-scale cannot be overlooked in the particular case of North America. It can be suggested, for example, that in general there is a tendency for areal scouring to be associated with shield rocks and minimal erosion or deposition to be associated with younger Palaeozoic sediments, However, examination of the shield boundaries reveals many anomalies which do not fit this simple generalization. A particular example is Victoria Island which is clearly affected by areal scouring and yet which is largely underlain by Palaeozoic rocks. At a more local scale the role of geological variation is clearer. Examples may be taken from northern Baffin Island and Somerset Island where there are extensive outcrops of Palaeozoic rocks, mainly limestones. Although there are no field measurements on which to base a generalization, one might predict that these rocks are more permeable than the adjacent shield and that therefore there would be less erosion. Such appears to be the case. On Somerset Island, the western boundary between areal scouring and areas with little or no sign of glacial erosion roughly coincides with the boundary between Palaeozoic limestones and shield rocks. Moreover, an area of little or no sign of glacial erosion on Brodeur Peninsula coincides almost exactly with an exposure of Silurian limestone, although in this case it is impossible to exclude the contributory role of topography since the area is also an upland. To the west and south of northern Baffin Island and Somerset Island, there is no such relationship between permeable rock and landscapes with little or no sign of erosion. Large areas of Melville Peninsula, Southampton Island and southern Baffin Island are underlain by Palaeozoic limestones and yet are ice-scoured.

A partial answer to this apparent paradox may concern the relationship of geological variations to other broad trends within the Laurentide area. The north Baffin and Somerset Islands areas lie close to a boundary between areal scouring to the south and little or no sign of erosion to the north. In such a marginal position, variation in rock type may be able to play a role. Farther south, any local variation introduced by changes in rock type may be insufficient to overcome the overall tendency for areal scouring. As with topography, geological variations then seem to be able to produce only limited local variations to a dominant overall pattern. Clearly, this is an hypothesis which needs much more stringent testing than has been attempted here.