One-Dimensional Sediment Transport Modeling of Pool Recovery Along A Mountain Channel After A Reservoir Sediment Release

REGULATED RIVERS: RESEARCH & MANAGEMENT
Regul. Ri6ers: Res. Mgmt. 17: 251– 273 (2001)

DOI: 10.1002/rrr.617
ONE-DIMENSIONAL SEDIMENT TRANSPORT MODELING OF POOL

RECOVERY ALONG A MOUNTAIN CHANNEL AFTER A RESERVOIR
SEDIMENT RELEASE
SARA L. RATHBURN* AND ELLEN E. WOHL
Department of Earth Resources, Colorado State Uni6ersity, Fort Collins, CO, USA
ABSTRACT
A reservoir sediment release along the North Fork Cache la Poudre River (North Fork) in northern Colorado
resulted in a massive fish kill and channel sedimentation that filled pools critical to fish as overwinter habitat.
Recognition of the hazards associated with a large influx of sediment into a riverine ecosystem is critical for a greater
understanding of the effects of sediment releases, and hence, future management of sediment within reservoirs. Two
one-dimensional sediment transport models, HEC-6 and GSTARS 2.0, were evaluated for applicability to predict
sediment removal along the steep gradient, bedrock-controlled pool-riffle North Fork. The ability of both models to
identify flushing discharges that assist channel recovery was also evaluated. Two modeling scenarios representing a
low and high flushing discharge were modeled. Within each scenario, two levels of simulations were conducted to
represent conditions of data availability, a default simulation for limited data input, and a robust simulation that
utilized the entire set of field data, collected over a 1-year period. The models were calibrated against quantitative
measurements of pool bed elevation obtained during field resurveys. Preliminary analyses were conducted to identify
the appropriate sediment transport equations. HEC-6 results indicate that long-term, robust simulations yield the
closest agreement between predicted and measured pool bed elevation change. More than 50% of the actual scour and
deposition within three pools was modeled using HEC-6. Modeling accuracy using GSTARS 2.0 was considerably
lower, and it appears that the present construct of the model does not reflect the physical processes operating along
the North Fork. Computer models are useful tools in the sediment management decision process, provided adequate
data collection and calibration are conducted. In situations where restrictions exist on available flushing discharges,
sediment transport modeling can assist management decisions, and modeling is always preferable to uncalibrated
estimates. Copyright © 2001 John Wiley & Sons, Ltd.
KEY WORDS: flushing discharges; pool recovery; sediment transport modeling; reservoir releases
INTRODUCTION
It has long been recognized that river regulation resulting from dams causes major alteration in sediment
dynamics and channel morphology in downstream reaches of the river (Williams and Wolman, 1984;
Collier et al., 1996). Upstream from the dam, sediment accumulation within the reservoir will eventually
threaten the storage capacity of the reservoir, and the operating efficiency and expected life of a dam. In
the semi-arid western US, where sediment fluxes from hillslopes may be greater than under any other
climate (Schumm, 1965), the hundreds of small irrigation reservoirs constructed in the last century are
undergoing siltation, forcing irrigation companies to address sediment management issues. The estimate
of world wide reservoir storage capacity loss, resulting from siltation alone, is approximately 4.9 E10 m3,
or the equivalent of $6 billion in replacement costs every year (Fan and Springer, 1990). Sediment
management practices within reservoirs may include flushing or sluicing reservoir sediments, sometimes
voluminously and catastrophically, into the downstream channel. In Colorado alone, five large-scale
sediment releases from reservoirs within the last decade have resulted in massive fish kills (Wohl, 1999).
Sediment released into the downstream channel can be highly detrimental to the aquatic biota and their
* Correspondence to: Department of Earth Resources, Colorado State University, Fort Collins, CO 80523, USA. E-mail:
rathburn@lamar.colostate.edu
Recei6ed 4 April 2000

Re6ised 30 July 2000
Copyright © 2001 John Wiley & Sons, Ltd. Accepted 25 August 2000
252 S.L. RATHBURN AND E.E. WOHL
habitat, can significantly alter the channel morphology, and can impair diversion and irrigation structures,
water quality, and recreational opportunities downstream (American Rivers, 1999; Wohl, 1999). The
sediment infilling may also affect the capacity of the channel to convey water and sediment, especially
during floods. In other cases, the flushing of reservoir sediment has had less detrimental, short-term
effects on benthic populations and subsequent selection for organisms with rapid life cycles (Gray and
Ward, 1982). Depending on the timing of the sediment release and the channel configuration, the
residence time of sediment in the downstream channel may range from months to years.
Given the current regulatory, environmental and recreational interests in rivers downstream from dams,
indiscriminate releases of large volumes of sediment are no longer acceptable. If sediment releases are
desirable to manage the recognized problem of sediment accumulation in reservoirs, then methods for
assessing the degree of channel change and channel recovery time are required to understand the potential
downstream impacts of sediment releases. This paper describes the application and assessment of two
widely used one-dimensional sediment transport models to a steep gradient mountain channel that
underwent a reservoir sediment release. The primary objective of this study is to test the ability of the
models to simulate field-measured erosion and deposition following the sediment release. In addition,
recommendations for reservoir operation are presented to identify minimum discharge requirements to
flush sediments from critical habitat reaches.
At a study site along the North Fork Cache la Poudre River (North Fork) in northern Colorado
(Figure 1), pool infilling and a massive fish kill occurred as a result of a reservoir sediment release during
September 1996. Because of the thriving, pre-release wild trout fishery downstream, flushing of pools to
recreate overwinter fish habitat was of prime concern. To date, most flushing flow research has been
directed at identifying discharges that effectively flush fine sediments from spawning gravels within riffles
and runs (Kondolf and Matthews, 1993; Wilcock et al., 1996). In the case of the North Fork, minimal
amounts of sediment from the reservoir release accumulated in riffles. As a result, the one-dimensional
modeling was directed at pool flushing to enhance overwinter pool volumes for fish. The models were
evaluated with regard to the ability of each model to simulate measured sediment aggradation and
Figure 1. Location map of the North Fork study site and modeling reaches. Roman numerals I and II indicate modeling reaches
with cross section locations shown as dashed lines perpendicular to the channel. Certain cross sections are numbered and are referred
to in the text
Copyright © 2001 John Wiley & Sons, Ltd. Regul. Ri6ers: Res. Mgmt. 17: 251 – 273 (2001)
ONE-DIMENSIONAL SEDIMENT TRANSPORT MODELING OF POOL RECOVERY 253
degradation within pools over a known annual hydrograph, as opposed to simulating hypothetical
scenarios of sediment transport and discharge.
One-dimensional sediment transport models have become increasingly useful predictive tools to assess
aggradation and degradation within channels, particularly in and around hydraulic structures. In
addition, where long-term predictions are required, numerical models are the only way to simulate
aggradation and degradation of the channel bed. We hypothesize that one-dimensional sediment transport
models are useful tools to estimate channel recovery times and flushing flow needs after a reservoir
sediment release. In situations of limited water availability, particularly, where any discharge release is
expensive and undesirable, models may be useful in designing a minimum flow regime necessary to flush
sediment and restore the aquatic habitat.
STUDY SITE
3
In September 1996, approximately 7000 m of fine-grained sediment (clay–gravel) were released from
Halligan Reservoir into the North Fork (Figure 1). The North Fork is a steep gradient (0.011–0.02),
bedrock-controlled mountain channel, with a well-developed pool-riffle sequence, along a strongly
armored cobble and boulder bed. The river is incised into granite of the Silver Plume Formation, and
flows within a broad gorge, 160 m deep, within the South Platte River drainage basin. Total drainage area
above Halligan Reservoir is 904 km2. The flow regime along the North Fork is strongly snowmelt driven,
with minimal flow throughout the winter. Once the reservoir fills in spring, spilling over the dam produces
a marked peak in the hydrograph, usually lasting from May to July. The range of mean annual discharge
for the period of record (1997 to present) is 0.1 to 40 m3/s.
Patterns of sediment transport, aggradation and degradation following the North Fork sediment release
are described in detail in Wohl and Cenderelli (2000). The main by-product of the 1996 release was the
infilling of pools, to varying degrees, for more than 8 km downstream, forming a thick veneer of
fine-grained sediment over the original cobble/boulder substrate. At 0.5 km downstream from the dam,
pools were completely filled, up to 3.5 m deep, with a uniform layer of gravel- to clay-sized sediment,
forming, essentially, a plane bed within the channel (Figure 2). At 3.2 km downstream from the dam,
pools were half filled with, primarily, sand, silt and clay. Grab samples of marginal pool deposits at this
Figure 2. Sediment filling of the North Fork after the reservoir sediment release from Halligan Reservoir, September 1996. Person
in the photo is in a pool filled with fine-grained sediment 3.5 m deep, and view is downstream. Photo courtesy of Janet R. Curran
distance downstream, collected in early October following the sediment release, contained approximately
8% sand, 82% silt, and 10% clay. Within the entire 8 km, deposition within riffles was minimal and
non-uniform, occurring as bars along the margins of the channel, and in the lee of large clasts within the
channel bed. Fine sediment also infiltrated the interstices of the coarse-grained riffle substrate to a depth
of approximately 6 cm (Wohl and Cenderelli, 2000).
An estimated 4000 fish, mostly rainbow trout (Salmo gairdneri ), brown trout (Salmo trutta) and white
suckers (Catostomus commersoni ), were killed by the sediment release. The fish loss was significant
because of the noted wild trout fishery within The Nature Conservancy’s Phantom Canyon Preserve that
occupies approximately 10 km along the river corridor downstream from the dam. In an attempt to flush
sediment downstream and clear pools critical to fish, a stepped-experimental discharge was released from
Halligan Reservoir for 1 month, from late February to late March 1997, with a peak of 3.4 m3/s (Figure
3). In the spring of 1997, flow increased naturally with the onset of snowmelt in the headwaters of the
North Fork, and peaked on 3 June at 10.1 m3/s.
Two study reaches were established for modeling purposes after the sediment release to monitor
sediment movement through the river over time. Study Reach I, located 0.5 km downstream from
Halligan Dam, is 120 m long and includes seven surveyed cross sections through one riffle and one pool
(Figure 1). Study Reach II, 3.2 km downstream from the dam, is 400 m long, with 19 surveyed cross
sections, including two pools and two riffles. Because wild trout populations in the North Fork rely
heavily on pools for overwinter habitat, the focus of this research was on sediment removal from the
pools. The pools of interest in this paper are informally called the Goose Pool (Study Reach I), the Ouzel
Pool, and the Tick Pool (Study Reach II; Figure 1).
MODEL CHARACTERISTICS
The two, one-dimensional models selected for application to the North Fork were HEC-6, developed by
the US Army Corps of Engineers, Hydrologic Engineering Center (US Army Corps of Engineers,
Figure 3. Hydrograph for the North Fork beginning on 22 February, the day flow began to rise from winter base flow and a
1-month experimental discharge was released from Halligan Reservoir with the intent of flushing sediment from the downstream
channel. Two modeling scenarios were conducted, which correspond to the delineations shown on the hydrograph: an experimental
discharge simulation and a snowmelt runoff simulation
1998a,b), and GSTARS version 2.0, created by the US Bureau of Reclamation (Yang et al., 1998a). The
models were selected because they are economical, widely accessible, and likely to be used as predictive
tools by water resource managers faced with channel infilling associated with released reservoir sediment.
It should be noted that neither model is strictly well suited for a steep gradient mountain channel with an
immobile bed. The question of whether or not moderately user-friendly, one-dimensional numerical
models can be used in such environments is a major component of this research, and one we tested by
applying the models to the North Fork sediment release.
The primary criteria used to evaluate HEC-6 and GSTARS was the ability of the model to reproduce
pool scour and fill measured in repeat field surveys under varied discharges to identify the appropriate
flushing flows to minimize or eliminate the hazards of pool infilling from reservoir sediment releases. An
additional aspect of the model evaluation was the time investment and level of expertise required to obtain
meaningful simulation results. A truly useful reservoir management tool would not require extensive
training in hydraulics and sediment transport, would have data requirements that are not too labor or
cost intensive, and could be developed and run in a timely manner, yielding useful raw model output.
However, the state of numerical models suggests that this will not be the situation for the foreseeable
future. Therefore, we applied existing models to the problem of pool infilling along the North Fork, and
interpreted the reasonableness of the output in the context of existing model limitations.
HEC-6
The HEC-6 model is a one-dimensional model that predicts scour and deposition within rivers and
reservoirs (US Army Corps of Engineers, 1998a). In river applications, HEC-6 simulates uniform changes
in river bed elevation over the entire width of the channel which is caused by erosion and deposition over
time under subcritical flow. The model has no provisions for simulating lateral channel changes, such as
meander migration, or lateral changes in bed slope. The governing equations in HEC-6 include the energy
equation, and conservation of mass for water and sediment. The momentum equation is not included in
HEC-6, so environments with rapid fluctuations between subcritical and supercritical flow are inappropri-
ate for modeling. In addition, HEC-6 assumes that sediment supply and demand are satisfied within each
reach at each time step, and the model takes into account the effects of sediment gradation. HEC-6 is one
of the most widely used and economical, commercially available sediment transport models. The most
recent version of HEC-6 can be downloaded from the Internet at http: //www.wrc-hec.usace.army.mil/.
Three model components comprise HEC-6, and require specification by the user. These include a
geometric component consisting of surveyed channel cross sections, a hydrologic component of discharge
at the upstream boundary, represented as a series of steady, uniform flows, and a sediment component,
including inflowing sediment load, sediment rating curve, and the gradation of bed material. Several
default options allow the user to select recommended input settings, should some of the input data be
unavailable, or should the user be unsure of which option to choose. In other cases, input settings offer
several choices, such as selecting one of 14 sediment transport equations. The output of interest in this
application of HEC-6 is the average or uniform bed change at a given channel cross section (Figure 4).
GSTARS 2.0
In contrast to HEC-6, GSTARS is more accurately described as a quasi-two-dimensional model that
utilizes a stream tube approach to accommodate differential scour and deposition over the width of a
cross section (Yang et al., 1998a). Stream tubes are conceptual tube-like surfaces whose walls are defined
by streamlines, imaginary lines which show the direction and magnitude of velocity as the tangent at every
point along the line, at each instant in time. In GSTARS, hydraulic parameters and sediment routing
computations are made for each stream tube, allowing the position and width of each stream tube to
change. In this way, vertical and lateral variations in cross sectional elevation can be simulated (Figure 4).
The governing equations are largely similar between HEC-6 and GSTARS, except that GSTARS
incorporates the momentum equation in backwater computations when the flow regime changes from
subcritical to supercritical or 6ice 6ersa. GSTARS was selected for testing along the North Fork to
Figure 4. Schematic cross section showing differences in model characteristics between HEC-6 and GSTARS 2.0. The solid black
pattern indicates scour along the bed of the channel, and the stippled pattern indicates deposition
evaluate the applicability of another comparable sediment transport model, to overcome suspected
limitations of the purely one-dimensional HEC-6 model, and to potentially improve the accuracy of the
output, without substantially increasing input effort.
Input for GSTARS is similar to HEC-6, but offers a broader range of options with very few default
choices built into the model. Other differences between the two models include the ability to specify the
number of stream tubes at each cross section (up to five), a different list of sediment transport equations
from which to choose, and determining whether non-equilibrium sediment transport and stream power
minimization procedures are appropriate for the application. Although both models include an option of
choosing among a variety of sediment transport formulas, neither provide guidelines on selecting the
appropriate formula. The most recent version of GSTARS can be downloaded from the Internet at
http://www.usbr.gov/srhg/gstars/2.0/.
Sediment transport model applications

Numerous one-dimensional sediment transport models have been developed and tested in flumes and
alluvial rivers of varied slopes, bed material and grain sizes. Most often, engineering applications of
one-dimensional sediment transport models test the resultant vertical scour and fill created by a proposed
structure to assist general engineering design work and flood control implementation (Pickup, 1980; Gee,
1984; Copeland, 1986; Fischenich, 1990). Other applications of one-dimensional sediment transport
modeling studies include assessing spawning habitat (Havis et al., 1996), especially downstream from
dams (Wick, 1997), predicting degradation of the stream bed below dams (Krishnappan, 1985; Carriaga
and Mays, 1995), and evaluating the effects of dam removal (Williams, 1977; Stoker and Williams, 1991).
GSTARS has been applied to alluvial rivers by Yang and Simoes (1998), and Yang et al. (1998b), to
assess the quasi two-dimensional changes in a channel bed resulting from scour and fill. Applications of
GSTARS to engineering problems include knickpoint migration that may undermine bridge piers and
other structures, bed degradation and armoring resulting from installation of a dam, and reservoir
sedimentation (Yang and Simoes, 1998). In the first two cases, GSTARS was tested against flume data of
other researchers, and against actual field data for the reservoir sedimentation study. The agreement
between measurements and simulation was good, although the model tended to over-predict reservoir
sedimentation in some areas.
METHODS
Fieldwork was conducted to monitor sediment movement within the North Fork after the 1996 sediment
release, and to establish a known condition for calibration of the models. Subsequent analyses were
carried out to develop input files for both HEC-6 and GSTARS, calibrate the models, and verify the
calibration on another pool within the North Fork system.
Calibration data set

Channel surveys were conducted along the study reaches in October 1996, immediately following the
sediment release, in March 1997 after the experimental discharge, and again in September 1997 after the
snowmelt runoff had receded. During all surveys, channel cross sections were surveyed, and the thickness
of reservoir sediment accumulation was determined by probing with a steel rod down to the original
cobble and boulder channel bed. Pebble counts (Wolman, 1954) were conducted along the riffles within
the study reaches to define the grain size distribution of the pre-reservoir release bed material. Grab
samples were collected from deposits along the margins of pools, and from the bed of pools, to quantify
the grain size of reservoir sediment in the pools (Figure 5).
A sampling cross section was established in a riffle at cross section 1 and at cross section 17 at Study
Reaches I and II, respectively (Figure 1). From 22 February to 3 September 1997 the cross sections were
sampled weekly. During the snowmelt peak (May– June), the sampling frequency was increased to twice
Figure 5. Grain size distribution of bed material from the Tick Pool and adjacent riffle during initial sampling in October 1996
immediately following the sediment release. The median diameter of bed material from the margin of the Tick Pool (0.092) was used
to evaluate various sediment transport equations prior to computer modeling
a week. Flow depth and velocity at 0.4 and 0.6 of the flow depth were measured, and bedload and
suspended load samples were collected at four 2-m intervals along the cross section (cross section widths
varied from 12 to 15 m). These data were used to derive sediment and water rating curves, inflowing
sediment loads, bed material gradation, and a starting water surface elevation, which are required inputs
for both sediment transport models. Details of the sampling methods and field procedures at the North
Fork are presented more fully in Wohl and Cenderelli (2000).
As noted previously, channel response to the sediment release was an initial uniform veneer of sediment
within pools, the thickness of the veneer decreasing with distance downstream from the dam. During the
pre-snowmelt, stepped-experimental discharge, scour along the thalweg of each pool within the study
reaches and additional deposition on the channel margins created a narrow deep cross section, with an
accompanying lateral shift in the thalweg (Wohl and Cenderelli, 2000). After the snowmelt runoff, lateral
scour within the pools increased channel widths, resulting in an overall pool volume recovery of
approximately 80% of the pre-release volume (Wohl and Cenderelli, 2000). Sediment storage in pools and
the transfer of sediment between pools was the key control on sediment volume stored within the channel,
and the major factor in determining habitat availability for fish within the North Fork.
Sediment transport modeling

Two flow scenarios were simulated using HEC-6 and GSTARS: (1) a short-term, 1-month experimental
discharge, that represented the stepped-experimental release from Halligan Reservoir, and (2) a longer-
term, 6-month snowmelt discharge that represented runoff occurring during the spring of 1997 (Table I).
Under both flow scenarios, two end member simulations were conducted: default and robust simulations.
Default simulations were used to determine the minimum input requirements to obtain reasonable results,
and were constructed by selecting pertinent default options available in the model input. In many
applications of sediment transport modeling, field-based measurements may be limited or unavailable,
and the default values within the model would be selected by a user. During the robust simulations, all
available field data were utilized, thereby constraining the model through field quantification of input
parameters. Table II lists the input parameters pertinent to the one-dimensional models HEC-6 and
GSTARS, as applied to the North Fork.
In HEC-6, the main default values substituted for field data that are difficult, or time intensive to
obtain, include: (1) elevation of the model bottom at each cross section, or the thickness of the sediment
deposit from the reservoir, and (2) transport of cohesive silt and clay (B 0.0625 mm). Standard
simulations of HEC-6 compute only deposition of silt and clay, and if erosion of silt and clay is desired,
then shear stress thresholds for fine-grained, cohesive sediments must be prescribed. Likewise, GSTARS
requires additional input if sediment size fractions less than 0.0625 mm are transported (recall that
approximately 92% of the initial grain size distribution of pool sediment following the release was silt and
clay). Theoretical data from Ariathurai and Arulanandan (1977) were used to specify the necessary values
of shear stress for cohesive sediment transport in both models.
Initially, model calibration was conducted on the Ouzel and Tick Pools within Study Reach II (Figure
1). Model calibration is the process by which adjustments are made to coefficients and parameters used
by the model with the objective, in this case, of minimizing differences between field-measured amounts
of pool scour and fill and the model-computed values. Subsequent verification simulations were
conducted on the Goose Pool (Study Reach I; Figure 1) to substantiate the credibility of the calibration
Table I. Discharge scenarios modeled using HEC-6 and GSTARS 2.0 to evaluate pool
recovery along the North Fork
Flushing flow Magnitude Duration
Experimental discharge 3.4 m3/s (dam outlet capacity) 1 month

Snowmelt runoff 10 m3/s (snowmelt peak) 6 months
Table II. Input parameters for HEC-6 and GSTARS 2.0 for simulations along the North Fork
HEC-6 Sensitivity range tested GSTARS 2.0 Sensitivity range tested
Manning’s ‘n’ values 0.02–0.04 pools, Number cross sections Not tested, see text
0.06–0.125 riffles*
Cross section location Not tested, see text Cross section properties N/A, known values
Cross section coordinates N/A, known values Number of subchannels N/A, known values
Moveable bed limits With and without Cross section geometry N/A, known values
Sediment properties N/A, known values Roughness coefficients 0.02–0.04 pools,
0.06–0.125 riffles*
Iterations for Exner 0, 25, 50 Number of iterations 1–3 days
equation
Cohesive sediment Yes/no Number of steam tubes 3–5
transport
Sediment transport for Ackers–White, Yang, Type of discharge input N/A, known values
sand transport DuBoys, Toffaleti and
(14 available) Schoklitsch, Meyer-Peter
and Muller
Parameters for clay, silt N/A, known values Type of stage input N/A, known values
and sand transport
Discharge–sediment load N/A, known values Sediment transport Ackers–White, Yang,
relationship equation (10 available) DuBoys, Toffaleti,
Meyer-Peter and Muller
Total sediment load N/A, known values Non-equilibrium sediment Yes, no
transport
Grain size fractions of N/A, known values Sediment discharge N/A, known values
sediment load
Bed material gradation N/A, known values Water temperature Not tested, see text
Water discharges N/A, known values Number of sediment size N/A, known values
fractions
Downstream boundary N/A, known values Sediment size groups N/A, known values
condition
Water temperature Not tested, see text Sediment size distribution N/A, known values
Flow duration 1 month, 6 months Transport parameters for Varied shear stress
cohesive sediments threshold, rates of erosion
* Sensitivity analysis of Manning’s ‘n’ and contraction and expansion coefficients were evaluated using HEC-RAS (US Army Corps
of Engineers, 1998b).
N/A is not applicable.
runs. Model settings established on the Ouzel and Tick Pools were applied without modification to the
Goose Pool for verification of the calibrated results. A model is not considered fully calibrated until the
model coefficients and parameters are verified for other sites under similar conditions (Beck, 1991).
Sensitivity analyses were then performed to identify the most sensitive coefficients, and parameters
requiring the most attention in the calibration and validation process. A range of values analyzed for
sensitivity is included in Table II.
PRELIMINARY ANALYSES
Step backwater computations

Both HEC-6 and GSTARS utilize the standard-step method for solving the energy equation to derive
a water-surface profile for the modeled reach. As a means of isolating the hydraulic component of the
models, HEC-RAS, the US Army Corps of Engineers Water Surface Profile program (US Army Corps
of Engineers, 1998b), was employed to test the effects of varied Manning’s ‘n’, and contraction and
expansion coefficients. HEC-RAS estimates of the friction slope were also obtained for subsequent
sediment transport calculations.
Sediment transport capacity

Sediment transport computations were completed for the Tick Pool using six sediment transport
equations to determine the best fit between measured and calculated sediment transport rates, prior to
selecting equations to be used in subsequent simulations within HEC-6 and GSTARS. In this way,
knowledge of the agreement between predicted and measured sediment transport rates could be acquired
a priori. This knowledge is useful because the models give no indication of an over or under-prediction
of sediment transport capacities through the reach of interest unless data exist for comparison.
Three bedload equations, including DuBoys, Meyer-Peter and Muller (in Julien, 1995), and Schoklitsch
(in Bathurst et al., 1987), and three total load equations, Ackers–White (in Julien, 1995), Yang’s (1973)
sand equation, and Engelund-Hansen (in Julien, 1995) were selected for the comparison (Appendix A).
The six equations were selected based on the similarity of conditions for which the equations were
developed (i.e. slope, grain size), availability of the equation as an option in HEC-6 and GSTARS, and
ease of spreadsheet manipulation.
The median diameter (d50) of sediment moving through the Tick Pool, as collected from a depositional
site within the pool, was used to apply the bedload and total load transport formulas. Bedload refers to
the transport of sediment particles that frequently maintain contact with the bed, and bedload equations
may be used to estimate bedload transport in the absence of actual transport rates. The equations can be
applied using either the median grain size (d50) of the bed material, or the size fractions of the bed
material, where bed material refers to those size fractions that reside in the bed of the channel, and which
may or may not become entrained and transported as bedload. Because we could not physically measure
bedload from the pools, an estimate of bedload transport was obtained from the selected equations. It was
assumed that the grain size distribution of sediment within a marginal channel bar of the Tick Pool (d50
of 0.0921 mm, or very fine sand) represents what is moving in transport over the immobile bed of the pool
(bed material of the pools is in the cobble/boulder size range). Bedload was measured within the adjacent
downstream riffle (Wohl and Cenderelli, 2000), at cross section 17 (Figure 1), providing a means of
comparing the calculated and measured bedload transport rates for sediment moving through the Tick
Pool.
Calculations using the bedload equations were compared with bedload quantities collected along
sample cross section 17 (Figure 6). Figure 6 indicates that the methods of DuBoys, Meyer-Peter and
Muller, and Schoklitsch over-predict the measured bedload transport of the North Fork by as much as
three orders of magnitude for a d50 of very fine sand. The best agreement between measured and predicted
quantities of bedload for the North Fork was obtained using the Schoklitsch equation (Figure 6), with the
data converging with the 1:1 line at higher values of measured unit bedload discharge. Although the
magnitude by which the equations over-predict may seem shocking, such an over-prediction represents the
state of the art in terms of current ability to quantify bedload transport within steep gradient, bedrock
channels using the selected equations. In a thorough analysis of twelve transport equations, Gomez and
Church (1989) acknowledged that most of the formulas they tested over-predict bedload transport. Smart
(1984) showed that the Schoklitsch equation overestimated transport for slopes greater than 3%, and
Blizard (1994) found that the Schoklitsch formula over-predicted bedload discharge by three orders of
magnitude on a snowmelt-dominated subalpine stream in the Rocky Mountains of Colorado. Bathurst et
al. (1987) evaluated the applicability of several sediment transport equations to steep gradient rivers.
Bedload discharge calculations using the Schoklitsch equation showed the lowest mean error for flume
data tested, and may be most applicable for small rivers with steep slopes (\ 1%) and relatively wide
ranges of sediment size (1– 1000 mm).
Because bedload quantities measured along the North Fork were a small percentage of the total
measured load (0.68– 32%), more accurate calculations of sediment transport capacity were anticipated
from total load equations. Total load can be classified several ways (see Julien, 1995, p. 205), and in this
case, total load is distinguished by the type of sediment movement, moving either as bedload or suspended
load. By calculating total sediment transport capacity using total load equations, the large proportions of
suspended sediment released from Halligan Reservoir, that filled pools along the North Fork, would be
Figure 6. Predicted versus measured unit bedload discharge for the Tick Pool obtained from estimates using three bedload
equations. The assumption that bedload discharge at cross section 17 (downstream riffle) equals the bedload discharge of the Tick
Pool is applied
accounted for in the formulas. The released reservoir sediment consisted of size fractions that were
transported predominantly as suspended sediment, sizes that are not found in significant amounts in the
pre-release bed material.
The total measured load of Figure 7 is the sum of bedload and suspended load quantities measured in
the field. Yang’s (1973) sand equation provided the closest agreement between measured and predicted
rates of sediment discharge (Figure 7). Total load, using Yang’s equation, is over-predicted by one order
of magnitude, an improvement over the bedload equations, with Engelund-Hansen also providing slightly
greater estimates. The Ackers– White equation over-predicts sediment transport rate to the point of
hyperconcentrated flow and, therefore, the data points were omitted from Figure 7. While hyperconcen-
trated flows are possible within mountain channels, they are not very common under the snowmelt-
dominated runoff regime of the Colorado Rocky Mountains.
As a result of these preliminary computations, Yang’s equation and a combination of the Toffaleti and
Schoklitsch equations were selected for the initial model simulations in HEC-6. Unfortunately, the
Schoklitsch equation was not an input option common to both HEC-6 and GSTARS, and could not be
evaluated via both models. The coupled Toffaleti–Schoklitsch was selected as an appropriate substitute.
The only equation deemed applicable to the North Fork, based on the preliminary sediment transport
calculations, and common to the two models, was Yang’s (1973) equation.
Scour and fill

Modeled results in HEC-6 estimate average change in bed elevation over the duration of the
hydrograph that was input into the model. As a one-dimensional model, HEC-6 calculates uniform bed
change reported as average elevation change of the bed in a cross section. For comparison with the
measured bed change in the pools obtained from field surveys, an estimate of the net bed change was
obtained from a program called ‘Scour and Fill’ (US Geological Survey, 1999), as was the maximum
amount of bed change within the thalweg. The ‘Scour and Fill’ program overlays two cross sections,
Figure 7. Comparison of measured and calculated sediment transport rates within the Tick Pool. The total measured load plot
represents the sum of bedload and suspended load measured in the field. Total measured load was compared with estimates from
various sediment transport equations to determine the most appropriate equation to select as input for the sediment transport
models
shows the area of scour and fill, and calculates the net areal change. This change in area was divided by
the width of the channel to obtain the net vertical change along the cross section, or the average bed
change, a quantity we considered to be more comparable with the average bed change estimated by
HEC-6.
HEC-6 RESULTS
Experimental discharge
Default simulations. All available HEC-6 default parameters for cross sections through the pools of the
modeled reach were selected. Moveable bed limits, or the elevation of the model bottom below which no
scour or deposition could occur, were set for the riffles. Only deposition of cohesive sediments is
accounted for in the default simulations, and the sediment transport equations used were Yang’s and a
Toffaleti–Schoklitsch combination.
During these default simulations, poor agreement was obtained between observed and modeled bed
changes (Table III). Columns 5 and 6 in Table III give estimates of the percentage of the measured bed
change that is predicted by the model, calculated by two methods, dividing column 4 by 2, and 4 by 3,
respectively. The model accuracy for two of the three cross sections through the Ouzel and Tick Pools
varied from 0 to 38% for Yang’s equation, and 0 to 49% for the Toffaleti– Schoklitsch equation for the
experimental discharge simulations with default settings (Table III). In other words, HEC-6 predicted
between 0 and 49% of the vertical bed change that was measured in the field within two pools, using two
different sediment transport equations. Modeling accuracy results using the maximum thalweg change
(column 5, Table III) are greater than those using the net bed change from ‘Scour and Fill’ (column 6,
Table III). The results of the average bed change calculated by ‘Scour and Fill’ consistently show
Table III. HEC-6 results for experimental discharge simulations (Qpeak = 3.4 m3/s) for default and robust data input within three pools along the North Fork
Default simulation Robust simulation
(1) Cross (2) October– (3) Average (4) Average (5) Model (6) Model (7) Cross (8) October– (9) Average (10) Average (11) Model (12) Model
section March ‘Scour and HEC-6 bed accuracy accuracy section March ‘Scour and HEC-6 bed accuracy accuracy
thalweg Fill’ bed change (%) (%) thalweg Fill’ bed change (%) (%)
change change (m) (4)/(2) (4)/(3) change change (m) (10)/(8) (10)/(9)
(m) (m) (m) (m)
Copyright © 2001 John Wiley & Sons, Ltd.

Yang’s Tick Pool 12 −0.63 0.05 −0.03 4.37 −57.32 12 −0.63 0.05 −0.50 80.10 −1050.96
13 −0.22 0.24 −0.04 16.44 −15.23 13 −0.22 0.24 −0.17 76.71 −71.07
14 0.50 0.15 0.00 −0.61 −2.09 14 0.50 0.15 −0.23 −46.01 −156.90
Ouzel Pool 2 1.20 0.36 0.45 37.66 124.37 2 1.20 0.36 0.03 7.80 9.24
3 0.49 0.02 0.04 8.70 254.55 3 0.49 0.02 −0.02 −4.97 −145.45
4 0.75 0.15 0.00 0.00 0.00 4 0.75 0.15 −0.09 −11.38 −57.73
Goose Pool 4 −0.45 −0.22 −0.35 76.16 156.89 4 −0.45 −0.22 −0.42 91.39 188.27
5 −0.57 −0.16 −0.24 42.25 148.78 5 −0.57 −0.16 −0.74 129.95 457.63
6 −0.74 −0.18 −0.16 21.40 86.24 6 −0.74 −0.18 2.69 362.96 1462.69
Toffaleti– Tick Pool −0.63 0.05 −0.03 4.85 −63.69 12 −0.63 0.05 −0.50 80.10 −1050.96
Schoklitsch
13 −0.22 0.24 −0.02 6.85 −6.35 13 −0.22 0.24 −0.17 76.71 −71.07
14 0.50 0.15 −0.02 −4.29 −14.64 14 0.50 0.15 −0.23 −46.01 −156.90
Ouzel Pool 2 1.20 0.36 0.58 48.60 160.50 2 1.20 0.36 0.17 14.50 47.90
3 0.49 0.02 0.14 27.95 810.18 3 0.49 0.02 0.00 0.62 18.18
4 0.75 0.15 00.00 0.00 0.00 4 0.75 0.15 −0.09 −11.38 −57.73
Goose Pool 4 −0.46 −0.22 −0.31 67.55 159.15 4 −0.46 −0.22 −0.49 106.62 219.85
5 −0.57 −0.16 −0.05 9.63 33.90 5 −0.57 −0.16 −0.50 88.24 310.73
6 −0.74 −0.18 0.00 0.00 0.00 6 −0.74 −0.18 −2.74 370.37 1492.54
Columns 5 and 6 represent model accuracy from HEC-6 using maximum thalweg change and average ‘Scour and Fill’ bed change, respectively. Negative values indicate a model predictor
of scour when aggradation occurs, or 6ice 6ersa. Percentages greater than 100 indicate over-prediction by HEC-6.
ONE-DIMENSIONAL SEDIMENT TRANSPORT MODELING OF POOL RECOVERY
Regul. Ri6ers: Res. Mgmt. 17: 251 – 273 (2001)

263
over- and under-predictions of bed change, hence, the negative values and values greater than 100% in
Table III. Because of this discrepancy, subsequent discussions of model accuracy refer to estimates that
incorporate the actual thalweg change (columns 5 and 11, Table III).
Robust simulations. Increased data input for robust simulations includes allowing for both deposition
and erosion of cohesive sediments, and setting the elevation of the model bottom to the depth of sediment
accumulated above the bed in the pools. In essence, the robust simulations utilized the entire suite of field
data that was collected over the course of 1 year.
Results of the robust simulations under the experimental discharge hydrograph show a noted improve-
ment over the default simulations. The percentage of bed change predicted by HEC-6 that was measured
in the field increased to between 77 and 80% for two of the three cross sections in the Tick Pool (Table
III). At cross section 14, scour was predicted by HEC-6 for both simulations over the experimental
discharge when 0.5 m of aggradation actually occurred between October 1996 and March 1997. Cross
section 14 is located at the exit slope of the Tick Pool, where a combination of the reverse gradient and
the low magnitude experimental discharge enhanced sediment deposition. Results of robust simulations
within the Ouzel Pool showed no improvement over the default settings.
Snowmelt runoff
Default simulations. Snowmelt runoff simulations for the 6-month hydrograph, using default parame-
ters, provided a more consistent match between the observed and predicted bed changes for the Tick Pool.
Between 23 and 75% of the observed bed change in the Tick Pool was modeled by HEC-6 for both Yang’s
and the Toffaletti– Schoklitsch equations (Table IV). Predictions of bed change for the Ouzel Pool again
were in poor agreement with field measurements.
Robust simulations. When the full set of field data is used, HEC-6 was able to predict 53–100% of the
actual bed change measured in both pools for the robust simulations over the longer, 6-month time period
(Table IV). This increased accuracy between observed and predicted values of bed change in the Tick and
Ouzel Pools for both sediment transport equations was derived by increasing the length of the simulated
flow duration and the discharge magnitude. These snowmelt simulations offer a vast improvement over
the results for the default runs.
Using all available field data, the model became insensitive to sediment transport formula selection, and
the resultant scour and deposition were nearly identical for the two equations selected (Table IV, column
11 for both Yang’s and the combined Toffaleti–Schoklitsch). These findings were similar for other
equations tested during the sensitivity analysis. Apparently, the HEC-6 model was sufficiently constrained
by the known field data that the variations between the transport formulas produced negligible bed
changes. At this point, we considered the model robustly calibrated.
Our interpretation of the modeling results is based on an expectation of accuracy that is relative to the
ability of commercially available numerical models to represent sediment transport within steep gradient,
immobile boundary channels; a greater than 50% agreement between measured and HEC-6 modeled bed
change is reasonable, given the stochastic nature of the processes being modeled along the North Fork.
Water resource managers need to make their own situation-dependent decisions regarding the reasonable-
ness of their model output, as guided by the management decisions of interest. For example, the HEC-6
modeling results can be viewed in terms of pool depth recovery to better understand and predict fish
re-establishment. For a given climate, water managers could use scenarios of fish populations (species,
size, number of individuals) along a river reach to estimate minimum pool volume needed for overwinter
habitat, setting guidelines for acceptable modeling results that fit with the objectives and uncertainties of
the modeling effort. In addition, the models can be applied to a series of flows and range of time periods
to test hypothetical situations that will answer questions regarding residence times of pool sediment.
Model 6alidation
Validation of the calibrated model settings in HEC-6 on the Goose Pool (Figure 1) yielded an accuracy
of 21 – 76% and 0–68% of the measured bed change for the default simulation of the experimental
Table IV. HEC-6 results for snowmelt runoff discharge simulations (Qpeak = 10.1 m3/s) for default and robust data input within three pools along the North
Fork
Default simulation Robust simulation
(1) Cross (2) October– (3) Average (4) Average (5) Model (6) Model (7) Cross (8) October– (9) Average (10) Average (11) Model (12) Model
section August ‘Scour and HEC-6 bed accuracy accuracy section August ‘Scour and HEC-6 bed accuracy accuracy
thalweg Fill’ bed change (%) (%) thalweg Fill’ bed change (m) (%) (%)
change change (m) (4)/(2) (4)/(3) change change (10)/(8) (10)/(9)
(m) (m) (m) (m)
Yang’s Tick Pool 12 −0.63 −0.12 −0.47 75.24 388.47 12 −0.63 −0.12 −0.63 100.00 516.29

13 −0.22 −0.17 −0.15 67.61 87.11 13 −0.22 −0.17 −0.17 78.87 101.63
14 −0.77 −0.20 −0.18 23.41 89.12 14 −0.77 −0.20 −0.41 53.17 202.42
Average =77.35
Ouzel Pool 2 –0.61 −0.11 0.04 −6.03 −32.43 2 −0.61 −0.11 −0.44 72.86 39.89
3 −0.80 −0.06 −0.01 1.53 20.30 3 −0.80 −0.66 −0.32 40.61 538.07
4 −0.30 0.02 −0.01 4.08 −66.67 4 −0.30 0.02 −0.16 55.10 −900.00
Average =56.19
Goose Pool 4 −0.50 N/A −0.20 40.24 N/A 4 −0.56 N/A −0.33 66.46 N/A
5 −1.12 −0.84 −0.22 19.62 26.00 5 −1.12 −0.84 −0.33 29.16 38.64
6 −2.64 −1.25 −0.04 1.62 3.41 6 −2.63 −1.25 −2.07 78.68 655.21
Average = 58.10
Toffaleti– Tick Pool 12 −0.63 −0.12 −0.47 74.76 385.96 12 −0.63 −0.12 −0.63 100.00 316.29
Schoklitsch 13 −0.22 −0.17 −0.15 67.61 87.11 13 −0.22 −0.17 −0.17 78.87 101.63
14 −0.77 −0.20 −0.20 25.40 96.68 14 −0.77 −0.20 −0.41 53.17 202.42
Average =77.35
Ouzel Pool 2 −0.61 −0.11 0.09 −15.08 −81.08 2 −0.61 −0.11 −0.20 32.66 175.68
3 −0.80 −0.06 −0.01 1.15 15.23 3 −0.00 −0.06 −0.21 26.05 345.18
4 −0.30 0.02 −0.02 5.10 −83.33 4 −0.30 0.02 −0.16 53.10 −900.00
Average =37.94
Goose Pool 4 −0.50 N/A −0.17 34.15 N/A 4 −0.50 N/A −0.48 95.23 N/A
5 −1.12 −0.84 −0.04 3.27 4.33 5 −1.12 −0.84 −0.57 51.23 67.89
6 −2.63 −1.25 0.05 −1.74 −3.65 6 −2.63 −1.25 −1.97 74.74 156.93
Average =73.90
ONE-DIMENSIONAL SEDIMENT TRANSPORT MODELING OF POOL RECOVERY
Columns 5 and 6 represent model accuracy from HEC-6 using maximum thalweg change and average ‘ Scour and Fill’ bed change. Respective negative values indicate a model predictor
of scour when aggradation occurs, or 6ice 6ersa. Percentages greater than 100 indicate over-prediction by HEC-6.
N/A is not applicable.

265
discharge using Yang’s and Toffaleti– Schoklitsch equations, respectively (Table III). Robust simulations
of the experimental discharge over-predicted the actual bed change within two of the three pool cross
sections.
The default and robust simulations for the snowmelt hydrograph generated a more consistent match
between field measurements and model predictions, with the strongest match for the robust simulations
(Table IV). Between 58 and 74%, on average, of the actual field scour or deposition was predicted by the
model, depending upon the sediment transport equation chosen. If HEC-6 calibrated model settings were
applied to yet another pool along the North Fork, at least 58% of the observed vertical bed change would
be predicted by the model, provided the requisite data for a robust simulation over the snowmelt runoff
were available. Pool depth recovery of 58% may be sufficient sediment removal to support fish over a
winter, and sediment transport modeling would then provide useful information to answer questions of
channel recovery. Depending on the target recovery depth or volume for the pools of interest, the results
of the HEC-6 simulations could be evaluated accordingly. However, the results and verification would be
obtained only after a year of data collection, the time involved in the case of the North Fork to gather
enough field data to calibrate and validate the models.
Sensiti6ity analysis
Sensitivity analyses were conducted on Manning’s ‘n’ value, and expansion and contraction coefficients
to assess the effect on water surface profiles calculated using HEC-RAS (US Army Corps of Engineers,
1998b). Less than 9 cm change in the elevation of the water-surface profile resulted from increasing or
decreasing ‘n’ or expansion and contraction coefficients by 9 10–50% beyond beginning values. Contrac-
tion and expansion values were taken from recommendations in the HEC-6 user’s manual as 0.1 and 0.3,
respectively (US Army Corps of Engineers, 1998b). A 9 cm fluctuation in water-surface profile is
acceptable for the North Fork, because we actually back-calculated Manning’s ‘n’ for various discharges,
and know a reasonable range of ‘n’ for the two flow discharges modeled, such that the 9 cm fluctuation
is a maximum.
Sensitivity analyses by O’Connor and Webb (1988) for paleofloods have shown that the location and
density of cross sections are the most important requirements for accurate water-surface profiles. Cross
sections should be spaced to adequately represent any abrupt geometry changes within the channel, with
closer spacing around bends and through expanded and contracted areas, and where bed slope changes.
We consider our choice of cross sections adequate for the configuration of the North Fork (Figure 1), and
sensitivity analyses were not completed on cross section number and placement. Likewise, water
temperature was not tested for sensitivity. Water temperature was set at 5°C, a temperature we know to
be reasonable for snowmelt runoff-dominated rivers during the spring and summer months.
Three additional sediment transport equations were assessed in a sensitivity analysis within HEC-6
(Table II) to evaluate the effects of different formulas for which preliminary computations were not
completed. We found that the modeling results were tremendously sensitive to selection of formula,
producing changes in bed elevation that varied by as much as a 400% over-prediction of field-quantified
bed changes. Selection of Ackers– White, DuBoys and Meyer-Peter and Muller sediment transport
equations within the HEC-6 input consistently over-predicted the measured bed elevation changes within
the North Fork study area.
GSTARS 2.0 RESULTS
Experimental discharge
Minimal default options are available for the user in GSTARS. As such, no default simulations
comparable to the HEC-6 results were conducted.
Robust simulations. Results of the GSTARS modeling for the Tick Pool indicate that between 13 and
90% of the measured bed scour at two of the three cross sections was predicted by GSTARS (Table V).
Similar to the HEC-6 results, bed scour, rather than aggradation, was predicted by GSTARS for cross
section 14. Model accuracy for the Ouzel Pool was lower, with a 1–14% match between model predictions
and field measurements of bed change over the three pool cross sections. This is consistent with the lower
accuracy findings using HEC-6. The geometry of the Ouzel Pool, with its large eddy pool of near-stagnant
flow, sets up more complex hydraulics that are outside the range of even a semi-two-dimensional flow
model.
Snowmelt runoff
Robust simulations. By increasing the length of flushing discharge from 1 month to 6 months to include
the snowmelt runoff hydrograph, GSTARS predictions of bed scour and/or aggradation were between 13
and 92% of actual field measurements for the Tick Pool. The model appears to be somewhat insensitive
to the increased time period over which a simulation is run because the GSTARS predictions of bed
change are nearly identical (columns 5 and 11, Table V) for two of the three cross sections.
Although other sediment transport equations were selected in subsequent simulations of GSTARS 2.0,
the resultant differences in bed elevation were minimal (B 5%), so only results using Yang’s (1973)
equation are presented in Table V.
DISCUSSION
Sediment transport equations appropriate to steep gradient rivers have not yet been developed and, by
necessity, relationships designed for lowland rivers are typically employed. Sediment transport equations
developed for alluvial rivers have limited applicability to the North Fork because of the effects of bed
armoring, spatially limited sediment supply, and the bimodal distribution of sediment sizes comprising the
bed compared with the sediment sizes in transport. Our preliminary analyses to determine the appropriate
sediment transport formula(s) for the steep gradient, bedrock-controlled North Fork indicate that the
total load formulas gave more satisfactory results than the bedload formulas tested. Our success with total
load equations is consistent with the flume experiment of Tingsanchali and Supharatid (1996), in which
the performance of HEC-6 was investigated. These investigators found that the Toffaleti and Yang total
load equations gave the most satisfactory prediction of actual bed profiles within the flume under various
conditions of flow and sediment transport.
Once the appropriate equation was identified, further analysis indicated that the modeling results of
HEC-6 and GSTARS are limited in ways specific to the conditions along the North Fork. The major
limitation of HEC-6 is that it is a purely one-dimensional model, which leads, by necessity, to
simplification of the complex, three-dimensional flow in rivers. HEC-6 cannot account for the following
hydraulic conditions present in pools along the North Fork that include: (1) lateral flow separation, or
eddies, where a majority of sediment was deposited during the sediment release from Halligan Reservoir,
(2) bank erosion or lateral channel migration, which occurred as a thalweg was re-excavated through the
pools after the release, (3) differential scour and deposition over the width of a cross section, which
occurred as sediment was simultaneously eroded out of the central, high velocity thalweg and deposited
along the channel margins in the pools, and (4) temporal changes in sediment supply and bed material
grain size distribution. In spite of these limitations, HEC-6 is a likely model to be selected by water
resource managers faced with assessing and mitigating the effects of a reservoir sediment release. As a
result of our testing the applicability of HEC-6 on a high gradient, bedrock fluvial system, we determined
that it can be used in such systems to answer questions about post-sediment release channel recovery.
A major limitation of GSTARS is the model’s insensitivity to long-term simulations and varied
sediment transport equations. The present configuration of the GSTARS code does not accommodate
large differences in grain sizes of bed sediment over short distances, such as between riffles comprised of
cobbles and boulders, and adjacent pools of silt and fine sand (F. Simoes, personal communication, 1999).
268
Table V. GSTARS 2.0 results for both the experimental and snowmelt discharge simulations using the robust data input within two pools along the North

Fork
Robust simulation, experimental discharge Robust simulation, snowmelt
(1) Cross (2) October– (3) Average (4) (5) Model (6) Model (7) Cross (8) October– (9) Average (10) (11) Model (12) Model
section March ‘Scour and GSTARS accuracy accuracy section August ‘Scour and GSTARS accuracy accuracy
thalweg Fill’ bed bed change (%) (%) thalweg Fill’ bed bed change (%) (%)
change change (m) (4)/(2) (4)/(3) change change (m) (10)/(8) (10)/(9)
(m) (m) (m) (m)
Yang’s Tick Pool 12 −0.63 0.05 −0.08 13.50 −177.07 12 −0.63 0.05 −0.08 13.50 −177.07
13 −0.22 0.24 −0.20 89.59 −82.99 13 −0.22 0.24 −0.20 92.11 −82.99
14 −0.50 0.15 −0.39 −77.67 −264.85 14 −0.77 0.15 −0.39 50.24 −264.85
Ouzel Pool 2 1.20 0.36 0.07 6.11 20.17 2 −0.61 0.36 0.07 −12.06 20.17
3 0.49 0.02 0.07 14.47 423.64 3 −0.80 0.02 0.07 8.93 423.64
S.L. RATHBURN AND E.E. WOHL
4 0.75 0.15 0.01 1.42 7.22 4 −0.30 0.15 0.01 −3.57 7.22
Results are derived using Yang’s (1973) sand transport equation.

The Bureau of Reclamation developed GSTARS for internal use, rather than to suit the needs of the
general public (F. Simoes, personal communication, 1999). As such, GSTARS assumes a much higher
level of user knowledge in hydraulics and sediment transport, and incorporates fewer default options. The
capabilities of GSTARS were severely underutilized in the North Fork application, and the power of the
model is best expressed in situations of sand transport where stream power minimization is important.
CONCLUSIONS
Simulations using HEC-6 produced reasonable, first-order approximations of the pool bed changes
resulting from scour and fill after the sediment release from Halligan Reservoir. Default simulations
produced minimal accuracy for all three pools modeled and, therefore, limit the management capabilities
of HEC-6 to situations with adequate field data. Average pool-wide trends of predicted versus observed
aggradation and degradation for three pools along the North Fork were greater than 50% for the
long-term, snowmelt runoff hydrograph using HEC-6 and Yang’s (1973) sand transport equation.
Average modeling accuracy was lower for the three pools when the combined Toffaleti–Schoklitsch
equation was used, ranging from 38 to 77% for the Tick and Ouzel Pools. All of these results, however,
were generated only after using an extensive, field-based data set collected during weekly or biweekly field
outings over the course of a 1-year period.
The calibration results were verified on the Goose Pool, with generally poor agreement for the
experimental discharge, but at least 58%, and up to 74% for the robust modeling, depending on the
sediment transport equation selected. In both calibrated and verification simulations, the longer-term
snowmelt runoff scenarios produced the closest agreement between modeled and measured bed change.
The HEC-6 User’s Manual (US Army Corps of Engineers, 1998a) recommends long-term simulations
over single event hydrographs, because of the bed instabilities that arise. A 1-month simulation, such as
the experimental discharge, was apparently insufficient to stabilize the bed elevation computations.
Overall, modeling results for the Ouzel Pool are in poor agreement with field measurements. The Ouzel
Pool has a large eddy (on channel right) of nearly stagnant flow. HEC-6 cannot account for the upstream
component of flow, and treats the entire cross section as the effective width. The increased width, for the
same depth, reduces velocity, and lower velocity translates into lower shear stress and, hence, lower
sediment transport at the bed. One option of handling the eddy was to decrease the effective pool width
in HEC-6 and obtain greater amounts of scour. We did not alter pool width in this study. A
two-dimensional finite element model would more effectively cope with transport through the eddy pools;
however, the increased expertise required to use multidimensional models is substantial, and not realistic
for most individuals without prior training in hydraulics and sediment transport.
The stepped, experimental discharge released from Halligan Reservoir produced improved pool volume
for fish habitat, much greater than was predicted by the models. It could be that after 5 months of 0.1
m3/s flow, the initial 3.4 m3/s release of water along the North Fork, over a bed of silt and fine sand in
the pools, entrained sediment by mechanisms that are not reproduced by one and quasi-two-dimensional
sediment transport models. We did not run the model for 5 months at 0.1 m3/s, but rather started the
model at the beginning of the experimental release in February. The field data attest to the fact that
experimental releases are effective at entraining and transporting sediment, as pulsed flow increases shear
stresses beyond that of gradually increased flow (Wohl and Cenderelli, 2000).
Model calibration on two pools and validation on one pool of the North Fork indicate that at least
58% of observed bed changes after a sediment release would be predicted by HEC-6. For model
applications where predictions of pool recovery for fish habitat are concerned, such as the North Fork,
58% accuracy renders the model a useful predictive tool to answer many management questions. In order
to use HEC-6, adequate data and extensive calibration are required. Two of the input parameters that
must be defined for each river system are thickness of sediment accumulated above the bed, in order to
specify the elevation of the model bottom, and allowance for deposition and erosion of cohesive
sediments, if appropriate. Also, the selection of sediment transport formula has the most substantial effect
on the computed results, and knowledge of the applicability of various equations to the system of interest
is critical. One-dimensional sediment transport models can only be useful long-term predictors of channel
bed elevation changes if adequate data are collected and calibration is performed.
Realistic evaluation of the time investment is necessary to determine the suitability of modeling channel
recovery scenarios. Once the field data are collected, a first time user of HEC-6 should expect to devote
substantially more time than someone with prior modeling experience and knowledge of hydraulic and
sediment transport processes. Although much of the input for GSTARS can be derived from HEC-6, with
minor format changes, the time investment using GSTARS is considerably longer given the greater
number of built-in options which give the modeler as much control as possible. Current releases of
GSTARS 2.0 are not constructed to reflect physical processes similar to those operating along the North
Fork and, therefore, GSTARS is not well adapted to steep gradient bedrock channels filled with
fine-grained material from a reservoir release.
For water resource managers faced with a newly-filled channel reach resulting from a reservoir sediment
release, depending on specific recovery needs, HEC-6 simulations could be conducted to established
flushing flows. In the case of the North Fork, more than 50% pool recovery was achieved at a discharge
of 10.1 m3/s. Simulations could be run, for example, until the desired pool recovery was achieved, and the
associated flow could be requested as the minimum flushing discharge for sediment mobilization and
transport. One constraint that must be recognized is the heavy data dependence of HEC-6, and that
without good field control the model predictions are subject to extreme error. However, computer models
are useful tools, and in situations with restrictions on available flow for transporting sediment associated
with a reservoir release, one-dimensional sediment transport modeling is still preferable to uncalibrated
estimates of flushing discharges.
ACKNOWLEDGEMENTS
Grateful appreciation is extended to Trout Unlimited, Colorado Water Conservation Board, US Bureau
of Reclamation, and The Nature Conservancy for initial funding and support to begin this project. Major
funding was provided by NSF Grant CMS-9727061. Daniel Cenderelli and numerous other people helped
with the data collection for this research, and we heartily thank them. Francisco Simoes, US Bureau of
Reclamation, Hydraulics Group, kindly assisted with the GSTARS 2.0 modeling. The manuscript was
substantially improved by the thoroughness of two anonymous referees, and Jim Finley provided
insightful comments on an earlier draft.
APPENDIX A. BEDLOAD EQUATIONS

(1) DuBoys Formula:
0.173
qb6 = ~o (~o −0.0125 −0.019ds )
d 3/4
s
(2) Meyer-Peter and Muller:

qb6 = 8(~ −~ c )3/2
(G − 1)gd 3s
(3) Schoklitsch:
2.5 3/2
qb6 = S (q − qc )
zs /z
where
qc = 0.26

zs
−1
5/3
d 3/2
40
z S 7/6
Total Load Equations:

(4) Ackers and White:
CW =cAW2 G
ds V cAW 1
cAW5
−1
n
cAW 4

h u cAW3
cAW5 =
u cAW 1 V n
1 − cAW 1

(G −1)gds
32 log(10h/ds )
for 1.0Bd B 60.0 (d = 6.8, in this case)
cAW1 + 1.0− 0.56 log d
log cAW2 =2.8c6 log d −(log d )2 −3.53
0.23
cAW3 = + 1.34
d 1/2
9.66
cAW4 + + 1.34
d
(5) Yang (1973) for sand:
ds
−0.457 log
u
log Cppm = 5.435 −0.286 log
w

+ 1.799 −0.409 log
ds
−0.314 log x log
u VS VcS
−

w
+0.66 for 1.2B s B 70

Vc 2.5 u d
=
log(u ds /w) −0.06 w

(6) Engelund and Hansen:
Cw = 0.05
G VSf
1/2
RhSf n1/2
G − 1 [(G −1)gds ] (G − 1)ds
NOTATION
Cppm Total sediment concentration by weight

ds Particle size, d50 of the bed material used, unless otherwise specified
d Dimensionless particle diameter
g Gravitational acceleration
G Specific gravity
h Flow depth
q Water discharge per unit width of flow
qc Critical value of q for initiation of sediment transport
qb6 Unit bedload discharge measured by volume (converted to mass for plotting on Figure 6)
Rh Hydraulic radius
S Channel slope
Sf Friction slope
u Shear velocity
V Depth-averaged flow velocity
Vc Average flow velocity at incipient motion
Vc / Dimensionless critical velocity at incipient motion
VS Unit stream power
w Kinematic viscosity
z Density of water
zs Density of sediment
Fall velocity
~o Boundary shear stress
~ Dimensionless shear stress (Shields parameter)
~ c Critical bed shear stress (critical values of Shields parameter)
REFERENCES
American Rivers. 1999. Dam removal success stories: restoring rivers through selective removal of dams that don’t make sense, http:
//www.amriver.org/successstories.html [accessed March 2000].
Ariathurai R, Arulanandan K. 1977. Erosion rates of cohesive soils. Journal of Hydraulic Engineering 104: 279 – 283.
Bathurst JC, Graf WH, Cao HH. 1987. Bed load discharge equations for steep mountain rivers. In Sediment Transport in Gra6el-bed
Ri6ers, Thorne CR, Bathurst JC, Hey RD (eds). John Wiley and Sons: Chichester; 453 – 491.
Beck MB. 1991. Principles of modelling. Water Science and Technology 24(6): 1 – 8.
Blizard CR. 1994. Hydraulic variables and bedload transport in East St. Louis Creek, Rocky Mountains, Colorado, Unpublished
MS Thesis. Colorado State University, Fort Collins, CO.
Carriaga CC, Mays LW. 1995. Optimal control approach for sedimentation control in alluvial rivers. Journal of Water Resources
Planning and Management 121(6): 408 –417.
Collier MP, Webb RH, Schmidt JC. 1996. Dams and rivers: a primer on the downstream effects of dams. US Geological Survey,
Circular 1126.
Copeland RR. 1986. San Lorenzo River sedimentation study: numerical model investigation. Army Corps of Engineers, Waterways
Experiment Station Technical Report HL-86-10.
Fan S, Springer FE. 1990. Major sedimentation issues and ongoing investigations at the Federal Energy Regulatory Commission.
In Hydraulic Engineering, Volume 2: Proceedings of the 1990 National Conference, San Diego, CA, Chang HH, Hill JC (eds).
American Society of Civil Engineers: New York; 1014 – 1020.
Fischenich JC. 1990. Cumulative impacts analysis on a midwest fluvial system. In Hydraulic Engineering: Proceedings of the 1990
National Conference. American Society of Civil Engineers: New York; 802 – 807.
Gee DM. 1984. Predictions of the effects of a flood control project on a meandering stream. Hydrologic Engineering Center
Technical Paper 97, US Army Corps of Engineers, Davis CA.
Gomez B, Church M. 1989. An assessment of bed load sediment transport formulae for gravel bed rivers. Water Resources Research
25(6): 1161 – 1186.
Gray LJ, Ward JV. 1982. Effects of sediment releases from a reservoir on stream macroinvertebrates. Hydrobiologia 96(2): 177 – 184.
Havis RN, Alonso CV, King JG. 1996. Modeling sediment in gravel-bedded streams using HEC-6. Journal of Hydraulic Engineering
122(10): 559 – 564.
Julien PY. 1995. Erosion and Sedimentation. Cambridge University Press: New York, NY.
Kondolf GM, Matthews WVG. 1993. Management of coarse sediment on regulated rivers. Report 80, Water Resources Center,
University of CA, Davis.
Krishnappan BG. 1985. Comparison of MOBED and HEC-6 river flow models. Canadian Journal of Ci6il Engineering 12: 464 – 471.
O’Connor JE, Webb RH. 1988. Hydraulic modeling for paleoflood analysis. In Flood Geomorphology, Baker VR, Kochel RC,
Patton PC (eds). John Wiley and Sons, Inc.: New York, NY; 393 – 402.
Pickup G. 1980. Hydrologic and sediment modelling studies in the environmental impact assessment of a major tropical dam project.
Earth Surface Processes 5: 61 –75.
Schumm SA. 1965. Quaternary paleohydrology. In The Quaternary of the United States, Wright Jr HE, Frey DG (eds). Princeton
University Press: Princeton; 783 –794.
Smart GM. 1984. Sediment transport formula for steep channels. Journal of Hydraulic Engineering 110: 267 – 276.
Stoker BA, Williams DT. 1991. Sediment modeling of dam removal alternatives, Elwha River, Washington. In Hydraulic
Engineering: Proceedings of 1991 National Conference, Shane RM (ed.). American Society of Civil Engineers: Nashville, TN;
674 –679.
Tingsanchali T, Supharatid S. 1996. Experimental investigation and analysis of HEC-6 river morphological model. Hydrological
Processes 10: 747 – 761.
US Army Corps of Engineers, Hydrologic Engineering Center. 1998a. HEC-6 Scour and Deposition in Ri6ers and Reser6oirs User’s
Manual, 6ersion 4.1.
US Army Corps of Engineers, Hydrologic Engineering Center. 1998b. HEC-RAS Ri6er Analysis System User’s Manual, 6ersion 2.2.
US Geological Survey. 1999. Scour and Fill for Windows, 6ersion 2.0. Redwood National Park: Arcata, CA.
Wick EJ. 1997. Physical processes and habitat critical to the endangered razorback sucker on the Green River, Utah, Unpublished
Doctoral Dissertation, Colorado State University, Fort Collins, CO.
Wilcock PR, Kondolf GM, Matthews WVG, Barta AF. 1996. Specification of sediment maintenance flows for a large gravel-bed
river. Water Resources Research 32(9): 2911 – 2921.
Williams DT. 1977. Effects of dam removal: an approach to sedimentation. Hydrologic Engineering Center Technical Paper 50, US
Army Corps of Engineers, Davis, CA.
Williams GP, Wolman MG. 1984. Downstream effect of dams on alluvial rivers. US Geological Survey Professional Paper 1286,
Government Printing Office, Washington, DC.
Wohl EE. 1999. Inheriting our past: river sediment sources and sediment hazards in Colorado. In Colorado Water Resources
Research Institute ‘Water in the Balance’, no. 7. Colorado Water Resources Research Institute: Fort Collins, CO.
Wohl EE, Cenderelli D. 2000. Sediment deposition and transport patterns following a reservoir sediment release. Water Resources
Research 36(1): 319 –333.
Wolman MG. 1954. A method of sampling coarse river-bed material. EOS Transactions 35: 951 – 956. American Geophysical Union
no. 6.
Yang CT. 1973. Incipient motion and sediment transport. Journal of Hydraulics Di6ision 99(HY10): 1679 – 1703.
Yang CT, Simoes FJ. 1998. Simulation and prediction of river morphologic changes using GSTARS 2.0. In Proceedings of Third
International Conference on Hydro-Science and -Engineering, Cottbus/Berlin, Germany, 31 August 31 – 3 September.
Yang CT, Trevino MA, Simoes FJ. 1998a. User’s Manual for Generalized Stream Tube model for Allu6ial Ri6er Simulation 6ersion
2.0 (GSTARS 2.0). US Bureau of Reclamation Technical Service Group, Sedimentation and River Hydraulics Group: Denver,
CO.
Yang CT, Trevino MA, Simoes FJ. 1998b. An enhanced generalized stream tube model for alluvial river simulation. In Proceedings
of First Federal Interagency Hydrologic Modeling Conference, Las Vegas, NV, 19 – 23 April; 8.143 – 8.150.

One-Dimensional Sediment Transport Modeling of Pool Recovery Along A Mountain Channel After A Reservoir Sediment Release

Uploaded by

Copyright:

Available Formats

One-Dimensional Sediment Transport Modeling of Pool Recovery Along A Mountain Channel After A Reservoir Sediment Release

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

One-Dimensional Sediment Transport Modeling of Pool Recovery Along A Mountain Channel After A Reservoir Sediment Release

Uploaded by

Copyright:

Available Formats

REGULATED RIVERS: RESEARCH & MANAGEMENT

Regul. Ri6ers: Res. Mgmt. 17: 251– 273 (2001)

ONE-DIMENSIONAL SEDIMENT TRANSPORT MODELING OF POOL

Recei6ed 4 April 2000

Sediment transport model applications

Calibration data set

Sediment transport modeling

Flushing flow Magnitude Duration

Experimental discharge 3.4 m3/s (dam outlet capacity) 1 month

HEC-6 Sensitivity range tested GSTARS 2.0 Sensitivity range tested

Step backwater computations

Sediment transport capacity

Scour and fill

Default simulation Robust simulation

Copyright © 2001 John Wiley & Sons, Ltd.

Regul. Ri6ers: Res. Mgmt. 17: 251 – 273 (2001)

Default simulation Robust simulation

Copyright © 2001 John Wiley & Sons, Ltd.

Regul. Ri6ers: Res. Mgmt. 17: 251 – 273 (2001)

GSTARS 2.0 RESULTS

Copyright © 2001 John Wiley & Sons, Ltd.

Robust simulation, experimental discharge Robust simulation, snowmelt

Results are derived using Yang’s (1973) sand transport equation.

Regul. Ri6ers: Res. Mgmt. 17: 251 – 273 (2001)

APPENDIX A. BEDLOAD EQUATIONS

(2) Meyer-Peter and Muller:

Total Load Equations:

+0.66 for 1.2B  s B 70

G − 1 [(G −1)gds ] (G − 1)ds

Cppm Total sediment concentration by weight

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

+0.66 for 1.2B s B 70