Twice during the climatic changes of the Glacial and post-Glacial periods, Lake Bonneville, described by Gilbert,(1) and Lakes Lahontan and [p.193] Mono, described by Russell,(2) have risen nearly or quite to overflowing. The climate of the Great Basin of interior drainage, which comprises the areas of these lakes, was marked at these times by considerably greater humidity than now, though to less degree than the present climate of the eastern half of the United States. The humid epochs were divided by a long interval of aridity, in which, as Gilbert and Russell have shown, the lakes were perhaps wholly evaporated, their soluble salt and alkaline mineral matter becoming intermingled and covered with playa silts, so that it could not be redissolved by the water of the lakes during their second rise, which therefore may have been nearly fresh.
        Lake Bonneville, the largest one of the many lakes which were formed during the Pleistocene period in the Great Basin, covered at its maximum stage an area of 19,750 square miles, lying mostly in northwestern Utah, but extending also into the borders of Nevada and Idaho. It was about ten times as large as its present representative, Great Salt Lake, which, having a mean height of 4,208 feet above the sea, lies 1,000 feet below the highest of the ancient shore-lines. The maximum depth of the Pleistocene lake was about 1,050 feet, while that of Great Salt Lake, in its range from the lowest to the highest stage within the past forty years, is from 36 to 49 feet. The hydrographic basin of Lake Bonneville comprised a fourth part of the Great Basin, whose total area is estimated to be 210,000 square miles; almost another quarter was tributary to the companion Lake Lahontan, which attained an extent of 8,422 square miles, occupying a very irregular tract of interlocking valleys among mountain ranges in western Nevada; and the remaining half of this arid region held some twenty-five smaller lakes, much exceeding, however, the saline lakes and playas to which they are now reduced.
        The first great rise of Lake Bonneville, lifting its level to within 90 feet of the lowest point of the inclosing watershed, is recorded by numerous beaches, marking the oscillations of the lake level under the varying influence of secular climatic changes, and by a thick deposit of yellow [p.194] clay. A long interlacustrine epoch is known by overlying alluvial gravel and sand. The second rise of the lake reached the level of overflow apparently after the water surface had been long held within 5 to 20 feet below that level, forming a widely spread deposit of white marl and the well-defined highest beach ridges and eroded cliffs, which Gilbert names the Bonneville shore-line. The time required for the great amount of wave work at this level would be made possible by long-continued underground drainage from the lake through the alluvial deposit of Cache Valley, over which a slightly higher rise of the lake finally gained a superficial outflow to the Columbia River, and then rapidly cut a channel 375 feet deep in the alluvium to a sill of limestone. At this lower level, marked by the Provo shore-line and deltas, the lake was held for a long time, perhaps occasionally interrupted by dry climate and fall of the water too low to maintain its outlet.
        Glaciers descending the canyons on the west front of the Wasatch Range attained their maximum extent, pushing their moraines into Lake Bonneville, during the time of formation of the Provo shore-line. From these moraines, and from those of the Sierra Nevada extending into the Pleistocene area of Lake Mono, the glaciation of the Cordilleran region is known to have been contemporaneous with the epochs of humid climate and extension of lakes in the Great Basin, the interlacustrine epoch being attended probably with a nearly or quite complete departure of the glaciers and ice-fields on the mountains.

        A glacial lake, according to my use of the term in this volume and elsewhere, is a body of water bounded in part by a barrier of land ice. The lake may be hemmed in by a glacier, as the Merjelen See, or by a continental ice-sheet, as Lake Agassiz. And the same name is also applicable to the lakelets, wholly bounded by ice, which are occasionally formed, attaining a considerable depth and extent and appearing in the same places during the summers of successive years, on the surface of [p.195] glaciers, as in the Himalayan Range, or on an ice-sheet, as observed by Nordenskjöld in Greenland.
        Very abundant and extensive development of glacial lakes attended the recession of the ice-sheet in the northern United States and in Canada, being due to the temporary damming of the waters of glacial melting and of rains on areas where the land has a northward descent. While the ice-sheet was melting away from south to north on such a slope, free drainage was prevented, and a lake was formed, overflowing across the lowest point of what is now the southern watershed of the basin. Many of these lakes were of small extent and short duration, being soon, by the continued retreat of the ice, merged into larger glacial lakes, or permitted to flow away where basins sloping northward are tributary to main river courses draining southward. Professor Chamberlin has well written of these lakes fringing the ice-sheet:

        They vary in areal extent from trivial valleys blocked by ice to the broad expanses of the great basins. If an attempt were made to enumerate all instances, great and small, and all stages, earlier and later, the list of localities and deposits would swell, not by scores and hundreds, but by thousands.(4)

        Five principal evidences of the former existence of glacial lakes are found, namely: (1) Their channels of outlet over the present watersheds; (2) cliffs eroded along some portions of the shores by the lake waves; (3) beach ridges of gravel and sand, often on the larger glacial lakes extending continuously through long distances; (4) delta deposits, mostly gravel and sand, formed by inflowing streams; and (5) fine sediments spread widely over the lacustrine area. A few words of general description may be given to each of these before proceeding to notice the areas of some of the more important glacial lakes formed by the waning North American ice-sheet, and to trace in detail the stages of the growth of Lake Agassiz and its final reduction to the present Lake Winnipeg.
        Outlets.--Among the evidences of glacial lakes, the one most invariably recognizable and most definite in its testimony is the outlet showing [p.196] distinct stream erosion across the rim dividing adjacent river basins, which now in many instances send their waters respectively to the Gulf of Mexico and to Hudson Bay or the Gulf of St. Lawrence. Obviously, watercourses could exist in these positions only as the outlets of lakes which were pent up by some barrier that is now removed. Shore-lines traceable northward from these deserted channels must therefore belong to a lake, and can not be regarded as the record of any marine submergence.
        Closely associated with such channels crossing watersheds, and at the same level, are the three following classes of proof cited, namely, eroded cliffs, beach ridges, and deltas; and below these shore records are the fine lacustrine sediments. These are found in hydrographic basins which are now drained by a continuous descent northward, presenting no indication that any land barrier ever existed across their lower portions to form these lakes, being afterward removed by erosion or by depression. The shore-lines, as shown thus by wave-cut cliffs, wave-built beaches, and deltas brought by inflowing rivers, extend far along both sides of the present hydrographic basin, often rising slightly and regularly northward, instead of sinking in that direction, as they would do if there had been a depression of the land at the north. When traced carefully with leveling, they are found, sometimes after an extent of hundreds of miles, as on the glacial Lake Agassiz and about the great lakes tributary to the St. Lawrence, to terminate abruptly where the basin attains its greatest width. Hence it is manifest that the barrier of these lakes could not have been land formerly raised higher than now, but was the receding ice-sheet, against which the land shores terminated.
        On slopes descending in parallelism with the retiring ice border, drainage from it in many places flowed in channels from which the streams became turned into new and more northerly courses as the ice retreated. Several glacial river courses of this kind I have observed between the Coteau des Prairies and the Minnesota River.(5) Others have been noted by G. M. Dawson,(6) McConnell,(7) and Tyrrell,(8) in various parts of Alberta [p.197] and Assiniboia. But these seldom were outlets of glacial lakes of large size. It was only when extensive hydrographic basins were inclined toward the ice-sheet that broad glacial lakes, as those named Lake Agassiz, Lake Souris, Lake Saskatchewan, and the greatly enlarged Laurentian lakes from Superior to Ontario, were held between the northwardly sloping land and the waning ice-sheet with long-continued outflow across the present main watersheds of the continent.
        The depth of erosion of these outlets varies from 50 feet or less to 150 feet or more. So far as known to me, they are cut through the easily eroded drift deposits, and sometimes beneath these, on the extension of the great plains in the Canadian Northwest, through Cretaceous shales or clays and soft, unconsolidated sandstones, which could be easily worn away. Nowhere is it found that a glacial river has channeled deeply into the harder rock formations. The time required for the work observed was brief.
        A noteworthy feature of many of the old watercourses which were outlets of glacial lakes, then carrying a much greater volume of water than now, is the occurrence of long and narrow lakes in such valleys, of which Long Lake, in Assiniboia, lying on the west side of a high remnant of the eroded Cretaceous strata called Last Mountain, is a conspicuous example. This lake, occupying one of the channels of outflow from the glacial Lake Saskatchewan, which thence continued down the Qu'Appelle Valley, is about 50 miles long from south to north and 1 to 2 miles wide. Its southern end is separated from the Qu'Appelle River by alluvial deposits only a few feet above Long Lake, which have been brought into the valley since its great glacial river ceased. Similar]y the Qu'Appelle Valley has been partly refilled by the postglacial deposits of its tributaries, and the present stream in its course through the Fishing Lakes flows at a level about 60 feet above the ancient river bed. Other rivers which thus flow through lakes produced by postglacial alluvium in the beds of the outlets of glacial lakes are the James River, formerly the outlet of Lake Souris; the Pembina River, which, with Langs Valley, afforded a later outlet from Lake Souris, now marked by Pelican, Rock, and Swan lakes, besides several other lakes of small size; the Minnesota River, with Browns Valley, [p.198] by which Lake Agassiz outflowed, where now are Lakes Traverse, Big Stone, and Lac qui Parle, the St. Croix River and Lake St. Croix, formerly the course of drainage from the west part of Lake Superior when that lake was held 500 feet higher than now by the barrier of the receding ice-sheet; and the Illinois River, the outlet of the glacial Lake Michigan, flowing through Lake Peoria.
        Eroded cliffs.--This type of shore-lines, denominated sea cliffs by Gilbert, is developed where a glacial lake has formed a terrace, usually in the unmodified glacial drift or till (see fig. 7, p. 26). Waves at these places have been efficient to erode, by undercutting at the base of the terrace; and shore currents have borne away the eroded material, excepting usually a considerable number of large bowlders. Only a small portion of the shores of Lake Agassiz examined by me consists of these wave-cut slopes of till; and they nowhere form conspicuous topographic features, their range in height being from 5 or 10 to 30 feet. Much higher cliffs of till of similar origin exist on some parts of the shores of the present great lakes of the St. Lawrence and Nelson rivers, where erosion has been in progress ever since the time of the glacial recession. Scarboro Heights, on Lake Ontario, near Toronto, extending 9 miles, with a height of 170 to 290 feet, consisting of till and interglacial beds, are cliffs thus produced by postglacial lake erosion. The duration of the glacial lakes appears to have been much shorter than the postglacial epoch.
        It is important, however, to note here that cliffs of preglacial erosion, which remained as prominent escarpments through the vicissitudes of the Ice age, became in some places the shores of glacial lakes. Of this class are the bold highlands of Pembina, Riding, and Duck mountains, which rise steeply 100 to 1,000 feet from the highest western shore-line of Lake Agassiz to form the margin of a plateau that stretches with a moderately undulating surface westward. Even where this lake washed the bases of the cliffs, it doubtless eroded them only to a slight extent. The horizontal Cretaceous beds of this great escarpment originally extended eastward a considerable distance, as believed by Hind and Dawson, probably so far as to cover the areas now occupied by Lake Winnipeg and the Lake of the Woods; and we must attribute the erosion of their eastern portion, leaving [p.199] this steep line of highlands, to river action during the time of epeirogenic elevation which closed the Tertiary and introduced the Quaternary era, not in any important degree to glaciation, and least of all to shore-cutting by the glacial lake.
        Beaches.--The course of the shore of a large glacial lake is usually marked by a smoothly rounded beach ridge of gravel and sand (see Pl. VI and fig. 6, p. 26) wherever the land is a slope of till sinking slowly beneath the ancient water-level. Like the shore accumulations of present lakes and of the seacoast, the glacial lake beaches vary considerably in size, having in any distance of 5 miles some portions 5 or 10 feet higher than others, due to the unequal power of waves and currents at these parts of the shore. Moderate slopes bordering the greater glacial lakes were favorable for the formation of beach ridges, and such ground frequently displays many beaches at successive levels which marked pauses in the gradual elevation of the land when it was relieved of its ice burden, and in the subsidence of the lake as its outlet became eroded deeper or as the glacial retreat uncovered new and lower avenues of discharge.
        Waves driven toward the shore by storms gathered the beach gravel and sand from the deposit of till or other drift which was the lake bed, and corresponding deposits of stratified clay, derived from the same erosion of the till, sank in the deeper part of the lake. But these sediments were evidently of small amount and are not commonly noticeable on the sheet of till which forms the greater part of the lacustrine areas. Where the beaches cross delta deposits, especially the fine silt and clay that lie in front of the delta gravel and sand, they are indistinctly developed or fail entirely. On the other hand, the most massive and typical beach ridges, often continuous several miles with remarkable uniformity of size, having a central thickness of 10 to 15 feet and a total width of 20 to 30 rods, are found on areas of till that rise with a gentle slope of 10 or 15 feet per mile. Under the influence of irregular contours of the shore, however, the beach deposits assume the form of bars, spits, hooks, loops, and terraces, of which Gilbert has given a careful classification, with analysis of the interactions of waves and currents by which they were made.(9)
        [p.200] Deltas.--A broad expanse of water exposed along a distance of many miles to strong winds is required for the formation of sufficiently large and powerful waves to erode cliffs or accumulate well-defined beach ridges; but the area of any glacial lake, small or large, may be partly occupied by deltas brought into its margin by tributary streams. These deposits at the mouths of small brooks are often only a few rods wide, while the deltas of rivers, especially those supplied with much englacial drift from the melting ice-sheet, sometimes extend many miles in a flat or moderately undulating plain of gravel and sand, lying at the level which the surface of the lake held during the accumulation of the delta, or within a few feet above or below that level. But at the mouth of the river forming the delta it was frequently built up in a fan-shaped mass to a considerable height, the head of the alluvial slope being in some instances 50 feet or more above the lake. The delta plain is generally bounded on its lakeward side by a somewhat steep descent, partly due to the ordinary conditions of delta formation, but often made more conspicuous by erosion of the outer portion of its original area by waves and shore currents when the lake fell to lower levels.
        Winds in many places have channeled and heaped the surface of the more extensive deltas, acting most efficiently as soon as they became uncovered from the lake and before they could be overspread by vegetation; and many of the resulting sand dunes (see Pl. VII, p. 28), which frequently range from 25 to 100 feet in height, though mainly covered by grass, bushes, and trees, are still undergoing slight changes of their form by wind erosion. All the dunes on the areas of the glacial lakes Agassiz, Dakota, Souris, and Saskatchewan, occur on delta deposits; but the great tracts of dunes about the south end of Lake Michigan belong wholly to beach accumulations, being sand derived from erosion of the eastern and western shores of the lake, whence it has been borne southward by shore currents, especially during northern gales. None of the beaches of our glacial lakes are large enough to make dunes like those on Lake Michigan, though the size and depth of Lake Agassiz, its great extent from south to north, and the character of its shores, seem equally favorable for their accumulation. It [p.201] is thus again indicated that the time occupied by the recession of the ice-sheet was comparatively brief.
        Lacustrine sediments.--In front of the delta plains of gravel and sand, the finer silt and clay brought into the glacial lake by the same tributaries were spread over the lake bottom, covering the till on large tracts adjacent to the great deltas. Only small contributions of fine sediment, usually inappreciable, as before stated, on the greater part of the lake basin, were supplied from the shore and sublittoral erosion of till, which yielded the gravel and sand of the beaches; but some of these areas of wave erosion, reaching a quarter of a mile off shore, are plentifully strewn with the residual bowlders.
        Because of their relation to the receding ice-sheet, the glacial lakes might be expected to receive noticeable deposits, including bowlders, from floating bergs and from floes of the ice foot which would be formed in winter along their northern barrier. It is certain, however, that no deposits which can be referred to such origin are spread generally over the lake basins. Bowlders are absent or exceedingly rare in the beaches, deltas, and finer lacustrine sediments. In a few places, however, I have observed bowlders in considerable numbers on esker ridges of gravel and sand (pp. 186, 188), where they were evidently brought and stranded by floating ice masses from the melting ice border, whose distance could not have exceeded a few miles at the farthest, and, indeed, probably was not so much as 1 mile while the bowlders were being stranded.
        Where terminal moraines cross a glacial lake, their knolly and hilly contour, as deposited on land, is changed to a smoothed, slightly undulating surface, and their proportion of bowlders exposed to view is diminished. The lake leveled the till that would otherwise have formed knobs and hills, in which process many of its bowlders were covered.
        After the drainage of the glacial lakes by the complete departure of the ice-sheet, the lower portions of their basins, in depressions and along the present river courses, have become filled to a considerable extent by fluvial beds of fine silt. These are similar in material with the lacustrine sediments bordering the deltas, from which they are distinguishable by their containing in some places shells like those now living in the shallow [p.202] lakes and streams of the region, remains of rushes and sedges and peaty deposits, and occasionally branches and logs of wood, such as are floated down by streams in their stages of flood. In the valley of the Red River of the North these recent fluvial deposits have commonly greater thickness and extent than the underlying silt of the glacial Lake Agassiz, which, however, in some portions, as near the deltas of the Sheyenne and the Assiniboine, occupies large areas.

        New England, Quebec, the eastern provinces, the Northeast Territory, and Labrador.--Attending the retreat of the ice-sheet from New England, Quebec, and the eastern provinces, many glacial lakes of small size and short duration were formed on areas declining toward the north or northwest, as in the valley of the Contoocook River, in New Hampshire;(10) on the western flanks of the Green Mountain range, in Vermont, where Mr. C. L. Whittle informs me that delta deposits of such origin occur up to heights of fully 2,000 feet; on head streams of the River St. John, in northern Maine; and in southern Quebec, between the Atlantic-St. Lawrence watershed and the receding ice front. Fewer and still smaller glacial lakes, usually leaving no well-marked records of their existence, doubtless also attended the glacial retreat in New Brunswick, Nova Scotia, Newfoundland, and Labrador. But soon the ocean-washed ice border was melted back from the Gulf of St. Lawrence and along the broad St. Lawrence Valley perhaps to Quebec, admitting the sea to the area of Lake Champlain, which, with the Hudson Valley, had been occupied during the recession of the ice by a long and narrow glacial lake, extending from near New York City to near Montreal, caused by the southward elevation and northward depression of the land.(11)
        North of the St. Lawrence the receding ice opposed no barrier to drainage from large areas until it withdrew across the height of land dividing the St. Lawrence waters from those tributary to James and Hudson [p.203] bays, when upon the country around Lake Mistassini and upon many other tracts glacial lakes of considerable size must have been formed. In the exploration of that region traces of these former lakes, especially of their channels crossing the watershed, should be carefully looked for, as not the least important of our records of the Ice age.
        Basins of the Laurentian lakes and of Hudson Bay.--As soon as the border of the retreating ice-sheet was withdrawn across the various parts of the watershed south of the Laurentian lakes, each considerable stream valley and embayment between the height of land and the ice front held a glacial lake. Doubtless hundreds of channels may be traced where these lakes outflowed. But the continuing glacial retreat merged these minor lakes into a few of large size overflowing at the lowest passes. In the States adjoining on the south, and in portions of Canada on the north, the shores of these glacial representatives of the present Laurentian lakes are recorded by eroded cliffs, beach ridges, deltas, and lacustrine sediments; but along other portions of their Canadian boundaries, where they were held in by the receding ice barrier on the northeast and north, the land shows no shore erosion nor beach deposits.
        The west part of Lake Superior stood about 500 feet higher than now, and outflowed by the St. Croix River. Lake Michigan outflowed by the low divide at Chicago to the Des Plaines and Illinois rivers. The glacial Lake Erie was at first some 200 feet above the present level of this lake, with overflow to the Wabash; but later it obtained lower outlets, the last being by Chicago, after the glacial lakes Erie, Huron, Michigan, and Superior had been merged into one expanse, which Spencer has named Lake Warren. Lake Ontario, or rather its glacial forerunner, named by Spencer Lake Iroquois, becoming by the retreat of the ice separated and distinct from the upper lakes, extended far to the north and northeast of its present limits and poured its waters into the Hudson, at first by the Mohawk and afterward by the way of Lake Champlain, while the continuing glacial recession uncovered the country north of the Adirondacks and along the great valley where it now outflows by the St. Lawrence.
        The watershed which divides the upper St. Lawrence basin from the basin of James Bay is crossed by many channels of outflow from glacial [p.204] lakes pent up between that watershed and the departing ice-sheet on the north. Kenogami or Long Lake, north of Lake Superior, having a length of about 54 miles from northeast to southwest and a width mostly between a half mile and 2 miles, forming the head of Kenogami River, tributary to the Albany, occupies the channel of outlet from a glacial lake in the Albany basin, passing southward by Trout Lake and Black River to Lake Superior.(12) The elevation of Kenogami Lake, according to the survey of the Canadian Pacific Railway, is 1,032 feet above the sea. Dr. Robert Bell states in a letter that the summit crossed by the Height of Land portage close south of this lake, and leading from it to Black River, is about 70 feet higher, being therefore approximately 1,100 feet above the sea. This portage "is about a half mile long, and is over an accumulation of well-rounded bowlders, with gravel and earth filling the interspaces in part; at other parts the bowlders are piled on each other quite naked. The valley between the rocky walls is about half a mile wide. The surface is somewhat level, and there is a subordinate valley or depression sweeping around on the west side between the bulk of the accumulation of bowlders and the rocky bluff on that side." The ancient watercourse thus described west of the portage is probably only a few feet above Kenogami Lake, having very nearly the same elevation as the divide between the Missinaibi and Michipicoten rivers, some 150 miles distant to the east. Both these low points of the watershed were doubtless occupied by rivers outflowing from glacial lakes on the north during the recession of the ice-sheet.
        Missinaibi Lake, near the head of Missinaibi River, the western branch of the Moose River system, is about 1,020 feet above the sea. This lake "bears south 48° west, is 24 miles long, nearly straight, and varies from a half mile to 11 miles in width."(13) Close southwest of Missinaibi Lake, in the continuation of this glacial river course, is Crooked Lake, at an elevation of about 1,038 feet. "It is 8½ miles long, and averages less than a quarter of a mile in width." Near the head of Crooked Lake, and only a few feet above it, is the Height of Land portage, approximately 1,042 [p.205] feet above the sea, and thence descending toward Lake Superior the old channel contains Dog Lake, having a height of about 1,026 feet, and Mattagaming or Mattawagaming Lake, which, according to the Canadian Pacific Railway survey, is 1,025 feet above sea-level.
        When the Kenogami, Missinaibi, and other glacial lakes of the James Bay region became merged in one of great extent, rivaling Lake Agassiz, the outlet of this confluent lake probably crossed the low watershed south of the eastern end of Lake Abittibi, passing to Lac des Quinze and the Ottawa River. The elevation of Lake Abittibi, according to observations of the Canadian Geological Survey, is about 857 feet above the sea, and the portage over the watershed rises only about 100 feet higher. Its present altitude is thus nearly a hundred feet less than that of the Kenogami and Missinaibi outlets, and it is probable that when the land was first uncovered from the ice-sheet the Abittibi outlet was relatively lower than the others by a much greater difference, and that with reference to the sea-level it was much less elevated than now.
        Basins of the Saskatchewan and the Red River of the North.--During the recession of the ice-sheet from Alberta small glacial lakes doubtless existed in the basins of the Bow and Belly rivers, outflowing from the former successively by the Little Bow River and the Snake Valley, and from the latter successively by the Verdigris, Etsi-kom, and Chin coulées, which Dr. Dawson describes as remarkable abandoned river courses now carrying little or no water. The glacial drainage from the present sources of the South Saskatchewan, and probably also of the North Saskatchewan and Athabasca, was thus carried southeastward, in parallelism both with the main Rocky Mountain range and with the retiring ice-border, to the Milk River, west and south of the Cypress Hills. The whole area of Alberta, partly land sloping northeastward and partly ice sloping southwestward, with glacial lakes here and there along the ice margin, seems then to have been tributary to the Missouri and the Gulf of Mexico.(14)
        From Lake Pakowki, through which this glacial drainage for a long time flowed southward to the Milk River, the ice front must have been [p.206] withdrawn more than 200 miles to the east, past the Cypress Hills and Wood Mountain, before a lower outlet from the Saskatchewan country north of these highlands would be obtained by Twelve Mile Lake and over the present continental watershed to Big Muddy Creek, which flows through the northeastern corner of Montana to the Missouri. But only a slight further retreat of the ice was sufficient to give still lower avenues of drainage. As soon as the Missouri Coteau was uncovered a glacial lake occupying the valley of the South Saskatchewan, in the vicinity of its elbow, outflowed by the way of Moose Jaw Creek, and through a glacial lake in the upper Souris or Mouse River basin, to the Missouri near Fort Stevenson. Later the outflow from the Lake Saskatchewan may have passed to the Lake Souris by way of the Wascana River, after flowing through a glacial lake which probably extended from Regina 60 miles to the westward in the upper Qu'Appelle basin.
        Through the whole period of the existence of the Lake Souris, which at first outflowed to the Missouri and afterward to Lake Agassiz, the glacial lake in the basin of the South Saskatchewan, doubtless also at last including the North Saskatchewan, was tributary to it, and the outlet of this Lake Saskatchewan was transferred to lower courses as the border of the ice-sheet receded from southwest to northeast. In the concluding part of this chapter detailed descriptions of these glacial lakes, and of their successive channels of outflow to Lake Agassiz, will be presented.
        Lake Agassiz, the largest of all the glacial lakes of North America, occupying the basin of the Red River of the North and Lake Winnipeg, covered extensive tracts of Minnesota and North Dakota, the greater part of Manitoba, and a considerable area of eastern Saskatchewan and southwestern Keewatin. The history of this lake, which increased in area as the ice-sheet decreased, forms the central theme of this chapter, succeeded by reviews of the associated glacial lakes of large size, two of which, lying in southern Minnesota and in South Dakota, had their brief existence before Lake Agassiz, the others being contemporaneous with this lake and several of them tributary to it.
        British Columbia, Athabasca, and the Northwest Territory.--Light-colored silt deposits, distinctly stratified and of considerable thickness, which seem [p.207] to me referable in some districts to glacial lakes and in others to river floods supplied from the melting ice-sheet, are reported by Dr. G. M. Dawson in many basins of the British Cordilleran region. His interpretation of their origin, however, is by a marine submergence since the latest glaciation of the region. No fossils, either of the sea or of fresh water, are found, though they are abundant in postglacial marine beds of the St. Lawrence Valley, on the southwestern side of Hudson Bay, and in Greenland and Grinnell Land; but lakes of only moderate size temporarily bordering the ice-sheet during its departure would probably be destitute of life, and this would certainly be true of rivers produced by the glacial melting. These deposits occur, up to heights 2,300 to 2,700 feet above the sea, in the basin of the Kootanie and upper Columbia, on the interior plateau of British Columbia, on the northward extension of the great plains crossed by the Peace River, and in the upper valleys of the Stikine, Liard, and Yukon rivers.(15)
        On the last-named river and the Lewes, tributary to it, Russell refers the formation of silt beds, fully 200 feet thick, and of higher terraces, to a glacial lake, named by him Lake Yukon, 150 miles long from north to south, with a maximum width of about 10 miles and depth of between 500 and 600 feet; and he suggests that this lake was probably caused by a depression of the upper part of the Yukon basin by the weight of the ice-sheet. The mouth of Lake Yukon, at its northern end, was near the northwestern boundary of the ice-sheet at its maximum extension, the whole lake being within the area that was ice-covered, as is known by the limits of glacial drift and striæ, which are first found in ascending the Yukon near the Rink Rapids, approximately in latitude 62° 20' north and longitude 136° 15' west, about 160 miles east of the line between British America and Alaska.(16)
        No other portion of the glaciated area of this continent presents a more interesting or more difficult problem in Pleistocene geology than these "white silts," as they are denominated by Dawson; and much further field work and study will be needed to demonstrate the conditions of [p.208] their deposition in each of the numerous basins in which they are found. But I believe that ultimately they will be shown to be everywhere attributable either to fluvial deposition attendant on the recession of the ice-sheet or to deposition as deltas in glacial lakes which owed their existence to ice dams or to depressions where the land had sunk beneath the ice weight and has since been reelevated. For example, the Kootanie basin may well have been filled by a glacial lake obstructed in the present course of drainage by the retreating ice-sheet and outflowing by the way of Pack River and Lake Pend d'Oreille, which Professor Chamberlin finds to have been covered by the maximum advance of the ice, while gravel-bearing floods from the glacial melting poured thence to the south and west.(17) Again, the silts on the Peace River east of the Rocky Mountains seem referable, as will be stated more fully on a later page, to a glacial lake held by the barrier of the departing ice-sheet on the north and northeast, with outflow southeastward into Lake Agassiz.

        On the west side of Lake Agassiz the Dakota lobe of the ice-sheet, from its junction with the Minnesota lobe near the Head of the Coteau des Prairies, 25 miles west of Lake Traverse and Browns Valley, at the beginning of the moraine-forming or Wisconsin division of the Glacial period, reached about 200 miles south along the valley of the James or Dakota River to Yankton and the Missouri; but it was gradually diminished in its extent until, at the time of formation of the Kiester, Elysian, Waconia, and Dovre moraines, it no longer retained its lobate outline. While these moraines were being formed in Minnesota the southwestern boundary of the ice-sheet in South and North Dakota passed from the vicinity of Big Stone Lake and Lake Traverse northwesterly along morainic belts which have been traced through Sargent, Ransom, Barnes, and Griggs counties, N. Dak., and by the sources of the James and Sheyenne rivers. During the later stages, represented by the Fergus Falls and Leaf Hills moraines, the Dakota ice front appears to have become [p.209] again lobate, extending from the west shore of Lake Agassiz southward and then westward and northward, between the lake area and the Sheyenne River, to the prominent and typical moraines that are found south of Stump and Devils lakes, on the Big Butte, about Broken Bone Lake and northward, and on Turtle Mountain. In their remarkable development these moraines are similar to the massive Leaf Hills, with which they seem to have been contemporaneous. The laving action of Lake Agassiz caused the thick portion of the ice-sheet filling the Red River Valley to melt back somewhat faster than its thinner portions on the higher land areas at each side.
        The highest of the Herman beaches of Lake Agassiz extends in Minnesota, as traced in this survey, to the north side of Maple Lake, 20 miles east-southeast of Crookston, and probably it continues thence into the forest region on the east, where it is impracticable to follow its course, to the vicinity of Red Lake; and on the west it reaches through North Dakota and at least 14 miles into Manitoba, terminating on the northern part of the Pembina escarpment somewhere between Thornhill and its northern end, that is, between 14 and 40 miles north of the international boundary. Before the formation of this beach was completed the ice-sheet had retired from the lake area as far north as the beach extends. During pauses of this glacial recession the Dovre, Fergus Falls, Leaf Hills, and Itasca moraines were formed, showing a northward retreat of the ice border from the Dovre moraine across a distance of about 150 miles in central Minnesota and 150 to 200 miles in North Dakota and southern Manitoba, with a maximum of probably not less than 300 miles in the Red River Valley, where Lake Agassiz produced a more rapid melting of the ice margin. Through this time the River Warren, outflowing from this glacial lake, fed by abundant ice-melting and rains, eroded a channel about 50 feet deep, approximately from 1,100 to 1,050 feet above the sea, or perhaps it eroded only the lower half of that depth, in the moderately undulating sheet of till which reached across the present valley of Lakes Traverse and Big Stone. The shortness of the time thus indicated as probably occupied in the formation of a single one of the beaches of Lake Agassiz, reaffirmed as it is by the small amount of the littoral erosion and resulting beach deposits, [p.210] may well astonish us in what it implies concerning the rapidity of the recession of the ice-sheet, and the brevity, geologically speaking, of the stages of pause or readvance when its moraines were accumulated.
        Between the times of accumulation of the successive terminal moraines, the ablation of the ice surface and the retreat of its border caused the portion of the drift which had been inclosed within the ice-sheet to be rapidly deposited on the land, partly as till and partly as stratified gravel, sand, and clay, brought by the streams that were produced by the glacial melting. Thus while the series of Herman beaches was being formed not only were several large moraines amassed, but also much englacial till was spread over the country between the moraines, and glacial rivers deposited a broad belt of modified drift that stretches from central Minnesota to Red Lake and the Lake of the Woods, and continues northward in Manitoba, as described in pages 181-183. The most southeastern part of this prolonged tract of plentiful modified drift, in the vicinity of St. Paul and Minneapolis and northwestward to St. Cloud, belongs to a time previous to Lake Agassiz; the portion of these stratified beds between St. Cloud and Lake Itasca represents the time of formation of the highest Herman beach; and the deposition of their northern half, continuing from the headwaters of the Mississippi to the southwest part of the Lake of the Woods and to the Birds Hill group of eskers, was contemporaneous with the lower Herman shores of Lake Agassiz. Toward this belt great areas of the ice-sheet sloped convergingly during its maximum extension, and in the early part of its time of recession rivers flowed thither from the ice-lobes on the northeast and northwest until this glacial lake began to exist and to grow northward, occupying the Red River Valley.

        The retreat of the ice between the Waconia and Dovre moraines (pp. 142, 147) began to uncover the southern end of the bed of Lake Agassiz, into which the inflowing glacial Sheyenne River, even at that early stage, brought much gravel and sand. This first delta deposit of the glacial lake is spread along its southwestern margin from near Taylor Lake to the bluff, in the northeast corner of South Dakota, that overlooks the valley of [p.211] Lake Traverse and the Bois des Sioux River, about 4 miles southwest of White Rock, lying 100 feet below this gravel and sand bluff. The same high tract was at that time continuous also southeastward across the present valley, which is 4 miles wide, to the plateau in Traverse County, Minn., between the Bois des Sioux and Mustinka rivers, which is crossed and cut into by the railway in section 26, township 128, range 47. A thickness of 12 feet of this delta of gravel and sand, having a surface 75 feet above Lake Traverse, is shown by the railway excavation, without exposing its plane of contact with the underlying till, which forms the basal part of the plateau and extended, before its erosion by the outflow from Lake Agassiz, in an inclined plane gradually rising to the bluff of till, 100 to 110 feet high, east of the northern end of Lake Traverse. In this incipient stage, contemporaneous with the accumulation of the Dovre moraine, Lake Agassiz stretched nearly 30 miles from northwest to southeast, with a width varying from 1 to 2 or 3 miles, being probably widest in the vicinity of Wheaton, Minn., at its southern end, where the River Warren flowed away southwestward. The lake in this stage was little more than a broad expansion of the glacial representative of the Sheyenne River, which deposited its delta sediments along the edge of the lacustrine area, being walled in by the front of the ice-sheet.
        With the glacial recession thence to the Fergus Falls moraine (p. 158) Lake Agassiz attained a length of about 120 miles from Lake Traverse north to Ada, Caledonia, and Hillsboro, with a width of 40 to 50 miles, occupying thus an area of about 5,000 square miles (Pl. XIX). Its depth at Breckenridge and Wahpeton was approximately 100 feet; at Moorhead and Fargo, 200 feet; and at Caledonia, 275 feet.
        In the earliest part of this extension of the lake its outlet by the River Warren seems to have been for a short time about 25 feet higher than during the later and much longer part of this stage of recession of the ice and growth of the lake, as is shown by the Milnor beach, a less distinct shore deposit than the Herman beach and 20 to 25 feet above it, which was observed near Milnor, N. Dak., and along a distance of about 10 miles thence northwest to the Sheyenne, but was not recognized farther north nor in Minnesota. The Sheyenne at the time of formation of the Milnor beach [p.212] had become established in the course which it now has to its debouchure into Lake Agassiz at the present most southern bend of the river. Its large delta there brought into the lake was already in progress of deposition during the accumulation of the Milnor beach-ridges and partly supplied the gravel and sand of which they are formed.
        But the River Warren quickly cut down its channel to the base of the earlier Sheyenne deposit of gravel and sand before described, lying above the present valley of the Bois des Sioux, until it reached the harder till, and there was stayed during the numerous stages of lacustrine extension and glacial retreat which are represented by the single Herman beach of this southern portion of the lake. The growth of the great Sheyenne delta continued, and the Buffalo delta was probably mostly completed, during the withdrawal of the ice-sheet to the Fergus Falls moraine and its pause or readvance by which that moraine was made. Through the same stage, excepting its very short early portion, represented by the Milnor beach, and for a long time afterward, Lake Agassiz held its Herman level, changing only very slightly in this southern area by slow erosion of the outlet, but experiencing northward a gradual uplifting of its basin, whereby its Herman beach, single at the south, becomes double and multiple in proceeding to the north.
        The next stage in the departure of the ice withdrew portions of its border to the ninth or Leaf Hills moraine (p. 163), which is closely associated with the Fergus Falls moraine, the two being merged together through much of their course. Lake Agassiz, therefore, gained only a small extension of its length and area (Pl. XIX). The most notable change was the formation of a northwestern bay of the lake, reaching in a reentrant angle of the ice-sheet to Larimore and McCanna, which received the Elk Valley delta, deposited by a large glacial river flowing from the depression on the ice surface where the descending slopes of its Minnesota and Dakota lobes met.
        After these contiguous and partly combined moraines were formed, the increasing warmth of the climate again pushed back the ice border a long distance, until its retreat was temporarily interrupted at the line marked by the tenth or Itasca moraine (p. 173). Advancing northward, [p.213] Lake Agassiz then expanded beyond the limit of the international boundary, reaching probably to Winnipeg and Birds Hill (Pl. XX). The entire area of this lake in North Dakota had become uncovered from the ice, a lobe of which, however, remaining on the Pembina Mountain plateau, closely bordered the shore along a distance of 50 miles south from the Manitoba line. In northwestern Minnesota the lake washed the base of ice cliffs that formed its eastern shore, beginning about 40 miles north of Lake Itasca and running north-northwesterly, as I have supposed, to an angle of the ice front at Birds Hill, from which a similar long, high coast of ice appears to have stretched southwestward to the Pembina Mountain in the vicinity of Thornhill, being the northwestern barrier of the widening and deepening lake. The water surface was about 290 miles in length, 110 miles in maximum width, and approximately 16,000 square miles in area; and the depth of water above St. Vincent, Pembina, and Emerson was about 450 feet, while its maximum above the site of the city of Winnipeg was not less than 550 feet. The extent of the portion of the lake in Manitoba at this time was probably about 3,500 square miles.
        Once more the margin of the ice-sheet recedes, and next halts at the eleventh or Mesabi moraine (p. 177), having relinquished the whole of its area in North Dakota, but still lingering on a large tract of northern Minnesota, from Red Lake and Lake Winnebagoshish eastward to Lake Superior near the international boundary. The great glacial lake has now extended north to the south end of Lakes Winnipeg and Manitoba, attaining a length of about 325 miles, a maximum width of 130 miles from the east end of the south half of Red Lake to Larimore, and an area not far from 26,000 square miles, of which fully one-third was comprised in Manitoba (Pl. XX). Its maximum depth, lying over the present mouth of the Red River, was about 650 feet, and its depth above the south end of Lake Manitoba was 525 feet, very nearly.
        These estimates of depths, it is to be noted, are derived from the determinations of the height of the shore-lines formed during the highest Herman stage, with allowance for the known north-northeastward differential elevation of the basin since the old plane of the lake surface was marked by the waves of storms. This earliest and highest level of Lake [p.214] Agassiz (excepting only the unimportant stage recorded by the Milnor beach) extended north along the Pembina Mountain into Manitoba and northeast to the south side of Red Lake, being contemporaneous with the accumulation of the Fergus Falls, Leaf Hills, Itasca, and Mesabi moraines, so that the single lacustrine plane of the uppermost in the series of the Herman beaches covered, at its final stage of greatest extent, all of the lake area to the latest of these moraines, which is the most northern one that has been definitely traced and mapped across this area.
        Yet again, and doubtless many times again, the ice-sheet was compelled to retreat across spaces of varying widths, sturdily resisting the encroachments of the warmer climate and of its product, the glacially dammed lake, pausing here and there long enough to heap up moraines, then shrinking and dissolving away from new tracts strewn with its drift deposits. When future researches shall enroll the numbers and delineate the courses of the probably many morainic belts lying still farther north, it will be possible to show the later stages of the gradual extension of this lake along the great Cretaceous escarpment and over the great lakes of Manitoba, across Rainy Lake, the Lake of the Woods, and the Winnipeg River, over a large region east of Lake Winnipeg, and to some now unknown distance down the Nelson River.
        Step by step, as fast as the ice-sheet waned, Lake Agassiz grew. The whole lacustrine area, as mapped provisionally for its northern and northeastern boundaries on Pl. III, was about 110,000 square miles or more, considerably exceeding the combined areas of the great Laurentian lakes. Although it was not entirely occupied by Lake Agassiz at any one stage of its existence, the beaches and terminal moraines indicate that the lake, during both its earlier and later stages, covered the greater part, probably three-fourths, of this area.
        The chief evidence of such great extension of the lake during the first half of its history is the observed extent of the higher and earlier Herman and Norcross beaches, which have been mapped from near Red Lake, Minnesota, southward to Lake Traverse, and thence northward through North Dakota to Riding and Duck mountains in Manitoba, a distance of about [p.215] 700 miles. Delta sand deposits, brought into Lake Agassiz by the Saskatchewan and referable to the Herman, Norcross, and later stages, reach from near Prince Albert, on the North Saskatchewan, about 40 miles west of the fork of the North and South branches, through a distance of more than 100 miles eastward to the head of the Seepanock Channel and the one hundred and third meridian.(18) The descent of the river in this distance is approximately from 1,275 or 1,300 feet to 950 feet, and the elevation of the west part of the delta is about 1,350 to 1,400 feet above the sea. As early as the time of the lower beaches of the Herman series, therefore, the recession of the ice-sheet had permitted the lake to extend along the whole front of the Manitoba escarpment to the latitude of the north end of Lake Winnipeg. The length of Lake Agassiz at that time was 550 miles or more, and I believe that its average width was not less than 150 miles, reaching east to the moraine which Mr. J. B. Tyrrell describes as forming the eastern shores and islands of Lake Winnipeg, with a height of 100 feet on Black Island.(19) This moraine would then have been deposited in water 600 to 700 feet deep, bordering the ice margin; its knolly and irregular accumulations of drift would not have been subjected to the leveling action of the lake waves until the further melting of the ice opened avenues of outflow to Hudson Bay and reduced the glacial lake nearly to the level of Lake Winnipeg; and the latest change of the northward outlets may have lowered the water surface so rapidly and to such vertical amount that it left no distinct marks of erosion or shore-lines on the upper portion of the moraine.
        Before the successive northeastward outlets began to drain Lake Agassiz below its channel of southward discharge at Lakes Traverse and Big Stone, the border of the ice-sheet had been gradually melted back from Lake Winnipeg doubtless far toward Hudson Bay, and perhaps even its thick central part, which occupied the basin of Hudson and James bays, had so far disappeared as to admit the sea there. At a time of halt or readvance, interrupting this recession, another terminal moraine appears to have been accumulated, crossing the Churchill and Nelson rivers, as observed by [p.216] Dr. Robert Bell.(20) If this belonged to the time of the Campbell or McCauleyville beaches, as seems most probable, the extent of the lake during these later stages of southward outflow was even greater than I have supposed it to be at the time of the Herman and Norcross beaches, and the area occupied by Lake Agassiz in its numerous stages much exceeded that of my map and estimate.
        Measured on the maps of this report, the portion of Lake Agassiz comprised within the limits of Minnesota has an area of approximately 15,000 square miles, and its portion in North Dakota is 6,800 square miles, very nearly, making together a tract of about 21,800 square miles in the United States, probably all uncovered from the ice and occupied by the lake during the time of formation of the Herman series of beaches. Within the limits of Manitoba and adjacent parts of Saskatchewan and Keewatin the extent attained by Lake Agassiz during its Herman and Norcross stages was probably at least 65,000 square miles. Somewhat more than three quarters of its expanse then was north of the international boundary, for while the lake expanded northward with the recession of the ice-sheet, the southern part of the basin was being uplifted and the lake was slowly cutting down its outlet, so that it had already relinquished the margins of its earliest area in Minnesota and North Dakota.
        During the stages of the lake represented by the Tintah, Campbell, and McCauleyville beaches, probably its area occupied by water at one time grew to exceed 100,000 square miles. Its southern portion, however, was meanwhile diminishing, until at that late time of maximum size of Lake Agassiz not more than a tenth or perhaps a fifteenth part of its water surface was in the United States. The decrease was in width, not in length, for at its maximum stage the outflow was doubtless still to the south by the River Warren.

By the melting away of the ice-sheet from the country northeast of Lake Agassiz this glacial lake at length obtained successive outlets lower than that through Lakes Traverse and Big Stone and the Minnesota River. Owing to the northeastward depression of the ice-laden area, the earliest of [p.217] these outlets may have flowed to the east and south, passing along the margin of the receding ice into Lake Superior, and thence into the Mississippi by the way of the Chicago outlet of the glacial Lake Warren, as Prof. J. W. Spencer has named the confluent glacial lake which is now reduced and separated into parts as the five great lakes of the St. Lawrence. After the glacial melting had proceeded so far as to open the great area of Hudson and James bays to the entrance of the ocean, Lake Agassiz was tributary for some time to this inland sea by outlets higher than the Nelson River, while the ice-sheet west of Hudson Bay was withdrawing northward.
        Some of the lowest and latest stages of Lake Agassiz during its decrease as it was drained away by its northeastern outlets, each in succession lower than the preceding, are shown by Mr. J. B. Tyrrell's observations of beaches on Kettle Hill close south of Swan Lake, on the portage between Lake Winnipegosis and Cedar Lake, and in the vicinity of the Grand Rapids of the Saskatchewan.(21) Between the time of formation of the Stonewall beach and that of the Niverville beach the surface of Lake Agassiz was lowered 45 or 50 feet, from a level slightly higher than Lake Winnipegosis to one slightly lower than Lake Manitoba. The former of these levels seems to be represented near the mouth of the Saskatchewan by a beach 140 feet above Lake Winnipeg, or 850 feet above the sea; and the latter becomes apparently double or triple, being represented by three beach ridges, 95, 90, and 80 feet above Lake Winnipeg. These beaches, if my correlation as thus stated is correct, are nearly horizontal throughout their observed extent of nearly 300 miles from south to north, and show that the differential northward uplift of the basin of Lake Agassiz was almost completed before the ice barrier was melted back from the area crossed by the Nelson River.
        According to my correlation of the five shore-lines noted by Mr. Tyrrell on Kettle Hill, successively, in descending order, 1,070, 1,015, 995, 955, and 920 feet above the sea, the highest belongs to the Hillsboro stage of Lake Agassiz; the next two to the Emerado stage, there divided because of northward uplifting of the land; while the lower two are, respectively, [p.218] the second of the Ojata beaches and the Gladstone beach. This locality is about 235 miles north of the international boundary, being 150 miles north from the latitude of Gladstone, Arden, and Neepawa, the most northern line upon which my own explorations supply a comparison of the beaches and determination of their northward ascent.
        At the time of formation of the Hillsboro beach, which had been already preceded by the three higher Blanchard levels of Lake Agassiz since it first began to outflow northeastward, the lake surface thus appears to have been about 140 and 240 feet, respectively, above the southern portions of the present Lakes Manitoba and Winnipeg, and approximately 240 and 360 feet above the northern portions of Lakes Winnipegosis and Winnipeg, the northward ascent of the Hillsboro beach being nearly 120 feet in the 150 miles between Gladstone and Kettle Hill. Lake Agassiz during this stage stretched south in the Red River Valley about 15 miles beyond Fargo and Moorhead; and its total length was probably not less than 650 miles, with a maximum width of about 200 miles.
        During the formation of the two Emerado beaches the lake on the latitude of Kettle Hill was about 185 and 165 feet, respectively, above the northern part of Lake Winnipegosis, to which 118 feet should be added for its depth above Lake Winnipeg, besides some undetermined amount of present northeastward ascent of the plane of that lake surface in the distance of more than a hundred miles to the north end of Lake Winnipeg. The Emerado level of Lake Agassiz began at the south about 5 miles north of Moorhead and Fargo, and stretched probably 600 miles to the north. Its width was little less than that of the Hillsboro stage; but the northward uplifting of the lower Emerado beach between Gladstone and Kettle Hill has been only 85 feet.
        When the lake held its two Ojata stages and Gladstone stage, the depth of water above Lake Winnipegosis was, successively, about 140, 125, and 90 feet; and its extension southward in the Red River Valley was for the lower Ojata beach to Caledonia, near the mouth of the Goose River, and for the Gladstone beach to the vicinity of Belmont, N. Dak., about 20 miles south of Grand Forks. The portion of Lake Agassiz extending into the United States at the Gladstone stage had a length of almost 100 miles; [p.219] and the total length of the glacial lake, then near the middle of its entire time of northeastward outflow, was more than 500 miles, with probably one-third as great width in its northern part. The amount of upward tilting toward the north upon the area extending 150 miles from Gladstone to Kettle Hill since the Gladstone beach was formed has been 40 or 45 feet. About twice as much tilting had occurred there between the times of formation of the Hillsboro and Gladstone beaches as since the date of the later one of these shore-lines. Lake Agassiz in its Gladstone stage had become reduced probably to half of its earlier maximum extent.
        Mossy portage, between Lake Winnipegosis and Cedar Lake, is about 60 miles northeast of Kettle Hill; and the Grand Rapids of the Saskatchewan, near its mouth, are about 25 miles farther east. Both these localities are nearly on latitude 53° 10' north, being some 50 miles north of the latitude of Kettle Hill and 285 miles north of the international boundary. The summit of the eastern Mossy portage is described by Mr. Tyrrell as a gravel ridge with crest 93 feet above Lake Winnipegosis or 921 feet above the sea. It is doubtless a beach formed by Lake Agassiz when it stood here at the level of about 910 feet. Descending southward to Lake Winnipegosis, the portage crosses another beach ridge with its crest 27 feet and its base about 15 feet above this lake, and it is therefore clearly referable to a level of Lake Agassiz about 845 feet above the sea. These stages of the glacial lake are quite surely the same which made the Burnside and Stonewall beaches near the south end of Lake Manitoba and the city of Winnipeg. An escarpment crossed by the portage midway between these beach ridges appears to mark the position of the intermediate Ossowa shore.
        The Burnside lake level reached south in the Red River Valley to Grand Forks, and had an entire length of nearly 500 miles thence to the latitude of 55° north, with a width from 150 to 175 miles in its northern half. Above the southern end of Lake Winnipeg the depth of Lake Agassiz at this stage was 150 feet, and above its northern end about 200 feet.
        The next lower level of Lake Agassiz, which is recorded by the Ossowa shore-line, lacked only 15 miles of reaching to Grand Forks, and had almost as great total length and width as the preceding. Its height above the south ends of Lakes Manitoba and Winnipeg was about 30 and 130 feet, respectively.
        [p.220] At the time represented by the Stonewall beach, lying next in descending order, the surface of Lake Agassiz was 10 to 20 or 25 feet above Lake Manitoba, 5 to 15 or 20 feet above Lake Winnipegosis, and about 110 and 140 feet, respectively, above the southern and northern ends of Lake Winnipeg. It yet extended nearly 40 miles south of the international boundary, to the vicinity of the mouth of Park River.
        In receding from the Stonewall to the Niverville stage Lake Agassiz sank below Lakes Winnipegosis and Manitoba, which remain as two of the three large remnants of this vast body of water. On the line of the tramway at the Grand Rapids of the Saskatchewan Mr. Tyrrell reports four beach ridges of gravel and sand, as already noted, at the heights of 850, 805, 800, and 790 feet above the sea. The first is referable to the Stonewall stage, and the three others to the Niverville stage, which is here compound, apparently on account of intermittent northward uplifting of the country. Mr. Tyrrell informs me that the Niverville beach on Black Island, in the southern part of Lake Winnipeg, is about 60 feet above the lake. At the Grand Rapids, 175 miles northwest from Black Island, its three ridges, in descending order, are 95, 90, and 80 feet above the lake, showing that there was a northward uplift of 15 feet along this distance during the Niverville stage, and that since then a further differential tilting of about 20 feet has taken place. The southern end of the Niverville level of Lake Agassiz was near Morris, Manitoba. It failed to reach into the United States by a distance of about 25 miles, being the first stage of this glacial lake that lay wholly in British America, and it was the latest stage held by the ice barrier and recorded by a well-marked shore-line. Lake Agassiz at this time, as during several preceding stages, reached far north and northeast of Lake Winnipeg, and up to its latest year it may have had an area of 20,000 or 30,000 square miles.
        Finally the retreat of the ice-sheet uncovered the land across which the Nelson outflows from Lake Winnipeg to Hudson Bay. The existence of the glacial lake was ended, and this largest of the great lakes of Manitoba was added to the number of its present representatives or descendants. Dr. Bell's descriptions of the outlet of Lake Winnipeg and the [p.221] topography of the adjoining country(22) show that no barrier of land so high as the Niverville beach can have been removed there by erosion. The original level of Lake Winnipeg, due to the height of the land upon which the Nelson River began to cut its channel in its present course, is doubtless that of the well-defined beach observed by Hind between the mouths of the Winnipeg and Red rivers, having "an elevation of 21 feet above the present level of Lake Winnipeg."(23) Traces of this shore-line will probably be found at nearly the same height around the whole lake.

        In the southern part of the area of this glacial lake, within 75 miles northward from its mouth at Lake Traverse, five principal beaches have been observed, and in their descending order have been named, from towns in Minnesota near which they are well exhibited, the Herman, Norcross, Tintah, Campbell, and McCauleyville beaches. These shore-lines, however, when traced farther north, are found to become double or multiple. The Herman beach in the vicinity of Maple Lake, Minnesota, is divided into five beaches, four besides the highest having been formed when the rise of the land, with the slight fall in the level of Lake Agassiz, amounted, successively, to 8, 15, 30, and 45 feet on the east side of the lake in that latitude. Still farther to the north, in Manitoba, we find seven beaches corresponding to the single Herman beach at the southern outlet. In like manner, the Norcross and Tintah beaches are each represented at the north by two, and the Campbell and McCauleyville beaches each by three distinct shore-lines, separated by slight vertical intervals. The northern part of the lake has thus no less than seventeen shore-lines, which were successively formed from the highest to the lowest during the time of the southward outflow through Lakes Traverse and Big Stone and the Minnesota River to the Mississippi.
        After the lake obtained its earliest outlet to the northeast, sinking below Lake Traverse, it formed fourteen shore-lines. The first three of these [p.222] pass near Blanchard, N. Dak., and thence are denominated the Blanchard beaches. The next in descending order is the Hillsboro beach, the succeeding two are the Emerado beaches, and the two next lower the Ojata beaches, named similarly from other towns of this State. The remaining six lower beaches are named from localities in Manitoba. In the same descending order, they comprise the Gladstone, Burnside, Ossowa, Stonewall, and Niverville beaches, the last being double. There are thus in total thirty-one separate shore-lines of this lake in the northern portion of its area explored by me; and all of them, excepting the lowest, extend south of the international boundary.

        Through the greater part of the duration of Lake Agassiz its outlet remained constantly in one position, and the stream of its overflow, named the River Warren, eroded during that time the remarkable valley, rather to be described as a trough-like channel, mostly 1 to 2 miles wide and 150 to 230 feet deep, which is now occupied by Lakes Traverse and Big Stone and the Minnesota River. There is evidence, however, in the terraces of modified drift along the Minnesota Valley, that in large part its erosion was effected in preglacial time and during stages of retreat and readvance of the ice-sheet previous to its final departure, when it was the barrier of this glacial lake.(24) The general surface of the moderately undulating drift sheet, having swells 10 to 25 feet above its hollows, which stretches away on each side from the top of the bluffs inclosing this valley at Lakes Traverse and Big Stone, is about 1,100 feet above the sea, and the heights of these lakes at their low stage of water are, respectively, 970 and 962 feet. Before the retreat of the ice uncovered this tract, a channel 40 or 50 feet deep probably existed here, nearly or quite continuous, along the course that was taken by the River Warren in its first discharge from the incipient Lake Agassiz; for this level, much below the even expanse [p.223] of drift through which the river flowed, is the height of the Herman beach, which was the shore of the glacial lake at an early stage and through a long time ensuing. The somewhat higher Milnor beach appears to have been due to the temporary barrier interposed at first by the delta gravel and sand of the glacial Sheyenne River, spread wholly across the southern end of the lake at its beginning. Over this barrier, to the west of the line between White Rock and Wheaton, the River Warren flowed for a short time with rapids, speedily cutting it down 20 or 25 feet to the bed of the previously existing channel along the distance of 50 miles above the present sites of Lakes Traverse and Big Stone. This channel, whose depth determined the level of the Herman beach while the lake expanded with the recession of the ice-sheet even to southern Manitoba, was, as I believe, a vestige of a preglacial and possibly interglacial river course not wholly filled by the drift deposits. Reasons for this belief are sufficiently stated in the memoir on the Minnesota Valley before referred to, and in the description of certain remarkable chains of lakes in Martin County, Minn.(25)
        Nearly all the changes in the relative heights of Lake Agassiz and the basin that held it, by which the Herman beach became fourfold and even sevenfold in proceeding northward, must be ascribed to epeirogenic uplifting of the land, with only a very small element of change in the lowering of the lake level by erosion of its outlet. The southern portion of this shoreline, as far to the north as the latitude of Moorhead and Fargo, is marked by a single beach ridge, very definite in form and course, but not massive in comparison with the present beaches of the ocean or of the great lakes tributary to the St. Lawrence, the Nelson, and the Mackenzie. While Lake Agassiz was forming the Herman beach, erosion probably lowered the channel of the River Warren and the level of the lake 5 or 10 feet. During the same time a much greater differential northward uplift, presently to be considered, was in progress.
        From the level of the Herman beach to that of the Norcross beach Lake Agassiz fell somewhat suddenly 15 or 20 feet. As this change of level affected the southern part of the lake, adjoining its mouth, it is evident that between the dates of these shore-lines the River Warren eroded [p.224] its bed to this additional depth. Through a comparatively long time, represented by the Herman beach, this large outflowing river, bearing the waters supplied by the progressive glacial melting upon a vast area, had only deepened its channel slightly; but at the close of this stage the division between it and the next following Norcross stage, though doubtless only a short interval of time, was marked by a considerable increase of depth of the channel. Why was the river able to erode so much faster then than during the time of formation of the Herman beach, or of the Norcross beach afterward, which likewise represents a nearly stationary period in the progress of erosion of the Lake Traverse Valley? The answer which seems best was suggested to me by Mr. G. K. Gilbert in a letter dated February 3, 1888, as follows:

        *    *    *    Retreat of the ice modified the geoid, and perhaps produced also a crustal change, and in consequence the baselevel assumed a new attitude to the land. The river adjusted its grade to the new conditions, and then remained stationary during the formation of the Norcross beach.

        When the outlet of Lake Agassiz was changed and an avenue of discharge toward the northeast was obtained, the south end of the lake at first fell only 15 feet below the McCauleyville beach and the bed of Lake Traverse. Its numerous stages, recorded by the shore-lines of the whole time of northeastward outflow, until the retreat of the ice-sheet uncovered the present course of the Nelson, were in succession each lower than the preceding by the following amounts, as determined mostly by leveling on the latitude of Gladstone, Manitoba, 308 miles north of Lake Traverse and 84 miles north of the international boundary: The first, second, and third Blanchard beaches, respectively, about 20, 15, and 15 feet; the Hillsboro beach, 12 or 15 feet; the Emerado beach, about 30 feet; the Ojata beach, 25 feet; the Gladstone beach, 20 feet; the Burnside beach, again 20 feet; the Ossowa beach, 15 feet; the Stonewall beach, 20 feet; and the Niverville beach, 45 feet. Thence to the earliest level of Lake Winnipeg there was again a fall of about 45 feet, and erosion by the Nelson River has since lowered this lake about 20 feet.
        As soon as the ice upon Hudson and James bays and the adjoining country had so receded as to give to Lake Agassiz an outlet lower than the River Warren, it began to be drained in that direction, perhaps flowing at first across the watershed between the Poplar and Severn, and later along lower courses, including the canoe route by the Hill and Hayes rivers. Each of its successive outlets was probably eroded to a considerable depth, being occupied by the outflowing river during the time of formation of two or more beaches, until the retreat of the southeastern border of the portion of the ice-sheet remaining west of Hudson Bay finally permitted drainage to take the course of the Nelson, the ice-dammed Lake Agassiz being thus changed to Lake Winnipeg. The northeastern outflow commenced when the lake at the latitude of the south end of Lake Winnipeg stood about 1,000 feet above the present sea-level, and it was gradually lowered to 730 feet, when the Nelson, between its successive lakes, began to erode the shallow channel of the upper part of its course.
        Inspection of the series of vertical intervals between the successive levels of the lake during its northeastern drainage suggests that probably [p.227] the earliest outlet in that direction was occupied during the time of the three Blanchard beaches and the Hillsboro beach, the channel being cut down about 45 feet. The comparatively large interval above the Emerado beach may be supposed to imply the transfer of the discharge to a new outlet; and the series of smaller intervals separating the next five, beaches may indicate that they all were formed while this channel was being cut down about 100 feet. Another large fall of the lake, to the Niverville beach, which is compound in its northern part, would again mark the occupation of a new outlet. This, however, was soon abandoned for the still lower course of the Nelson. Exact heights of these old river courses, crossing present lines of watershed, and the depths of their erosion, will doubtless be determined at some future time by exploration and leveling, though probably not until after many years, on account of the difficulty of carrying instrumental surveys through that wooded and uninhabited region.

        The five or six distinct beaches that were formed by the southern part of Lake Agassiz during its outflow southward are represented in the northern part of its basin by seventeen separate shore-lines, which are marked by definite beach ridges. The individual beaches at the south, being traced northward, become double or triple, and the highest or Herman beach expands into seven successive shore-lines. During the earlier years of my exploration of this glacial lake I believed that this duplication and multiplication of the beaches observed in advancing from south to north was referable to the diminution of the attraction of the ice-sheet as its final melting progressed. Gravitation of the lake toward the vast mass of the ice, and its decrease with the glacial recession, I then supposed to be adequate to explain the observed northward ascent of the beaches, amounting for the highest Herman beach to 5 or 6 inches per mile for its first 50 miles at the south, but thence increasing northward to 1 foot and 1½ feet per mile; for the succeeding beaches, of gradually diminishing amount; and for the McCauleyville beach, the latest formed during the southward outflow, [p.228] ranging from 1 inch per mile at the south to about 2½ inches per mile from the international boundary to the latitude of Gladstone.
        With further exploration and study, including the portion of the lake area examined by me in Manitoba, I became convinced that this explanation is inapplicable to the problem, because the highest beach of the Herman series (formed contemporaneously with the six large deltas which were dependent for their formation on the accompanying retreat of the ice-sheet supplying their sand and gravel) is found to be continuous along an extent of nearly 250 miles from south to north, reaching from Lake Traverse at least to Thornhill, in Manitoba, across an area which has several prominent moraines of recession, denoting important stages of decrease of the ice-sheet. These moraines extend to the borders of Lake Agassiz, and the ice front at the time of their formation traversed the lake basin. Therefore, if diminution of the ice-attraction were the principal cause of the changes of the levels of the lake, we should expect the highest beach to cease at the successive morainic belts, and another somewhat lower to take its place thence northward.

        For aid in the investigation of this and other movements of elevation of the land following the departure of ice-sheets and the evaporation of Lake Bonneville, Mr. R. S. Woodward, of the United States Geological Survey, made a careful mathematical computation of the effects of such masses of matter formerly existing upon portions of the earth's surface to deform the geoid or level of the water of lakes or the sea.(28) His result, agreeing approximately with conclusions from similar computations by European mathematicians and physicists, shows that the North American ice-sheet, with its known area and its maximum probable thickness, would be capable of drawing the level of Lake Agassiz upward to the north not more than a quarter, or perhaps no more than an eighth or tenth, as much as the ascent of the Herman beach. It is thus evident that we must look to some other cause for explanation of these changes of level, and this is found in a differential uplifting of the lake basin, increasing in amount from south to north upon all the area where we have determined the heights of the beaches.
        [p.229] The departure of the sheets of land ice which spread drift formations over the northern part of North America, northwestern Europe, and Patagonia, was in each of these great and widely separated areas attended by a depression of the land. While each of these ice-sheets was melting away, the land on which it had lain was somewhat lower than now, and its coasts were partially submerged by the sea. These are the only extensive regions of the earth which have lately borne ice-sheets that have now melted, and it seems to be a most reasonable inference that the vast weight of their burdens of ice was an important element in the causes of their subsidence. Since the disappearance of their ice-sheets, each of these continental areas has been uplifted, probably in large measure because of the withdrawal of the ice-load. In Europe these epeirogenic movements of depression and reelevation seem to have been more nearly proportionate to the volume and extent of the ice-sheet than on our continent. Both in North and South America, other great epeirogenic movements, affecting large areas which were never glaciated, have been in progress, apparently during the same time and in close association with the oscillations of the glaciated regions. In another chapter, treating more fully of the causes of the changes in level of the beaches of Lake Agassiz, these complex movements of our continent and other parts of the world during the Quaternary era will be reviewed for the purpose of learning, if possible, how and why such subsidences and uplifts of great areas take place.
        At present we need only to inquire what were the amounts of depression of the basin of Lake Agassiz and of contiguous parts of this continent, since these would affect the history of this lake in its reduction from its highest level to its lower shore-lines and to Lake Winnipeg; and what was the manner of the reelevation, whether by regular and continued movement, or by intermittent uplifting and stages of repose, and whether the basin was uplifted differentially as a whole or in successive portions.
        Depression of the continent shown by coastal submergence.--Answering the first part of these inquiries so far as we may by the known extent of oceanic submergence of the land when it became uncovered from the ice, we have the testimony of marine fossils in beds overlying the glacial drift, which show that the country southwest of Hudson and James bays then stood [p.230] 300 to 500 feet below its present level; that the Ottawa basin was depressed 400 to 500 feet; the St. Lawrence Valley, about 250 feet at the mouth of Lake Ontario, at least 560 feet at Montreal, and 375 feet opposite the Saguenay; and the country bordering the Gulf of St. Lawrence, about 200 feet at the Bay of Chaleurs, with diminishing amount thence to the east and south, ceasing in Nova Scotia and southeastern Massachusetts. In the Mackenzie basin, evidences of marine submergence since the Glacial period have not been found; but they are discovered, up to heights of 100 to 300 feet, on the Pacific coast of the drift-bearing area. It is probable, however, that these elevations of marine deposits are not full measurements of the depression under the ice-load. The nearly complete uplifting of the basin of Lake Agassiz while the ice-sheet was retreating from it and was still the barrier of the waning glacial lake proves that the reelevation closely followed the departure of the ice, and suggests that in the districts of these marine beds some uplifting may have been done while the ice above was becoming thin, but had not wholly disappeared, or at least before its retreat had opened ways of ingress for the sea.
        Depression and reelevation of the basin of Lake Agassiz shown by differentially uplifted beaches.--If we next seek a measure of the subsidence of the basin of Lake Agassiz while it was ice-burdened, no marine beds above the drift in this district can aid in giving an answer, but we must look to the known amount of northward uplifting of the once level beaches, and from this differential elevation it seems well-nigh sure that the maximum depression which this basin underwent failed to sink its lowest part, the shallow bed of Lake Winnipeg, to the sea-level. The central and northern portion of the area of Lake Agassiz, where the great lakes of Manitoba are now outspread, was depressed apparently 400 or 500 feet, carrying the present shores of Lake Winnipeg down to an altitude of only 300 or 200 feet above the sea; but the bed of this lake, which is less than 100 feet deep, was still above the ocean. The amount of subsidence here is thus found to be harmonious with that of other parts of our glaciated area which bordered the oceans and Hudson Bay. As a whole, the ice-enveloped portion of the continent is seen to have sunk slightly more in its central region than on its boundaries. The vertical extent of the maximum known depression, [p.231] determined by marine fossils of the Champlain epoch and by the inclined beaches of this glacial lake, ranged from no subsidence in the greater part of Nova Scotia to probably 600 feet at Montreal, nearly the same at Ottawa and about James Bay, approximately 500 feet in Manitoba, none or little on the Mackenzie, and from 300 to 100 feet, probably decreasing from north to south, on the shores of the Queen Charlotte Islands and British Columbia.
        Some addition to these figures, but probably nowhere exceeding a quarter or third more, is required to give the earlier extreme extent of the subsidence of the ice-weighted land, thus including its rise before the ice above had wholly melted, or before the sea was admitted to Hudson and James bays, and to the St. Lawrence, Lake Champlain, and Ottawa valleys. But this small added amount was offset in part or entirely by the effect of gravitation, which raised the levels of the ocean and lakes toward the ice-sheet. These two causes of changing levels acted in conjunction in their relationship to the series of shore-lines of Lake Agassiz, and to the position and course of its outlets after it fell below the channel at Lake Traverse; but the effect of ice attraction must be deducted from the total, if we ask the extent of epeirogenic subsidence and reelevation, which therefore are probably closely expressed in the figures before stated, having a maximum of 500 to 600 feet.
        Improbable hypothesis of an outlet from Lake Agassiz to the Mackenzie River.--We may therefore dismiss as untenable the supposition that the outflow of Lake Agassiz, after falling below Lake Traverse and the McCauleyville beach, and being still obstructed from going to Hudson Bay by the presence of a large remnant of the ice-sheet there, could have passed for a time across the divide between the Churchill and Athabasca rivers, thus being discharged into the Mackenzie and the Arctic Ocean. Such northwestward outflow would have crossed the present watershed near the Methy portage, or by way of Wollaston or Hatchet Lake, which has two outlets, one to the Churchill and the other to the Athabasca. The altitude of the summit of the Methy portage, according to Richardson's observations, with correction for the now better-known heights of Lake Winnipeg and the Saskatchewan, appears to be about 1,750 feet above the [p.232] sea; and Methy Lake, at the head of a series of lakes and connecting streams tributary to the Churchill, is about 50 feet lower. According to Dr. Robert Bell, "there is said to be a continuous watercourse" near this portage, passing from the Clearwater, a branch of the Athabasca, into the Churchill basin;(29) but its height forbids the inference that the waters of Lake Agassiz ever outflowed there, for a subsidence of more than 700 feet would be required to reduce Methy Lake and the divide in its vicinity to the level of Lake Traverse. Instead, as was stated on page 64, it is my belief that this channel was the outlet of a lake in the Athabasca basin, dammed by the barrier of the receding ice-sheet on the north and thus made tributary to Lake Agassiz. The other pass over the watershed, by the way of Hatchet and Jackfish lakes, situated 300 to 375 miles northeast of the Methy portage, is probably 1,300 or 1,400 feet above the sea, and presents nearly equal difficulty to the hypothesis of an outlet from the Winnipeg basin to the McKenzie; but again there is much likelihood that this course also served, at a later date, as an important avenue of inflow to Lake Agassiz from the Athabasca glacial lake.
        Probable hypothesis of the discharge from the northeastward outlet being tributary successively to the Mississippi and Hudson rivers.--When the discharge by the River Warren ceased, the new outlet flowed northeastward. Perhaps, as before stated, it turned back for some time to the south along the border of the waning ice-sheet and thus still passed into the Mississippi by the way of Lakes Superior and Michigan. Stranger yet, through the effect of subsidence, which greatly modified the conditions of drainage in the Champlain epoch, as pointed out by Mr. Gilbert,(30) this overflow, prevented by the ice-barrier from going in the direction of the land slope to Hudson Bay, may have been later carried into the Atlantic by the Mohawk and Hudson rivers; or, at a still later stage, it may have taken its course past the mouth of Lake Ontario to the sea near the head of the greatly enlarged Gulf of St. Lawrence, which then filled the St. Lawrence Valley, the basin of Lake Champlain, and the Ottawa Valley to Allumette Island [p.233] or higher.(31) Perhaps, last of all, before the glacial recession admitted the sea to Hudson Bay, this discharge from Lake Agassiz, flowing through a great glacial lake in the southwestern part of the basin of Hudson and James bays, would find its lowest and final outlet to the south by the way of Lakes Abittibi, des Quinze, and Temiscaming, then entering the Ottawa arm of the Gulf of St. Lawrence. It seems even possible that the vicissitudes of the changing courses of drainage produced by the gradual retreat of the ice may have included for the outflow from Lake Agassiz, after it began to pass first northeastward, all of these four ultimate routes; first, by Chicago to the Mississippi; second, by the glacial Lakes Algonquin, Lundy, and Iroquois, and the Mohawk and Hudson rivers, to the Atlantic; third, by the lake portion of this route, and perhaps for some short time by Lake Nipissing and the Mattawa River, to the head of the Gulf of St. Lawrence, then filling the St. Lawrence Valley and a part of the Ottawa basin; and fourth, by Lake Abittibi, crossing the lowest point of the watershed between James Bay and the St. Lawrence. The known epeirogenic subsidence of the Champlain epoch and the probable manner of recession of the ice border from south to north make each and all of these courses, diverting the northeastward outflow to the south, far more probable than either of the courses before considered by which this glacial lake might be supposed to send its overflow to the Mackenzie.
        Division of the ice-sheet into parts east and west of Hudson Bay.--It seems to me most probable, however, that long before the complete departure of the ice-sheet it became melted in twain by the laving action of Lake Agassiz and of a great glacial lake in the southwestern and southern part of the basin of Hudson Bay, and on its other side by the sea washing its ice-cliffs in Hudson Strait and the northern part of Hudson Bay, so that the latest general glaciation of our continent was confined to two areas, one east and the other west of this vast mediterranean sea. Several of the lower shorelines of Lake Agassiz, formed during its northeastward drainage, if not all of them, doubtless mark levels of outlets which flowed into this inland sea, until the northward recession of the remnant of the ice-sheet on its west side laid bare the course of the Nelson. Careful collection and study [p.234] of observations of the bearings of glacial stræ on all portions of Canada to the far north, and examination of the lowest points of watersheds as to their glacial river courses, will be the means of displacing these speculations by definite knowledge and proofs of what were the fortunes of the departing ice-sheet and of the late outlets of this lake.(32)
        Amount of differential elevation between Lake Traverse and Gladstone.--How the Red River Valley and the lake country of Manitoba were uplifted from the late glacial or Champlain depression is told by the inclination of the Lake Agassiz shore-lines. So far as exploration and determination of the heights of the beaches have extended including both my own and Mr. Tyrrell's work, there is found to be a northward ascent of the old lake levels, greatest in amount along the earlier and higher beaches, and diminishing almost to horizontality in the latest and lowest. Comparing the heights of the beaches at or near the mouth of the lake with their heights about 300 miles to the north, on the latitude of Gladstone, which is near the northern limit of my observations, it is seen that the epeirogenic uplifting of this part of the lake basin, increasing gradually from south to north, and the fall of the lake surface, also greatest at the north on account of the decreasing effect of gravitation toward the diminishing and receding ice-sheet, were together approximately 265 feet in this distance, averaging nearly 1 foot per mile, after the formation of the second Herman beach, which is the highest found on that latitude. Of this combined uplift of the land and fall of the lake, about 80 feet had taken place before the formation of the Norcross beach; 50 feet more before the upper Tintah beach; about 45 feet more before the Campbell beach; and again some 25 feet more before the McCauleyville beach; leaving only 65 feet of the whole 265 feet of changed level to take place after the lake began to outflow northeastward, and it appears that all but about 20 feet of this remaining change had been accomplished before the formation of the lowest or Niverville beach. While the ice was departing from the country and still was the barrier of the lake, this part of its basin was uplifted nearly to its [p.235] present altitude, and the total amount of the differential elevation in this extent of 300 miles, after subtracting a quarter part for the probable or possible effect of ice attraction, was about 200 feet.
        Alternate stages of elevation and rest.--The considerable number of definite additional shore-lines observed in proceeding to the north indicates, like the stages of erosion by the River Warren between the times of formation of the beaches near Lake Traverse, that there were periods of comparatively rapid uplifting which alternated with others of repose or of very slow progress of the general epeirogenic movement. Vertical uplifts of 10 to 20 or 25 feet were many times repeated, and were separated by longer intervals of rest. But the initiation of these stages of uplift was delayed until long after Lake Agassiz began to exist. The ice-sheet had retreated from Lake Traverse to Manitoba, and three or four conspicuous moraines of recession had been formed, while the lake level reposed undisturbed during the formation of the first and highest beach of the Herman series and the accumulation of the contemporaneous deltas. Such tardiness in the beginning of the elevation of this area, as it is recorded by the inclined shore-lines, implies, and, indeed, makes it almost certain, that very little uplifting, if any, had taken place during the time of melting away of the ice above. If the restoration of the land to its wonted height had already begun under the thinned edge of the ice, it would probably have gone forward more promptly, while the Red River Valley was being gradually occupied by Lake Agassiz, following upon the retreat of the ice-front.
        Later and greater inclination of beaches along the base of Riding and Duck mountains.--On the area of 300 miles extent from south to north between the mouth of the lake and Gladstone, the epeirogenic differential uplift was mostly done before the times of formation of the Campbell and McCauleyville beaches, the last two belonging to the southward outlet; but farther to the north, within the area at the base of the escarpment of Riding and Duck mountains, where Mr. Tyrrell has mapped the beaches of this lake and determined their heights, a very important differential elevation, amounting to about 3 feet per mile along a distance of 50 miles between Valley and Duck rivers--that is, between latitudes 51° 15' and 52° north--took place after the Campbell and McCauleyville beaches were formed, since they are [p.236] thus remarkably changed from their original horizontality. It is clearly shown here that the uplifting was not uniformly proportionate and regular for the whole area of Lake Agassiz. The chief movements of elevation of its southern and central part, as far to the north as Gladstone, seem not to have extended farther, at least in their full proportion. The district next to the north along an extent of 120 miles to Pine, Duck, and Swan rivers, at the north end of Duck Mountain, was perhaps only so far disturbed by these movements as was necessitated to connect the rise of the country on the latitude of Gladstone with the continuing condition of maximum subsidence on the latitude of the lower part of the Saskatchewan and the north end of Lake Winnipeg. But there ensued in this district, after the date of the Campbell beach, a great differential elevation, giving to these late shore-lines two to three times more northward ascent than that of the Herman beach from Lake Traverse to Gladstone; and the total change in level of the highest observed beach, probably representing the upper Norcross stage, situated at Pine River, on latitude 51° 50' to 52° north, is approximately 400 feet as compared with this shore-line at Lake Traverse, about 420 miles distant to the south. Nearly the whole uplift of the northern part of the basin was accomplished, however, while the ice-sheet was still a barrier of the lake, for the Niverville beach at the Grand Rapids of the Saskatchewan is only slightly higher than on the Red River, 250 miles to the south.
        Review of the epeirogenic uplifting.--After the recession of the ice from its vicinity, the mouth of Lake Agassiz by the River Warren was uplifted apparently at least 50 or 75 feet, and perhaps as much as 90 feet, by several small stages of elevation, separated by comparatively long pauses. Thence to the latitude of Gladstone, in a distance of 300 miles northward, such small uplifts, increasing in number and in aggregate vertical amount from south to north, raised the lake basin in southern Manitoba not less than 200 feet; and, in combination with the fall of the lake level northward, due to decreasing ice attraction, the change in level was 265 feet. To these figures we must add the uplift of the Lake Traverse region, which was probably between 50 and 100 feet, to obtain the total epeirogenic elevation at Gladstone. Later epeirogenic movements of the same kind raised a more [p.237] northern part of the basin, on the latitude of 52° north, about 400 feet, if we neglect the fall of the lake level, in comparison with Lake Traverse. At the same time, or possibly still later, the northern end of the area of Lake Agassiz and the adjoining portion of the Churchill basin were uplifted a similar amount. Last of all, when Lake Winnipeg and the Nelson River had come into existence, the shores of Hudson and James bays were raised 300 to 500 feet from their temporary postglacial marine submergence.
        The elevation of the eastern shore-lines of Lake Agassiz, in Minnesota, exceeded that of the western shores, in North Dakota; and the ratio of this eastward ascent of the old lake levels to their doubly greater northward ascent implies that the tilting of this area was from south-southwest to north-northeast. Again, at the north end of Duck Mountain, the west-to-east portions of beaches observed by Mr. Tyrrell, between the Swan and Duck rivers, show a similar eastward ascent, about half as much as the northward rise along the eastern base of this highland. It thus appears true of the greater tilting of that northern district also, after the formation of the Campbell beach, that its maximum ascent was toward the north-northeast; but, like the elevation between Lake Traverse and Gladstone, this movement was doubtless of limited extent, so that the country adjoining Hudson Bay retained nearly or quite its maximum depression until the somewhat later time when the sea was admitted to that basin.

        Fossils have been found in the deposits of Lake Agassiz at two localities. They are all fresh-water shells of species now living in this district, occurring in beach ridges where excavations have been made to obtain sand for masons' use. The Campbell beach, about 6 miles southwest of Campbell, Minn., at an elevation approximately 985 feet above the sea, has thus yielded shells of Unio ellipsis Lea, a common species of the Upper Mississippi region. In the Gladstone beach, a half mile northeast of Gladstone, Manitoba, about 875 feet above the sea and 165 feet above Lake Winnipeg, four species occur in considerable abundance from 2 to 4 feet below the surface, namely, Unio luteolus Lamarck, Spærium striatinum [p.238] Lam., Spærium sulcatum Lam., and Gyraulus parvus Say. These species from both localities were kindly determined by, Prof. R. Ellsworth Call, who states that Unio lateolus is one of the most widely distributed representatives of the genus, its range being from Lake Winnipeg to Texas, east to New York, and west to Montana. It is generally abundant in Minnesota. Both these species of Spærium are reported by Dr. Dawson from the Lake of the Woods and Pembina River; and the first is the most common species of its genus in Minnesota, while its range northward extends at least to Great Playgreen Lake and York Factory, where it has been collected by Dr. Bell. The Campbell beach was formed in the later part of the time of the lake's southward outflow; and the Gladstone beach belongs to the middle portion of the time of its outflow toward the northeast, its south end being then about 90 miles south of the international boundary.

        If the question be asked, How many thousand years ago did the recession of the ice-sheet take place, causing Lake Agassiz to fill the Red River Valley and the basin of Lake Winnipeg? a reply is furnished by the computations of Prof. N. H. Winchell,(33) that approximately eight thousand years have elapsed during the erosion of the postglacial gorge of the Mississippi from Fort Snelling to the Falls of St. Anthony; of Dr. Edmund Andrews,(34) that the erosion of the shores of Lake Michigan, and the resulting accumulation of dune sand drifted to the southern end of that lake, can not have occupied more than seven thousand five hundred years; of Prof. G. Frederick Wright,(35) that streams tributary to Lake Erie have taken a similar length of time to cut their valleys and the gorges below their waterfalls; of Mr. G. K. Gilbert,(36) that the gorge below Niagara Falls has required only seven thousand years or less; and of Prof. B. K. Emerson,(37) on the rate of [p.239] deposition of modified drift in the Connecticut Valley at Northampton, Mass., from which he believes that not more than ten thousand years have elapsed since the Ice age.
        Making such inquiry also concerning the glaciation of Europe, we find that in Wales and in Yorkshire, England, the amount of denudation of limestone rocks on which bowlders lie has been regarded by Mr. D. Mackintosh as proof that a period of not more than six thousand years has elapsed since the bowlders were left in their positions.(38) The vertical extent of this denudation, averaging about 6 inches, is nearly the same with that observed in the southwest part of the Province of Quebec by Sir William Logan and Dr. Robert Bell, where veins of quartz marked with glacial striæ stand out to various heights not exceeding 1 foot above the weathered surface of the inclosing limestone.(39)
        Another indication that the final melting of the ice-sheet upon British America was separated by only a very short interval, geologically speaking, from the present time is seen in the wonderfully perfect preservation of the glacial striation and polishing on the surfaces of the more enduring rocks. Of their character in one noteworthy district Dr. Bell writes as follows: "On Portland promontory, on the east coast of Hudson's Bay, in latitude 58°, and southward, the high, rocky hills are completely glaciated and bare. The striæ are as fresh-looking as if the ice had left them only yesterday. When the sun bursts upon these hills after they have been wet by the rain, they glitter and shine like the tinned roofs of the city of Montreal."(40) Again, Professor Macoun writes of the red Laurentian gneiss in the vicinity of Fort Chipewyan, at the west end of Lake Athabasca: "The rocks around the fort are all smoothed and polished by ice action. When the sun shines they glisten like so much glass, and a person walking upon them is in constant danger of falling."(41) It seems impossible that these rock exposures can have so well withstood weathering in the severe climate of those northern regions longer than a few thousand years at the [p.240] most; and they even suggest that remnants of the continental ice-sheet may have lingered there considerably later than the time, computed to be six to eight thousand years ago, when its southern portion retreated.

        The foregoing measures of time, surprisingly short, whether we compare them on the one hand with the period of authentic human history or on the other with the long record of geology, carry us back to the date when the ice-sheet was melting away from the basins of the Upper Mississippi, of the Red River of the North, and of the Laurentian lakes. If the postglacial epoch has been so short, we must infer that the final recession of the ice was very rapid and that its barrier between the Red River Valley and Hudson Bay was soon melted away. Though Lake Agassiz attained vast areal extent, its duration or extent in time was geologically brief, as is shown by the small volume of its beach deposits and lacustrine sediments in comparison with the Pleistocene lakes of the Great Basin and with the amount of postglacial erosion and deposition on the shores of the great lakes tributary to the St. Lawrence and Nelson rivers. The geologic suddenness of the final melting of the ice-sheet, proved by the brevity of existence of its attendant glacial lakes, presents scarcely less difficulty for explanation of its causes and climatic conditions than the earlier changes from mild or warm preglacial conditions to prolonged cold and ice accumulation.
        Comparison with postglacial lakes.--The disappearance of the greater part of the vast North American ice-sheet probably occupied not more than two or three thousand years; and half of this time may measure the duration of Lake Agassiz, with the formation of its beaches marking more than thirty successive stages in the concurrent subsidence of its surface and rise of the earth's crust. But even these short estimates may be too long. The shores of Lake Michigan, similar with those of Lake Agassiz, in the drift of which they are formed, in their north and south trends, and in the adjoining depths of water, have suffered an amount of erosion by the lake waves during postglacial time which very far exceeds the total erosion that was effected upon the shores of Lake Agassiz during all its stages, the [p.241] proportion between them being surely not less than ten to one; and Lake Michigan has a similarly greater amount of beach deposits, which upon a large area about its south end are raised by the wind in conspicuous dunes. This contrast, indeed, suggests that the duration of Lake Agassiz and the recession of the ice-sheet from Lake Traverse to the lower part of the Nelson River may have been included within less than one thousand years.

        Likewise, as Mr. Tyrrell remarks, beach ridges of larger size than those of Lake Agassiz, and composed of coarser shingle, occur on each of the three great lakes of Manitoba, although these lakes are far smaller than their glacial predecessor and therefore surely have less powerful waves.
        Comparison with Lakes Bonneville and Lahontan.--During the first high stage of Lake Bonneville, a fine, laminated, olive-gray clay, which weathers to a pale-yellow color, was spread throughout all the lower parts of its basin, ascending also in the shallower bays toward the upper shore-lines. In two typical deep sections this member of the lacustrine sediments has an exposed thickness, respectively, of 90 and 100 feet, but its base is not seen. Again, during the second rise of this lake it deposited a similarly widespread stratum of light-gray or cream-colored marl or calcareous clay, weathering nearly white, and passing upward into a fine sand; and typical sections show that this marl and associated sand range from 20 to 50 feet or more in thickness. In like manner, Lake Lahontan during its two high-water periods deposited over all its bed fine marly clays, which in the earlier flood stage attained a thickness of more than 100 feet, their base not being exposed by the deepest sections, and in the later stage an average of 50 to 75 feet. These thick sediments occupying the basins of the Pleistocene lakes of Utah and Nevada indicate, like their great amount of shore erosion and correspondingly massive beach deposits, that the term of existence of these lakes during each of their high stages, and especially the first, far exceeded that of Lake Agassiz. No such lacustrine beds are generally spread over the basin of this glacial lake, which upon large tracts, even of its lower portion, as on and near the Red River, near the lower Assiniboine between Poplar Point and Winnipeg, and adjacent to Lake Manitoba, Shoal Lake, and Lake Winnipeg, consists of till, with frequent bowlders, the [p.242] direct product of the ice-sheet, very slightly changed by its deposition in the lake, and not covered by any aqueous sediments.
        Only where tributaries entered this lake and brought, besides the ordinary alluvium of their erosion, a much larger volume of modified drift from the melting ice-sheet, were any important lacustrine sediments laid down; and these appear in the form of extensive deltas of gravel and sand, with fine silt spread over adjoining parts of the lake bottom. Other inflowing streams, though in several instances important rivers, as the Red River itself above Fergus Falls, the Wild Rice River of Minnesota, and the Red Lake River, formed no noteworthy delta accumulations, which, however, could not have failed to be conspicuously developed if the lake had long remained at any of the levels of its many successive shore-lines. The sediments in Lakes Bonneville and Lahontan were evidently derived in large part from wave erosion of their shores; but in the case of Lake Agassiz, though its shores are the easily eroded drift, no appreciable lacustrine beds were supplied from this source. The duration of Lake Agassiz was very short in comparison with either of the Pleistocene humid epochs of the Great Basin and Cordilleran mountain belt.
        Brevity of time required for the formation of terminal moraines.--The shortness of the existence of Lake Agassiz seems, at first thought, to present a difficulty in the brevity of the time which would be allowed, if the glacial lake endured only a thousand years or less, for the accumulation of the moraines described in the preceding chapter. By the probable ratios of time deduced from the extent of the upper Herman beach and the sequence of all the lower and later beaches, we see that the formation of even the great moraines of the Leaf Hills and of the south side of Devils Lake could have occupied only a small fraction of the whole duration of Lake Agassiz, perhaps not more than fifty or even twenty-five years for amassing these morainic hills 100 to 350 feet high on a belt 3 to 5 miles wide! For the Dovre, Fergus Falls, Leaf Hills, and Itasca moraines were apparently formed before the completion of the highest one in the series of four principal beaches which unite in the Herman beach along the southern 75 miles of Lake Agassiz. But this may be easily accepted when we recall the rapidity of motion of the thick and wide glaciers of Greenland and [p.243] Alaska, 30 to 100 feet per day.(42) Doubtless the continental ice-sheet moved faster than the glaciers of the Alps, but the waste from its border by melting must evidently have been less than the discharge of ice from these Arctic glaciers where they terminate in the sea and are broken into bergs and floated away.
        The two factors on which the accumulation of the terminal moraines depends are the rate of motion of the ice-sheet and the amount of the englacial drift which was thus brought forward to its margin. In the region of Lakes Benton, Shaokatan, and Hendricks, in southwestern Minnesota, we have evidence that the drift contained within the ice amounted to a sheet at least 40 feet thick.(43) As great volume of englacial drift is also indicated in Manitoba by the relation of the Birds Hill esker to the adjoining sheet of till.(44) The inequalities in the aggregate mass of the drift forming different portions of the morainic belts, causing these to rise in great prominence upon some areas, while in other places they are low and scanty, seem due to unequal distribution of this drift within the basal part of the ice-sheet.(45) It was most abundant in those portions where glacial currents had converged between the great lobes of the ice border during the time of maximum area of this ice-sheet, as from the vicinity of Minneapolis and Lake Minnetonka northwestward to the Leaf Hills, to Lake Itasca, and to Birds Hill, and in the country west and northwest of Cooperstown, N. Dak., to the Washington Lakes, Devils Lake, the Big Butte, Broken Bone Lake, and to Turtle Mountain. Upon these areas the morainic belts show a close relationship, not only by their parallelism and the similar positions of [p.244] reentrant angles of the ice border through several stages in its retreat, but also by remarkably massive accumulations of drift in corresponding portions of the successive moraines. If the average amount of englacial drift thus supplied by the ice-sheet where its moraines are largest was equal to a thickness of 40 feet, or even 20 or 10 feet, it will be seen that these moraines would require, with a steep frontal gradient of the ice due to the marginal melting and a consequent rate of glacial motion probably several times faster than that of the Alpine glaciers, only a few decades or years for their formation.

By T. C. CHAMBERLIN.