Outburst flood origin of the Central Kalamazoo River Valley, Michigan, USA

ARTICLE IN PRESS

Quaternary Science Reviews 24 (2005) 2354–2374

Outburst flood origin of the Central Kalamazoo River Valley, Michigan, USA Andrew L. Kozlowskia,, Alan E. Kehewb, Brian C. Birdb a

Department of Geological and Environmental Science, 514 University Ave, Susquehanna University, Selinsgrove, PA 17870-1001, USA b Department of Geosciences, Western Michigan University, Kalamazoo, MI 49008, USA Received 13 January 2004; accepted 2 March 2005

Abstract Geomorphic evidence and stratigraphic information from boreholes suggest that the oversized Central Kalamazoo River Valley (CKRV) in southwest Michigan resulted from a catastrophic outburst flood emanating from subglacial channels under the Saginaw lobe of the Laurentide Ice Sheet. The CKRV occurs as a deeply incised trench over 2 km wide and in excess of 50 m deep situated in a reentrant formed by the Lake Michigan, Saginaw and Huron-Erie lobes. The course of the CKRV follows an irregular flow path that bisects the Kalamazoo Moraine of the Lake Michigan lobe. Erosional terraces near the mouth of the channel indicate that Lake Michigan lobe meltwater drained eastward prior to the westward Saginaw outburst. Prior to valley formation the Lake Michigan lobe had retreated westward to at least the Lake Border Moraine. With the Lake Michigan lobe absent to impede flow, drainage from the CKRV proceeded southwesterly until draining into glacial Lake Chicago near St. Joseph, Michigan. The outburst originated from a system of Saginaw tunnel channels that display convex-up flow profiles and contain eskers. Meltwater drainage transitioned from subglacial-to-ice marginal and proglacial environments. During the interval represented by the outburst, the Saginaw Lobe appears to have been in a relatively stationary position. r 2005 Published by Elsevier Ltd.

1. Introduction 1.1. Location of the study area and nature of the problem The present-day Kalamazoo River represents a major drainage course for surface water in the Great Lakes Region, USA. The 250 km river drains an area in excess of 5000 km2 in southern Michigan originating near the Town of Albion and flowing to the northwest until emptying into Lake Michigan (Fig. 1). The morphology of the central section of the Kalamazoo River Valley (CKRV) occurs as a trench in excess of 2 km wide and exceeds 50 m in depth. The channel size and morphology of the CKRV is in striking contrast to eastern and Corresponding author. Tel.: +1 570 372 4211; fax: +1 570 372 2726. E-mail address: [email protected] (A.L. Kozlowski).

0277-3791/$ - see front matter r 2005 Published by Elsevier Ltd. doi:10.1016/j.quascirev.2005.03.016

western valley segments and other streams in the region. Additionally, the stream channel follows an anomalous flow path that turns abruptly north and bisects a prominent upland of a former ice margin that should have obstructed or otherwise diverted flow to the south. The CKRV is located in a complex glaciated terrain associated with a reentrant formed by three Late Wisconsin sublobes of the Laurentide Ice Sheet, that include the Lake Michigan, Saginaw, and Huron-Erie lobes (Fig. 2). Data will be presented to show that the anomalous size and course of this segment of the river represents a major discharge event that significantly influenced drainage development in south-central and southwestern Michigan. The chronology of events is thought to be between the conclusion of the Erie Interstade (Morner and Dreimanis, 1973) and the Crown Point Phase of deglaciation (Hansel and Johnson, 1996).

ARTICLE IN PRESS A.L. Kozlowski et al. / Quaternary Science Reviews 24 (2005) 2354–2374

Fig. 1. Location of study area in Great Lakes Region, USA. Presentday course of Kalamazoo River system drains over 5000 km2 originating near the Town of Albion and emptying into Lake Michigan at Saugatuck.

2355

Near the village of Plainwell, in southeast Allegan County, the 2.5 km wide CKRV channel empties westward into a narrow northeast trending glaciofluvial/glaciolacustrine basin (Fig. 3). Immediately upstream from Plainwell, the channel cuts through the Kalamazoo Moraine of the Lake Michigan lobe (LMKM) in northwest Kalamazoo County, maintaining its uniform channel width and steep walled sides for 35 km to approximately Galesburg, Michigan (Fig. 3). The morphology and size of the CKRV in this reach is similar to those described for glacial-lake spillways in the upper Great Plains of the United States and Canada produced by catastrophic floods. (Kehew and Lord, 1986, 1987; Teller and Thorliefson, 1987; Clayton and Attig, 1987; Lord, 1991; Kehew and Teller, 1994). Further east near Augusta, Michigan, the channel morphology becomes more irregular and at Battle Creek in Calhoun County (Fig. 3) the oversized CKRV deviates from the modern river and continues northeasterly through the Kalamazoo Moraine of the Saginaw lobe (SKM) toward the Village of Bellevue in southwestern Eaton County (Fig. 3). Northeast of the SKM, the CKRV passes through an upland dissected by a series of NE–SW oriented, anastomosing channels 0.5–1.3 km in width containing eskers (Figs. 3 and 4). Our hypothesis is that the CKRV is a channel from a catastrophic outburst flood that originated from subglacial Saginaw lobe tunnel channels. The ensuing floodwaters then proceeded subaerially, channeling through ice marginal positions and draining westward into glacial lake basins, and eventually draining into glacial Lake Chicago (Fig. 4). The purpose of this paper is to evaluate a flood hypothesis as an explanation for the unique morphology of the central valley, and to propose a subglacial origin for floodwater. Additionally, our aim is to evaluate the regional drainage history, timing, ice dynamics and role of meltwater associated with the formation of the CKRV. 1.2. Previous studies

Fig. 2. Extent of the Laurentide Ice Sheet at approximately 15.5 ka BP. Note the development of distinct lobes of ice impacting Michigan. Arrows indicate ice flow direction. Great lake basins outlined with dashed lines and study area outlined by gray polygon.

The concept of catastrophic meltwater drainage is not new to southern Michigan. Early studies by Ekblaw and Athy (1925) and later elaborated on by Fisher and Taylor (2002) and Fisher et al. (2005) proposed that the Kankakee outwash plain in Illinois resulted from high magnitude meltwater flows (Kankakee Torrent) originating from the drumlinized topography in Branch County, Michigan, flowing southwestward across northwest Indiana and into Illinois. Leverett and Taylor (1915), Bretz (1952) and most recently Kehew (1993) completed studies investigating catastrophic westward drainage of the glacial Grand River resulting from overflow of glacial Lake Saginaw in north-central Michigan.

ARTICLE IN PRESS 2356

A.L. Kozlowski et al. / Quaternary Science Reviews 24 (2005) 2354–2374

Fig. 3. (A) Color DEM of study area with low elevations in blue and high elevations in yellow. Letters correspond to cities and towns: K—City of Kalamazoo; O—Village of Otsego; P—Village of Plainwell; G—Galesburg; A—Augusta; B—Battle Creek; and Bv—Bellevue. (B) Regional landform map displays prominent regional landforms including end moraines as mapped and tunnel channels mapped by Kozlowski et al. (2003). Note that Plainwell is located in the basin referred to in the text.

ARTICLE IN PRESS A.L. Kozlowski et al. / Quaternary Science Reviews 24 (2005) 2354–2374

2357

Fig. 4. (A) Color DEM of southwestern Michigan showing location of study area and regional elevation relationships. Highest elevations occur in the southeast and decrease to the northwest. White arrow displays drainage path proposed in this paper for Kalamazoo outburst flood. Note the intertwined channels feeding into the head of the oversized CKRV and large set of radial channels in the northeast section of the map interpreted as tunnel channels by Kozlowski et al. (2003). Letters correspond to referred locations: L—Lawrence; D—glacial Lake Dowagiac and SL—Saginaw Lowlands. (B) Mapped end moraines occur as concentric ridges transverse to ice flow directions. (C) Figure locations shown on DEM for spatial reference.

Leverett and Taylor (1915) published the first map of moraines, lakebeds and outwash fans and provided the most thorough initial review of surface formations and landforms. Revised maps were produced by Martin (1955) and later Farrand (1982) produced a map that differentiated units on the basis of sediment texture and

geomorphology. Glacial terrain and surficial deposits maps have been produced for a few selected counties including: Kalamazoo County (Monaghan et al., 1983), Van Buren County (Terwilliger, 1954; Kehew et al., 2002) and St. Joseph County (Kehew et al., 1999a). More recently, investigations regarding till stratigraphy

ARTICLE IN PRESS 2358

A.L. Kozlowski et al. / Quaternary Science Reviews 24 (2005) 2354–2374

and mineralogy have been completed (Monaghan and Larson, 1986; Monaghan et al., 1986; Monaghan, 1990; Finkbeiner, 1994; Gardner, 1997; Kozlowski, 1999; Beukema, 2003; Kehew et al., 2005). Correlation and chronology of Wisconsin glacial landforms in this region were proposed by Zumberge (1960), Monaghan (1990) and more recently by Kehew et al. (1999b). Recent studies regarding tunnel channels in the region have concentrated on their formation (Kehew et al., 1999b; Sjogren et al., 2002), deposits and stratigraphy (Fisher and Taylor, 1997, 1999, 2002; Taylor et al., 1998; Kozlowski, 1999; Fisher et al., 2003), mapping and distribution (Kehew et al., 1999b; Kozlowski et al., 2001, 2003; Fisher and Taylor, 2002; Fisher et al., 2005). 1.3. Geologic setting The study area is located in the Great Lakes Region of the Central Lowlands geologic province of the United States. Late Wisconsin glacial sediment, 2–180 m thick, covers the study area and overlies Mississippian to lower Pennsylvanian age shales, sandstones and carbonate rocks gently dipping northeast on the southwestern rim of the Michigan Basin (Catacosinos et al., 2001). The Mississippian Coldwater Shale comprises the majority of bedrock underlying the study area (Fig. 5) and limited exposures of Marshall Sandstone and Bayport Limestone outcrop along the Kalamazoo River valley near Battle Creek and Bellevue Michigan. The majority of landforms in southern Michigan have been customarily interpreted as constructional in nature and occur as a series of concentric uplands oriented transverse to ice flow from their respective lobes

Fig. 5. Map of bedrock formations occurring in the study area. Study area occurs on southwestern rim of Michigan Basin. Note how some segments of oversized CKRV outline coincides with bedrock valleys mapped.

(Leverett and Taylor, 1915; Zumberge, 1960). Traditionally, these landforms have been interpreted as successively younger end moraines (Fig. 4B) produced as glacial ice retreated or readvanced across southern Michigan during the Late Wisconsin Stage. The Sturgis Moraine located just north of the Indiana State border in St. Joseph County forms the southernmost moraine of the Saginaw lobe (Fig. 4B). Approximately 10 km up, ice from the Sturgis Moraine, the topography transitions from a high-relief hummocky ice-marginal landscape to a 30-km-long by 44-km-wide drumlinized plain of Saginaw lobe origin (Dodson, 1985). Broad shallow channels with irregular margins are interpreted as tunnel channels (Kozlowski, 1999; Sjogren et al., 2002; Fisher et al., 2005) that parallel the Saginaw drumlins. The Tekonsha Moraine (Fig. 3) occurs in eastern Kalamazoo and Calhoun Counties and has traditionally been interpreted as an interlobate moraine of the Saginaw and Michigan lobes of ice (Leverett and Taylor, 1915; Martin, 1955; Farrand, 1982; Monaghan et al., 1983, 1986). However, unobscured palimpsest Saginaw lobe tunnel channels (Kehew et al., 1999b, 2005) in the central lowland (Fig. 3) between the LMKM and the Lake Michigan Tekonsha Moraine (LMTM) suggest that the steep walled western arm of the Tekonsha Moraine may simply be an erosional scarp resulting from glacial meltwater drainage. This would imply that the Lake Michigan lobe ice did not advance as far east as previous studies suggested. The Tekonsha Moraine (Figs. 3 and 4B) has been remapped several times with little consensus on its origin (Kozlowski, 1999; Fisher and Taylor, 2002). Some of the eastern areas traditionally mapped as a Saginaw end moraine are drumlinized (Kozlowski, 1999; Sjogren et al., 2002; Fisher et al., 2005) and crossed by irregular NE–SW linear linked depressions or shallow channels interpreted as tunnel channels of Saginaw lobe origin (Kozlowski, 1999; Sjogren et al., 2002). Such features are typically attributed to subglacial origin rather than ice-proximal positions (Menzies, 1987; Boulton 1987; Shaw et al., 1989; Shaw and Sharpe, 1987). However, the Climax outwash fan (Fig. 3B) and other large heads of outwash buries and partially obscures drumlins on and adjacent to the mapped western arm of the moraine and clearly indicates the presence of a Saginaw ice margin in this region during deglaciation (Monaghan et al., 1983; Dodson, 1993). The Kalamazoo Moraine occurs as the next interlobate moraine of the Saginaw and Lake Michigan lobes. The LMKM forms a prominent north-northeast trending ridge in southwestern Michigan (Figs. 3 and 4). High-relief hummocks composed predominantly of sand and gravel comprise the surface materials with thick sequences of diamicton occurring at depth (Kozlowski et al., 2001). The physiographic significance of the

ARTICLE IN PRESS A.L. Kozlowski et al. / Quaternary Science Reviews 24 (2005) 2354–2374

LMKM is in part due to elevations exceeding 335 m above sea level and an enhanced (scarp-like) proximal slope produced from wave action and or glaciofluvial erosion (Kehew et al., 2001). The distal margin consists of enormous glaciofluvial fans, some of which emanate from gaps in the moraine interpreted as deposits from Lake Michigan lobe tunnel channels (Kehew et al., 1999b). These topographic attributes make the LMKM the most discernable in the region and allow it to be easily traced from northern Indiana to Barry County (Fig. 4). On the contrary, the eastern SKM (Fig. 3) occurs as a vague ridge obscured within hummocky terrain that in places appears to have been dissected (Figs. 3 and 4). The glaciofluvial fans of the Lake Michigan lobe dwarf those of the Saginaw lobe, indicating that the LMKM represents either a sustained ice margin or a more active margin with rapidly deposited fans. Kehew et al. (1999b, 2005) documented cross-cutting relationships of palimpsest Saginaw lobe tunnel channels and Lake Michigan lobe fans that reinforce earlier suggestions by Zumberge (1960) that the eastern and western segments of the Kalamazoo Moraine formed at different times. Accordingly then, it appears that the Lake Michigan lobe encroached upon terrain previously occupied by the Saginaw lobe. West of the Kalamazoo Moraine is a broad upland that has traditionally been mapped as the Valparaiso/ Kendall Moraine system. The Kendall Moraine (Leverett and Taylor, 1915) consists of a steep escarpment of predominately hummocky topography along the easternmost edge of the larger Valparaiso system (Fig. 3). Westward of the Valparaiso Moraine the Lake Border Moraine occurs as a narrow contorted ridge much lower in elevation than the Kalamazoo or Valparaiso Moraines. Detailed descriptions of the Valparaiso and Lake Border Moraine is discussed elsewhere (Kehew et al., 2005). Throughout the last several decades a deficiency of dateable material has plagued Quaternary scientists in southern Michigan and hampered attempts to constrain the glacial chronology. At present, the chronology is largely based on correlation of till units, cross-cutting end moraines and radiometric dates outside of Michigan. Fig. 6 provides a summary of the proposed landform chronology and regional chronostratigraphy of southern Michigan.

2. Data/results 2.1. Channel morphology Studies of late Pleistocene catastrophic floods and modern jo¨kulhlaups can document specific geologic and geomorphic attributes associated with such high magnitude discharges (Malde, 1968; Baker, 1973; Kehew and

2359

Fig. 6. (A) Plan view of distribution and proposed chronology of ice marginal positions: LBM—Lake Border Moraine; VAL—Valparaiso Moraine; LMKM—Lake Michigan lobe Kalamazoo Moraine; SKM—Saginaw lobe Kalamazoo Moraine; CHAR—Charlotte Moraine; LAN—Lansing Moraine. Schematic diagrams of chronostratigraphic relationships to landforms produced by the (B) Lake Michigan Lobe and (C) the Saginaw Lobe. Note: USP denotes the Union Streamlined Plain. Sources: Monaghan and Larson (1986), Monaghan et al. (1986), Monaghan (1990) and Hansel et al. (1996).

Lord, 1986; Lord and Kehew, 1987; Maizels, 1991; Kehew, 1993; O’Connor, 1993; Russell et al., 1999; Cutler et al., 2002; Fisher and Taylor, 2002). Diagnostic indicators of these floods include: (i) a deep, wide, steep walled channel morphology; (ii) the presence of largescale bars; (iii) scour zones and erosional terraces; (iv) streamlined erosional residuals; (v) inversely graded boulder-gravels; and (vi) coarse-grained fan deposits where channels terminate. The CKRV has many of these characteristics. Under-fit streams by their nature imply that nonequilibrium conditions existed at some point in a valley’s history. Width-to-depth ratios (W/D) measured for several stream valleys in the study area quantify the disparity in channel form of the CKRV (Fig. 7). Schumm (1977) determined that much of the relationship between W/D ratios is largely dependent on the nature of the sediment comprising the stream channels. In light of the variable surface deposits in this region, stream valley cross-sections from multiple locations in the study area were compared and illustrate the striking contrast in morphology of the CKRV to other stream valleys and even other segments of Kalamazoo valley (Fig. 8). W/D relationships as described by Schumm (1977) have been used primarily to evaluate channel shape of streams from bank-to-bank and not typically

ARTICLE IN PRESS 2360

A.L. Kozlowski et al. / Quaternary Science Reviews 24 (2005) 2354–2374

Fig. 7. Graph of width to depth relationships for western Michigan stream valleys. Note how all but CKRV cluster as one population and the CKRV occurs on its own, clearly illustrating the variation in channel morphology.

valley wall-to-valley wall. Nevertheless, the W/D relationships provide a useful measure to evaluate morphological variations in regional fluvial systems. In this instance, the shape of the CKRV has more in common with that of an enormous channel than it does with a ‘‘V’’-shaped stream valley produced by normal fluvial erosion (Fig. 8). The enormous channel of the CKRV provides some of the most compelling evidence in support of a flood origin. However, little variation exists between the average annual peak discharges for the streams evaluated and the present discharges of the Kalamazoo River are inadequate to account for the oversized channel dimensions. Further, entrenchment and downcutting of streams are usually attributed to regional uplift or fluctuations in base level. Explanation of the CKRV as the product of downcutting due to isostatic rebound is precluded because such processes generally affect an entire stream and most streams in a region as opposed to just one segment of a stream. Thus, the explanation of valley dimensions through lateral planation by meanders is unsuitable to account for the size of the CKRV channel. 2.1.1. Depositional bars Large-scale bars and other bedforms have been described in association with catastrophic floods, especially Lake Bonneville and glacial Lake Missoula (Malde, 1968; Baker, 1973). Lord and Kehew (1987) and Kehew and Lord (1987) found evidence for glacial lake outburst floods on the Souris Spillway in North Dakota. A characteristic feature associated with the outburst events are large-scale bar deposits composed of largecaliber sediment (boulders) within the channel.

Fig. 8. (A) DEM showing location of cross-sectional river valley profiles in southwest Michigan. (B) Topographic profiles of selected stream valleys and segments of the Kalamazoo River Valley in southwest Michigan. Note the change in channel morphology for Central Kalamazoo River Valley (vertical exaggeration 25
).

West of Galesburg along the northern wall of the Kalamazoo River Valley, is a large bar form approximately 2 km long, 0.5 km wide and 15 m thick (Fig. 9). The landform occurs in a point bar position as the channel makes a bend northward. The surface of the bar is mantled with cobbles and sporadic boulders. An absence of exposures prevents evaluation of the composition and sedimentologic details. However, well logs indicate a composition of sandy–cobble gravel. The description given above is similar to descriptions of pendant bars occurring in the Souris Spillway and other glacial-lake spillways of the upper Great Plains (Lord and Kehew, 1987). Another bar is located about 10 km upstream near Battle Creek (Fig. 10). Located just south of the channel center is a large streamlined bar 2 km

ARTICLE IN PRESS A.L. Kozlowski et al. / Quaternary Science Reviews 24 (2005) 2354–2374

2361

Fig. 9. (A) Inset schematic diagram showing types of large-scale bars formed within Souris Spillway ND: 1—pendant bars and 2—alcove bar deposits on the spillway side of a landslide. Source: Kehew and Lord (1987). (B) Hillshade DEM displaying (P) large-scale pendant bar deposit located near Galesburg, Michigan, within the Central Kalamazoo River Valley.

long, 1 km wide and 10 m higher than the valley floor (B in Fig. 10). Exposures from a gravel pit located on the northwest end of the bar reveal a poorly sorted, sandy matrix supporting cobble gravel with rip-up clasts of laminated silt (lake sediment), tabular pieces of local bedrock (Marshall Sandstone), and clasts of diamicton interpreted as till. A 20 m deep hollow-stem auger exploration borehole in the floor of the pit indicates the sediment below the bar fine downwards from cobblegravel-to-sand, consistent with descriptions of flood deposits elsewhere (Fisher and Taylor, 2002). 2.1.2. Erosional residuals Obstacles to flood flow in the Channeled Scabland of Washington, glacial-lake spillways of the northern Great Plains, and the Lake Bonneville floods of Idaho evolve into erosional resisduals that often approach a minimal-drag equilibrium form (Malde, 1968; Baker, 1973; Kehew and Lord, 1987). Elliptical and streamlined hills, interpreted as erosional residuals, occur in and along the outer margin of the CKRV, parallel to flow direction (E in Fig. 10). Additionally, the residuals found in association with the CKRV display length-towidth ratios similar to those described by Komar (1984) for minimal-drag equilibrium forms. Some of the CKRV residuals are found on an outer marginal zone (Fig. 10) similar in position to the outer scour zone (OSZ) of erosional development detailed by Kehew and Lord (1986). The OSZ has a limited extent along the CKRV. The zone is discontinuous and occurs typically at point bar locations along the valley near Augusta and Battle Creek (Fig. 3). The scour zone, where present, typically exists 15–24 m above the bottom of the present-day inner channel, grades to the

Fig. 10. (A) Hillshade DEM of Central Kalamazoo River Valley displaying large streamlined bar (B) west of Battle Creek within the main inner channel (INC). The elevation of the bar 8 m above the present-day valley floor, while erosional residual (E) in the center of dashed ellipse occurs on an outer scour zone (OSZ) 33 m above the top (B). Note the loss of clarity in channel margins and hummocky topography along flanks most likely due to melting of buried ice after valley formation. (B) Topographic profile indicates surface of OSZ grades eastward.

west and appears as terrace scarps on maps by Monaghan et al. (1983). The scour zone described in this paper is interpreted to represent early stage erosional development when large discharges overwhelmed a smaller nonintegrated drainage system. The residuals in the OSZ lack the idealized equilibrium form seen in other spillway systems which may be explained by shorter flow duration before entrenchment of an inner channel. 2.1.3. Stream terraces Several terraces exist along the CKRV, some represent an OSZ as described above near Augusta and Battle Creek and others may have originated from a different erosional episode. The most striking example exists as a laterally offset pair of high-elevation stream terrace treads in northern Kalamazoo County (Fig. 11). Both sets of terraces are approximately 37 m above the valley floor and 30 m below the crest of the Kalamazoo Moraine at an elevation of 260 m AMSL. The western

ARTICLE IN PRESS 2362

A.L. Kozlowski et al. / Quaternary Science Reviews 24 (2005) 2354–2374

terrace has an elliptical shape 8.5 km long and 2.5 km wide, and the eastern terrace is 4.5 km wide and 6.5 km long. The terrace treads grade to the southeast (Fig. 11B), and cross-bedded sand and gravel outwash exposures in the western terrace (Havira, 1969; Straw, 1976) indicate a southeasterly flow direction. The large terraces are well defined by steep scarps formed on the Kalamazoo Moraine (Figs. 4 and 11) and are interpreted as erosional in origin (see Fig. 11C, model 1.) A continuous-core rotosonic boring on the western

terrace tread revealed that the upper 6 m consists of poorly sorted sand and gravel overlying 20 m of massive gray diamicton (Kozlowski, 2001). Diamicton with the same characteristics outcrops 0.5 km west in a gravel-pit along the west wall of the valley. Geomorphic and stratigraphic relationships strongly suggest that the terraces represent a higher elevation eastward meltwater flow of Lake Michigan Lobe origin (Kozlowski, 2004). 2.1.4. Cross-cutting relationships: Evidence for rapid incision Non-oriented hummocky topography is common in southwestern Michigan. Most often, this topography is explained by topographic inversion from the melting of buried ice blocks (Clayton and Moran, 1974; Johnson et al., 1995; Ham and Attig, 1996). Alternative explanations include ice-pressing of soft-bedded, fine-grained deformation till (Stalker and Mac, 1960; Eyles et al., 1998), erosional remnants of pre-existing deposits by subglacial sheetfloods (Munro and Shaw, 1997) or as depositional bedforms (Fisher et al., 2003). In Kalamazoo County, and surrounding counties in southwest Michigan, many kettles occur as a series of linear, linked depressions that trend northeast to southwest. Kehew et al. (1999b, 2005) interpret these linear kettles as palimpsest tunnel channels of the Saginaw lobe. In many cases, these features provide evidence for relative timing of glacial lobes in the region. One such tunnel channel identified by Kehew et al. (1999b) and associated with the CKRV occurs in the vicinity of Galesburg, Michigan. Close examination of the USGS Galesburg 7.5 min quadrangle topographic map reveals an interesting relationship between a series of north to south trending tributaries and a NE–SW trending tunnel channel along the northern edge of the CKRV (Fig. 12). The tributaries cross-cut the tunnel channel at nearly right angles indicating that the tunnel channel was filled with buried ice and debris and not expressed at the surface during incision of the tributaries. Buried ice is capable of persisting for long periods of time (Evans and England, 1993; Sugden et al., 1995), and the buried ice in the tunnel channel could not have melted before being cross-cut at right angles by the tributaries. Support for this hypothesis is rapid incision of the CKRV. 2.2. Stratigraphic framework of the CKRV

Fig. 11. (A) Hillshade DEM displaying large terraces (highlighted and numbered 1 and 2) along the Central Kalamazoo River Valley. Black arrow shows flow direction of the Kalamazoo River. (B) Topographic profiles displays large terrace cut through the Kalamazoo Moraine along river valley (A–A0 ). Terrace elevations of 260 are 30 m above the valley floor and 30 m below the crest of the moraine. Profile B–B0 indicates terrace treads slope southeastward to the central lowland. (C) Schematic diagram displays models of erosional and depositional terraces.

Within the CKRV the water table is usually within a few meters of the surface and exposures are not available to evaluate the stratigraphy and composition of the valley fill. Twenty exploration boreholes were completed with hollow-stem auger, direct rotary and rotosonic methods in combination with near surface geophysics to characterize the deposits and stratigraphy of the CKRV (Kozlowski, 2004). Geophysical methods utilized include electrical resistivity methods, ground

ARTICLE IN PRESS A.L. Kozlowski et al. / Quaternary Science Reviews 24 (2005) 2354–2374

Fig. 12. DEM of palimpsest Saginaw lobe tunnel channel labeled TC. South-southeast trending valleys are tributaries to the Kalamazoo River Valley. Tributaries were cut at nearly right angles across buried tunnel channel filled with ice and debris. Interrupted stream gradients (profiles A and B) support the stagnant ice model proposed by Kehew et al. (1999a, b) for Saginaw lobe tunnel channel preservation. Crosscutting relationships indicate that incision of Central Kalamazoo River Valley must have occurred rapidly. Contour interval is 3 m.

penetrating radar (GPR) and gamma-ray logging of exploration boreholes and newly completed water wells. In this paper, only the exploration boreholes and downhole geophysics are presented. For a detailed discussion on resistivity and GPR results regarding this study see Kozlowski (2004). The subsurface information collected was used to construct geologic cross-sections to evaluate the history and composition of the valley-fill deposits.

2363

2.2.1. Valley-fill description The first few meters of sediment consist of flood-plain silts and sandy alluvium, some of which is derived from Holocene alluvial fans. The majority of this upper alluvial sediment is finer grained than deposits found at depth. Coarse gravels constitute the bulk of sediment from 2 to 12 m within the channel fill and continuous rotosonic cores revealed repeating sequences of reversely graded beds. Borehole samples and difficult drilling conditions (loss of circulation) suggest that these deposits are primarily clast-supported gravel. Coarsergrained deposits interpreted as clast-supported openwork boulder gravel also occur and were encountered in several locations within the valley. Encounters with boulder gravels while drilling and near surface geophysics indicate that the boulder-gravel deposits occur with a lens geometry at, or near the base of the channel. Beneath the channel fill within the CKRV is a nearly continuous, massive, gray diamicton unit averaging 5 m thick (Fig. 13), with a composition of 34% sand, 48% silt, 16% clay, and with a substantial gravel component. The upper diamicton encountered in the KV-02 borehole contained wood fragments in the matrix, greater than 50 ka BP in age, suggesting that this diamicton is either pre-late Wisconsin in age or contains reworked wood fragments. Two other deep diamicton units were also encountered in borehole KV-02; most likely preWisconsin in age. Deposits underneath the upper diamicton (Figs. 13 and 14) are variable and discontinuous, although thick units of sand and gravel are not uncommon and a number of boreholes display sequences of laminated, fine-grained silt and clay material. Some of these laminated sections were as much as 9 m thick. From other locations in the valley deposits underlying the upper diamicton unit consist of silty, fine-grained sand. West of where the channel breaches the LMKM the CKRV enters a valley with a complex glaciofluvial/ lacustrine drainage history, and continues southwesterly (Fig. 4). Continuous rotosonic cores and water well data near Plainwell (Fig. 3) indicate 1–3 m of fine sand and isolated pockets of laminated fine-grained sediment above 10 m of coarse gravel deposited above a 5 m massive gray diamicton. The gravel deposits at Plainwell are more broadly distributed than further upstream where they are confined within the channel. Further south and west along the proposed drainage route of the CKRV, Bird et al. (2000, 2001) and Kehew et al. (2001) recognized a large infilled channel that partially incised through a brown diamicton thought to be the Saugatuck till of the Lake Michigan lobe. Geologic cross-sections (Fig. 15) and continuous core rotosonic exploration boreholes in the Paw Paw River Valley near Lawrence, Michigan, indicate that the channel is infilled with 12 m of gravel similar in character and thickness to those described within in the CKRV further upstream.

ARTICLE IN PRESS 2364

A.L. Kozlowski et al. / Quaternary Science Reviews 24 (2005) 2354–2374

Fig. 13. (A) Cross-section locations and (B) photograph of typical channel fill deposits of the Kalamazoo River valley (15 cm OD auger for scale). (C) Stratigraphic profile of Central Kalamazoo River Valley constructed from exploration boreholes (KV—designation), water well records (UV— designation) and municipal water wells. Profile displays nearly continuous diamicton underlying coarse gravel deposits. Note the strong correlation between gamma-ray logs in explorations boreholes. Vertical lines to the right of the gamma-ray logs labeled 1000 and 2000 correspond to the horizontal scale of natural gamma radiation, measured in counts per minute (CPM). Finer-grained geologic materials (clay–silt) yield strong signals to the right while coarser materials (sand and gravel) yield weaker signals that deflect to the left of the signal center. Note also the undulating Coldwater Shale surface, and the deep trough (City Well Field) in the center of cross-section which corresponds to a Saginaw lobe tunnel channels at surface.

Exploration drilling and construction of geologic cross-sections (Figs. 13–15) suggest that the outburst event that formed the CKRV scoured the valley to an average depth of 12 m below the present-day valley floor. The subsequent valley fill or channel fill is comprised of Holocene and Pleistocene glaciofluvial deposits. The repeated coarsening-upward sequences, and coarse nature of the deposits within the CKRV, is consistent with descriptions of Maizels (1991) for jo¨kulhlaup deposits as opposed to normally graded deposits derived from ablation flows. The rhythmic nature may represent velocity fluctuations, pulses of meltwater or perhaps multiple Late Wisconsin flood events. For clarity, it is important to note that these deposits within the channel either postdate the outburst event that scoured out the valley or represent waning flows during the same incision event. The boulder-gravel deposits may represent buried bar forms or erosional

residuals. Cross-sections derived from multiple boreholes indicate that incision was into a widely distributed diamicton.

3. Discussion 3.1. Subglacial origin of meltwater outburst Outburst flood origins are typically prescribed to three possible origins that include: (1) failure of ice dams or over spilling of proglacial lakes (Baker, 1973; Kehew and Lord, 1987; Clague and Evans, 1994); (2) enormous meltwater flows derived from subglacial volcanism (Russell et al., 1999; Roberts et al., 2000); (3) sudden drainage of stored subglacial meltwater (Wright, 1973; Shaw et al., 1989; Shoemaker, 1991; Sugden et al., 1991; Brennand and Shaw, 1994). The absence of volcanic

ARTICLE IN PRESS A.L. Kozlowski et al. / Quaternary Science Reviews 24 (2005) 2354–2374

2365

Fig. 14. Geologic cross-sections B–B0 and C–C0 across the Central Kalamazoo River Valley based on exploration boreholes (KV-), water well data (MLC, OD, and NC) and downhole geophysics; locations shown in Fig. 13A. Cross-cutting relationships observed on the Morrow Lake Crosssection (B–B0 ) suggests the outburst event incised fully through the upper sand and gravel facies into more cohesive diamicton. Deposits beneath valley fill consist of multiple diamictons (till) and silty laminated sediment (lake sediment?). The Cooper geologic cross-section (C–C0 ) displays a more complicated stratigraphy than (A).

processes in southern Michigan precludes it as a formational mechanism for the CKRV. Additionally, at the head of the oversized channel of the CKRV in southwestern Eaton County glaciolacustrine sediment or landforms have not been mapped to support the presence of a large proglacial lake in the area. Instead,

the terrain is composed of rolling hummocky topography containing large coalescing anastamosing channels at the head of the CKRV (Fig. 16). Channels range in width from 0.5–1.3 km. The channels appear to be part of a larger system of northeast–southwest trending channels described by Kozlowski et al. (2003) and Fisher

ARTICLE IN PRESS 2366

A.L. Kozlowski et al. / Quaternary Science Reviews 24 (2005) 2354–2374

Channels feeding into the upstream end of the CKRV are interpreted as tunnel channels rather than subaerial valleys on the following basis: (1) channels have undulating convex-up profiles with reverse gradients in some segments (Fig. 17); (2) channels cross modern drainage divides (Fig. 16); (3) channel cross-sections are wide and steep walled, and box-shaped rather than ‘‘V’’shaped channels; (4) channels feeding into the CKRV are a subset of a system of radial channels that trend oblique to the regional slope and parallel the flow lines of the Saginaw lobe (Fig. 4); (5) the majority of channels contain eskers (Fig. 18). Some of the eskers that occupy tunnel channels in Barry County terminate at fans that may indicate the margin of a short-lived stable ice front. These inset eskers may be a continuation of a larger esker system mapped in Livingston County to the east. 3.2. Formation of the central Kalamazoo River Valley and regional drainage

Fig. 15. (A) Hillshade DEM of Paw Paw River Valley near Lawrence, Michigan, showing location of geologic cross-section (B). Most of the escaping drainage proceeded through the Paw Paw spillway (large arrow). However, some drainage may have escaped through other outlets and possibly the glacial Lake Dowagiac plain (dashed arrow). (B) Cross-section constructed from exploration boreholes (VBdesignation) and water wells (L-designation) displays an infilled channel scoured out by southwestward discharge from Kalamazoo River Valley outburst flood. Thickness of channel fill is similar to that measured in other valley cross-sections east of the Lake Michigan Lobe Kalamazoo Moraine. Note: depth of channel scour approaches that of the Glennwood level of glacial Lake Chicago.

et al. (2005). Channels in the region and especially those at the head of the CKRV display a wide variation in morphology. Some channels display box-shaped stream cut channels similar to those formed by incision of spillways. Others are highly irregular channels, composed of a series of depressions with vague margins that constrict and expand with chains of lakes. Channels occur at different elevations, and some are abandoned with no relation to the modern drainage (Kozlowski et al., 2003). The channels have many similarities to channels described elsewhere (Shaw et al., 1996; Beaney, 2002) and especially the extensive channel systems in Minnesota (Wright, 1973; Mooers, 1989; Patterson, 1994), and tunnel channels in Wisconsin (Clayton et al., 1999; Johnson, 1999; Cutler et al., 2002) and those recently described near Albion, Michigan (Fisher and Taylor, 2002; Sjogren et al., 2002; Fisher et al., 2005).

Determining why the CKRV flow path developed through the regional drainage divide of the LMKM rather than draining down the lower elevation central lowland is problematic. Two possible scenarios might explain the course of the CKRV. First, Dodson (1993) proposed that Lake Michigan lobe ice crossed the central lowland plain in an easterly direction, stagnated and was subsequently buried by the Saginaw lobe outwash delivered from southwesterly meltwater flows. Therefore buried ice might have blocked the southerly path of floodwater. An alternative scenario might include a large proglacial lake between the LMTM and LMKM. If present, such a lake could have drained catastrophically if the lake were to spill over the Kalamazoo Moraine westward, incising a path for future floodwaters to follow. The first scenario of buried ice as a broad sheet of stagnant ice is unlikely on the basis of multiple lines of evidence. Isolated pockets of diamicton exist at the surface exposures south and west of Kalamazoo; however, exploration boreholes completed to bedrock did not encounter a near surface diamicton that would be expected from direct glacial action or ice stagnation processes. Secondly, the scarp-like proximal margin of the LMTM does not continue on the north side of the Kalamazoo River valley as would be expected from a Lake Michigan lobe advance. Lastly, buried ice of this model would more than likely generate hummocky icecollapse topography, which is not present in the central lowland. Scenario two is also refuted on the basis that lake sediment was not encountered in exploration boreholes or at the surface in the central lowland plain, and there are no barriers or continuous moraines to the south of the plain to impound meltwater to form a lake, and even if there were, they would be lower in elevation than the LMKM and overflow would have occurred

ARTICLE IN PRESS A.L. Kozlowski et al. / Quaternary Science Reviews 24 (2005) 2354–2374

2367

Fig. 16. Hillshade DEM displaying a network of channels coalescing into the head of central Kalamazoo River Valley. Channels (black lines) are interpreted as tunnel channels of the Saginaw lobe. Complex topographic profiles within channels (Fig. 17) cross present-day drainage divides (white dashed lines). Additional support for a subglacial interpretation is the presence of inset eskers and the fact the channels parallel the hydraulic axis of the Saginaw lobe and trend oblique to the regional slope (Fig. 4).

through the southern end of the central lowland as opposed to through the LMKM. Therefore, it seems unlikely that either of these initial scenarios can satisfactorily explain the anomalous flow path of the CKRV. In our subglacial outburst hypothesis, we envision a different scenario based on the observed field evidence to explain the origin of the CKRV. In our model the

Lake Michigan lobe is at the Kalamazoo Moraine and the Saginaw lobe is at its correlative Kalamazoo Moraine at 15.5 ka BP as proposed by Monaghan (1990). The Lake Michigan lobe may have surged to its position (Kehew et al., 2005) where as the Saginaw lobe may have readvanced to the Kalamazoo Moraine or already have been there (Fig. 19A). Ice-contact marginal positions traced from the proximal side of the

ARTICLE IN PRESS 2368

A.L. Kozlowski et al. / Quaternary Science Reviews 24 (2005) 2354–2374

Fig. 17. Topographic profiles of a few tunnel channels feeding the Central Kalamazoo River Valley (see Fig. 16 for location). Channel system must have been pressurized because longitudinal topographic profiles display undulating flow paths that are in some cases convex-up with segments having reversed gradients.

Kalamazoo Moraine to a well-defined ice-contact position in Allegan County indicate that the Lake Michigan lobe margin near Plainwell, Michigan, readvanced or retreated non-uniformly and created a pinched high-elevation proglacial lake (Fig. 19B). This hypothesis is supported by the presence of 43 m of laminated silt and clay overlying several meters of eastward dipping gravel foreset beds at the highest elevation along the Kendall Moraine (Kozlowski, 2004). The highest elevation of the lake sediment along the Kendall Moraine is 264 m, whereas the high-elevation stream terraces bisecting the LMKM are at an elevation of 260 m. Presumably, the lake drained eastward through the LMKM, forming the terraces. The final spillway elevation is unknown but could not be below the elevation of the central lowland (Figs. 11 and 19C). Alternatively, supraglacial meltwater may have eroded through this topographic low while Lake Michigan lobe ice was still present at the LMKM and created a similar drainage way through normal glaciofluvial erosion. Subsequent retreat of the Lake Michigan lobe from the Kalamazoo Moraine opened a southern outlet for

Fig. 18. Hillshade DEM of a portion of Barry County adjacent to the coalescing tunnel channels (Fig. 16). NE–SW oriented tunnel channels of the Saginaw lobe (dashed lines) contain inset eskers (arrows). Esker-tunnel channel relationship is unclear but may represent a reorganization of subglacial hydraulics as suggested by Brennand and Shaw (1994) or as a late-stage conduits for englacial and superglacial sediment-charged drainage.

ARTICLE IN PRESS A.L. Kozlowski et al. / Quaternary Science Reviews 24 (2005) 2354–2374

2369

Fig. 19. Ice dynamics and developmental sequence for formation of Central Kalamazoo River Valley. (A) The Kalamazoo ice marginal systems are established at 15.5 ka BP in southern Michigan. (B) Development of a confined high-elevation proglacial lake or concentrated drainage occurs on the proximal side of the Lake Michigan Kalamazoo Moraine. (C) Incision of a wide spillway occurs where the proglacial lake overtops a sag in the crest of the moraine and drains eastward. (D) Continued retreat of the Lake Michigan lobe allows the development of multiple proglacial lakes between the Valparaiso and Kalamazoo systems. Eventually, drainage escapes down the glacial Lake Dowagiac corridor into northern Indiana. (E) Sometime around or shortly after 14,000 ka BP an outburst flood occurs from the Saginaw lobe and ensuing floodwaters capitalize on a predisposed weakness in the Lake Michigan Lobe Kalamazoo Moraine and drain into the glacial lake basins west of the Kalamazoo Moraine. The floodwaters are able to escape through the Paw Paw River Valley at Lawrence and continue southwestward to the vicinity of St. Joseph, Michigan. (F) The drainage event that created the large Central Kalamazoo River Valley drainage system proceeded through multiple glacial settings.

ARTICLE IN PRESS 2370

A.L. Kozlowski et al. / Quaternary Science Reviews 24 (2005) 2354–2374

Lake Michigan lobe glacial drainage, through glacial Lake Dowagiac (Fig. 4) (Martin, 1955) and out to South Bend, Indiana (Fig. 19D). Instability in the Lake Michigan lobe may have caused proglacial lakes to drain and re-fill episodically as it retreated westward and some drainage down the Lake Dowagiac channel may have been catastrophic in nature. Rates of downcutting and headward erosion would have been high in the un-vegetated soils of the moraine as the tributaries on the proximal side of the LMKM tried to incise through the spillway floor to adjust to new base levels of lower proglacial lakes behind the Kalamazoo Moraine. Finer-grained sand and gravel deltaic deposits below coarser gravel deposits in exposures on the Alamo Moraine (Fig. 3) indicate that lower-energy, glaciofluvial activity proceeded the outburst flood, similar to the observations of Cutler et al. (2002) for tunnel channels in Wisconsin. The combination of downcutting and headward erosion of gulleys and meltout of buried ice would be more than sufficient to enhance a preexisting topographic low through the LMKM. Throughout this sequence of events braided outwash streams driven by normal ablation meltwater from the Saginaw lobe most likely continued draining southward out through the central lowland. At the next phase the Lake Michigan lobe continues its retreat to at least the Lake Border Moraine, perhaps further into the Lake Michigan Basin. When the Lake Michigan lobe is at this position 14,000 ka BP (Hansel et al., 1985), we suggest that the subglacial outburst flood from the Saginaw lobe discharged westward (Fig. 19E) because a drainage way through the Paw Paw River Valley near Lawrence, Michigan, was open. The meltwater emptied into glacial Lake Chicago in the vicinity of St. Joseph, Michigan. The floodwaters that emanated from the Saginaw lobe appear to have funneled into the head of the main channel subglacially from tunnel channels supplying water from some yet unknown source. The floodwater scoured bedrock and continued uphill along undulating flow paths until exiting the ice at a major vent in the SKM northeast of Battle Creek. Floodwater was redirected to the northwest by a near surface bedrock valley at Battle Creek. Along this area large volumes of buried and stagnating Saginaw ice are indicated by hummocky topography of the moraine and irregular channel flanks as enormous flows continued through this ice marginal position. High-relief hummocky topography is present along much of the region between the Saginaw Kalamazoo Moraine and Augusta suggesting the possibility of a broad region of buried ice along the CKRV. From Galesburg to the mouth near Plainwell the channel margins are more in keeping with those of glacial-lake spillways resulting from normal subaerial erosion in a proglacial setting (Fig. 19F).

It seems only logical that some of the initial Saginaw outburst flow dispersed southward down the central lowlands prior to intercepting gulleys incised into the floor of the former eastward draining spillway through the LMKM. However, once piracy occurred and the westerly course was initiated, the predominately sandy soils of the region posed little challenge to the erosive power these flows must have had. The enormous CKRV channel is the product of a channel adjusting to the immense discharge of outburst flooding from underneath the Saginaw lobe and a lower base level west of the LMKM. Where the channel emptied into the lake basin near Plainwell it formed a coarse debris fan as cataclysmic flow deviated south and west through proglacial lake basins and subsequently lower outlets near Lawrence, Michigan, rather than cutting through the Kendall/ Valparaiso systems to the northwest as it does today. Coincidentally, the elevation to which the channel is incised near Lawrence approaches the elevation of the Glenwood level of glacial Lake Chicago (Hansel et al., 1985). After the initial outburst flood, sporadic discharges with high aggradation rates most likely continued down this drainage course partially infilling the large channels until the northwesterly segment of the modern Kalamazoo River pirated the southwesterly flowing course. Stream piracy of the northwestern branch of the Kalamazoo River most likely resulted from rapid downcutting and expansive headward erosion due to large base level drops within the Lake Michigan basin such as the Chippewa Lowstand at 9.8 ka BP (Hansel et al., 1985). Winters and Rieck (1982) and Bird (pers. comm., 2001) have shown evidence that the northwest segment of the Kalamazoo River Valley is coincident with a bedrock valley. The bearing this circumstance has on piratization of the ancestral CKRV is unclear at this time. 3.3. Ice dynamics If our hypothesis is correct about a pressurized subglacial origin for the CKRV, then it has implications for past ice dynamics in the region. If the CKRV formed by an outburst flood while the Saginaw lobe was at the Kalamazoo position and drainage was to the Michigan Basin it would require that the Lake Michigan lobe had retreated a minimal distance of 35 km to the Lake Border Moraine or perhaps even further back into the Lake Michigan basin. The cross-cutting relative age relationships of the younger Kalamazoo River outburst event are constrained by the 15.5 ka BP date of Monaghan (1990) for ice at the Saginaw lobe Kalamazoo Moraine, and the 14,000 ka BP date of Hansel et al. (1985) for the Lake Michigan lobe Lake Border Moraine. This requires that the Saginaw lobe remained stationary at the Kalamazoo position while the Lake

ARTICLE IN PRESS A.L. Kozlowski et al. / Quaternary Science Reviews 24 (2005) 2354–2374

2371

Michigan lobe was actively retreating westward across Van Buren County. The younger relative age of the Kalamazoo River outburst event in association with the stagnant Saginaw lobe suggests that either the Saginaw lobe persisted into southern Michigan longer than previously thought or that the ages assigned to the Lake Border and/or Kalamazoo Moraines in southern Michigan are questionable.

reorganization as proposed by Brennand and Shaw (1994) or alternatively they may indicate contribution of englacial and supraglacial meltwater from stagnating ice. Further work is clearly needed to understand the tunnel channel esker relationship.

3.4. Summary

In conclusion, morphologic and stratigraphic evidence strongly suggests that the unique morphology and anomalous flow path of the CKRV is the result of an outburst flood from the Saginaw lobe. Further, the CKRV and larger Kalamazoo System is a remarkable example of the role meltwater has on the development of glacial drainage systems that transition from subglacial, to ice marginal and proglacial environments. The tunnel channels that feed this system as well as others in the region have only recently been recognized and need further study. Yet it is clear that subglacial drainage was a powerful mechanism in shaping the landscape near the headwaters of the CKRV and establishing drainage routes in southwest Michigan. The source of meltwater for subglacial outbursts remains an important subject that can only be resolved by further study. However, discoveries of more subglacial lakes under the Antarctic Ice sheet (Siegert and Bamber, 2000) indicate that their presence is not as uncommon as was once thought. Further, researchers investigating large Pleistocene channels systems in the US Midwest (Wright, 1973; Clayton et al., 1999) have suggested that rapid drainage of subglacial lakes may provide a viable mechanism to explain the supply of meltwater necessary to generate high magnitude floods capable of large-scale erosion. In Michigan the Saginaw Lowland (Fig. 4) seems a likely candidate for a subglacial reservoir capable of generating pressurized meltwater necessary to cut tunnel channels that fed the outburst flood responsible for the CKRV. This study has documented the morphologic and stratigraphic evidence associated with the drainage history of the Kalamazoo River system and the early influence Saginaw drainage may have had on the development of regional drainage. Several other large channels exist west of the LMKM that appear to have drainage histories just as complicated as those described in this paper. Further study is needed to resolve these drainage events in order to more adequately understand the effects drainage may have had on glacial Lake Chicago and the Michigan Basin.

In summary, morphological and stratigraphic evidence in conjunction with cross-cutting relationships of tunnel channels suggest that the CKRV formed as a result of a transient, high magnitude outburst flood originating from under the Saginaw lobe. The anomalous flow path through the LMKM was most likely controlled by a gap resulting from drainage of an earlier proglacial lake, lateral meltwater streams, or from melting ice in the Kalamazoo Moraine. The eastward drainage event incised the gap to an elevation of 260 m. Subsequent westward retreat of Lake Michigan lobe ice from the Kalamazoo Moraine lead to declines in base level of proglacial lakes enhancing downcutting and headward erosion further facilitating a preferential flow path through the moraine. Stored subglacial meltwater under the Saginaw lobe must have reached some threshold value and converged toward the margin of the Saginaw lobe via pressurized tunnel channels. The configuration of the outburst vent may have been controlled by ice fracturing as suggested by Roberts et al. (2000) or similar to ice walled outlets (Russell et al., 1999) for jo¨kulhlaups in Iceland. Highrelief hummocky topography between the Saginaw Kalamazoo Moraine and east of Battle Creek, in combination with irregular channel margins suggests that buried ice was likely present. From Augusta to Plainwell, Michigan, the channel morphology and associated features are consistent with proglacial spillways of the upper Great Plains that resulted from subaerial catastrophic floods. High magnitude westward discharge exploited the opening through the moraine previously created and scoured out the oversized channel to compensate for the immense discharges that must have been present. These discharges emptied into proglacial lake basins behind the Lake Michigan Kalamazoo Moraine and proceeded south and west through the Paw Paw River Valley outlet near Lawrence and continued until emptying into glacial Lake Chicago in the vicinity of present-day St. Joseph, Michigan. The channel appears to have incised down to an elevation approaching the Glenwood level. The effect of such a discharge on lake levels of glacial Lake Chicago is unknown at this time. The eskers within the tunnel channels near the head of the CKRV may represent some kind of hydraulic

4. Conclusions

Acknowledgements We would like to thank the following agencies for financial support provided during this project: The

ARTICLE IN PRESS 2372

A.L. Kozlowski et al. / Quaternary Science Reviews 24 (2005) 2354–2374

Geological Society of America, NASA’s Michigan Space Grant Consortium, USGS EDMAP program, Kalamazoo Foundation and the Graduate College of Western Michigan University. In addition, we would like to thank Steve Brown for helpful discussions regarding the study area and the construction of DEMs. The manuscript was greatly improved thanks to helpful comments by Grahame Larson, Thom Wilch, and Timothy Fisher.

References Baker, V.R., 1973. Paleohydrology and sedimentology of lake Missoula flooding in eastern Washington. Geological Society of America Special Paper 144, 1–79. Beaney, C.L., 2002. Tunnel channels in southeast Alberta, Canada: evidence for catastrophic channelized drainage. Quaternary International 90, 67–74. Beukema, S.P., 2003. Stratigraphy of Lake Michigan lobe deposits in Van Buren County, Michigan. Unpublished M.S. Thesis, Western Michigan University. Bird, B.C., Kozlowski, A.L., Kehew, A.E., 2000. Interpretations of 3dimensional glacial mapping of the Paw Paw and Decatur 7.5 minute quadrangles, Michigan. Geological Society of America Abstracts with Programs 32 (7), 19–20. Bird, B.C., Kozlowski, A.L., Kehew, A.E., 2001. Three-dimensional glacial mapping of the Lawrence 7.5 minute quadrangle, Van Buren County, southwest Michigan. Geological Society of America Abstracts with Programs 33 (6), 317. Boulton, G.S., 1987. A theory of drumlin formation by subglacial sediment deformation. In: Menzies, J., Rose, J. (Eds.), Drumlin Symposium. Balkema, Rotterdam, pp. 25–80. Brennand, T.A., Shaw, J., 1994. Tunnel channels and associated landforms, south-central Ontario: their implications for ice-sheet hydrology. Canadian Journal of Earth Sciences 31, 505–522. Bretz, J.H., 1952. Glacial Grand River, Michigan. Papers of the Michigan Academy of Science Arts and Letters 38, 359–382. Catacosinos, P.A., Harrison III, W.B., Reynolds, R.F., Westjohn, D.B., Wollensak, M.S., 2001. Stratigraphic Lexicon for Michigan. Michigan Department of Environmental Quality Geological Survey Division and Michigan Basin Geological Society, vol. 8, 56pp. Clague, J.J., Evans, S.G., 1994. Formation and failure of natural dams in the Canadian Cordillera. Geological Survey of Canada Bulletin 464. Clayton, L., Moran, S.R., 1974. A glacial process-form model. In: Coates, D.R. (Ed.), Glacial Geomorphology, vol. 3. State University of New York, Binghamton, pp. 88–119. Clayton, L., Attig, J.W., 1987. Drainage of Lake Wisconsin near the end of the Wisconsin Glaciation. In: Mayer, L., Nash, D. (Eds.), Catastrophic Flooding—The Binghamton Symposia in Geomorphology: International Series, pp. 139–153. Clayton, L., Attig, J.W., Mickelson, D.M., 1999. Tunnel channels formed in Wisconsin during the last glaciation. In: Mickelson, D.M., Attig, J.W. (Eds.), Glacial Processes Past and Present: Geological Society of America Special Paper 337, pp. 69–82. Cutler, P.M., Colgan, P.M., Mickelson, D.M., 2002. Sedimentologic evidence for outburst floods from the Laurentide Ice Sheet margin in Wisconsin, USA: implications for tunnel channel formation. Quaternary International 90, 23–40. Dodson, R.L., 1985. Topographic and sedimentary characteristics of the union streamlined plain and surrounding morainic areas. Unpublished Ph.D. Thesis, Michigan State University.

Dodson, R.L., 1993. Reinterpretation of the northwest portion of the Tekonsha Moraine, south-central Michigan. Physical Geography 13, 139–154. Ekblaw, G.E., Athy, L.F., 1925. Glacial Kankakee torrent in northeastern Illinois. Geological Society of America Bulletin 36, 417–428. Evans, D.J.A., England, J., 1993. Geomorphological evidence of Holocene climate change from northwest Ellesmere Island, Canadian high arctic. The Holocene 2, 148–158. Eyles, N., Boyce, J.I., Barendregt, R.W., 1998. Hummocky moraine: sedimentary record of stagnant Laurentide Ice Sheet lobes resting on soft beds. Sedimentary Geology 123, 163–174. Farrand, W.R., 1982. Quaternary geology of Southern Michigan. State of Michigan Department of Natural Resources. Map Scale 1:500,000. Finkbeiner, K., 1994. Subsurface glacial geology of the area between the Tekonsha and Kalamazoo Moraines, Kalamazoo County, Michigan. Unpublished M.S. Thesis, Western Michigan University. Fisher, T.G., Taylor, L.D., 1997. Subglacial tunnel valley sedimentation, Saginaw lobe. Geological Society of America Abstracts with Programs 29 (4), 15. Fisher, T.G., Taylor, L.D., 1999. Glacial landforms and sediment landform assemblages, Saginaw Lobe. In: Brown, S.E., Fisher, T.G., Kehew, A.E., Taylor, L.D. (Eds.), Guidebook for the 45th Midwest Friends of the Pleistocene Field Conference. Indiana Geological Survey, Bloomington, 82pp. Fisher, T.G., Taylor, L.D., 2002. Sedimentary and stratigraphic evidence for subglacial flooding south-central Michigan, USA. Quaternary International 90, 87–115. Fisher, T.G., Taylor, L.D., Jol, H.M., 2003. Boulder-gravel hummocks and wavy basal till contacts: product of subglacial meltwater flow beneath the Saginaw Lobe, south-central Michigan, USA. Boreas 32, 328–336. Fisher, T.G., Jol, H.M., Lahners, A.M., 2005. Saginaw Lobe tunnel channels and their significance in south-central Michigan, USA., this volume. Gardner, C.R., 1997. Lithologic and stratigraphic analysis of glacial diamictons, Sturgis, Michigan. Unpublished M.S. Thesis, Western Michigan University. Ham, N.R., Attig, J.W., 1996. Ice wastage and landscape evolution along the southern margin of the Laurentide Ice Sheet, northcentral Wisconsin. Boreas 25, 172–186. Hansel, A.K., Johnson, W.H., 1996. Wedron and Mason Groups: lithostratigraphic reclassification of deposits of the Wisconsin Episode, Lake Michigan Lobe Area. Illinois State Geological Survey Bulletin, 104pp. Hansel, A.K., Mickelson, D.M., Schneider, A.F., Larson, C.E., 1985. Late Wisconsinan and Holocene history of the Lake Michigan basin. In: Karrow, P.F., Calkin, P.E. (Eds.), Quaternary Evolution of the Great Lakes. Geological Association of Canada Special Paper 30, pp. 39–53. Havira, R.D., 1969. An application of statistical methods to crossbed features in the Kalamazoo Moraine. Unpublished Report, Western Michigan University. Johnson, M.D., 1999. Spooner Hills, northwest Wisconsin: high relief hills carved by subglacial meltwater of the Superior Lobe. In: Mickelson, D.M., Attig, J.W. (Eds.), Glacial Processes Past and Present: Geological Society of America Special Paper 337, pp. 83–102. Johnson, M.D., Mickelson, D.M., Clayton, L., Attig, J.W., 1995. Composition and genesis of glacial hummocks western Wisconsin, USA. Boreas 24, 98–116. Kehew, A.E., Lord, M.L., 1986. Origin and large-scale erosional features of glacial-lake spillways in the northern Great Plains. Geological Society of America Bulletin 97, 162–177.

ARTICLE IN PRESS A.L. Kozlowski et al. / Quaternary Science Reviews 24 (2005) 2354–2374 Kehew, A.E., Lord, M.L., 1987. Glacial-lake outbursts along the midcontinent margins of the Laurentide ice-sheet. In: Mayer, L., Nash, D. (Eds.), Catastrophic Flooding—The Binghamton Symposia in Geomorphology: International Series, pp. 95–120. Kehew, A.E., 1993. Glacial-lake outburst erosion of the Grand Valley, Michigan, and impacts on glacial lakes in the Lake Michigan Basin. Quaternary Research 39, 544–553. Kehew, A.E., Teller, J.T., 1994. Glacial-lake spillway incision and deposition of a coarse-grained fan near Waterous, Saskatchewan. Canadian Journal of Earth Science 31, 544–553. Kehew, A.E., Nicks, L., Smous, A., Gardner, C., Kendzursky, S., 1999a. Glacial Terrain Map of St. Joseph County, Michigan. Unpublished Map, U.S.G.S. STATEMAP Program. Kehew, A.E., Nicks, L.P., Straw, T.W., 1999b. Palimpsest tunnel valleys: evidence for relative timing of advances in an interlobate area of the Laurentide ice sheet. Annals of Glaciology 28, 47–52. Kehew, A.E., Bird, B.C., Kozlowski, A.L., 2001. Origin of the Valparaiso Moraine, Van Buren County, Michigan. Geological Society of America Abstracts with Programs 33 (6), 268. Kehew, A.E., Bird, B.C., Beukema, S.P., 2002. Glacial Terrain Map of Van Buren, County Michigan. Unpublished Maps, U.S.G.S. STATEMAP Program. Kehew, A.E., Beukema, S.P., Bird, B.C., Kozlowski, A.L., 2005. Fast flow of the Lake Michigan Lobe: evidence from sediment-landform assemblages in southwestern Michigan, USA. Journal of Quaternary Science Reviews, this volume. Komar, P.D., 1984. The Lemniscate loop — comparisons with the shapes of streamlined landforms. Journal of Geology 92, 133–146. Kozlowski, A.L., 1999. Three dimensional mapping of the East Leroy and Union City 7.5 minute quadrangles in Southwest Michigan. Unpublished M.S. Thesis, Western Michigan University. Kozlowski, A.L., 2004. Origin of the Central Kalamazoo River Valley, Southwest Michigan, USA. Unpublished Ph.D. Thesis, Western Michigan University. Kozlowski, A.L., Kehew, A.E., Bird, B.C., 2001. An outburst hypothesis for the origin of the Kalamazoo river valley, Kalamazoo and Allegan counties, Michigan. Geological Society of America Abstracts with Programs 33 (7), 269. Kozlowski, A.L., Kehew, A.E., Bird, B.C., 2003. Tunnel channels of the Saginaw Lobe, southern Michigan. In: Proceedings, 34th Annual Binghamton Geomorphology Symposium, Binghamton, NY. Leverett, F., Taylor, F.B., 1915. The Pleistocene of Indiana and Michigan and the history of the Great Lakes. United States Geologic Survey Monograph 53, 0. Lord, M.L., 1991. Depositional record of a glacial-lake outburst: Glacial Lake Souris, North Dakota. Geological Society of America Bulletin 103, 290–299. Lord, M.L., Kehew, A.E., 1987. Sedimentology and paleohydrology of glacial-lake outburst deposits in southeastern Saskatchewan and northwestern North Dakota. Geological Society of America Bulletin 99, 663–673. Maizels, J.K., 1991. Origin and evolution of Holocene sandurs in areas of Jokulhlaup drainage, south Iceland. In: Maizels, J.K., Caseldine, C. (Eds.), Environmental Change in Iceland: Past and Present, pp. 267–300. Malde, H.E., 1968. The catastrophic late Pleistocene Bonneville flood in the Snake River Plain, Idaho. US Geological Survey Professional Paper 595. Martin, H.M., 1955. Map of the surface formations of the southern Peninsula of Michigan. Michigan Department of Conservation Geological Survey Division, Publication 49, Scale 1:500,000. Menzies, J., 1987. Towards a general hypothesis on the formation of drumlins. In: Menzies, J., Rose, J. (Eds.), Drumlin Symposium. Balkema, Rotterdam, pp. 9–24.

2373

Monaghan, G.W., 1990. Systematic variation in clay-mineral composition of till sheets: evidence for the Erie Interstade in the Lake Michigan Basin. In: Schneider, A.F., Fraser, G.S. (Eds.), Late Quaternary History of the Lake Michigan Basin. Geological Society of America Special Paper 251, pp. 43–50. Monaghan, G.W., Larson, G.J., 1986. Late Wisconsinan drift stratigraphy of the Saginaw ice lobe in south-central Michigan. Geological Society of America Bulletin 97, 324–328. Monaghan, G.W., Larson, G.J., Forstat, D.W., Sorenson, H.O., 1983. Selected geological maps of Kalamazoo County. Michigan Department of Natural Resources, Geological Survey Division. Monaghan, G.W., Larson, G.J., Gephart, G.D., 1986. Late Wisconsin drift stratigraphy of the Lake Michigan lobe in southwestern Michigan. Geological Society of America Bulletin 97, 329–334. Mooers, H.D., 1989. On formation of the tunnel valleys of the Superior lobe, central Minnesota. Quaternary Research 32, 24–35. Morner, N.A., Dreimanis, A., 1973. The Erie Interstade. In: Black, R., Goldthwait, R., Willman, H. (Eds.), The Wisconsinan Stage. Geological Society of America Memoir 136, pp. 107–135. Munro, M.J., Shaw, J., 1997. Erosional origin of hummocky terrain in south-central Alberta, Canada. Geology 25, 1027–1030. O’Connor, J.E., 1993. Hydrology, hydraulics, and geomorphology of the Bonneville flood. Geological Society of America Special Paper 274. Patterson, C.J., 1994. Tunnel-valley fans of the St. Croix moraine, east-central Minnesota, USA. In: Warren, W.P., Croot, D.G. (Eds.), Formation and Deformation of Glacial Deposits. Balkema, Rotterdam, pp. 69–87. Roberts, M.J., Russell, A.J., Tweed, F.S., Knudsen, O., 2000. Ice fracturing during jokulhlaups: implications for englacial floodwater routing and outlet development. Earth Surface Processes and Landforms 25, 1429–1446. Russell, A.J., Knudsen, O., Fay, H., Marren, P.M., Heinz, J., Tronicke, J., 1999. Morphology and sedimentology of a giant supraglacial ice-walled, jokulhlaup channel, Skeidararjokull, Iceland: implications for esker genesis. Global and Planetary Change 28, 193–216. Schumm, S.A., 1977. The Fluvial System. Wiley, New York. Shaw, J., Sharpe, D.R., 1987. Drumlin formation by subglacial meltwater erosion. Canadian Journal of Earth Sciences 24, 2316–2322. Shaw, J., Kvill, D.R., Rains, R.B., 1989. Drumlins and catastrophic subglacial floods. Sedimentary Geology 62, 177–202. Shaw, J., Rains, R.B., Eyton, R., Weissling, L., 1996. Laurentide subglacial outburst floods: landform evidence from digital elevation models. Canadian Journal of Earth Sciences 33, 1154–1168. Shoemaker, E.M., 1991. On the formation of large subglacial lakes. Canadian Journal of Earth Sciences 28, 1975–1981. Siegert, M.J., Bamber, J.L., 2000. Subglacial water at the heads of Antarctic ice-stream tributaries. Journal of Glaciology 46, 702–703. Sjogren, D.B., Fisher, T.G., Taylor, L.D., Jol, H.M., Munro-Stasiuk, M.J., 2002. Incipient tunnel channels. Quaternary International 90, 41–56. Stalker, A., Mac, S., 1960. Ice-pressed drift forms and associated deposits in Alberta. Geological Survey of Canada Bulletin 57. Straw, T.W., 1976. Guidebook for a Field Trip on Some Aspects of the Glacial Geology in the Kalamazoo Area, North-central Section of the Geological Society of America Field Guide, vol. 28, Western Michigan University. Sugden, D.E., Denton, G.H., Marchant, D.R., 1991. Subglacial meltwater channel systems and ice sheet overriding, Asgard Range, Antarctica. Geografiska Annaler 73A, 109–121. Sugden, D.E., Marchant, D.R., Potter Jr., N., Souchez, R.A., Denton, G.H., Swisher III, C.C., Tison, J.L., 1995. Preservation of Miocene glacier ice in East Antarctica. Nature 376, 412–414.

ARTICLE IN PRESS 2374

A.L. Kozlowski et al. / Quaternary Science Reviews 24 (2005) 2354–2374

Taylor, L.D., Fisher, T.G., Jol, H.M., Okraszewski, M.A., 1998. Stratigraphic evidence for a late Pleistocene glacial outburst flood sequence in south-central Michigan. Geological Society of America, North-Central Section, Program with Abstracts 32 (4), 74. Teller, J.T., Thorleifson, L.H., 1987. Catastrophic flooding into the Great Lakes from Lake Agassiz. In: Mayer, L., Nash, D. (Eds.), Catastrophic Flooding—The Binghamton Symposia in Geomorphology: International Series, pp. 121–138. Terwilliger, F.K., 1954. The glacial geology and groundwater resources of Van Buren County, Michigan. Occasional Papers for 1954, Michigan State Geological Survey, vol. 95.

Winters, H.A., Rieck, R.L., 1982. Drainage reversals and transverse relationships of rivers to moraines in southern Michigan. Physical Geography 3, 70–82. Wright Jr., H.E., 1973. Tunnel valleys, glacial surges, and subglacial hydrology of the Superior Lobe, Minnesota. In: Black, R., Goldthwait, R., Willman, H. (Eds.), The Wisconsinan Stage. Geological Society of America Memoir 136, pp. 251–276. Zumberge, J.H., 1960. Correlation of Wisconsin drifts in Illinois, Indiana, Michigan, and Ohio. Geological Society of America Bulletin 71, 1177–1188.