Summary and Conclusions

This study, done in cooperation with the U.S. Army Corps of Engineers, assessed spatial and temporal trends in channel change and bed-material transport for 350 km of alluvial and semi-alluvial river channel in the Umpqua River basin. Basin network structure and channel geomorphology led to subdivision of the river system into six contiguous analysis reaches. The North Umpqua reach includes 47 km of channel extending upstream of the North Umpqua River confluence with the South Umpqua River. The Days Creek reach encompasses 47 km of the South Umpqua River from the upstream extent of the study area near Tiller, Oregon, to the Cow Creek confluence. The Roseburg reach continues 76 km downstream of Cow Creek to the South Umpqua River confluence with the North Umpqua River. The Garden Valley reach contains the Umpqua River for the 19 km from the confluence of the North Umpqua and South Umpqua Rivers to the entrance of the Coast Range, from where the Coast Range reach of the Umpqua River extends another 116 km downstream to the head of tide near Scottsburg, Oregon. The much lower gradient and partly estuarine Tidal reach encompasses the final 45 km of river channel and through Winchester Bay to the Pacific Ocean at Reedsport. These reaches have distinct physiographic and bed-material transport conditions, as well as distinct histories of instream gravel mining, thereby providing an efficient analysis and discussion framework.

The findings reported here draw largely upon two components: (1) historical analyses, including detailed mapping of the active channel using aerial photographs and repeat surveys, to document spatial and temporal changes in channel morphology and bed-material storage and (2) quantitative investigation of the bed-material flux through the study reaches. These analyses provide a basis for understanding the recent history of the active channel and also allow for inferences regarding the spatial and temporal variation of production, fluxes, and routing of bed material through the study reaches.

Primary Findings

The overall character of the Umpqua River reflects its geologic history. For the past 10,000 years, the overall trend for fluvial reaches of the Umpqua River has been incision, where transport capacity has exceeded the supply of coarse bed-material sediment, as indicated by abundance of exposed bedrock in and flanking the active channel throughout the study area. This channel characteristic, as well as the sparse gravel cover, was specifically noted by 19th and 20th-century Euro-American explorers. Repeat mapping from multiple aerial-photograph sets spanning 1939–2009 shows that the fluvial reaches of the Umpqua, South Umpqua, and North Umpqua Rivers flow within largely stable, single-thread channels of bedrock or coarse boulder and cobble substrates. Coarse bed-material sediment locally mantles the bedrock, forming shallow bars in and flanking the low-flow channel, whose position and overall size are dictated primarily by valley geometry rather than channel migration processes.

Gravel bars have historically been most abundant on the South Umpqua River within the Roseburg and Days Creek reaches, where there has been as much as 1.3–4 times the area of gravel bars per unit length of stream (approximately 12.7–31.8 m²/m) compared with the Coast Range and Garden Valley reaches (where specific bar area has ranged from 5.0 to 13.7 m²/m). Although bedrock rapids and channel-flanking bedrock shoals are common features throughout the study area, they are most abundant along the Umpqua and North Umpqua Rivers, where 2005 aerial photographs show 3–5 times more exposed bedrock (by area) than mapped gravel. Most of the gravel in the study area is stored in large bars with areas greater than 20,000 m², many of which apparently become active areas of bed-material transport only during exceptionally large floods, such as the December 1964 flood. Although many numerous smaller gravel patches (less than 2,000 m²) flank the river at the heads of rapids and immediately downstream, these smaller depositional zones account for less than 6 percent of the total mapped gravel in the study area in 2005.

The abundance of gravel bars along the lower South Umpqua River most likely results from Klamath Mountains source areas underlying much of the South Umpqua River basin. The tectonically deformed and metamorphosed Mesozoic rocks of the Klamath Mountains terrain, together with its steep slopes and dense stream network, enhance production and delivery of bed material to the South Umpqua River. High bed-material fluxes from this terrain have been documented for the Chetco River (Wallick and others, 2010) and Smith River (MFG, Inc. and others, 2006) to the south. Additionally, clasts from this terrain are probably more resistant to abrasion than bed material from the High Cascade and Western Cascade terrains, and consequently are a persistent component of Umpqua River bed material as far downstream as the Tidal reach. Cow Creek, a large tributary draining Klamath Mountains terrain, probably is a major supplier of gravel to the South Umpqua River, judging from the extensive gravel deposits near its mouth, the increased abundance of bars downstream of its confluence, and the low armoring ratio of bars within Cow Creek. Historically, several Klamath Mountains tributaries, including Cow Creek, Myrtle Creek, and Lookingglass Creek, were subject to extensive placer mining, which may have further enhanced sediment output from these streams, although historical photographs of the Roseburg reach do not indicate significantly greater gravel volumes during the early 20th century.

Bed-material sediment from Cascade Range streams originates mainly in the Western Cascades, because the much younger lava flows of the High Cascades are highly porous and have little capacity for sediment transport. Although the Western Cascades terrain yields more bed-material sediment than the High Cascades terrain, sediment production from the Western Cascades probably is small compared to that from Klamath Mountains terrain as evidenced by: (1) the North Umpqua reach, which exclusively drains the Cascade Range, had less than one-half of the gravel bars per unit stream length in 1939 than the South Umpqua reaches, and (2) the South Umpqua River upstream of the Days Creek reach drains only the Western Cascades terrain, and unlike the more gravel-rich lower reaches downstream of Klamath Mountains terrain tributaries, the river within the Western Cascades terrain is a narrow bedrock stream with boulder-dominated rapids and few gravel bars.

Farther downstream of its confluence with the South Umpqua River, sedimentary rocks supplied to the main stem Umpqua River by Coast Range tributaries are highly erodible, and although this region produces high suspended-sediment loads (for example, Beschta [1978]), bed-material clasts from these geologic units disintegrate readily. Therefore, although the terrain of Klamath Mountains comprises only 21 percent of the Umpqua River basin, it probably supplies a disproportionately large amount of bed-material sediment to the channel system. Further, the importance of this terrain to total basinwide sediment production is even larger because of the effects of dams on sediment transport in the North Umpqua River.

The primary observation from the repeat channel mapping and surveys is the overall stability of the Umpqua River planform. All fluvial reaches showed little change in sinuosity or channel width throughout the 70-year analysis timeframe, mainly because of lateral and vertical bedrock control. Consistent with this, repeat stage measurements at USGS streamflow-gaging stations show only local areas of slight channel deepening (on the order of 0.1–0.2 m), some of which may be associated with bedrock erosion.

The main temporal trend evident from repeat channel mapping from aerial photographs is a 29-percent decrease in the area of mapped gravel bars between 1939 and 2005. Most of this decrease was between 1967 and 2005, and was partly due to vegetation colonization on formerly active, upper bar surfaces, converting some of these high bar surfaces to floodplain. Also important was erosion of lower elevation bars to bedrock, particularly for the Coast Range and Garden Valley reaches. The decrease in mapped gravel bar area probably resulted from a combination of factors, including decreasing peak flows, gravel extraction, and dam construction. Several unregulated tributary streams, as well as the South Umpqua River gaging station at Brockway, show significant trends of decreasing peak flows since the 1950s, which is probably due mainly to decadal-scale climate cycles. Because three of the five streams that show this trend drain Klamath Mountains terrain, even small decreases in peak flows on these tributaries may have a disproportionate effect on overall gravel transport in the study area. The cumulative effects of instream gravel extraction in recent decades likely also affects bed-material storage in the active channel because mined volumes in some years probably constituted a substantial portion of the overall gravel flux.

For the North Umpqua River, the 59-percent decrease in gravel between 1967 and 2005 is probably due to a combination of trapping of bed material by hydropower dams constructed in 1952–55 and climate-driven decreases in peak flows, as detected for the gaging station at Winchester. For this reach, decreased gravel bar area has led to much more exposure of active channel bedrock.

Although the overall trend was of decreasing bar area, many bars have episodically grown, mainly as a consequence of large floods. This is particularly the case for the major flood of December 1964, which had an annual exceedance probability of about 1 percent. Total bar area throughout the fluvial reaches increased by more than 11 percent between 1939 and 1967, which probably is attributable mainly to the 1964 flood. Evident in the 1967 photographs are (1) removal of vegetation and bed-material deposition on upper bar surfaces, and (2) bed-material deposition extending the margins of low-elevation bars into areas mapped as water in 1939. Later but smaller floods in December 1996 and December 2005 resulted in smaller increases in bar area, but these increases have been offset by erosion and bar diminishment during intervening and subsequent years.

Bed-material flux was estimated for fluvial reaches of the Umpqua and South Umpqua Rivers by two independent approaches, supplemented by bed-material recruitment measurements at six sites of past gravel mining and by earlier measurements of suspended-sediment transport. Bed‑material transport capacity estimates at 44 sites throughout the South Umpqua and main stem Umpqua Rivers for the period 1951–2008 result in transport capacity estimates that vary spatially and temporally. The temporal variations relate to flow history, with most transport associated with large peak flows. The diminishment in peak discharges over the last three decades, at least partly to climate cycles, has led to an overall temporal trend of reduced gravel transport.

The even wider spatial variations in calculated bed-material transport rates reflect the more fundamental difficulty of applying equations of bed-material transport capacity to a supply-limited river, where bar textures chiefly reflect local hydraulics rather than reach-scale supply conditions. Nevertheless, the transport capacity values should provide an indication of maximum possible bed-material transport rates; reach-averaged median transport capacity values calculated by the bed-material surface-based equations of Parker (1990a, 1990b) and Wilcock–Crowe (2003) equations for 1951–2008 yields a transport capacity of 7,000–27,000 metric tons/yr for the South Umpqua River and 20,000–81,000 metric tons/yr for the main stem Umpqua River upstream of the head of tide (tables 11 and 12). The values of bed-material transport capacity values for the intermediate mobility sites, generally ranging between 500 and 20,000 metric tons/yr as predicted by the Parker (1990a, 1990b) equation, may be the best estimate for actual bed-material transport rates, although confidence in this assessment would be bolstered substantially by actual transport measurements.

These estimates of bed-material transport capacity are broadly consistent with an empirical bed-material yield analysis developed from regional bed-material transport measurements. The most satisfactory regional relation predicts bed-material yield as a function of source area slope and precipitation (fig. 43). Adopting this relation in conjunction with estimates of in-channel attrition, results in predicted bed-material fluxes of as much as 25,000 metric tons/yr on the Days Creek reach, increasing to nearly 50,000 metric tons for the Roseburg, Garden Valley, and Coast Range reaches, but then decreasing to approximately 30,000 metric tons/yr at the entrance to the Tidal reach.

Both of these approaches—the transport capacity estimates and the regional bed-material sediment yield analysis—give results consistent with site surveys at individual bars within the Days Creek and Roseburg reaches. These surveys indicate minimum local bed-material flux rates of up to 30,600 metric tons/yr in high-flow years, but more typically less than 10,000 metric tons/yr.

The two approaches adopted by this study give estimates less than those predicted by Curtiss (1975) from suspended-sediment transport measurements made during 1956–1973. By applying an assumed bedload transport ratio to measured suspended-sediment loads, the Curtiss (1975) analysis predicts bedload transport rates of 8,400 metric tons/yr at Tiller, near the upstream end of the Days Creek reach at FPKM 273; 46,000 metric tons/yr at the Brockway streamflow measurement site on the South Umpqua River within the Roseburg reach near FPKM 195.3; and 160,000 metric tons/yr at the Elkton measurement site on the main stem Umpqua River in the Coast Range reach at FPKM 72.1. Although these bed-material transport values for the Days Creek and South Umpqua reaches are slightly higher than those we infer from the sediment yield and capacity analyses, they are within realistic uncertainty bounds. The estimate of 160,000 metric tons/yr of bedload at the Elkton measurement site on the main stem Umpqua River greatly exceeds likely bed-material transport rates for this reach as estimated from the capacity and yield analyses, and is almost certainly high as a result of substantially elevated suspended loads derived from Coast Range sedimentary rocks, which produce little bed material.

In consideration of all these analyses, together with the depositional volumes measured by individual gravel bar surveys, we judge that the actual bedload flux in most years is probably less than 25,000 metric tons/yr in the Days Creek and Roseburg reaches, although Cow Creek probably adds substantial bed material to the South Umpqua River at its confluence. Bed-material transport in the Garden Valley and Coast Range reaches may be similar or slightly less because of bed-material attrition exceeding tributary addition. For comparison, the estimated annual volume of commercial gravel extraction from the South Umpqua River was 9,260 metric tons in 2001, 610 metric tons in 2003, and 36,570 metric tons in 2004, based on data supplied by the two main operators in the South Umpqua River—which indicates that historical instream gravel extraction may have been a substantial fraction of the total bed-material flux in the Umpqua River system.

The Tidal reach has a distinctly different morphologic character and transport regime. The Umpqua River along the Tidal reach contains the largest bars in the study area, particularly at the expansive valley bottom near the confluence of the Smith River. These bars are mainly composed of sand and mud, contrasting with the gravel bars upstream. Commercial dredging has historically focused on the section between the Smith River confluence and upstream to the head of tide at FPKM 40, where there are few bars and repeat surveys show persistent channel deepening even in areas that had not been mined for several years.

Like other coastal streams in Oregon, the lower Umpqua River has been strongly affected by the 130 m of sea-level rise after the culmination of the last maximum glacial period 18,000 years ago, resulting in long-term aggradation and trapping of bed material and suspended coarse sand transported from upstream. Consequently, it is unlikely that substantial bed material from the upstream fluvial reaches (and the upper Smith River) is transported into the Pacific Ocean. The long Tidal reach (and lack of graded profile to the Pacific Ocean mouth) is evidence that upstream sediment supply has not kept pace with Holocene sea level rise inundating the lower Umpqua River valley.

The sediment yield analysis indicates that about 30,000–40,000 metric tons of bed-material sediment enters the Tidal reach annually, but bed-material accumulation within the lower Tidal reach may be substantially greater, because much of the sand transported in suspension upstream is likely transported as bedload in the Tidal reach due to the lower gradients. Consequently, while annual commercial instream mining averaged 140,000 m³ annually during 1949–2002, this volume is not indicative of bed-material transport rates in the upstream fluvial reaches because much of this material probably entered the Tidal reach as sand transported as suspended load from the upstream reaches.

Implications Regarding Future Trends and Monitoring Strategies

For a mixed bedrock and alluvial river such as the Umpqua River, the physical character of the channel is mainly the result of its geologic history and physiography. Throughout the Holocene, transport capacity has exceeded the supply of bed-material sediment, causing the Umpqua River to incise through Pleistocene valley fill and bedrock, resulting in a modern channel that flows mostly on bedrock. The character of individual bars is highly variable and depends on the history of flow and sediment transport, time lags involved in eroding and redepositing sediment, and other local and drainage-basin-scale disturbances that might affect the channel directly or indirectly.

Although many factors influence the abundance and character of Umpqua River gravel bars, the decreases in bar areas observed on all of the fluvial reaches between 1967 and 2005 will likely continue if future gravel removal exceeds bed-material influx. Continued decreases in bar area may also be accompanied by the coarsening of low-elevation active bars that currently have low armoring ratios. In the absence of future mining, bar building will probably be greatest in the lower Days Creek reach and throughout the Roseburg reach, as this area has historically had the greatest concentration of gravel bars because of the high influx of bed-material sediment from tributaries draining the Klamath Mountains terrain.

However, even prior to gravel extraction and dam construction, transport capacities throughout the Umpqua River study area were much greater than sediment supply, so bar building may proceed slowly following cessation of gravel extraction, and the rate of bar growth will depend on the timing and magnitude of peak flows and the sediment influx accompanying these floods. Although gravel augmentation on the North Umpqua River began in 2004 (Stillwater Sciences, 2006), this additional gravel is not likely to have a substantial effect on bar area in the lower North Umpqua River and main stem Umpqua River because the total augmentation volume is small relative to the long-term gravel deficit introduced by the hydropower dams (based on data provided by PacifiCorp, 2002 and Stillwater Sciences, 2006).

To better understand variation in bed-material storage under different management scenarios, actual bed-material influx to the Umpqua River study reaches must be accurately quantified. However, it is difficult to characterize bed-material fluxes in gravel-rich settings, and even more so for the supply-limited Umpqua River, where bar characteristics and sediment transport are highly variable. Improving our understanding of bed-material fluxes on the Umpqua River will require a variety of independent methodologies, bolstered by high-resolution datasets. The approaches that will potentially be most useful for future characterization of bed-material storage in the Umpqua River study area include (1) the application of transport capacity equations, similar to the methodology used here, but updated using a detailed hydraulic model and up-to-date bar texture information, and (2) direct measurements of bedload transport, which could be difficult to obtain and interpret, but as part of a sustained monitoring program would significantly aid in characterizing bed-material fluxes across a range of flows.

A detailed hydraulic model, along with several key datasets, would support these approaches and form the basis for a future adaptive management program. The nonlinear response of calculated transport capacities to variation in grain size and slope indicates the need for accurate, detailed data describing Umpqua River bar textures and hydraulics. The hydraulic model encompassing the South Umpqua and Umpqua Rivers above the head of tide could be developed from LIDAR topography, and bathymetric surveys would provide a more accurate method of calculating energy slope under a variety of discharge scenarios. At a minimum, the modeling approach would entail a 1D hydrodynamic model with closely spaced cross sections to characterize the highly variable channel. Ideally, the approach would entail a blend of both 1D and 2D models so that the complex hydraulics at large key bars in sharp bends (such as Maupin Bar and Days Creek Bar) are accurately characterized. Such a modeling framework could be used to more accurately calculate energy slope under a variety of discharge scenarios, enabling better understanding of transport conditions and refining our overall understanding of longitudinal patterns in bed-material transport. A detailed hydraulic model, if coupled with a spatially discrete sediment transport model, could also be used to simulate morphological changes to the channel bed under different management and flood scenarios.

If repeated at regular intervals or following large floods, the LIDAR and bathymetric surveys underlying the hydraulic model would provide a comprehensive basis for evaluating future changes to channel morphology and bar topography. Such data, combined with detailed measurements of bar thickness, could also be used to calculate volumetric sediment flux and deposition rates throughout the study area (similar to a morphology-based approach applied on alluvial rivers). Because the channel is primarily underlain by bedrock and is in many places shallow, the bathymetric survey could consist of depth soundings along the centerline of deep pools, as the LIDAR acquired at low flows would provide an adequate approximation of bed elevation in rapids. Future monitoring could also incorporate sampling of bar textures at regular intervals (perhaps every 2–5 years, or following a flood of a certain magnitude) in order to improve our transport capacity calculations. Textural information, combined with repeat mapping of vegetation densities from aerial photographs would also aid in evaluating temporal evolution of bars in response to different management scenarios and floods.

First posted September 29, 2011