The western part of the Cypress Hills is entirely unglaciated, and must have formed an island in glacial times projecting about 400 feet above the surface, as no drift or other mark of glacial action was observed within that distance of the summit, and as this part has a height of about 4,800 feet above the sea, this would give the surface of the glacial sea or glacier, disregarding Post-Tertiary changes in elevation, [p.118] a height of 4,400 feet above the present sea-level. The Hand Hills are stated by Mr. Tyrrell to be unglaciated above a height of 3,400 feet,(20) and as these hills are situated N. 40° W. from the western end of the Cypress Hills, at a distance of about 150 miles, a line connecting the bases of the driftless parts of the two plateaus would incline toward the northwest at a rate of 6.7 feet per mile, and would have an average elevation above the present surface of about 1,550 feet. Drift was also observed by Dr. G. M. Dawson on the West Butte [of the Sweet Grass Hills] at an elevation of 4,660 feet, or 1,260 feet above the level at which it disappears in the Hand Hills, which are in nearly the same meridian, and 260 feet above the same point in the Cypress Hills. These differences in level, divided by the difference in latitude of the several elevations, afford evidence of a Post-Tertiary depression of the plains to the north in this region, relatively to those in the vicinity of the forty-ninth parallel, of about 7.2 feet per mile. The glacial sea or continental glacier is also shown, by subtracting the elevations given above from the present level of the surface, to have had a maximum depth in the plains surrounding the Cypress Hills of 2,000 feet, and to have averaged about 1,500 feet.(21)

        The prevailing courses of glaciation and dispersal of the drift lead me to recognize, with Dr. Dawson, the existence of two central areas upon which the ice was accumulated in greater depth than elsewhere, and from which consequently it flowed outward on all sides.(25) One of these areas [p.120] embraced the Laurentide highlands, James Bay, a portion of Hudson Bay, and the western part of the Archean region from Lakes Superior and Winnipeg to Great Slave and Great Bear lakes. From this large northeastern or Laurentide center of outflow the ice-sheet crept southward, eastward, and northward to the limits of glaciation before noted. Westward the ice from this area outflowed, as I believe, to the limit of Archean bowlders on or near the base of the Rocky Mountains, where I find, from Dr. Dawson's observations of the drift in Alberta and on the Peace River, that it abutted against and was confluent with ice outflowing eastward and southeastward from the Rocky Mountains. The other area whence currents of the ice-sheet flowed radially in every direction was the northern central part of British Columbia; and the portions of the ice-sheet pouring outward respectively from these two centers have been named by Dawson the Laurentide and Cordilleran glaciers. Toward the south, west, and northwest the Cordilleran outflow extended to the boundaries of our glaciated area; but eastward, pouring through passes of the Rocky Mountains, and in the Peace River region probably overtopping the highest summits, which there are only about 6,000 feet above the sea, the Cordilleran ice pushed across a narrow belt adjoining the mountains to a maximum distance of nearly a hundred miles, and there (on land about 2,500 feet above the sea) became confluent with the Laurentide ice; the two united currents thence passing in part to the south and in part to the north from the interior tract where the confluent ice was thickest.

        Taking up the particular description of localities where the junction of the Laurentide and Cordilleran drift has been observed, we may begin at the international boundary and proceed northward. Laurentian erratics and drift are stated by Dawson to extend quite to the foot of the Rocky Mountains near the forty-ninth parallel, and to occur between the forty-ninth and fiftieth parallels, "stranded on the surface of moraines produced by the large local glaciers of the Rocky Mountains."(26)
        In Montana, within 30 miles southward from the forty-ninth parallel, Prof. G. E. Culver finds that ice was accumulated so thickly west of the [p.121] main eastern range of the Rocky Mountains that it outflowed eastward through the passes, carrying diorite bowlders from ledges west of the watershed to a distance of several miles on the plains at the eastern base of the mountains. No Laurentian drift was observed there, but in the valley at the head of St. Marys River, a tributary of the Belly River, on longitude 113° 30', 5 to 20 miles south of the international boundary, shore-lines of a glacial lake, which was probably formed by the neighboring barrier of the Laurentide ice-sheet on the northeast, occur up to the height of at least 800 feet above the present St. Marys lakes, or approximately 5,400 feet above the sea.(27)
        In the neighborhood of Calgary, which is the western limit of Laurentian bowlders and till, Dawson reports somewhat farther westward a deposit resembling bowlder-clay, in which the stones "are entirely those of the mountains or sandstone blocks from the underlying beds." Accordingly, he declares that the absence of Laurentian erratics west of Calgary is probably to be accounted for "by the existence of Rocky Mountain glaciers of sufficient size in this region to fend off the eastern glaciating agent." Again, be mentions, west of Calgary, "heavy glacial striation in a southward or southeastward direction    *    *    *    about 13 miles east of the mountains, in a region of wide valleys and low foothills."(28)
        On the Peace River, in its course close east of the Rocky Mountains, and on its tributary, Pine River, Dawson reports drift containing a large proportion of "hard quartzite pebbles like the more resistant materials of the axial range of the Rocky Mountains. These are mingled with a preponderating number of fragments of the softer sandstones of the country, and embedded in a whitish or cream-colored silty clay, not unlike the material representing the bowlder-clay over wide districts west of the Rocky Mountains. No Laurentian or other fragments of eastern origin were observed in this region." Continuing eastward, these drift deposits become more conspicuous, attaining in places a thickness of 150 feet. On reaching the D'Echafaud River, about 100 miles from the mountains, [p.122] though "no change in the character of the drift deposits was noted,    *    *    *    Laurentian pebbles and bowlders were for the first time seen in considerable abundance.    *    *    *    East of this point    *    *    *    the surface is thickly covered with drift deposits, so much so that exposures of the underlying rocks are, as a rule, only found in the larger river valleys."(29) No better evidence could be desired by a glacialist, accounting for the formation of the bowlder-clay by the agency of land ice, to demonstrate the confluence here of two currents of the ice, one flowing eastward from the Cordilleran area and the other flowing westward from the Archean area, whose nearest portion is on Lake Athabasca, about 400 miles distant.
        Near the divide between the Liard and Yukon River systems, Dawson found drift on the summit of an isolated mountain 4,300 feet above the sea and about 1,000 feet above this part of the Pacific-Arctic watershed.(30) This, however, is on the west side of the Rocky Mountains proper, which, as defined by Dawson, constitute the northeastern marginal range of the broad mountainous Cordilleran belt. With this definition, the Rocky Mountains are intersected by the Mackenzie River south and west of Great Bear Lake.
        Farther northward the Laurentide or eastern portion of the ice-sheet pushed northwestward to the extreme limit of the drift. "The till near the lower ramparts of the Mackenzie," according to Mr. R. G. McConnell, "is in approximately the same latitude as the northern boundary of the Archean area on the east, and the gneissic bowlders which it contains must have traveled either directly west or northwest in order to reach their present situation." He therefore infers that "the ice from the Archean gathering grounds to the east poured westward through the gaps and passes in the eastern flanking ranges of the Rocky Mountains until it reached the barrier formed by the main axial range, when, being unable to pass this, it was deflected northwestward in a stream from 1,500 to 2,000 feet deep down the valley of the Mackenzie and thence out to sea."(31)
        All the testimony thus gathered concerning the line of junction and the limits of the eastern and western drift seems to the present writer to [p.123] amount to full and convincing proof that the ice of the Laurentide and Cordilleran areas of outflow became confluent, and at its culmination stretched as one continuous ice-sheet from the Atlantic to the Pacific, enveloping the northern portion of the Rocky Mountains in their comparatively low development within the basin of the Mackenzie and Peace rivers, and overspreading the whole of the Dominion of Canada southward, except the highest parts of the Rocky, Selkirk, and Coast ranges.

        An ice-sheet similar to that of North America in the Glacial period now covers the Antarctic lands, and another is spread over the interior of Greenland. The latter has been so far explored within the past ten years by Nordenskjöld, Peary, and Nansen as to give us a knowledge of its slopes and the altitude of its surface, with which the ancient ice-sheets of North America and Europe may be most instructively compared.
        The first long journey on the Greenland ice-sheet was accomplished by Nordenskjöld in 1883, going eastward from Aulatsivik Fjord, close south of Disco Bay, near latitude 68° 20' north. At a distance of about 73 miles from the head of this fjord and edge of the inland ice Nordenskjöld reached an altitude of 4,950 feet; and at a probable distance of 45 or 50 miles farther, crossed by Lapps on the peculiar snowshoes called "ski," the barometers indicated a height of 6,386 feet. The average ascent of the ice surface here in the first 73 miles, including the more rapid rise near the margin, is about 68 feet per mile, or slightly less than three-quarters of a degree; but in the next 45 miles of estimated distance it is reduced to 32 feet per mile, or about a third of a degree.(32)
        A second important journey on the inland ice of Greenland was by Lieut. R. E. Peary and Christian Maigaard in 1886, going east from the head of Pakitsok Fjord, on the northeast part of Disco Bay, in latitude 69° 30' north. These explorers advanced to an estimated distance of about 100 miles from the edge of the ice, attaining an altitude of about 7,500 feet.(33)
        [p.124] Two years later, in 1888, Dr. Fridtjof Nansen crossed this ice-sheet from east to west between latitude 64° 10' and 64° 45' north. The width of the ice here is about 275 miles, extending into the ocean on the east, but terminating on the west about 14 miles from the head of Ameralik Fjord and 70 miles from the outer coast-line. For the first 15 miles in the ascent from the east, rising to the altitude of 1,000 meters, or 3,280 feet, the average gradient was nearly 220 feet per mile. In the next 35 miles an altitude of 2,000 meters, or 6,560 feet, was reached; and the average gradient in this distance, between 15 and 50 miles from the margin of the ice, was thus about 94 feet per mile, or a slope very slightly exceeding 1 degree. The highest part of the ice-sheet, about 112 miles from the point of starting, was found to have an altitude of 2,718 meters, or about 8,920 feet. Its ascending slope, therefore, in the distance from 50 to 112 miles was about 38 feet per mile. Thence descending westward, the gradients are less steep, averaging about 25 feet per mile for nearly 100 miles to the altitude of 2,000 meters, about 63 feet per mile for the next 52 miles of distance and 1,000 meters of descent, and about 125 feet per mile for the lower western border of the ice.(34)
        Lieutenant Peary, in an expedition from Inglefield Gulf, near latitude 78°, on the northwest coast of Greenland, starting early in May and returning August 6, 1892, crossed the northwestern and northern parts of this ice-sheet, reaching altitudes of 5,000 to 8,000 feet, and determining approximately the northern boundary of the ice from Petermann Fjord to the eastern coast at Independence Bay, in latitude 81° 37' and longitude 34° west from Greenwich.
        In comparing the slopes and altitudes of the upper limits of glaciation on mountains in Maine, New Hampshire, and New York, with the ice in Greenland, we observe the remarkable contrast that the former show gradients only about half as steep as the latter. Mount Washington, as before noted, indicates an average gradient of only about 25 feet per mile for the rise of the ice surface along a distance of 220 miles from its margin during the principal part of the Glacial period; to Mount Katahdin in a similar distance it appears to have risen somewhat less steeply, or perhaps [p.125] nearly the same; and to the Catskills the apparent ascent was only 30 feet per mile for the distance of 100 miles from the ice border. But in Greenland all of the four journeys on the inland ice find it to ascend with much steeper slopes, attaining the altitude of the summit of Mount Washington at distances which vary from 50 to 125 miles from its edge. Nor does the less area of the Greenland ice explain its steeper gradients, for it probably has a length of more than 20 degrees from south to north, or over 1,400 miles, with a width of 200 to 600 miles, and an area of about 600,000 square miles, or one-seventh as much as the later ice-sheet of North America.
        Apparently the conditions for outflow of the ice from this area are similar and equally favorable with those which prevailed on our continent in the Glacial period. The comparison therefore suggests that the present elevation of the glaciated portion of this continent is probably much changed from that which it had during its epochs of glaciation. If the North American ice-sheet during its stages of growth and culmination attained steep slopes and high altitudes near its borders comparable with the Greenland ice, the records of glaciation on our mountains show that during the time of accumulation of the ice and until it attained its maximum extent the glaciated area was uplifted as a high continental plateau, with the same principal topographic features of mountains, valleys, and general contour as in preglacial and postglacial times, but having in its outer 100 or 200 miles slopes of probably 20 to 30 feet per mile, descending from the plateau of the interior of the ice-enveloped country to its margin.(35)
        Similar uplifting seems also to have affected the glaciated northwestern portion of Europe, for there, too, the slopes and height of the limits of the drift resemble those of North America rather than the Greenland ice-sheet. Prof. James Geikie finds that the surface of the ice which moved westward from northern Scotland across the Minch and the Hebrides had a descent of 25 feet per mile; "but slight as that incline was," he remarks, "it was probably twice as great as the slope of the mer de glace that filled up the German Ocean."(36) Mr. T. F. Jamieson therefore concludes that when the [p.126] ice-sheets of Scandinavia and Scotland were being accumulated these countries stood far above their present height, the maximum uplift being at least equal to 4,080 feet, which is the depth of Sogne Fjord, the longest and deepest fjord of Norway.(37)
        At the end of the Glacial period, however, the glaciated regions are known to have been mostly depressed somewhat below their present level. This change seems to be well accounted for by the vast weight of the ice-sheet itself, causing the land to sink finally beneath its load; and the subsequent rise of the land is an expression of the buoyancy of the earth's crust when it had been relieved by the disappearance of the ice. The preglacial elevation may well have produced a cool climate throughout the year, with abundant snowfall and resulting ice accumulation; and the subsidence of the burdened land would cause rapid melting of the ice upon its borders and thence backward progressively over its whole area.

        Table of courses of glacial striæ.--The directions of the currents of the ice-sheet are shown by its tracks, the furrows and striæ which bowlders and gravel frozen in the base of the moving ice engraved upon the bed rocks over which they passed. From these courses of movement of the ice, the areas of its thickest accumulation and consequent outflow are known. In some districts, also, changes in the outlines of the ice border and in its slopes and currents during its final retreat are indicated by deflected glacial striæ which run across the earlier courses. Occasionally two or more sets of striæ are found intersecting on the same rock surface, but more frequently the earlier and later sets are preserved on separate portions of the same or contiguous rock-outcrops.(42) The testimony of these records is so important concerning the barrier which held Lake Agassiz that a table is presented as an appendix of this volume, noting the courses of striæ which have been reported upon all the country from Hudson Bay, Lake Superior, and Minnesota westward and northward across the basin of this glacial lake.
        Converging lobes of the ice-sheet in Minnesota and Manitoba.--The southwestward striation in northeastern Minnesota and the southeastward striation in the central and southern part of that State belong to two convergently flowing lobes of the ice-sheet, partly corresponding to its portions which earlier inclosed the driftless area of southwestern Wisconsin and united in a continuous area of ice farther south. The central [p.130] line of the western of these ice-lobes coincided nearly with the Red and Minnesota rivers and the upper Des Moines, its southern end being near Des Moines, in central Iowa. This may be named the Minnesota lobe of the ice-sheet. Farther west the Dakota lobe stretched from the Souris basin and the region of Turtle Mountain south across the east half of North and South Dakota to Yankton, its central line being along the valley of the James or Dakota River. In Manitoba the glacial currents, passing to the Minnesota and Dakota ice-lobes, moved to the south-southeast and south, as noted at several localities on the Winnipeg River above Lac du Bonnet, on Lakes Winnipeg, Winnipegosis, Manitoba, and St. Martin, at Stonewall and Stony Mountain, and on the Assiniboine; and these currents are remarkably contrasted with the southwestward striæ of the contiguous region of the Lake of the Woods and the country extending thence east and north. These converging striæ in western and eastern Manitoba probably were engraved mostly during the recession of the glacial boundary, when Lake Agassiz was extended over the greater part of the Red River Valley. On the east this lake appears to have been bounded by a vast ice-lobe outflowing from the region of Lake Superior and James Bay southwest and south to the Lake of the Woods and Lake Itasca, representing the earlier convergent ice-lobes of the northeastern and of the western and southern portions of Minnesota, while on the west it was bounded by the representative of the Dakota ice-lobe, then outflowing from the region of Lake Manitoba and Riding Mountain southward to the terminal moraine of Pilot Knob, the north side of Devils Lake, Turtle Mountain, and the Tiger, Brandon, and Arrow hills.
        Transportation of bowlders.--Nearly everywhere the greater part of the drift is derived from formations not far distant, varying from a few miles to 25 or 50 miles away, in the direction from which the ice-sheet moved; but mingled with this material from comparatively near sources are other portions, both of the fine detritus and of the small and large rock masses, which have been transported longer distances, as the Archean bowlders of northern Montana and the upper Saskatchewan district, derived from the Archean belt east and north of Lake Winnipeg and about Reindeer Lake. [p.131] The least distance from the most western of these bowlders to the margin of the Archean belt is about 550 miles. Other bowlders of Archean origin which must have traveled nearly or quite as far occur in Kansas, Missouri, and Illinois, on the southwestern part of the drift-bearing area of the United States. The method of transportation of all these is believed by the writer to have been wholly by the slow currents of land ice.
        Dr. Robert Bell observes that the bowlders and pebbles of the drift on the west coast of Hudson Bay, near the mouth of the Churchill, and on the lower part of the Nelson, consist largely of rocks like those of the opposite eastern coast of Hudson Bay, which is 500 miles distant.(43) But the farthest known transportation of rock fragments in the drift, recorded in part by Dr. Bell, whose observations are supplemented by my own, is from James Bay southwest to North Dakota and Minnesota. The rock thus recognized is a "dark gray, granular, siliceous felsite or graywacke,    *    *    *    characterized by round spots, from the size of a pea to that of a cricket ball or larger, of a lighter color than the rest of the rock, which weather out into pits of the same form." It occurs in situ, as reported by Dr. Bell, on Long Island, off Cape Jones, on the east coast of Hudson Bay where it is narrowed to form James Bay, having there a Southwestward strike and probably continuing under the sea for some distance in that direction. He notes that the abundance of pebbles and bowlders of this rock is the most remarkable feature of the drift on the west coast of James Bay and along the Attawapishkat, Albany, and Kenogami rivers and that its fragments have been found by him as far west as Lonely Lake and southward to Lake Superior.(44) Farther to the southwest and south I have observed fragments of it, usually only a few inches but in some instances a foot or more in diameter, occurring very rarely in the drift in the northeastern part of North Dakota, where the largest piece ever found by me was about 30 miles south of the international boundary and 50 miles west of the Red River, and at numerous localities in Minnesota, where it extends at least as far south as Steele County, 75 miles south of St. Paul and 1,000 miles southwest of its outcrop north of James Bay.

        Till, also called bowlder-clay, constitutes the greater part of the entire sheet of superficial deposits, both within the area of Lake Agassiz and upon the adjoining country. It usually lies on the striated bed-rock, and upon large areas it reaches thence upward to the surface; but elsewhere this unmodified glacial drift is covered by modified drift, the stratified gravel, sand, and clay deposited by streams which flowed down from the ice-sheet during its melting, or by lacustrine and fluvial sediments. Fully half of the area of Lake Agassiz in Minnesota and North Dakota has a surface of till. In the part of this lake area examined by me in Manitoba its proportion is less, because much of this district is covered by the Assiniboine delta and its associated lacustrine beds. Extensive tracts of till, however, occupy the surface on the north and east portions of the Manitoba area, as north of Neepawa, on the east side of the Big Grass Marsh, from the south end of Lake Manitoba eastward by Shoal Lake [p.135] nearly to the Red River and Winnipeg and south to the Canadian Pacific Railway, from East Selkirk eastward along this railway, and 10 miles east of Emerson, where the flat plain of the Red River Valley is bordered by slightly higher land. Till also forms the surface of the terrace along the foot of the Pembina Mountain escarpment between the international boundary and Thornhill. Beneath the delta deposits of gravel and sand, and along the central portion of the Red River Valley, where the surface is commonly fine silt or clay, a sheet of till lies between these sediments and the bed-rock.
        The till is the direct deposit of the ice-sheet, as is shown by its consisting of clay, sand, gravel, and bowlders, mingled indiscriminately in an unstratified mass, without assortment or transportation by water. Very finely pulverized rock, forming a stiff, compact, unctuous clay, is its principal ingredient, whether at great depths or at the surface. It has a dark, bluish-gray color, except in its upper portion, which is yellowish to a depth that varies from 5 to 50 feet, but is most commonly between 15 and 30 feet. This difference in color is due to the influence of air and water upon the iron contained in this deposit, changing it in the upper part of the till from protoxide combinations to hydrous sesquioxide. Another important difference in the till is that its upper portion is commonly softer and easily dug, while below there is a sudden change to a hard and compact deposit, which must be picked and is far more expensive in excavating. The probable cause of this difference in hardness was the pressure of the vast weight of the ice-sheet upon the subglacial till, while the upper part of the till was contained in the ice and dropped loosely at its melting. Upon each side of Lake Agassiz the till has a moderately undulating and rolling surface. Within the area that was covered by this lake it has a much smoother and more even contour, and its upper portion, owing to its manner of deposition in this body of water, sometimes shows an imperfect stratification, with a scantier intermixture of bowlders and gravel. Yet even where it has distinct lamination it usually is more like till than like ordinary modified drift, and contains stones and gravel through its entire mass.
        [p.136] The chief characters of the englacial upper portion of the till, as compared with the subglacial lower portion, are its looser texture, its more plentiful and larger bowlders, the prevailingly angular and subangular shapes of its bowlders and smaller rock fragments, whereas they are mostly worn smooth by glaciation in the lower till, and the usually more gravelly and sandy and less clayey composition of the englacial till, owing to the washing away of much of its finer material by superglacial drainage. To these originally inherent characters we must add the very noticeable postglacial change of color of the upper till already mentioned. This change has generally extended through the englacial till, stopping at the more impervious subglacial deposit. Between the two there is also frequently a layer of subglacial stratified gravel and sand, from a few inches to several feet thick. The extremes of thickness of the englacial till appear to range from almost nothing or only a few feet for minima to 40 feet or more for its maxima near massive terminal moraines and where great currents of the ice-sheet converged.(46)
        Rock fragments and other drift inclosed in the ice at a considerable height above the ground were borne forward without attrition. The higher part of the englacial drift is thought by the present writer to have supplied most of the material forming the terminal moraines, which, therefore, have a remarkable profusion of bowlders and angular gravel. When the ice-sheet was finally melted, its inclosed bowlders were dropped, and they now lie frequently as conspicuous objects on both the lower and higher parts of the land. Scattered here and there in solitude on an expanse of prairie, or perched on the sides and tops of hills and mountains they at first suggest transportation and stranding by icebergs or floe ice.
        Bowlders and gravel from Archean and Paleozoic formations.--Bowlders are frequent or plentiful in the till throughout the area of Lake Agassiz, their abundance being nearly the same as in the least rocky parts of the till of New England, New York, and the country surrounding the Laurentian lakes. Their usual range in size extends up to a diameter of 4 or 5 feet; but in a few localities, especially in the course of morainic belts, they [p.137] were observed of all sizes up to 10 or 12 feet cube. Generally as large a proportion as 99 per cent of the bowlders exceeding 1 foot in diameter consists of Archean granite, gneiss, and schists, being derived from the Archean area on the northeast and north. With these are occasional limestone blocks, derived from the belt of Paleozoic limestones, constituting on the average perhaps nearly 1 per cent of the large rock fragments of the drift. The bedded and jointed character of the limestones has prevented their supplying many large bowlders in comparison with the more massive crystalline Archean rocks, while yet usually about half of the smaller cobbles and pebbles in the till and in gravel and sand deposits are from these Paleozoic limestones. Upon the Cretaceous area a considerable proportion of the gravel and cobbles is derived from the Fort Pierre shale, but this formation supplies no large blocks.
        Northeastern limit of limestone drift.--East of Lake Winnipeg and northeast of a line drawn from this lake southeastward by Lac du Bonnet on the Winnipeg River and across the Lake of the Woods to the west end of Rainy Lake and onward to Vermilion Lake, both bowlders and gravel of limestone are absent or exceedingly rare. This line probably marks the farthest extent ever attained by the glacial currents which moved south-southeast in the vicinity of Winnipeg and at Black Bear Island, near the Narrows of Lake Winnipeg, carrying debris from the limestone region of the Manitoba lakes.
        It is also very remarkable that the same line divides an area of very thin drift on its northeast side from the area of very thick drift which thence extends southwestward across all western Minnesota, the southern part of Lake Agassiz, and the region of the Sheyenne and James rivers to the Missouri Coteau.
        Localities of very abundant and large bowlders.--The following localities may be mentioned as having especially abundant bowlders: On the slope of the Pembina Mountain, in township 3, range 6, Manitoba, between Morden and Thornhill, very plentiful and large bowlders are spread upon an area of several square miles, as noted in the description of the Tintah beaches. The sides of Star Mound, Manitoba, especially those facing the north and northeast, are strewn with a multitude of bowlders, nearly all [p.138] granitic, of all sizes up to 5 feet in diameter or rarely larger. These were probably combed out of the ice-sheet in its passage over this hill. Comparatively few bowlders occur on the small flat area at its top. Pilot Mound, an equally prominent hill seen from this in looking northwest, is, like Star Mound, a knob of Cretaceous shale with thin covering of drift, but it has no such unusual profusion of bowlders on its slopes. Rock Lake, through which the Pembina flows, derives its name from the remarkable abundance of bowlders, mostly granitic, up to 6 feet or more in diameter, bordering its shores; and along a distance of 1 or 2 miles west from this lake the Pembina Valley is much encumbered with bowlders, which in some places are accumulated upon small morainic ridges and knolls.
        The largest bowlder observed within the area of Lake Agassiz south of the international boundary has given name to White Rock station, in the northeast corner of South Dakota, 11 miles north of Lake Traverse. This bowlder, lying 50 feet west of the railway, at a distance of about 25 rods north of the station, measures 18 by 12 feet, with a height of 5½ feet. It is a medium-grained, massive, flesh-colored granite, weathering to a whitish gray.
        Another bowlder of nearly equal size lies about 50 rods west of the Herman beach, in or near section 12, township 140, range 46, Minnesota, some 6 miles north of Muskoda. Its dimensions are 15 by 12 by 5 feet, and its top is 1,095 feet above the sea. It is gneiss, minutely porphyritic, with white feldspar crystals up to an eighth or a quarter of an inch long.
        A somewhat larger block, exceeding any other noted during my survey of Lake Agassiz and the adjoining region, lies in the northwest quarter of section 9, township 1, range 4 east, Manitoba, on the low ridge 10 miles east of Emerson. It is dark-gray granitoid gneiss 22 feet long, 8 to 14 feet wide, and projecting 2 to 5 feet above the surface. Among the other plentiful bowlders of that vicinity none was seen exceeding 7 or 8 feet in dimensions. Like many of the smaller bowlders throughout this prairie region, this block is surrounded by a slight depression 1 to 3 feet below the adjoining ground; and a careful examination shows that some of its projecting corners and edges are smoothly polished. These depressions [p.139] were formed by the trampling and pawing of buffaloes in rubbing on the bowlders, which were thereby sometimes worn and polished as perfectly as could be done by art.(47)

        Exploration of the terminal moraines in the northeast edge of South Dakota, accumulated on the west margin of the Minnesota lobe of the ice-sheet, northward to the Head of the Coteau des Prairies, was included in the work of the writer during 1880 for the Minnesota Geological Survey. The three outer moraines are typically developed and distinctly separated in that portion of their course, as well as through the adjoining southwest part of Minnesota; and from their description for this district(48) they have been denominated by Professor Chamberlin the Altamont, Gary, and Antelope moraines. Besides these, nine others, to a total of twelve in all (as shown on Pls. III and XVII) lying along a large part of their extent in successive order from south to north, and apparently marking consecutive stages in a wavering recession of the ice-sheet, are recognized in Minnesota and receive names in the annual and final reports of the State survey from localities where they are notably prominent or distinct. In western Minnesota they seem to constitute a simple series, each in order advancing from south to north and northeast being of somewhat later formation than the one preceding; but in the central and eastern portions of the State, from the Leaf Hills southeast to Minneapolis and St. Paul, and in their course eastward into northern Wisconsin, consecutive moraines are merged together, and even the later are found overlapping the earlier in the series.
        [p.140] Farther to the west the terminal moraines of South Dakota, and of North Dakota west of the James River and north to the Northern Pacific Railroad, have been mapped by Prof. J. E. Todd for the United States Geological Survey. To give a more complete view of these moraines through the region of Lake Agassiz, notes based on his map and several published papers are included in the present monograph. My observations of the outer moraines on the Coteau des Prairies from Iowa to the north east part of South Dakota and on the Coteau du Missouri in the northwest part of North Dakota are thus connected and correlated through Professor Todd's exploration of the successive boundaries of the intervening Dakota lobe of the ice-sheet.
        The moraines of the Tiger, Brandon, and Arrow hills in Manitoba, mapped by the writer in 1887, seem probably contemporaneous with the most northern morainic belts in Minnesota, and it is evident that in both districts they belong to the time of the uppermost in the series of the Herman beaches, the first and highest of the well-marked shore-lines of this glacial lake. Though these moraines of southwestern Manitoba and northern Minnesota are the tenth and eleventh, the latest formed, in the series of moraines here described, they mark a much earlier stage of the glacial retreat than the terminal moraine observed by Dr. Robert Bell(49) as crossed by the Hill, Nelson, and Churchill rivers, about midway between Lake Winnipeg and Hudson Bay, which may well belong to the time of the Campbell beaches or later, while the line of morainic islands reported by Mr. A. P. Low(50) along an extent of about 200 miles from south to north and north-northwest in James Bay was certainly formed after Lake Agassiz began to outflow northeastward, perhaps after it was lowered to its present representative, Lake Winnipeg.
        The twelve moraines of Minnesota are doubtless correlative with the similarly numerous moraines, partly simple, with approximately parallel courses, and partly complicated in their arrangement by interblending and overlapping, which have been recently traced by Mr. Frank Leverett, passing southward from Wisconsin along the east side of the driftless area [p.141] and running in great loops, the boundaries of lobes of the ice-sheet, across Illinois, Indiana, and Ohio. These are embraced within a strip of country of similar width with that of Minnesota, Iowa, and the Dakotas, which, though 200 or 300 miles wide, is yet only a minor part of the drift-covered area of this continent. For the interior of this area the observations of Bell and Low give us good assurance that nearly an equal profusion of marginal moraines, recording step by step the wavering departure of the ice-sheet, await exploration in all the region northeast and north from Minnesota and the Great Lakes to Hudson Bay, and from northern Pennsylvania, New Jersey, Long Island, Marthas Vineyard, Nantucket, and Cape Cod to the Laurentide highlands, north of Montreal and Quebec.

        When the North American ice-sheet attained its greatest area, and during its later Iowan and Wisconsin stages, its southern portion, from Lake Erie to the Missouri River, consisted of vast lobes, one of which, at the beginning of the Wisconsin stage of accumulation of moraines, reached from central and western Minnesota south to central Iowa. This Minnesota lobe then ended near Des Moines, and its margin was marked by the first or Altamont moraine, lying upon the Coteau des Prairies and in part forming its crest. When the second or Gary moraine was formed, it terminated on the south at Mineral Ridge, in Boone County, Iowa. At the time of the third or Antelope moraine it had farther retreated to Forest City and Pilot Mound, in Hancock County, Iowa. The fourth or Kiester moraine was formed when the southern extremity of the ice-lobe had retreated across the south line of Minnesota and halted a few miles from it in Freeborn and Faribault counties. The fifth or Elysian moraine, crossing southern Lesueur County, Minn., marks the next halting place of the ice. At the time of formation of the fifth moraine the south end of the ice-lobe had been melted back 180 miles from its earlier extent, shown by the Altamont moraine, and its southwest side, which at first rested on the Coteau des Prairies, had retired 30 to 50 miles to the east side of Big Stone Lake and the east part of Yellow Medicine County.
        During its next stage of retreat the Minnesota ice-lobe was melted away from the whole of Lesueur County, and its southeast extremity was [p.142] withdrawn to Waconia, in Carver County, where it again halted, forming its sixth or Waconia moraine. This records the position of the front of the ice-sheet immediately before its continued recession gave place for the beginning of Lake Agassiz. It will therefore be described somewhat in detail along its course adjacent to this glacial lake.

        Between the fifth and sixth moraines the southeast end of the Minnesota ice-lobe retreated from Elysian to Waconia, a distance of about 40 miles from south to north, uncovering the lower portion of the Minnesota Valley and finally draining the glacial lake of the Blue Earth and Minnesota basins, which bad outflowed southward in its highest, early stages by Union Slough in Iowa to the East Des Moines River, and later to the east by the Cannon River. The advance of the east side of this ice-lobe at the time of the Kiester and Elysian moraines beyond its previous limit, by an incursion from Wright County to Chisago County and the edge of Wisconsin, had been followed by a withdrawal from the greater part of the area thus acquired, until at the time of the sixth moraine the most eastern portion of the ice margin was accumulating the prominent drift hills close east and north of Elk River, in Sherburne County. The glacial recession there from east to west and southwest between the Elysian and Waconia moraines appears to have been also about 40 miles. A long indentation of the ice-sheet, between its Minnesota and Lake Superior lobes, was melted back during the same interval, the apex of this reentrant angle being carried from southeastern Stearns County 50 miles west to Lake Whipple and Glenwood, in Pope County. But in some places the ice border north and east of Waconia had probably retreated no more than a few miles, and on the southwest side of the Minnesota lobe, in Redwood, Yellow Medicine, Chippewa, Swift, and Big Stone counties, there was only slight recession of the ice, and the Elysian and Waconia moraines seem to be blended, though they form together only inconspicuous marginal deposits.(51)
        [p.143] After passing northwest across Lake Traverse and the Head of the Coteau des Prairies, the Waconia moraine appears to be merged with the two preceding Elysian and Kiester moraines in the conspicuous belt of drift hills that extends from the line dividing South and North Dakota northward between Straubville and Crescent Hill, between Nicholson and Oakes, and along the east side of Bear Creek to the southeast part of township 135, range 59. Thence it turns west and northwest a few miles, beyond which it runs again northward through the west part of township 136, range 59, the most northeastern of Lamoure County, where it forms a narrow belt of knolls and hills, rising 40 to 60 feet above the nearly level plain on each side.
        In Barnes County, running 42 miles from south to north, this moraine is distinct and well developed, being divided from the next earlier and later moraines of the series by smoothly undulating and in large part nearly level belts of till, which vary in width from 2 to 3 miles to a maximum of about 8 miles on the west and 12 miles on the east, the separation from the seventh or Dovre moraine being on the average the wider of the two. The Waconia moraine enters Barnes County at the middle of the south side of township 137, range 59, and curving northeastward passes through sections 34, 26, and 24, in a belt of typical knolls and hills 25 to 75 feet high, very rough in their outlines and profusely strewn with bowlders, to the east line of this township, where the apex of a reentrant angle of this belt almost touches the Dovre moraine, which rises to equal or greater prominence in the adjoining township. A plain of overwashed gravel and sand, deposited just outside the ice border in the indentation of the Waconia moraine, is crossed by the road on the west line of sections 23, 14, and 11, township 137, range 59, thinly covering the underlying till, which is occasionally exposed, with its projecting bowlders, in slight depressions. Turning by a right angle, this moraine runs northwestward through sections 13, 11, and 3, rising 25 to 50 or 75 feet above the general level. In the northeast quarter of section 3, a lake bed, wholly dry in August, 1889, lies at the northeast base of these hills, and a belt about a half mile wide, next to the north, is moderately rolling till, beyond which a second belt of morainic hills similar to the foregoing and parallel with it, a fourth to a third of a [p.144] mile wide, runs northwestward through the southwest quarter of section 35, township 138, range 59. This twofold condition of the Waconia moraine is observable along a distance of 3 or 4 miles. In sections 32 and 29 the moraine turns to the north and continues through the middle of the west half of this township in hills and short south-to-north ridges 25 to 60 feet above the adjoining nearly level inter-morainic surface of till. The width of the hilly belt here and north-northeastward through sections 33, 28, 21, and 22, township 139, range 59, is about a mile; but for the next 4 miles north it expands to a width of 2 or 3 miles and is conspicuously displayed in steep hills 50 to 150 feet high, to the apex of another reentrant angle in sections 33 and 34, township 140, range 59, 2 to 3 miles south of Hobart. Thence the moraine again turns by a right angle, taking a westward course, parallel with the Northern Pacific Railroad and about 2 miles south of it, through the south edge of township 140, range 60, where its hills cover an average width of 1 mile and rise 50 to 75 feet above the smoothly undulating expanse of till on each side.
        The Northern Pacific Railroad crosses the Waconia moraine close west of Eckelson, where it has a width of about a mile, marked by a rolling and partly knolly contour, with elevations 25 to 50 feet above the hollows. Within a mile west of the morainic belt the railroad crosses an ancient watercourse, a fourth to a half of a mile wide, extending from north to south, occupied by a lake on the north side of the railroad and by a marsh on the south. Both are bordered by bluffs which rise steeply about 40 feet to the general level. This lake is one of a series that that extends 6 miles south-southwest, occupying portions of this old watercourse, but intervening portions and its farther continuation southward are mostly filled with the glacial drift. Lake Eckelson and a series of smaller lakes, reaching 5 miles south to Walker Lake, mark a second and parallel watercourse, similarly enveloped in other portions by the general drift sheet. The north end of a third series or chain of lakes of the same kind, about 6 miles long, is crossed by the railroad a mile east of Sanborn; and a fourth is indicated by a long lake extending south from the railroad near Hobart. The first of these chains of lakes lies wholly outside the Waconia moraine, but the others are crossed by the east-to-west portion of [p.145] this moraine south of Hobart, Sanborn, and Eckelson. They probably have had a history like that of the Spiritwood series of lakes, several miles farther northwest, and of the similar chains of lakes extending from north to south in Martin County, Minn., which are believed to occupy the unfilled parts of preglacial or perhaps interglacial channels of drainage.(52)
        Beyond Eckelson this moraine extends north-northwestward as a rolling and knolly belt, inconspicuous in any distant view, to the southeast part of township 142, range 61, about 10 miles east of Spiritwood Lake. There it is moderately rolling and occasionally hilly, a third to a half of a mile wide, with more bowlders than the adjoining lower and only slightly undulating surface, both being till. Curving north-northeastward, it passes through sections 14, 12, and 1 of this township, being well marked in the east part of section 1 as a belt of low morainic knolls, a quarter of a mile wide. Onward through the next 6 miles northeast to the south part of section 11, township 143, range 60, this moraine forms a belt, a half mile to 1 mile wide, of knolls and scattered steep hills, partly composed of kame gravel and sand, with few bowlders, rising 40 to 75 feet above the general level, as conspicuously seen from Dazey. In section 11 its course seems again to be deflected nearly by a right angle, passing thence northwest and north through township 144, range 60, to the vicinity of the Helena farm in section 29, township 145, range 60; but along this distance it is marked only by a rolling contour, with no prominent elevations.
        Continuing northward through Griggs County, the Waconia moraine becomes gradually more knolly and hilly, with increasing proportion of bowlders, to its magnificent development in townships 147 and 148, range 60, passing close east of Lake Sibley and west of the beautiful Lakes Addie and Jessie, to Red Willow Lake. Along the distance of 10 miles adjoining these lakes the irregularly piled masses of morainic drift, strewn with many bowlders, rise 100 to 200 feet above the lakes, giving a measure of boldness and even grandeur to the scenery, such as is rare in this plain and prairie region. West of Lake Sibley the fine agricultural tract of Blooming Prairie has a nearly level surface upon a width of about 10 miles, including [p.146] township 146 and the south part of township 147, range 61, beyond which, northward, this smooth area of lowland, dividing the Waconia and Elysian moraines, narrows into a belt only about a mile wide in the southeast part of township 148, range 61, called Colemans Valley. At the north end of this valley, on the west side of Red Willow Lake, these two moraines meet, and thence pass in a united morainic belt west and northwest to the Washington Lakes, the Sheyenne River, and the Indian reservation south of Devils Lake.

        The seventh or Dovre moraine marks a pause in the glacial recession when the southeast end of the Minnesota ice-lobe rested on Kandiyohi County. At this time Lake Agassiz had begun to exist, the south end of the Red River Valley having been uncovered from the ice. Probably nearly all of the southern half of Minnesota was then divested of its ice mantle, while nearly all of the northern half was still ice-covered, the glacial boundary across the State passing in an approximately east-to-west course.
        By its next recessions the ice border was withdrawn to the eighth or Fergus Falls moraine and the ninth or Leaf Hills moraine. These are partly merged together in the prominent accumulations of the Leaf Hills, which reach in a semicircle from Fergus Falls to the southeast, east, and northeast, a distance of 50 miles, marking the southern limits of this ice-lobe when it terminated half way between the south and north borders of Minnesota. During the formation of the tenth or Itasca moraine, and of the eleventh or Mesabi moraine, the ice border crossed the lake region at the head of the Mississippi. Farther north the twelfth or Vermilion moraine, discovered and mapped by the present writer in 1893 during work for the Minnesota Geological Survey, passes by the south side of Vermilion, Pelican, and Net lakes. Later moraines, formed at times of halt or readvance, interrupting the recession of the ice-sheet between northern Minnesota and Hudson Bay, have been observed in only a few places; but I believe that they exist and will be continuously mapped when the glacial drift of that wooded and very scantily inhabited region shall be [p.147] fully explored. The many beaches of Lake Agassiz, all showing an ascent northward when compared with the level of the present time, but with this ascent gradually decreased during the successive stages of the lake, probably find their explanation in the manner of retreat of the ice in Canada, interrupted there, as farther south, by pauses and the formation of moraines.
        The following are notes of the five moraines already mapped which cross the expanse of Lake Agassiz, being conspicuous upon each side of this lake, but faintly developed or lost on the lacustrine area:

        The Dovre moraine is prominent in Stearns, Douglas, Pope, and Kandiyohi counties, Minn. Its distinctive name is taken from its hills in Dovre, Kandiyohi County, to which the southeast extremity of the Minnesota lobe of the ice-sheet had been withdrawn west-northwestward about 70 miles from Waconia during the interval between the sixth and seventh moraines. In Pope and Douglas counties an area about 25 miles wide, from Glenwood north to Miltona and Spruce Hill townships, was probably uncovered by this glacial recession. But considerable portions of the ice border in its general course at this time from east to west across central Minnesota had receded only a few miles between these moraines. Indeed, they seem to be merged together north of Richmond, in Stearns County, and from Barsness, in Pope County, to Mount Tom, about 9 miles north-northeast of the Dovre Hills. Again, in Big Stone County, a single belt of somewhat rolling till, 5 to 8 miles wide, seems representative of the Elysian, Waconia, and Dovre moraines combined. In the wooded country east from Little Falls, Minn., to the sources of the St. Croix River in Wisconsin, this moraine has not been definitely traced.(53)
        Crossing Richland and Sargent counties, in the southeast corner of North Dakota, the Dovre moraine is well developed in knolls, hills, and short ridges of till, covering a belt from a half mile to 2 miles in width, with abundant bowlders and characteristically rough contour. On the southwest side of Taylor Lake, near Hankinson railway station, these rough [p.148] drift hills rise to heights 50 to 150 feet above this lake, or 1,100 to 1,200 feet above the sea. Thence a bowlder-strewn, rolling, and knolly surface, with numerous small lakes, extends west along the south side of the Great Northern Railway to prominent morainic hills, 50 to 100 feet in height, which extend about 7 miles from east to west close south of Geneseo and Cayuga. In the east part of township 130, range 54, this moraine curves to the north, passing about a mile west of Cayuga and Ransom, and north-northeastward through the northwest part of township 131, range 53. From near the northwest corner of this township it runs to the northwest diagonally across township 132, range 54, passing close west of Milnor, where its knolls and hills are 20 to 50 feet high, with abundant bowlders. The same northwestward course is continued through Ransom County, passing by Lisbon as a belt of knolls and hillocks crowning the southwest bluff of the Sheyenne Valley, to the conspicuous morainic hills (including "Bears Den Hillock") in the vicinity of Fort Ransom, rising 50 to 100 feet above the general level and 250 to 300 feet above the river.
        At the time of accumulation of these hills the ice-sheet had retreated a few miles north from the Head of the Coteau des Prairies and 10 to 25 miles eastward from the moraine referred to the Kiester, Elysian, and Waconia stages, near Straubville and Nicholson and along Bear Creek. The northeastwardly sloping surface of the greater part of Sargent County was covered by a glacial lake, whose silt beds, confluent southwestward with those of Lake Dakota, are about 1,300 feet above the sea from Sargent and Straubville southward into South Dakota, to Newark, Kidder, and Burch, but decline eastward to about 1,250 feet on the south side of Silver and Sprague lakes. The surface of this glacial lake was 1,300 feet, or probably at first 1,310 feet, above the present sea-level, its outflow being southwestward across the bed of Lake Dakota to the James River. The channel of this outlet is doubtless distinctly traceable. On the north this lake received a large inflowing stream, the representative of the present Sheyenne River, which brought the waters that were discharged from the border of the receding ice-sheet and from the drainage of a considerable belt of the adjoining land along all the distance northward to the vicinity of Devils Lake and thence northwestward to the head of the Sheyenne. [p.149] Even farther northwest, the glacial Lake Souris outflowed by this stream as previously at its beginning it had found outlet during the time of the Elysian and Waconia moraines into the upper part of the James River, flowing through Arrow Wood and Jim lakes to Lake Dakota, so long as that lake existed.
        This great affluent, which may be called the glacial Sheyenne River, is marked by a flat or in part moderately undulating belt of stratified gravel and sand, extending from the central part of the Fort Ransom military reservation southward by Marshall and Nicholson. It includes a width of 1½ miles to the west and an equal distance to the east of Marshall, where it is bounded on each side by higher tracts of smooth till. Its height above the sea at Marshall is 1,343 feet, and at Nicholson, where it widens into the glacial lake of Sargent County, 1,309 feet. Two comparatively small channels, probably occupied by the stream in winter when glacial melting was at its minimum, were seen near the west side, on the wide alluvial belt, about 1 1/3 miles and 1 mile west of Marshall, each having a width of an eighth of a mile and a depth of about 15 feet. One of these channels, or the two interlocking and here and there separated by islands, is commonly known as the Big Slough, and has an extent of many miles from north to south.
        During the recession of the ice from the compound moraine on the west line of Sargent and Ransom counties to the Dovre moraine, before described, the glacial lake grew as fast as the land became uncovered, extending gradually east around the northern base of the Coteau des Prairies to Skunk Lake (recently called Lake Tewaukon), northeast over the smoothly undulating surface of till, very abundantly sprinkled with bowlders, about Forman and Lake Kandiota, to the Stormy Lakes and adjacent moraine near Milnor, and northward along the moraine and ice front into Ransom County. Its depth at Forman was 50 feet; at Perry, 6 miles east, nearly 100 feet; and farther east and northeast, beside the Dovre moraine, about 150 feet, if it continued tributary to the James River through the whole time of this glacial retreat.
        It is more probable, however, that when the recession of the ice uncovered Lake Tewaukon and the country eastward, an outlet was found in [p.150] that direction along the front of the ice-sheet and the Dovre moraine, flowing into Lake Agassiz in the northwest part of township 129, range 49. The belt of stratified gravel and sand, 1 to 2 miles wide, which there and for a distance of 15 miles southeastward constitutes the border of this lacustrine area,(54) seems to have been deposited by this great river, while the ice-sheet lay on its northeast side, terminating where the edge of this level or somewhat undulating tract descends like a terrace and is bordered by the slightly lower Herman and Norcross beaches. Since the deposition of these stratified beds, the River Warren, outflowing from Lake Agassiz, has eroded and carried away their continuation across an extent of several miles southeast to a remnant of the same gravel and sand which, with underlying till, forms the plateau cut by the Fargo and Southern (Chicago, Milwaukee and St. Paul) Railway in the southeast part of township 128, range 47, Traverse County, Minn., about half way between White Rock and Wheaton. The outline of the ice margin along the extreme southwestern edge of this glacial lake at the time of its accumulating the Dovre moraine and forming the northeast bank of the glacial Sheyenne River at its entrance to the area of Lake Agassiz may therefore be somewhat confidently traced around the little plateau between the Bois des Sioux and Mustinka rivers and southward by Wheaton to the rolling land about the Tokua Lakes at Graceville. The prominent morainic hills west of Taylor Lake, according to this interpretation of our observations, were massed in an angle of the ice margin, the usual place for plentiful drift accumulations.
        Windy or Airy Mound, on the northern end or Head of the Coteau des Prairies, close south of the line between North and South Dakota, is a slight elevation above the general surface of this drift-covered Cretaceous ridge. Its height is about 1,950 feet above the sea, and by estimate 100 feet lower than the crest of this ridge a few miles farther south. Thence gentle slopes descend 750 feet to Sprague and Skunk lakes, near the northern base of this highland; and the whole view east, north, and west from Windy Mound sweeps over a broad, nearly flat expanse of till and lacustrine silt, ranging in altitude from 960 feet at Wahpeton to 1,250 feet at Forman and 1,300 feet on the area of Lake Dakota adjoining the James [p.151] River. After extending as a continuous massive ridge nearly 200 miles from south-southeast to north-northwest through southwestern Minnesota and the northeast corner of South Dakota, with an elevation increasing northward from 1,600 to 2,050 feet above the sea, the Coteau des Prairies is thus terminated, and along the next 175 miles northward to the south end of the Pembina Mountain escarpment no conspicuous rise of the surface is observable from a great distance on the west side of Lake Agassiz and the flat, low plain of the Red River Valley. There is, however, a slow ascent of several hundred feet from this lacustrine area west to the plain-like expanse of Cretaceous shales and overlying drift, which rises northward from 1,200 to 1,300 feet above the sea in Sargent County to about 1,500 feet in the region surrounding Devils Lake.
        Into this plain the Sheyenne and James rivers have cut narrow and trough-like channels or valleys that vary from a third or a half of a mile to commonly 1 mile and rarely 2 miles in width. These channels, like the narrow morainic belts of knolls and low hills, are thus minor features of the general topography. The Sheyenne channel or valley is 100 to 200 feet deep, mainly cut in the Cretaceous shales for its lower half or more, though the faces of the bluffs are usually covered by a talus of drift, while the James Valley, ranging from 75 to 125 feet in depth, is mostly eroded in the drift sheet, there thicker than along the Sheyenne. From the vicinity of Valley City northward by Cooperstown to Devils Lake, Langdon, and a large part of southwestern Manitoba, stretching west from the crest of the Pembina Mountain, the depth of the drift is only from 10 to 50 feet. Over extensive tracts of Griggs and Cavalier counties it varies from 10 to 30 feet. Its average northward on this belt is small, not probably more than as 1 to 4 or 6 in comparison with its thickness in the Red River Valley, throughout western and southwestern Minnesota and on most parts of the great Cretaceous ridge of the Coteau des Prairies.
        Between the Dovre moraine and the compound moraine next west the general level of the northwestern part of Sargent County and of southwestern Ransom County is diversified by three massive swells or hills, which, like the surrounding nearly flat country, have a smooth surface of till. One of these, rising 75 to 100 feet and extending 2 or 3 miles from south [p.152] to north, is in the west part of township 131, range 57, close west of Harlem. Another, also trending with the meridian, lies 12 to 15 miles farther north, between Marshall and Elliott, above which railway stations it rises about 60 feet. The third is White Stone Hill, which extends about 4 miles from east to west in the north part of township 132, range 56, having a somewhat crescentric and oval form, convex to the south, with a height of about 150 feet. None of these elevations consists of morainic drift, but they seem to be due to the prominence of the underlying shale, which, however, has no outcrops, the thickness and character of the drift being nearly the same as on the surrounding intermorainic area.
        Multitudes of bowlders, mostly Archean gneiss and granite, seldom exceeding 3 feet in diameter, are scattered on the gently undulating expanse of till north of the Head of the Coteau. The bowlder-strewn tract extends west to Belle Plaine and 2 or 3 miles west of Forman, and northward through townships 130 and 131, in ranges 54 and 55, occupying an area of about 150 square miles. The bowlders are fully fifty times more plentiful than their average numbers on similarly smooth areas of till in western Minnesota and North Dakota, their proportion being nearly the same as on the morainic belts of this region. They may belong to morainic drift which has been smoothed by a subsequent advance of the ice over it, having been, perhaps, deposited during the time of the Kiester or Elysian moraines, and covered by a glacial advance at the time of the Waconia moraine; or they may have been gathered in unusual numbers in the englacial drift of this part of the ice-sheet, because of convergent currents from the east and west, together with the influence of the highland so near on the south. Further observations seem needed for determining satisfactorily the reasons for their abundance. They lie upon the central and eastern part of the lacustrine area of Sargent County, unless that lake became almost wholly drained away eastward, as is probable, before this surface was uncovered from the ice.
        The Dovre moraine extends north from its high hills in the Fort Ransom Reservation and forms a belt of lower rolling and knolly till, with plentiful bowlders, through sections 34, 27, and 22, township 136, range 58. On each side it is bounded by flat tracts, about 50 feet lower than the [p.153] highest points of the moraine and 200 feet above the Sheyenne, that on the west being till and that on the east stratified gravel and sand 1 to 2 miles wide, dividing the moraine from the trough-like Sheyenne channel or valley. From near the center of this township the morainic belt, continuing with inconspicuous development, passes northeastward across the Sheyenne, and at the northeast corner of this township and in section 6 of that next east the drift, heaped in a reentrant angle of the ice border, forms Standing Rock Hill, rising 100 feet above the general level and about 300 feet above the river. Thence this moraine turns back by a right angle to the west and northwest, recrossing the Sheyenne and rising in hills 50 to 100 feet high, through sections 34, 27, 28, 21, 20, and 17, township 137, range 58, approaching within a mile of the equally prominent hills of the Waconia moraine at its reentrant angle in the next township on the west.
        Continuing thence to the north, the Dovre moraine is represented by a narrow series of knolls, 20 to 30 or 40 feet above the general level, through township 138, range 58, veering slightly to the east, and closely skirting the west side of the Sheyenne Valley across the northern half of the township. Its drift covers the east slope and top, and its bowlders are strewn abundantly on the west slope of a hill of the Fort Pierre shale which rises nearly on the line between sections 32 and 33, township 139, range 58, to a height 50 feet above the average of the adjoining country, or about 225 feet above the river. Thence the moraine appears to turn northeastward and to lie concealed across a distance of about 4 miles beneath the high flood-plain of gravel and sand adjoining the Sheyenne, out of which the apex of a sharply reentrant angle projects in two typically morainic hills about 40 feet high in or near section 12 of this township, some 2 miles east of the river. Returning thence west-southwestward to the prominent and massive hill in the north part of section 19 of this township 139, range 58, 2 miles west of the river, abundant bowlders, apparently marking the course of the ice front at the time of the Dovre moraine, are strewn over the east, northeast, and southwest sides of this hill, which, however, is principally a rounded projection or boss of the Fort Pierre shale, seen in the slight ravines of its east side to 40 feet below its top. The higher part of its southwesterly sloping but somewhat plateau-like top, less encumbered [p.154] with bowlders than its sides, is about 125 feet above the general level, or 300 feet above the Sheyenne. In the view from this hill I traced faint indications of the course of the Dovre moraine in a curve passing northwest, north, and northeast to the similarly prominent hill 2 miles west-northwest of Valley City. The highest hills of the moraine in this distance of 7 miles rise about 60 feet above the general level in sections 11 and 2, township 139, range 59. All the surface from this loop eastward 3 or 4 miles to the Sheyenne Valley, is a very smooth, fertile tract, apparently the flood-plain of the river when it flowed about 175 feet above its present level, at the time of the retreat of the ice-sheet from this moraine.
        In section 18, township 140, range 58, 1½ miles north of the Northern Pacific Railroad and slightly farther northwest of Valley City, a prominent tract of morainic drift, probably owing part of its height to an underlying hummock of the Cretaceous shale, rises 100 feet above the old gravel and sand flood-plain of the Sheyenne on the east or about 300 feet above the river. The surface of this elevation is abundantly strewn with bowlders and is very irregularly broken by ravines, hillocks, and small ridges, trending from south to north to its highest point and thence trending toward the west-northwest, indicating that there was here another reentrant angle of the ice margin. Looking across the country west-northwestward, I observed low knolls and ridges of this moraine, scarcely above the general level, extending at least 2 or 3 miles; but no prominent hills are visible in this direction. Two to 3 miles north of Hobart this moraine curves northward, and passes as a narrow belt of knolly drift north and northeast through the east half of township 141, range 59, and across the Sheyenne to the northeast part of township 142, range 58, where it becomes partially merged with the Fergus Falls moraine. Thence turning back by a right angle, it recrosses the Sheyenne about a mile above the mouth of Bald Hill Creek and extends northwestward along the east side of this creek. Its most prominent portion, called Bald Hill, lies 5 miles east of Dazey and extends along a distance of 2 miles or more from southeast to northwest, rising some 300 feet above the Sheyenne or fully 100 feet above the general level.
        Through Griggs County the Dovre moraine is very well developed and forms especially conspicuous hills west and north of Cooperstown, [p.155] rising 75 to 150 feet above the adjoining country. From Bald Hill it takes a very straight course a little west of north for 15 miles to the center of township 146, range 59, 2 to 3 miles west of Cooperstown, having through this distance an average width of a mile and consisting of many knolls and small hills of till, with abundant bowlders, rising 25 to 50 feet above the nearly level surface on each side. Toward the west this belt is bordered along much of its extent by an overwashed plain of gravel and sand, about a mile wide, descending by a very gentle slope away from the moraine, which thus has a very definite boundary; but on the east there is a gradual change through a decreasingly knolly and rolling contour to the slightly undulating expanse of intermorainic till, with only few bowlders which stretches 4 to 6 miles east to the Sheyenne Valley and 20 miles north between the moraine and the Sheyenne from Bald Hill to the scattered hills of a reentrant angle of this Dovre moraine 4 to 6 miles north of Cooperstown.
        About 5 miles farther north another angle of the moraine is marked by the conspicuous hill called Butte Mashue, from the name of an Indian who was buried in a mound on its summit. This hill, situated in the east half of section 35, township 148, range 59, rises 150 or 175 feet above the general level east and north, or nearly 350 feet above the Sheyenne River, which is only 1 mile distant at the northeast. It is a typical morainic drift hill of small area irregularly knolly contour, and very abundant bowlders. Its diameter of base is only a third to a half of a mile. On the township line south of this section 35 there is a hollow a quarter of a mile wide, through which a road runs from east to west, nearly 150 feet below the top of the Butte Mashue. Immediately to the south, in the north edge of section 2, a very rocky and typically morainic north-to-south ridge rises about 125 feet; and thence a series of hills and short morainic ridges, 75 to 125 feet high, extends south to the irregularly scattered hills of similar heights which occupy most of the area between Cooperstown and Clear Lake.
        Unusual numbers of limestone bowlders were observed in this vicinity on the morainic hills and adjacent to them, including many 6 to 8 feet in diameter, one mass 10 by 15 feet in dimensions, and another, seen close east of the south-to-north road about a mile south of the Butte Mashue, [p.156] measuring 20 by 30 feet and projecting 1 to 2 feet above the surface, and shown by digging to be more than 3 feet thick. All these are grayish-yellow, very compact magnesian limestone, similar to that which is found forming bowlders of equal abundance near Audubon and the White Earth Agency in Becker County, Minn., and again on the massive morainic hills near Fort Totten, on the south side of Devils Lake. In these localities the limestone bowlders sometimes occur in equal or even greater numbers than those of the Archean rocks; but generally throughout this region the proportion of such limestone masses in the drift is very small, averaging probably not more than a hundredth part of the large bowlders, though the portions of the gravel of the drift supplied respectively by the limestone and the Archean rocks are commonly about equal. They are like the outcrops of magnesian limestone in the neighborhood of Winnipeg, in Manitoba, 165 to 180 miles distant to the north, which are its nearest natural exposures.
        West of Butte Mashue the Dovre moraine is represented by knolls and hills 25 to 75 feet high, which continue to the prominently rolling land surrounding Norway Lake, in the northeast corner of township 147, range 60. Next this moraine extends northward in an inconspicuous belt of knolly and rolling till to low hills in or near section 23, township 149, range 60; and thence, turning to the west-northwest, it holds this general course through a distance of 30 miles to the Devils Heart Hill. It crosses the Sheyenne River in the southeast part of township 150, range 61, and forms a conspicuous belt of hills, 100 to 150 feet high, along the south side of township 151, range 63, 4 to 5 miles south of Free Peoples Lake. A lower but well-marked series of morainic hills and knolls, mostly 40 to 60 feet high, with very abundant bowlders, curves thence northwest and north, passing through the center of township 151, range 64, to the Devils Heart. Again, only a few rods north of the base of that hill, a typically irregular belt of very rocky morainic knolls, 30 to 50 feet high, occupying a width of only 20 or 30 rods, trends from east-southeast to west-northwest, and forms the frontal line of the chief morainic tract bordering the south side of Devils Lake (Pl. XVIII). The Dovre moraine south of the Devils Heart Hill would thus join a larger morainic tract on the north, which [p.157] appears referable to the Fergus Falls and Leaf Hills moraines. At this intersection there is heaped the largest and most remarkable kame that has ever come under my observation, known by its aboriginal appellation as the Devils Heart. This mound of gravel and sand appears to have been deposited where a glacial river descended from the convergent slopes of the ice-sheet to the open land contemporaneously with the accumulation of the Dovre moraine.
        Devils Heart Hill rises in steep slopes of 20° to 30°, being thus steep on all sides excepting the south, where the otherwise nearly round-topped, conical hill is somewhat drawn out into a narrow, more slowly descending ridge. It consists of gravel and sand, mostly not showing pebbles on the surface larger than 1½ inches in diameter. A few bowlders however, a score or more in all, are seen on the sides of this hill to its top, and one a foot long (the only one seen at the crest) is embedded in the gravel a rod south of the highest point and less than 1 foot lower. The gravel is of threefold origin, being derived in about equal proportions from granitic and gneissic Archean rocks, from the Silurian limestones, and from the Cretaceous shales. Nearly all of the bowlders seen on the hill are granite or gneiss, but two or three on its west side are limestone. It is situated in section 4, township 151, range 64, about a mile southwest from the head of Donahues Bay of Devils Lake. Its height above its base is about 175 feet, and above Devils Lake, according to Nicollet's barometric determination, 290 feet, which appears to be about 15 feet higher than Sullys Hill, the culminating point of the very prominent hills of morainic till south of Devils Lake. From the altitude of the lake, 1,430 to 1,434 feet above the sea, as known by railway surveys, the top of the Devils Heart is approximately 1,722 feet, and of Sullys Hill, 1,707 feet, above the sea.
        The Dovre moraine is blended with the later Fergus Falls and Leaf Hills moraines from the Devils Heart northwest and west by Fort Totten and the Crow Hills to the northwest part of township 151, range 66, in the west edge of the Indian reservation, where it again becomes a separate belt. Thence it passes in a general west-northwestward course along the north side of the Antelope Valley to the area of the glacial Lake Souris. Immediately north of Oberon its very irregular knolls, hills, and small short [p.158] ridges border the Minnewaukan branch of the Northern Pacific Railroad along a distance of 3 miles. Next to the north there is a width of nearly 2 miles of moderately undulating, smooth till, in the central part of which is the disused Fort Totten station. Westward this smooth tract extends about 2 miles, dividing the Dovre moraine on the south from the compound Fergus Falls and Leaf Hills moraines on the north; but beyond this it is interrupted by morainic hills surrounding Long Lake, where the two belts seem to be once more united. The further course of the Dovre moraine and its relationship with these later moraines upon the very scantily settled country extending from Long Lake and Minnewaukan westward have not been definitely traced and are laid down on the map (Pl. XVII) only in a provisional manner. It is probable that the Dovre moraine is well developed 5 to 10 miles south of Rugby Junction, where the surface bears numerous small lakes, and that it is also represented on the southeast part of the Lake Souris area by swells about 60 feet above the general level, seen 1 to 3 miles southeast of Berwick. The continuation of the ice border thence northwest may have been approximately coincident with the sand hills lying on both sides of the Souris or Mouse River, in townships 157 and 158, ranges 75 and 76. Little Medicine Lodge, so named from its being formerly the scene of dances of the Indians, with incantations of their medicine men, is one of these dunes without vegetation, which rises about 75 feet in height on the west bank of the Souris River, 5 miles north of Towner.

        The eighth and ninth or Fergus Falls and Leaf Hills moraines unite in the south part of Ottertail County, Minn., to form the Leaf Hills, 100 to 350 feet high, which reach from Fergus Falls in a semicircle 50 miles southeast, east, and northeast, to the Leaf Lakes. These are the most conspicuous morainic hills of this State; and on account of their prominence above all the adjoining country they have been commonly called the Leaf Mountains. Eastward from these hills the Fergus Falls moraine seems to be merged with the Dovre moraine through Miltona and Spruce Hill, in the northeast part of Douglas County, and through southwestern and southern Todd County. Next it runs northward in a well-marked belt [p.159] of drift hills through the east edge of Todd County to the magnificent development of this moraine, apparently united again with the ninth or Leaf Hills moraine, about Fish Trap Lake and Lake Alexander, in northwestern Morrison County. Crossing to the east side of the Mississippi, the Fergus Falls moraine passes south along the east side of this river's Glacial flood-plain to Hole-in-the-Days Bluff and the massive hills east of Little Falls, where it is probably combined with a reentrant angle of the Dovre moraine. Thence it passes northeastward through the southeast part of Crow Wing County, skirting the northwest side of Mille Lacs; and curving next southeastward around this lake, it appears to be represented by morainic hills observed in the southeast part of Aitkin County and in northwestern Pine County. Northward from Fergus Falls, this moraine passes by Lakes Lida and Lizzie to Detroit, the White Earth Agency, and White Earth Lake; and thence it turns nearly by a right angle west-northwest to the Frenchmans Bluff, in township 143, range 43, Norman County, west of which it enters the area of Lake Agassiz, close South of the Wild Rice River.(55)
        The course of the ice front where it formed the northern barrier of Lake Agassiz at the time of its accumulation of the eighth or Fergus Falls moraine is marked by hilly and knolly drift deposits, with plentiful bowlders, both east and west of the lake near the latitude of 47° 10', which passes 20 miles north of Fargo; by an unusual abundance of bowlders near this latitude and farther north on portions of the slightly undulating or nearly level till forming each side of the lacustrine area; and by a tract of till several miles in width, probably representing both the eighth and ninth or Fergus Falls and Leaf Hills moraines combined, which stretches across the Red River Valley at Caledonia, from Ada, Rolette, and Beltrami west to Reynolds, Buxton, Cummings, and Blanchard, constituting the bed and banks of the river along the Goose Rapids. In Lake Agassiz the morainic till was spread with a generally even surface, but it has many small inequalities, the higher portions being 3 to 5 feet or rarely 10 feet above adjoining hollows. Bowlders and gravel are plentiful on its surface, [p.160] this being the only interruption of the lacustrine and alluvial clayey silt which elsewhere continuously occupies the central part of the Red River Valley plain from near Breckenridge to Winnipeg.
        In North Dakota the Fergus Falls moraine passes southward in a loop outlining a lobe of the ice-sheet which lay between Lake Agassiz and the Sheyenne River. This lobe reached about 40 miles south from the latitude of Caledonia, which marks approximately the apex of a glacial reentrant angle in the Red River Valley, where the laving action of Lake Agassiz appears to have caused the melting of the ice border to progress faster than on the land surface at each side. Emerging from the lacustrine area, on which its drift was leveled by the waves, the Fergus Falls moraine presents a prominently rolling surface of till west-southwest of Galesburg, and forms typical morainic hills in the northwest part of township 142, range 53, 1 to 3 miles west of Erie, rising 50 to 75 feet above the intervening hollows and the adjoining surface of smoothly undulating till westward, and 100 feet or more above the highest shore of Lake Agassiz. Thence the east boundary of the ice-lobe is marked by rolling and knolly till, with plentiful bowlders, and by low kames of gravel and sand, occupying generally a belt about 1 mile wide, which runs south-southwestward close north and west of Ayr and southward across the Northern Pacific Railroad 1 to 2 miles east of Buffalo. In the northwest part of township 139, range 54, a few miles south of this railroad, this morainic belt, continuing with similar features, turns to the west-southwest, crosses the Maple River, and holds this course about a dozen miles, passing through sections 25 to 28 and 32 and 31, township 139, range 56. This portion defines the southern extremity of the ice-lobe, the distance of its retreat from the Dovre moraine at Taylor Lake and near Geneseo and Cayuga having been about 55 miles. In Minnesota the recession of the ice from Dovre to the Leaf Hills was 60 miles, and along the Red River Valley it was approximately 100 miles from the southeast corner of North Dakota to Caledonia.
        Occasional kame knolls and small plateaus of gravel and sand, 15 to 25 feet high, were observed along a distance of a dozen miles south of the end of this morainic loop, in townships 138 and 137, range 57, showing where glacial streams had flowed down from the ice-sheet during its retreat [p.161] between the Dovre and Fergus Falls moraines, and there are small areas of undulating or nearly level gravel and sand associated with these kames. With these exceptions, all the expanse there and southward between the Dovre moraine and the area of Lake Agassiz consists of smoothly undulating intermorainic till. A similar sheet of till also occupies the area 15 to 20 miles wide inclosed on the north between the sides of the morainic loop, from Tower City and Oriska to Page City, Hope, and Sherbrooke.
        From sections 36, 26, and 26, township 139, range 57, the Fergus Falls moraine takes a north-northwest course through section 14 in this township, the