The Sequence and Timing of Large Late Pleistocene Floods from Glacial

Home , Glacial Lake Columbia, Lake Hitchcock, Lake Missoula, Missoula floods

Quaternary Science Reviews 31 (2012) 67e81

Contents lists available at SciVerse ScienceDirect

Quaternary Science Reviews

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

The sequence and timing of large late Pleistocene ﬂoods from glacial Lake Missoula

Michelle A. Hanson a,*, Olav B. Lian b, John J. Clague a a Department of Earth Sciences, Simon Fraser University, 8888 University Drive, Burnaby, British Columbia, Canada V5A 1S6 b Department of Geography, University of the Fraser Valley, 33844 King Road, Abbotsford, British Columbia, Canada V2S 7M8 article info abstract

Article history: Glacial Lake Missoula formed when the Purcell Trench lobe of the Cordilleran ice sheet dammed Clark Received 24 March 2011 Fork River in Montana during the Fraser Glaciation (marine oxygen isotope stage 2). Over a period of Received in revised form several thousand years, the lake repeatedly filled and drained through its ice dam, and floodwaters 28 October 2011 coursed across the landscape in eastern Washington. In this paper, we describe the stratigraphy and Accepted 10 November 2011 sedimentology of a significant new section of fine-grained glacial Lake Missoula sediment and compare Available online 30 November 2011 this section to a similar, previously described sequence of sediments at Ninemile Creek, 26 km to the northwest. The new exposure, which we informally term the rail line section, is located near Missoula, Keywords: Glacial Lake Missoula Montana, and exposes 29 units, each of which consists of many silt and clay couplets that we interpret to fi Outburst floods be varves. The deposits are similar to other ne-grained sediments attributed to glacial Lake Missoula. Stratigraphy Similar varved sediments overlie gravelly flood deposits elsewhere in the glacial Lake Missoula basin. Sedimentology Each of the 29 units represents a period when the lake was deepening, and all units show evidence for Varves substantial draining of glacial Lake Missoula that repeatedly exposed the lake floor. The evidence Optical dating includes erosion and deformation of glaciolacustrine sediment that we interpret happened during draining of the lake, desiccation cracks that formed during exposure of the lake bottom, and fluvial sand deposited as the lake began to refill. The floods date to between approximately 21.4 and 13.4 cal ka ago based on regional chronological data. The total number of varves at the rail line and Ninemile sites are, respectively, 732 and 583. Depending on lake refilling times, each exposure probably records 1350e1500 years of time. We present three new optical ages from the rail line and Ninemile sites that further limit the age of the floods. These ages, in calendar years, are 15.1 0.6 ka at the base of the Ninemile exposure, and 14.8 0.7 and 12.6 0.6 ka midway through the rail line exposure. The sediment at the two sections was deposited during later stages of glacial Lake Missoula, after the largest outburst events. Ó 2011 Elsevier Ltd. All rights reserved.

1. Introduction a critical depth, the east end of the ice dam became buoyant, tunnels opened beneath the dam, and the lake drained cata- The Purcell Trench lobe of the Cordilleran ice sheet filled the strophically (Waitt, 1985). The maximum discharge of the largest basin now occupied by Lake Pend Oreille, Idaho, during the Fraser flood was at least 17 million m3 s 1 (Baker, 1973; Clarke et al., 1984; Glaciation (late Wisconsinan; marine oxygen isotope stage 2). The O’Connor and Baker, 1992). Floodwaters flowed into glacial Lake ice lobe blocked Clark Fork River and impounded glacial Lake Columbia to the west, depositing a thick layer of gravel, sand, and Missoula in the intermontane basins of western Montana (Fig. 1; silt on top of laminated silty glaciolacustrine sediments (Fig. 1; Pardee, 1910, 1942). At its maximum, glacial Lake Missoula had Atwater, 1984, 1986, 1987). Floodwaters also flowed across the a volume of 2200e2600 km3 and a depth at the ice dam of Columbia Plateau, eroding Miocene basalt and Pleistocene loess 610e640 m (Pardee, 1942; Bretz, 1969; Clarke et al., 1984; Craig, (Fig. 1; Bretz, 1923, 1925, 1928). The floodwaters formed giant flood 1987; Smith, 2006). When water in glacial Lake Missoula reached bars and current ripples, deposited fine-grained sediments in some Columbia River tributaries in southern Washington, and recharged Columbia River Basalt aquifers with isotopically depleted water * Corresponding author. Present address: Saskatchewan Geological Survey, Sas- katchewan Ministry of Energy and Resources, 200-2101 Scarth Street, Regina, (Bretz, 1925, 1928, 1929; Bretz et al., 1956; Baker, 1973; Waitt, 1980, Saskatchewan, Canada S4P 2H9. Tel.: þ1 306 798 4211; fax: þ1 306 787 1284. 1985; Brown et al., 2010). They also backed up southward into E-mail addresses: [email protected], [email protected] (M.A. Hanson), Willamette Valley of western Oregon and deposited slackwater [email protected] (O.B. Lian), [email protected] (J.J. Clague).

Fig. 1. Study area showing locations mentioned in the text. The figure shows the maximum extent of the Cordilleran ice sheet as well as the maximum extent of glacial Lake Columbia. LCFR: Lower Clark Fork River; MV: Mission Valley; NC: Ninemile Creek; RL: Rail line. Darker gray areas were swept by floodwaters from glacial Lake Missoula. (Figure adapted from Waitt, 1985; Atwater, 1986; Levish, 1997). sediments there (Allison, 1978; Waitt, 1980, 1985; O’Connor et al., bedded slackwater flood sediments in southern Washington to these 2001). Within a few weeks, the floodwaters had reached the drainings and thus inferred that the drainings were catastrophic. mouth of Columbia River, over 700 km from their source, where Baker and Bunker (1985), however, argued that the slackwater beds plumes of freshwater spread out into the northeast Pacific Ocean are not proof that each draining was catastrophic. Levish (1997) (Baker, 1973; Craig, 1987; O’Connor and Baker, 1992; Lopes and Mix, studied exposures of rhythmically bedded sediments similar to 2009; Denlinger and O’Connell, 2010). Turbidity currents triggered those at Ninemile in Mission Valley, Montana (Fig. 1). He inferred 66 by the floods travelled an additional 1100 km southward along the sedimentation units, each of which is recorded by approximately floor of the eastern North Pacific Ocean (Zuffa et al., 2000), while 30e60 varves that thin up-section and, collectively, represent a plume of finer sediment travelled northward along the coast 3240e3610 years of deposition in glacial Lake Missoula. He found no (Gombiner et al., 2010). Glacial Lake Missoula may have drained evidence of unconformities between the units and concluded that and refilled approximately 90 times during the Fraser Glaciation there was no evidence for repeated drainage of glacial Lake Missoula. (Waitt, 1985; Atwater, 1986). Most recently, Smith (2006) described large gravel flood bars that Most of the evidence for drainings of glacial Lake Missoula has underlie rhythmic silt and clay beds in several places along the lower come from study of flood deposits in Idaho, Washington, and Oregon Clark Fork River (Fig. 1). He attributed these features to catastrophic (Bretz,1925,1928; Bretz et al.,1956; Baker,1973; Allison,1978; Waitt, draining of glacial Lake Missoula and the finer beds at the Ninemile 1980, 1985; Atwater, 1984, 1987; O’Connor et al., 2001). In contrast, site to shallower, later stages of the lake when drainings were less relatively little research has been done in the glacial Lake Missoula energetic (Smith, 2006; Alho et al., 2010). basin itself. Pardee (1910, 1942) described the extent of the lake and This paper describes the stratigraphy and sedimentology of an landforms that led him to conclude that very large and rapid currents important new section of rhythmically stratified sand, silt, and clay were involved in the draining of the lake. Chambers (1971, 1984) in the heart of the glacial Lake Missoula basin and compares it to the studied a section of cyclic, rhythmically bedded glaciolacustrine Ninemile section. The new section, which we refer to as the rail line sand, silt, and clay units exposed in a highway cut near Ninemile section, provides evidence for several tens of substantial drainings Creek, Montana (Fig. 1). Most of the units1 at the Ninemile exposure of glacial Lake Missoula. These drainings repeatedly exposed the comprise a thick layer of very fine sand and silt that grades upward lake floor in the Clark Fork River valley at Missoula. Three new into rhythmically laminated silt and clay beds that Chambers (1971, optical ages, two from the rail line site and one from the Ninemile 1984) interpreted to be varves. The silteclay couplets thin upward site, better constrain the timing of glacial Lake Missoula flooding. within units that are separated by unconformable contacts. Thin beds of pebble gravel, weathering, and small frost wedges at the top 2. Section locations, stratigraphy, and sedimentology of 22 of the 40 units led Chambers (1971, 1984) to conclude that glacial Lake Missoula drained at least 22 times to below 959 m above The Ninemile section is a 250-m-long exposure along Interstate sea level (asl), which is 306e321 m below the highest elevation Highway 90 near the confluence of Ninemile Creek and Clark Fork reached by the lake. Waitt (1980, 1985) ascribed rhythmically River in western Montana (N 47.0, W 114.4; Fig. 1). The base and top of the section are, respectively, 936 m and 959 m asl. The base of the section is approximately 281 m above the base of the former ice 1 In this paper, we define “unit” as an ensemble of lithofacies recording a single dam, 200 km to the west. The rail line section is a 1.4-km-long, lake phase from filling through maximum stage to drainage. northwest-trending exposure along the former track of the Chicago M.A. Hanson et al. / Quaternary Science Reviews 31 (2012) 67e81 69

Milwaukee St. Paul and Pacific Railroad west of the city limits of 2.1. Ninemile Creek section Missoula and approximately 26 km upstream of the Ninemile section (N 46.9, W 114.1; Fig.1). The base and top of the section are, The Ninemile Creek section comprises 34 units of rhythmically respectively, 963 m and 975 m asl. The highest glacial Lake Missoula stratified, graded beds of fine sand, silt, and clay (Figs. 2 and 3A) shorelines around Missoula are 1265e1280 m asl, 290e315 m above and, at the base of the section, at least 1 m of coarse to fine sand and the top of the rail line section (Pardee, 1942; Smith, 2006). pebble-cobble gravel. The uppermost 3 units were too weathered

Fig. 2. Stratigraphic log of the Ninemile Creek section, Montana. 70 M.A. Hanson et al. / Quaternary Science Reviews 31 (2012) 67e81 to describe in detail. The sediments comprise three lithofacies: 2.1.2. Sandesilt facies a gravelesand facies, a sandesilt facies, and a silteclay facies. The The sandesilt facies is present in each of the 34 rhythmically sequence of lithofacies in most units, from bottom to top, is grav- stratified units and comprises very fine sand or silt (Figs. 2 and 3D). elesand, sandesilt, and silteclay facies. Individual occurrences of the facies range from 9e220 cm thick. Lower contacts are commonly sharp and planar or undulating; 16 of 2.1.1. Gravelesand facies the contacts are clearly erosional. Some of the lower contacts, This facies comprises thin beds of coarse sand and, in some however, are loaded, producing convoluted bedding above or cases, granule to medium pebble gravel. It lies at the base of nine of below the contact (Figs. 2 and 4A). the 34 units (Figs. 2 and 3B) and ranges from less than 1e20 cm The sandesilt facies is characterized by planar or undulating thick; the thinnest beds are discontinuous. Most gravel clasts are laminae ranging from <1e4 mm thick. About half of the strata of subrounded to subangular. The gravelesand facies also fills wedge- the sandesilt facies contain ripples (Figs. 2 and 3D); in one case, the shaped cracks at the top of the five lowest units (Fig. 3C). ripples are overturned. Both type-A and type-B ripples are present

Fig. 3. Rhythmically bedded sand, silt, and clay units at the Ninemile section. (A) One unit comprising sandesilt facies (below) and silteclay facies (above). The contact between the two facies is gradational. Unit is 320 cm thick. Varves thin upward as the silt content decreases. (B) Gravelesand facies at the base of a unit. Sand is w6 cm thick. (C) Wedge-shaped feature at the top of a unit, inﬁlled with gravel and sand. Feature is w15 cm long. (D) Type-A and type-B climbing ripples in sandesilt facies. (E) Loaded silt laminae and water escape structures in sandesilt facies. M.A. Hanson et al. / Quaternary Science Reviews 31 (2012) 67e81 71

(Jopling and Walker, 1968). Most ripples indicate flow to the internal laminations than thicker couplets. Near the top of three of northwest; in one instance, flow was to the southeast (Fig. 2). Other the units, couplets are separated by anomalously thick (1.5e5 cm) common sedimentary structures in the sandesilt facies are loaded silt beds (Fig. 4D). The three silt beds have sharp, planar lower and laminae and beds, diapirs, dish structures, and flame structures upper contacts, and include planar and undulating laminae and (Figs. 2, 3E, 4A, and B). Thin (w2 mm) clay laminae are present ripples. locally within the facies, commonly near the top and generally associated with laminated silt. 2.1.4. Correlation with Chambers’ work Chambers (1971, 1984) counted six more units at the Ninemile 2.1.3. Silteclay facies site than we did, although he was only able to describe 32 in detail; The silteclay facies is also present in each of the 34 rhythmically the remaining units were at the top of the exposure and were too stratified units and comprises sets of silt and clay couplets that weathered to describe in detail. Based on unit thickness, elevation, resemble varves (Figs. 2 and 4C). The number of couplets per and sedimentological characteristics, such as the presence of the rhythmically stratified unit ranges from two to 44 and averages 19. sand-gravel facies and ripple-drift sequences in the silt, we are able, Couplets are 0.3e20 cm thick. Within each unit, couplet thickness with few exceptions, to correlate our units with Chambers’ units. decreases upward from an average of 4.3e0.8 cm (Fig. 3A). The Just above our 10-m level, Chambers identifies one additional unit. lower part of each couplet consists of massive silt with few clay We are confident, however, in our decision to include his thin unit laminae; the upper part is clay, in some cases with thin silt inter- (18.5 cm thick) within the unit that underlies it. Above the 17.5-m laminae (Fig. 4C). Lower contacts of individual couplets are sharp level, correlation is more difficult. Chambers’ 22nd unit from the but the boundary between the silt layer and overlying clay layer in base is 1.5 m thick, which does not match the thickness of any of our each couplet is gradational (Fig. 4C). The thickness of the clay layer units. A possible explanation for this discrepancy is that Chambers’ changes little upward within a unit; thin couplets have a higher 22nd unit represents two of our units, which occur between 17.4 m ratio of clay to silt than thick couplets and the thinnest ones are and 18.6 m. Chambers also described one more unit than we were approximately 80 percent clay. Thinner couplets also have fewer able to near the top of the exposure, below the weathered zone.

Fig. 4. Sedimentary features at the Ninemile section. (A) Loaded lower part of unit with diapiric structure in gravelesand facies and dish structures in the sandesilt facies. (B) Flame structures within lower part of sandesilt facies. (C) Nine varves with sharp lower contacts, gradational internal contacts, and internal laminations. (D) Event beds at the top of silteclay facies. 72 M.A. Hanson et al. / Quaternary Science Reviews 31 (2012) 67e81

Finally, above the 17.5-m level, Chambers also has one occurrence of sand and gravel, whereas we identiﬁed six. A possible explanation for the discrepancy is the discontinuous nature of this facies. Chambers’ total varve count is 716; ours is 583. His counts are higher for nearly all units that we can conﬁdently correlate. We offer two explanations for this discrepancy: (1) weathering of the upper part of the exposure over the past several decades did not allow us to observe all of the varves in the uppermost units; (2) we erred by grouping varves into composite varves, or Chambers erred by subdividing composite varves. The latter explanation is more likely in the sandesilt facies, where thin clay laminae might be interpreted to be part of a composite varve or the top of a non- composite varve.

2.2. Rail line section

We studied the rail line section in detail where the exposed sediments are thickest (11.8 m; Fig. 5). All but the lowest sediments are well exposed on the northeast-facing side of the railway cut; the lowest sediments were documented on the southwest-facing side of the cut. The exposed sediments include 29 upward-thinning units of rhythmically bedded ﬁne sand, silt, and clay (Figs. 5 and 6A). A zone of highly disturbed sediments approximately 6 m above the base of the exposure is thicker than the units directly above and below it (Fig. 5). Judging from the average thickness of adjacent units, we estimate that the disturbed zone includes 3 units, bringing the total count to 31. Units consist of three lithofacies similar to those at the Ninemile site. The three lithofacies, in the order in which they generally occur from bottom to top, are the sand, silt, and silteclay facies. We describe the facies in detail because some of their characteristics differ from those at the Ninemile site.

2.2.1. Sand facies The sand facies, which resembles the gravelesand facies at the Ninemile site, occurs below 10 units, all in the lower half of the exposure (Fig. 5). The facies comprises well-sorted, very fine to fine sand (Fig. 6A and B). The lower contact of each of the 10 beds is erosional, and most contacts are planar (Fig. 6B). The facies ranges in thickness from <1 cm to 6 cm; the thinner occurrences are discontinuous. The sand is normally graded, commonly rippled at the top, and, in some cases, laminated at the base. Type-A ripples, commonly containing silt stringers or clay rip-ups that were eroded from underlying sediments, indicate flow to the northeast (Fig. 5). The thick sand at the base of the third lowest unit (Fig. 5)is different in character from the sand layers at the base of the other 9 units. It is 64 cm thick and mostly massive, but with some laminations and type-A ripple-drift cross-laminae in the upper 9 cm of the bed.

2.2.2. Silt facies The silt facies is limited to two beds, 14 cm and 35 cm thick, in the two lowest units (Fig. 5). The two silt beds resemble most of the beds of the sandesilt facies at the Ninemile site. Lower contacts are sharp and erosive. The facies consists of thinly laminated, coarse to Fig. 5. Stratigraphic log of the rail line section, Montana. See Fig. 2 for legend. ﬁne silt with type-B ripple-drift cross-laminae (Fig. 7A). A few thin (w2 mm) clay laminae occur within silt beds near the top of the facies. The number of couplets per unit ranges from 15 to 42 and 2.2.3. Silteclay facies averages 24. Couplets range in thickness from 0.4e12 cm. Couplet The dominant facies at the rail line section is silteclay couplets thickness decreases upward within each unit from an average of similar to those at the Ninemile site. In most of the units, the 3.4e0.7 cm (Fig. 6A). The lower contact of each couplet is sharp silteclay facies either drapes the sand facies (Fig. 6B) or uncon- (Fig. 6D). Each couplet comprises a lower lamina or bed of formably overlies silteclay couplets of the next lower unit (Fig. 6C). massive silt, in some cases with a few thin laminae of clay near M.A. Hanson et al. / Quaternary Science Reviews 31 (2012) 67e81 73

Fig. 6. Rhythmically bedded sand, silt, and clay units at the rail line section. Arrows indicate unit contacts. (A) A sedimentary unit comprising sand facies (below) and silteclay facies (above). Varves thin upward as the silt content decreases. Square shows location of photo B. (B) Sand facies with type-A ripples and an erosional lower contact. Silt and clay varves are draped over ripples. (C) The basal sediment of the upper unit unconformably drapes the sediments of the unit below. The uppermost 2 cm of each unit were disturbed during draining of the lake. (D) Silt and clay varves. Note clay laminations within silt beds. Contacts between couplets and internal contacts are sharp and planar. Undulation of laminations is caused by post-depositional desiccation. the top, and an upper clay cap (Fig. 6D). Most contacts between laterally in a down-lake direction; and (5) couplets are found at all the massive silt and the clay cap are sharp (Fig. 6D); only in a few elevations within the lake basin (Ashley, 1975, 2002; Smith and cases are they gradational. The clay caps range in thickness from Ashley, 1985). 0.2e0.7 cm and appear to be normally graded. Few of the clay caps contain silt laminae. As at the Ninemile site, the ratio of the 2.2.4. Boundaries between units thickness of the clay cap to that of the associated silt layer The contacts between successive units at the rail line section increases upward within units; the uppermost couplets are are unconformable. Even where the basal sand facies is absent, approximately 80 percent clay. Thinner couplets also have fewer there is evidence of erosion of up to seven varves at the top of internal laminations. The uppermost 15 cm of ten of the units are each unit. Tilted and broken varves are also present in the upper brecciated and include rip-ups of silt and clay couplets (Figs. 5 few centimetres of many units. These broken varves form small and 6C). These disturbed zones have a sharp and probably concave-up features 2e3 cm wide and less than 2 cm deep erosive lower contact. (Fig. 6A and B). Eight units are characterised by small, Silteclay couplets in other glacial Lake Missoula sediments downward-tapering clastic wedges that do not cross the contact identical to those described here have been interpreted to be varves with the overlying unit (Figs. 5 and 7B). These wedge-shaped by Chambers (1971,1984), Waitt (1980), Fritz and Smith (1993), and structures are filled with silt and clay derived from the adja- Levish (1997). We interpret these silteclay couplets, as well as cent sediment; in some cases brecciated fragments of varves those as the Ninemile site, to be varves. The silteclay couplets meet occur within the casts (Fig. 7B). The cracks extend down to five criteria of varves: (1) they are thin, regular, fine-grained, and 26 cm from unit contacts. The tops of the cracks range from commonly contain no current structures; (2) there is little or no 0.5e15 cm wide (Fig. 7B). Varves in the adjacent sediment are fining in the silt layer, whereas the clay cap fines upward; (3) the bent downward at the margins of the cracks, and micro-faults in thickness of the clay cap changes little from couplet to couplet, sediment adjacent to the largest crack parallel the crack walls unlike the thickness of the silt layer; (4) couplets thin and fine (Fig. 7B). 74 M.A. Hanson et al. / Quaternary Science Reviews 31 (2012) 67e81

Fig. 7. Unique characteristics of sediments at the rail line section. (A) Laminated and rippled silt facies with clay laminations. Arrow indicates lower contact. (B) Largest wedge- shaped crack, measuring 15 by 26 cm. Crack is inﬁlled with adjacent varved silt and clay. Varves are downturned toward the crack. Micro-faulting can be seen to the right of the wedge-shaped crack (squares). A secondary similar feature from the overlying unit crosscuts the main feature (arrows).

3. Optical dating separated quartz grains were mounted on 9.8-mm-diameter aluminum discs using silicone oil as an adhesive. In principle, each Optical dating provides an estimate of the time elapsed since aliquot should be as small as possible, but in practice aliquot size is deposition and burial of quartz or feldspar grains (see Lian and dependent on the sensitivity of the quartz grains. For our samples, Roberts, 2006 and Wintle, 2008 for detailed descriptions of the each aliquot contained 50 to 100 grains, therefore it is likely that method). For this work, quartz was prepared from two samples of the measured luminescence was dominated by that originating the sand facies at the rail line section (samples MAH-3 and MAH-5) from fewer than 10 grains (Duller et al., 2000). and from one sample of the gravelesand facies at the Ninemile section (sample MAH-8). These facies were chosen for dating based 3.2. Environmental dose rate determination on our interpretation that they were deposited in shallow fluvial environments on the exposed floor of the lake between ponding The radiation dose absorbed by the sample grains after depo- events and before lake refilling (see Section 4.1). sition comes from alpha (a), beta (b), and gamma (g) radiation produced by decay of U, Th, their daughter products, and 40K. The 3.1. Sample collection and preparation radiation absorbed by the mineral grains of interest originates in the surrounding sediment and within the sampled grains them- We collected samples from horizons free of gravel indicative of selves; there is also a contribution from cosmic rays. Samples used high-energy deposition and avoided sediments showing evidence for dosimetry were dried, milled, and sent to a commercial labo- of post-depositional disturbance that could have mixed grains of ratory for neutron-activation analysis to determine 40K, U, and Th different ages. Each sample was collected by inserting an aluminum contents; results are presented in Table 1. From these data, dose cylinder (w40 cm long and 10 cm in diameter) into a cleaned face; rates (in Gy ka 1) caused by g and b radiation were calculated using the sampled sediment included both the target sand bed and standard formulae (Aitken, 1985; Berger, 1988; Lian et al., 1995) and bounding silt and clay. The cylinder then was excavated and its ends the dose rate conversion factors of Adamiec and Aitken (1998; were filled with fabric and taped to keep the contents immobile Table 2). In each case, we assumed that the U and Th decay chains and to retain water content. have been in secular equilibrium since deposition of the sediment. In the laboratory, the outer 0.5e1 cm of the sediment from each This assumption generally has been found to be valid for well- end of the cylinder was removed under controlled lighting condi- drained mineral sediments (Prescott and Hutton, 1995; Olley tions. Some of this sediment was dried and used for dosimetry and et al., 1996). Pore water in the sediment matrix attenuates or the remainder was discarded. The sand was then separated from absorbs radiation differently from mineral matter, and therefore it the surrounding silt and clay for optical dating. All sand samples has to be accounted for in the dose rate calculations. When the were treated with hydrochloric acid to dissolve carbonates and sediments were overlain by a lake, it is likely they were saturated then sieved to isolate the 180e250 mm size fraction. Samples then with water. The sediments, however, have spent the vast majority were treated with concentrated hydrofluoric acid for 40 min to of the time since deposition as well-drained subaerial deposits, dissolve feldspar grains and to etch from quartz grains outer thus we used the as-collected water contents with an appropriate material that had been affected by alpha radiation in the environ- uncertainty to account for any reasonable variance (Table 1). For ment. Quartz grains were separated from heavier minerals using samples near the ground surface, cosmic-ray radiation can a sodium polytungstate solution with a density of 2.68 g ml 1. The contribute significantly to the environmental dose rate, but its M.A. Hanson et al. / Quaternary Science Reviews 31 (2012) 67e81 75

Table 1 Sample depths, K, U, and Th contents, and water content.

Sample Site d (m)a K (%)b U(mg/g)b Th (mg/g)b Dw c MAH-3 Rail line 7.6 3.61 0.17 3.06 0.12 13.0 0.5 0.172 0.017 MAH-5 Rail line 8.5 3.46 0.18 3.27 0.13 13.4 0.5 0.172 0.017 MAH-8 Ninemile Creek 2.9 3.29 0.16 3.18 0.13 13.0 0.5 0.115 0.012

a d: sample depth beneath the ground surface. b K, U, and Th contents were determined using neutron-activation analysis of a dried and milled subsample of the bulk material sampled for dating. c Dw: water content ¼ water mass/mineral mass. For each sample, the as-collected water content was used in the dose rate calculation, with an uncertainty that is thought to account for any reasonable variation at 2s.

effect diminishes rapidly with burial depth (Prescott and Hutton, 1994). For the samples dated here, cosmic-ray radiation consti- tutes less than three percent of the total dose rate (Table 2).

3.3. Equivalent dose determination

Luminescence measurements were made using a Risø TL/OSL DA-20 reader. The separated quartz grains were stimulated using light-emitting diodes that delivered 45 mW cm 2 of blue (470 20 nm) light to the sample. Ultraviolet emissions (w350 nm) were detected by an Electron Tubes Ltd. 9235QB photomultiplier tube placed behind a 7.5-mm-thick Hoya U-340 optical filter. Laboratory irradiations were applied with a calibrated 90Sr/90Y b- source mounted on the reader that delivered 6.38 0.12 Gy min 1 to fine sand quartz grains mounted on an aluminum substrate. The equivalent dose (De; Table 2) of each sample was determined using the standard single-aliquot regenerative dose (SAR) protocol of Murray and Wintle (2000) but with the inclusion of an infrared wash to remove any luminescence resulting from K-feldspar contamination (Olley et al., 2004; Wintle and Murray, 2006; see Fig. 8 for details), which typically was found to be less than one percent of that measured from the quartz fraction. A zero-dose point was included to assess the severity of preheat-induced thermal transfer and one of the regenerative-dose points was repeated to assess the efficacy of the test dose to correct for sensitivity change. Of the 73, 49, and 48 aliquots prepared for samples MAH-3, MAH-5, and MAH-8, respectively, 58, 40, and 35 aliquots were accepted. For each sample, De values determined using SAR were plotted on radial plots (Galbraith et al., 1999; Galbraith, 2010), which display the distributions of De values about mean values (Fig. 9). The De values for the samples are distributed approximately evenly about their respective mean values and thus show no obvious discrete De populations. Therefore, a representative De used in the calculation of sample burial age was determined using the central age model (CAM) of Galbraith et al. (1999). This model determines the weighted De and its associated standard error and takes into account over-dispersion of the data, which is the variation between Fig. 8. (A) Flow chart showing the modified single-aliquot regenerative dose (SAR) individual De values above and beyond that associated with method used to date the rail line and Ninemile samples. Li is the luminescence measured after administration of a regenerative dose, Di. In this case the regenerative doses (i ¼ 1 to 5) were 15.9, 31.8. 63.6, 127.2, and 0 Gy; the 0 Gy data point is used to Table 2 check for thermal transfer resulting from the preheat, which was 220 C for 10 s. An Dose rates, equivalent doses, and optical ages. aliquot was rejected if the luminescence measured at the 0 Gy step was more than 5 _ a _ b c d Sample DcðGy=kaÞ DTðGy=kaÞ De (Gy) Optical age (ka) percent of that of the natural. Ti is the luminescence measured following a test dose of 12.7 Gy and a cut-heat of 160 C. For each aliquot, the 15.9 Gy dose (arbitrary choice) MAH-3 0.067 0.003 4.38 0.3 65.0 2.0 14.8 0.7 was repeated to determine the effectiveness of the test dose for correcting for sensi- MAH-5 0.060 0.003 4.32 0.3 54.6 1.4 12.6 0.6 tivity change; only aliquots that had a recycling ratio within 10 percent of unity were MAH-8 0.129 0.006 4.45 0.3 67.2 1.1 15.1 0.6 accepted. (B) Example of a luminescence decay curve for an aliquot of sample MAH-3, a _ Dc : dose rate due to cosmic rays, calculated using present burial depths and the which is typical of those generated from natural aliquots of the three samples. The relationship of Prescott and Hutton (1994). initial measured signal is dominated by the thermally stable fast component, which b _ DT : total dose rates (that due to cosmic rays plus that due to a, b, and g radi- decays to about 10 percent of its initial intensity in w2 s. The inset graph shows dose- ation). Only an internal dose rate applies for a radiation, and an estimate of 0.03 Gy/ response data for this aliquot, determined from the sensitivity corrected luminescence ka was used (Olley et al., 2004). (Li/Ti) measured over the first 0.4 s of stimulation, minus that recorded during the last c De: equivalent dose found using the central age model. 20 s. The data are fitted with a saturating exponential curve, onto which the natural d Apparent optical ages and their analytical uncertainties at 1s. luminescence is interpolated in order to determine the equivalent dose (De). 76 M.A. Hanson et al. / Quaternary Science Reviews 31 (2012) 67e81

analytical uncertainties (Galbraith et al., 1999; Lian and Roberts, 2006; Jacobs and Roberts, 2007). Even in cases where aliquots consisted only of grains that were exposed to sufficient sunlight prior to burial, over-dispersion values can be as high as 20 percent (see references cited in Lian and Roberts, 2006, p. 2459). All of the samples that we analyzed have over-dispersion values of approximately 20 percent (sample MAH-3) or much less than 20 percent (samples MAH-5 and -8; Fig. 9). These relatively low values suggest that the aliquots used to determine De had not been significantly contaminated with older grains that were insufficiently exposed to sunlight prior to burial. The results are consistent with our interpretation that the sands were deposited in a shallow, non-turbid fluvial environment (see Section 4.1). It is possible, although unlikely, that the size of the aliquots has masked the presence of younger or older grain populations. When one cannot be certain that the luminescence signal measured from the sampled grains was properly reset prior to burial or when the nature of the distribution of De values suggests the presence of more than one age population, it is common to apply the minimum age model (MAM) of Galbraith et al. (1999). MAM calculations were performed on all three samples, and in each case the MAM yielded De values that are consistent with those found using the CAM. These results support our supposition that the sampled quartz grains were exposed to adequate sunlight before burial. To test whether or not the SAR protocol used here is appropriate for our quartz samples, we performed dose recovery experiments (Roberts et al., 1999; Murray and Wintle, 2003) on 24 aliquots each of samples MAH-3 and MAH-8. Aliquots were exposed to natural sunlight for 10 min to remove the natural light-sensitive luminescence signal and then each aliquot was given a laboratory radiation dose similar to the De determined previously using the CAM. We then attempted to recover the laboratory dose using the same SAR protocol that we used to determine natural De values. In both cases, the recovered dose was at least 99 percent of the applied dose.

3.4. Optical ages

The two samples collected from the rail line section yielded burial ages, in calendar years, of 14.8 0.7 ka (MAH-3) and 12.6 0.6 ka (MAH-5; Table 2). The samples were taken from directly below the 5th and 7th units of glacial Lake Missoula sediments, numbered from the base of the section (Fig. 5). The single sample from the Ninemile section (MAH-8) returned an age of 15.1 0.6 ka (Table 2). It was collected directly below the 1st unit of glacial Lake Missoula sediments, as numbered from the base of the exposure (Fig. 2).

4. Discussion

4.1. Sedimentology

The rhythmic units at the rail line site are similar to those at the Ninemile site, as well as to those in Mission Valley described by Levish (1997), except for the lack of unconformities between units in Mission Valley. Thin gravelesand and sand beds separate rhythmic units in the Ninemile and rail line sections. These are likely the gravel-ﬁlled channels described by Chambers (1971, 1984). Such sediments can be deposited by underﬂows, but based

Fig. 9. Radial plots showing the distribution of equivalent doses determined from consistent with the mean at a 95 percent confidence interval. De values that showed aliquots of samples (A) MAH-3, (B) MAH-5, and (C) MAH-8. More precise data plot poor recycling ratios or unacceptably high thermal transfer (recuperation) were not farther from the origin. In each case, the gray bands, which are generated by the radial included in the CAM calculations. The uncertainty in the weighted mean De used in the plot software, are centered on the weighted mean equivalent dose (De) values provided age calculation is provided by CAM. Over-dispersion values are 25 4, 16 2, and by the central age model (CAM). De values that plot within the gray bands are 9.5 1.3 percent for samples MAH-3, -5, and -8, respectively. M.A. Hanson et al. / Quaternary Science Reviews 31 (2012) 67e81 77 on evidence for exposure of the underlying unit (see below), we the cracks implying secondary infill; and (3) micro-faults occur in interpret the gravel and sand to be non-channelized fluvial sedi- adjacent sediment (Harry and Gozdzik, 1988; Murton and French, ments deposited in shallow water on the floor of the lake before 1993; Gao, 2005). Chambers (1971, 1984) described similar cracks refilling; Chambers (1971) reached a similar conclusion at the at the Ninemile site that he interpreted to be frost wedges filled Ninemile site. with the gravelesand facies. Shaw et al. (2000) did not observe Thick, laminated and rippled basal silt beds are present cracks at Ninemile, but argued that they could be dewatering throughout the Ninemile section but were found only in the two structures associated with loading caused by rapid deposition of lowest units at the rail line section. At Ninemile, these beds have the overlying sandesilt units. The cracks at Ninemile, however, are been interpreted by Shaw et al. (1999) as deposits of turbidity filled with the gravelesand facies, which nowhere appears currents generated by periodic jökulhlaups from beneath the substantial enough to load the underlying varved sediment. Simi- Cordilleran ice sheet in the Rocky Mountain Trench. This explana- larly at the rail line site, what we interpret to be ice-wedge casts are tion does not explain, however, why upwardly thinning varves filled with adjacent sediment, not overlying sediment as would be above this facies indicate a deepening lake over time after depo- expected from injection of liquefied sediment into fractures caused sition of the sand and silt. Nor does this explanation account for by rapid sedimentation. Unfortunately, due to the nature of the what we interpret as evidence for subaerial exposure (see below) exposure, we could not examine these features in plan view, which below this facies. The climbing ripple-drift sequences indicate bears on our interpretation of them as ice-wedge casts. If the a high influx of suspended sediment and high accumulation rates. wedges are ice-wedge casts, the lake floor must have been exposed Soft-sediment deformation structures, which are present at Nine- between units. mile but absent from the two similar units at the rail line section, The interpretation of the depositional environment of each of are evidence of rapid deposition causing differential loading and the facies and of the boundaries between units is subject to consequent dewatering of the sediments (Lowe and LoPiccolo, different interpretations, but considering all the evidence together, 1974; Rust, 1977). These characteristics, combined with the we believe we have made a credible case for repeated draining and erosive lower contact, leads us to conclude that these two thick silt filling of glacial Lake Missoula at the Ninemile and rail line sites. beds were likely deposited by turbidity currents but at relatively Desiccation in the upper few centimetres and possible ice-wedge shallow depths as glacial Lake Missoula began to fill, a conclusion casts at the tops of units indicate subaerial exposure of varved also reached by Chambers (1971, 1984) at the Ninemile site. units. Overlying laminated and rippled sand and gravel represent The sand and silt beds at both sites grade up into silteclay varves initial infilling of the lake; laminated and ripple-drift sand and silt that were deposited in deeper water. The varves thin upward sequences represent increased sedimentation in a shallow but within each unit, evidence of continued deepening of the lake over deepening lake. Lastly, varved sediment with couplets that time and increasing the distance from sediment sources. The decrease in thickness upward indicate that the lake deepened until general decrease in the number of varves per unit up-section at the ice dam failed, after which some of the lake sediment became each site is consistent with a thinning ice dam toward the end of disturbed and the lake bottom exposed. the Fraser Glaciation. The 1.5e5-cm-thick silt beds that interrupt Subaerial exposure at the top of every unit at the rail line section the thinning-upward sequence near the top of some units at indicates that glacial Lake Missoula drained or partially drained at Ninemile (Fig. 4D) are deposits of rare, high-energy events, possibly least 29 times. Each time, the lake lowered below 960 m asl. The turbidity currents. They are not separate, thin, lake-ponding units existence of possible ice-wedge casts in some units suggests that because they are abruptly and conformably overlain by varves and the lake floor was exposed for at least several years. This length of because the varves above the silt beds are similar in thickness and exposure is longer than one would expect from simple lake-level in clay-silt ratios to those directly below them. Deformed beds at fluctuations and is consistent with the fact that it might take the top of 10 units at the rail line section (Fig. 6C) are likely related several years after the lake drained for the lake to refill to the to the unconformity directly above. We exclude the possibility that elevation of this exposure. Twenty-two of the units at the Ninemile these deformed beds could be caused by rapid loading and sedi- site show evidence of subaerial exposure, including erosional mentation from underflows because these beds have erosive lower contacts, frost wedges, or overlying fluvial sand and gravel. This contacts and are not consistently overlain by the silt facies that evidence indicates that the majority of the lake drainings were could have produced the loading. These beds possibly result from significant enough to lower the lake to below 936 m asl. Loaded erosion and rapid re-deposition of saturated glaciolacustrine sedi- contacts between varves and overlying sandesilt beds at three of ment during draining of the lake. the contacts at Ninemile possibly support the argument that glacial We cannot conclusively correlate the rail line and Ninemile Lake Missoula did not drain completely each time the ice dam sections on the basis of their stratigraphy and sedimentology. Units failed. at the base of the rail line section, however, are similar to units The hypothesis of repeated draining of glacial Lake Missoula is throughout the Ninemile section, indicating that at least the lower consistent with the conclusions drawn by Chambers (1971, 1984) part of the rail line section possibly overlaps with the Ninemile and Waitt (1980) at the Ninemile site. It is contrary, however, to section, which also appears to be older based on our ages. If there is the conclusions of Levish (1997) based on his study of similar glacial no overlap between the two sections, then the 31 units at rail line Lake Missoula sediments in Mission Valley. He argued that these are missing from the top of the Ninemile section. In this scenario, it sediments show no evidence of repeated subaerial exposure and would be difficult to explain how earlier glacial Lake Missoula units that the silt beds at the base of rhythmic beds could be attributed to were preserved at the Ninemile site, but later ones were eroded, episodic surging of the Flathead lobe of the Cordilleran ice sheet, given that the later lake draining events probably were less erosive which increased sedimentation in Mission Valley. Levish did not because the ice sheet was thinning and the lakes were smaller. describe any features at the contacts between units that would Thus, there is likely some overlap between the two sites. indicate subaerial exposure, thus we assume that his interpretation The broken and tilted varves and the concave-up features at the of continuous sedimentation is correct. Levish’s interpretations are top of the rail line units probably record desiccation caused by sedimentologically plausible, but there are two other explanations subaerial exposure. We tentatively interpret the large wedge-like for rhythmic sedimentation with no evidence of subaerial exposure cracks as ice-wedge casts because: (1) they are infilled with adja- that would be more consistent with evidence of repeated glacial cent varved sediment; (2) adjacent varves are bent down toward Lake Missoula draining found elsewhere in the basin. First, it is 78 M.A. Hanson et al. / Quaternary Science Reviews 31 (2012) 67e81 possible that the Mission Valley sediments were deposited at 1985). Using this tephra and paleomagnetic secular variation a different time than the sediments at the Ninemile and rail line records from three slackwater sediment exposures in Washington, sites; Levish’s optical ages (19.2e16.0 ka ago) would seem to Clague et al. (2003) placed the period of glacial Lake Missoula support this interpretation. Levish argued on the basis of regional flooding between either 18.5 and 13.1 or 21.0 and 16.1 ka ago, topography, however, that no more than approximately 100 years depending on the age of the tephra and the number of years of sediment could be missing from the top of the record, implying between floods. Radiocarbon ages on turbidites in the North Pacific that the exposures represent the later phases of glacial Lake Mis- extend flooding to between 13.0 and 12.6 ka ago (Zuffa et al., 2000). soula, as do the sediments at the Ninemile and rail line sites. The last flood from glacial Lake Missoula, however, must have Second, Levish’s sites are in the deepest part of the glacial Lake occurred before approximately 13.7e13.4 ka ago (Kuehn et al., Missoula basin and thus conceivably remained inundated when 2009), when Glacier Peak tephras G and B blanketed parts of part of the lake drained. Clarke et al. (1984) considered this possi- Washington and Montana, including Mission Valley (Fryxell, 1965; bility when modelling glacial Lake Missoula outburst floods and Porter, 1978; Mehringer et al., 1984; Foit et al., 1993; Sperazza et al., found that not all models resulted in complete lake drainage. Some 2002; Hendrix et al., 2004; Kuehn et al., 2009; Hofmann and model runs left up to 22 percent of the water in the lake. A water Hendrix, 2010). Hofmann and Hendrix (2010) recovered a pine depth of greater than 220 m, approximately 35 percent the needle from a varved sediment sequence in Flathead Lake in maximum lake depth, would have inundated most of the exposures northern Montana; it yielded an age of 14,150 150 yr BP. They in Mission Valley (Levish, 1997; Smith, 2006). Significant fluctua- inferred that this varve sequence, which dates back to at least tions in water level at these sites, caused by partial drainage, may 14,475 150 yr BP, was deposited in a proglacial lake dammed by explain the cyclicity of the units and the rippled sand and silt facies. a terminal moraine, the Polson moraine, that was built after glacial In contrast, the rail line and Ninemile sites require a minimum lake Lake Missoula ceased to exist. depth of about 370 m, or approximately 60 percent of the The three new optical ages reported in this paper contribute to maximum lake depth, to remain inundated during a potential the discussion of the age and duration of glacial Lake Missoula and partial draining and thus would more likely be exposed during such its outburst floods. The stratigraphic positions of the optical ages events. are shown in Figs. 2 and 5. The ages indicate that the Ninemile Although there is strong evidence for repeated draining of section is slightly older than the rail line section, consistent with glacial Lake Missoula in both the rail line and Ninemile sections, we the stratigraphy, which shows that the lower beds at rail line are cannot conclusively demonstrate that these drainings were similar to most beds at the Ninemile site. The two ages at the rail complete or catastrophic. Excellent preservation of sediment at the line site do not differ statistically when their analytical uncer- two sites suggests that flow out of the basin at these localities did tainties are taken into account at two standard deviations, although not significantly erode the lake floor. Hydraulic modelling by Alho the mean ages are stratigraphically reversed. We see no reason to et al. (2010) indicates that the sediments at the rail line section favour one age over the other and thus adjust the age of the upper could be preserved during most lake drainings, even during higher- sample to that of the lower by assuming that w100 years separate energy floods. It is likely, however, that the sediments at Ninemile the two dated horizons. With this adjustment, the weighted mean would only be preserved during smaller, less energetic lake of the two ages, referenced to the lower sample, is 13.5 0.5 ka draining events (Alho et al., 2010), which is consistent with our (Fig. 5). optical ages that situate these exposures toward the end of the The main potential source of uncertainty in optical dating arises existence of glacial Lake Missoula (see Section 4.2). from the assumption that effective bleaching of the relevant electron traps resets the measured luminescence signal of all the grains immediately before final burial. Anomalously old ages would arise 4.2. Chronology if a significant number of electron traps had not been emptied before deposition and burial. In particular, dating of glaciolacus- Radiocarbon ages associated with the advance of the Cordilleran trine sediments can be problematic because the grains are ice sheet into eastern Washington, with glacial Lake Missoula flood commonly transported in turbid water (Berger et al., 1987; Berger, sediments in Oregon, with low salinity anomalies inferred from 1988; Lian and Hicock, 2001). Our strategy was to avoid this freshwater diatom abundances off the southern Oregon coast that problem by dating sand deposited in fluvial environments on the have been attributed to glacial Lake Missoula flooding, and with exposed floor of glacial Lake Missoula between ponding events. Our glacial Lake Missoula sediment in the North Pacific all place the supposition was that the sand grains would be well exposed to onset of glacial Lake Missoula flooding at or after 23.0 to 19.0 ka sunlight in this environment before being buried. Rendell et al. ago2 (Clague et al., 1980; Benito and O’Connor, 2003; Lopes and (1994) concluded that the optical signal from quartz and feldspar Mix, 2009; Gombiner et al., 2010). Based on varve counts and grains in relatively clear water approximately 10 m deep is thor- optical dating of the clay portion of glacial Lake Missoula varves in oughly reset in 3 h. If, however, the water is turbid, as is common in Mission Valley, Levish (1997) concluded that the lake existed from proglacial environments, the optical signal might not be reset. approximately 19.2e16.0 ka ago. Similarly, on the basis of a radio- Ditfelsen (1992) found that feldspar grains only 75 cm deep in carbon age and varve counts, Atwater (1986) argued that the 89 turbid water are not reset after 20 h of exposure to light. The glacial Lake Missoula floods he identified occurred over a period of problem is exacerbated by the fact that fluvial grains are not always 2000e3000 years between 19.4 and 14.7 ka ago. Mount St. Helens equally reset e coarser grains are more likely to be reset than finer set S (Sg and So) tephra, dated to between 18.5 and 13.3 ka ago ones (Murray et al., 1995; Olley et al., 1998). The sand grains that we (Mullineaux et al., 1975, 1978; Crandell et al., 1981; Baker and dated, however, were probably deposited in water no deeper than Bunker, 1985; Waitt, 1985; Clynne et al., 2008) lies at the top of 10 cm. The sediments sampled for dating came from thin sheets, one flood bed, at least 11 flood beds from the top of slackwater rather than thick channel deposits, and were more likely deposited flood sediment sequences in southern Washington (Waitt, 1980, by non-channelized sheet flow on the exposed lake floor than in channels. The sands are well sorted and thus were probably deposited in relatively clear water. The radial plots (Fig. 9) do not 2 All calibrated radiocarbon ages discussed are reported directly from the source fi or have been calibrated with OxCal 4.1 using the IntCal09 calibration curve (Bronk display multiple age populations that would indicate a signi cant Ramsey, 2009; Reimer et al., 2009). number of grains had not been properly reset, as would be expected M.A. Hanson et al. / Quaternary Science Reviews 31 (2012) 67e81 79 in turbid water, thus reaffirming our interpretation of the deposi- lake to re fill to the elevations of the sections. Ice-wedge casts can tional environment of these facies. If these sediments had been commence growth within a year after a surface is exposed and deposited by turbidity currents, we would expect a larger spread expand up to 3 cm per year (Mackay and Burn, 2002), thus indi- (over-dispersion) in the single-aliquot De values. cating that some surfaces were exposed for a few years. It is possible that our optical ages underestimate the true burial Estimated refilling times for glacial Lake Missoula range from ages of the sediments of interest, in spite of the fact that our quartz w20 years to 110 years. Atwater (1986) counted the number of samples passed all quality-control tests mentioned above. Reports varves between successive glacial Lake Missoula flood beds in of this scenario are rare, but one example was provided recently by a sequence of glacial Lake Columbia sediments. The numbers Steffen et al. (2009) who attempted to date quartz collected from decrease from 45 to 55 near the base of the sequence to two alluvial fan and terrace sediments in Peru using a SAR protocol between the last two flood beds. Similarly, Waitt (1984, 1985) similar to the one that we used in this study. Steffen et al. (2009) counted 20e55 varves between glacial Lake Missoula flood beds concluded that their calculated quartz ages were too low when in three exposures in Idaho and Washington. Many of these counts compared to independent age information, probably due to the are minima for the number of years between floods because they influence of an unstable component in the measured luminescence do not include varves eroded by the floodwaters. Waitt (1985) signal. This unwanted component can be seen in their lumines- calculated a refilling time for a maximum glacial Lake Missoula of cence decay data, and it is especially evident when compared to less than 65 years. He based this estimate on the assumption that that measured from a standard, well-behaved quartz sample the rate of discharge into glacial Lake Missoula during the Fraser (Steffen et al., 2009, their Fig. 9). It is clear, however, that the Glaciation was twice the present-day rate of discharge into the luminescence decay data recorded from our samples (a represen- basin. He also estimated that it would take 40 years to fill glacial tative curve is shown in Fig. 8B) closely resembles that of the Lake Missoula to half its maximum volume. Waitt (1985) estimates standard quartz sample; both are dominated by the thermally of refilling times seem reasonable in light of the glacial Lake stable fast component over the first w2 s of stimulation. It is Columbia varve counts of Waitt (1984, 1985) and Atwater (1986). therefore unlikely that our samples suffer from this effect. Levish (1997) counted 20e107 varves per unit in glacial Lake Despite these favourable results, additional confidence in our Missoula sediments in Mission Valley; most units have 30e60 optical ages could be attained via three supplementary methods: varves. Assuming that the cyclicity of the units in Mission Valley (1) comparison of our ages with those derived from chemically and corresponds to the draining cycle of glacial Lake Missoula, Levish’s physically similar quartz sediment in the region for which the age data provide constraints on glacial Lake Missoula refilling times. has been determined previously by an independent method or Precise stratigraphic correlation of the rail line section and the methods; (2) the analysis of single grains of quartz, and subse- upper part of Levish's (1997) Landslide Bend section is not possible, quently looking for and eliminating data derived from grains that but based on our previous discussion, we argue that these expo- show abnormal luminescence behaviour (cf. Jacobs et al., 2006; sures record the same time interval. In an attempt to improve our Demuro et al., 2008). Additionally, if single-grain ages can be estimate of the age of the rail line sediments, we attempted to successfully determined, then the finite mixture model (Roberts correlate the 31 rail line units and the last 41 Landslide Bend units et al., 2000; Jacobs and Roberts, 2007; Arnold and Roberts, 2009) based on the number of varves per unit. In this scenario, all units at may be applied to distinguish age populations, if there are any, from the rail line site have fewer varves than corresponding units at the sample. Single-grain analysis may, however, be impractical due Landslide Bend owing to erosion during lake draining. Some gaps to the sensitivity of the quartz in the region; and (3) comparison of between units at rail line are required to make this scenario our quartz ages with those derived from the feldspar fractions of feasible. The last 41 units at Landslide Bend record 1348 years of the samples (cf. Steffen et al., 2009) after correction for anomalous glacial Lake Missoula sedimentation. If the overlap is valid, 616 fading. varves are missing from the rail line site, an average of 20 varves per The three optical ages lie within the range of ages that constrain unit. It is likely, however, that there are also missing units at the top the time when glacial Lake Missoula could exist (21.4e13.4 ka ago). of the rail line section because of erosion and weathering. Using They are near the young end of this age range, which is consistent this number and the weighted mean of the two optical ages, the with the stratigraphy (Smith, 2006). The rail line samples, for base and top of the rail line section would date to, respectively, example, are from beneath the 25th- and 27th-to-the-last glacial 13.7 0.5 and 12.4 0.5 ka ago. We make a similar provisional Lake Missoula sediment units there. match of the 34 Ninemile units with units 18e51, numbered from Three wells drilled on the surface adjacent to the top of the rail the top, at Landslide Bend, which record 1505 years of glacial Lake line section indicate that glacial Lake Missoula sediments extend Missoula sedimentation. Assuming this amount of time is recorded another 30e40 m below the base of the section (www.mbmggwic. at Ninemile, the top of that section would date to approximately mtech.edu). The units at rail line systematically increase in thick- 13.6 0.6 ka ago. ness downward and this trend probably continues below the base Our chronology accords with the known maximum age for the of the section. Thus a reasonable estimate of the number of units advance of the Cordilleran ice sheet into northern Idaho (21.4e19.6 below the base of the section, based on the average thickness of the ka; Clague et al., 1980) and with an age that pre-dates glacial Lake lowest 5 units (0.80 m), is 38e50, for a total of 69e81 units in the Missoula flooding in Oregon (23.3e22.3 ka; Benito and O’Connor, Missoula basin. These numbers are consistent with Levish (1997) 2003). It is consistent with radiocarbon ages on glacial Lake Mis- and Atwater (1986) estimates of, respectively, the number of soula sediment in the Pacific Ocean that indicate flooding occurred glacial Lake Missoula units and flood events. between 19.6 and 12.6 ka ago (Zuffa et al., 2000). Although on the The total number of varves exposed at the rail line section, based young side, our chronology is consistent with Smith's (2006) on the average count for each unit, adding a reasonable number of interpretation that the glacial Lake Missoula rhythmic silt and varves for the disturbed units in the middle of the section, and clay deposits date to the later stages of the lake. Waitt (1980, 1985) assuming that the lower silt facies capped by clay is one varve, is inferred, and Atwater (1987) attempted, bed-for-bed correlations 732. The total number of varves at the Ninemile section is 583. between beds at Ninemile and, respectively, the Touchet Beds and These are minimum estimates, because some varves probably were alternating glaciolacustrine and glacial Lake Missoula flood beds in eroded from the top of each unit and because they do not include glacial Lake Columbia. Based on the age of the set S tephra, known the length of time it took the ice dam to re-establish itself and the limits of the Cordilleran ice sheet advance and retreat, and the 80 M.A. Hanson et al. / Quaternary Science Reviews 31 (2012) 67e81 estimated period between floods, Waitt (1985) postulated that Thoughtful and careful reviews by V. Baker and an anonymous glacial Lake Missoula existed for 2000e2500 years between w18.8 reviewer greatly improved the paper. and 14.2 ka ago, which is fairly consistent with our chronology. Based on varve counts and constrained by a radiocarbon age in References glacial Lake Columbia sediment, Atwater (1986) proposed that the fl Adamiec, G., Aitken, M.J., 1998. Dose-rate conversion factors: update. Ancient TL 16, Ninemile beds correlate with ood beds in glacial Lake Columbia 37e49. that were deposited during the early half of his record (19.4ew17.1 Aitken, M.J., 1985. Thermoluminescence Dating. Academic Press, New York. ka), which is older than the inferred ages of our sections. Our Alho, P., Baker, V.R., Smith, L.N., 2010. Paleohydraulic reconstruction of the largest e chronology corresponds well with the younger of the intervals glacial Lake Missoula draining(s). Quaternary Science Reviews 29, 3067 3078. Allison, I.S., 1978. Late Pleistocene sediments and floods in the Willamette valley. proposed by Clague et al. (2003) based on paleomagnetic secular The Ore Bin 40, 177e191. and 193e202. variation correlations and different ages of the Mount St. Helens set Arnold, L.J., Roberts, R.G., 2009. Stochastic modelling of multi-grain equivalent dose S tephra (w18.5e13.1 ka). The inferred ages of the two exposures (De) distributions: implications for OSL dating of sediment mixtures. Quater- nary Geochronology 4, 204e230. are considerably younger than the optical ages of Levish (1997; Ashley, G.M., 1975. Rhythmic sedimentation in glacial Lake Hitchcock, Massachu- 19.2e16.0 ka) and Lopes and Mix (2009) chronology for low salinity setts-Connecticut. In: Jopling, A.V., McDonald, B.V. (Eds.), Glaciofluvial and anomalies in the Pacific Ocean that are likely due to inputs of water Glaciolacustrine Sedimentation. Society of Economic Paleontologists and e e Mineralogists Special Paper, vol. 23, pp. 304 320. from glacial Lake Missoula (19.0 17.0 ka). Levish dated glaciola- Ashley, G.M., 2002. Glaciolacustrine environments. In: Menzies, J. (Ed.), Modern and custrine sediments, and it seems likely that a significant number of Past Glacial Environments. Butterworth-Heinemann Publications, Oxford, grains in his samples were not sufficiently reset prior to burial, pp. 335e362. Atwater, B.F., 1984. Periodic floods from glacial Lake Missoula into the Sanpoil Arm leading to apparently old ages. Radiocarbon-dated material in the of glacial Lake Columbia, northeastern Washington. Geology 12, 464e467. cores analysed by Lopes and Mix (2009) could be considerably Atwater, B.F., 1986. Pleistocene glacial lake deposits of the Sanpoil River valley, older than the sediment in which it lies. Our chronology is also not northeastern Washington. U.S. Geological Survey Bulletin 1661, 1e39. Atwater, B.F., 1987. Status of glacial Lake Columbia during the last floods from glacial consistent with Hofmann and Hendrix's (2010) calibrated radio- Lake Missoula. Quaternary Research 27, 182e201. carbon age of 14,150 150 yr BP from glaciolacustrine sediments in Baker, V.R., 1973. In: Paleohydrology and Sedimentology of Lake Missoula Flooding Flathead Lake. It is possible, if not likely, that the needle they dated in Eastern Washington. Geological Society of America Special Paper, vol. 144. fl is older than the enclosing sediments. Notably, the age assigned to Baker, V.R., Bunker, R.C., 1985. Cataclysmic late Pleistocene ooding from glacial Lake Missoula: a review. Quaternary Science Reviews 4, 1e41. the top of the rail line section is younger than the most likely age for Benito, G., O’Connor, J.E., 2003. Number and size of last-glacial Missoula floods in Glacier Peak G and B tephras (13.7e13.4 ka; Kuehn et al., 2009) the Columbia River valley between the Pasco Basin, Washington, and Portland, e before which glacial Lake Missoula had ceased to exist. This Oregon. Geological Society of America Bulletin 115, 624 638. fl fi Berger, G.W., 1988. Dating Quaternary events by luminescence. In: Easterbrook, D.J. discrepancy may re ect our inability to precisely de ne the inter- (Ed.), Quaternary Sediments. Geological Society of America Special Paper, vol. vals represented by these two exposures because of the amount of 227, pp. 13e50. time missing from each unit due to erosion and lake refilling. The Berger, G.W., Clague, J.J., Huntley, D.J., 1987. Thermoluminescence dating applied to glaciolacustrine sediments in central British Columbia. Canadian Journal of sediment at the rail line site was deposited near the end of the Earth Sciences 24, 425e434. Fraser Glaciation. As the ice sheet began to retreat, a thinning ice Bretz, J.H., 1923. Glacial drainage on the Columbia Plateau. Geological Society of dam would have produced more frequent floods, decreasing the America Bulletin 34, 573e608. fl fl Bretz, J.H., 1925. The Spokane ood beyond the Channeled Scablands. Journal of ood recurrence interval and thus the number of years per unit Geology 33, 97e115. and 236e259. toward the top of the exposures. Unfortunately, there is no way to Bretz, J.H., 1928. Bars of Channeled Scabland. Geological Society of America Bulletin accurately model this decreasing recurrence interval. 39, 643e701. Bretz, J.H., 1929. Valley deposits immediately east of the Channeled Scabland of Washington. Journal of Geology 37, 393e427. and 505e541. fl 5. Summary Bretz, J.H., 1969. The Lake Missoula oods and the Channeled Scablands. Journal of Geology 77, 505e543. Bretz, J.H., Smith, H.T.U., Neff, G.E., 1956. Channeled Scabland of Washington: new The rail line section in Missoula, Montana, exposes 29 units of data and interpretations. Geological Society of America Bulletin 67, 957e1049. rhythmically bedded sand, silt, and clay similar to other deposits Bronk Ramsey, C., 2009. Bayesian analysis of radiocarbon dates. Radiocarbon 51, 337e360. that have been attributed to deposition in glacial Lake Missoula. Brown, K.B., McIntosh, J.C., Baker, V.R., Gosch, D., 2010. Isotopically-depleted late Each of the units records partial or complete emptying of glacial Pleistocene groundwater in Columbia River Basalt aquifers: evidence for fl Lake Missoula, and the entire sequence spans approximately 1350 recharge of glacial Lake Missoula oodwaters. Geophysical Research Letters 37. doi:10.1029/2010GL044992. years. The Ninemile section exposes 34 units of rhythmically Chambers, R.L., 1971. Sedimentation in glacial Lake Missoula. M.Sc. Thesis, Univer- bedded sand, silt, and clay and spans approximately 1500 years. sity of Montana, Missoula, MT. Recurrent draining and subaerial exposure of the lake floor at both Chambers, R.L., 1984. Sedimentary evidence for multiple glacial lakes Missoula. In: McBane, J.D., Garrison, P.B. (Eds.), Northwest Montana and Adjacent Canada. sites are indicated by unconformities, desiccation, possible ice- Montana Geological Society, Billings, pp. 189e199. wedge casts, and fluvial gravel and sand beds. Three new optical Clague, J.J., Armstrong, J.E., Mathews, W.H., 1980. Advance of the late Wisconsinan ages limit the time of marine oxygen isotope stage 2 glacial Lake Cordilleran ice sheet in southern British Columbia since 22,000 yr BP. Quater- fl nary Research 13, 322e326. Missoula ooding. Based on a tentative correlation between our Clague, J.J., Barendregt, R., Enkin, R.J., Foit Jr., F.F., 2003. Paleomagnetic and tephra two sites and Levish's (1997) Landslide Bend site, the Ninemile evidence fortens of Missoula floods in southern Washington. Geology 31,247e250. section spans the period 15.1 to 13.6 0.6 ka ago, and the rail line Clarke, G.K.C., Mathews, W.H., Pack, R.T., 1984. Outburst floods from glacial Lake e section spans the period 13.7 to 12.4 0.5 ka ago. Missoula. Quaternary Research 22, 289 299. Clynne, M.A., Calvert, A.T., Wolfe, E.W., Evarts, R.C., Fleck, R.J., Lanphere, M.A., 2008. The Pleistocene eruptive history of Mount St. Helens, Washington, from 300,000 to 12,800 years before present. In: Sherrod, D.R., Scott, W.E., Acknowledgements Stauffer, P.H. (Eds.), A Volcano Rekindled: The Renewed Eruption of Mount St. Helens, 2004e2006. U.S. Geological Survey Professional Paper, vol. 1750, e Our research was supported by the Natural Sciences and Engi- pp. 593 627. Craig, R.G., 1987. Dynamics of a Missoula flood. In: Mayer, L., Nash, D. (Eds.), Cata- neering Research Council (NSERC) of Canada through Discovery strophic Flooding. Allen and Unwin, London, pp. 305e332. grants to Clague and Lian and through an NSERC PGS B Scholarship Crandell, D.R., Mullineaux, D.R., Rubin, M., Spiker, E., Kelley, M.L., 1981. Radiocarbon to Hanson. M. Al Suwaidi, J. Dykstra, and C. Reiss assisted in the Dates from Volcanic Deposits at Mount St. Helens. U.S. Geological Survey Open- e fi File Report, Washington. 81 844. eld, and N. Cox, L. Day, and W. Moon assisted with analyses in the Demuro, M., Roberts, R.G., Froese, D.G., Arnold, L.J., Brock, F., Bronk Ramsey, C., 2008. Luminescence Dating Laboratory at University of the Fraser Valley. Optically stimulated luminescence dating of single and multiple grains of M.A. Hanson et al. / Quaternary Science Reviews 31 (2012) 67e81 81

quartz from perennially frozen loess in western Yukon Territory, Canada: Murray, A.S., Olley, J.M., Caitcheon, G.G., 1995. Measurement of equivalent doses in comparison with radiocarbon chronologies for the late Pleistocene Dawson quartz from contemporary water-lain sediments using optically stimulated tephra. Quaternary Geochronology 3, 346e364. luminescence. Quaternary Science Reviews 14, 365e371. Denlinger, R.P., O’Connell, D.R.H., 2010. Simulations of cataclysmic outburst floods Murton, J.B., French, H., 1993. Thaw modifications of frost-fissure wedges, Richards from Pleistocene glacial Lake Missoula. Geological Society of America Bulletin Island, Pleistocene Mackenzie Delta, western Arctic Canada. Journal of 122, 678e689. Quaternary Science 8, 185e196. Ditfelsen, C., 1992. Bleaching of K-feldspars in turbid water suspensions: Olley, J.M., Murray, A., Roberts, R.G., 1996. The effects of disequilibria in the uranium a comparison of photo- and thermoluminescence signals. Quaternary Science and thorium decay chains on burial dose rates in fluvial sediments. Quaternary Reviews 11, 33e38. Science Reviews 15, 751e760. Duller, G.A.T., Bøtter-Jensen, L., Murray, A.S., 2000. Optical dating of single sand- Olley, J., Caitcheon, G., Murray, A., 1998. The distribution of apparent dose as sized grains of quartz: sources of variability. Radiation Measurements 32, determined by optically stimulated luminescence in small aliquots of fluvial 453e457. quartz: implications for dating young sediments. Quaternary Geochronology 17, Foit Jr., F.F., Mehringer Jr., P.J., Sheppard, J.C., 1993. Age, distribution, and stratig- 1033e1040. raphy of Glacier Peak tephra in eastern Washington and western Montana, Olley, J.M., Pietsch, T., Roberts, R.G., 2004. Optical dating of Holocene sediments United States. Canadian Journal of Earth Sciences 30, 535e552. from a variety of geomorphic settings using single grains of quartz. Geomor- Fritz, W.J., Smith, G.A., 1993. Revisiting the Ninemile section: problems with relating phology 60, 337e358. glacial Lake Missoula stratigraphy to the Scablands-floods stratigraphy. Eos O’Connor, J.E., Baker, V.R.,1992. Magnitudes and implications of peak discharges from (Transactions American Geophysical Union) 74, 302. glacial Lake Missoula. Geological Society of America Bulletin 104, 267e279. Fryxell, R., 1965. Mazama and Glacier Peak volcanic ash layers: relative ages. Science O’Connor, J.E., Sarna-Wojcicki, A., Wozniak, K.C., Polette, D.J., Fleck, R.J., 2001. Origin, 147, 1288e1290. Extent, and Thickness of Quaternary Geologic Units in the Willamette Valley, Galbraith, R.F., 2010. On plotting OSL equivalent doses. Ancient TL 28, 1e9. Oregon. U.S. Geological Survey Professional Paper, vol. 1620. Galbraith, R.F., Roberts, R.G., Laslett, G.M., Yoshida, H., Olley, J.M., 1999. Optical Pardee, J.T., 1910. The glacial Lake Missoula, Montana. Journal of Geology 18, dating of single and multiple grains of quartz from Jinmium rock shelter, 376e386. northern Australia. Part I: experimental design and statistical models. Pardee, J.T., 1942. Unusual currents in glacial Lake Missoula, Montana. Geological Archaeometry 41, 339e364. Society of America Bulletin 53, 1570e1599. Gao, C., 2005. Ice-wedge casts in Late Wisconsinan glaciofluvial deposits, southern Porter, S.C., 1978. Glacier Peak tephra in the north Cascade range, Washington: Ontario, Canada. Canadian Journal of Earth Sciences 42, 2117e2126. stratigraphy, distribution, and relationship to Lateglacial events. Quaternary Gombiner, J., Hendy, I.L., Hemming, S.R., Fleisher, M.Q., Pierce, E., Mesko, G., Dale, Research 10, 30e41. C.L., 2010. Spatial and temporal variation of the last ice age mega-floods in the Prescott, J.R., Hutton, J.T., 1994. Cosmic ray contributions to dose rates for lumi- Pacific Northwest: sediment provenance using single-aliquot K/Ar dating. nescence and ESR dating: large depths and long-term time variations. Radiation American Geophysical Union 2010 Fall Meeting. American Geophysical Union, Measurements 23, 497e500. San Francisco, California, Abstract PP21A-1663. Prescott, J.R., Hutton, J.T., 1995. Environmental dose rates and radioactive disequi- Harry, D.G., Gozdzik, J.S., 1988. Ice wedges: growth, thaw transformation, and librium from some Australian luminescence dating sites. Quaternary Science palaeoenvironmental significance. Journal of Quaternary Science 3, 39e55. Reviews 14, 439e448. Hendrix, M.S., Bondurant, A.K., Timmerman, G., Salmon III, E.L., 2004. Preliminary Reimer, P.J., Baillie, M.G.L., Bard, E., Bayliss, A., Beck, J.W., Blackwell, P.G., Bronk results from geologic mapping and sedimentologic analysis of late Pleistocene Ramsey, C., Buck, C.E., Burr, G.S., Edwards, R.L., Friedrich, M., Grootes, P.M., glacial and post-glacial deposits, southern Flathead and northern Mission Guilderson, T.P., Hajdas, I., Heaton, T.J., Hogg, A.G., Hughen, K.A., Kaiser, K.F., valleys, western Montana. Geological Society of America Abstracts with Kromer, B., McCormac, F.G., Manning, S.W., Reimer, R.W., Richards, D.A., Programs 36, 287. Southon, J.R., Talamo, S., Turney, C.S.M., van der Plicht, J., Weyhenmeyer, C.E., Hofmann, M.H., Hendrix, M.S., 2010. Depositional processes and the inferred history 2009. IntCal09 and Marine09 radiocarbon age calibration curves, 0e50,000 of ice-margin retreat associated with the deglaciation of the Cordilleran ice years cal BP. Radiocarbon 51, 1111e1150. sheet: the sedimentary record from Flathead Lake, northwest Montana, USA. Rendell, H.M., Webster, S.E., Sheffer, N.L., 1994. Underwater bleaching of signals Sedimentary Geology 223, 61e74. from sediment grains: new experimental data. Quaternary Science Reviews 13, Jacobs, Z., Roberts, R.G., 2007. Advances in optically stimulated luminescence dating 433e435. of individual grains of quartz from archeological deposits. Evolutionary Roberts, R.G., Galbraith, R.F., Olley, J.M., Yoshida, H., Laslett, G.M., 1999. Optical Anthropology 16, 210e223. dating of single and multiple grains of quartz from Jinmium rock shelter, Jacobs, Z., Duller, G.A.T., Wintle, A.G., 2006. Interpretation of single grain De northern Australia: part II, results and implications. Archaeometry 41, 365e395. distributions and calculation of De. Radiation Measurements 41, 264e277. Roberts, R.G., Galbraith, R.F., Yoshida, H., Laslett, G.M., Olley, J.M., 2000. Dis- Jopling, A.V., Walker, R.G., 1968. Morphology and origin of ripple-drift cross-lami- tinguishing dose populations in sediment mixtures: a test of single-grain nation, with examples from the Pleistocene of Massachusetts. Journal of Sedi- optical dating procedures using mixtures of laboratory-dosed quartz. Radia- mentary Petrology 38, 971e984. tion Measurements 32, 459e465. Kuehn, S.C., Froese, D.G., Carrara, P.E., Foit Jr., F.F., Pearce, N.J.G., Rotheisler, P., 2009. Rust, B.R., 1977. Mass flow deposits in a Quaternary succession near Ottawa, Can- Major- and trace-element characterization, expanded distribution, and a new ada: diagnostic criteria for subaqueous outwash. Canadian Journal of Earth chronology for the latest Pleistocene Glacier Peak tephras in western North Sciences 14, 175e184. America. Quaternary Research 71, 201e216. Shaw, J., Munro-Stasiuk, M., Sawyer, B., Beaney, C., Lesemann, J.-E., Musacchio, A., Levish, D.R., 1997. Late Pleistocene Sedimentation in Glacial Lake Missoula and Rains, B., Young, R.R., 1999. The Channeled Scabland: back to Bretz? Geology 27, Revised Glacial History of the Flathead Lobe of the Cordilleran Ice Sheet, 605e608. Mission Valley, Montana. Ph.D. Thesis, University of Colorado, Boulder, CO. Shaw, J., Munro-Stasiuk, M., Sawyer, B., Beaney, C., Lesemann, J.-E., Musacchio, A., Lian, O.B., Hicock, S.R., 2001. Lithostratigraphy and limiting optical ages of the Rains, B., Young,R.R., 2000. TheChanneledScabland: back to Bretz?Geology 28, 576. Pleistocene fill, Fraser valley, near Clinton, south-central British Columbia. Smith, L.N., 2006. Stratigraphic evidence for multiple drainings of glacial Lake Canadian Journal of Earth Sciences 38, 839e850. Missoula along the Clark Fork River, Montana, USA. Quaternary Research 66, Lian, O.B., Roberts, R.G., 2006. Dating the Quaternary: progress in luminescence 311e322. dating of sediments. Quaternary Science Reviews 25, 2449e2468. Smith, N.D., Ashley, G., 1985. Proglacial lacustrine environment. In: Ashley, G.M., Lian, O.B., Hu, J., Huntley, D.J., Hicock, S.R., 1995. Optical dating studies of Quater- Shaw, J., Smith, N.D. (Eds.), Glacial Sedimentary Environments. Society of nary organic-rich sediments from southwestern British Columbia and north- Economic Paleontologists and Mineralogists Short Course, vol. 16, pp. 135e216. western Washington State. Canadian Journal of Earth Sciences 32, 1194e1207. Sperazza, M., Thomas, G., Hendrix, M.S., Moore, J.N., 2002. Record of late Pleistocene Lopes, C., Mix, A.C., 2009. Pleistocene megafloods in the northeast Pacific. Geology through Holocene climate change in a regional lake system: Flathead Lake basin, 37, 79e82. northwestern Montana. Eos (Transactions American Geophysical Union) 83, F941. Lowe, D.R., LoPiccolo, R.D., 1974. The characteristics and origins of dish and pillar Steffen, D., Preusser, F., Schlunegger, F., 2009. OSL quartz age underestimation due structures. Journal of Sedimentary Petrology 44, 484e501. to unstable signal components. Quaternary Geochronology 4, 353e362. Mackay, J.R., Burn, C.R., 2002. The first 20 years (1978e1979 to 1998e1999) of ice- Waitt, R.B., 1980. About forty last-glacial Lake Missoula jökulhlaups through wedge growth at the Illisarvik experimental drained lake site, western Arctic southern Washington. Journal of Geology 88, 653e679. coast, Canada. Canadian Journal of Earth Sciences 39, 95e111. Waitt, R.B., 1984. Periodic jökulhlaups from Pleistocene glacial Lake Missoula e new Mehringer, P.J., Sheppard, J.C., Foit Jr., F.F., 1984. The age of Glacier Peak tephra in evidence from varved sediment in northern Idaho and Washington. Quaternary west-central Montana. Quaternary Research 21, 36e41. Research 22, 46e58. Mullineaux, D.R., Hyde, J.H., Rubin, M., 1975. Widespread Late Glacial and post- Waitt, R.B., 1985. Case for periodic, colossal jökulhlaups from Pleistocene glacial glacial tephra deposits from Mount St. Helens volcano, Washington. Journal of Lake Missoula. Geological Society of America Bulletin 96, 1271e1286. Research of the U.S. Geological Survey 3, 329e335. Wintle, A.G., 2008. Luminescence dating: where it has been and where it is going. Mullineaux, D.R., Wilcox, R.E., Ebaugh, W.F., Fryxel, R., Rubin, M., 1978. Age of the Boreas 37, 471e482. last major Scabland flood of the Columbia Plateau in eastern Washington. Wintle, A.G., Murray, A.S., 2006. A review of quartz optically stimulated lumines- Quaternary Research 10, 171e180. cence characteristics and their relevance in single-aliquot regeneration dating Murray, A.S., Wintle, A.G., 2000. Luminescence dating of quartz using an improved protocols. Radiation Measurements 41 396e391. regenerative-dose protocol. Radiation Measurements 32, 57e73. Zuffa, G.G., Normark, W.R., Serra, F., Brunner, C.A., 2000. Turbidite megabeds in an Murray, A.S., Wintle, A.G., 2003. The single aliquot regenerative dose protocol: oceanic rift valley recording jökulhlaups of late Pleistocene glacial floods of the potential for improvements in reliability. Radiation Measurements 37, 377e381. western United States. Journal of Geology 108, 253e274.