John Pitlick
Geography Department
University of Colorado
Boulder, CO 80309
|
John Pitlick
Geography Department
|
|
Introduction
The total sediment load of a river consists of two time-varying components: the wash load and the bed-material load. Einstein [1950] proposed that the wash load consists of relatively fine-grained sediment supplied from diffuse sources within a watershed, whereas the bed-material load consists of coarser sediment derived primarily from the channel bed. For convenience, the cutoff between wash load and bed-material load is taken as D10, the grain size for which 10 percent of the bed material is finer. In gravel rivers this cutoff typically occurs in the sand-size range, 0.0625 < Ds < 2 mm, where the subscript s refers to the subsurface bed material (substrate). Some if not much of the sand carried in gravel-bed rivers is thus derived from the bed, and it cannot be assumed that all of this sediment is wash load or ìthroughput loadî, with the mass flux determined only by the supply.
This paper examines seasonal patterns of sediment transport in a gravel-bed river. Comparison of suspended sediment samples at a US Geological Survey (USGS) gaging station with sediment samples collected in stream-bed traps suggests that sand continues moving as bed load long after the peak in the annual hydrograph, and that this material actively exchanges with the channel bed substrate.
Study Area and Methods
|
The present study examines sediment transport in a gravel-bed reach of the Colorado River in western Colorado (Fig. 1). The study reach has the following characteristics: average slope, S = 0.0020; average channel width, B =134 m; and average bankfull depth, H = 2.5 m. Measurements of discharge and suspended sediment have been taken frequently at a USGS gauge on the Colorado River near Cameo, 25 km upstream of the study reach (station 09095500, Fig. 1). From 1982 through 1998 the USGS made 576 measurements of discharge and suspended sediment concentration at this location; 449 of these samples were analyzed to determine the fraction of material finer than 0.0625, the cutoff between silt and sand. |
Figure 1. Location of study area and USGS gauging stations. |
Figure 2. Sediment traps. A six-inch ruler is shown for scale. |
In 1998, as part of a study to assess the geomorphic and ecologic effects of upstream reservoir releases, I installed a series of stream-bed traps to monitor the movement of sand and granules over gravel bars and riffles. The traps consist of 20-cm cans mounted within a piece of plastic pipe, both placed flush with the bed surface (Fig. 2). The cans were filled with clean gravel > 32 mm in size. At various times after the peak in the annual snowmelt hydrograph the cans were retrieved, emptied of fine sediment, and replaced. Conditions limited us from retrieving cans in flows more than ~0.5 m deep. Samples taken from the traps were subsequently sieved at 1/2 phi intervals. |
Point counts of 100-200 particles were taken at a number
of locations to determine the reach-average bed-surface (pavement) size
distribution. Bulk samples consisting of 50 -100 kg of the subsurface
bed material were taken at 3 locations to determine the size distribution
of the substrate. The coarser fraction of this sediment (> 32 mm)
was sieved in the field and the finer fraction (< 32 mm) was sieved
in the laboratory.
Field Observations and Results
Gaging Station Measurements: Paired measurements of discharge and suspended sediment load at the Cameo gage show that the Colorado River consistently carries higher suspended sediment loads on the rising limb of the hydrograph than it carries on the falling limb (Fig. 3a). The silt-clay fraction of the total suspended load generally dominates over the sand fraction (Fig. 3b & 3c); however, the separation between rising- and falling-limb samples is not as apparent in the silt-clay fraction as it is in the sand fraction. Sand loads also appear to follow a more consistent relation to discharge than silt loads, suggesting that sand transport rates are governed to some extent by flow hydraulics. Least squares regression of the sand data yields the following relations:
Rising limb: Qs = 0.007Q^2.35 (r2 = 0.49)Falling limb: Qs = 0.001Q^2.44 (r2 = 0.74)
The exponents in the above relations are quite similar to each other, and not far from the value of 2.5 expected if transport rates followed a typical bed load equation. The difference in coefficients reflects a change in the load for a given discharge, due either to a change in supply or a change in the grain sizes available for transport.
Figure 3. Observed relations between discharge and suspended sediment load, Colorado River near Cameo, Colorado.
(a) Total suspended sediment load; (b) load of sediment finer than 0.0625; c) load of sediment coarser than 0.0625 mm.
Figure 4 below shows seasonal patterns in water discharge, sediment concentration, and percent sand constructed by arranging all of the flow and sediment measurements in chronological order regardless of year. The smoothed curve running through the data is fit using a locally weighted least squares method. The trends suggest that in typical years the peak in suspended sediment concentration occurs 2-3 weeks prior to the peak in water discharge (Fig. 4a). The peak in sand concentration, however, occurs 2-3 weeks after the peak in discharge (Fig. 4b). The peak in sand concentration coincides with a discharge of ~280 m3/s. At this flow the local shear velocity, u*, estimated on the basis of a calibrated 1-d flow model, is 0.15 to 0.18 m/s. Using the relations for fall velocity, ws, proposed by Dietrich [1982], and the criterion suspension, u* > ws, a discharge of 280 m3/s should be sufficient to suspend sand finer than about 0.5 mm.
Figure 4. Seasonal trends in (a) suspended sediment concentration and (b) sand load.
Bed Material Characteristics: Surface and substrate grain size distributions determined from samples of low lying gravel bars are shown in Figure 5. The median grain size, D50, of the bed surface is 58 mm, and the D50 of the substrate is 29 mm. Approximately 20% of the substrate is finer than 2 mm (the break between sand and gravel), reflecting a grain size distribution that is bimodal overall.
Figure 5. Size distributions of bed sediment.
Sediment Trap Data: The first row of panels below (Fig. 6) shows grain size distributions of sediment collected in channel bed traps from 1998-2001. For comparison substrate size distributions are also shown. The second row of panels shows corresponding hydrographs for each year. The first point to note is that the size range of sediment caught in the traps overlaps considerably with the finer fraction of the substrate (< 20 mm). Second, it appears that the percentage of coarse sand and granules caught in traps is related to hydrograph characteristics (magnitude and duration). In 1998, the year with the highest peak discharge, coarse sand and granules continued moving as bed load, and these sizes were collected in traps long after the peak in the hydrograph. In 2001, the year with the lowest discharge, nearly all of the sediment caught in traps was finer than 1 mm.
Figure 6. Upper row, size distributions of sediment collected in traps; lower row, annual hydrographs with gray lines indicating collection dates.
Conclusions
Sand-sized sediment is transported through gravel-bed
reaches of the Colorado River over a wide range of discharges. The
peak concentration of suspended sand (30-40%) occurs two to three weeks
after the peak in the hydrograph. Sediment caught in stream-bed traps
during this time period varies widely in size according to the magnitude
of the annual peak flow. Overlap of the size distributions of sediment
caught in traps with samples of the subsurface bed material indicates that
sand in transport after the peak in the hydrograph actively exchanges with
material in the bed.
Acknowledgements
This work was funded by the US Fish and Wildlife Service
(USFWS) Recovery Implementation Program for Endangered Fish Species in
the Upper Colorado River Basin. I would like to thank George Smith,
Doug Osmundson and Chuck McAda of the USFWS for advice and logistical support.
I would also like to thank Jillian Aldrin, Aaron Cloud, Ingrid Corson,
Bill McCaslin, Jennifer Nissenbaum, and Rebecca Thomas for help with field
work.
References
