WATERRESOURCES

RESEARCH,
VOL. 29,NO. 8, PAGES2925-2939,
AUGUST1993

FlumeSimulation
of Recirculating
Flow andSedimentation
JOHN C. SCHMIDT

Departmentof Geography
andEarthResources,
Watershed
ScienceUnit, Utah StateUniversity,
Logan
DAVID M. RUBIN

U.S. GeologicalSurvey,Menlo Park, California

HIROSHI IKEDA

EnvironmentalResearchCenter, Universityof Tsukuba,Ibaraki, Japan

Experimentswere conductedin a 4-m-wideflumeto simulaterecirculatingflow and sedimentation
in a lateral eddywithin a channelexpansion.The percentageof mainstem sedimentthat was captured
by the eddy decreasedfrom 37% (whenthe eddywas empty)to 24% (when sandfilled approximately
32% of the eddyvolume).The reattachment
barwithinthe eddygrewin an upstreamdirection,andthe
finest size sedimentwas depositedin the lee of the obstruction;both observationsare consistentwith
field observations.
Measurements
of reattachment
lengthduringsedimenttransport(0.5-1.0 kg/s)at

constant
discharge
(0.60m3/s)showthatreattachment
lengthdepends
notonlyoncharacteristics
of
the expandingjet, but alsoon the topographyof the channelbed downstream;reattachmentlength
decreasedwhenpart of the channelexpansionwasfilledby an aggrading
midchannelbar. Comparison
of these resultswith measurements
in the ColoradoRiver in GrandCanyon suggeststhat downstream
channelirregularitiesplay a largerole in controllingthe lengthof eddiesin natural rivers.

INTRODUCTION

Recirculatingcurrents,or eddies,developin rivers wher-

have also been few physical model experimentswhich

measuredrecirculatingflow conditionsduring bar aggradation. This paper summarizesthe results of experiments
intended to simulate flow and sedimentation in a bedrock

ever the orientation of downstream flow and the channel

wereundertaken
in the 4-m-wide
banks are sufficientlydivergent, such as at sharp meander canyon;theseexperiments
bends [Leeder and Bridges, 1975] and downstreamfrom flume at the Environmental Research Center, University of
channel confluences[Best, 1986]. These currents also de- Tsukuba. We describe (1) flow patterns, velocities, and
velopwhereconstricted
flowentersa widerreach,suchasin

changes
in reattachmentlengthin a largerecirculationzone,

the lee of debris fans or bedrock that partly obstructs (2) evolutionof bar topographyand bed forms, (3) areal
downstreamflow. Recirculatingflowsare weakerthan adja- sortingof sediments,and (4) efficiencyof this zone in
sediment.The experimental
resultsare relatedto
centdownstreamflow, whichcausessedimentto accumulate capturing

in recirculationzones; sedimentstoragein recirculation

zonesmay comprisea high proportionof total sediment

storagein somebedrockcanyons.Barsdeposited
at high
discharge
in recirculating
currents(andsubsequently
emergent)may becomesubstrate
for riparianvegetation
or be
usedas campsites.
In canyons
affected
by upstream
dams,
recirculating
currentbarsmaybeeroded
[Schmidt
andGraf,
1990],and futurerestorationmanagement
strategies
may
includehigh-discharge,
regulated
releases
intended
to reconstruct these bars.

field studies.

LITERATURE REVIEW

We have observedlarge recirculationzones and associated sandbars along the Colorado River in Grand Canyon,
Arizona,andCataractCanyon,Utah; alongthe SnakeRiver
in Hells Canyon,Idaho and Oregon;alongthe River of No
Return reach of the Salmon River, Idaho; along the Green

River in Lodore and Desolation Canyons, Utah; and on

Flow separation
and associated
recirculation
are long- smaller streams in the western United States and New
studied
phenomena
in thelaboratory
[EatonandJohnston,England.Accountsof river runnersindicatethat suchzones
1981;Simpson,1989]and coastalsettings
[Signelland occurin bedrockcanyonsthroughoutthe world, and recirhasbeendescribedin paleoGeyer,1991;Wolanski
et al., 1984],buthavereceived
less culatingcurrentsedimentation
flood
studies
throughout
the
western
United States[Baker,
attention
inrivers.Thegeneral
hydraulic
andsedimentologic

characteristicsof recirculationzones and associatedbars 1984].In the ColoradoRiver in GrandCanyon,recirculation

have beendescribedin GrandCanyon[Rubinet al., 1990; zonesare largeand persistent,typicallyoccupyingonethird

Schmidt,1990;Schmidt
and Graf, 1990],but few field

to one half of the channel width downstream from constrict-

observations
have been madeof the rate and styleof bar ing debrisfans. In the upstream195km of GrandCanyon,
development
andhydraulic
conditions
inrivereddies.
There about400largeeddiesexistat mostdischarges[Schmidtand
Graf, 1990].

Copyright
1993bytheAmerican
Geophysical
Union.
Papernumber93WR00770.
0043-1397/93/93WR-00770505.00

Experimentsshow that wakeswith simplerecirculating

flow developin the lee of cylindersat Reynoldsnumbers
greaterthanabout4 [Chang, 1966].At Reynoldsnumbers
2925

2926

SCHMIDTET AL.: FLUME SIMULATIONOF FLOW AND SEDIMENTATION

typesthatoccurin narrow,deepbedrock
stream
channels,

Shearlayer

'

includingthe expansionbars and eddy bars that are the

subjectof this paper. The categoryof eddy bars, formed
beneathrecirculatingcurrents,was subdividedby Schmidt

Time-Averaged

Dividing

....
.Hi
-,._,,/1
Streamline

[1990],basedon observations
in GrandCanyon(Figure2).

Recirculation

Separation Zone
Point
Fig. 1.

"'

Reattachment
Zone

Plan view of flow features at a backward facing step

adapted from Driver et al. [ 1985]. H is step height.

Reattachmentbars form in the reattachment zone and beneath the primary eddy. Separation bars mantle the con-

stricting
debrisfan andareformedat higherdischarges
by
secondary eddies that submerge parts of these fans
[Schmidt, 1990]. Sorting occurs within recirculation zones

[Pageand Nanson, 1982].In GrandCanyon,thoseseparation bars formed by high annual peak dischargesin 19831985 were composed of finer sediment than were reattach-

greater than 40, flow within the wake becomes unsteady, and
vortices develop along the shear zone between the main flow
and the wake. Two flow zones develop: a near wake of
closed

recirculation

and

a far

wake

within

which

shed

vortices move downstream. In open-channelflow, the near

wake is referred to as the separation bubble or recirculation

ment bars formed by the same flows, presumablybecause

sortingprocesses
causedthe coarsersedimentto be deposited in the area where sediment first entered recirculation
zones, the reattachment zone [Schmidt, 1990]. Bed load
transportdirectionsbeneath the primary eddy, inferredfrom

zone.

The simplest reattaching flow occurs at the backward

facing step at the boundary of an otherwise semi-infinite or
uniform flow field (Figure 1). The point of flow separation,
which Simpson [1989] has shown to be a zone of varying
detachment conditions, is generally considered fixed in
space where the step corner is sharp, although Eaton and
Johnston [1981] have documentedunsteadinessof the separation point even at this geometry. Near the separation
point, the detached boundary layer forms a free shearlayer
which is initially thin and parallel to the orientation of the
constricted flow. Further downstream, the separated shear
layer curves sharply toward the wall. Simplified drawingsof
experimental conditions typically show a time-averaged
mean streamline impinging the wall at the time-averaged
reattachment point, which is located within a zone through
which the instantaneousreattachment point fluctuates.Reattachment point fluctuation has been related to the migration of large turbulent structures along the shear layer, and
the length of the reattachment zone is of the order of 4 step
heights. Simpson [1989] also reported that the maximum
backflow velocity is typically greater than 20% of the free
stream velocity. Experiments and numerical modeling have
demonstrated that reattachment length is highly sensitiveto
the longitudinal pressure gradient and to bottom friction
[Eaton and Johnston, 1981; Yeh et al., 1988].
Schmidt and Graf [1990] measured great variability in
reattachment length in Grand Canyon, and found that recirculation zones lengthen with increasing discharge until discharge becomes so great that the constricting debris fan is
overtopped and the surface expression of the recirculation
zone appears to become washed out. Flow in recirculation
zones is organized into a simple one-cell primary eddy
rotating suchthat upstream flow occurs along the fiver bank.
At higher discharges, primary eddies lengthen downstream
and secondary eddies develop upstream from the primary
eddy on flooded parts of the constricting debris fan. Schmidt
[ 1990] showedthat there is a maximum length to which each

Fig. 2. Photograph,lookingdownstream,of the ColoradoRiver

in Grand Canyon about 1 km upstream from Blacktail Canyon.

Three reattachmentbarson fiver right existat this discharge;oneis

at the bottom, one in the middle, and one near the top of the
photograph.Each reattachmentbar occursin associationwith a flow

recirculation zone can extend, controlled at the upstream separationinducingdebrisfan. A small rapid in the centerof the
hasbeenformedby a debrisfan on river right;a small
end by the constrictionitself and at the downstreamend by photograph
irregularitiesin channelgeometry. Schmidt [ 1990]found that patchof sandon top of this debrisfan is a separationbar. Note that
reattachment
barsarehighestin elevationon theirdownstream
and
narrow and deep constrictions create longer recirculation
zones than do wider, shallower constrictions.

Baker [1984] distinguishedand describedfour major bar

shoreward side, and that each reattachment bar has a small remnant

channelseparatingthe low-elevationpart of the bar from the talus

bank. This channel is maintainedby backflow at higher discharges.

SCHMIDT
ETAL.:FLUME
SIMULATION
OFFLOW
aNDSEDIMENTATION
TABLE 1. ExperimentalConditions

2927

METHODS

Experiment

Theseexperimentswere conductedin the 160-m-long,

4-m-wide,2-m-deepflume at the University of Tsukuba;
EmptyFlume BedAggradation
detailsof theTsukubafacilityareprovidedby Ikeda [1983].
0.15-0.61
0.60
Water surfaceslopewas controlledby a tailgateat the
22
30
downstream
end of the flume. Bed topographyand bed
13
4
formswereexaminedandmeasured
duringexperiments
by
0
0.5-1
raisingthe tailgate, pendingthe flow, and then slowly

Flowrate,m3/s
Total run time, hours
Number of runs
Sedimentfeed rate, k/s
Sediment feed characteristics:
mean,

mm

standarddeviation, phi units

draining the flume. After measurementswere made, flow

'"

1.3

---

0.96

wasreintroducedandexperimentsresumed.Flow velocities
Number of runsindicatesnumberof timesflumewas drainedand were measuredwith a two-dimensionalelectromagneticcurbed observations were made.
rent meter and from time-lapse photographyof surface
floats.Electromagneticcurrent meter measurements
at 16
sitesweremade0.1 m beneaththe water surfaceevery0.75
sfor a durationof 5 min. Becauseflowdepthincreasedin the
bed form orientationsand sedimentarystructures,mimic downstreamdirection,the positionof the measuringpoint
surfaceflowpatternsandindicaterotarymotion[Rubinet relativeto thebed differedlongitudinally.Resultanthorizonal., 1990].The only largearea of structures
indicatingtal flow vectorsand rose diagramsof the relative amount
downstreamtransportis in the downstream
part of the (durationtimesvelocity) of flow in differentdirectionswere
reattachmentzone. Typical sedimentarystructuresfound computed.The vertical componentof the flow field was not
withinreattachment
andseparation
barsareforesets
depos- determined,althoughNelson [1991] has shown that the
is an importantpart of the flowfieldnear
itedby thebar asit migrates
onshoreandforesets
deposited verticalcomponent
by ripplesanddunesmigratingin a rotarypattern.Symmet- the shear layer. Locations in the flume are referenced to
rical straight-crestedripplesare commonin the reattachment distancedownstreamfrom the headgatewhere water and
zone[Rubinet al., 1990].Separation
bar development
has sedimentwere supplied;for example,station120 is located
beendescribed
by Schmidtand Graf[1990,Figure13],who 120m downstreamfrom the headgate.All referencesto left
documentedupstreammigrationof a separationbar within a or right side of the flume are made as if the viewer were
secondaryeddy duringa regulatedfloodin GrandCanyon. facing downstream.
Field studiesof topographicchangesand sedimentologic We report the resultsof two experimentsin this paper
characteristicsof selectedbars in Grand Canyon are the (Table 1). In both, a semicircularobstruction(constructed
basisof conceptualmodelsof bar buildingwithin recirculat- with sandbagsand coveredby large plasticsheets)coning currentsproposedby Rubin et al. [1990] and Schmidt strictedflowwidth to about1.5 m (Figure3). This obstruc[1990]. These modelspredictthat reattachmentbarsbuild in tionwaslocatedbetweenstations90 and95. For purposesof
an upstreamdirection from the reattachmentzone and that comparison
with laboratorystudiesof backwardfacingstep
sediment deposited at high dischargeis reworked during flow, the stepheightduringour experimentsis taken as the
upstream retreat of the reattachment zone during flood distancefrom the rightwall to the obstructionapex, 2.5 m.
In order to simulate the increased elevation of the bed that

recession.Separationbarsare formedat highdischargesand

are reworked during flood recessionto a lesserextent than
are reattachment

occursat constrictionsin Grand Canyon [Kieffer, 1988], a

0.1-mcinderblockstepwasplacedat theconstriction
during

bars.

decelerating
Jet

90
-..,

,>

z,

95 /
..,-,-,..
......

100

..... :.:...,r;'.,4:-:..
;..,

/"-.'.....-.,'.-.',.'..,.'..'.-...':...
,.:.....x,_
"
-: - ,,,

/'-'
'"..::.:.'-'""'"'"'"::'
"-'"'

. \..

-N-"

,.,
- -

105 /

,,r-_

....

.'

shear
layer

_ .....

....

,.

....... '".t.5..--,2:.

._._-_

.,

,.

cross-over
ofmndownstream
flow

110
-.

l ..- v,,Y,'

,..-_ ,,, r, _,.,

.... _,,,

119

--X.-

.--'-<-2',J-'-,','' /">-'--',-.'-,.,.'"K'X

\,,\,

"' :r...... a -- q,t. %, .X',-,, ..

"--,'---=-- - --/',.7_ . "/a

.__

115/

'2;.'
,. ,"'":,3:,..'?,',
%

--

_,.,,,

/ ,. t

--

;, A.,, ',.,

, :,,. ,,,,.,

RPw, RPp
stagnant
flow

reattachmen!
zone
strongbackflow

Fig.3. Pathlines
ofsurface
floats
during
run13(discharge
0.30m3/s).
Length
of individual
paths
arefor1 s

duration.
Somepathlinesintersect
oneanother
because
fieldofviewforeachphotograph
wasabout5 m andfieldsof
view overlapped.
Dashedline through
pathlinesshowsaveragelocationof dividingstreamline
for purposes
of
measurement
of A d. RPws is locationof the instantaneous
reattachment
point determined
from water surface

measurements,
andRPpislocation
of theinstantaneous
reattachment
pointdetermined
fromphotographs.
Dashed

curvenearreattachment
pointsis thefullrangeof thereattachment
zoneduringthisrun.Valuesalongtheleft sideof
the flume are station locations.

2928

SCHMIDTET AL FLUMESIMULATIONOF FLOWAND SEDIMENTATION

was narrowerthan the mean flow width constrictionratioin

Grand Canyon which Kieffer [1985]found to be 0.5, but the
experimentalrangewas within the rangeof all measuredfield
sites. The flow-width expansionratio was 2.7 in all runs

1.6

total

1.4

energy

range of locationof mean

reattachment
point

because
the flumewallswere neveraltered;Schmidt[1990]
ater

surface

;"T;,'bar-crest
leVatsiOn

.....

after hr

o.
0.6

empty flume experiment;

Initialcondition
for bar-building
experiment
0.4

8O

90 100 110 120 !30

140

5O

DISTANCE,IN METERS,FROMHEADGATE

found that the mean flow-width expansion ratio in Grand

Canyon is 2.9. The flow area expansion ratio (downstream
flow area at station 130 divided by constricted flow area at
station 92) during the experiments varied between 3.4 and
14.2 and greatly exceeded the flow width expansionratio
(Table 2). Becauseflow depth increasedlongitudinally,the
flow area expansion ratio was even greater at points downstreamfrom station 130. Some expansionsin Grand Canyon
are as large as those in the experimentsbecausedepthsof
individualscourholesare as much as 9 times the depthof the

constrictedchannel.Theseexperimentsare thereforerepresentative of some of the field conditions observed in Grand

Canyon; however, these experimentsmay not be representative of sites where the downstream increase in flow area is
1.6

water-surface
Indicated

elevation
hours

small. In the latter case, deceleration of streamwise flow due

after

to increasingflow area is smaller, and the relative impor-

of run time

1,4

tance

to deceleration

of lateral

diffusion

of momentum

across the shear zone is greater.

In the first experiment, changes in recirculating flow

characteristics

were measured for different clear water dis-

charges and tailgate elevations. During this experiment,

22 5

flume
floor
0.6

0.4

'

30

barcrest

ind
. i
90

sediment formed a linear bar with maximum relief of 0.18 m

in the center of the flume downstream from station 120. In

lactat
ed
i

100

110

120

130

rntime
140

150

DISTANCE,IN METERS,FROMHEADGATE

Fig. 4. Longitudinalprofileof water surfaceelevationand maximum bed elevation during bar-building experiment for different
durations. (a) Water surface elevation and total energy grade line
after 14 hours and bed conditionsat start of experimentand after 11
hours. Total energy grade line calculatedfrom v a, using(1) from
text. (b) Water surface elevation after 20 and 26 hours and bed
elevation after 22.5 and 30 hours.

the secondexperiment, a heterogeneoussand mixture was

addedto a 0.60m3/sflowat a ratebetween0.5 and1.0kg/s,

and the rate and style of bar aggradation was measured5
times during 30 hours.
Water surface elevations were measured from a movable

platform above the flume. Mean downstreamvelocity was

computedusingthe following equation:

va = Q/Aa

(1)

whereva is the meandownstreamvelocity, Q is discharge,

andA a is the cross-sectional
areaof downstreamflowbased
on estimates of the location of the time-averaged mean

separation
surfacethat dividesdownstream
andrecirculating
flow. Estimatesof va are only a general approximation
becausethe straightline definingthe mean dividing stream-

the secondexperiment and for some of the runs of the first

experiment(Figure 4). Downstreamfrom the obstruction,a
zone of recirculating current extended to within about 20 m
from the tailgate. Bed slopewas 0.01 for the first experiment
and was variable due to bed aggradationduring the second
experiment. Water surface slope was always less than the
bed slope, and water depth increased greatly downstream.
This physical model is an analogueof some of the conditions in large rivers that flow within bedrock canyons. In
these canyons, the longitudinal prof-fie consists of steps
(debris fans) and pools (channel expansions) [Leopold,
1964]. Debris fans function as natural tailgates, often acting
as local controls upstream to the base of the preceding step
[Kieffer, 1985]. Recirculation in bedrock gorges occurs
within the channel and is not restricted to overbank settings
investigatedby Tamai et al. [1986].
The size of the experimental channel expansion was
greater than the mean size of expansionsin Grand Canyon.
In the experiments, the flow width constrictionratio (downstream flow width divided by constricted flow width) of 0.38

line can only be an approximaterepresentationof a highly

unsteadythree-dimensionalsurface(Figure 3). Froudenumberswerecalculatedbasedon va anddepthsmeasured
0.5m
fromtheleft wall. Hydraulicjumpsoccurredat the constriction where the Froude number of constricted flow was as

smallas0.8 because
va is lessthanmaximumvelocityin the
constriction.

The location of the reattachmentpoint during each run

was determinedin three ways. First, 1-s-exposurephotographsof surfacefloatswere usedto identify the instanta-

neousreattachment
pointat the time whenthe photograph
was taken. Second,digitizedlongitudinaltracesof the water

surface0.5 m from the fight wall were usedto identifythe

reattachmentpoint by determiningthe most downstream

extentof adverse(upstream)
watersurfaceslopeat thetime
of the water surface measurements. The time of these

measurements
wasgenerally
within15minfromthetimeof
the photography.
Third, visual observations
of the timeaveragedlocationof the reattachment
pointwere made.

Bedtopography
wasdetermined
by repeatedly
surveying

SCHMIDT
ETAL.: FLUMESIMULATiON
OFFLOWANDSEDIMENTATION

2929

2930

SCHMIDT ET AL.' FLUME SIMULATION OF FLOW AND SEDIMENTATION

105

109

113

117

Scale of resultant flow vectors

0.25

Fig. 5.

cross sections

Rose diagrams showing resultant flow vectors and amount of flow (speed times duration) toward each
indicated direction during run 7.

located

at 2-m intervals

and 125 and at 4-m intervals

each cross section
accurate

m/s

to within

were

between

elsewhere.

taken

at 0.2-m

0.1 cm. Volume

stations 89

Measurements
intervals

calculations

at

and are

were

made

using surface and volume modeling software, and mass

transport rates were calculated assuming that the specific
gravity of the sediment was 2.65 and the porosity of the
sediment deposits was 35%. Samples of bed sediment were
collected

when the flume was drained and were later sieved.

Sediment transport rates past the recirculation zone could

not be determined

from the sediment

feed rate because the

deposition rate upstream from the constriction was variable.

We determined the transport rate through the constriction by
measuring the total volume of sediment aggraded in the
flume downstream from the constriction during each run.
The resulting transport rates are accurate because no sediment washed out of the flume; in fact, little aggradation
occurred in the most downstream 20 m of the flume. Aggradation rates in different parts of the flume were used to
determine the efficiency of the eddy in capturing sediment
transported by the main downstream current. These rates
were calculated by determining the difference in sediment
volume between successive measurements for five areas,
each of constant dimension. This procedure eliminated bias
in estimatingthe dimensionsof the recirculation zone during
each measurement period but may slightly overestimate the
proportion of sediment depositedin the eddy during the final
period because the recirculation zone decreased in size.
RESULTS

vortices

increase

in

diameter

in

the

downstream

direction, (3) nearly stagnant flow in the lee of the obstruction and extending 1-1.5 channel widths downstream, and
(4) strong backflow in an area immediately downstream from
the stagnant flow area (Figure 3). Vortices eventually increased in diameter such that they extended acrossthe entire
flume. The recirculation zone typically was truncated where
the high-velocity core of downstream current crossed to the
right side of the flume. The location of the reattachment
point fluctuated over distances as great as 5 times the step
height, similar to fluctuation ranges reported by Simpson
[1989] (Table 2).
Although the flow field can be described by average flow
conditions, instantaneous vectors were widely variable, es-

pecially in the shear layer. Figure 5 showsrose diagramsof

instantaneous velocity at selected locations using the twodimensional current meter. In the upstream part of the
expandingjet, downstream velocities were so much greater
than the rotary flow of the vortices that the instantaneous
flow directions

of the combined

flow remained

within 45

degreesof directly downstream. In the center of the recirculation zone, mean velocities were weaker and flow directions were more variable. Farther downstream where the jet
slowed and combined with the vortices, instantaneous vec-

tors varied over 360 degrees. Instantaneousvelocies in the

reattachment

zone were bidirectional,

and upstream-

downstreamflow reversals occurred. Along the wall, flow

directionwasrestricted,but speedsvariedgreatly. Backflow
along the wall was typically about half of v,/, but instantaneous speeds ranged from a minimum of nearly zero to a

maximum that approached the mean speed of the down-

Flow Patterns and Velocity

The flow field downstream

where

streamcurrent. Upstreamvelocity was between0.00 and

from the constriction

can be

0.10 m/s in the lee of the constriction. Backflow speeds

subdivided into (1) the deceleratingjet, (2) the shear layer

greatly exceededvalues reported by Simpson[1989] and

SCHMIDT
ETAL.:FLUMESIMULATION
OFFLOWANDSEDIMENTATION
3.5

'

'

'

'

'

'

'

'

'

'

''

'

' ''

'

'

'

'

'

2931
'

short recirculation zone

orlong
thin

, 3

recirculationzone

13

withpoorly- O

developed
circulation
(13.4-13.))

2
n-

well-developed
circulation

1.5

10.2

10.2-13

- 0.5

0 (9.4-9.8)

9-10.6 13/

8.2-11
O

(7.4-9.4)
,

(7-10.6)

O (10.2) O 8.6
,

rm

7-1 1

poorly-developed
circulation
,

10

FLOW-AREA EXPANSION

12

14

16

RATIO

Fig. 6. Recirculationzone lengthat differentexperimentalconditions.Numbers are time-averagedreattachment

length divided by step height. Numbers in parenthesesare the range of instantaneousreattachmentlength where
time-averagedvaluesare not available.Squaresare runswith 0.1 m stepat constriction.Solid squaresare bar-building
experiment.

exceededthe values (0.2-0.4 times vd) measuredby Schmidt

[1990] at three sites in Grand Canyon.

channel width downstream from the constriction, a Froude

Reattachtnent Point Migration and Fluctuation

During Clear Water Experiment

these conditions resulted in narrow recirculation

number of constricted flow exceeding 2, and a flow area

expansionratio lessthan 8 (Figure 6). Very slightchangesin

Over the range of conditions of the runs of the first

experiment,reattachmentlengthswere bimodal:either very
short or were much longer; intermediate conditionsdid not
exist (Table 2). The very short reattachment lengths (less
than 2 step heights) were associatedwith shootingflow
throughthe constriction,a hydraulicjump locatedabout 1

zones with

reattachmentlengths 13 times the step height (Figure 7). At

Froude numbers less than 1.5, reattachment lengths were
between 9 and 16.6 times step height. At these conditions, a
recirculation zone similar to that shown in Figure 3 developed; this condition best representedfield conditions observed in Grand Canyon. At Froude numbers typically less
than 0.5, poorly developedrecirculation existed and time-

Contraction Froude number > 2

area expansionratio< 8

Contraction Froude number > 2.2

area expansion ratio > 5

Contraction

Froude number 0.8 - 1.6

Contraction Froude number < 0.5'

Fig 7. Schematic
diagram
offlow
patterns
andrecirculation
zone
length
atdifferent
flow
conditions
Hatchured
line
is location
of hydraulic
jumpanddashed
lineis average
shearsurface.

2932

SCHMIDT ET AL.: FLUME SIMULATION OF FLOW AND SEDIMENTATION

0hrs
100

110

120

linear
ridge

130

11 hrs

mid-channel

aggradation
rateswithinthe recirculation
zonewerealways
less than 2 cm/h. The highest rates occurred in the downstream part of the recirculation zone, and these rates were
between 2 and 3 times as great as rates that occurred at the
linear ridge. Deposition within the recirculation zone did not
occur at the reattachment zone, but instead always occurred
well upstream.
Although the linear tidge was a persistent feature, its
volume increased at a slower rate than that of downstream

bar

parts of the recirculationzone. The location of the ridge

175 hrs
flatbed

ripples

dunes

changedwith time as parts of the ridge were scoured while

other parts aggraded.

Bedformsthat developedon the ridgeandplatformduring

the bar-building experiment are also mapped on Figure 8.

Within the recirculation zone, upstream-migrating dunes
,.% %,,. . 'y......
.-""-"*'22.5
hrs existednearthe reattachmentpoint, and upstream-migrating
ripples developedelsewhere. With time, dunes migratedinto
areas previously covered by ripples, so that a vertical
sequence exposed in the bar would have passed upward
from tipples to dunes. Areas of highest recirculation zone
..., .....
,::_____
,
30hrs aggradation were areas covered by dunes. Symmetrical
tipples (Figure 10) orientedwith creststransverseto the long
RP
axis of the flume existed near the center of the eddy.
Fig. 8. Bed topography and distribution of bed forms during Evidently, lateral shifts in the location of the eddy center
bar-buildingexperiment. Contour interval is 10 mm. Dark patterned subject this region to flow that reverses in direction.
...

areas were covered with dunes migrating in the directions indicated

by the arrows. Intermediate pattern areas are rippled areas, and
lightest patterned area are fiat bed areas. Zone over which reattachment point oscillated during experiments also shown. Cumulative
run time indicated at right.

Sediment Sorting and Capture Rates

Aggraded sediments within the eddy were finer than the

sediment feed. Within the eddy, the coarsest sedimentswere
deposited near the reattachment point and the finest sediments

collected

in

the

lee

of

the

obstruction

near

the

averaged reattachment length was between 7 and 11 times separationpoint (Figure 11). Sorting within the recirculation
step height; vortices migrated slowly downstream, and re- zone was similar to that reported for flood deposits in Grand
circulation was poorly developed. We never observed in- Canyon [Schmidt, 1990] and for other recirculation zone
stantaneous reattachment lengths between 2 and 6.2 times deposits [Page and Nanson, 1982].
step height, and only rarely were reattachment lengthsless
The rate of deposition within the recirculation zone was
than 8.2 times step height. Variation in the instantaneous
roughly proportional to the transport rate through the conlocation of the reattachment point during runs where con- striction (Figure 12). The relation in Figure 12 (which is not
stricted flow Froude numbers were less than 2 was nearly of
temporally ordered) could be expected to depend in part on
the same order as variation in time-averaged location among
the extent to which the eddy was filled at the time of each
all these runs. The large variation in instantaneousreattachmeasurement. Based on stratigraphic evidence in bars in
ment length in the first experiment (Figure 6, open circles Grand Canyon [Rubin et al., 1990], we suspectedthat the
and squares) masked any relation between reattachment
length and the flow area expansion ratio [Eaton and
Johnston, 1981].

110

1,20

"'

Topography and Bed Forms During

Bar-Building Experiment

1,30

...........
::...--

:::.-.
,:.........:.......:..::::.,

During the secondexperiment, sedimentwas deposited(1)

in the pool upstream from the constriction, (2) in and near
the recirculation zone, and (3) downstream from the recirculation zone. Downstream from the constriction, sediment
was first deposited beneath the shear zone as a linear ridge,
near the tight wall near station 120, and as a midchannel
expansion bar near station 130 (Figure 8).
The highest vertical aggradationrates in the entire flume
occurred at the midchannel expansion bar downstream from
the recirculation zone (Figure 9). These rates were as high as
5 cm/h and occurred over a larger area than the area of
recirculation zone aggradation. With time, the midchannel
bar became attached to the left wall and may have steered
downstream flow from the left to the right wall. Vertical

1,00

RP

========================

Fig. 9. Aggradationratesduringruns, determinedby subtracting previousbed topographyfrom resultantbed topography.Contour interval 10 mm/h, with supplemental5 mm/h dashed contours.

(a) Zeroto elevenhours;(b) 11-17.5hours;(c) 17.5-22.5hours;(d)

22.5-30

hours.

SCHMIDT
ETAL.:FLUME
SIMULATION
OFFLOW
ANDSEDIMENTATION

Fig. 10.

2933

Photograph
of symmetrical
ripplesthatdeveloped
beneaththe centerof theeddy.

capturerateswould be greatestwhenthe eddywas empty Changesin Reattachrnent Length and Do)t,nstrearn

and would decreaseas the eddy becamefilled. To evaluate Changesin Mean Velocity During
the effect of suchfilling on the capturerate, the data were Bar-Building Experiment
plotted in temporal order in Figure 13, which showsthat the
During the bar-buildingexperiment, reattachmentlength
cumulativepercentageof sedimentcapturedby the eddy decreasedand velocity increased as deposition occurred.
decreasedthrough time. The percentageof main stem sediDuring the initial part of this experiment, the reattachment

ment that was captured by the eddy between observations

(Table 3, column I0) decreasedfrom 37% (when the eddy

was empty) to 24% (when sand filled approximately32% of
the eddy volume). We hypothesizethat if the experimenthad
continued long enough, the mean capture rate would have
approached zero as the eddy filled to such an extent that
circulating currents and sedimentinput were restricted or
eliminated.

i 0.25

lOO

length was 13 to 16.6 times step height' after 17.5 hours the
reattachmentlength decreasedto 10.2 to 13.8 times step
height, and at the end of the experiment the reattachment
length was 9 to 10.6 times step height (Figure 8). The
location of the reattachment zone was compared to the
longitudinal pattern of deceleration of streamwise flow.
Mean downstreamvelocity in the channel expansionwas

7.5 HRS

y - 0.0192+ 0.226x
r2 - 0 933

80

/ C
'

_ 0.2
(separation
point)
/
/
/

//

leaIra
.... t point
.

,'7

0.15

5 HRS

11 HRS
O

/
::)

//

0.1

/ sediment
feed

20

uj

o
o.{

o.1

lO

0.05 0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

MAIN CHANNEL TRANSPORTRATE, IN KILOGRAMSPER SECOND

SIZE IN MILLIMETERS

Fig. l 1. Size distributionof composite

samples
of sedimentcollectedduringdifferentrunsat indicatedlocations.

Fig. 12. Eddy depositionrate and main channeltransportrate

for bar-buildingexperiment.Measurementperiod, in cumulativerun
time, is shown for each value.

2934

SCHMIDTET AL.: FLUMESIMULATION

OFFLOWANDSEDIMENTATION

volume
recirculation

stream tributary source of sediment. Schmidt [!987] concluded, basedon textural characteristics,that all deposition
had occurred during this one Paria flood event. He useda
locally developedstage-to-discharge
relation at the reattach-

of
zone

ment bar to estimate the duration of inundation of the bar

andestimated
themassof aggraded
sediment
fromsurveying
and trenching.Basedon thesemeasurements
and assumptions, Schmidt [!987] estimatedan aggradationrate of 4.2-

y=0.0824
+0.340x
- 0.00236x
2

30HRS

r2 = 0.999
11 HRS

-e '-' "-

--.-,

.-- --- -"

22.5HRS

17.5 HRS
,

..

....

....

5
10
15
20
25
CUMULATIVE VOLUME OF SEDIMENT TRANSPORTED

30

THROUGH CONSTRICTION,IN CUBIC METERS

Fig. !3. Cumulativevolume of recirculationzone filled with

sediment in relation to cumulative sediment transportedthrough
constrictionduring bar-buildingexperiment,both divided by the

volumeof theemptyoriginalrecirculation
zone.The totalvolumeof

7.5 kg/s (verticalaggradationrate of 0.6-0.9 cm/h) during

4.5-8.5

hours of inundation.

Schmidt [ 1987] estimated a capture rate by the recirculat-

ing current by comparinghis estimateof the massof sediment delivered by the Paria with the estimate of the massof
sedimentdeposited in the recirculation zone. The massof
delivered sediment was estimated by using the average
sedimenttransportrelationfor the Paria River gage[Randle
and Pemberton, 1987], the 25% value suggestedby Randle
and Pemberton [1987] for the proportion of transported
sediment within the sand size fraction (0.0625-2 mm), and

hourlywater dischargevalues. Basedon the estimatedrate

theoriginal
recirculation
zonewas25.2m3(28mlengthx 2 mwidth of sedimentdelivery to the Colorado River, the maximum
x 0.45 m average water depth between stations97 and 125). suspendedsedimentconcentrationof the ColoradoRiver
Cumulative run time for each measurement is shown.
immediately downstreamfrom the Paria was 55,000 rag/L,
and concentrations exceeded 10,000 mg/L for the entire

calculatedfrom (1) by calculatingA d from measurements

of periodof bar aggradation.The total deliveredsandloadto
thefiverwasbetween5.0 x 10?and10.0x 107kg.Themass
bed topographyand water surfaceelevationand estimating

ofsediment
aggraded
atthereattachment
bar,1.2x 105kg,

the location of the dividing streamlinebetween streamwise

was at least 0.1 and 0.2% of the total estimated sand load
and recirculatingflow. The longitudinalpatternof very high
deliveredby the Paria River flood, and at least 0.03 to 0.06%
velocity at the constrictionand steadydecreasedownstream
of the total suspended
load. The averageaggradationratefor
shown in Figure 14 exists becauseconservationof mass
the entire bar of 4.2-7.5 kg/s was not less than two to three
dictates that flow accelerate through the constrictionand
orders of magnitudeless than the estimatedsedimentdisdecelerate as flow cross-sectional area increases down-

stream; after 4 hours cumulative run time, v, decreased chargerate of the ColoradoRiver. Theseestimatesare only
minimumvalues of eddy capture rate because(1) the procontinuously downstream. Velocity increased with time

on the channel
between stations 120 and 130 because bed aggradation portionof the deliveredsedimentdeposited
decreasedA d; followingthe initial measurement
period,
decreased to a minimum but increased further downstream

between stations 120 and 130. The magnitudeand location of

the cross section of minimum streamwise velocity increased
and migratedupstream.Reattachmentoccurredwhere mean

bedupstreamfromthe measurement
siteis unknownand(2)
the actual delivered sediment load and its size distribution is
unknown.

Schmidtand Graf[1990] estimatedan aggradation

rateof
0.3 kg/s (vertical aggradationrate of 0.04 cm/h) duringa
33-dayperiodof inundationof a separationbar at Eighteen

downstreamvelocity was about0.3 m/s (4 hours),0.4 m/s(14

Mile Wash in Grand Canyon. Using the sand transport
hours), 0.5 m/s (20 hours), and 0.55 m/s (26 hours), respecrelationfor the ColoradoRiver at Lees Ferry, located30 km
tively. The location of the reattachmentzone was between
the cross section of minimum downstream velocity and the
cross section where accelerated flow reached its maximum

value at the two intermediate measurementsperiods (14 and

20 hours),and was locatedupstreamfrom the locationof the
minimum value in the last (26 hours) measurementperiod.

upstreamfrom this site [Randleand Pemberton,1987],the

average
waterdischarge
duringtheperiodof inundation,
and
the local aggradationrate, this site accumulatedat least
0.04% of the total main stem transported sand.
DISCUSSION

FIELD ESTIMATES OF AGGRADATION

Bar Construction

and Bed Forms

Available estimates of deposition rates in recirculation

Thereis generalsimilaritybetweenthe topographic
form
zones in Grand Canyon provide minimum values of the
percentageof sedimenttrappedin eddiesthat are at leastone of reattachmentbars in Grand Canyon and the bar that
duringthe bar-building
experiment.In bothcases,
order of magnitudeless than those measuredin theseexper- developed
iments. Schmidt [1987] estimated an aggradation rate at a

thehighest
partof thebarexistsin thedownstream
partof

reattachmentbar along the Colorado River 1.5 km downstream from the Paria River (see Schmidt and Graf[!990] for
locations). At this site, topographic surveys and recovered

the recirculationzone, althoughin GrandCanyon,thereis

greatvariabilityin the degreeto whichthesebarsfill the

upstream
partsof recirculation
zones.Field evidence
is
consistent
with the progression
of bar construction
during
Sediment
transport
directions
interpreted
gradedbetweenOctober 1985and January !986. During this theexperiments.
of bed formsand the progression
of
period, only one dischargeevent occurredin the Paria River from the orientation
change
in theflumeshows
thatthebarbuiltin
(October10, 1985,peakdischarge
41 m3/s),the onlyup- topographic
scour chains showed that 0.18 m of brown silty sand ag-

SCHMIDT
ETAL.'FLUME
SIMULATION
OFFLOW
ANDSEDIMENTATION

2935

an upstreamdirection.Upstreammigrationof the reattachment zone also caused scour of the farthest

downstream

areasdeposited
within the initiallylongerrecirculationzone.
There is inconsistency
betweenfield and flume in aggradationpatternsin the reattachmentzone, however. In the
field, we have observed the results of reattachment zone

aggradation
in theformof thicksequences
of climbingripple
structureswith migration directions consistentwith reat-

tachmentzone flow [Rubin et al., 1990]. In the flume,

depositiondid not occurin the reattachmentzone adjacent
to the wall; in fact, this zone was an area of local scour.

This discrepancymay be related to the experimental

channelgeometry.First, in these experimentsthere were
only smalldifferencesbetweenthe streamwiseand upstream
velocity in the vicinity of the reattachmentzone becauseof
the large magnitudeof streamwisedecelerationupstream
from the reattachmentzone (Figure 14). In Grand Canyon,
thesedifferencesare muchgreater[Schmidt,1990,Table 2
and Figure !1], such that more sediment-ladenwater is
delivered to reattachment zones in the field. Second, there
may be differencesbetween flume and field in the relative
position of main channel scour hole and the reattachment
zone leading to differencesin the location of shoalingof
streamwiseflow. Third, the area of low velocity surrounding

the instantaneous
reattachmentpoint was much smallerin
the flume than is observed in the field. This is because the

rough slopingbank of the river slows the flow more than

does the smooth vertical wall of the flume. In the field, the

areaof low velocity is largerandthe averagevelocity of the

reattachmentzone may be lower than in the flume. Scour in
downstreamparts of the reattachmentzone in the flume
could occur if high velocity downstreamflow occasionally
exists

in the reattachment

zone

when

the instantaneous

reattachmentpoint is at the upstreamend of the zone.

Characteristicsof RecirculatingFlow

The courseof floating objectsreflectsthe sum of all the

instantaneousflow directionsover the object's path. Observations of these path lines are consistentwith bed form
migrationdirectionsinterpretedfrom analysisof sedimentary structures[Rubinet al., 1990].However, both field and
experimentalresultsindicatethat there is greatvariation in
the instantaneous
velocity field, especiallynear the centerof
the recirculationzone and near the reattachmentpoint.
These instantaneousvariations are great enough to even
causesymmetricalripples to develop beneath areas where
there is net onshore

flow.

Some flow pattern characteristicsof these experiments

were influencedby the strongstreamwisedeepeningand are
consistent
with experimentsconcerningjets debouchinginto
still water. A high rate of decelerationleads to relatively
rapid decreasein the velocity head (Figure 4), increasein
water surface elevation downstream, and existence of an

adversepressuregradient. First, current meter measurements showed that instantaneous

backflow velocities

were

about 0.5 times vd, values much greater than typically

observedin the field. The high ratio of backflow to downstreamvelocity may simplybe due to fact that deceleration
of vd was sogreat. Second,unusualcrossoverof streamwise
flow from left to right wall occurredin someruns, suchas is
depictednearstation119in Figure3. Vorticesadvectedfrom
upstreamdevelopedwherevelocity and cross-stream
shear

2936

SCHMIDT ET AL.: FLUME SIMULATION OF FLOW AND SEDIMENTATION

2.5

1.5

I1
imO

Wm'"'XX

O3
1 01

RANGE
IN
LOCATION
OF
REATTACHM
ENT
POINT
AFTER
INDICATED

14

--0.5
SED
R
U
NTIME
area;
'f
d
h
"-.
14.,
'

vvn;3',...,'nr-a.

'

80

,,

I,

90

."-.

..:

bar deposition

100

110

120

130

140

DISTANCE, IN METERS

Fig. 14. Mean section velocity during different runs of bar-building experiment. Range of location of reattachment

point also shown.Values of v, determinedby estimatingarea of downstreamflow.

was high, and thesevorticesmay have beenstrongenoughto

produce upstream flow along the left wall. Third, reattachment lengthswere very long for the experimentalflow width
expansiongeometry.

and more adverse pressure gradients. In these runs, Froude

number was higher at higher discharges (Table 2, ninth
column) because water depth through the constriction did

Controls on Reattachment Length

because

Our laboratory experiments and field observationshave

documentedboth shorter and longer recirculationzonesthan

would be predictedusingpublishedrelationsbetweenstep
height and reattachment length. Relative to the relations of
Abbott and Kline [1960] for clear water flow at a single
backward step (Figure 15, line c), the reattachmentlengths
in our empty flume experimentsare too long. Reattachment
lengthsin the bar-buildingexperimentsare shorterand tend
to be similar to those of publishedrelations; field reattach-

not change(Table 2, eleventhcolumn). The magnitudeof the

adverse pressure gradient was higher at higher discharges
the elevation

of the water

surface

downstream

(Table 2, twelfth column) increased with increasing discharge.

At constant dischargeand nearly constant flow conditions

in the constrictionduringthe bar-buildingexperiment(Table

2, ninth column), reattachment length decreased as the
magnitudeof streamwisedecelerationdecreased(Figure 14).
The decrease in the rate of deceleration

was due to the

decrease in flow area expansion ratio (Figure 6, solid

squares;Figure 16, solid squares); there was no changein
ment lengthstypically are too short (Figure 15). As will be the magnitudeof the adverse water slope (Table 2, runs
discussedbelow, these deviationsresult from topography 15-17, eleventh to twelfth columns).
Reattachment length shortening was also related to the
that influencesthe pressuregradient. In flows with obstacles
fact
that streamwisevelocity(v d) nearthe initially very long
downstream(suchas the bar that was depositedin our flume
experiments or cobble bars in Grand Canyon) the flow reattachment zone nearly doubled due to midchannel bar
accelerates,resultingin a diminishedadversepressuregra- deposition.At the beginningof the bar-buildingexperiment,
dient and a shorter reattachment length. In contrast, where streamwisevelocity was about 0.3 m/s oppositethe reattachdownstream flow is a deep pool (as in our empty flume ment zone (Figure 14, station 130, 4 hours). At the time of
vctwas more than0.5 m/s at this same
experimentand in very deep pools in rivers), flow deceler- the lastmeasurement,
ates, resultingin an increasedadversepressuregradientand site. As the channel filled with sediment and velocity necessarilyincreased,areasthat had formerlybeen of low streamlonger reattachmentlength.
Seven runs were conductedat different dischargesbut at wise velocity or stagnantbecame areasof higher streamwise
the same tailgate elevation (Table 2, thirteenth column) and velocity.The midchannelbar essentiallyfilled areasthat had
thereforesimulatechangingflow conditionsin an expansion been part of the reattachment zone.
whose water surface elevation is set by a downstream
These results suggestthat there shouldbe wide variation
control. The experimentalresults of these runs (1, 6, 10, in reattachment lengths in natural rivers, due in part to
14--17) are consistent with field data that show that time- presenceor absenceof downstreambars and narrowingof
averagedreattachmentlength increaseswith discharge.Al- downstream walls which would also accelerate flow. Figure
though instantaneous fluctuations in the location of the 15 showsGrandCanyonreattachmentlengthdata, reattachreattachment point varied greatly in these runs, the time- ment lengthsduring the bar-buildingexperiment, and other
averaged reattachment length was greatest at highest dis- experimentaldataplotedin relationto stepheight.Data from
charge (Figure 16). Lengthening at higher dischargewas specificfield sitesare shownas separatelinesextendingover
associatedwith higher Froude numbers of constrictedflows the range of different measuredreattachment lengths and

SCHMIDT
ETAL.:FLUME
SIMULATION
OFFLOW
ANDSEDIMENTATION

2937

I
I
i
I

I'-'

3::

I--

I
!

-=

!
0

30

z 2O
0

range
for
first
experiment

j,/

bar-building
/ ! ._.,,,t
,'""
oxporimort
! !/'' -

c
-

71

o
z

o10

large
mid-channel

,*

iTM

,EXPLA-']

.... !

?,,

.;.,;

/- \
11/'"'
deb;isfan
shortens

':'/

,
0

--

I
7

SCALED-STEP
HEIGHT(V-Wc)AN
c

Fig. 15. Changein recirculation

zonelengthin relationto changein channelexpansiongeometryshowingthatfield
siteshave shorterscaledrecirculationzonesthan in flumeexperiments.Lines A, B, and C are for flume studiesof
double-backward
steps(maximumobservedvalues),double-backward
steps(minimumobservedvalues),and singlebackward steps,respectively,from Abbott and Kline [1962].Solidverticallines above abscissavalue of 1.7 are range
in reattachment lengthsfor first experiment(longer line) and range in recirculation zone length for bar-building
experiment(shorter line) after indicatedrun time. Inset diagramshowshow field measurementswere made. Number
of eachfield site refersto Schmidt[1990,Table 1]. Scaled-stepheightchangesin field with changingdischargebecause
we is held constantat value for lowest observeddischarge.

geometric conditions. Reattachment lengths were deter-

1990, Table 1]. The parameter unconstrainedreattachment

minedby mappingrecirculationzoneflow patternsat differ- lengthwas proposedto permit comparisonof reattachment

ent dischargesonto 1:5000 scale air photos and then esti- length in channel expansions of different curvature and
mating the unconstrained reattachmentlength [Schmidt, shape. These data show that unconstrainedreattachment
length increasesas step height increases. There is great
variabilityin the rate of lengtheningat eachfield site, but all
2O

reattachment lengths are shorter than those reported by

Abbott and Kline [1962] for double backward facing steps
(Figure 15, A and B) and are typically shorterthan for single
backward facing steps (Figure 15, C).
What field conditions might explain the shorter reattach.4.9
mentlengthsin Grand Canyon?In Grand Canyon, midchan.3-4.9
nel cobble and gravel bars are common downstream from
many constrictions,especiallyrapids with the steepestgra3.9
dients [Webb et al., 1988]. These bars are analogousto the
expansionbars describedby Baker [1984]and analogousto
3.4
the midchannelbar that developedin the secondexperiment.
In other field cases, a downstream debris fan or talus cone
narrows the channel and shortensthe length of the channel
...
I
0.1
0.2
0.3
0.4
0.5
0.6
0.7
expansion.Of the field sites plotted on Figure 15, the
DISCHARGE,
INCUBIC
METERS
PERSECOND
channel expansionsat sites 5, 10, and 11 are strongly
influencedby a midchannelbar (site 10, Granite Rapids)or a
Fig. 16. Time-averaged
reattachment
length
forthesame
down- debris fan (site 5, Eighteen Mile Wash, and site 11, One
stream
gateheight(0.800m aboveanarbitrary
point,seeTable2,
thirteenth
column).
Errorbarsindicate
totalrangein instantaneousHundred Twentytwo Mile Creek). These three sitesplot the
reattachment
length,in stepheights.
Numbers
adjacent
to circles furthest from the experimental values of Abbott and Kline

areflow-area
expansion
ratioforeachrun.Open
circles
areforclear [1962],suggesting
that downstream
channelirregularities
do
water,emptyflumeexperiment,
andsolidsquares
arefor bar- constrainthese reattachment lengths.
building
experiment.
Arrowindicates
theprogress
of thesecond
experiment.

Midchannel bars not only shorten the length of recircula-

2938

SCHMIDTET AL.: FLUMESIMULATIONOFFLOWAND SEDIMENTATION

tion zonesat high discharges

by necessitating
acceleration
over them, but at low discharges
reattachmentlengthmaybe
shortenedbecausebars are emergent.Schmidt [1990] implied that recirculationzone length is controlledby the

the sediment-transporting
event. In the planningof any

regulated flood intended to reconstruct eroded reattachment

and separationbars, one must be sure that there is sufficient
sedimentsupply available to maintain high main stemtranschance intersection of the shear zone and irregularly ori- port rates for durationssufficientto result in net deposition.
Field evidence showsthat some recirculating current bars
ented downstreamchannelbanks.Theseexperimentsshow

that accelerationcausedby narrowingof the channel,shoal- are scouredduringrise of a regulatedflood [Schmidt,1990;

ing over midchannelbars, and associated
decreasein flow Schmidt and Graf, 1990]. These observations,if representaarea expansionratio also influencesthe point whereflow tive, imply that reattachmentbars are either a primary
reattaches. In rivers where midchannel bars are larger and sourceof sedimentfor the main stemor a relativelyquickly
more numerous, reattachment lengths should be even mobilized source. The bars can not begin to rebuild until
shorter than those measuredin Grand Canyon. The results sediment concentration in the main stem exceeds that in the
alsoindicatethat successfulmodelingof reattachmentlength eddy. To achieve net aggradation, any flow designedto

in riverswill dependon predictionof flow conditionsin the

constriction,identificationof the downstreamchannelcontrol, predictionof the watersurfaceelevationupstreamfrom
that controlinto the expansionin question,andpredictionof
main channelbed behavior during the modeled flows.

Implicationsfor Restorationof Eddy Bars

in Regulated Rivers

These experiments show that deposition rates within

recirculation zones increase as main channel transport increases due to increased sediment concentration at constant

discharge.Aggradation rates within the recirculation zone

were highly correlatedwith main current sedimenttransport
rates and indicate, within the range of these experimental
conditions, that about 20% of main channel transport was
depositedwithin the recirculationzone. We only conducted
our bar-building experiment at one constantflow, and we
have no data to suggestwhether the percent of sediment
trapped in an eddy increasesat higher main channeltransport rates causedby (1) further increasesin sedimentconcentration at constant discharge or (2) increased water
discharge and increased sediment concentration, such as
indicated by the calculationsof Andrews [1991].
Although we found a direct correlation between cumulative transportthroughthe constrictionanddegreeof fillingof
the recirculation zone, our experiment only continueduntil
the recirculation zone was filled to about 40% of its original
volume. These data suggestthat this relation decreasesin
slope;sucha relation shouldexist becauseas the recirculation zone fills it can be expectedthat circulationwithin the
eddy and delivery of sediment into the eddy would both
decrease. This idea is consistent with stratigraphic sequences observed in Grand Canyon reattachment bars. In
these sequences,grain size decreasesas reattachmentzones
fill and sedimentarystructureschangefrom those indicating
vigorouscirculation (dunes and erosionalflutes filled with
coarse sand depositedfrom turbulent suspension)to those
indicating weak circulation (ripples and mud drapes). An
importantunresolvedquestionconcernshow the proportion
of capturedsedimentvaries as a functionof the strengthof
recirculatingflow, which is not only a functionof the extent

reconstruct

bars must first rebuild bars to the level that

existedprior to initial scour, then continueto add sediment.

The role of sediment availability in the design of bar
reconstructionfloods places great importance on accurate
accounting of sediment sources prior to design of such
floods. If it turns out that most sediment in bedrock gorgesis
stored in recirculation zones, rather than in main channel

pools,than it may be impossibleto sustainnecessarysediment transport rates for sufficient durations without augmentingnaturally availablesupplies.Otherwise, net erosion
of recirculation zones in upstream sediment-deficient
reaches will result. Planning of bar reconstruction floods
must be based on predicting eddy depositionrates, predicting the duration of high main channel sedimenttransport,
and accountingfor the wide variation in channel and recirculationzone geometrythat exists in natural rivers.
CONCLUSIONS

During a 30-hourexperimentwhen a mixed size distribution of sediment was added to the flow at a rate of 0.5-1.0

kg/s, a bar formed within the zone of recirculating current

downstream from a channel constriction. The general topo-

graphicform of this bar was that of a linear ridgewidening

downstreaminto a broad platform. Within the recirculation
zone, the highestdepositionrates occurrednear the downstream end, but these rates were always less than rates on a
midchannelbar that formed just downstreamfrom the reattachment zone. These results are consistent with field inves-

tigationsof the sedimentology

of reattachmentbarsin Grand
Canyonwhichshowthat depositionoccursin the vicinityof
the reattachment zone and fills the primary eddy in an
upstreamdirection.

Bed forms on the surfaceof the ridge and platform and

repetitive topographicsurveys show that the entire bar
aggradedin an onshoreand upstreamdirectionduringthe
experiment.Sort'ragof sedimentwithin the recirculation
zoneis sufficientto form a depositthat is generallyfinerthan
the size distributionin transportby the main channel;the

finestsizesare depositedin the lee of the debrisfan. This

areal distribution is similar to that described in Grand

Canyonfor separationand reattachmentbars, and the pro-

gressionof bar deposition

is consistent
with the modelof
of the spatial structure of the deceleratingjet. Flume and GrandCanyonbar depositiondevelopedfrom interpretation
structuresand sequentialfield topographic
field measurementsof eddy capture rates likely differ by at of sedimentary

to which the recirculation

zone is filled but is also a function

least one order of magnitude;however, available field measurements are subject to substantial error.
While these data show that the rate of eddy deposition
dependson the rate of main channeltransport,the extentto
which a recirculation zone fills depends on the duration of

surveys.The percentageof main stem sedimentthat was

capturedby the eddy during the bar-buildingexperiment

decreased
from 37% (whenthe eddywas empty)to 24%
(whensandfilledapproximately
32%of the eddyvolume).
The reattachment
pointoccurswhereexpandingvortices

SCHMIDT
ETAL.:FLUME
SIMULATION
OFFLOW
ANDSEDIMENTATION

2939

impingeon the flumewall, andthis pointfluctuatedovera

Koster and R. J. Steel, pp. 87-98, The CanadianSociety of

PetroleumGeologists,Calgary,Alta., 1984.
recirculationzone had instantaneous
velocitiesthat varied Best, J. L., The morphologyof river channel confluences,Prog.
Phys. Geog., 10, 157-174, 1986.
overan entire360-degree
range.Resulting
bedformsthat Chang,P. K., Separationof Flow, Pergamon,New York, 1966.
developed
nearthecenteroftheeddyresembled
oscillatoryDriver, D. M., and H. L. Seegmiller,Featuresof a reattacl-dng
turbulentshear layer in divergent channel flow, AIAA J., 23,
ripples.
Oscillatory
ripplesthatin thefieldareformedby

rangeof about5 timesthe stepheight.Somepartsof the

reattachment
pointfluctuation
did not developduringour

163-171, 1985.

experiments.In fact, depositiondid not occurnear the wall Eaton, J. K., and J.P. Johnston,A review of researchon subsonic
turbulentflow reattachment, AIAA J., 19, !093-1100, 1981.
in the reattachment zone.
Ikeda, H., Experiments on bedload transport, bed forms, and
sedimentarystructures using fine gravel in the 4-meter-wide
Recirculation
lengthsweresimilarwithinbroadranges
of
hydraulicconditions.
Whereconstricted
flow was highly flume, Univ. of TsukubaEnviron. Res. Cent. Pap. 2, Ibaraki,
Japan, 1983.
supercritical,the length of the recirculationzonewas about
Kieffer, S. W., The 1983hydraulicjump in Crystal Rapid: Implicatwo step heights, but where Froude numberswere lessthan
tions for river-running and geomorphicevolution in the Grand
about2, the zone of recirculatingcurrentwas of similarsize
Canyon, J. Geol., 93,385-406, 1985.
over a range of area expansionratios. To someextent, Kieffer, S. W., Hydraulic maps of major rapids, Grand Canyon,
Arizona, U.S. Geol. Surv. Geophys.Invest., Map, 1-1897, 1988.
reattachmentlengths were longer, and recirculationflow
Leeder, M. R., and P. H. Bridges, Flow separation in meander
patternsbetter developed, at intermediateFroude numbers
bends, Nature, 253, 338-339, 1975.
between 0.8 and 1.6. As aggradationof a midchannelbar Leopold, L. B., The rapids and the pools--Grand Canyon, U.S.
proceededand the bar migrated upstream,reattachment Geol. Surv. Prof. Pap., 669-D, 1964.
length decreased and the recirculation zone shortened. Nelson, J. M., Experimental and theoretical investigationof lateral
separationeddies (abstract), Eos Trans. AGU, 72,218-219, 1991.
These processeswere linked becauseflow acceleratedover
Page, K., and G. Nanson, Concave-bank benches and associated
the midchannelbar, which preventedstagnationalongthe
floodplainformation, Earth Surf. ProcessesLandforms, 7, 529543, 1982.
wall in areas previously of adversewater slope. As the
midchannelbar retreated upstream, accelerationoccurred Randle,T. J., and E. L. Pemberton,Resultsand analysisof STARS
modelingefforts on the Colorado River in Grand Canyon, Glen
where decelerationhad previouslybeen the case.

Canyon Environ. Study GCES-38, U.S. Bur. of Reclam., Salt

Sedimentation can occur in bedrock gorgeswhere flow

Lake City, Utah, 1987.
separationand flow stagnationexist [Baker, 1984].The areal Rubin, D. M., J. C. Schmidt, and J. N. Moore, Origin, structure,
and evolution of a reattachment bar, Colorado River, Grand
extent and magnitude of' deposition depend on the areal
Canyon, Arizona, J. Sediment. Petrol., 60, 982-991, 1990.
extent of the stagnatedflow, the areal extent of the zone over
Schmidt,J. C., Geomorphologyof alluvial-sanddeposits,Colorado
which reattachment point oscillationoccurs,the flow strucRiver, Grand Canyon National Park, Arizona, Ph.D. dissertation,
ture of the remainder of the recirculation zone, the rate of
main channel sediment transport, and the duration of the

Johns Hopkins Univ., Baltimore, Md., 1987.

Schmidt, J. C., Recirculating flow and sedimentation in the Colorado River in Grand Canyon, Arizona, J. Geol., 98, 709-724,

transportingevent. In regulatedrivers where degradationof

1990.
reattachment and separationbars has occurred,the task in Schmidt, J. C., and J. B. Graf, Aggradation and degradation of
planningbar reconstructionfloodsis to predict the location
alluvial sand deposits, 1965 to 1986, Colorado River, Grand
of stagnationpoints, the size of the zone over which these
Canyon National Park, Arizona, U.S. Geol. Surv. Prof. Pap.,
1493, 1990.
stagnationpoints oscillate, the nature and stability of the
generalrecirculatingflow field, the rate of sedimentdelivery Signell, R. P., and W. R. Geyer, Transient eddy formation around
from main channel to recirculation zone, and the duration

during which that delivery from main channelto recirculation zone will occur.

headlands,J. Geophys. Res., 96, 2561-2575, 1991.

Simpson,R. L., Turbulent boundary-layerseparation,Annu. Rev.
Fluid Mech., 21,205-234,

1989.

Tamai, N., T. Asaeda, and H. Ikeda, Study on generation of

periodicallarge surfaceeddiesin a compositechannelflow, Water
Resour. Res., 22, 1129-1138, 1986.

Webb. R. H., P. T. Pringle, S. L. Reneau, and G. R. Rink,

Acknowledgments.Field work wassupported

by the GlenCanMonument Creek debris flow, 1984:Implicationsfor formation of
yon EnvironmentalStudiesprogramof the U.S. Bureauof Reclarapidson the ColoradoRiver in Grand Canyon National Park,
mation. Travel funds were providedby the MiddleburyCollege
Geology, 16, 50-54, 1988.
FacultyDevelopment
Fund(J.C.S.)andtheU.S. Geological
Survey Wolanski, E., J. Imberger, and M. L. Heron, Island wakes in
G. K. GilbertFellowship(D.M.R.). T. O. Manley,Department
of
shallow coastal waters, J. Geophys. Res., 89, 10553-10569, 1984.

Geology,MiddleburyCollege,conducted
the surfaceandvolume Yeh, H. H., W. Chu, and O. Dahlberg, Numerical modeling of
modelingcalculations.Assistance
duringthe experiments
wasprovidedby Y. KodamaandH. Iijima. We thankJ. G. Bennett,J. M.
Nelson,andT. L. Vallierfor carefulreviewof themanuscript
and

separationeddies in shallow water, Water Resour. Res., 24,

607-614, 1988.

suggestionsfor improvements.

H. Ikeda, Environmental Research Center, University of

Tsukuba, Ibaraki, 305 Japan.
D. M. Rubin, U.S. Geological Survey, MS 999, 345 Middlefield
REFERENCES
Road, Menlo Park, CA 94025.
J. C. Schmidt, Department of Geography and Earth Resources,
Abbott,D. E., and Kline, S. J., Experimental
investigation
of WatershedScienceUnit, College of Natural Resources,Utah State
subsonicturbulentflow over singleanddoublebackwardfacing University, Logan, UT 84322.

steps.J. BasicEng., 84, 317-325,1962.

Andrews,E. D., Deposition

rateof sandin lateralseparation
zones,
Colorado
River(abstract),
EosTrans.AGU, 72,219,1991.

Baker,V. R., Floodsedimentation

in bedrock
fluvialsystems,
in
Sedimentology
of Gravels
andConglomerates,
editedby E. H.

(Received
October
8, 1992;
revised March 1, 1993;

accepted March 16, 1993.)