Bed-Material Characterization and Transport

Bed-Material Characterization and Source

Three primary objectives motivated sampling of bed material throughout the Umpqua River study area. First, a detailed dataset of grain-size distributions collected at closely spaced intervals throughout the study area provides a foundation for evaluating longitudinal trends in transport capacity (for example, Wallick and others, 2010). Secondly, collection of particle-size data from both the surface and subsurface of gravel bars enables calculation of armoring ratios, which can be used to assess the spatial patterns in sediment supply relative to transport rates (Dietrich and others, 1989; Bunte and Abt, 2001). Third, spatial patterns in clast lithology can be used to assess bed-material contributions from tributary basins (Wallick and others, 2010).

Gravel Distribution and Textures

Throughout most of the study area, the Umpqua River above the head of tide, along with the North Umpqua and South Umpqua Rivers, flows directly on bedrock alternating with boulder-cobble substrates. Locally flanking the channel are gravel bars, commonly small, thin, and discontinuous above and adjacent to bedrock outcrops. Some bars, however, are large, with areas that exceed 120,000 m² and thicknesses of possibly several meters. As mapped from 2005 aerial photography, the total area of roughly 336 gravel bars (minimum mappable area 300 m²) along the main stem Umpqua and South Umpqua Rivers between the head of tide (FPKM 40) and Tiller (FPKM 273.1) was approximately 2.7 km² (fig. 21), accounting for only 10.5 percent of the total active channel area and covering much less area than the 19 km² of low-flow channel area.

Positions of most bars are fixed by valley physiography and bedrock outcrops, but although their locations are constant over time, aerial photographs show that bar texture and overall appearance can change in response to flow conditions. Bar height above the low-water surface, as determined from field observations and LIDAR topography (which covers part of the Coast Range reach), ranges from below the low-flow water surface on the low-elevation bars to more than 1 m on the high surfaces of stable bars.

Sampling

Bed-material textures on gravel bars along the Umpqua River system were measured by sampling 51 bars throughout the study area in August 2009. Of these, 27 were on the South Umpqua River, 5 on the North Umpqua River, 14 on the main stem Umpqua River, and 5 on other tributaries (table 9). Along the main stem Umpqua and South Umpqua Rivers between Scottsburg (FPKM 40) and Tiller (FPKM 273.1), the average distance between sampling sites was 6.5 km. The distance between sample sites was greater, reaching intervals of as much as 20 km, along the lower reaches where bars are sparse. Sampling sites were selected on the basis of bar size, accessibility, and their ability to represent reach-scale conditions. All five sample sites on the North Umpqua River were located upstream of Winchester Dam (FPKM 180.9), as no substantial gravel deposits were found downstream of the dam. Sites on three major tributaries (Calapooya, Myrtle, and Cow Creeks) also were sampled to characterize bed-material sediment entering the Umpqua River system.

Surface-particle sizes at each of the sampling sites were measured by a modified grid technique (Kondolf and others, 2003). At each site, 200 particles were measured at 0.3-m increments along two parallel 30-m tapes using an aluminum template (Federal Interagency Sediment Project US SAH–97 Gravelometer). The tapes were spaced 1–2 m apart and were aligned parallel to the long axis of the bar (fig. 36).

Although most sampling was conducted at bar apices to enable consistent comparisons, bedrock outcrops, vegetation, and irregular bar topography resulted in some bars not having a clearly defined apex (which we defined as the topographic high point along the upstream end of the bar). In such instances, a section of the bar that appeared active and representative of the overall bar was measured. Few of the bars, however, had uniform surface textures; many either had irregular patches of different-sized clasts, varying amounts of exposed bedrock and vegetation, or had been disturbed by vehicle traffic. Hence, some variation among bars can be attributed to local depositional conditions and post-deposition disturbance.

Bed-material substrate was sampled at 30 of the 51 surface-material sites to evaluate textural differences between the surface and subsurface material (a measure of “armoring”) and to support sediment transport calculations. The samples were collected by removing the surface layer to a depth approximately equal to the maximum grain size, and then collecting approximately 40 L of sediment from an area approximately 30–50 cm in diameter and 30–50 cm in depth (fig. 36). The bed-material substrate was analyzed by the USGS Sediment Laboratory in Vancouver, Washington, where the samples were dried and sieved into half-phi intervals. Total sample weights ranged from 53 to 83 kg, with an average sample weight of 69 kg. At many sites, the sample weight did not quite meet the criteria suggested by Church and others (1987), although 22 of the 27 samples had sample weights that were at least 50 percent of the recommended weight. These same 22 samples were judged to have medium accuracy (whereby the largest particle represented no more than 1 percent of the total sample mass).

Assessment of Bed-Material Sizes

For the surface samples of the Umpqua and South Umpqua Rivers, the median particle diameter (D₅₀) ranged in size from 16 to 122 mm (fig. 37; table 9). As for most rivers, median bed-material particle size diminishes downstream. The coarsest samples were measured along the Days Creek reach between FPKM 240 and FPKM 268, where the channel flows through a series of large, alternating bars downstream of the confluence of Coffee Creek. Variability is greatest within the Days Creek and Roseburg reaches, where surface material median grain-size diameter (D₅₀) can differ by more than 30 mm between bars spaced 2–3 km in distance (fig. 37; appendix C). This heterogeneity in surface textures reflects the wide ranging differences in the size and character of gravel bars along the Umpqua River system, as adjacent sampling sites varied considerably with respect to local hydraulic conditions, gravel thickness, abundance of bedrock, and degree of vegetation.

The subsurface samples were considerably finer and had less spatial variability than the surface-material samples measured at the same locations (fig. 37; appendix C). Subsurface D₅₀ (D_50s) increased along the Days Creek reach, decreased sharply from 35 to 10 mm at the confluence of Cow Creek (FPKM 232), then remained nearly constant along the Roseburg reach until coarsening to about 35 mm at the confluence with the North Umpqua River. Downstream of the confluence of the North Umpqua and South Umpqua Rivers, D_50s was relatively constant at about 20 mm for more than 100 km as the Umpqua River traverses the Coast Range.

Particle-size distributions show that although the coarsest fractions of the subsurface and surface samples were similar in size, the bed-material subsurface was dominated by a finer matrix of sand to pebble-sized particles (ranging in size from 1 to 10 mm), whereas the bar surfaces were dominated by cobble-sized clasts greater than 30 mm (table 9; appendix C). Disparity between surface and subsurface particle size is commonly attributed to an imbalance between sediment supply and transport capacity, with the surface layer coarsening when the transport capacity of the fine fraction exceeds its supply (Dietrich and others, 1989; Buffington and Montgomery, 1999).

Hence, the ratio of D₅₀ to D_50s (or the “armoring ratio”) can be used to infer the balance between sediment availability and transport capacity. Armoring ratios close to 1, indicating similar surface and subsurface sediment median grain size, indicate high sediment supply, whereas channels with excess transport capacity typically have armoring ratios closer to 2 (Bunte and Abt, 2001). Along the Umpqua and South Umpqua Rivers, armoring ratios ranged from approximately 1 to 4.7, but more than half of the measured bars had armoring ratios greater than 2, indicating excess available shear stress and transport capacity relative to bed-material supply (fig. 37; table 9). The mean armoring ratio for the 25 measurement sites on the South Umpqua and Umpqua Rivers was 2.3, slightly higher than the 2.0 value measured for three sites on the Chetco River (Wallick and others, 2010).

The limited sampling conducted on tributary streams indicates that surface material entering the South Umpqua and Umpqua Rivers from Cow Creek, Myrtle Creek, and Calapooya Creek generally is finer than the bed-material sediment in the main stem channel. The two sampling sites on Cow Creek had D₅₀ values of 27.7–30.9 mm, compared to an average D₅₀ of 76 mm for the Days Creek reach of the South Umpqua River upstream of the Cow Creek confluence. Surface-material samples from the North Umpqua River were coarser than main stem Umpqua River gravel bars, but this may be partly due to the location of the North Umpqua River sampling sites, which were more than 15 km upstream of the confluence because of the absence of suitable bars for sampling along the lower reaches of the North Umpqua River.

Although there are too few samples to determine trends regarding sediment supply imbalances on tributary streams, the tributary sites generally are less armored than the Umpqua and South Umpqua Rivers. Cow Creek had the lowest armoring ratio (1.01), consistent with a balance between sediment supply and shear stress, whereas Myrtle Creek and the Calapooya Creek had armoring ratios of 1.78 and 1.66, respectively. The armoring ratios for the North Umpqua River were relatively low (1.17 and 1.34) and probably not representative of overall reach conditions, as these sites had large areas of recently deposited gravel, and essentially no armor layer, whereas all other sites on the North Umpqua River where only bar surface material was sampled appeared substantially armored.

Bed-Material Lithology and Sources

In addition to measuring sediment texture, clast lithologies were characterized at most surface bed-material sampling sites in order to support inferences of major sources of bed material (see section below, “Basin-Scale Bed-Material Sediment Yield”). Although clasts of many lithologies are present in Umpqua River gravel bars, reflecting the varied source terrains (fig. 1), the assessment was simplified into three broad categories readily distinguished by field inspection: (1) intermediate to coarse-grained felsic igneous and metamorphic rocks (here termed felsic intrusive rocks, chiefly from the Klamath Mountains terrain), primarily light‑colored granitic and gneissic rocks, (2) brown sandstones and shales, mainly from the Tyee Formation and equivalents (termed sandstones and mainly from the Coast Range terrain), and (3) all others, which mainly included igneous rocks derived from the Cascade Range. The sandstone category was only assessed at sites downstream of the confluence of the North Umpqua and South Umpqua Rivers; upstream of the confluence, sandstone clasts were rare if present at all.

For the surface-sampling sites, clasts were inspected at 400 points during the particle-count measurements, but classified only if greater than a 16-mm diameter (resulting in total assessed sample sizes at each measurement site ranging between approximately 200 and 400, depending on the surface texture of the bar). Similarly, for the subsurface samples, we classified all sieved clasts greater than 16 mm. The surface samples were done over the 4-week course of field sampling and involved different crew members. Consequently, the categorization of the surface samples may not be as consistent as that for the subsurface samples, which was done in a concentrated effort in the laboratory by a single crew. Sandstone clasts were not categorized for the subsurface samples because they did not reliably survive sieving and transport.

For the surface clast counts along the main stem Umpqua and South Umpqua Rivers, the percentage of felsic intrusive clasts ranged from less than 1 percent to as high as 22 percent. Similarly, the subsurface percentages range from 0.6 to 20.9 percent. The North Umpqua River has few sources of felsic intrusive clasts, reflected in the counts of 1 percent or less. Myrtle Creek and Cow Creek, both of which drain parts of the Klamath Mountains terrain (fig. 1), have higher percentages of felsic intrusive clasts, with Myrtle Creek having one surface sample exceeding 20 percent (table 9).

On the main stem Umpqua and South Umpqua Rivers, felsic intrusive clasts were most abundant along the upper reaches of the South Umpqua River and decreased in abundance downstream (fig. 37). The highest concentrations were along the South Umpqua River within the Days Creek reach, where bar surfaces between FPKM 257 and 269.1 all had felsic intrusive clasts composing more than 10 percent of the surface samples (table 9; fig. 37; appendix C). These high percentages in part owe to the South Umpqua River traversing outcrops of gneiss and coarse-grained schist in this reach, but also to the increasing percentage of total area granitic source terrains such as that drained by Elk Creek, which enters the South Umpqua River at FPKM 272. Downstream, the percentage of felsic intrusive clasts diminishes to values consistently between about 1 and 4 percent.

The percentage of sandstone clasts in the surface samples ranged up to 8.6 percent (table 9; fig. 37; appendix C), but typically were less than 5 percent. Although not specifically counted upstream of the confluence of the North Umpqua and South Umpqua Rivers, they were exceedingly rare if present at all, almost certainly accounting for less than 1 percent of surface clasts. The overall distribution of sandstone clasts reflects the near absence of Paleogene sedimentary rocks upstream of the confluence of Lookingglass Creek at FPKM 119.4, and the increasing area of sandstone sources downstream within the Coast Range terrain.

Estimation of Bed-Material Transport Capacity from Transport Equations

Equations of bed-material transport use channel hydraulics and sediment characteristics to estimate sediment fluxes on streams. Although subject to certain assumptions and limitations, such equations can be applied for any stream where information on flow, channel geometry, and bed‑material characteristics is available (Collins and Dunne, 1989; Gomez, 1991; Hicks and Gomez, 2003). Moreover, these formulas provide a relatively rapid means of estimating sediment flux across a range of flow scenarios, from individual storm events to decades. For the Umpqua River study area, multiple transport equations were applied for 39 of the sediment sampling sites between FPKM 43 and 270, as well as for sites of three long-term gaging stations. These calculations encompassed the period 1951–2008, aligning with the flow record available from all three gaging stations. The approach applied to the Umpqua River study area largely follows from that applied to the Chetco River in southwestern Oregon (Wallick and others, 2010).

Although several empirical and semi-empirical transport equations are available for bedload transport (Gomez and Church, 1989), all these relations actually predict sediment transport capacity, defined as the “maximum load a river can carry” (Gilbert and Murphy, 1914, p. 35). For situations where there is unlimited bed material available from upstream sources, as well as local erosion from the channel bed and banks, a correct relation for transport capacity coupled with accurate descriptions of flow and bed material should result in accurate estimates of bed-material flux. For the Umpqua River system, however, the assumption of unlimited sediment supply is not valid, as indicated by (1) the extensive and bare bedrock surfaces in and flanking the channel (fig. 16; Howard, 1998; Klingeman, Professor Emeritus, Water Resources Engineering, Oregon State University, written commun., 2010) and (2) the abundant bars with armor values exceeding 2, thereby suggesting either high transport capacities relative to the supply of fine-grained bed-material or low-transport capacity relative to the supply of coarse-grained bed material (fig. 37). This situation contrasts with that of the similar analysis conducted for the Chetco River in southwestern Oregon, where the low-flow channel is formed in gravel and flanked by voluminous gravel accumulations forming a nearly continuous swath of tractively transported bed-material sediment for the lowermost 18 km (Wallick and others, 2010). For the more sediment-limited Umpqua River, consequently, calculated transport capacities are best considered indicative of maximum plausible bed-material transport rates, and likely overestimate actual fluxes by substantial margins.

Even in more sediment-rich situations where river conditions satisfy the requirement that bed-material transport is a function of flow, channel, and bed texture rather than sediment availability, large uncertainties still arise because bed-material transport is highly variable in time and governed by highly nonlinear relations between local flow and bed‑material transport—both of which are difficult to characterize at high resolution (Gomez, 1991; Wilcock and others, 2009). These challenges, in conjunction with the wide variety of field situations and few measurements, in part explain the large number of transport equations available and the variation in their forms and data requirements (Hicks and Gomez, 2003). For this study, we assess and possibly mitigate for these factors by (1) evaluating multiple transport relations for multiple cross sections, (2) where possible, characterizing flow at individual cross sections using the results from a one‑dimensional flow model, and (3) evaluating the results in the context of other information on sediment transport rates.

Equation Selection and Analysis

The bedload transport calculations for the Umpqua River were implemented by the software package Bedload Assessment in Gravel-bedded Streams (BAGS), a program operating within a Microsoft® Excel® workbook (Pitlick and others, 2009; software and documentation available at http://www.stream.fs.fed.us/publications/bags.html). The BAGS software enables users to select from six semi-empirical transport formulas that were developed and tested using data from gravel or sandy-gravel streams (Wilcock and others, 2009). Users specify a transport equation and provide information describing channel geometry, flow, and sediment parameters. With this information, bed-material transport rates are calculated for a specific flow and cross-section geometry.

The bedload transport formulas implemented in BAGS are:

• Parker–Klingeman–McLean, a subsurface-based equation (Parker and others, 1982)
• Parker–Klingeman, a subsurface-based equation (Parker and Klingeman, 1982)
• Bakke and others, a calibrated equation version of the Parker–Klingeman formula (Bakke and others, 1999)
• Parker, a surface-based equation (Parker, 1990a, 1990b)
• Wilcock, a two-fraction calibrated model for sand and gravel, (Wilcock, 2001)
• Wilcock and Crowe, a surface based equation (Wilcock and Crowe, 2003)

Although all six formulas are substantially similar and have been successfully applied to gravel-bed rivers, key attributes differentiate the equations, elaborated in Wilcock and others (2009). The subsurface-based methods (Parker–Klingeman–McLean and Parker–Klingeman) rely on grain-size data from the bed subsurface, beneath the coarser cobble-pavement forming the surface of most bars. Both subsurface-based approaches were developed from measurements made by Milhous (1973) at Oak Creek, a small gravel-bed stream in the Oregon Coast Range. By contrast, two surface-based methods are based on grain-size distributions from bed surfaces. The Parker (1990a, 1990b) equation is a surface-based method also developed from grain‑size distributions and transport rates at Oak Creek, whereas the second surface-based method implements the Wilcock and Crowe (2003) equation, which is partly based on the Parker (1990a, 1990b) approach, but is supplemented by flume experiments evaluating the role of sand content on gravel transport.

The main distinction between the two surface-based approaches is in the determination of the reference Shields shear stress (τ*_rsg); in the Parker (1990a, 1990b) equation, τ*_rsg is assumed to be a constant value of 0.0386, but in the Wilcock and Crowe (2003) equation, τ*_rsg varies with the sand content of the surface material. The two calibrated equations of Bakke and others (1999) and Wilcock (2001) require measurements of bedload transport to calibrate reference shear stress, and thus improve the overall transport estimates. These relations are not applicable to this study because of the absence of direct measurements, resulting in the implementation being restricted to the four uncalibrated transport capacity relations.

On the Umpqua River, bedload transport capacity was estimated at a total of 42 sites along the main stem Umpqua and South Umpqua Rivers, including the 39 bed-material sampling sites and the 3 gaging station locations (table 9). Transport also was estimated for two sites on lower Cow Creek, including the gaging station on lower Cow Creek and a nearby bed-material sampling site. Each calculation requires information on flow, bed-material size distribution, cross-sectional geometry, and water or energy-surface slope. No accurate and consistent sources of these measurements were available for all sites. Consequently, information was derived from various sources: flow data were obtained from the USGS gaging stations; sites within the Coast Range and Garden Valley reaches were assigned discharges from the gaging station at Elkton (14321000); and flow records from the gaging stations at Brockway (14312000) and Tiller (14308000) underlie the calculations for the Roseburg and Days Creek reaches, respectively. The gaging station near Riddle on Cow Creek (14310000) was used to calculate transport for the two sites on Cow Creek.

Because nearly all transport capacity calculations were made at August 2009 bed-material measurement sites, these bar-texture measurements were used directly in the transport equations and were applied to the entire cross section. For the four sets of calculations at the gaging stations for which there were no sediment texture measurements, we averaged grain-size distributions from adjacent measurement sites (table 9). At the 15 analysis sites where only bar surface material was sampled, only the surface-based equations of Parker (1990a, 1990b) and Wilcock–Crowe (Wilcock and Crowe, 2003) were applied, whereas at the remaining 24 sites where bar subsurface material was sampled (as well as for the gaging stations where subsurface grain sizes were estimated from nearby sample locations), we also applied the subsurface‑based formulas of Parker–Klingeman (1982) and Parker–Klingeman–Mclean (1982).

Cross-section geometry information was limited by the lack of continuous and high-resolution topographic and bathymetric data. For transport calculation locations between FPKM 83.5 and FPKM 155 (Coast Range reach and lower part of the Garden Valley reach), channel cross sections were extracted from a 2009 LIDAR survey (table 7). Elsewhere, cross sections were extracted from the USGS National Elevation Data 1/3 arc second digital elevation model (approximately 10 m resolution, U.S. Geological Survey, 2010c). For all sites except the Days Creek reach, a trapezoidal cross-section shape was assumed, and the cross sections extracted from these elevation data were adjusted to reflect actual streambed topography using early 1970s thalweg elevations from USGS flood studies of the Umpqua and South Umpqua Rivers (Oster, 1972, 1973; table 7). The highly stable, chiefly bedrock channel (fig. 30) reduces the uncertainty introduced by using such old channel-depth data. These flood studies did not include the Days Creek reach; therefore, streambed elevations within this reach were estimated to be 1 m below water surface indicated by the digital elevation models—a value broadly consistent with observations by field personnel. At the three analysis sites located at streamflow-gaging stations, channel cross sections were obtained from recent USGS measurement surveys (U.S. Geological Survey, 2010a).

A key hydraulic variable for computing transport rates is the energy slope (Sf ), which was approximated using water surface slope (Sw). For the Coast Range, Garden Valley, and Roseburg reaches (FPKM 41–231), Sw was obtained from the 0.1 annual exceedance probability flood profiles calculated by a one-dimensional step-backwater model (Oster, 1972; 1973; 1975). From these calculated water-surface profiles, we extracted water-surface slope for distances ranging between 550 and 850 m and spanning the transport capacity calculation location (in nearly all cases, sites of bed-material size analyses). For the Days Creek reach, Sw was determined for 1- to 2-km-long channel segments spanning the analysis sites using low-flow water-surface elevations, as depicted by 5-ft contour intervals on the USGS 1914 Plan and Profile surveys (Marshall, 1915).

On the basis of these morphologic, bed-texture, and hydraulic characterizations, we calculated bed-material transport rates for 26 discharges spanning the range of historical flows recorded at nearby USGS streamflow-gaging stations. Using an approach similar to that applied to the Chetco River (Wallick and others, 2010), the results for each discharge produced a relation between discharge (Q) and bed-material transport rate (Qs), which were fitted by 5th order polynomial curves to produce sediment-discharge rating curves for each analysis site (fig. 38). Although sediment discharge rating curves are typically modeled using power law relationships (Wilcock and others, 2009), the 5th order polynomial curves provided a better approximation of the Q–Qs relationship than was achieved using a power law relationship.

In conjunction with discharge records of October 1, 1988–September 30, 2008, from the corresponding USGS streamflow-gaging stations, the calculated Q–Qs relations enabled calculations of annual sediment transport fluxes at each of the analysis sites. Although annual transport volumes typically are calculated using mean daily values (for example, Collins and Dunne, 1989), the combination of highly nonlinear transport rates and the rapid flow changes in the Umpqua River basin during transport events, cause annual bed-material transport volumes determined from mean daily values to likely underestimate true values. Therefore, annual bed-material transport volumes were based on the unit discharge values acquired every 30 minutes and archived electronically since 1988 (although 15-minute flow data are available for WYs 2006–08, to simplify the calculations, these data were not used). For WYs 1988–2006, transport rates were calculated for each analysis site using the 30-minute unit-flow data and summing total transport for each day.

To extend the record back through WY 1951 and to fill recent periods when unit flow data were not available, relations for each calculation site were developed between daily transport volumes calculated from the unit-flow measurements and mean daily flow for all days of predicted transport. These regressions, which typically had correlation coefficients ranging between 0.97 and 0.99, were applied to all days to permit calculations for the entire record of October 1, 1951–September 30, 2008. This analysis period coincides with the construction of the North Umpqua hydropower dams (completed by 1955) and encompasses the construction of the Galesville flood control reservoir in the upper Cow Creek watershed in 1985.

In addition to total transport volumes for each site for each of the four bedload transport capacity equations, we calculated the reference discharge, Q_T,P, as predicted by the Parker (1990a, 1990b) transport capacity equation. The reference discharge is the flow for which shear stress is equivalent to the reference shear stress required for very small but measurable transport (Parker and others, 1982; Parker, 1990a, 1990b). This value can be compared to annual flow characteristics to assess the frequency of transport or degree of mobility for each measurement site, thereby providing a measure of how frequently a bar may be subject to mobilization. For the measurements here, we calculated the ratio of the reference discharge to the flood with a 0.67 annual exceedance probability (1.5-year flow) to determine a mobility index (table 9). Sites with ratios of unity or less are predicted to have measurable transport for the 1.5-year flood discharge, but those sites with mobility indexes greater than 1 require larger and less frequent flows for mobilization.

Uncertainty and Limitations

Bedload transport calculations are sensitive to grain size, slope, depth, and discharge (for example, Wilcock and others, 2009). Hence, uncertainties in these parameters affect our calculations of transport capacity for the South Umpqua and Umpqua Rivers. These uncertainties arise from difficulties in measuring or calculating many of these factors in the complicated river channels of the Umpqua River basin, especially in locations where high-resolution topographic data are lacking. Although grain size is easily measured, many of the Umpqua River gravel bars are heterogeneous, and characterized by patches of various sized clasts and intermittent bedrock. Accordingly, the sampling site location within a particular bar, and the resulting measured grain-size distribution, influences the transport capacity calculations for that site. Additionally, the one-dimensional flow model used to calculate Sw is most valid along straight reaches, yet many of the largest bars in the study area were situated along bends, which are not well represented in a one-dimensional model because the hydraulics are dominated by strong secondary flow currents and bedload transport is almost certainly nonuniform across the channel (for example, Dietrich and Smith, 1984). Other sources of uncertainty stem from the limited bathymetric data available for the study area and our resultant dependence on coarse-scale topographic information to characterize channel geometry and bathymetry at some sites.

Calculations that used ranges of values for grain size, slope, depth, and streamflow were used to evaluate the sensitivity of the calculated bed-material transport capacity values (fig. 39). Using flow data from WY 1999 as the base case scenario, annual transport capacity was calculated at the Umpqua Sand and Gravel Bar (FPKM 171.4) by applying the Parker (1990a, 1990b) equation and individually varying each input parameter. Water year 1999 was selected as a typical year for the sensitivity analysis because although it had a mean annual flow approximately 30 percent higher than the average flow for the period 1955–2004, the peak flow during this year was similar in magnitude to the 1.5-year exceedance event.

Of the four parameters evaluated, annual transport capacity was most sensitive to variation in grain size and energy slope, both of which are affected by the calculated transport rate by about a factor of 2 to 4 when increased or decreased by 25 percent. The transport calculations were less sensitive to mean daily flow values and flow depth, for which a ±25 percent variation affected annual transport capacity totals by about a factor of 2 or less (fig. 39). Although this assessment is specific to the Parker (1990a, 1990b) equation, results are likely to be similar for all transport equations because of their similar forms. These results indicate that reasonable uncertainties in the primary factors affecting bed-material transport stem from the heterogeneous nature of gravel bars in the Umpqua River system, which can lead to uncertainties in calculated annual transport volumes approaching a factor of 4 for a specific transport capacity equation. These parameter-related uncertainties are in addition to that resulting from the choice of bed-material transport capacity equation, which is difficult to assess in the absence of direct measurements.

Results of Bed-Material Transport Equations

Application of the 4 bed-material transport formulas to the 42 sites in the Umpqua River study area for 57 years shows wide temporal and spatial variability in the predicted annual bed-material transport capacities, ranging from negligible transport capacity in some years for many sites to bed-material transport capacities as great as 600,000 metric tons/yr for some sites in high-flow years (table 9; figs. 40 and 41).

The large annual variation at a site owes to the nonlinear relation between flow and bed-material transport capacity (figs. 38 and 41). This is evident by considering the Umpqua Sand and Gravel Bar at FPKM 171.4, for which the mean annual transport capacity for 1951–2008 is 9,070 metric tons/yr as calculated by the Parker (1990a, 1990b) equation, which is similar to the 8,850 metric tons/yr median value for all 27 calculation locations in the Days Creek and Roseburg reaches. At the Umpqua Sand and Gravel Bar, two-thirds of the years have calculated transport capacities less than the mean value of 9,070 metric tons/yr (fig. 41C). The high frequency of low transport years is compensated by a few years of much greater transport capacity: 9 years with values greater than 20,000 metric tons/yr, and exceptional years, such as 1956, 1965, 1974, and 1997 with bed-material transport capacities exceeding 30,000 metric tons/yr (fig. 41C). Considering the calculated daily transport values for the 58-year period of record, the Parker (1990a, 1990b) equation predicts transport on about 15 percent of all days. Half, however, of the total cumulative calculated transport capacity for the Umpqua Sand and Gravel Bar occurred over 80 days (less than 0.4 percent of total days), and 10 percent of the total transport capacity for this 58-year period was over just 6 days, including October 29, 1950, December 23, 1964, and January 16, 1974, which all had transport capacities exceeding 10,000 metric tons.

The annual bed-material transport capacity values for the Umpqua Sand and Gravel Bar decrease significantly (P <0.05 based on Parker [1990a, 1990b] calculations) over the 58-year period of record. This decrease corresponds to the overall decrease in peak flows since the early 1950s (P <0.05 for USGS streamflow gage at Brockway; fig. 5, table 2), particularly after 1974 (fig. 41C) and probably reflects regional and longer term climate cycles controlling flow volumes and peak discharges.

Understanding the spatial variations in calculated transport rates is more challenging. Although the largest bar complexes (including Maupin Bar and those near the confluence of Cow Creek, fig. 20) broadly coincide with zones of decreasing transport capacity, there is considerable site-to-site variability. Average annual transport capacity calculated by the 4 equations and among the 42 calculation locations (39 bars and 3 gaging-station locations) ranges from 0 to 623,000 metric tons/yr (table 9). The two surface‑based transport equations (Parker and Wilcock–Crowe) generally predicted greater transport capacity than the subsurface-based equations (Parker–Klingeman–Mclean and Parker–Klingeman) for sites within the Days Creek, Roseburg, and Garden Valley reaches, but less transport capacity than the subsurface-based calculations for the Coast Range reach. These results are consistent with bar-surface particle sizes decreasing faster with respect to river location than that for subsurface bed material (fig. 37). The two subsurface-based methods (Parker–Klingeman–McLean and Parker–Klingeman) produce similar bedload rating curves (fig. 38), and at most sites, the mean annual transport capacities predicted by the two equations differ by less than 20 percent (table 9). For most sites, the two surface-based equations of Parker (1990a, 1990b) and Wilcock and Crowe (2003) generally agree within an order of magnitude (table 9), with the Wilcock–Crowe equation predicting higher levels of bedload transport at locations where there is a higher proportion of sand.

Median values among the 4 equation-based calculations of average annual transport for the 1951–2008 period for all 39 main stem and South Umpqua River sites for which we had local measurements of bed texture (including the 24 sites where subsurface-based calculations were applied) range from 12,800 to 27,200 metric tons/yr depending on the transport equation. Assessed by reach, the median values increase downstream from 4,450 to 12,620 metric tons/yr in the Days Creek reach to 20,280 to 56,440 metric tons/yr in the Coast Range reach.

Consideration of only these median values, however, masks the six-orders-of-magnitude variation in calculated bed-material transport rates among sites (fig. 40). Although the sensitivity analysis showed that transport capacity calculations are sensitive to both grain size and slope, plausibly affecting the calculated capacity values in this study by as much as a factor of four, the much wider scatter in the computed values for the Umpqua River indicates that other factors are also important. Most of the variation in transport values is the likely result of applying transport capacity equations to a channel system for which bed-material transport is limited by sediment supply rather than transport capacity.

For river systems in which bed-material transport at all locations is limited by flow capacity and for which the channel is in steady state with respect to channel and flood-plain storage of bed material, the expectations are (1) the active channel would be chiefly alluvial and consist of alternating bars flanking a low-flow channel formed in alluvial bed material, (2) all bars would be subject to gravel transport during approximately the same flows, and (3) calculated transport rates would be similar from site to site and variations would chiefly reflect changes in supply resulting from tributary inputs and particle attrition. By contrast, bedrock forms much of the active channel for the Umpqua River, particularly in the Coast Range reach, where the area of bedrock almost everywhere exceeds the area of gravel bars (fig. 18). Moreover, for all reaches of the Umpqua River other than the Tidal reach, the channel flows mostly over bedrock (fig. 19). As a consequence, bar locations and their textural characteristics reflect local hydraulic conditions established by valley and bedrock morphology rather than broad-scale transport conditions (fig. 16).

The diversity of bar types with correspondingly wide‑ranging textures is the main reason for the large range of calculated transport capacity values. For example, some bars appear to be largely relic or only active during exceptional flows. Maupin Bar, at FPKM 94.5, for which sequential photographs show little change since 1967, is likely an example of a bar only rarely subject to substantial bed-material transport, which probably occurs during flows similar to the December 1964 flood (fig. 23B, table 9). Aerial photographs and field observations show that although this type of sampling site may appear bare and recently scoured in aerial photographs, many sites (like Maupin Bar) are overlain by a coarse armor layer (table 9). Consistent with this observation, these bars have high mobility indexes; the reference discharge (QTP) for Maupin Bar, is 4,400 m³/s as calculated by the Parker (1990a, 1990b) transport capacity equation, which is twice the magnitude of the 1.5-year annual peak flow, and similar in magnitude to the 10-year annual peak flow (fig. 38B, table 9).

Infrequent mobility results in low calculated transport capacities. For the case of Maupin Bar, the mean annual transport rate was 50 metric tons/yr as calculated by the Parker (1990a, 1990b) transport capacity equation. Such calculations associated with a stable bar probably do not reflect reach-scale bed-material transport conditions. In contrast, other bars have very low calculated reference discharges, including six with mobility indexes less than 0.1, indicating that measureable transport is predicted at flows less than one-tenth the 1.5‑yr flood. These bars are associated with the greatest calculated annual fluxes, such as the 305,420 metric tons/yr calculated by the Parker (1990a, 1990b) transport capacity equation for the Myrtle Creek Bridge Bar at FPKM 219.9 (table 9). Many of these bars, including the Myrtle Creek Bridge Bar (fig. 26A), Willis Creek Bar (FPKM 204), and Hutchinson Wayside Bar (FPKM 112.3) are in the lee of bedrock protrusions or are otherwise in locations of complicated hydraulics for which bar compositions may more closely reflect local hydraulic and depositional conditions rather than reach-scale bed-material transport conditions (fig. 16).

The wide range of calculated transport capacity values and the apparent supply limited bed-material conditions for the Umpqua River hinder understanding as to how the calculated transport capacity values might relate to actual bed-material transport rates. Pitlick and others (2009) provide guidance for evaluating such calculations pointing out, on the basis of data reported by Mueller and others (2005), that mobility indexes for most gravel-bed rivers for which flux is inferred to be transport limited range between 0.3 and 1.25. For the Umpqua River bars for which the reference discharge is within this range, the annual transport capacity values range between 600 and 50,000 metric tons. The median value of mobility index for the 45 datasets considered by the Mueller and others (2005) analysis is 0.67. For the 39 sites on the Umpqua River, the relation between transport capacity (as calculated by the Parker [1990a, 1990b] equation) and site mobility index indicates that a mobility index value of 0.67 correlates to an annual flux of 4,000 metric tons/yr, although the 2-sigma range (95-percent confidence interval) bracketing this prediction ranges from 500 to 33,000 metric tons/yr (fig. 42).

The high degree of spatial variability in bed textures and transport rates demonstrates the challenges of applying bedload transport formula to the Umpqua River, where bed-material transport is supply limited. Of the 4 transport formulas applied to the 39 bed-material sampling sites, the surface-based equations of Parker (1990a, 1990b) and Wilcock and Crowe (2003) probably are most applicable because many sites are armored. If bed-material conditions and bar textures were to change, however, than the subsurface-based equations of Parker–Klingeman and Parker–Klingeman–McLean would become more applicable to the study area. Selecting an appropriate range of values to characterize reach-average transport rates is difficult because of the nonuniform nature of Umpqua River gravel bars and transport rates. Consideration of bars with intermediate mobility indexes (0.3<Q_T,P/Q_1.5yr <1.25) as proposed by Pitlick and others (2009) provides one approach for characterizing actual flux rates, although these rates probably are best considered maximum plausible bed‑material transport rates because of the semi-alluvial character of the Umpqua River. In our judgment, the resulting range of annual transport values of 500–20,000 metric tons/yr, as predicted by the Parker (1990a, 1990b) equation for these intermediate mobility sites, most plausibly reflects overall bed‑material transport rates, although transport at some sites and for some years will vary substantially from this range.

Basin-Scale Bed-Material Sediment Yield

An empirical and independent approach to estimating bed-material flux along the Umpqua River was derived from relations between measured sediment yield and predictor variables, such as basin slope, precipitation, and drainage density. This approach is modeled after that used for assessing the sediment yield in the Deschutes River basin of central Oregon (O’Connor and others, 2003) and globally by analyses such as Milliman and Syvitski (1992). In conjunction with estimates of bed-material abrasion rates, this approach permits spatially explicit estimates of bed-material flux for each of the study reaches as well as assessment of the effects of sediment trapping by dams and diversions.

The approach used in this study follows from the premise that basin slope exerts a primary influence on sediment yield. Specifically, in steady-state landscapes dominated by diffusive processes of surficial sediment mobilization (such as biogenic activity, soil creep, freeze–thaw action), sediment flux per unit stream length is proportional to the gradient of the flanking hillslope (Culling, 1960, 1963; Hirano, 1968), although this relation may not be linear in steeper terrains (Andrews and Buckman, 1987; Roering and others, 1999). Consequently, sediment yield per unit area should be proportional to the product of the average slope gradient and drainage density, termed the sediment production index (SPI) by O’Connor and others (2003).

Alternative approaches rely on correlations between sediment yield and precipitation (Langbein and Schumm, 1958; Douglas, 1967; Walling and Webb, 1983, p. 81), and lithology (for example, Schmidt [1985] and Hicks and others [1990]). Hooke (2000) in summarizing available data and analyses concluded that slope steepness and precipitation (as it affects runoff) are the key factors positively correlated with sediment yield but that precipitation is more difficult to assess because it also controls vegetation cover, which is inversely correlated to sediment yield. This study examined correlations between sediment yield and (1) mean basin slope, (2) the product of basin slope and drainage density (SPI of O’Connor and others, 2003), and (3) the product of basin slope and mean annual basin precipitation (following the analysis of Hooke, 2000).

This analysis does not explicitly consider lithology, which has been shown to be a strong predictor of sediment yield in some studies (Aalto and others, 2006; Syvitski and Milliman, 2007), and which is a very important factor for the Umpqua River basin, judging from the correspondence of gravel-bar abundance with Klamath Mountains source areas. The rationale is that geology is difficult to parameterize objectively in a manner relevant to producing bed material, and for many areas in the region is strongly correlated with slope and drainage density (O’Connor and others, 2003; Jefferson and others, 2010).

Sediment Yield Measurements

Underlying this analysis are 33 measurements of sediment yield from 26 basins in the Cascade and Coast Ranges of Oregon and northern California. Sediment-yield measurements were compiled from reservoir surveys, reservoir delta surveys (surveys of sediment volumes deposited at river or stream entrances into reservoirs), bedload sampling, and bedload transport equations confirmed by sampling. The yield measurements and calculated totals encompass durations of 1 to 95 years and basins ranging from 0.7 to 6,901 km² (table 10).

This analysis focused on compiling bed-material yield rates from previous studies, which required estimation of the bed-material volumes from the reported total sediment volume measurements. Bed-material transport rates were measured and calculated directly for Oak Creek, Chetco, and Smith Rivers, as well as the measurements from the H.J. Andrews Experimental Forest, so these reported values were used without modification. For studies where the total fluvial load was measured, such as for the Pistol River and Redwood Creek, the bed-material load was assumed to be 20 percent of the suspended load, in broad agreement with the ratio of bed‑material load to suspended load measurements at Smith River (which drains similar terrain as Redwood Creek), Redwood Creek, and at the H.J. Andrews Experimental Forest (table 10).

Most of the yield data is from reservoir surveys, for which estimating the percentage of the reported value representing bed material is more challenging, especially for instances where sediment size data are lacking. It was assumed, on the basis of the size composition of bed-material samples collected along the Umpqua River (this study) and the Chetco River (Wallick and others, 2010), that bed material is composed primarily of clasts with diameters greater than 0.5 mm (coarse sand and gravel) and that finer particles were transported primarily as suspended load. For the full reservoir surveys for which the data sources give no estimate of the portion representing bed material or indication of sediment‑size distributions (such as for reservoirs in the North Umpqua and Deschutes River basins), it was assumed that bed material composes 20 percent of the total volume. For the McKenzie River reservoirs, a bed-material percentage of 21 percent was estimated by Stillwater Sciences (2006). For the Clackamas River basin reservoirs, we assigned a bed-material percentage of 30 percent on the basis of reservoir sediment samples reported in Wampler (2004).

Several reservoir surveys focused on deltas formed at major points of inflow, allowing estimates of sediment volumes delivered by as many as three basins draining into a reservoir. For most of these surveys of the typically coarser deposits found at tributary deltas, 50 percent of the total volume was assumed to be bed material, based on sediment‑size analyses for the delta sediment accumulations in Lake Billy Chinook (Deschutes River; O’Connor and others, 2003) and Iron Gate Reservoir (PacifiCorp, 2004). For Detroit Lake, the volumes in the three surveyed reservoir arms were assumed to be 80 percent bed material, based on grain-size analyses conducted by Tetratech (2009).

Specific Bed-Material Yield

Analyses were performed with respect to specific bed-material yield (Y) in units of metric tons per square kilometer of contributing drainage basin area. For sediment accumulations originally reported as volumes, such as for all reservoir surveys, bed-material mass was calculated by multiplying the estimated bed-material volume by 2.1 metric tons/m³. Contributing drainage area was determined for each of the measurement sites through the U.S. Geological Survey (2010d) Streamstats website.

Predictor Variables

On the basis of previous analyses in the Deschutes River basin (O’Connor and others, 2003) and Hooke’s (2000) analysis, correlations between specific bed-material yield and predictor variables involving combinations of mean basin slope (S; in percent), drainage density (dl; in kilometer per square kilometer), and mean basin precipitation (P; in millimeters) were assessed. Mean basin slope was calculated from a slope raster map derived from the U.S. Geological Survey National Elevation Dataset 1/3-arc-second (approximately 10 m resolution) raster digital elevation data (obtained from the U.S. Geological Survey [2010c] The National Map Seamless Server ). Drainage density was calculated from the total stream length for each basin as mapped in the high-resolution (based on 1:24,000-scale topographic mapping) National Hydrologic Dataset (U.S. Geological Survey, 2010a). Mean annual basin precipitation for 1971–2000 was calculated from 30-arc-second (approximately 800-m resolution) gridded data provided by the PRISM Climate Group (2007) at Oregon State University.

Correlations

Specific correlations were made between the natural logarithm of specific bed-material yield and (1) mean basin slope, (2) the product of mean basin slope and mean annual basin precipitation, and (3) the product of mean basin slope and basin drainage density (fig. 43). The logarithmic transformation normalized the observations and enabled linear regression fits. Seven of the bed-material measurements were excluded from the correlation analyses: two exceptionally high values from the Redwood Creek basin, where sediment yield was substantially increased by land-use practices in the basin prior to the 1970s (Madej, 1995); and the five measurements from the H.J. Andrews Experimental Forest because of the very small size of the basins (all less than 1 km²) and their unrepresentative experimental treatments. The resulting exponential correlations between specific bed-material sediment yield (Y, in metric tons per square kilometer of drainage area) and three predictor variable sets are statistically significant, although the correlations are much stronger for S and S–P than for S∙dl:

(2)

(3)

(4)

Application to the Umpqua River

All three correlations between specific sediment yield and basin-scale predictor variables were applied to the Umpqua River basin. To examine spatial trends in sediment yield explicitly relative to river position, we subdivided the basin into 62 subbasins ranging from 0.16 to 3,190 km², accounting for major tributaries, dam locations, and bed-material sampling locations. For each of these subbasins, total bed-material yield was calculated using equations 2–4 as well as for the bounding 95-percent confidence limits for the regression (fig. 43). Total yield calculations allow for coarse predictions of bed-material flux to the South Umpqua River and continuing downstream along the main stem Umpqua River channel for the entire study area (fig. 44).

The 95-percent confidence intervals of the regressions imply that uncertainties range from 30 to 70 percent about the calculated specific sediment flux. For example, the predicted sediment flux for the South Umpqua River immediately upstream of the North Umpqua River confluence at FPKM 175.1 is 212,600 +140,600/–129,700 metric tons/yr for the regression based on the product of slope (eq. 1) and 70,000 +29,200/–20,800 metric tons/yr for the regression based on the product of slope and precipitation (eq. 2). These flux calculations also show that the regression based solely on mean basin slope (eq. 1) predicts approximately twice the bed‑material volume as those resulting from the slope–drainage density (eq. 3) and slope–precipitation (eq. 2) combinations. Although all three predictions of bed-material sediment yield to the Umpqua River exceed local transport capacity in most locations, on the basis of local particle size and bed-material transport equations (fig. 44; table 9), equations 3 and 4 provide mutually consistent bed-material sediment yield predictions more closely matching the transport capacity estimates. From these observations, we judge equation 3, based on the product of slope (S) and precipitation (P), to be most appropriate for estimating sediment delivery to the Umpqua River channel.

The predictions shown in figure 44 do not account for bed-material attrition by abrasion or for discontinuities in sediment yield and transport because of dams or other special circumstances. Bed-material attrition along the main stem Umpqua River resulting from fracture, abrasion, dissolution, and weathering and the resulting transformation of bed material to suspended load is accounted for in a similar manner as for the Chetco River analysis of Wallick and others (2010). For sediment generated from the Western Cascades and Klamath Mountains geomorphic provinces (fig. 1), we applied a mass loss rate of 0.51 percent per river kilometer, adopting the rate determined by Shaw and Kellerhals (1982; calculated from their estimate of a fraction-diameter reduction rate of 0.0017/km) for quartzites in natural rivers. For the much softer rocks produced from the Coast Range geomorphic province, we assigned a mass loss rate of 18 percent per kilometer on the basis of tumbler experiments with Tyee Formation sandstone clasts conducted by Benda and Dunne (1997). Although this rate seems high relative to that for quartzites, it is consistent with our observations that there were few Coast Range clasts in the gravel bars within reaches supplied by sedimentary rocks of the Coast Range. At the surface sampling sites within the lower reaches, many clasts of Paleogene sedimentary rocks have disintegrated in place (fig. 45) and typically composed less than 5 percent of the bar surface clasts despite these rocks accounting for 10–25 percent of the contributing drainage area at the sampling sites. The rapid attrition of Paleogene sedimentary rock clasts is also supported by the high suspended sediment loads in Coast Range streams (for example, Beschta, [1978]).

By contrast, the fraction of felsic intrusive rocks in surface and subsurface bar sediment samples is consistent with the overall fraction of the basin underlain by felsic intrusive rocks as well as the expected percentage of felsic intrusive clasts as predicted from the sediment yield analysis in conjunction with mapped areas of felsic intrusive rocks (Wells and others, 2001; Ma and others, 2009; table 9, fig. 46). The result of applying the abrasion rates proportionally to the volume of bed material delivered from the different terrains, as predicted from the bed-material sediment yield correlation of equation 3, indicates that the bed-material flux generally increases downstream to the North Umpqua confluence, but decreases downstream from the North Umpqua confluence as attrition reduces bed-material volume faster than it is replenished by downstream tributaries that furnish mainly soft clasts of Paleogene sedimentary rocks (fig. 47).

We further modified predictions of cumulative sediment yield to account for present river conditions by assuming no bed-material input from the dammed portions of Cow Creek and the North Umpqua River, and from the Smith River, of which the lower 6 km is estuarine as a consequence of Holocene sea-level rise, thereby inhibiting gravel transport to the Umpqua River (fig. 47). The influence of the Galesville Dam on Cow Creek is small, reducing the total predicted yield at the Cow Creek confluence with the South Umpqua River by 6 percent. The dams on the North Umpqua River likely have had larger effects on total bed-material yield to the Umpqua River: the calculated volume of bed-material yielded by the North Umpqua River in the absence of impoundments accounts for as much as 65 percent of the total bed material downstream of the confluence with the South Umpqua River. This analysis probably overestimates the historical contribution of the North Umpqua River because of the river’s smaller peak flows and the smaller area underlain by Klamath Mountains terrain rocks relative to the South Umpqua River, and neither factor is explicitly accounted for in the bed-material sediment yield model.

Combining the empirical bed-material sediment yield estimates with estimates of downriver attrition rates and the impoundments of Cow Creek, North Umpqua River, and Smith River gives an overall prediction of bed-material flux along the Umpqua and South Umpqua Rivers (fig. 47). Flux rates generally increase within the Days Creek and Roseburg reaches from approximately 13,400 metric tons/yr at the upstream end of the Days Creek reach to nearly 50,000 metric tons/yr within the Roseburg and Coast Range reaches. Downstream, predicted flux rates diminish as attrition exceeds input of bed material, gradually diminishing to 30,000–40,000 metric tons/yr at the entrance to the Tidal reach.

Summary of Bed-Material Characterization-and Transport

Measurements of bed-material sediment at 51 sites in the Umpqua River study area provide a basis for characterizing longitudinal patterns in bar texture and also support calculations of transport capacity (table 9). Surface particle sizes at 41 bars on the South Umpqua and main stem Umpqua Rivers indicate a general downstream fining of surface bar texture, with median grain size diminishing from about 100 mm in the Days Creek reach to less than 50 mm in the Coast Range reach (table 9; fig. 37). Subsurface grain sizes diminished much less, from median grain sizes of 20–40 mm in the Days Creek reach to about 15 mm in the Coast Range reach (fig. 37; table 9). Tributary bed material typically is finer than that in the main stem Umpqua River.

Armoring ratios computed for 25 sites on the main stem Umpqua and South Umpqua Rivers range between 1.08 and 4.73, with a mean value of 2.3, which is consistent with overall conditions of excess transport capacity relative to bed-material supply. Armoring ratios on tributaries generally are lower; a value of 1.0 computed for Cow Creek indicates that for that stream, sediment supply may approximate or exceed transport capacity.

Clast lithologies assessed at each of the sampling sites document the persistence of felsic intrusive rocks (which originate in Klamath Mountains terrain) in the bed material. The much softer sedimentary rocks of the Coast Range terrain were either scarce or absent at most sites, which is consistent with the greater abundance of gravel bars in the Days Creek and Roseburg reaches compared to downstream reaches and the overall conclusion that Klamath Mountains terrain rock types are a major source of bed material for the Umpqua River.

Calculated bedload transport capacity for WYs 1952–2008 varies markedly temporally and spatially among 44 computation sites on the South Umpqua River and Umpqua River, and a single site on Cow Creek (figs. 40 and 41). Variation among the four applied transport equations is due primarily to differences in surface and subsurface sediment‑size distributions among the sites and the minor differences in the equation forms. The largest discrepancies between the surface and subsurface-based equations are at armored bars, where the transport capacities computed by the subsurface-based equations can be 2–3 orders of magnitude greater than those computed using surface grain sizes (table 9).

Annual transport capacity calculated for individual sites ranges from essentially zero transport during low-flow years, such as 1977 and 2001, to as much as 600,000 metric tons/yr for high-flow years, such as 1964 and 1997. For the Umpqua Sand and Gravel Bar (FPKM 171.4), more than 50 percent of the total bed-material transport capacity for this 57-year period occurred over a total of just 80 days. There is also considerable spatial variation, as computed transport capacity values throughout the study area span more than six orders of magnitude. Armored and rarely mobilized sites such as Maupin Bar (FPKM 94.5) have little or no transport capacity in most years, whereas other highly mobile sites, such as Willis Creek Bar and Myrtle Creek Bridge Bar (FPKM 204 and 219.9), have calculated mean annual transport capacities of 100,000–300,000 metric tons. With the exception of the large bars at the mouth of Cow Creek, many of the largest bar complexes in the study area are in areas of decreasing transport capacity.

Most sites with high calculated annual transport capacities are small patches of gravel deposited in the lee of bedrock rapids (fig. 16) and likely are not representative of overall reach conditions. The wide-ranging variability in estimated transport capacities over time and between sites illustrates the highly nonlinear dependence of bed-material transport on key parameters of grain size, slope, and discharge. Sensitivity trials, combined with field observations, show that although some of the spatial variations in transport capacity may arise from uncertainties in calculation parameters, most of the variation for the Umpqua River system probably results from the diversity of bar types and the supply-limited character of the Umpqua River.

Considering the limitations of the equation-based approach to the Umpqua River system, these calculations probably are not a reliable means of estimating actual bedload fluxes; nonetheless, given the range of calculations and associated bar mobilities (figs. 40 and 42), we judge that the overall transport capacity along the South Umpqua and Umpqua Rivers is probably less than 25,000 metric tons/yr, a value consistent with the minimum bed-material transport rates of 2,810 to 30,600 metric tons during WYs 2002–08 determined from the site-specific gravel bar surveys.

A second approach to estimating rates of bed-material transport in the Umpqua River basin is an analysis of regional measurements of bed-material yields. This analysis indicates that basin attributes such as slope, drainage density, and precipitation correlate with bed-material yield. The correlation between bed-material yield and the product of basin slope and precipitation is most applicable to the Umpqua River basin. Application of this sediment-yield relation in conjunction with (1) estimates of particle attrition, and (2) the effects of dams and tidal influence, which markedly reduce or eliminate bed-material supply from the North Umpqua River, Smith River, and part of the Cow Creek basin, results in flux rates that generally increase in the Days Creek and Roseburg reaches, from approximately 13,400 metric tons/yr at the upstream end of the Days Creek reach to nearly 50,000 metric tons/yr within the Roseburg and Coast Range reaches. Downstream, predicted flux rates diminish as attrition exceeds input of bed material, gradually diminishing to 30,000–40,000 metric tons/yr at the entrance to the Tidal reach.

The bedload transport capacity estimates are broadly consistent with the bed-material yield analysis and site-specific surveys at instream gravel mining sites along the Roseburg and Days Creek reaches (tables 11 and 12) which indicate local bed-material flux rates of up to 30,600 metric tons/yr in high-flow years, but less than 10,000 metric tons/yr in more typical years.

First posted September 29, 2011