Record of glacial Lake Missoula floods in glacial Lake Columbia,
Washington
Michelle A. Hanson a, *
, John J. Clague b
a
Saskatchewan Geological Survey, 1000e2103 11th Avenue, Regina, Saskatchewan S4P 3Z8, Canada
b
Department of Earth Sciences, Simon Fraser University, 8888 University Drive, Burnaby, British Columbia V5A 1S6, Canada
a r t i c l e i n f o
Article history:
Received 31 July 2015
Received in revised form
12 December 2015
Accepted 14 December 2015
Available online 29 December 2015
Keywords:
Glacial Lake Missoula
Glacial Lake Columbia
Outburst floods
Stratigraphy
Sedimentology
a b s t r a c t
During the last glaciation (marine oxygen isotope stage 2), outburst floods from glacial Lake Missoula
deposited diagnostic sediments within glacial Lake Columbia. Two dominant outburst flood lithofacies
are present within glacial Lake Columbia deposits: a flood expansion bar facies and a finer-grained
hyperpycnite facies. We conclude that the flood sediments have a glacial Lake Missoula source
because: (1) current indicators indicate westward flow through the lake, and upvalley flow followed by
downvalley flow in tributary valleys; (2) no flood sediments are found north of a certain point; (3) there
is a dominance of BeltePurcell Supergroup clasts in a flood expansion bar; and (4) some of the finergrained
beds have a pink colour, reflective of glacial Lake Missoula lake-bottom sediments. A new
radiocarbon age of 13,400 ± 100 14
C BP on plant detritus found below 37 flood beds helps constrain the
timing of outburst flooding from glacial Lake Missoula.
© 2015 Elsevier Ltd. All rights reserved.
1. Introduction
During the Fraser Glaciation (marine oxygen isotope stage 2),
outburst flooding from glacial Lake Missoula significantly modified
the landscape of eastern Washington (Bretz, 1928, 1923; Waitt,
1980). Within this landscape was glacial Lake Columbia, which
existed due to damming of the Columbia River by the Okanogan
lobe of the Cordilleran ice sheet (Fig. 1A; Atwater, 1986, 1984; Bretz,
1932; Flint,1935; Waitt and Thorson,1983). Farther east, the Purcell
Trench lobe dammed the Clark Fork River in northern Idaho,
impounding glacial Lake Missoula in Montana (Fig. 1A; Pardee,
1942, 1910). Glacial Lake Missoula had no subaerial outlet, but it
drained when water reached a threshold depth against the
damming ice. Failure of the dam and emptying of the lake producied
outburst floods with peak discharges of up to 17 million m3
sÀ1
that lasted days to weeks (Baker, 1973; Clarke et al., 1984; Craig,
1987; Denlinger and O'Connell, 2010; O'Connor and Baker, 1992).
After the lake emptied, the ice dam reformed and the process
repeated itself. As many as 89 floods occurred over a period of
approximately 2000 years during the Fraser Glaciation (Atwater,
1986; Clague et al., 2003).
Glacial Lake Columbia was in the direct path of these outburst
floods. Floodwaters flowed through glacial Lake Columbia along
different routes and with different outlets, depending on the depth
of the lake and the configuration of the Okanogan and the Columbia
River ice lobes. For most of the lake's existence, the Okanogan lobe
dammed Columbia River downstream of Grand Coulee (Fig. 1A),
forming a low-level lake at ~500 m above sea level (asl; Fig. 1B;
Atwater, 1987, 1986). At this low level, floodwaters drained from
glacial Lake Columbia via Grand Coulee (465 m asl; Atwater, 1987,
1986). At its maximum, however, the Okanagan lobe terminated
at the Withrow moraine, blocking Grand Coulee (Fig. 1A; Atwater,
1987, 1986; Flint, 1937, 1935; Flint and Irwin, 1939; Kovanen and
Slaymaker, 2004), and glacial Lake Missoula floodwaters drained
at higher elevations along the southern margin of the lake, at the
Cheney divide (702 m asl) and the Hawk Creek divide (710 m asl;
Fig. 1A; Atwater, 1987, 1986). This high-level lake, which may have
lasted for about two centuries, attenuated the effects of glacial Lake
Missoula floodwaters (Atwater, 1986). The presence of ice-rafted
erratics at elevations up to ~750 m asl indicates that the lake
swelled during flooding from glacial Lake Missoula (Atwater, 1986).
For a short period, a maximally extended Columbia River lobe
blocked the confluence area of the Columbia and Spokane rivers
and temporarily divided the lake in two (Atwater, 1986; Flint, 1937,
1936; Pardee, 1918; Richmond et al., 1965; Waitt and Thorson,
1983). Atwater (1986) argued that anomalously thin flood beds
* Corresponding author.
E-mail addresses: Michelle.Hanson@gov.sk.ca (M.A. Hanson), jclague@sfu.ca
(J.J. Clague).
Contents lists available at ScienceDirect
Quaternary Science Reviews
journal homepage: www.elsevier.com/locate/quascirev
http://dx.doi.org/10.1016/j.quascirev.2015.12.009
0277-3791/© 2015 Elsevier Ltd. All rights reserved.
Quaternary Science Reviews 133 (2016) 62e76
Fig. 1. (A) Study area showing locations mentioned in the text and the maximum extents of the Cordilleran ice sheet and of glacial Lake Columbia; darkest grey highlights glacial
Lake Missoula floodpaths; dashed line under the Okanogan lobe shows the extent of the Grand Coulee (adapted from Atwater, 1986; Levish, 1997; Waitt, 1985). (B) Franklin D.
Roosevelt Lake, Washington (white); the grey area is the areal extent of glacial Lake Columbia at the 500 m asl stage. Representative sites are shown. Symbols refer to the dominant
lithofacies association at each site; other lithofacies associations may be present. (Sanpoil Valley area of figure adapted from Atwater, 1986.)
M.A. Hanson, J.J. Clague / Quaternary Science Reviews 133 (2016) 62e76 63
near Manila Creek were the product of attenuation of floodwaters
by a fully extended Columbia River lobe and inferred that the lobe
sat at this maximum position for approximately one to two centuries
around 15,350 14
C yr BP.
Pardee (1918) was the first to recognise glaciolacustrine sediments
in the glacial Lake Columbia area. He named them the
“Nespelem silt” and described them as fine white silt with interbedded
gravel. Other early researchers presented evidence for the
advance of the Columbia River lobe southward (Flint, 1936; Jones
et al., 1961) and described coarse sediments interbedded with
glaciolacustrine sediments that would later be recognized as evidence
for the flow of glacial Lake Missoula floodwaters into glacial
Lake Columbia. Subsequent research of exposures of glacial Lake
Columbia sediment focussed on beds attributed to outburst floods
from glacial Lake Missoula (Richmond et al., 1965; Waitt, 1984;
Waitt and Thorson, 1983). The most significant body of work
done on glacial Lake Columbia sediments and interbedded glacial
Lake Missoula sediments is that of Atwater (1987,1986,1984) in the
Sanpoil River valley and Manila Creek area (Fig. 1B). He presented
evidence for 89 glacial Lake Missoula flood-laid beds interbedded
with fine-grained glaciolacustrine sediment.
This paper describes and discusses glaciolacustrine and glacial
Lake Missoula flood-laid beds in the west part of former glacial Lake
Columbia. It is based on examination of 128 exposures, including
six exposures previously described by Atwater (1986) in the Sanpoil
River valley. One new radiocarbon age further constrains the timing
of glacial Lake Missoula flooding during the Late Pleistocene.
2. Study area
Sediments deposited in glacial Lake Columbia are extensively
exposed along Franklin D. Roosevelt Lake in northeast Washington
(Fig. 1B). The lake is impounded behind Grand Coulee Dam near
where the Okanogan lobe dammed the Columbia River at the peak
of the Fraser Glaciation. Lake Roosevelt is 107 m deep at the dam
and extends upstream almost to the CanadaeUnited States International
Boundary. It has an area of approximately 325 km2
and a
perimeter of 965 km. The major streams flowing into the lake
include the Sanpoil River, Hawk Creek, the Spokane River, and the
Colville River (Fig. 1B). We surveyed the lake shoreline over a distance
of 170 km from Grand Coulee Dam upstream to just north of
the Colville River, 16 km up the Sanpoil River valley, and 45 km up
the Spokane River valley (Fig. 1B).
3. Methods
We accessed most exposures along Lake Roosevelt by boat and a
few sites by roads. We described exposures, noting unit thickness,
the nature of contacts between units, grain size, sorting, clast
support, and sedimentary structures, including flow direction
indicators.
Lithofacies types presented below (Section 4) follow the terminology
of Miall (1978, 1977) and Eyles et al. (1983), but without
the lithofacies codes. The terminology applied to flood sediments
(Section 5) is based on the j€okulhlaup (glacier outburst flood)
lithofacies types of Maizels (1997, 1993). We group lithofacies into
associations where appropriate to interpret depositional environments.
For descriptive purposes, we divide the outburst flood
sediments deposited in glacial Lake Columbia into two groups
based on the magnitude of the floods.
4. Description and interpretation of glacial Lake Columbia
sediment
4.1. Diamicton-gravel-sand facies association
The diamicton-gravel-sand facies association is exposed at
several places along the shoreline of Roosevelt Lake north of the
mouth of the Spokane River (‘subaqueous fan sediments’, Fig. 1B).
We examined several exposures of these sediments north of the
Fig. 2. Subaqueous fan deposits. (A) Diamicton (site 38). The diamicton is massive, poorly sorted, and moderately compact, with ~5% subangular to subrounded clasts. (B) A
horizontal diamicton bed overlain by five diamicton beds dipping ~9 southward (site 38). The diamicton beds are separated by outwash beds of poorly sorted gravel, sand, and
minor silt. (C) Massive, poorly sorted diamicton overlain and underlain by trough and planar cross-bedded gravelly sand and silt (site 37). The diamicton pinches out to the south
(right). The lower sand and silt unit contains lenses of similar diamicton that truncate the sand and silt and contain rip-ups of silt. One of the diamicton lenses has downvalleyoriented
cross-beds and clasts along its lower contact.
M.A. Hanson, J.J. Clague / Quaternary Science Reviews 133 (2016) 62e7664
Colville River (Fig. 1B). The diamicton contains lenses of silt and
sand, and is moderately compact, massive, poorly sorted, has a
sandy silt matrix, and contains 5e25% subangular to subrounded
clasts ranging from granules to boulders up to 1.5 m in diameter
(Fig. 2A). Some clasts are striated and faceted. We describe the
facies at two well-exposed representative sites.
At site 38 (Figs. 1 and 2B), five diamicton beds ranging in
thickness from 1 to 3.5 m dip approximately 9 southward. These
beds are separated conformably by bedded to graded layers of
poorly sorted gravel, sand, and minor silt ranging from a few centimetres
to ~1 m thick.
At site 37 (Figs. 1B and 2C), a similar diamicton sharply overlies
trough cross-bedded and inclined planar-bedded sand and silt with
gravel lenses. The diamicton pinches out to the south (down-ice)
and is conformably overlain by inclined and trough cross-bedded
gravelly sand and silt. This upper stratified unit coarsens upward
and contains a few lenses of gravel up to 20 cm long that deform
underlying beds. The lower stratified unit contains lenses of diamicton
with clasts of silt up to 20 cm long. The diamicton lenses
truncate the sand and silt. Most of the diamicton lenses are
massive, but one has faint downvalley cross-bedding and a concentration
of clasts along its lower contact.
The diamicton-gravel-sand facies association was likely deposited
on grounding-line subaqueous fans (‘subaqueous fan sediments’,
Fig.1B) in glacial Lake Columbia at the front of the Columbia
River lobe as it retreated northward. There is no evidence that these
deposits were overridden by an advancing glacier.
The diamicton beds at site 38 were deposited by non-erosive,
non-channelised subaqueous debris flows (Benn, 1996; Eyles and
Clague, 1991; Eyles and McCabe, 1989), flowing away from the ice
margin into the lake. The source of the diamictons may have been
till emerging at the grounding line or pre-existing sediment close to
the glacier margin. The interbedded layers of stratified sediment
are outwash.
The trough cross-bedded and inclined sand and silt at site 37 are
outwash deposited on the distal part of a subaqueous fan (Rust and
Romanelli, 1975). The diamicton layers record subaqueous debris
flows, probably from failures on the slope of the fan. Rip-up clasts
from underlying glaciolacustrine sediments, clasts concentrated
along the base, and crude bedding in one of the diamicton layers
Fig. 3. Subaqueous outwash deposited in front of the Columbia River lobe. (A) Clast-supported, planar-inclined, imbricated cobble gravel. The gravel grades up into pebble-cobble
gravel and contains interbeds of massive pebbly sand (MS). This sequence is overlain by glaciolacustrine sand and silt (GL). (B) Normally graded planar-inclined beds of pebblecobble
gravel and sand overlain by normally graded horizontal beds of pebble-cobble gravel and sand (site 60). (C) Interbedded pebble-cobble gravel, sand, and silt. (D) Trough
cross-bedded coarse to medium sand (CS) and interbeds of massive fine sand (MS). (E) Bedded and massive sand and silt containing ice-rafted gravel (arrows). (F) Massive and
bedded sand cross-cut by a vertically laminated dike (arrow).
M.A. Hanson, J.J. Clague / Quaternary Science Reviews 133 (2016) 62e76 65
indicate movement in a fluid state (Benn, 1996). The overlying
upward-coarsening gravelly sand records southward progradation
of the fan.
4.2. Gravel-sand-silt facies association
We documented 14 outcrops of associated gravel, sand, and silt
north of Coyote Creek along the Columbia River (‘subaqueous fan
sediments’, Fig. 1B). This facies association unconformably overlies
Fig. 4. Distal glaciolacustrine sediment. (A) An approximately 6-m-high exposure of sand, silt, and clayey silt varves. (B) Varves. Each varve comprises a massive lower layer of sandy
silt and an upper thin layer of clayey silt. The exposed part of the measuring tape is approximately 45 cm long. (C) Coarse-grained varves; one conspicuous bed (arrow) was traced
for several kilometres along the shoreline of Lake Roosevelt. The exposure is approximately 3 m high. (D) Faulted, folded, and tilted varves. The box highlights some of the folds. The
exposure is approximately 4 m high.
Fig. 5. Gravelly sediment deposited in glacial Lake Columbia by glacial Lake Missoula floodwaters. (A) Clast-supported cobble-boulder gravel with downvalley planar-inclined
bedding. The exposure is approximately 5 m high. (B) Normally graded, sandy, matrix-supported pebble-cobble gravel with downvalley, planar-inclined bedding. The exposure
is approximately 20 m high. (C) Normally graded, matrix-supported, pebble-cobble gravel with horizontal and downvalley, planar-inclined bedding. Note the small boulder lag at
the top of the gravel sequence (arrows highlight boulders). The gravel is overlain by glaciolacustrine bedded sand and silt. The exposure is approximately 19 m high. (D) A large
sandy rip-up clast (~1 m long; arrow) in trough-cross-bedded pebbly coarse sand.
M.A. Hanson, J.J. Clague / Quaternary Science Reviews 133 (2016) 62e7666
diamicton or bedrock and is conformably or unconformably overlain
by planar-bedded and planar-laminated sand and silt (Section
4.3).
The gravel facies are normally graded and include: clastsupported,
planar-inclined, and imbricated cobble gravel
(Fig. 3A); and matrix-supported, poorly to well-sorted, inclined or
horizontally planar-bedded, pebble-cobble gravel with some small
outsized boulders (Fig. 3AeC). Planar-inclined beds dip 8e23
downvalley (south) or cross-valley (southwest or southeast).
The gravel facies are interbedded with pebbly sand and silt
facies (Fig. 3A and C). At several sites, gravel grades vertically or
laterally into the finer-grained facies and the two are interbedded.
These finer-grained deposits include trough and planar crossbedded
sand (Fig. 3D), horizontally planar or undulatory bedded
sand (Fig. 3E), massive sand, ripple-drift cross-laminated pebbly
sand, and massive silt (Fig. 3D). Most ripples indicate downvalley
flow, but some suggest up- or cross-valley flow. Gravel beds or
pockets of gravel and isolated outsized clasts are common within
the sand (Fig. 3E). Structures such as flames, ball-and-pillows, and
rip-up clasts are present but uncommon. At some exposures, these
sediments are faulted and folded, and vertically laminated clastic
dikes cut the strata (Fig. 3F). At other exposures the beds are steeply
inclined.
At site 60, coarse gravel (Fig. 3B) is conformably overlain by a
matrix-supported diamicton containing 25e30% clasts ranging
from pebbles to boulders up to 30 cm in diameter. The diamicton is
unconformably overlain by a 1e3 m of cobble gravel. The entire
sequence is overlain by irregularly bedded sandy silt.
The gravel-sand-silt facies association records deposition on icemarginal
subaqueous fans as ice retreated up the Columbia River
valley (Cheel and Rust, 1982; Paterson and Cheel, 1997; Rust and
Romanelli, 1975). In Fig. 1B, the diamicton-gravel-sand facies association
and the gravel-sand-silt facies association are grouped
together as ‘subaqueous fan sediments.’ The coarsest sedimentsdthe
clast-supported gravel facies (Fig. 3A)dwere deposited
close to the glacier margin, possibly at the mouth of a
meltwater conduit. Finer gravel (Fig. 3C) was deposited farther
from the ice margin. Clast imbrication indicates transport by
saltation, which implies relatively high-energy currents that are
common in ice-marginal settings. The crude stratification, normal
grading, local erosion, and poor sorting of the matrix-supported
gravel indicate deposition by high-density cohesionless debris
flows or hyperconcentrated flows (Benn, 1996). Finer gravel and
sand beds (Fig. 3C) were deposited by non-channelized sheet flows
moving across the fans.
Site 60 records advance and retreat of the Columbia River lobe.
We infer that the lower unit is subaqueous outwash deposited in
front of the glacier as it advanced southward into the lake. The
gravel was overridden by the Columbia River lobe, which deposited
subglacial till. As the glacier retreated, subaqueous outwash
deposited a thin bed of cobble gravel on the till. Sandy silt at the top
of the sequence is glaciolacustrine sediment deposited in glacial
Lake Columbia (Section 4.3).
Distal slopes of the subaqueous fans are blanketed by finer
grained sediments (Fig. 3DeF). Bedded and laminated sand and silt,
locally with cross-beds and ripple-drift cross-laminae, record
deposition from turbulent underflows. Massive sand and silt record
high-concentration density underflows or suspension deposition.
Current indicators indicate dominantly downvalley flow, but there
is also evidence of cross-valley and upvalley flows, which is common
in ice-marginal settings. The upward fining of sand and silt in
this facies association indicates retreat of the ice margin upvalley.
We attribute coarser lenses within the fine-grained deposits and
isolated outsized clasts to ice-rafting (Fig. 3E). Faulting and folding
are characteristic of ice-marginal settings where sediment becomes
unstable as glaciers retreat upvalley. The clastic dykes are indicative
of loading and rapid pore-water expulsion from underlying
sediments.
4.3. Sand-silt-clay facies association
The sand-silt-clay facies association is present throughout the
study area, but is best exposed along the north-south reach of the
Columbia River (‘varves’, Fig. 1B). It typically consists of a thick
sequence of rhythmically bedded sand/silt and clayey-silt couplets
(Fig. 4A and B); some exposures contain up to 300 couplets. Individual
couplets range from 0.02 to 3 m thick but are commonly
2e20 cm thick (Fig. 4B). Most couplets comprise a lower layer of
massive sand or sandy silt and an upper thinner layer of clayey silt
(Fig. 4B). The clayey-silt layer commonly constitutes 5e20% of the
couplet, but can form up to 75% of thinner couplets. Basal contacts
of couplets are sharp; contacts between the sand/silt and clayey silt
layers are gradational or sharp (Fig. 4B). Thicker couplets contain
coarser sediment than thinner couplets (Fig. 4C), as well as
microlaminations, soft-sediment deformation structures (flames
and ball-and-pillows), silt/clay rip-ups, and climbing or draped
ripples that indicate down- or cross-valley currents. The sand
content, thickness of couplets, and frequency of current structures
all increase to the north in Lake Roosevelt. Some conspicuous
thicker couplets can be traced for several kilometres along the lake
shore (Fig. 4C). At some exposures, rhythmically bedded sediments
contain lenses of gravel with small outsized boulders. At others, the
sediments are faulted or dip up to 20 (Fig. 4D).
This facies association unconformably or conformably overlies
the gravel-sand-silt facies association and unconformably overlies
the diamicton-gravel-sand facies association. It is overlain by
postglacial sediment.
The couplets are interpreted to be varves (Ashley, 2002, 1975;
Smith and Ashley, 1985). Thicker varves with a larger array of
sedimentary structures are the product of high-density turbidity
currents. Thinning of varves up-section indicates deepening of the
lake over time, an increase in distance from the sediment source, or
both. We interpret the outsized clasts and gravel lenses to be,
respectively, dropstones and iceberg dumps. Deformation of the
varves is attributed to destabilization of sediment when nearby or
buried ice melted.
5. Description and interpretation of glacial Lake Missoula
outburst-flood sediment
5.1. Gravel-sand facies association
Sediments of the gravel-sand facies association are discontinuously
exposed along a 5-km reach of the Spokane River (sites
10e14, 18, and 19; Fig. 1B). Gravel facies include unsorted to poorly
sorted, clast-supported cobble gravel with outsized (up to 2 m)
boulders (Fig. 5A); and unsorted to poorly sorted sandy, matrixsupported,
pebble-cobble gravel with outsized (up to 2 m) boulders
(Fig. 5B). Individual beds range up to 2 m thick and are
commonly normally graded (Fig. 5B and C), but a few beds are
reversely graded or massive. Contacts between beds are conformable
or erosive. Boulder or cobble lags are present at the top of
many of the gravel beds (sites 13,14; Fig. 5C). Gravel beds are planar
and either horizontal or inclined up to 35 downvalley (Fig. 5AeC).
Trough cross-bedding was noted at a few sites (Fig. 5D). Most gravel
clasts are subrounded, but some are subangular to angular. More
than half of the clasts are argillite, quartzite, or other metamorphic
rocks of the Precambrian BeltePurcell Supergroup. Other lithologies
include granitic rocks, vein quartz, and gneiss. Gravel beds
contain rip-ups of sand, silt, or clay up to 2 m long (Fig. 5D); small
M.A. Hanson, J.J. Clague / Quaternary Science Reviews 133 (2016) 62e76 67
rip-ups tend to concentrate in specific beds. Interbeds of massive
pebbly coarse sand or silt are present within the gravel. At a few
sites, the gravel fines upward through the entire exposure. The
gravel fines downvalley to the northwest along the Spokane River
from poorly sorted, matrix-supported, pebble-cobble gravel to
coarse pebbly sand. The gravel-sand facies association underlies the
finer-grained facies association described in Section 5.2.
The gravel-sand facies association was deposited by glacial Lake
Missoula floodwaters as they flowed through glacial Lake
Columbia. The sediments are similar to those deposited on flood
expansion bars and include: clast-supported gravel indicative of
deposition from high-energy traction carpets; poor sorting and
crude stratification reflecting deposition from turbulent flow with a
high sediment load; and large-scale planar beds and cross-beds
indicative of deposition from sheets of bedload on a downstreammigrating
bar (Baker, 1973; Maizels, 1997; O'Connor, 1993). We
Fig. 6. Examples of sand and silt deposited in glacial Lake Columbia by glacial Lake Missoula floodwaters. (A) Five sand flood units bounded by silty varves near Hawk Creek. Flood
units have loaded lower contacts, silt rip-ups, planar beds, and massive beds. Box highlights photo C. (B) A pebbly coarse sand flood bed at Hawk Creek (site HC). The flood bed has
an erosive planar lower contact, contains beds with abundant silt rip-ups, and is sharply overlain by laminated silt. (C) Close-up of part of photo A. The flood unit has a loaded lower
contact and silt diapirs and silt rip-ups in the lower part of the unit. The flood unit grades upward into planar-bedded sand and silt. The entire sequence is draped by silty varves that
were loaded and eroded by a subsequent flood. (D) Trough cross-beds in a flood unit with laminae of silt at the base. The tool is 26.5 cm long. (E) Climbing ripple-drift sequence
sharply overlain by laminated silt (Z). (F) Climbing ripple-drift sequence grading upward into laminated silt and interlaminated sand, which is overlain by bedded and massive silt
(displayed portion of tape is ~70 cm).
M.A. Hanson, J.J. Clague / Quaternary Science Reviews 133 (2016) 62e7668
attribute inverse grading in some beds to grain collisions in a
hyperconcentrated flow whereby coarser grains move to zones of
lower shear at the edges of the flow (Baker, 1973; Maizels, 1997,
1989) or to increasing flow velocity and sediment availability during
the waxing stage of a flood (Maizels, 1997). The decrease in
sediment size up-section and downstream indicates a decrease in
flow and is characteristic of flood bars, particularly expansion bars.
Most clasts are sub-rounded, but some are angular, which indicates
local entrainment of clasts, possibly due to erosion or collapse of
nearby bedrock. The boulder lags developed by winnowing of the
surface of the bar during waning stages of flow (Baker, 1973;
Maizels, 1997; O'Connor, 1993; Rushmer, 2006; Russell, 2007).
Rip-up clasts were eroded from interflood glaciolacustrine sand
and silt. The ~2-km-wide, ~7-km-long expansion bar that hosts the
gravel-sand facies association was deposited where floodwaters
decelerated as they flowed out of a constriction and expanded to
the northwest (Fig. 1B). The bar abuts bedrock along the western
side of the lake. The large scale of the gravel foresets suggests that
this bar was deposited during the early or maximum stage of a
flood (Rushmer, 2006).
5.2. Sand-silt facies association
The sand-silt facies association crops out along much of the
shoreline of Lake Roosevelt (Fig. 1B). ‘Units’, which we interpret to
be sand and silt deposited during a single event, range from a few
centimetres to ~5 m thick, but are commonly between 0.5 m and
2 m thick. Units generally thin up-section and in the down-flow
direction. Numerous units are commonly exposed at most sites,
interbedded with the distal glaciolacustrine sediments (Fig. 6A).
The maximum number of units at any one site is 37 exposed over a
96.5-m-thick section at Hawk Creek (site HC, Fig. 1B). Based on the
thickness of exposed units, we estimate that six to nine more units
are covered by colluvium in the lower part of the Hawk Creek
section.
Fig. 7 is a schematic drawing of two typical units of the sand-silt
facies association separated by glacial Lake Columbia varves. Units,
including those shown in Fig. 7, are normally graded, ranging from
cobbly to pebbly very coarse sand at the base to silt at the top. The
contact between a unit and underlying varves is loaded (Figs. 6A
and 7), planar (Figs. 6B and 7), undulatory, and erosive. Varved
sediments separating successive units range from a few centimetres
to 2 m thick (Fig. 6B and C). In some cases, varves are
discontinuous or missing between two units, having been eroded.
The number of varves between sand-silt units could not be counted
at every site, but where counted, range from three to 87. Some units
have loaded the underlying varves, producing flame structures and
diapirs up to 2.5 m high (Fig. 6A and C). Silt rip-ups are common in
the lower part of units and can have the form of stringers <10 cm
thick up to 1.5 m long (Fig. 6AeC). Dish structures, pillow structures,
and micro-faults are less common.
The base of some units is massive and poorly sorted sand. Above
this massive zone is planar- or laminated bedded sand (Fig. 6C),
followed by planar- or trough cross-bedded sand (Fig. 6D), then
climbing ripple-drift sand that grades from type-A and type-B
ripples to draped ripples (Fig. 6E), locally with flame structures
and pillows. Above this is laminated silt and minor rippled sand
(Fig. 6F) capped by one to several massive silt beds that have a
combined thickness of 1 m and are slightly pink (Figs. 6F, 8A and
8B). The massive silt beds are most common along the upper
Fig. 7. Schematic drawing of two typical glacial Lake Missoula sand and silt flood units
separated by glacial Lake Columbia varves. Not all flood units contain every characteristic
depicted here. Italicized letters represent traditional turbidite sequence units.
M.A. Hanson, J.J. Clague / Quaternary Science Reviews 133 (2016) 62e76 69
Columbia River, Sanpoil River, and Hawk Creek. They are
conformably overlain by varves (Figs. 7, 8A and 8B).
There are exceptions to the common sequence of sediments in
units, as described above. Some units include reversely graded beds
with gradational basal and upper contacts (Figs. 7 and 8B). Some
units contain massive, brecciated or convoluted beds of finergrained
material within the overall fining-up sequence (Figs. 7
and 8B). Others have inset channel fills comprising massive, clastsupported,
cobble-boulder gravel with angular to sub-angular
clasts.
Along the Spokane River and the Columbia River below the
SpokaneeColumbia River confluence, uncommonly thick units, up
to 5 m thick, contain planar-bedded or cross-bedded, pebbly coarse
sand with sharp upper and lower contacts and rip-up clasts; the
finer-grained facies are absent (Fig. 8C). These units are unconformably
overlain by either massive silt beds or silty varves
(Fig. 8C).
Sand-silt units at sites 7 and 91 (Fig. 8D) contain an uncharacteristically
large number (6e15) of beds. The beds are massive or
normally graded, but can be reversely graded in the middle
(Fig. 8E). They contain laminations, cross-beds, or ripples, and
commonly load underlying beds, producing rip-ups and flame
Fig. 8. Glacial Lake Missoula sand and silt flood units in glacial Lake Columbia. (A) Sandy climbing ripples in a flood bed (site HC). The flood bed is sharply overlain by massive silt
(MZ), which grades upward into silty varves (V). The tool is 26.5 cm long. (B) An upward-fining flood bed (site HC). Planar-bedded sand grades upward successively though rippled
sand, interlaminated sand and silt, laminated silt, and massive silt into silty varves. The sequence is interrupted by a massive, brecciated silt bed (ZB) and a massive, reverse-graded
sand bed (SB). The shovel is ~1 m long. (C) Three sand flood beds (F1, F2, and F3) separated by thin loaded varved silt (GL). The fine-grained flood facies are absent at this site. (D)
Three flood beds (F1, F2, and F3) separated by varves (GL) (site 91). The lower flood bed comprises approximately 15 sand beds. (E) Close-up view of two beds in photo D. The lower
bed contains normal and reverse grading (arrows), cross-beds, and silt rip-ups (circles). Flame structures at the top of the bed are attributed to loading by the overlying coarser sand
bed. (F) One of the four anomalous beds at site 67. The bed has an erosive lower contact, type-A ripples indicating upvalley flow (to the left), and a pinkish colour different from that
of the bounding varves. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
M.A. Hanson, J.J. Clague / Quaternary Science Reviews 133 (2016) 62e7670
structures oriented down-flow (Fig. 8E). Some of these beds thin in
the down-flow direction.
Ripples and cross-beds in sand-silt units exposed along the
Spokane River and the Columbia River below the SpokaneeColumbia
river confluence indicate downvalley flow. In exposures
along the upper Columbia River and Sanpoil River, ripples,
cross-beds, flame structures, and rip-up stringers indicate both
up- and downvalley flow. Where only downvalley ripples are present,
they are common only in the upper part of the unit. In the
Sanpoil Valley, two dominant paleocurrent directions were noted
in the same units, with downvalley ripples above upvalley ripples,
cross-beds, or flame structures (Atwater, 1986; e.g., site SP-8,
Fig. 1B). Most ripples in the Hawk Creek embayment indicate
cross-valley flow in both directions or downvalley flow towards the
Columbia River. There are two exceptions to these current directions:
below the Sanpoil River, two exposures (sites WLR-1 and
WLR-2; Fig. 1B) include cross-bedded, cobbly to pebbly coarse sand
inclined toward the east (upvalley).
One to four anomalous sand beds were noted in five, closely
spaced exposures dominated by varves (sites 65e69, Fig. 1B),
~33 km north of the ColumbiaeSpokane river confluence. The beds
are 4e37 cm thick, much thicker than the bounding varves and
have either an erosive or loaded basal contact (Fig. 8F). They
comprise type-A (Fig. 8F) or type-B rippled fine to medium sand,
which is not present in the bounding varves, and are draped by
fining-upward sand, silt, and clay. At the site with four anomalous
sand beds (site 67, Fig. 1B), the ripples in three of the beds indicate
upvalley flow; ripples in the second lowest bed indicate downvalley
flow. The sand has a pink tinge that contrasts markedly with the
olive grey colour of the bounding sediment (Fig. 8F). The sand beds
can be traced over a distance of 3 km and are separated by
97e152 cm of silt and clayey silt.
The sand-silt association was deposited by glacial Lake Missoula
floodwaters. The flood units were deposited by a continuum of
discharges and flow velocities, but can be divided into two groups:
those along the main path of the Missoula floods that travelled
generally westward through glacial Lake Columbia; and those in
back-flooded tributary valleys such as the upper Columbia River,
Hawk Creek, and the Sanpoil River.
The eastewest reach of glacial Lake Columbia contains exposures
of the flood gravel-sand facies association but is dominated
by outcrops of the flood sand-silt facies association (Fig. 1B), in
particular, the up-to-5-m-thick units that lack the finer-grained
facies and that are unconformably overlain by massive silt or silty
varves (Fig. 8C). Loaded lower contacts and rip-up clasts indicate
that these sediments were likely deposited during early rising or
maximum stages of floods. The unconformable upper contact indicates
that either flow velocities were high enough to prevent
deposition of the finer facies or that the upper part of the flood
deposits was eroded before the silt was deposited. It is likely that
some floods eroded, not only the underlying varves, but also the
underlying deposits of one or more floods. At one location in Sanpoil
Valley, Atwater (1986) observed that a flood eroded at least
three varved units and their intervening flood beds.
The upvalley cross-beds at two exposures (sites WLR-1 and
WLR-2, Fig. 1B) are anomalous. This counterintuitive flow direction
might be due to upvalley flow after floodwaters encountered the
Okanogan lobe that dammed glacial Lake Columbia. Two units at
site WLR-2 indicate that, if true, this occurred during more than one
flood. The coarseness of the sedimentdcobbly to pebbly coarse
sanddand the distance traveled from the ice dam (~15 km) indicate
that these floods were likely earlier, larger floods.
Preserved flood beds in the back-flooded channels (Hawk Creek,
upper Columbia River, and the Sanpoil River valley) were deposited
by currents with lower velocities that resulted in preservation of
more of the intervening varves. The presence and preservation of
the finer-grained facies of these units are also consistent with
deposition in calmer arms of the lake. The flood units described in
these back-flooded channels are similar to glacial Lake Missoula
flood beds in glacial Lake Columbia described by Atwater (1987,
1986, 1984) and Waitt (1985, 1984): (1) the flood units rest across
erosional contacts on underlying varves or flood beds, or are convoluted
with the underlying varves; (2) the base of some units
contains rip-up clasts from underlying varves; (3) units are normally
graded from granule gravel or coarse sand to silt or clay; (4)
lower parts of the units consist of planar-laminated sand, in some
instances with outlier cobbles or boulders; (5) planar laminations
are succeeded upward by rippled sand that in the lower parts of
units in tributary valleys indicate upvalley flow, and in the middle
and upper parts of units indicate downvalley flow; (6) ripples are
overlain by draped, planar-laminated, and massive very fine sand
or silt; (7) the uppermost parts of the units consist of normally
graded or massive silt or clay that sharply overlies sand; (8) the
graded or massive silt at the top of some units grades into varves;
(9) some of the fine-grained sediment is pink in colour, unlike
adjacent varves; and (10) flood units thin and fine up-section. These
units are also similar sedimentologically to the subaerially deposited
glacial Lake Missoula back-flooded, slackwater rhythmites of
the Touchet Beds in southern Washington (Atwater, 1984; Waitt,
1985).
Erosion of underlying sediment and the widespread occurrence
of rip-up clasts attest to the high energy involved in deposition of
the flood sediments. Planar bedding at the base of flood units is
indicative of the upper flow regime and possibly the waxing phase
of the flood. The vertical sequence of the rippled bedforms indicates
very rapid sedimentation and lower energy flows during waning of
the flood. Draped and laminated silt near the top of flood units was
deposited from suspension in the lower flow regime, but the
presence of interbeds of sand indicates fluctuating flow. Once the
turbulent floodwaters had passed, the silt in the water column
settled out, producing the massive silt beds. These silt beds were
likely deposited over the days to months following the flood. For
example, it would have taken eight to 22 months for fine particles
in the water column to reach the lake floor at Manila Creek,
assuming that the surface of glacial Lake Columbia was 715 m above
sea level (asl) at its highest stand (not flood swollen), that the lake
floor was between 457 and 502 m asl at Manila Creek (Atwater,
1986), and that silt-to clay-sized particles (4.5e3 mm) settle at a
rate of 0.27e0.62 mm/min Atwater (1983) and Waitt (1984) also
concluded that these silt beds are part of the flood-laid unit,
deposited before the return to normal glaciolacustrine
sedimentation.
Differences between the flood beds described here and those
described by Atwater and Waitt include: (1) an absence of reverse
grading in the lowest part of the flood units, which was noted by
Atwater (1987) and Waitt (1984) in flood beds deposited in glacial
Lake Priest, Idaho, and by Atwater (1987) at the Steamboat Rock
exposure in Grand Coulee; (2) some of the units described here
have massive, poorly sorted sand at their base; and (3) two of the
units have numerous beds.
Atwater (1987) and Waitt (1984) attributed reverse grading at
the base of flood units to deposition during the waxing part of the
flow. Reverse grading was noted at some of the sites described here,
but we attribute the finer sediments at the base of the flood units to
sand being mixed with silt and clay eroded from the underlying
varves, because these parts of the units are dominated by convolute
bedding, diapirs, and silt rip-ups (Fig. 6A and C). The absence of true
reverse grading at our sites is likely due to: (1) high-energy
floodwaters eroding earlier-deposited sediments; (2) ice dam
tunnels enlarging rapidly enough that the peak discharge caught up
M.A. Hanson, J.J. Clague / Quaternary Science Reviews 133 (2016) 62e76 71
with the lesser earlier discharges by the time floodwaters entered
glacial Lake Columbia (Atwater, 1986); or (3) indirect flood pathways,
such that the waxing part of the flood hydrograph is not
present at all sites. The latter two explanations are favoured at sites
where loading of the underlying silt and rip-up clasts are present
because there is no evidence of erosion of coarser flood sediments.
The waxing or peak phase of the flood may be recorded, however, at
sites where there is a massive, poorly sorted sand bed at the base of
a unit. These beds could reflect large sediment concentrations and
insufficient time for the sediment to become sorted or graded while
it was being deposited (Maizels, 1997; Rushmer, 2006).
The anomalous number of beds in a few of the flood units (sites
7 and 91; Fig. 8D and E) are likely the product of velocity pulsing of
turbulent flows within a single flood, something that has been
described in Icelandic j€okulhlaups (Maizels, 1989; Russell and
Knudsen, 1999). A steady underflow can transform into an unsteady,
non-uniform current, and thus bed shear stresses and
sediment entrainment and deposition would vary during a flood
(Best et al., 2005). Non-uniform flow could produce beds of
different grain sizes, loading of underlying beds, and the variety of
sedimentary structures seen in individual beds within one flood
unit. Sites 7 and 91 are in the eastewest reach of the lake and are
not overlain by the ripple-drift sequence found in the back-flooded
valleys. These numerous beds are possibly equivalent to other sandsilt
facies association sediments in the eastewest reach and
therefore were likely deposited by high flow during waxing and
early waning flood stages.
The massive, brecciated, or convoluted beds of finer-grained
sediment that interrupt the fining-upward sequence of a flood
unit (Figs. 7 and 8B) are attributed to mass flows triggered by the
floodwaters during waning flow stages. Any mass movement triggered
by waxing or peak flows probably would have been entrained
in the flood waters and not deposited locally. Because a mass flow
would move slower than the turbulent current that it is moving
through, its deposit lies within the fining-upward sequence. We
attribute the channelized, clast-supported cobble-boulder deposits
to mass wasting from nearby bedrock that was destabilized by the
floodwaters.
6. Chronology
Delicate detrital plant tissue was collected from bedded silt at
the base of the 96.5-m-high section at Hawk Creek (site HC, Fig.1B).
The sample was collected from an intact block of silt directly below
the exposure. Because it was intact, we assumed that the block had
not traveled far, and therefore we associated it with varves
immediately below all 37 flood units logged at the site. The sample
yielded an age of 13,400 ± 100 14
C BP (Beta-225100).
7. Discussion
7.1. Source of flood sediments in glacial Lake Columbia
Atwater (1987, 1986, 1984) and Waitt (1984) have previously
described glacial Lake Missoula flood sediments in glacial Lake
Columbia. Not all researchers, however, have been convinced of
glacial Lake Missoula's influence on glacial Lake Columbia. Shaw
et al. (1999) discussed one of Atwater's (1986) Sanpoil Valley sites
(Sage Trig) and argued against glacial Lake Missoula flooding and in
favour of outburst floods from beneath the Cordilleran ice sheet to
the north. They concluded that floods from the north were
responsible for gravel and sand cross-beds, diapiric injections of silt
and clay into overlying gravelly sand, glaciolacustrine rip-up clasts,
and soft-sediment deformation structures. They overlooked, however,
the implication of the fact that Atwater documented flow to
the east-northeast at that site, as well as flow to the north at four
sites farther north in Sanpoil Valley. Similarly, Peters (2003) and
Peters et al. (2002) described two exposures of gravel foresets in
the lower Spokane River valley (our sites 11e14, 18, and 19) and
upper Columbia River valley (our site 60), respectively, which they
interpret as evidence of floods from sources other than glacial Lake
Missoula. We interpret these exposures, respectively, as part of a
flood expansion bar (Section 5.1) and a proglacial subaqueous
outwash sequence (Section 4.2). Gaylord et al. (2007) confirmed,
using detrital zircon geochronology, that stratigraphically low flood
beds in the Sanpoil River valley record floods from glacial Lake
Missoula. However, they attributed stratigraphically higher flood
beds (not specifically identified) to meltwater from the retreating
Cordilleran ice sheet, thus implying that not all of Atwater's 89 beds
can be attributed to glacial Lake Missoula flooding. Using magnetic
properties of fine-grained sediments, Hanson et al. (2015) also
confirmed a glacial Lake Missoula source for some of the beds in the
Sanpoil River valley.
We found no evidence of large-scale flooding from the north at
the sites that we studied, although we cannot conclusively rule it
out. Evidence supporting a glacial Lake Missoula source for the
flood beds in western glacial Lake Columbia includes: (1) current
indicators; (2) the areal distribution of flood beds; (3) clast lithology;
and (4) the colour of some of the fine-grained sediments.
Nearly all current structures indicate flow of floodwaters from east
to west in the lake, clearly demonstrating a glacial Lake Missoula
source. Anomalous current indicators (sites WLR-1 and -2) have
been addressed (Section 5.2) and cannot conclusively be attributed
to flooding from a source other than glacial Lake Missoula. Current
structures also indicate northward flow up tributary valleys; where
present, southward current indicators commonly overlie northward
indicators within the same bed, reflecting flow back downvalley
after the upvalley surge. In the Columbia River valley, flood
beds are not found north of site 66 (Fig. 1B); exposures to the north
Fig. 9. Glacial Lake Missoula bottom sediment, displaying 22 varves. The pink tinge is
attributable to BeltePurcell Supergroup rocks in the glacial Lake Missoula basin. (For
interpretation of the references to colour in this figure legend, the reader is referred to
the web version of this article.)
M.A. Hanson, J.J. Clague / Quaternary Science Reviews 133 (2016) 62e7672
are dominated by varved sediments and do not contain evidence of
large-scale flooding down the Columbia River. Gravel samples from
the expansion bar in the Spokane River (sites 10e14, 18, and 19)
indicate a dominance of clasts from the BeltePurcell Supergroup,
which is the bedrock surrounding the glacial Lake Missoula basin
and is not present in Washington or to the north of glacial Lake
Columbia, emphasizing a glacial Lake Missoula origin for the flood
that deposited that bedform. Lastly, the pinkish tinge of the sand
and silt in the upper part of the sand-silt flood association is
reflective of the colour of the lake bottom sediments of glacial Lake
Missoula (Fig. 9).
7.2. Flood sediments as turbidites
The sequence of sedimentary structures in the flood sand-silt
facies association is similar in many respects to a classic turbidite
(Baker, 1973; Bjornstad, 1980; Bretz et al., 1956; Bunker, 1982;
Waitt, 1985, 1980), which is characterised by an upward sequence
of massive medium sand, planar laminated fine-medium sand,
ripple laminated fine sand, parallel laminated silt, and massive clay
(Fig. 7). There is at least one significant difference, however, between
turbidites and these flood sediments: the processes by
which turbidites are deposited.
Turbidity currents commonly occur in response to a triggering
event, such as an earthquake- or storm-induced landslide (Mulder
et al., 2003). They are characterised by highly turbid water with
large amounts of suspended sediment that increase in density as
they flow downward through less dense water in marine environments.
This process differs from the passage of sediment-laden
flood through a relatively shallow lake. Mulder et al. (2009, 2003)
have shown, however, that a type of turbidity current (hyperpycnal
surge) can form when a flood with a sufficiently high sediment
concentration enters a standing body of water. Indeed, Zuffa
et al. (2000) attribute deposition of beds of glacial Lake Missoula
origin, >1100 km from the mouth of the Columbia River, to
hyperpycnal gravity flows. These quasi-steady flows can last hours
to weeks and contain very high sediment concentrations.
Mulder et al. (2003) describe a hyperpycnite sequence caused by
a glacial outburst flood. During the waxing stage of the flood, the
flow deposits a coarsening-upward basal unit (Ha), whereas during
the waning stage, the flow deposits a fining-upward unit (Hb,
Fig. 7), commonly with climbing ripples, parallel and crossbedding,
and intrasequence erosional contacts. Mass flows can
form simultaneously with the hyperpycnal surge, and their deposits
can be intercalated with the hyperpycnite or superimposed
on it (Figs. 7 and 8B). A complete hyperpycnite would consist of
both the coarsening and fining sequences, and the transition between
the two would correspond to the maximum grain size and
the peak discharge of the flood (Mulder et al., 2003). Mulder et al.
(2003) suggest that complete hyperpycnite sequences may be rare
in the stratigraphic record because discharge and velocity at flood
peaks are high enough to erode previously deposited sediments, a
scenario that likely existed in glacial Lake Columbia. Based on this
description, we thus conclude that the sand-silt facies association
in glacial Lake Columbia is composed of incomplete hyperpycnite
sequences.
7.3. Characteristics of glacial Lake Missoula flooding in glacial Lake
Columbia
Glacial Lake Columbia was clearly a dynamic environment,
subject to frequent, high-magnitude outburst floods from glacial
Lake Missoula. The characteristics of the flood sediments within
glacial Lake Columbia are dependent on several factors: the
magnitude and frequency of the floods; the shape of the flood
hydrograph; the abundance of entrainable material, and the geometry
of the lake.
Flood magnitude is dependent on the volume of water in glacial
Lake Missoula ( 2600 km3
; Smith, 2006), which itself depends on
the thickness of the ice dam. As the Purcell Trench lobe began to
retreat, the ice dam became thinner between successive floods,
impounding less water each time and flooding more frequently
(Atwater, 1986; Waitt, 1985). At the Manila Creek composite section,
Atwater (1986) noted an overall decrease in the thickness,
maximum grain size, and erosiveness of the flood beds up-section,
and a coincident decrease in the number of varves between flood
beds, all of which reflect changes in the magnitude and frequency
of glacial Lake Missoula flooding. Although none of our exposures is
as comprehensive as Atwater's, we noted similar trends at many
sites, mostly in the back-flooded valleys.
Evidence of the higher-energy floods is evident in the grain-size
(boulders and cobbles) of some of the flood beds (e.g., the expansion
bar along Spokane River) and the erosiveness of the floodwaters,
with very little, or nothing, left of the interbedded varves in
the eastewest reach of the lake. The Spokane River expansion bar
represents the largest glacial Lake Missoula flood recorded in
western glacial Lake Columbia. This landform, however, does not
compare in size to some of the landforms in the Channeled Scablands
(Baker, 1973). It could be that much of the record of glacial
Lake Missoula flooding in glacial Lake Columbia is not exposed or
that many of the earlier floods were eroded by subsequent floods
with higher flows.
The coarser facies of the sand-silt flood facies association present
in the eastewest reach also represent high-magnitude flood
events, although smaller than the flood that created the expansion
bar. The shortness of the flood record in these areas is consistent
with high energy, highly erosive floods. These high-magnitude
flood events of short duration dominate the sedimentary record,
representing nearly 100% of the sediment in exposures. In the backflooded
valleys, the sedimentary record is more complete because
of the relatively calmer nature of these environments.
The four sand beds 33 km upvalley from the ColumbiaeSpokane
River confluence are the most northern distal glacial Lake Missoula
flood deposits in glacial Lake Columbia and thus are likely associated
with large flood events, at least while the Columbia River lobe
was not fully advanced. It is not certain that these four sand beds
record four individual floods. It is difficult to estimate the time
between these beds based on the silt between them. The best estimate
is that, at most, three varves were deposited between two of
the beds, which is inconsistent with what is known about the
timing of glacial Lake Missoula floods. Atwater (1986) counted
1e20 varves between the last 17 floods, which he attributed to a
decrease in the period between floods as the ice dam thinned toward
the end of the Fraser Glaciation. This scenario is not
compatible with the interpretation that the four sand beds are
products of larger floods because they would have occurred earlier
in the flooding sequence than the ones in the Sanpoil Valley that
have few varves between them. It is possible that there are no
varves between these four beds and that the beds are the product of
a single flood, with intermittent surging or pulsing. In this scenario,
the second bed with the downvalley ripples would record the
recession of floodwater during one surge and the bedded to
massive silt could be deposited rapidly from high-concentration
sediment flows between pulses.
Changes in the ice dam over time would affect the shape of the
flood hydrograph. A typical glacial outburst flood hydrograph
commonly exhibits a prolonged rising limb followed by a rapid
falling limb, reflecting gradual expansion of tunnels until the lake
has drained to the level of the tunnels (Carling, 2013; Maizels,1997).
But most hydrographs will also exhibit irregularities caused by
M.A. Hanson, J.J. Clague / Quaternary Science Reviews 133 (2016) 62e76 73
temporary blockages in the ice dam and floodway paths, differential
expansion rates of ice tunnels, access to en- or subglacial water
storage, and/or different flood paths, all of which can cause separate
flood waves or pulsed hydrographs (Carling, 2013; Maizels, 1997,
1993). Furthermore, it is possible that the lake drained due to mechanical
collapse of the dam, rather than enlargement of subglacial
tunnels. In any case, by the time the floodwaters reached the eastern
edge of the study area, they had travelled ~100 km from the ice dam,
largely through eastern glacial Lake Columbia and likely following
multiple pathways. Thus it is not reasonable to assume that the
flood hydrograph would still have its original shape. The absence of
reverse grading in all of the units observed, or why there is very little
record of the waxing or peak surge in the sediments, could be
explained by the peak surge catching up to the initial flood surge.
Erosion by the peak flood surge could also explain the lack of
reversely graded deposits. Because reverse grading is so noticeably
absent, and, in many cases, interflood sediment is preserved and
loaded, we prefer the former explanation.
Because glacial Lake Missoula floods were large and occurred in
a glacial environment, sediment availability was not likely an issue.
Earlier floods likely had access to more glacially derived sediment
than later floods, but later floods had more access to glaciolacustrine
sediment in both glacial lakes Missoula and Columbia. In
Manila Creek, for example, flood sediments appear to be dominated
by locally derived glaciolacustrine sediment with only a small
contribution of glacial Lake Missoula sediments (Hanson et al.,
2015).
The division of glacial Lake Columbia into relatively narrow,
eastewest and north-south reaches clearly affected the distribution
and characteristics of the flood sediments. The narrowness of the
lake reaches would have had increased flow depths and thus shear
stress, producing greater flood power and higher transport and
erosive capacities. This result is evident in the eastewest reaches,
which have the largest calibre sediment and where very little
interflood sediment has been preserved. When floods flowed up
the tributaries, shear stress, flood power, and thus transport and
erosive capacities decreased, allowing for more preservation of
interflood varves and flood beds.
7.4. Chronology
Chronological control on glacial Lake Columbia and the advance
and retreat of the Okanogan, Columbia River, and Purcell Trench
lobes of the Cordilleran ice sheet is poor. An age of 17,240 ± 330 14
C
yr BP on a piece of wood approximately 100 km north of the ice
limit is a maximum age for the advance of the Columbia River lobe
into Washington (Clague et al., 1980). Using varve counts and a
radiocarbon age of 14,490 ± 290 14
C yr BP on a piece of wood
recovered from a varve at Manila Creek, Atwater (1986) estimated
that the Okanogan lobe dammed the Columbia River from shortly
before 15,550 ± 450 14
C yr BP to approximately 13,050 ± 650 14
C yr
BP. The presence of Glacier Peak tephra G in the lowermost 20 km of
the Okanogan Valley, Washington, indicates that the Okanogan
lobe had retreated 80 km from its maximum position at the
Withrow moraine, had left the Columbia River valley, and no longer
dammed glacial Lake Columbia before 11,600 ± 50 14
C yr BP (Kuehn
et al., 2009, and references therein; Porter, 1978).
The new radiocarbon age from this study (13.4 ± 0.1 14
C ka BP)
fits well with what is known of the chronology of glacial Lake
Columbia. At least 37 glacial Lake Missoula flood units overlie the
glaciolacustrine sediment from which the dated sample came. This
age indicates that these flood beds were deposited later in the
sequence of glacial Lake Missoula flooding, and thus represent
some of the smaller floods that were produced by a thinning Purcell
Trench lobe.
Because of the erosive nature of the floodwaters at this site,
numbers of varves between flood units are low, commonly less
than 10, thus it is difficult to estimate the total time spanned by the
exposure. Estimates on the recurrence interval of glacial Lake
Missoula floods range from a minimum of 20 to a maximum of 107
years (Atwater, 1986; Hanson, 2013; Levish, 1997; MacEachern and
Roberts, 2013; Waitt, 1985, 1984). A correlation based on paleomagnetic
secular variation of glacial Lake Missoula flood sediments
deposited in glacial Lake Columbia (Hanson, 2013) provides an
average recurrence interval of 55 years, which is consistent with
varve counts at other locations in glacial Lake Columbia where
floodwaters were less erosive (Atwater, 1986; Waitt, 1985, 1984). At
Manila Creek, however, Atwater (1986) reported an upward
decrease in numbers of varves between glacial Lake Missoula flood
units, thus 55 years is probably not appropriate for flood beds in the
upper part of the Hawk Creek exposure. MacEachern and Roberts
(2013) estimated a tetrapod-recolonization time between some of
the later floods in southern Washington to be about 20e30 years.
An estimated recurrence interval of 25 years, applied to the entire
section, means that the exposure might span ~900e1150 years,
depending on the number of flood units covered by colluvium. By
this line of reasoning, we produce an age of 12.2e12.5 ± 0.1 14
C ka
BP for the top of the exposure and presumably the end of glacial
Lake Missoula flooding into glacial Lake Columbia. At Manila Creek
and the Steamboat Rock section in the Upper Grand Coulee, decades
to centuries of varved sediment were deposited after the last
glacial Lake Missoula flood, indicating that the Purcell Trench lobe
that dammed glacial Lake Missoula started to retreat before the
Okanogan lobe, which dammed glacial Lake Columbia (Atwater,
1986; Waitt et al., 2009). These sediments are not present at the
top of the Hawk Creek exposure, but the estimated age of the top of
the exposure is within two sigma of Atwater's estimate for the
retreat of the Okanogan lobe and is consistent with retreat of this
lobe based on the Glacier Peak tephra G age.
8. Conclusions
Sediments deposited by outburst floods dominate the sediment
fill in western glacial Lake Columbia. We established the source of
the floods to be outbursts of glacial Lake Missoula based on: current
indicators (east to west in the main part of the lake, and up and
then down tributary valleys); BeltePurcell Supergroup clasts; the
pink colour of silt associated with glacial Lake Missoula lake bottom
sediments; and the absence of flood sediments in the northern
reach of the lake. Although we found no evidence for flooding from
another source into glacial Lake Columbia, we cannot conclusively
rule it out.
The eastewest reach of glacial Lake Columbia is dominated by
coarse flood facies, including a flood expansion bar. The rarity of
intervening glaciolacustrine sediment along this reach indicate that
the floodwaters were very erosive. The north-south reaches of the
lake are dominated by back-flooded sediment. Sedimentary structures
and thicker beds of varves between floods indicate that these
reaches were calmer and less subject to erosion by floods than the
eastewest reach. The sand-silt flood facies association is very
similar to flood sediments described elsewhere in glacial Lake
Columbia and to slackwater sediments deposited by glacial Lake
Missoula floodwaters in southern Washington. These fine-grained
sediments show some similarities to classic turbidite sequences,
but here are classified as deposits of hyperpycnal flows, a specific
type of turbidity current.
The timing of later, smaller floods from glacial Lake Missoula
into glacial Lake Columbia is constrained by a radiocarbon age on
plant detritus of 13.4 ± 0.1 14
C ka BP. The dated sediment is overlain
by at least 37 flood beds. An estimated flood recurrence interval of
M.A. Hanson, J.J. Clague / Quaternary Science Reviews 133 (2016) 62e7674
25 years indicates that these 37 floods span ~900 years. This age
helps constrain the timing of glacial Lake Missoula flooding during
marine oxygen isotope stage 2 and is consistent with other published
ages on the advance and retreat of the Okanogan, Columbia
River, and Purcell Trench lobes of the Cordilleran ice sheet.
Acknowledgements
This research was supported by the Natural Sciences and Engineering
Research Council (NSERC) of Canada through Discovery
Grant (24595) to Clague and an NSERC PGS B Scholarship to Hanson.
Access to some of the field sites was granted by the Confederated
Tribes of the Colville Reservation. J. Koch, T. Lakeman, and C.
Reiss assisted with the field work. R. Waitt offered helpful suggestions
for improving the manuscript.
References
Ashley, G.M., 1975. Rhythmic sedimentation in glacial Lake Hitchcock, Massachusetts-Connecticut.
In: Jopling, A.V., McDonald, B.V. (Eds.), Glaciofluvial and
Glaciolacustrine Sedimentation, Soc. Econ. Paleontol. Mineral. Special Pap. 23,
pp. 304e320.
Ashley, G.M., 2002. Glaciolacustrine environments. In: Menzies, J. (Ed.), Modern and
Past Glacial Environments. Butterworth-Heinemann Publications, Oxford,
pp. 335e362.
Atwater, B.F., 1983. Guidebook for 1983 Friends of the Pleistocene field trip to the
Sanpoil River Valley, northeastern Washington. U.S. Geol. Surv. Open File Rep.
83e456.
Atwater, B.F., 1984. Periodic floods from glacial Lake Missoula into the Sanpoil Arm
of glacial Lake Columbia, northeastern Washington. Geology 12, 464e467.
Atwater, B.F., 1986. Pleistocene glacial lake deposits of the Sanpoil River Valley,
northeastern Washington. U.S. Geol. Surv. Bull. 1661.
Atwater, B.F., 1987. Status of glacial Lake Columbia during the last floods from glacial
Lake Missoula. Quat. Res. 27, 182e201.
Baker, V.R., 1973. Paleohydrology and sedimentology of Lake Missoula flooding in
eastern Washington. Geol. Soc. Am. Spec. Pap. 144.
Benn, D.I., 1996. Subglacial and subaqueous processes near a glacier grounding line:
sedimentological evidence from a former ice-dammed lake, Achnasheen Scotland.
Boreas 35, 23e36.
Best, J.L., Kostaschuk, R.A., Peakall, J., Villard, P.V., Franklin, M., 2005. Whole flow
field dynamics and velocity pulsing within natural sediment-laden underflows.
Geology 33, 765e768.
Bjornstad, B.N., 1980. Sedimentology and Depositional Environment of the Touchet
Beds, Walla Walla River Basin, Washington (M.Sc. thesis). Eastern Washington
University, Cheney, WA.
Bretz, J.H., 1923. Glacial drainage on the Columbia Plateau. Geol. Soc. Am. Bull. 34,
573e608.
Bretz, J.H., 1928. Bars of Channeled Scabland. Geol. Soc. Am. Bull. 39, 643e701.
Bretz, J.H., 1932. The Channeled Scabland. In: International Geological Congress
Guidebook 22: Excursion C-2.
Bretz, J.H., Smith, H.T.U., Neff, G.E., 1956. Channeled Scabland of Washington: new
data and interpretations. Geol. Soc. Am. Bull. 67, 957e1049.
Bunker, R.C., 1982. Evidence of multiple late-Wisconsin floods from glacial Lake
Missoula in Badger Coulee. Wash. Quat. Res. 18, 17e31.
Carling, P.A., 2013. Freshwater megaflood sedimentation: what can we learn about
generic processes? Earth Sci. Rev. 125, 87e113.
Cheel, R.J., Rust, B.R., 1982. Coarse grained facies of glacio-marine deposits near
Ottowa [sic], Canada. In: Davidson-Arnott, R., Nickling, W., Fahey, B.D. (Eds.),
Research in Glacial, Glacio-fluvial, and Glacio-lacustrine Systems. Proceedings e
Guelph Symposium on Geomorphology. University of Guelph, Guelph, ON,
pp. 279e295.
Clague, J.J., Armstrong, J.E., Mathews, W.H., 1980. Advance of the late Wisconsinan
Cordilleran ice sheet in southern British Columbia since 22,000 yr BP. Quat. Res.
13, 322e326.
Clague, J.J., Barendregt, R., Enkin, R.J., Foit Jr., F.F., 2003. Paleomagnetic and tephra
evidence for tens of Missoula floods in southern Washington. Geology 31,
247e250.
Clarke, G.K.C., Mathews, W.H., Pack, R.T., 1984. Outburst floods from glacial Lake
Missoula. Quat. Res. 22, 289e299.
Craig, R.G., 1987. Dynamics of a Missoula flood. In: Mayer, L., Nash, D. (Eds.), Catastrophic
Flooding. Allen and Unwin, London, pp. 305e332.
Denlinger, R.P., O'Connell, D.R.H., 2010. Simulations of cataclysmic outburst floods
from Pleistocene glacial Lake Missoula. Geol. Soc. Am. Bull. 122, 678e689.
Eyles, N., Clague, J.J., 1991. Glaciolacustrine sedimentation during advance and
retreat of the Cordilleran ice sheet in central British Columbia. Geogr. phys.
Quat. 45, 317e331.
Eyles, N., McCabe, A.M., 1989. The Late Devensian (<22,000 BP) Irish Sea Basin: the
sedimentary record of a collapsed ice sheet margin. Quat. Sci. Rev. 8, 307e351.
Eyles, N., Eyles, C.H., Miall, A.D., 1983. Lithofacies types and vertical profile models;
an alternative approach to the description and environmental interpretation of
glacial diamict and diamictite sequences. Sedimentology 30, 393e410.
Flint, R.F., 1935. Glacial features of the southern Okanogan region. Geol. Soc. Am.
Bull. 46, 169e194.
Flint, R.F., 1936. Stratified drift and deglaciation of eastern Washington. Geol. Soc.
Am. Bull. 47, 1849e1884.
Flint, R.F., 1937. Pleistocene drift border in eastern Washington. Geol. Soc. Am. Bull.
48, 203e232.
Flint, R.F., Irwin, W.H., 1939. Glacial geology of Grand Coulee dam, Washington.
Geol. Soc. Am. Bull. 50, 661e680.
Gaylord, D.R., Pope, M.C., Cabbage, P.R., Glover III, J.F., Anfinson, O.A., Baar, E.E.,
Vervoort, J.D., 2007. Provenance of glacial outburst flood deposits in the
Channeled Scabland, WA; influence of glacial Lake Missoula, meltwater, Snake
River, and Bonneville flood sources. Geol. Soc. Am. Abstr. Progr. 39, 82.
Hanson, M.A., 2013. Sedimentological and Paleomagnetic Study of Glacial Lake
Missoula Lacustrine and Flood Sediment (Ph.D thesis). Simon Fraser University,
Burnaby, BC.
Hanson, M.A., Enkin, R.J., Barendregt, R.W., Clague, J.J., 2015. Provenance and
deposition of glacial Lake Missoula lacustrine and flood sediments from rock
magnetic properties. Quat. Res. 83, 166e177.
Jones, F.O., Embody, D.R., Peterson, W.L., 1961. Landslides along the Columbia River
Valley, northeastern Washington. U.S. Geol. Surv. Prof. Pap. 367.
Kovanen, D.J., Slaymaker, O., 2004. Glacial imprints of the Okanogan Lobe, southern
margin of the Cordilleran ice sheet. J. Quat. Sci. 19, 547e565.
Kuehn, S.C., Froese, D.G., Carrara, P.E., Foit Jr., F.F., Pearce, N.J.G., Rotheisler, P., 2009.
Major- and trace-element characterization, expanded distribution, and a new
chronology for the latest Pleistocene Glacier Peak tephras in western North
America. Quat. Res. 71, 201e216.
Levish, D.R., 1997. Late Pleistocene Sedimentation in Glacial Lake Missoula and
Revised Glacial History of the Flathead Lobe of the Cordilleran Ice Sheet,
Mission Valley, Montana (Ph.D. thesis). University of Colorado, Boulder, CO.
MacEachern, J.A., Roberts, M.C., 2013. Ichnological evidence of j€okulhlaup deposit
recolonization from the Touchet Beds, Mabton, WA, USA. Quat. Res. 79, 37e48.
Maizels, J., 1989. Sedimentology, paleoflow dynamics and flood history of
j€okulhlaup deposits: paleohydrology of Holocene sediment sequences in
southern Iceland sandur deposits. J. Sediment Petrol. 59, 204e223.
Maizels, J., 1993. Lithofacies variations within sandur deposits: the role of runoff
regime, flow dynamics and sediment supply characteristics. Sediment. Geol. 85,
299e325.
Maizels, J., 1997. J€okulhlaup deposits in proglacial areas. Quat. Sci. Rev. 16, 793e819.
Miall, A.D., 1977. A review of the braided river depositional environment. Earth Sci.
Rev. 13, 1e62.
Miall, A.D., 1978. Lithofacies types and vertical profile models in braided rivers. In:
Miall, A.D. (Ed.), Fluvial Sedimentology. Can. Soc. Petrol. Geol., Calgary, AB,
pp. 597e604.
Mulder, T., Syvitski, J.P.M., Migeon, S., Faugeres, J.-C., Savoye, B., 2003. Marine
hyperpycnal flows: Initiation, behavior and related deposits: a review. Mar.
Petrol. Geol. 20, 861e882.
Mulder, T., Zaragosi, S., Jouanneau, J.-M., Bellaiche, G., Guerinaud, S., Querneau, J.,
2009. Deposits related to the failure of the Malpasset Dam in 1959: an analogue
for hyperpycnal deposits from j€okulhlaups. Mar. Geol. 260, 81e89.
O'Connor, J.E., 1993. Hydrology, hydraulics, and geomorphology of the Bonneville
flood. Geol. Soc. Am. Spec. Pap. 274.
O'Connor, J.E., Baker, V.R., 1992. Magnitudes and implications of peak discharges
from glacial Lake Missoula. Geol. Soc. Am. Bull. 104, 267e279.
Pardee, J.T., 1910. The glacial Lake Missoula, Montana. J. Geol. 18, 376e386.
Pardee, J.T., 1918. Geology and mineral deposits of the Colville Indian Reservation,
Washington. U.S. Geol. Surv. Bull. 1e186.
Pardee, J.T., 1942. Unusual currents in glacial Lake Missoula, Montana. Geol. Soc. Am.
Bull. 53, 1570e1599.
Paterson, J.T., Cheel, R.J., 1997. The depositional history of the Bloomington Complex,
an ice-contact deposit in the Oak Ridges Moraine, southern Ontario, Canada.
Quat. Sci. Rev. 16, 705e719.
Peters, E.K., 2003. Evidence for two outburst floods of non-Missoula origin in NE
Washington state. Geol. Soc. Am. Abstr. Progr. 35, 217.
Peters, E.K., Gaylord, D.R., Pope, M., 2002. Large-scale foreset, boulder-gravel beds in
the lower Spokane River valley, northeastern Washington. Geol. Soc. Am. Abstr.
Progr. 34, 109.
Porter, S.C., 1978. Glacier Peak tephra in the North Cascade Range, Washington:
stratigraphy, distribution, and relationship to Late-glacial events. Quat. Res. 10,
30e41.
Richmond, G.M., Fryxell, R., Neff, G.E., Weis, P.L., 1965. The Cordilleran ice sheet of
the northern Rocky Mountains, and related Quaternary history of the Columbia
Plateau. In: Wright Jr., H.E., Frey, D.G. (Eds.), The Quaternary of the United
States. Princeton University Press, Princeton, NJ, pp. 231e242.
Rushmer, E.L., 2006. Sedimentological and geomorphological impacts of the
j€okulhlaup (glacial outburst flood) in January 2002 at Kverkfj€oll, northern Iceland.
Geogr. Ann. A 88, 43e53.
Russell, A.J., 2007. Controls on the sedimentology of an ice-contact j€okulhlaupsdominated
delta, Kangerlussuaq, west Greenland. Sediment. Geol. 193,
131e148.
Russell, A.J., Knudsen, O., 1999. An ice-contact rhythmite (turbidite) succession
deposited during the November 1996 catastrophic outburst flood (j€okulhlaup),
Skeidararj€okull. Icel. Sediment. Geol. 127, 1e10.
Rust, B.R., Romanelli, R., 1975. Late Quaternary subaqueous outwash deposits near
M.A. Hanson, J.J. Clague / Quaternary Science Reviews 133 (2016) 62e76 75
Ottawa, Canada. In: Jopling, A.V., McDonald, B.V. (Eds.), Glaciofluvial and Glaciolacustrine
Sedimentation. Soc. Econ. Paleontol. Mineral. Spec. Pap. 23,
pp. 177e192.
Shaw, J., Munro-Stasiuk, M., Sawyer, B., Beaney, C., Lesemann, J.-E., Musacchio, A.,
Rains, B., Young, R.R., 1999. The channeled Scabland: back to Bretz? Geology 27,
605e608.
Smith, L.N., 2006. Stratigraphic evidence for multiple drainings of glacial Lake
Missoula along the Clark Fork River, Montana, USA. Quat. Res. 66, 311e322.
Smith, N.D., Ashley, G., 1985. Proglacial lacustrine environment. In: Ashley, G.M.,
Shaw, J., Smith, N.D. (Eds.), Glacial Sedimentary Environments, Soc. Econ.
Paleont. Mineral. Short Course 16, pp. 135e216.
Waitt, R.B., 1980. About forty last-glacial Lake Missoula j€okulhlaups through
southern Washington. J. Geol. 88, 653e679.
Waitt, R.B., 1984. Periodic j€okulhlaups from Pleistocene glacial Lake Missoula e new
evidence from varved sediment in northern Idaho and Washington. Quat. Res.
22, 46e58.
Waitt, R.B., 1985. Case for periodic, colossal j€okulhlaups from Pleistocene glacial
Lake Missoula. Geol. Soc. Am. Bull. 96, 1271e1286.
Waitt, R.B., Thorson, R.M., 1983. The Cordilleran Ice Sheet in Washington, Idaho, and
Montana. In: Wright Jr., H.E. (Ed.), Late-Quaternary Environments of the United
States, Volume 1: the Late PleistoceneUniversity of Minnesota Press, Minneapolis,
MN, pp. 53e70.
Waitt, R.B., Denlinger, R.P., O'Connor, J.E., 2009. Many monstrous Missoula floods
down Channeled Scabland and Columbia Valley. In: O'Connor, J., Madin, I. (Eds.),
Volcanoes to Vinyards: Geological Field Trips through the Dynamic Landscape
of the Pacific Northwest, GeolSoc. Am. Field Guides 15, pp. 737e774. http://
dx.doi.org/10.1130/2009.fld019(32).
Zuffa, G.G., Normark, W.R., Serra, F., Brunner, C.A., 2000. Turbidite megabeds in an
oceanic rift valley recording j€okulhlaups of late Pleistocene glacial floods of the
western United States. J. Geol. 108, 253e274.
M.A. Hanson, J.J. Clague / Quaternary Science Reviews 133 (2016) 62e7676