Study Area

The study area includes the lower 27.3 km of the Middle Fork Willamette River below Dexter Dam and the lower 11.5 km of Fall Creek downstream from Fall Creek Dam, as well as the 10-km long Fall Creek Lake (fig. A6). Downstream from Dexter Dam, the Middle Fork Willamette River is predominantly an alluvial, gravel-bed river, whereas Fall Creek downstream from Fall Creek Dam flows over bedrock and alluvium, reflecting relatively greater valley confinement along Fall Creek compared to the Middle Fork Willamette River. Both river corridors are flanked by broad floodplains and locally impinge upon valley walls composed of bedrock or early Quaternary terrace gravels. The width of the primary active channel is typically greater than 100 m along the Middle Fork Willamette River with an average gradient of about 0.0024 m/m. Fall Creek downstream from Fall Creek Dam is steeper and narrower, with typical active channel width of about 20–60 m and average gradient of about 0.0027 m/m. Fall Creek is a predominantly straight, single-thread channel for 5 km below Fall Creek Dam, then adopts a more sinuous planform, with intermittent secondary channel features formed prior to streamflow regulation. The Middle Fork Willamette River below Dexter Dam has sinuous, single thread segments, but some segments of the active channel and floodplain contain numerous active gravel bars, secondary channels, and islands, so that the overall planform of the Middle Fork Willamette River in the study area is similar to a “wandering gravel-bed river” (Church, 1983).

Like the Willamette River and other major gravel-bed rivers of the Willamette Valley, the river channels and floodplains of both Fall Creek and Middle Fork Willamette River have fundamentally changed since Euro-Americans first settled in the Willamette Valley in the mid-19th century (for example, Sedell and Froggatt, 1984; Dykaar, 2005, 2008; Gregory and others, 2002a, 201954; Wallick and others, 2007, 2013159; Keith and others, 2023a). Flood control, bank stabilization, large-wood removal, conversion of riparian forests to agriculture, and other large-scale alterations have substantially changed the basin’s streamflow, sediment, and large-wood regimes, resulting in narrower floodplain corridors and less complex assemblage of landforms in present-day floodplains compared with conditions in the mid-19th century (Wallick and others, 2013; Keith and others, 2023a). The underlying geology and bank stabilization structures constructed by USACE and local partners in the mid-20th century (Gregory and others, 2002b) both affect bank erodibility and resulting patterns of bank erosion, sediment exchange between the channel and floodplain, and overall channel morphology on Fall Creek and Middle Fork Willamette River (Wallick and others, 2006, 2013; Keith and others, 2023a; fig. A6). Prior to dam construction, a 10.5 km-reach on the Middle Fork Willamette River below Dexter Dam evolved through the development of large meander loops as islands coalesced, avulsions that created and abandoned channels, and rapid gravel bar growth in newly created channels (Dykaar, 2005, 2008); by 2004, river channel length and avulsions had decreased, exposed gravel patches have been reduced by 70 percent, and island area decreased by 57 percent (Dykaar, 2005, 2008).

Dividing the study area into reaches (fig. A6; table A1) provides a framework for systematically evaluating the geomorphology of the river corridors and their responses to sediment releases from the Fall Creek streambed drawdowns. The study area is divided into seven reaches based on channel morphology, slope, and location of major tributaries (fig. A6; table A1).

Fall Creek

Fall Creek Lake (fig. A8A) encompasses the 10-km-long reservoir reach upstream from Fall Creek Dam inundated at maximum pool (elevation 830 ft). The main channels of the reservoir flow over variable substrate including bedrock, gravel, and finer materials. The average water-surface slope for the Fall Creek channel through the reservoir from WY 2012 lidar (Watershed Sciences, Inc., 2012; topographic lidar collected during full streambed drawdown) is about 0.0038, with the sections above and below minimum conservation pool elevation being 0.0044 and 0.0030, respectively. Even steeper is the Winberry Creek arm below minimum conservation pool at 0.0063.

Upper Fall Creek (fig. A8B) is a mixed-bed reach with bedrock and coarse alluvium and encompasses 5.9 km of Fall Creek between base of Fall Creek Dam to the confluence of Little Fall Creek. The reach was historically, laterally stable, likely due to local bedrock controls, relatively smaller basin, and low sediment supply (O’Connor and others, 2014; Keith and others, 2023a). Reductions in peak streamflows and sediment supply since upstream dam construction have resulted in less pronounced geomorphic transformation in this reach (related to the historical stability) relative to other study reaches (Keith and others, 2023a). Presently (during the period of this study), Fall Creek streamflows on sections of bedrock and alluvium. Volcanic and volcaniclastic rocks bound the channel at river kilometer (rkm) 10.5–10.4. A series of steep and flat sections are likely controlled by bedrock occurrence but overall, this is the second steepest (0.0029) and least sinuous (1.08) reach within the study area. The average primary active channel width is about 37 m, although it ranges from about 20 to 180 m.

Lower Fall Creek (fig. A8C) is an alluvial reach with sections flowing on bedrock and includes the lowermost 5.6 km of Fall Creek from the confluence with Little Fall Creek to mouth of Fall Creek. Like Upper Fall Creek reach, the Lower Fall Creek reach primarily was historically stable (Keith and others, 2023a); however, additional sediment supply and streamflow from Little Fall Creek likely supports the presence of the gravel bars in this reach. The channel bed in this reach is composed of sections of bedrock and alluvium. Volcanic and volcaniclastic rocks provide lateral control the channel at rkm 2.6–2.4 and 1.1–0.0. This reach has a similar gradient (0.0026) to the Dexter reach and the greatest sinuosity (1.36) in the study area. Primary active channel widths are similar to that in the Upper Fall Creek reach (average of 40 m; range from about 20 to 125 m).

Geology and Sediment Supply

The Fall Creek Basin is almost entirely underlain by the steeply dissected, less permeable Tertiary volcanic rocks of the Western Cascades (Tague and Grant, 2004; fig. A5); in contrast, the Middle Fork Willamette River Basin also drains Western Cascades (66 percent), and there is also a substantial contributing area (20 percent) underlain by highly permeable Pliocene and Quaternary lava flows of the High Cascades (Tague and Grant, 2004; fig. A5). Within the Western Cascades, Eocene to Miocene volcanic and volcaniclastic rocks of the Little Butte Volcanics are dominant (Peck and others, 1964; Sherrod, 1991; Smith and Roe, 2015). The Holocene floodplains of Fall Creek and Middle Fork Willamette River downstream from the Fall Creek Lake and Dexter Reservoir (respectively), are bordered by Pliocene to Pleistocene terrace gravels, upper Pleistocene Missoula Flood deposits, and Western Cascades volcanic rocks (O’Connor and others, 2001).

The geology and steep terrain underlying the Western Cascades historically supported high rates of bed-material sediment entering the river corridors of the study area. Landslides (Lyons and Beschta, 1983) and forest fires (U.S. Forest Service [USFS], 1995) influence sediment supply to Middle Fork Willamette River and Fall Creek. Processes that may have influenced sediment supply to Fall Creek Lake include clearcutting that began in the 1940s and peaked in the 1960s (USFS, 1995) and forest fires (Rakestraw and Rakestraw, 1991; USFS, 1995; USGS, 2018a). However, dams prevent much of this sediment from reaching the lower river corridors of the study area (O’Connor and others, 2014). A regional model for bed-material loads in Western Oregon provides estimates of 23,000 t/year of coarse sediment input to Fall Creek Lake and 82,000 t/year coarse sediment to the Lookout Point Lake and Dexter Reservoir and assumes a 0.69 coefficient of variation determined from Monte Carlo assessment that approximates the standard error (O’Connor and others, 2014; O’Connor and others, 2021). Similarly, regional models of suspended-sediment loads representative of WY 2012 conditions suggest about 1,300 t (standard error of 1,700 t) of fine sediment is entering the Fall Creek annually from Fall and Winberry Creeks, and that 62,000 t (standard error of 31,000 t) of fine sediment from the Middle Fork Willamette River (excluding reservoir tributaries) is entering Lookout Point Lake (Wise, 2018; O’Connor and others, 2021). Since construction in the mid-20th century, USACE dams in the Middle Fork Willamette River Basin (Hills Creek, Lookout Point, Dexter, and Fall Creek Dams; fig. A5) trap nearly all coarse sediment entering the lower reaches of Fall Creek and Middle Fork Willamette River and reduce the amount of fine-grained sediment transport. Reduction in the amount of bedload due to dams is about 95 percent at the mouth of Fall Creek and about 94 percent near the mouth of the Middle Fork Willamette River (O’Connor and others, 2014). Sediment trapping efficiency for fine-grained sediment at Fall Creek Lake calculated from measurements 1 month prior to the WY 2013 drawdown is about 46 percent (updated estimate from Schenk and Bragg, 2014, based on corrected sediment inflows to the reservoir).

Hydrology and Streamflow Regulation by Dams

Major tributaries within the study area (fig. A6) include Winberry Creek (rkm 12.6), which joins Fall Creek just upstream from Fall Creek Dam and forms the primary southern arm within Fall Creek Lake, and Little Fall Creek (rkm 5.7). Lost, Rattlesnake, Hills, Wallace, and Pudding Creeks all enter the Middle Fork Willamette River between rkm 21.4 and 5.5. Additionally, since 1852, the Mill Race (fig. A6) diverted streamflow from the Middle Fork Willamette River near Clearwater Park (rkm 6.3) to the city of Springfield for grist and lumber mills. Restoration efforts on the Mill Race have been completed, and hydraulic modeling used 300 ft³/s as a “normal” streamflow condition for the Mill Race (Schall, 2017). Other minor withdrawals from Fall Creek and Middle Fork Willamette River for irrigation and other purposes were not quantified for this study.

Three USGS streamflow-gaging stations within the study area (table A2; figs. A5, A6) have recorded streamflow since WY 1936 on Fall Creek (14151000; rkm 10.1; herein referred to as Fall Creek streamgage) since WY 1947 near Dexter on the Middle Fork Willamette River (14150000; rkm 22.5; herein referred to as Dexter streamgage) and since WY 1906 at Jasper (14152000; rkm 13.1; herein referred to as Jasper streamgage) on the Middle Fork Willamette River (table A2; figs. A5, A6; USGS, 2021). Other streamgages referenced for this study are summarized in table A2.

The USACE operates three major multi-purpose reservoirs and one re-regulating reservoir within the Middle Fork Willamette River Basin (figs. A5, A6) that have altered streamflow to downstream river reaches since 1953. Lookout Point Dam (84 m-tall), Hills Creek Dam (93 m-tall), and Fall Creek Dam (55 m-tall) operate primarily as flood-control reservoirs that capture high streamflows during winter months, reducing peak streamflows for downstream communities, then release stored water during spring and summer months to support other authorized purposes. As a result of dam operations, summer streamflows are typically higher than they were in pre-dam periods (Gregory and others, 2007). Dexter Dam (35.7 m-tall) re-regulates power-generating water releases from Lookout Point Dam. Each dam is also operated to meet multiple seasonally varying streamflow objectives described in the 2008 Biological Opinion (BiOp) for the WVP (NMFS, 2008). Typical mean daily streamflows (fig. A9; since WY 2008 following implementation of the BiOp) vary from 1,200 to 8,000 ft³/s at the Dexter streamgage on the Middle Fork Willamette River, though greater streamflows regularly occur in the winter and spring months. Since dam construction, baseflows at the Fall Creek streamgage during July and August are typically 50–300 ft³/s, but short-term operational releases have exceeded 3,000 ft³/s since WY 2008. In the autumn and winter months from October to March, Fall Creek daily streamflows are typically greater than 500 ft³/s with peak streamflows from 2,000 to 4,000 ft³/s, and a maximum peak streamflow of 5,820 ft³/s (April 10, 2019; USGS, 2021).

The river corridors downstream from Dexter and Fall Creek Dams have had substantial reductions in peak streamflows since the dams became fully operational in the mid-19th century. Prior to the construction of Fall Creek Dam in 1965, the 0.5 annual exceedance probability flood (often referred to as the 2-year recurrence interval) at the Fall Creek streamgage was about 10,300 ft³/s and the peak of record was 24,700 ft³/s (WY 1957; table A2; fig. A10). Since dam construction, flows at the streamgage had not exceeded 5,000 ft³/s during dam operations until the WY 2019 peak-flow event. At the Dexter streamgage on the Middle Fork Willamette River, the peak of record prior to streamflow regulation is 62,200 ft³/s (WY 1953; table A2; fig. A10), and the peak streamflow for the regulated period (beginning in 1961 after Hills Creek, Lookout Point, and Dexter Dams were fully operational) is 29,500 ft³/s (WY 1965; table A2; fig. A10). Farther downstream at the Jasper streamgage, where peak streamflows have been regulated by all four dams since 1965 (WY 1966), pre- and post-flow regulation period maximum annual peak streamflows were 94,000 ft³/s (WY 1910) and 24,400 ft³/s (WY 2019), respectively (table A2; fig. A10). Gregory and others (2007) calculated a 50 percent reduction in the 0.5 annual exceedance probability streamflow for the Middle Fork Willamette River at the Jasper streamgage.

In addition to the streambed drawdown operation, dams in the Middle Fork Willamette River and Fall Creek Basins are managed to support a variety of other biological objectives for recovery of spring Chinook salmon, including streamflows on the main-stem Willamette River that may benefit winter steelhead (NMFS, 2008). The 2008 BiOp was a primary driver of streamflows in spring and summer during this study, but other environmental streamflow programs were in effect in the Willamette River Basin. In 2002, The Nature Conservancy (TNC) and USACE created the Sustainable Rivers Project (SRP) to identify and implement streamflows at dams that would be ecologically beneficial for downstream aquatic species and riparian vegetation communities while continuing to meet congressionally authorized purposes at those dams (Warner and others, 2014). For the Middle Fork Willamette River, “The ecological goals include re-establishment of physical processes creating and connecting in-channel and off-channel habitats, recruitment of cottonwood on floodplains, promoting salmon migration, and mitigating thermal impacts of the dam releases” (Konrad, 2010). Streamflow recommendations and targets were developed to align with those ecological goals (Gregory and others, 2007). Implementation of SRP streamflow recommendations at Lookout Point and Dexter Dams on the Middle Fork Willamette River began in 2015 and have been evaluated by Jones and others (2016) and White and others (2023). Comparison of measured streamflows (2008–2022) to SRP target streamflows reveals that small autumn and spring pulse events were met in most years, whereas baseflows were exceeded every year, and winter bankfull and flood events were not achieved (White and others, 2023).

Dam Infrastructure and Streambed Drawdown Operations at Fall Creek Dam

Characteristics of the Fall Creek Dam infrastructure and operations are pertinent to understanding geomorphic evolution of the reservoir and downstream reaches in response to annual streambed drawdowns. Fall Creek Dam is a rock fill dam with concrete spillway and three sets of three nested fish horns at different elevations that pass water and fish downstream (fig. A7; U.S. Army Corps of Engineers [USACE], 2015; Hansen and others, 2017). Unique to this infrastructure are two regulating outlets (1.7 m wide by 3.0 m high) at the base of the dam. The 55-m-high, 1,554-m-long dam is operated primarily for flood risk management, and secondarily for water quality, irrigation, recreation, and habitat, and has a storage capacity of 115,000 acre-feet (USACE, 2018). Flood-control season lasts from November to March where pool levels are typically held elevations of 728 ft, whereas conservation season from April to November is restricted to a maximum elevation of 830 ft (USACE, 2018; fig. A11). More typically, pool levels are about 758 ft from November to January and near maximum pool levels of 830 ft from May to August (Hansen and others, 2017).

During the streambed drawdowns, Fall Creek essentially becomes a free-flowing river through the lower regulating outlet. Draining of the reservoir for short periods to support downstream fish passage through the regulating outlets was assessed for multiple years (Smith and Korn, 1970; Downey and Smith, 1992) prior to implementation of annual streambed drawdowns that began in WY 2012. In 2008, the Biological Opinion (BiOp) for the USACE’s WVP mandated a series of improvements to address fish passage and other adverse effects of the dams (NMFS, 2008). The BiOp identified Reasonable and Prudent Actions (RPAs) for Fall Creek Dam to improve the downstream passage and survival of juvenile spring Chinook salmon including drawing down lake levels to at least 714 ft elevation (NGVD 1929) or lower (fig. A7) and building a new adult fish collection facility downstream from the dam (NMFS, 2008). In 2011 (during WY 2012), the USACE began annual implementation of a full streambed drawdowns, lowering the reservoir pool elevations to 680 ft, resulting in successful downstream fish passage and survival (Nesbit and others, 2014; Northwest Fisheries Science Center, 2015; Hansen and others, 2017). Nesbit and others (2014) assessed fish passage during the WY 2013 drawdown, finding 95 percent of juvenile spring Chinook salmon migrated out of the reservoir between elevations 720 and 700 ft. Evaluation of fish capture data from the rotary screw trap downstream from the Fall Creek Dam, suggest that drawing down the lake fully to streambed also helped to expel warm-water, invasive species, which predate upon and compete with native species in the lake (Murphy and others, 2019c). By 2015, streambed drawdowns became part of the standard operations (G. Taylor, USACE, written commun., October 2021).

Pool levels at Fall Creek Lake have been lowered to elevations near streambed multiple times since dam construction (fig. A12). Between the late 1960s and 1990s, Fall Creek Lake has been lowered well below minimum conservation pool (728 ft) to an elevation 700 ft (or lower) nearly 40 times. After a pause in drawdowns during WY 2000–11, annual streambed drawdowns were resumed in WY 2012 under the 2008 BiOp (fig. A12). As evident from Fall Creek Lake level records (USGS streamgage 14150900, Fall Creek Lake near Lowell, Oregon; USACE, 2017; fig. A12), some of these drawdown operation resulted in pool elevations that reached streambed (680 ft; WY 1969–75, 1977, 1982) or were within about 3 m (10 ft) of streambed (690 ft; WY 1976, 1978–80, 1983–84, 1990, 1992)—an elevation below which recent drawdown activities have been observed to mobilize large amounts of sediment (Gregory Taylor, USACE, oral commun., November 2015). The streambed drawdowns evaluated in this study (WY 2012–18) occurred in late autumn or early winter and typically lasted 1–2 weeks. However, in 2021 (after this study was completed), streambed drawdown operations were modified in response to a court injunction that directed USACE to improve conditions for fish passage and water quality to better support spring Chinook salmon and winter steelhead trout (U.S. District Court for the District of Oregon, 2021). As a result, the streambed drawdown operations for WY 2022–24 were extended nearly 6 weeks from December 1 to January 15, and were followed by period of delayed refill with exceptions for flood risk management actions.

Historical Data Analysis: Storage Curves, Maps, and Channel Surveys

Several historical datasets were evaluated to estimate sediment volume and distribution in Fall Creek Lake. Historical (pre-dam) morphology, land cover, land use, geology, and channel substrate within the footprint of Fall Creek Lake was evaluated from historical maps (USGS, 1942, 1955a, b; USACE, 1947, 1966), aerial photographs (1936), and habitat surveys (1938; McIntosh and others, 2009). Long-term spatial and temporal patterns of sediment deposition in Fall Creek Lake were mainly determined from reservoir storage curves (relation between reservoir storage capacity and lake elevation) augmented with information from historical surveys. Comparison of reservoir storage curves of Fall Creek Lake from 1988 and 2012 as well as a partial storage curve from 1965 (for elevations above 710 ft NGVD29; provided by Jacob Macdonald, U.S. Army Corps of Engineers, written commun., July 6, 2016) provides estimates of total storage loss and changes in the storage-elevation profile of the reservoir over time. Historical survey data, maps, and cross-sections of reservoir topography were also reviewed (U.S. Army Corps of Engineers [USACE], 1966; Jacob Macdonald, U.S. Army Corps of Engineers, written commun., January 29, 2019; app. 1) to provide rough estimates of reservoir sediment thickness and distribution.

Aerial Photograph Acquisition and Digital Surface Model Development

DSMs and orthophotographs of the empty Fall Creek Lake (fig. B2) created with structure-from-motion techniques (SfM; for example, Micheletti and others, 2015) serve as base layers for reservoir geomorphic mapping. Methods for aerial photograph collection, processing, and resulting datasets are summarized here, with more complete descriptions provided within the metadata for each dataset (Keith and Mangano, 2020). Prior to photograph acquisition, 40 ground-control targets were placed throughout Fall Creek Lake and surveyed with real-time kinematic global positioning system (RTK-GPS with reported horizontal and vertical precision ranging from 0.01 to 0.05 m; fig. B3) to project photogrammetry models into real world coordinates (fig. B2). Aerial photography was collected by unoccupied aerial systems (UAS) on November 8 and November 9, 2016 (Sony ILCE-5100; 6,000 by 4,000 pixel resolution), and from a Cessna aircraft with a wing-mounted digital camera (Ricoh GRII; 4,982 by 3,264 pixel resolution) on November 10, 2016, similar to the methods used by Randle and others (2015) and Anderson and others (2017). Agisoft Photoscan Professional (version 1.4) software was used to generate models of dense point clouds, DSMs, and orthophotographs for each day of acquired photographs. The models resulted in ground resolutions of 4.18 cm/pixel or smaller (table B3). Average point densities for each model ranged from of 58.2 to 76.6 points/m² (table B1).

Reservoir Geomorphic Mapping Methods

Detailed geomorphic mapping of WY 2017 reservoir landforms and substrate (general description provided in table B1 and more fully described in appendix 2 and Keith and Stratton Garvin, 2021) provides a basis for evaluating geomorphic processes and patterns of sediment transfer within Fall Creek Lake and to quantify short-term changes from WY 2012 to 2017 streambed drawdown operations. The mapping framework has three main components: process domains (3), landforms (17), and substrate classification (7). The over-arching classification structure is tied to the process domains, which correspond to regions of the reservoir that have distinct landforms and broadly similar suites of geomorphic processes (table B2; app. 2). Mapped landforms and substrate character provided information about (1) sediment regime and processes that formed the landform, (2) general context for the categories of sediment sizes particles being mobilized during drawdown processes, and (3) susceptibility to erosion and re-distribution in future drawdowns (table B3; app. 2).

Geomorphic mapping is based on 2016 SfM DSM and orthophotographs and supplemented with 2012 lidar in upper segments of Fall and Winberry Creeks (app. 1; Keith and Mangano, 2020). Landforms were digitized at a scale of 1:500 with 2016 DSM and orthophotograph coverage and 1:1,500 in areas with lidar-only coverage within Fall Creek Lake. All in-reservoir landforms greater than 10 m² were digitized according to mapping protocols specific to each landform based on their morphology and topographic characteristics. The mapping hierarchy was created to specifically identify landforms created and shaped by streambed drawdown operations (table B3) and landforms that had a greater potential to supply sediment to reaches downstream from Fall Creek Dam compared with those created through longer-term sedimentation processes (app. 2). Therefore, landforms in the main channel domain, such as channel banks and slumping banks, were mapped in greater detail, whereas landforms less likely to drive sediment export during streambed drawdown operations within the reservoir floor and hillslope domains, such as broad areas of reservoir that span multiple pre-dam terraces at higher elevations than the main channel, were lumped into one landform. A general category defining substrate was then attributed to each landform based on the appearance in the high-resolution base imagery. Field verification of substrate was limited and conducted during other reservoir-based field work during the 2016 and 2017 streambed drawdowns.

Topographic Change Assessment

Net volumetric changes and patterns of sediment deposition and erosion for the period 2012–16 were assessed using digital elevation models (DEMs) and digital surface models (DSMs) of Fall Creek Lake during full streambed drawdowns acquired in January 2012 (lidar; Watershed Sciences, Inc., 2012) and November 10, 2016 (SfM data acquired for this study); see appendix 1 for resolution and accuracy details. Quantitative comparisons between the datasets were made with Geomorphic Change Detection (GCD) Software version 6.1.14 (Riverscapes Consortium, 2018). Change detection analyses were focused on the reservoir floor and main channel domains within the lower 2.5 km of the reservoir where ground control coverage was greatest and where most observed changes have occurred. Within this area of interest, the higher-resolution 2016 DSM was also resampled to 0.5-m resolution cell size and snapped to the lidar raster for spatial concurrency between datasets.

The GCD software calculated total erosion, total deposition, and net volumetric change, along with propagated uncertainty analyses within the area of interest. Propagated uncertainty analyses consider the relative error for each survey (Brasington and others, 2000; Lane and others, 2003; Wheaton, 2008). An error surface of 12 cm (default for photogrammetric surveys) was applied to the 2016 DSM and is deemed reasonable given the total model error for control points was 0.08 m and the total error for check points was 0.12 m. For the 2012 lidar data, an error surface of 0.15 m (default for lidar surveys) was used. Comparatively, reported overall project vertical accuracy root mean square error (RMSE) for the lidar data was 0.02 m (Watershed Sciences, Inc., 2012), so this default error surface is considered a relatively conservative estimate of survey error. Spatial homogeneity in surface uncertainty was assumed for the SfM and lidar datasets for the analyses completed here, though approaches for more spatially explicit approaches exist (for example, see Wheaton and others, 2010) and might be more appropriate if doing a more in-depth analysis of topographic change in the reservoir. Overall, the propagated error approach resulted in an uncertainty threshold of about 0.19 m. No account of wetted area uncertainty was made due to the absence of lidar penetration of water, modeled shallow depths, depth distortion from SfM techniques, and variance in discharge during data acquisition (daily mean discharge of 4,200 ft³/s in WY 2012 and 155 ft³/s in WY 2017). Without streamgages in the reservoir to illustrate the relation between discharge and water depth, it is difficult to quantify uncertainty due to discharge, though at the Fall Creek streamgage downstream from the dam, the stage difference between 4,200 ft³/s in WY 2012 and 155 ft³/s is about 5.2 ft (1.6 m), which is quite substantial. Adjusting total erosion volumes for discharge would likely decrease the overall estimated total amount of erosion.

Magnitude of Erosion and Deposition in the Fall Creek Lake Water Years 2012–17

Within the lower reservoir of Fall Creek Lake, topographic change analyses show 143,200 m³ of change for erosion and 13,700 m³ of change for deposition (table B1; fig. B18). Total net erosion is 129,500 m³, which suggests an annual average net erosion rate over the 5-year analyses period of about 25,900 m³/year. However, field measurements downstream (Schenk and Bragg, 2014, 2021; G. Taylor, USACE, oral commun., 2017) indicate that larger amounts of sediment were eroded during earlier streambed drawdown periods (WY 2012–13) than during later drawdowns suggesting later streambed drawdowns were shifting toward being supply limited as sediments from accessible reservoir supply are exhausted, similar to patterns of reservoir sediment change observed following dam removals (for example, Major and others, 2012; Collins and others, 2017)

Comparison of the reservoir elevation changes from 2012 to 2016 with mapped reservoir landforms shows erosion was focused in the main channels of Fall and Winberry Creeks (fig. B18). The area with greatest lowering of the reservoir topography was about 0.7 km upstream from the dam where the change in channel bed elevations decreased by up to 3.8 m (point A on fig. B18). Areas with a net increase in vertical elevation were few and mainly located along the channel margins (for example, rkm 13.4, point B on fig. B18) where deposition of less than a meter was caused by bar formation or sediment delivery from slumping banks. In some cases, drawdown terraces, situated above the channel bed, also accumulated sediment (for example, rkm 12.3, point C on fig. B18). Few vertically stable areas within the main channel of Fall Creek were measured, and where present, typically coincide with bedrock outcrops (for example, rkm 14; point D on fig. B18).

Much of the pelagic reservoir floor appeared vertically stable with no changes greater than the uncertainty threshold. Prominent areas of erosion within the reservoir floor domain were associated with drawdown channels, drawdown terraces, and drawdown distributary zones. The large distributary zone near the confluence of Winberry Creek (rkm 12.6; fig. B18) showed changes in elevation ranging from +0.32 to −0.9 m, though the overall feature was dominantly erosional with large areas of no detectable change. Some drawdown channels showed incision up to 1.8 m near the confluences of Fall and Winberry Creek channels (point E on fig. B18).

Patterns of Erosion and Deposition During the 2016 Streambed Drawdown

In addition to the volumetric analyses, direct comparison between the three orthophotographs acquired during the November 2016 (WY 2017) streambed drawdown were used to qualitatively assess patterns and processes of planimetric geomorphic change within Fall Creek Lake during a single streambed drawdown event. Comparison of 2016 orthophotographs between November 8 and 9, and November 9 and 10, shows the main channel was laterally stable along much of the reservoir except for the lowest 1.5 km of the reservoir where lateral adjustments in bank location were systematically greater than 0.5 m over distances of 20–50 m. For the period between November 8 and 9, the maximum observed lateral erosion resulted in about 2.5 m of bank retreat near rkm 13.4; over the following days (November 9–10) the maximum observed lateral erosion was about 0.75 m near the same location (figs. B19A–C). For both days, these findings suggest average lateral erosion rates of 1.1 m/day are possible at discharges of about 155–194 ft³/s at the end of the low precipitation drawdown period. Other localized erosion areas over November 8–10 were observed in drawdown surface landforms and along the transition between channel and reservoir floor (for example, fig. B19D–F). At different discharges or under other conditions, such as with substantial rain or below freezing temperatures, lateral erosion through reservoir floor deposits would likely vary. Additionally, bank stability also varies with water table elevation, so erosion rates are likely higher in the days immediately following lake lowering, when slumping may reflect positive pore pressure in exposed channel banks.

Suspended-Sediment Monitoring and Calculation of Loads Water Years 2013–18

Continuous discharge and turbidity were used in conjunction with discrete SSC samples to develop ordinary least squares regression models following USGS techniques and methods described in Rasmussen and others (2009). At the Fall Creek streamgage (14151000), regression models were typically developed for three distinct phases within each yearly monitoring period, including pre-streambed drawdown, streambed drawdown, and post-streambed drawdown periods, reflecting changes in the grain size (sand versus finer-grained sediments) of suspended sediment that affected turbidity sensor response and relations of turbidity with SSC. However, for some years, a single regression model was developed owing to a limited number of samples or non-distinguishable grain-size characteristics of sediment between the phases (Schenk and Bragg, 2021). The regression models produced continuous records of SSC (at 15-minute intervals), which were combined with continuous discharge records (also at 15-minute intervals) to compute SSL for each monitoring location. For periods of missing turbidity, SSC was estimated using various methods and multiplied by the discharge to compute SSL; turbidity and SSC sampling and SSL computations followed established protocols (Edwards and Glysson, 1999; Rasmussen and others, 2009), and additional details can be found reports by Schenk and Bragg (2014, 2021).

Bedload Sediment Monitoring and Calculation of Transport Rates Water Years 2013 and 2017

Bedload measurement collection for WY 2013 is documented in Schenk and Bragg (2014) at the Fall Creek streamgage (fig. A6). Similar collection methods were employed in WY 2017. Sample collection in WY 2017 was timed to characterize transport rates at a variety of streamflows during and following streambed drawdown. Sampling followed the multiple equal-width-increment (MEWI) method described in Edwards and Glysson (1999) for two of the WY 2017 samples collected with two passes across the channel and a minimum of 20 total verticals. Two additional samples followed a slightly modified sampling protocol and consisted of 18 verticals. Bedload samples were collected by wading with a BLH-84 handheld sampler (7.6 by 7.6 cm opening, 1.4 expansion ratio) during lower streamflows on November 4, 2017, and from the cableway with a BL-84 sampler secured with a stayline about 15 m upstream during higher streamflows (fig. C2). Bedload samples were analyzed at the Cascades Volcano Observatory Sediment Laboratory for grain size and mass, as well as loss-on-ignition (LOI) for organic content. Samples were dried, sieved into full-phi size class and each size class weighed.

A transport rate for each MEWI measurement was calculated following guidance in Edwards and Glysson (1999), using the formula

Each bedload sample pass was analyzed separately for grain size and LOI, but only the full samples were used to calculate transport rates. Bedload rates reported in Schenk and Bragg (2014) were computed with the English units form of equation 1 and are converted to SI units for this report.

Water Year 2013 Suspended-Sediment Budgets from Schenk and Bragg (2014)

Schenk and Bragg (2014) assessed suspended-sediment loads using computed SSC values and time series of streamflow at six sites in WY 2013 to develop sediment budgets between the Fall Creek Lake inflow and outflow and between the Fall Creek and Jasper streamgages. Over a 93-day period spanning the streambed drawdown, more than 6 times as much sediment was computed passing the Fall Creek streamgage (14151000; 54,300 metric tons [t], error ranging from 40,400 t [−25.7 percent] to 73,500 t [+35.2 percent]) downstream from the Fall Creek Dam as was entering the reservoir from Fall Creek (Fall Creek above North Fork near Lowell, Oregon, 14150290; 6,450 t; error ranging from 4,300 t [−33.3 percent] to 10,350 t [+60.5 percent]) and Winberry Creek (Winberry Creek near Lowell, Oregon, 14150800; 2,060 t; error ranging from 1,710 t [−17.1 percent] to 2,410 t [+16.9 percent]), most of which was mobilized during the 6-day streambed drawdown operation when turbidity and SSC were elevated (Schenk and Bragg, 2014; values for 14150290 updated from original report). Over a 106-day period, Schenk and Bragg (2014) also calculated a balance of about 7,440 t more suspended sediment at the Jasper streamgage (14152000) than at the Fall Creek streamgage (14151000) accounting for sources from Little Fall Creek (Little Fall Creek near Lowell, Oregon, 435730122483201) and the Middle Fork Willamette River near Dexter Dam (Middle Fork Willamette River near Dexter, Oregon, 14150000), but about 25,000 t more sediment passing the Fall Creek over a shorter 6-day period coinciding with the streambed drawdown. The authors note:

Suspended-Sediment Transport at the Streamflow-Gaging Station Fall Creek Below Winberry Creek, Oregon (14151000), Water Years 2013–18

Suspended-sediment loads computed for the full monitoring periods WY 2013–18 at the Fall Creek streamgage 1.4 km downstream from Fall Creek Dam range from 54,700 t in WY 2013 to 13,900 t in WY 2018 (table C1; figs. C3, C4). Calculated SSL were greatest in WY 2013 measured for the 119 day monitoring period, and 73 percent of that mass occurred during the 5.0-day drawdown period when lake levels were below 690 ft (40,200 t; table C1; figs. C3, C4). Substantially less suspended sediment was calculated for WY 2014 (25,100 t for the 133-day monitoring period) when cumulative discharge was about a third of that observed in WY 2015 and when air temperatures were below freezing conditions in the first few days of the 9.1-day streambed drawdown period (table C1; figs. C3, C4). Extensive ice within the reservoir during the beginning of the WY 2014 streambed drawdown combined with low discharges led to reductions in reservoir erosion and subsequent sediment export to downstream reaches (Schenk and Bragg, 2021). In WY 2015, 33,000 tons of suspended sediment were mobilized during the 103-day monitoring period, including 7,160 t during the 3.8 days when reservoir elevations were less than 690 ft; 1,740 t (error ranging from 1,040 t [−40.1 percent] to 2,170 t [+25.1 percent]) were transported in the period when the reservoir elevation was raised and lowered in the middle of the drawdown, for a total of 8,900 t during the drawdown timeframe. In WY 2016, the computed load over the 206-day monitoring period (25,600 t), and the sediment mobilized during the streambed drawdown below 690 ft (5,900 t) were slightly less than WY 2015 even though the WY 2016 streambed drawdown durations was about twice as long and cumulative discharge was about 25 percent greater than in WY 2015; although daily median discharge during the drawdown was much lower (table C1). The reservoir elevation also was raised and lowered in the middle of the WY 2016 drawdown; SSL export for lake elevations below 690 ft (5,900 t) and the 4.6 days above 690 ft (2,980 t; error ranging from 1,680 t [−43.7 percent] to 5,260 t [+76.7 percent]) during the streambed drawdown timeframe resulted in the similar SSL (8,900 t) as the WY 2015 streambed drawdown timeframe including SSL from above and below 690 ft. In WY 2017, 18,600 t of suspended sediment was calculated for the full 198-day monitoring period. The second greatest loads calculated for the streambed drawdown period occurred in WY 2017 and 2018, which both had 11,300 t of SSL during the respective drawdowns, at low and moderate median streamflow observed during the drawdown monitoring program (table C1).

Despite considerable year-to-year variability in SSL, streamflow conditions, dam operations, and partial-year monitoring programs that varied in duration and preclude calculation of annual loads, the SSL data from WY 2013 to 2018 reveal three key findings. First, total loads for each WY monitoring periods were much higher in WY 2013 compared with WY 2014–18, indicating diminishing sediment supply (table C1). Although WY 2014 data likely reflect anomalously low sediment export due to freezing conditions and low inflows to Fall Creek Lake, subsequent years also resulted in suspended-sediment loads that were 40–66 percent lower than those computed for WY 2013. Second, the suspended-sediment loads passing the Fall Creek streamgage during streambed-drawdown conditions (lake level below 690 ft) comprise a major portion (between 18 and 73 percent) of each-year’s full monitoring period sediment load (table C1; fig. C4). Third, seasonal patterns of suspended-sediment transport followed a similar overall pattern each year whereby sediment loads are low prior to the drawdown, then increase rapidly as lake levels approach the streambed drawdown elevation of 680 ft (fig. C5). Following the drawdown period (as lake levels rise above 690 ft), calculated loads remain slightly elevated above pre-drawdown levels but decline through the following months (fig. C4).

Site-Scale Changes in Channel-Bed Elevation at Streamgages from Specific Streamgage Analyses, Water Years 1972–2020

Vertical changes in bed-elevation can be evaluated with specific streamgage analyses (Blench, 1969; Klingeman, 1973; Juracek and Fitzpatrick, 2009; Cashman and others, 2021). Discharge values at many USGS streamflow-gaging stations are computed from stage measurements based on a stage–discharge relation (rating curve; for example, see Rantz, 1982). Over time, that relation can change with local changes in hydraulics or channel geometry. Changes in channel geometry due to vertical adjustments from channel bed lowering (incision) or raising (aggradation) are easier to detect at low streamflow than at higher streamflows that are more sensitive to changes in channel width. Field measurements of stage and discharge collected over time are used to create a rating curve. Persistent adjustments in the channel bed may warrant creation of a new rating curve, whereas temporary adjustments due to episodic fill or scour related to sources of streamflows (floods) or sediments (drawdown operations, upstream bank failure, or debris flows) would deviate from an existing rating curve but not require a whole new rating curve. The methods for specific streamgage analyses at the USGS streamflow-gaging stations at Fall Creek, as well as at Dexter and Jasper (Middle Fork Willamette River; fig. A6), are presented here with findings focused on the period spanning streambed drawdowns at Fall Creek Lake to identify potential changes in channel bed elevation relating to reservoir sediment releases. Longer-term changes in bed elevations at the Jasper gage were evaluated with specific streamgage analyses (spanning 1953–2012) by Wallick and others (2013) and based on earlier analyses by Klingeman (1973).

Specific Streamgage Analyses Results

At the Fall Creek streamgage, 261 field measurements (from January 25, 1972, through September 24, 2020) were less than 6,000 ft³/s and available in NWIS. Of those measurements, 56 percent were categorized as good, 39 percent were categorized as fair, and 5 percent were categorized as poor (1 measurement was unspecified). Stage residuals between those measurements and predicted stages from the benchmark rating curve 26.0 range from −0.07 to 0.32 m where negative values are lower than current rating predicted stage and positive values are higher (fig. D1A). Overall, measured stages at the streamgage are relatively stable with 252 of the field measurement residuals evaluated falling within 2 standard deviations (plus or minus [±] 0.07 m) of the mean (0.00 m) from the rating curve prediction. The other nine residual values all were positive and four of those values coincide with to Fall Creek Lake streambed drawdown events (fig. D1B). Since the streambed drawdowns began in WY 2012, field measurements show slightly greater deviations from the current rating that occur during or shortly after (a few months) the official streambed drawdown. The most apparent of these deviations is for three measurements made during the WY 2012 drawdown where the relation between streamgage height and discharge suggests an increase in local bed elevation of 0.12 m (December 1, 2011) relative to the current rating and reaching as much as 0.32 m on January 25, 2012. During the WY 2013 streambed drawdown, measurements made on December 12 and 13, 2012, were nearly 0.09 m greater than predicted from the current rating curve. Other measurements collected during subsequent streambed drawdown periods of WY 2014–18 were generally closer to predicted values.

Specific streamgage analyses at the Fall Creek streamgage indicate that, although channel bed elevations varied minimally between 1972 and 2020, temporary increases in the bed elevation occurred in the months following streambed drawdowns in WY 2012–20. The overall trend of vertical stability is supported by minimal net variation in the stage-discharge relation, coinciding with field observations of bedrock near the Fall Creek streamgage and other historical analyses that indicate long-term stability (Keith and others, 2023a). The few measurements with the greatest stage residuals were recorded during streambed drawdowns in WY 2012 and 2013 (fig. D1B) likely reflect actual increases in water surface elevations driven by temporary increases in channel bed elevation and are consistent with observations of temporary sediment storage on the channel bed during the drawdowns (fig. D2). These apparent increases in bed elevation during the early streambed drawdown periods do not persist throughout the year, and sediment dissipates quickly (days to weeks) depending on magnitude of sediment released following the streambed drawdown. The temporary increases in bed elevation during the drawdowns were greater in WY 2012 and 2013 than during the WY 2014–20 drawdowns, consistent with patterns of SSL for WY 2012–18 (Chapter C. Sediment Delivery from Fall Creek Lake and Transport through Downstream Reaches).

Channel bed elevations along the Middle Fork Willamette River were evaluated from gaging station measurements available in NWIS that date back to the early 1980s. Specific streamgage analyses indicate a trend of gradually decreasing bed elevations at the Dexter streamgage and generally stable bed elevations at the Jasper streamgage. At the Dexter streamgage (located upstream from the Fall Creek confluence and 4.8 km downstream from Dexter Dam), there have been 212 measurements at discharges lower than 7,500 ft³/s between December 7, 1983, and September 3, 2020 (post-regulation). Thirty-five percent of the measurements were categorized as good, 53 percent were fair, and 11 percent were poor (1 measurement was unspecified). Stage residuals from the benchmark rating (23.0) range from −0.09 to 0.36 m (fig. D3A), but generally, the comparison suggests a decrease in bed elevation of about 0.30 m since the early 1990s. For the short period of WY 2012–20 spanning streambed drawdowns, the specific streamgage analyses at Dexter shows decreasing bed elevations that are consistent with longer-term trends of apparent bed lowering since 1983.

In contrast to the Dexter specific streamgage analyses showing gradual bed lowering, the 175 measurements made at the Jasper streamgage between January 25, 1985, and September 23, 2020, were compared to the benchmark rating curve 16.0 and indicate general bed stability since the mid-1980s (fig. D3B). Measurements at the Jasper streamgage were categorized as excellent (1 percent), good (78 percent), fair (20 percent), poor (1 percent), or unspecified (1 percent). After the streambed drawdowns were implemented at Fall Creek Dam, there have been a few large stage residuals (for example, October 2014, June 2017, February 2020, which range from 0.04 to 0.06 m; highlighted in fig. D3) that potentially signify increases in channel bed elevation resulting from reservoir sediment releases. However, these residuals fall within the full range of variability (from −0.05 to 0.07 m) observed for the overall analysis period; therefore, any assessment of aggradation is uncertain and cannot be attributed to sediment deposition from the Fall Creek drawdown. Additionally, these stage residuals are on the order of a size of gravel to cobble and may reflect natural variability in bed conditions. Even if these stage residuals are representative of actual aggradation, the increases in bed elevations are temporary (subsequent measurements do not indicate persistent change in stage for that discharge), and the magnitude of potential aggradation is very modest (less than several centimeters). Wallick and others (2013) compared rating curves over a longer period at the Jasper streamgage (WY 1953–2012) for low to moderate streamflows (1,150–11,500 ft³/s). They found the incision prior to 1975 was followed by relative stability through WY 2012. Although, re-evaluation of that streamgage (Wallick and others, 2013) suggests some minor (less than 0.3 m) aggradation between WY 1953 and 1973 followed by relative stability.

Overall, findings from specific streamgage analyses indicate sediment released from the Fall Creek Lake streambed drawdowns is not causing persistent, aggradation in the low-flow stream channel of the Fall Creek or Middle Fork Willamette River near the Fall Creek and Jasper streamgages (respectively). Reservoir sediments may temporarily be deposited in the channel near the gaging station at Fall Creek, but this sediment is rapidly carried downstream in subsequent weeks and months following the drawdown. Additionally, evaluation of the decadal-scale relations between streamflow discharge and stage suggests that the channel bed at the locations of the Fall Creek and Jasper streamgages have been relatively stable since the late 1970s to early 1980s. At these streamgages, overall bed stability and the presence of bedrock in the reach suggest that transport capacity and sediment supply are somewhat balanced or that transport capacity exceeds sediment supply (at least in post-dam regime). Conversely, potential incision at the Dexter streamgages since the early 1980s suggests transport capacity exceeds sediment supply.

Site-Scale Characterization of Bed-Material Grain Size, Water Years 2015–16

Bed-material substrates of gravel-bed rivers reflect sediment transport conditions balanced by sediment supply and transport capacity (Church, 2002, 2006; Lisle, 2012). Evaluation of bed-material grain sizes is useful for characterizing morphology and sediment transport conditions along a river. Where repeat measurements are made over time using similar methods to collect a representative sample, bed-material grain size can indicate local changes that may have resulted from variation in streamflows or sediment supply. Increased sediment supply from Fall Creek Lake during streambed drawdowns can potentially change overall size of bed-material within the channel bed and gravel bars downstream from the dam. For example, the release of fine-grained sediments into these gravel-bed rivers might be expected to result in an overall fining of the channel bed and reduction in overall grain size or could result in local sand deposition on gravel bars, which has implications for bar mobility, aquatic habitats, and vegetation succession. For this study, bed-material grain sizes were measured at multiple gravel bars in 2015 and 2016 along Fall Creek and Middle Fork Willamette River to assess overall sediment transport that reflects decadal-scale condition as well as short term (annual) changes that may relate to the transport and deposition of reservoir sediments released during streambed drawdowns at Fall Creek Lake.

Bed-Material Surface Grain-Size Measurement Results

Bed-material grain-size measurements from Fall Creek and Middle Fork Willamette River show that bed-material substrates are primarily composed of cobble to gravel-sized particles with median grain sizes (D50) ranging from 34 to 90 mm and no apparent longitudinal patterns in bed-material textures (table D1; fig. D4). The two sites within the Lower Fall Creek reach had D50 values of 40.4 and 59.0 mm (figs. D4, D5A). Within the Jasper reach of Middle Fork Willamette River, D50 for the three sites ranged from 43.8 to 90.0 mm (figs. D4, D5B). In the Clearwater reach of the Middle Fork Willamette River, D50 at 6 sites ranged from 36.3 to 69.1 mm (figs. D4, D5C), but surficial measurements by Jones and others (2016) along other bar features using different methods found finer gravel distributions (D50 less than 32.0 mm, rkm 5.6). The sand fraction at many sites within the study area was 0 and did not exceed 6 percent for any of the sampled sites (table D1; fig. D5). Bed-material grain sizes were moderately-well to moderately sorted (classification from Folk and Ward, 1957; table D1; fig. D5) except at four sites that had localized sand accumulations and non-zero sand fractions—rkm 5.1 on Fall Creek (Downstream Little Fall Creek) and rkm 17.1, 16.6, and 9.5 on the Middle Fork Willamette River (Sand Mountain, River Right Island, and Orchard Bars, respectively).

In 2016, five additional sites in the Clearwater reach of the Middle Fork Willamette River were measured to better characterize spatial variability in surface grain size within and among bars. These measurements reveal distinctly coarser (Upstream The Nature Conservancy [TNC]) and finer (Downstream TNC) sections of a single large gravel bar located at rkm 5.8–5.7, where D50 values range from 59.1 to 34.1 mm. Bed-material textures on closely spaced bars did not show an apparent longitudinal pattern in D50 values; however, one of the finer grain-size distribution bars (Wide Bar, where D50 was 36.3 mm) was located in a relatively wide, laterally active area and was moderately well sorted (table D1; fig. D5C; Keith, 2019). All sites in this reach had a D16 greater than 19.0 mm and a D84 less than 113 mm. Sand content at Orchard and Upstream TNC Bars was 6 and 4 percent, respectively. Orchard Bar was poorly sorted whereas most other sites were moderately sorted.

Repeat measurements of bed-material substrates from 2015 and 2016 within the Lower Fall Creek, Jasper, and Clearwater reaches do not reveal substantial changes between years in bed-material textures at the sampling sites; although, a larger sampling size for the particle counts would be needed to accurately compare sites over time (Bunte and Abt, 2001). At rkm 5.1 in Lower Fall Creek reach (Downstream of Little Fall Creek [DSLFC] site) just downstream from Little Fall Creek, very similar bed-material textures were observed in both 2015 and 2016, with D50 values varying by about 3 mm, while particle size distributions, sand fractions (6 percent) and sorting (poorly sorted) were consistent for both years (fig. A6; table D1; fig. D5A, Keith, 2019). The three sites on the Middle Fork Willamette River in the Jasper reach showed minimal changes. Potential coarsening was detected at Downstream of Confluence (DSC; rkm 17.7) and Sand Mountain (rkm 17.1) Bars where D50 value increased by nearly 20 mm in 2016. But for the Rail Road Island Bar (rkm 16.6), the 2016 distribution was only slightly finer the 2015 distribution where the sand fraction increased from 0 to 3 percent although changes are typically within the range of uncertainty determined for error due to sampling bias. Sorting changed from moderately sorted to poorly sorted at Sand Mountain and Rail Road Island Bars where sand content increased slightly (by less than 3 percent; table D1; fig. D5B, Keith, 2019). Within the Clearwater reach, one repeat bed-material grain-size measurement was made at Clearwater Park (rkm 6.1) showing similar grain-size distributions in both years.

Overall, the surficial grain-size data reveal several findings regarding sediment transport conditions in the study area and effects of Fall Creek Lake streambed drawdowns on downstream substrates. First, all of the sampling sites had little to no sand and were dominated by cobble to gravel size substrate, indicating minimal persistent deposition of sands and fine-grained sediment sourced from Fall Creek Lake on the channel bed; however, these analyses do not address infiltration of fine sediments into the channel bed. Because grain-size distributions vary considerably from site to site and no apparent longitudinal trend in surface grain size was detected along Fall Creek and the Middle Fork Willamette River, the observed patterns of grain-size variation likely reflect local conditions, such as hydraulics and availability of coarse sediment liberated from nearby bank erosion rather than reach-scale trends in transport capacity. Broadly similar patterns of river-specific transport capacity and bed-material supply for Fall Creek and Middle Fork Willamette River are expected because (1) bed-material supply to Fall Creek and Middle Fork Willamette River is essentially depleted due to upstream dams and lack of large unregulated tributaries (O’Connor and others, 2014), (2) streamflow magnitudes are dominated by regulated dam releases, and (3) reach-scale slope values within the Lower Fall Creek, Jasper, and Clearwater reaches are similar (table D1). Although there are only five sites from which to evaluate temporal changes in bar substrate, the lack of substantial, systematic changes in bed-material textures between 2015 and 2016 suggests that sediment releases from the Fall Creek Lake streambed drawdown in WY 2015 did not produce system-wide deposition of fine-grained sediment on the gravel bar surfaces sampled within this period.

Site-Scale Deposition of Fine-Grained Sediment from Clay-Horizon Markers, Water Years 2016–17

Sand and finer sediment deposition along channel margins or on low elevation overbank surfaces can influence channel conveyance (potentially increasing flood hazards), aquatic and riparian habitats (potentially leading to losses in low-velocity areas accessible at high streamflows), and riparian vegetation (through burial of existing stands or creation of new establishment sites). To assess magnitudes and patterns of overbank deposition that may have resulted from Fall Creek Lake streambed drawdowns, a short-term, event-based, monitoring program in WY 2016–17 was established to measure local deposition within the active channel and floodplain areas at twelve sites along Fall Creek and the Middle Fork Willamette River with clay-horizon markers (fig. A6; Keith, 2019). Repeat site-scale measurements of deposition in the days and months following the streambed drawdowns at Fall Creek Lake provide a dataset from which to assess depositional patterns that may relate to the streambed drawdowns and other influences such as variation in streamflow, sediment supply from other sources, and local hydraulics. Clay-horizon markers are a tool to measure aggradation of sediment depths on surfaces that typically are only subject to low velocity streamflows (for example, Kleiss, 1996; Gellis and others, 2009; Curtis and others, 2013; Hupp and others, 2008, 2015). Bright white feldspar clay powder forms a distinct stratigraphic layer in the sediment record at the time of installation; if the layer is not locally redistributed or eroded, it can be used to measure the depth of deposition above the clay marker following subsequent flooding or other sedimentation events. The advantage of using clay-horizon markers for sedimentation monitoring is that they provide a distinct, pre-event (prior to WY 2016 and 2017 Fall Creek Lake streambed drawdowns for this study) surface from which to measure fine-grained deposition and have a limited effect on the local hydraulics that influence local sedimentation (Steiger and others, 2003).

Clay-Horizon Marker Methods

Clay-horizon markers were installed at multiple sites within Fall Creek and Middle Fork Willamette River in early WY 2016 and 2017 prior to the annual Fall Creek Lake streambed drawdown of each respective year. Approaches for this study were broadly similar to those used by Kleiss (1996) to assess bottomland hardwood wetland sedimentation in the Cache River Basin, Arkansas, and Hupp and others (2008) for floodplain sedimentation in the Atchafalaya River Basin, Louisiana. Each marker consisted of feldspar clay powder placed onto ground surfaces cleared of sticks or other large debris (grasses or exposed soil were left intact) in 0.5 by 0.5 m squares about 0.5 cm thick. At each site (fig. A6), clay markers were placed in transects spanning an array of geomorphic conditions such as: low-elevation unvegetated bars to high-elevation, vegetated bars; off-channel features inundated at low streamflows and adjacent vegetated bars (for example, fig. D6); and along low elevation floodplains. Multiple markers were installed at 10 sites in the autumn of WY 2016 prior to the annual Fall Creek streambed drawdown. Clay-horizon markers were replaced at 9 of those sites in autumn of WY 2017 and installed at one new site. Upon deployment, each clay-horizon marker was flagged on an outside corner with PVC or rebar so that sites could be relocated after burial. Because local hydraulics near the PVC or rebar appeared to affect markers at some sites, space between poles and clay markers was adjusted to 0.5 m or greater at two sites in WY 2017. Measurement sites along Fall Creek were generally accessible by foot, while most sites on the Middle Fork Willamette River were only accessible by boat, so timing of measurement observations varied with streamflow and other site access logistics. For most sites along Fall Creek, 500 ft³/s was considered the maximum discharge for accessing markers (Fall Creek streamgage, 14151000). For the Middle Fork Willamette River, the streamflow threshold for accessing marker sites was about 2,000 ft³/s (Dexter streamgage, 14150000) upstream from the Fall Creek confluence and 2,500 ft³/s (Jasper streamgage, 14152000) downstream from the Fall Creek confluence. For each marker, the measurement dates and sediment thickness were recorded, as were streamflow data (including average mean daily streamflow for period between measurements and number of days exceeding site access threshold) from the nearest streamgage. If clay horizon makers showed evidence of disturbance from erosion, wildlife, or vegetation growth, the markers were not replaced. Clay markers submerged under water were not measured.

Clay-horizon markers are useful for measuring very localized sediment accumulation, but as with other techniques for measuring sedimentation, there are many sources of uncertainty that can confound interpretation of results. Uncertainty can arise from implementation (quality of marker deployment, geomorphic setting and site selection, number of markers deployed) and measurement (number and locations of measurement within a marker, precision, observer accuracy). Additionally, interpretation of deposited material can also lead to some uncertainty in the magnitude or source of sediment observed on the clay-horizon marker; for example, fluvially-deposited sediment should be distinguished from sediment introduced by wildlife disturbance, hillslopes, or other sources. Even within an individual marker during one measurement period, measured deposition can be spatially variable across the marker, making it difficult to characterize overall patterns of deposition. Uneven ground conditions underlying each marker and spatially heterogeneous deposition patterns could affect measured depths, so multiple depth measurements were used at each clay marker to characterize variability in deposition to the nearest millimeter using a ruler. Over time, settling and compaction of sediment deposits could account for reductions in measured sediment depths; especially considering most sites underwent inundation and fluvial modification after initial sediment deposition. However, reductions in measured sediment depths over time related to compaction or fluvial or wind erosion of the deposit were not directly evaluated in this study. Additionally, accurate sampling of coarser sands proved to be difficult. Sandy deposits measured in August 2017 lacked sufficient cohesion to permit excavation and measurement down to the clay marker surface; to mitigate these conditions, water was gently poured over the deposits to increase cohesion, which also resulted in some minor compaction and reduced sediment depths. Observer repeatability was not directly accounted for in this study. Interpolation between marker sites or extrapolation of marker findings to other similar geomorphic landforms across a reach introduces additional uncertainties because local aggradation reflects complex local hydraulics. Direct comparisons of deposition within and among sites is challenging because of the diversity of geomorphic settings where floodplain sedimentation was monitored for this this study. For example, a transect of multiple clay markers deployed at different elevations across a gravel bar produces a different suite of depositional conditions than a nested cluster of markers within an off-channel feature affected by backwater from the main channel. In order to better characterize the magnitudes of deposition that may be occurring throughout the study area, sites that spanned multiple geomorphic conditions and assumed to be generally representative of those conditions were prioritized over multiple sites with similar geomorphic conditions.

Clay-Horizon Marker Results

The range of measured sediment depths at clay-horizon markers along Fall Creek and Middle Fork Willamette River varied considerably within and among sites over the 2 years of monitoring to assess deposition associated with the WY 2016 and 2017 streambed drawdowns at Fall Creek Lake. Sediment deposited on clay-horizon markers was primarily silt to medium sands (fig. D7), although coarse sand and some gravels were also observed (fig. D8). Overall, cumulative deposition in WY 2016 (July–September 2016 measurements) varied from less than 6 mm at individual markers within the Unity, Row Tree, Place Road, and The Nature Conservancy (TNC) sites to 185 mm at Sand Mountain, although maximum sediment thickness measured earlier in the season was sometimes greater. Five sites displayed more than 50 mm of cumulative deposition at individual markers (Tufti, Row Tree, Fall Creek Confluence, Sand Mountain sites; figs. D9A, E, F, H; Keith, 2019). In WY 2017, 6 sites showed less than 6 mm of deposition at individual markers (Fall Creek Gage, Unity, Place Road, Row Tree, Sand Mountain, and TNC), but no sites exceeded 50 mm in depth (maximum recorded depth was 49 mm at Sand Mountain). Maximum depth at most sites in WY 2017 was much lower than was observed in 2016, ranging from 20 to 40 mm. Comparisons of sediment depths at markers within individual sites reveal considerable within-site variability, with typical marker average spanning less than 6 mm at higher elevation markers and those farther from the main channel to more than 20 mm at low-elevation sites near the channel (fig. D9; Keith, 2019).

During WY 2016, most clay-horizon marker sites showed a general increase in floodplain deposition in the days and months since the November 2015 streambed drawdown. In the days immediately following the initiation of the streambed drawdown (drawdown began November 5, 2015; measurements made on November 9, 2015) when streamflows were modest (daily-mean discharge between 71 and 696 ft³/s), only 4 of the 28 clay-horizon markers showed measurable deposition. At these four clay markers, deposition was relatively modest (less than 2.7 mm) and observed only at low elevation markers nearest to the channel at the two sites nearest to Fall Creek Dam (Tufti Park and Fall Creek Gage; fig. D9A, B; Keith, 2019). The next measurement effort in January 2016 followed a longer period of higher streamflows (about 2 months since previous measurement on Fall Creek; first measurements on the Middle Fork Willamette River) that spanned the maximum mean daily streamflows for WY 2016 (3,270, 13,100, and 18,300 ft³/s at the Fall Creek, Dexter, and Jasper streamgages, respectively) observed during the total clay-horizon marker monitoring period. During the January 2016 measurement, up to 267 mm of sediment was measured at Sand Mountain on the Middle Fork Willamette River (subsequent erosion at the marker resulted in 185 mm cumulative sediment thickness near the end of WY 2016; fig. D9H), and most clay markers in the Upper and Lower Fall Creek and Jasper reaches showed minor to moderate deposition (fig. D9). For comparison, clay markers at the Chub Pond site in the Dougren reach (reference site on the Middle Fork Willamette River upstream from the Fall Creek confluence and not influenced by Fall Creek Lake streambed drawdowns) also showed minor deposition up to 2.5 mm in this period (fig. D9G; Keith, 2019). Most of the sites revisited along Fall Creek in March 2016 continued to show minor increases in sediment thickness at one or more markers, though the Tufti and Fall Creek sites were dominated by decreases in sediment depths. Six sites showed continued deposition of more than 10 mm at a minimum of one marker through the summer of 2016, although the Middle Fork Willamette River sites had not been sampled since January 2016 and reductions in sediment depths were apparent at most markers along the Fall Creek sites. Clay markers where measured sediment depths decreased over time often showed evidence of erosion and there were several instances where the entire markers were lost. For example, Tufti, Row Tree, Fall Creek Confluence, and Sand Mountain sites all contained markers with more than a 10-mm decrease in sediment thickness from the previous visit. At the time of the last measurement WY 2016, nine markers had been destroyed or were missing and vegetation had begun to displace the remaining markers so that any future measurements were not reliable. Therefore, new markers were deployed in WY 2017 slightly upstream from the WY 2016 markers.

Field observations in WY 2017 reveal similar temporal and longitudinal patterns in floodplain sedimentation to those observed in WY 2016, although the magnitudes of sediment deposition were much lower in WY 2017. Maximum marker depth observed in WY 2017 was 78.5 mm compared with 267 mm observed in WY 2016 (fig. D9). During the measurement period in January 2017, nearly all markers showed deposition along Fall Creek ranging up to 78.5 mm at the Row Tree site, though the Unity site only showed 1.5 mm or less of sediment (fig. D9). By March, sites along Upper and Lower Fall Creek reaches showed variable depositional patterns generally with losses in sediment depth at markers near the main channel and little to no increase in depths for the Tufti, Fall Creek streamgage, and Unity sites (figs. D9A–C). A similar pattern for the Row Tree site was also measured for the March measurement on the side of the bar closest to the channel; however, more substantial increases in sediment depth (increase of up to 28 mm) were observed farthest from the main channel (fig. D9E). Sediment depths also increased at the Fall Creek Confluence markers that were not underwater (fig. D9F). By May 2017, most sites along Fall Creek showed decreases in sediment depths with several of the clay-horizon markers being unrecoverable. Sites on the Middle Fork Willamette River were first measured in March 2017, 6 months following the WY 2017 streambed drawdown. Downstream from the confluence with Fall Creek deposition on clay-horizon markers showed smaller magnitude depths (ranging from 0 to 9.4 mm the TNC Downstream site) compared with WY 2016. Additionally, the Chub Pond site upstream from the confluence with Fall Creek showed deposition up to 5.3 mm. Later measurements in August 2017 showed substantial deposition at Sand Mountain and TNC Downstream (up to 45.3 and 10.4 mm, respectively) while most other sites showed no to little change in measured sediment thickness at the clay-horizon markers.

Landform-Scale Erosion and Deposition from Lidar Comparison, 2009–15

Repeat landform-scale assessment of floodplain topography was conducted to document cumulative changes and spatial patterns in erosion and deposition across 3- to 6-year periods that may relate streambed drawdowns or other geomorphic processes. The measurements at the clay-horizon markers provide a method for tracking small (millimeter-scale) variation in local deposition that may occur in the days and months following an individual drawdown operation, whereas the topographic change analyses characterize erosion and deposition across entire landforms for multiple years. The topographic change analyses focused on four sites along Fall Creek and the Middle Fork Willamette River downstream from the Fall Creek Dam where substantial changes in landforms have been observed following the implementation of streambed drawdowns in WY 2012.

Topographic Change Methods

Topographic change analyses were completed at four gravel-bar features co-located with clay-horizon marker sites (table D2; fig. D10) including the Unity and Row Tree sites on Fall Creek and the Sand Mountain and Downstream TNC site on the Middle Fork Willamette River. Lidar datasets were used to characterize an initial “baseline” topography at each site, which was compared against topographic-bathymetric lidar data collected September 2015 (Quantum Spatial, 2016; app. 1). For the Unity, Row Tree, and Sand Mountain sites, the initial topographic lidar dataset was acquired in February 2012 (Watershed Sciences, Inc., 2012), and for the Downstream TNC site, the initial topographic lidar surface was acquired March 2009 (Watershed Sciences, Inc., 2009) before annual streambed drawdowns began. Lidar rasters for each site were resampled to 1-m spatial resolution and clipped to a shape coinciding with excluding water depicted in the baseline (2009 or 2012 depending on coverage) topographic lidar dataset. Baseline surfaces were then compared to the 2015 topo-bathymetric lidar with GCD software (version 6.1.14; Riverscapes Consortium, 2018) similar to methods described in the “Magnitude and Distribution of Sediment Erosion and Deposition Water Years 2012–17” section in Chapter B. Reservoir Morphology and Evolution Related to Dam Operations at Fall Creek Lake with direct comparison (raw) and propagated uncertainty analyses. Default errors for topographic (0.15 m for 2009 and 2012 data) and topo-bathymetric (0.18 m for 2015 data) lidar from the GCD software were applied as uniform errors to account for uncertainty in the surveys (overall, minimum detection threshold of 0.23 m) and were used for the propagated threshold analyses. Comparatively, overall project vertical accuracy RMSE reported for the lidar datasets ranges from 0.04 (Watershed Sciences, Inc., 2009) to 0.02 m (Watershed Sciences, Inc., 2012) for the topographic lidar and from 0.03 to 0.04 m (Quantum Spatial, 2016) for the topo-bathymetric lidar (app. 1); therefore, the default error surfaces are considered a relatively conservative estimate of survey error associated with the datasets.

Table D2.    

Deposition, erosion, and volume change summary from topographic change analyses with Geomorphic Change Detection (GCD) software in ArcGIS between 2009–12 and 2015 for select sites along Fall Creek and the Middle Fork Willamette River, Oregon.

[Change analyses completed with Geomorphic Change Detection (GCD) software in ArcGIS. Default uniform error surfaces were used for topographic lidar data (0.15 m) and topobathymetric lidar data (0.18 m). Additional survey details are included in appendix 1 of this study. Average annual net volume change and Annual average net volume change normalized by area: provided for normalization, however, it is likely changes varied annually over the periods analyzed and average estimates are not representative of real conditions. Dates given in month/day/year format. Abbreviations: ±, plus or minus; rkm, river kilometer; cm, centimeter; TNC, The Nature Conservancy]

Table D2.    Deposition, erosion, and volume change summary from topographic change analyses with Geomorphic Change Detection (GCD) software in ArcGIS between 2009–12 and 2015 for select sites along Fall Creek and the Middle Fork Willamette River, Oregon.

Reach	Site	Site description	River kilometer	Initial survey date	Second survey date	Total area analyzed, in square meters	Area of detectable change, in square meters	Percent of area with deposition (thresholded)	Percent of area with erosion (thresholded)	Maximum deposition, in meters	Maximum erosion, in meters	Raw volume change, in cubic meters	Thresholded volume change, in cubic meters	± Error volume, in cubic meters	Percent error	Average annual net volume change, in cubic meters per year	Average annual net volume change normalized by analysis area, in meters per year
Upper Fall Creek	Unity	Large alcove and low floodplain area occupying former side channel behind floodplain island	7.9–7.7	2/23/2012	9/15/2015	24,800	5,590	19	3.1	1.55	−0.93	2,250	1,560	1,040	67	520	0.02
Upper Fall Creek	Row tree	Downstream segment from a low flow alcove and high-flow side channel	6.8–6.4	2/23/2012	9/15/2015	11,900	4,130	30	4.9	1.80	−1.69	1,940	1,470	843	57	490	0.04
Jasper	Sand Mountain	Vegetated gravel bar that grades to low floodplain and alcove along unvegetated bar	17.3–17.0	2/23/2012	9/15/2015	4,080	1,410	34	0.4	1.14	−0.42	792	526	325	62	175	0.04
Clearwater	TNC	Large bare gravel bar that grades into vegetated bar and low floodplain	5.9–5.6	3/2/2009	9/15/2015	20,300	15,500	75	1.2	1.75	−0.99	7,780	7,240	3,580	49	1,207	0.06

Topographic Change Comparison Results

Overall, detectable topographic changes at the four sites along Fall Creek and the Middle Fork Willamette River primarily showed deposition; although, smaller localized instances of erosion were also identified at all sites. Comparison of the 2012 and 2015 lidar at the Unity, Row Tree, and Sand Mountain locations shows broad-scale deposition encompassing substantial parts of the gravel bars and low-elevation floodplain (table D2; fig. D10) with depositional volumes ranging from about 526±325 m³ at the Sand Mountain site to about 1,560±1,040 m³ at the Unity site. Locally, the greatest deposition detected over the 2012–15 measurement period within a single cell (1 m² area) within the Unity, Row Tree, and Sand Mountain sites was 1.55, 1.80, and 1.14 m thick, respectively. At the Downstream TNC site where topographic change was measured for the 2009–15 period, detected volumetric change are much greater, revealing 7,240±3,590 m³ of deposition across the site and local (grid cell) deposition as great as 1.75 m.

At the Unity site, 3.8 km downstream from Fall Creek Dam (table D2; fig. D10A), the majority of detected deposition was distributed along the margins of the small floodplain island and in a small former side channel, whereas erosion occurred in local patches. Both the average depositional depth and average erosional depth across the site was 0.38±0.21 m; however, volumetrically, deposition dominates due to the much larger spatial area where sediment was deposited, resulting in 1,850±1,020 m³ of deposition compared with total erosion of 291±164 m³, indicating an average annual net deposition of about 520 m³/year for 2012–15. A submerged alcove adjacent to the area of change analyses and present in the 2012 lidar was excluded from the measurement area due to a lack of elevation data. This alcove had largely filled with sediment by the time of 2015 topo-bathymetric lidar acquisition (as verified by field observations in WY 2016; fig. D11). Extending the change analyses to include this sediment filled alcove by the time of the 2015 lidar (fig. D10A) would result in at least an additional 340 m³ of deposition for this site. Calculated depositional depths were more than a meter in the alcove without accounting for any deposition below the 2012 water surface. At the Unity site, three clay-horizon markers were included in the topographic change analysis area; depositional changes detected for 2012–15 from lidar topographic analyses ranged from 0.26 to 0.76 m—greater than the 5 mm net deposition measured on the markers in WY 2016 and 2017.

Deposition along the Row Tree site located 5.1 km downstream from Fall Creek Dam was distributed across the broad, heavily vegetated gravel bar and low elevation floodplain, with much of the total area showing no elevation change (fig. D10B). Within areas of the Row Tree site that show changes in elevation, average depositional depths (0.50±0.23 m) were slightly less than the calculated average erosion (0.55±0.23 m) although deposition occurred over a larger area. The total volume change of sediment (1,940 m³) suggests an average annual deposition rate from 2012 to 2015 at the Row Tree site of 490 m³/year—similar to that at Unity site (table D2). An alcove at the downstream end of the site (not included in the analysis area) had undergone deposition after 2012 such that channel areas that were submerged in 2012 lidar had accumulated enough sediment to support deployment of the clay-horizon markers to support measurements of overbank deposition by 2015.

At the Sand Mountain site on Middle Fork Willamette River, 0.8 km downstream from the Fall Creek confluence, local patterns of deposition were highly variable. Sediment depths were greatest at the upstream end of the bar near an entrance to a slough and also along the downstream end of the bar at the head of an alcove, but the largest areal distributions of newly deposited sediment were along the lower and central areas of the main bar (fig. D10C). Average deposition thickness calculated was 0.38±0.23 m, just slightly greater than the average erosional thickness (0.31±0.23 m). Similar to other sites evaluated with lidar comparisons, total erosion volumes were minimal because of the relatively small areas undergoing erosion. Field observations in October of 2015, when the clay-horizon markers were deployed prior to the WY 2016 streambed drawdown, showed that recent fluvially transported sand deposits were locally about a meter in depth. This site continued to accumulate sediment after the WY 2016 streambed drawdown, and at the time of the January 2016 measurement, an additional 260 mm of sand had accumulated on 1 clay-horizon marker. However, despite substantial local deposition, net overall deposition across the bar complex of the Sand Mountain site (175 m³/year) between 2012 and 2015 indicates this site on the Middle Fork Willamette River has undergone less overall topographic change than the Unity or Row Tree sites on Fall Creek (520 and 490 m³/year, respectively).

At the Downstream TNC site, on the Middle Fork Willamette River (fig. D10D), deposition was distributed across the bar with few localized patches of erosion. The average annual depositional thickness across the site is slightly greater than erosional depths at 0.48±0.23 m/year and 0.43±0.23 m/year, respectively. Of the four landform-scale sites evaluated for this study, average annual deposition was the greatest at the Downstream TNC site (1,210 m³/year) by nearly 30–60 percent, even after normalizing for differences in site area and the longer period between surveys (Downstream TNC site spans 2009–2015, whereas other sites span 2012–15). Substantial changes in the channel planform at the TNC site limit change analyses to the 2009 bar area that excludes a large portion of the same site that was present at the time of the 2015 lidar survey. The distance of the Downstream TNC site from the sediments sourced at Fall Creek Lake (greater than 23 km) and the confounding presence of other sediment sources preclude clear linkages between reservoir sediment export from Fall Creek Lake and deposition at the Downstream TNC site. However, sediment sources could not be determined definitively at the TNC site given the presence of fine-grained depositional patches distributed along Fall Creek and Middle Fork Willamette River upstream, relatively low volumes of sediment entering from tributaries (O’Connor and others, 2014), or sediment supplied from bank erosion (as described in the following section).

Channel Mapping Methods and Units

Geomorphic maps of the active channel and adjacent floodplain were developed for the reaches downstream from Dexter and Fall Creek Dams to document reach-scale changes in channel planform that may have resulted from annual streambed drawdowns at Fall Creek Lake. Repeat mapping was conducted for 5 years (2005, 2011, 2012, 2014, and 2016) to characterize planform changes at 1–2 year intervals prior to and after annual streambed drawdowns that began in 2011 (WY 2012). For this study, the active channel was defined as the area typically inundated during annual high streamflows and includes features such as the low-flow channel, side channels, and both vegetated and unvegetated gravel bars. These features likely evolved through frequent bed-material transport prior to dam construction (Church, 2006) but are more stable due to reductions in coarse sediment supply and the magnitude and frequency of peak streamflow events following dam construction (Wallick and others, 2013).

A mapping corridor was defined for this study to provide a static spatial boundary for repeat mapping including the active channel and adjacent floodplain areas. For Fall Creek, the mapping corridor was defined using methods similar to those described in Wallick and others (2011) and includes the Holocene floodplain, extending from Fall Creek Dam to the confluence with the Middle Fork Willamette River. For the Middle Fork Willamette River, the mapping corridor was modified from a previous mapping corridor developed by Jones and others (2016) and encompasses channel-adjacent portions of the Holocene floodplain downstream from Dexter Dam.

The aerial photograph mapping methods used for this study have similar hierarchy and protocols to previous channel mapping studies for other gravel-bed rivers in Oregon, including the Chetco, Umpqua, and Sprague Rivers (Wallick and others, 2010, 2011; O’Connor and others, 2014). Additionally, the mapping presented herein expands upon mapping frameworks developed specifically for the Middle Fork Willamette River in previous studies (Dykaar, 2008; McDowell and Dietrich, 2012; Jones and others, 2016) with increased spatial, temporal, and mapping unit resolution to characterize potentially subtle planform changes that may relate to sediment releases from Fall Creek Lake streambed drawdown operations. The mapping datasets produced for each period include information on reaches, large-scale process domains, landforms, water-features, vegetation density, and vegetation cover (table D3). Features were digitized at 1:2,500 scale in ArcGIS with a minimum mapping unit of 200 m². Additionally, wetted channel centerlines were mapped for the 2005 and 2016 periods to aid in documenting changes in channel position. A minimum of two project team members reviewed all mapping to ensure similar mapping quality among reaches and mapping periods.

Results of Repeat Geomorphic Mapping from Aerial Photography

The active channels of Fall Creek and Middle Fork Willamette River include landforms shaped by annual bed-material transport and scouring streamflows that preclude vegetation encroachment, as well as vegetated bars that were active in the pre-dam era but are currently stabilized with herbaceous and woody vegetation (Keith and others, 2023a). Repeat geomorphic mapping from aerial photography spanning 2005–16 (figs. D12–D14) shows that these landforms include the low-streamflow wetted channel which flows over bedrock and unvegetated gravel bars signifying frequent scour or substrate mobilization. Repeat geomorphic mapping and changes over the longer-term historical scale between 1936 and 2016 are discussed in the companion paper (Keith and others, 2023a); the 1936 mapping is shown repeat mapping figures for this chapter to provide some historical context. Although both rivers occupy primarily single-thread, laterally stable channels, there are substantial differences in alluvial character and channel planform among reaches. Fall Creek occupies a narrow channel (about 40 m wide) that occasionally flows over bedrock with gravel bars and secondary channel features concentrated in lower sections of the mapped reach. In contrast, the Middle Fork Willamette River occupies a broader (about 90–150 m wide), fully alluvial channel throughout the study area with fewer bedrock outcrops (mainly located in Jasper reach) and larger, more abundant gravel bars and secondary water features (especially in downstream reaches). The dominant features within the 2016 active channels of Fall Creek and Middle Fork Willamette River are the main, low-flow wetted channel, which comprise 20–61 percent of the active channel and gravel bars, which comprise 9–29 percent of the active channel mapped for all reaches in the study area. The distribution, areal extent, and vegetative cover of gravel bars are highly variable throughout the study area, ranging from unvegetated gravel bars to bars covered in dense, mature woody vegetation. The reaches immediately downstream from Fall Creek and Dexter Dams (Upper Fall Creek and Dexter reaches, respectively) have relatively narrow active channels with smaller, intermittent unvegetated gravel bars (fig. D12; app. 3) compared with downstream reaches. With increasing distance from the dams, the active channels of each river widen and unvegetated bars become larger and more abundant (fig. D12; app. 3). Floodplain islands and bedrock outcrops were less common in the study area than gravel bars, with bedrock especially prominent on the Upper Fall Creek and Jasper reaches whereas floodplain islands were most common in the Dexter and Dougren reaches (fig. D12; app. 3). Mapping results are further described by reach.

Temporal Changes in Bar Landforms and Secondary Water Features by Reach, Water Years 2005–16

Upper Fall Creek Reach

The Upper Fall Creek reach (fig. D15) was laterally stable between 2005 and 2016 with most variation in mapped features owing to changes in the locations, morphology, and vegetative cover of gravel bars. Bar landforms flanking the main channel were predominantly covered in herbaceous and woody vegetation during that period, as well (figs. D12, D13, D15; app. 3). From 2005 to 2016, total bar area (including vegetated and unvegetated bars) increased by about 9 percent, from 64,300 to 70,400 m², which was primarily driven by increases in vegetated bar area as areas that mapped as main channel or secondary water features in 2005 became bars with either herbaceous or woody vegetation by 2016 (fig. D13; table D5; app. 3). From 2005 to 2011, prior to implementation of annual streambed drawdowns, changes in mapped bar areas were negligible and likely within the range of mapping uncertainty. From 2011 to 2012 (spanning the WY 2012 streambed drawdown), mapped bar area increased by 2,500 m² (or about 4 percent), due mainly to increases in unvegetated bars and bars with herbaceous cover. The unvegetated gravel bars mapped in the 2012 photographs were not detected in the 2014 photographs, leading to small decreases in overall mapped bars (about 2 percent) during that period, which spanned two streambed drawdowns. Between 2014 and 2016, also encompassing two streambed drawdowns (WY 2014 and 2015), mapped vegetated and unvegetated bar area collectively increased by 5,200 m² (about 8 percent; fig. D13; app. 3).

Table D5.    

Landform area, unit area (landform area divided by reach length), and percent of landform within the whole active channel area by reach for geomorphic mapping along Fall Creek and the Middle Fork Willamette River, Oregon, from geomorphic mapping from 2016 aerial photography.

[Abbreviations: km, kilometer; m², square meters; m²/m, square meters per meter; --, no data]

Table D5.    Landform area, unit area (landform area divided by reach length), and percent of landform within the whole active channel area by reach for geomorphic mapping along Fall Creek and the Middle Fork Willamette River, Oregon, from geomorphic mapping from 2016 aerial photography.

Mapping unit		Fall Creek			Middle Fork Willamette River
		Reach
		Upper Fall Creek	Lower Fall Creek*	All of Fall Creek*	Dexter	Dougren	Jasper	Clearwater	All of the Middle Fork Willamette River
Centerline	Reach length (km)	6.0	5.1*	11.1	5.9	3.5	6.7	11.2	27.2
Active channel Area	Total area (m2)	246,600	258,900	506,000	971,800	972,000	806,000	1,703,000	4,453,000
	Unit area (m2/m)	41	51	46	166	280	120	153	164
All bars	Total area (m2)	70,400	68,100	138,000	88,800	202,000	106,000	341,000	738,000
	Unit bar area (m2 bars/m centerline length)	11.7	13.4	12.5	15.2	58.2	15.7	30.6	27.1
	Percentage active channel	28.55	26.30	27.27	9.14	20.78	13.15	20.02	16.57
Unvegetated bars	Total area (m2)	412	4,640	5,050	15,000	28,000	19,200	101,000	164,000
	Unit bar area (m2 bars /m centerline length)	0.1	0.9	0.5	2.6	8.1	2.8	9.0	6.0
	Percentage active channel	0.17	1.79	1.00	1.54	2.88	2.38	5.93	3.68
Vegetated (woody and herbaceous) bars	Total area (m2)	69,970	63,500	133,000	73,800	173,500	87,200	239,400	574,000
	Unit area (m2/m)	11.8	12.4	12.0	12.6	50.0	12.9	21.4	21.1
	Percentage active channel	28.37	24.53	26.28	7.59	17.85	10.82	14.06	12.89
Bedrock	Total area (m2)	5,460	3,980	9,430	790	0	511	1,590	2,900
	Unit area (m2/m)	0.9	0.8	0.9	0.1	--	0.1	0.1	0.1
	Percentage active channel	2.21	1.54	1.86	0.08	--	0.06	0.09	0.07
Floodplain Islands	Total area (m2)	17,900	32,900	50,700	333,000	372,000	105,000	92,400	902,000
	Unit area (m2/m)	3.0	6.4	4.6	56.9	107.1	15.6	8.3	33.1
	Percentage active channel	7.26	12.71	10.02	34.27	38.27	13.03	5.43	20.26
Main wetted channel	Total area (m2)	141,000	138,000	280,000	457,000	198,000	495,000	673,000	1,822,000
	Unit area (m2/m)	23.7	27.0	25.3	78.0	57.0	73.5	60.3	66.9
	Percentage active channel	57.18	53.30	55.34	47.03	20.37	61.41	39.52	40.92
Secondary channel features	Total area (m2)	1,880	11,300	13,200	48,900	118,000	65,900	110,000	343,000
	Unit area (m2/m)	0.3	2.2	1.2	8.4	34.0	9.8	9.9	12.6
	Percentage active channel	0.76	4.36	2.61	5.03	12.14	8.18	6.46	7.70

^*

The lower 0.54 km of the Fall Creek centerline Falls within the Dougren reach floodplain. All features along this segment are included in the Dougren reach and the reach length for Lower Fall Creek reach is shorter than that used in table A2.

Some of the adjustments in bar area between 2005 and 2016 likely reflect higher discharges represented in the aerial photographs from 2005 and 2014 compared with other years. For example, no unvegetated gravel was mapped in 2014 when aerial photograph discharge was 415 ft³/s, whereas mapped unvegetated gravel was greater in 2012 and 2016 when discharge was lower (150 and 225 ft³/s, respectively; table D4; fig. D13). However, the 2011 and 2016 aerial photographs were collected at relatively similar discharges (222 and 225 ft³/s, respectively), so planimetric changes from those periods are expected to represent actual geomorphic change; the maps show very similar bar features with main change being an increase in woody vegetation (table D4; fig. D13). Considering the relatively modest changes in the mapped bars and various sources of mapping uncertainty, the repeat mapping of bars on Upper Fall Creek reach indicates that widespread bar building and bar growth did not occur in the period spanning five streambed drawdowns (WY 2012–16) and that the main changes potentially attributable to sediment releases from Fall Creek Lake were localized increases in vegetated bar area, particularly where unvegetated channel margins mapped in 2005 and 2011 were converted to herbaceous or woody bars in subsequent mapping years.

Although secondary water features are sparse and relatively small along Upper Fall Creek reach (compared with other study reaches), these features underwent major losses in area between 2005 and 2016 (figs. D14, D15; app. 3). The total area of all secondary water features decreased by 52 percent (1,990 m²) between 2005 and 2016, and mainly occurred through an 89 percent reduction in the area of alcoves between 2014 and 2016 and much of that attributable to losses in a single feature at the Unity site (rkm 7.9–7.7; figs. D14, D15; app. 3). The losses in secondary water features at the Unity site also coincide with locations of calculated net deposition from lidar comparison between 2012 and 2015 (table D2; fig. D10A). The apparent net loss in secondary water features during the 2005–16 period may partly owe to lower discharges at the time of aerial photograph acquisition in 2016 compared with 2005 (464 and 225 ft³/s, respectively; table D4). Likewise, the modest decrease in area of main wetted channel between 2005 and 2016 (less than 3 percent) could also relate to lower aerial photograph discharge in 2016; however, actual narrowing of the wetted channel where bar areas increased, and mapping uncertainty related to channel margins being obscured by vegetation or shadows may have also contributed this decrease. Despite various sources of mapping uncertainty, the mapped losses in secondary water features were validated with visual inspection of aerial photographs and field visits, suggesting that at least some of the mapped loss in secondary water features relates to deposition.

Lower Fall Creek Reach

Like the Upper Fall Creek reach, the Lower Fall Creek reach (fig. D16) also maintained a stable planform with similar locations and sizes of channel features between 2005 and 2016. Bars within this reach were dominated by woody-type vegetative cover, but unlike the Upper Fall Creek reach, secondary water features primarily consisted of side channels rather than alcoves. Net changes between 2005 and 2016 included an increase in total bar area of 14.8 percent (8,800 m²) and 7 percent decrease (770 m²) in the area of secondary water features (fig. D13; app. 3), although the area of both features fluctuated slightly over time. From 2005 to 2011, before the streambed drawdowns were implemented, total area of mapped bars decreased by 3.7 percent (similar to the changes on Upper Fall Creek reach 2005–11). From 2011 to 2012, spanning the WY 2012 streambed drawdown, total mapped bar area increased 9.6 percent (from about 61,400 to 67,400 m²), mainly through increases in unvegetated bars. In 2011, 2012, 2014, and 2016 (spanning multiple drawdowns), the area of vegetated bars remained approximately constant but the area of unvegetated bars in each period fluctuated substantially. From 2011 to 2012, unvegetated bar area increased 259 percent (4,110 m²), followed by a 57 percent decrease from 2012 to 2014 (3,260 m²), coinciding with higher discharge in 2014 aerial photographs and 90 percent increase from 2014 to 2016 (2,200 m²). Considering only mapping from periods with similar discharge (2011 and 2016; table D4), increases in total unvegetated bar areas from 1,590 to 4,640 m² indicates an actual net increase in bars following the implementation of streambed drawdowns (fig. D13; app. 3). Mapped area of secondary water features increased 10 percent from 2005 to 2011, then decreased by 24 percent from 2011 to 2014 and increased 11 percent from 2014 to 2016 (app. 3). The decrease in mapped secondary water features 2012–14 and subsequent increase 2014–16 was primarily attributable to a fluctuation of inundation extent within single feature (side channel at rkm 2.0 downstream from Pengra Road bridge). This side channel was mapped as part of the main channel in 2014 because higher discharge at the time of aerial photograph acquisition obscured a portion of a bar that separates these channel features at low streamflow conditions when subsequent aerial photographs were acquired.

Considering the overall trends in bar and channel features amid mapping uncertainty and variation in discharge at the time of aerial photograph acquisition (table D4), the main planimetric changes from 2011 to 2016 on Lower Fall Creek reach that coincide with the Fall Creek Lake streambed drawdowns include (1) a reduction in the area of secondary water features by 6.4 percent reflecting decreases in the area of side channels, local conversions of side channels to alcove (7.9–7.7; figs. D12, D14, D16), and local adjustments in the perimeter of main and side channels, and (2) a nearly twofold increase in the area of unvegetated bars, reflecting bar growth near the confluence with Little Fall Creek while vegetated bar area remained approximately similar.

Dexter Reach

Repeat mapping of the Dexter reach of Middle Fork Willamette River from 2005 to 2016 shows that this reach (downstream from Dexter Dam and upstream from Fall Creek confluence) had a stable channel planform with minor changes in unvegetated gravel bars and secondary water features (fig. D17). Mapped gravel bars in the Dexter reach were predominantly covered in herbaceous vegetation (fig. D13), and secondary water features were predominantly alcoves (fig. D14; app. 3). From 2005 to 2016, the net change in total area of mapped bars was minor (6.4 percent increase, or 5,320 m²) although the total area of gravel bars between each mapping period varied by up to 27 percent (fig. D17; app. 3). That variation mainly occurred in the unvegetated gravel bars and likely reflects differences in discharge in the underlying aerial photographs. For example, the 59 percent decrease in unvegetated bar area (10,100 m²) between 2005 and 2011 may reflect higher streamflow in the 2011 photographs (3,300 ft³/s compared with 2,290 ft³/s) and the subsequent 108 percent increase in unvegetated bar area (7,760 m²) from 2011 to 2012 may reflect increased bar exposure at a lower discharge (1,840 ft³/s) in 2012 (table D4; app. 3). The mapped area of secondary water features in the Dexter reach increased by about 7 percent (3,290 m²) between 2005 and 2016, but much of this change is attributable to features near rkm 24.6–24.1 (fpkm 21) where bar growth led to expansion of an alcove behind the bar and creation of a side channel. Because the Dexter reach is not affected by streambed drawdowns at Fall Creek and there are limited sources of sediment downstream from the dam in this reach, the minor changes in mapped water features most likely reflect differences in aerial photograph discharge as field observations did not reveal substantial changes in water features. For example, the years with greatest area of mapped water features (2011 and 2014) were also years with substantially greater discharge (table D4).

Dougren Reach

The Dougren reach of Middle Fork Willamette River is immediately upstream from Fall Creek confluence and although much of the reach was laterally stable between 2005 and 2016 (fig. D18), channel changes near the confluence area resulted in mapped changes in the reach-aggregated area of gravel bars and wetted features. Throughout the 2005–16 period, gravel bars on the Dougren reach were predominantly covered by woody-type vegetation, with very few mapped unvegetated bars (fig. D13), and side channels were the main type of secondary water features (fig. D14; app. 3). Net changes in mapped features from 2005 to 2016 included a 17 percent increase (28,900 m²) in the total area of mapped bars when the area of unvegetated bars more than doubled (17,600 m²), and smaller decreases in herbaceous bars and increases in woody bars were mapped (fig. D13; app. 3). Area of mapped secondary water features decreased by 14 percent during that same period (19,900 m²), which was mainly attributable to changes at the confluence with Fall Creek and could reflect the wider active channel in this location and the presence of actively shifting mid-channel bars. Though local changes in gravel bars and secondary water features that owe to erosion and deposition are apparent from the mapping, the reach-aggregated trends in bar area almost certainly also reflect variation in discharge at the time of aerial photograph discharge rather than reach-wide channel change because the years, with greatest area of unvegetated gravel bars (2012 and 2016) coinciding with lowest discharges at the time of aerial photograph acquisition (table D4; figs. D13, D14).

Jasper Reach

The Jasper reach of the Middle Fork Willamette River is immediately downstream from the confluence with Fall Creek and was laterally stable with few changes in overall planform between 2005 and 2016 (fig. D19). Like most other reaches of the study area, gravel bars on the Jasper reach were predominantly covered with woody vegetation and secondary water features were primarily side channels (figs. D13, D14; app. 3). Most of the changes in gravel bars from 2005 to 2016 were mapped in the upper part of Jasper reach between rkm 17.9 and 15.3 (fpkm 15.5 and 13.5), while the lower portion of the reach that flows along the base of hillslopes and adjacent to bedrock outcrops was more stable (fig. D19). From 2005 to 2016, the total area of mapped bars increased by 8.9 percent (8,690 m²), when the area of unvegetated bars more than tripled (from 5,690 to 19,200 m²) and smaller increases (10 percent) in woody vegetated bars and decreases (40 percent) in herbaceous bars occurred (fig. D13; app. 3). Between 2005 and 2011 (prior to the first drawdown in autumn of 2011 [WY 2012]), the area of unvegetated bars increased by 83 percent (4,750 m²; fig. D13; app. 3). Trends in unvegetated bar area following the implementation of annual streambed drawdowns include a 112 percent increase in area from 2011 to 2012 (11,700 m²), a 25 percent decrease in area from 2012 to 2014 (5,620 m²), and a 17 percent increase in area from 2014 to 2016 (2,710 m²). Trends in total area of secondary water features include an 11 percent increase in area from 2005 to 2011, less than 9 percent increases for the 2011–12 and 2012–14 periods, and a 14 percent decrease from 2014 to 2016 (fig. D14; app. 3). Because discharges in the aerial photographs in 2012 and 2016 post-dating drawdown implementation were similar (varying by less than 190 ft³/s; table D4), the mapped changes in bars and secondary water features between 2012 and 2016 likely reflect actual geomorphic changes as well as other sources of uncertainty introduced during mapping.

Considering overall mapping results from the Jasper reach and associated uncertainty, key findings of geomorphic changes between 2005 and 2016 include: (1) negligible changes in the location and areas of vegetated bars and the main wetted channel indicate overall geomorphic stability of the reach in the present (post-dam) streamflow and sediment regime, and (2) the main detectable change from the repeat mapping was a twofold increase in unvegetated bars partially owing to substantial increases between 2005 and 2011 (pre-dating streambed drawdowns at Fall Creek Lake) followed by another twofold increase in unvegetated bar area between 2011 and 2012 (spanning the first streambed drawdown). Negligible changes in bars and secondary water features in the lower Jasper reach (downstream from about rkm 15.0) also suggest that reach-scale aggradation and associated geomorphic responses are not occurring in response to sediment releases from Fall Creek Lake in this segment of the Middle Fork Willamette River. Where increases in unvegetated bars were most apparent, changes were concentrated in a more laterally active segment that coincides with lateral channel change near the confluence of Fall Creek (where mapping from the Dougren reach also shows greater amounts of channel change) and pre-date the Fall Creek Lake streambed drawdowns; it is therefore difficult to attribute changes in gravel bars after 2012 solely to deposition from Fall Creek Lake streambed drawdowns. However, the concentration of bars at upstream end of the Jasper reach near the Fall Creek confluence, and persistence of these bars over time, likely provides suitable areas for deposition of reservoir sediment.

Clearwater Reach

Although much of the Clearwater reach of the Middle Fork Willamette River was laterally stable from 2005 to 2016 (fig. D20), local segments were laterally dynamic and reach-aggregated changes in the area of gravel bars and secondary water features were the greatest in this reach compared with other reaches in this study. Like other reaches, gravel bars were predominantly covered with woody vegetation, but the Clearwater reach also had a substantially greater proportion of unvegetated bars compared with other reaches (comprising about 30 percent of total bar area in 2016, figs. D12, D13; app. 3). Between 2005 and 2016, the total area of bars increased by 18.5 percent (53,100 m²), due mainly to increases in unvegetated bars (82,300 m²) while herbaceous and woody bars underwent net losses in area of 19,800 m² and 9,390 m² (respectively; fig. D13; app. 3). Between 2005 and 2011, prior to streambed drawdowns, at Fall Creek Lake there was a minor (4.6 percent, 13,300 m²) decrease in total bar area, yet the mapped area of unvegetated bars increased by 23,300 m² despite higher discharge in the 2011 aerial photographs (table D4) that would bias these features toward smaller areas if no actual geomorphic change had occurred, whereas herbaceous and woody bars decreased by 26,900 and 9,710 m², respectively. For the mapping periods bracketing the first full streambed drawdown at Fall Creek Lake (2011 and 2012), the 18 percent increase in total bar area (50,000 m²) was mainly due to mapped increases in unvegetated bars. From 2012 to 2014, unvegetated gravel bar area decreased by 30,100 m² followed by an increase of 35,000 m² during the 2014–16 period. Although the mapped area of secondary water features within the Clearwater reach increased by 18 percent (16,700 m²) from 2005 to 2016, the 2005–11 and 2011–12 mapping intervals each show minor losses in these landforms (app. 3; fig. D14). The largest increase reach-aggregated area of mapped secondary water features occurred between 2012 and 2014 (14 percent, 12,700 m²), with additional increases in mapped area observed between 2014 and 2016 (fig. D14; app. 3). Although minor fluctuation in aerial photograph discharge (table D4) may partly explain some of the variation in mapped area of gravel bars and secondary water features, distinct changes in locations of gravel bars and perimeter of side channels are apparent from the aerial photographs, suggesting that some of the computed changes relate to local erosion or deposition. For example, lateral channel shifting resulted in erosion of a vegetated bar between 2005 and 2016 at rkm 7.0–6.7, while at the same time, the area between rkm 7.1 and 6.8 underwent deposition and bar growth (fig. D20).

Altogether, the repeat mapping of the Clearwater reach from 2005 to 2016 shows that this reach has laterally active areas where modest, but detectable, bank erosion along bars covered in herbaceous and woody vegetation is coupled with steady increases in the area of unvegetated bars. Despite uncertainty in the repeat channel maps, increases in unvegetated bars from 2005 to 2016 are approximately matched by losses in vegetated bars over the same period, indicating sediment released from Fall Creek Lake streambed drawdowns is not the only source of sediment contributing to bed-material deposition and bar growth in this reach. Also, similar to the Jasper reach, the Clearwater reach underwent increases in unvegetated bars (and losses in vegetated bars) between 2005 and 2011—prior to the Fall Creek Lake annual streambed drawdowns—suggesting that erosion and re-deposition of gravel stored in channel-flanking bars is an ongoing process pre-dating streambed drawdown operations.

Fall Creek Lake

Fall Creek Lake (fig. A6) was created when Fall Creek Dam was constructed in 1965, impounding Fall Creek and resulting in the inundation of relatively narrow (50–1,000 m-wide) valleys bisected by Fall and Winberry Creeks (figs. A7, B4A). Aerial photographs from 1936 and other historical datasets show that steep, valley-edge hillslopes confining the reservoir were heavily forested (fig. B4); the valley floor was dominated by agricultural and residential areas in the mid-19th century prior to dam construction; and that much of Fall Creek flowed on bedrock, with occasional gravel bars and patches of coarse substrates (Keith and others, 2023a). After nearly six decades of impoundment, the valley floor had evolved to reservoir floor with predominantly fine-grained, unconsolidated sediment deposits mantling the pre-dam topography. Within the reservoir floor, thicker deposits (about 3 m) were located in the lower reservoir near the dam where pre-dam topography is completely obscured by overlying sediments (fig. B5B). Repeat surveys and detailed geomorphic mapping in the reservoir show that reservoir deposits thin upstream (or up-reservoir), and many pre-dam valley features and infrastructure remain identifiable despite varying amounts of burial by sediment (fig. B5). Exposed in-channel bedrock and overall channel stability upstream from minimum pool indicate fluvial processes are still active, at least intermittently, within the stream channel, and sand and finer-grained sediments deposited during lacustrine conditions at higher lake levels can be mobilized from the upper reservoir to the lower reservoir or downstream from the dam. Although the Fall Creek stream channel is presently (during the period of this study) flowing through a reservoir and is bordered by broad reservoir floor surfaces that evolve by lacustrine sedimentation, the stream channel in the middle and upper reservoir was historically and is presently a stable bedrock channel with intermittent gravel bars that may be episodically mobilized.

Fall Creek

Downstream from Fall Creek Dam, Fall Creek flows through a relatively confined valley (compared with the Middle Fork Willamette River) reflected by a narrow, semi-alluvial channel that efficiently conveys water and sediment at typical regulated streamflows. Unlike many gravel-bed rivers below large dams (Grant, 2012), the morphology of Fall Creek has changed little in the six decades following dam construction (Keith and others, 2023a). The channel within the study area had an overall planform and distribution of landforms with little change between 1936 and 2016 (figs. D15, D16). The most substantial morphologic change detected over this period was a reduction in unvegetated gravel bars mainly attributed to vegetation encroachment (94 percent loss in the Upper Fall Creek reach; 63 percent loss in the Lower Fall Creek reach; Keith and others, 2023a). As of 2016, the channel has both bedrock and alluvial segments with greater alluvial cover downstream from the confluence with Little Fall Creek (figs. D12–D13; app. 3). The presence of in-channel bedrock (fig. D12) and lack of full alluvial cover on the channel bed in 1936 and 2016 suggests transport capacity along Fall Creek has historically and recently exceeded sediment supply, particularly in the Upper Fall Creek reach between Fall Creek Dam (rkm 11.6–11.8) and near the confluence with Little Fall Creek (rkm 5.6), where in-channel bedrock is most extensive.

Middle Fork Willamette River

Downstream from Dexter Dam, the Middle Fork Willamette River flows through a relatively less confined valley than Fall Creek and is a large, gravel-bed river that remains a fully alluvial river notwithstanding substantial transformations in channel morphology and reduction in lateral dynamism following the construction of three upstream dams. The Middle Fork Willamette River Basin was historically one of the largest producers of streamflow and gravel of all Willamette River Basin tributaries (Gregory and others, 2007; O’Connor and others, 2014). The lower 27.5 km of the Middle Fork Willamette River encompassing the reaches downstream from Dexter Dam were laterally dynamic with multi-thread and single-thread channels flanked by large, shifting gravel bars prior to dam construction (panel A of figs. D17–D20; Keith and others, 2023a). After streamflow regulation and other channel modifications in the mid-20th century, the lower reaches of the Middle Fork Willamette River have become more stable and encompass a narrower floodplain corridor. By 2016, former gravel bars were converted to low-elevation floodplains colonized by young, dense forests that flank much of the present-day channels (fig. E1; Keith and others, 2023a). Compared with Fall Creek study reaches, the Middle Fork Willamette River study reaches have about twice the area of bars per unit length of channel, 13 times the area of unvegetated gravel bars per unit length of channel, and 11 times the area of secondary water features per unit length of channel (in 2016; fig. D12; table D5). Although all study reaches along the Middle Fork Willamette River have a much greater abundance of alluvial related features (bars, secondary water features, and floodplain islands) than Fall Creek, these reaches have relatively smaller and fewer actively shifting gravel bars than other unregulated, fully alluvial large rivers of western Oregon (O’Connor and others, 2014). For example, the area of mapped gravel bars per unit length of channel for fluvial reaches of the Rogue River was 63.2 m²/m (Lobster Creek reach in 2009; Jones and others, 2012) and the Chetco River was 51.6 m²/m (Chetco River in 2008; Wallick and others, 2010), whereas mapped bar area per unit length channel on Fall Creek downstream from Fall Creek Dam was 12.5 m²/m and 27.1 m²/m in the Middle Fork Willamette River downstream from Dexter Dam in 2016 (table D5).

Reservoir Sedimentation, Streambed-Drawdown Related Erosion, and Sediment Supply to Downstream Reaches

Findings from the reservoir mapping and analyses are consistent with regional estimates of sediment yield (O’Connor and others, 2014; Wise and O’Connor, 2016; Wise, 2018; Wise, 2020; O’Connor and others, 2021), and pre-dam (USACE, 1954) studies indicating that, although reservoir sediment accumulation in Fall Creek Lake is relatively low compared with other U.S. reservoirs such as the Missouri River Basin (for example, Graf and others, 2010; Reservoir Sedimentation Database [RESSED] database, https://water.usgs.gov/osw/ressed/db_doc2013/index.html), reservoir sediment is concentrated near the dam where it is most susceptible to erosion and export during streambed drawdown operations. The spatial distributions and thicknesses of sediment deposited in the reservoir reflect many factors including pre-dam morphology, sediment supply to the reservoir (and for example, relation to land use changes or fire), location of the regulating outlets, historical streambed drawdowns, longer-term typical operations for seasonal flood management, and annual streambed drawdowns. Comparison of historical datasets (pre-dam) to more recent topography (2012 and 2016) shows most of the reservoir deposition is concentrated in the lower reservoir at elevations below minimum conservation pool (728 ft). Annual streambed drawdown events since WY 2012, in combination with earlier intermittent streambed drawdowns (fig. A12), have helped to create the conditions necessary to maintain a well-defined and actively evolving channel in much of Fall Creek Lake. During periods of low pool levels (including streambed drawdowns), bedload transport and re-deposition of coarse-grained sediment is mainly constrained to the channel. Higher pool levels create lacustrine conditions that support deposition of sand and finer-grained sediment on the former pre-dam floodplain and terrace surfaces (presently, the reservoir floor during the period of this study) and lowermost areas of the reservoir. Hence, most of the reservoir floor is mantled by unconsolidated sands and finer material. Upstream from minimum conservation pool where the reservoir landforms are often exposed during seasonal flood-control operations coinciding with winter floods, the Fall Creek channel continues to undergo active bedload transport and has high bedload transport capacities relative to incoming sediment supply as indicated by bedrock and gravel reaches within the reservoir main channel. The broader morphology of the upper reservoir largely reflects pre-dam topography, as surfaces outside of the main channel have been minimally buried by reservoir sedimentation; the overall lack of thick, erodible fine-grained sediment deposits and distance of those deposits from the dam or other tributary and drainage channels together indicates that the upper reservoir only contributes minor amounts of sediment to reaches downstream of the dam during streambed drawdowns.

During streambed drawdown operations at Fall Creek Lake of WY 2012–18, pool levels were lowered nearly 15 m (50 ft) below minimum conservation pool (728 ft). Incremental lowering of pool levels, and ultimately a free-flowing channel, provided a suite of conditions for erosive processes to interact with sands and finer sized, unconsolidated sediment deposits in the lower reservoir that have accumulated in preceding decades. Drawdown-influenced erosion through the reservoir floor deposits is mainly limited to areas adjacent to the main reservoir channel (and to some degree within drawdown channels), as most of the reservoir floor is topographically higher than or farther from the main channel during streambed drawdown and inaccessible to major erosion and reworking. Areas immediately upstream from the dam, where reservoir sedimentation has buried the pre-dam channel, are subject to substantial downcutting and lateral migration through sands and finer-grained material during streambed drawdowns. When the reservoir was completely drained and streamflows through the reservoir were confined to the main channel, broad areas of reservoir floor deposits were largely inaccessible to erosive flows, because those deposits are perched on paleo-floodplain and terraces. However, drawdown-related landforms such as channels, terraces, and distributary zones developed within the broader reservoir floor as a result of streambed drawdowns show evidence of sediment entrainment and mobilization. Although the stream channel may entrain gravel and other clasts larger than sand that are transported as bed-material during annual streambed operations or other periods of lowered lake levels (such as during flood control operations in winter months), much of the gravel transport is occurring over highly resistant channel bed because many sections of the main channel flow on or adjacent to bedrock, particularly upstream from minimum pool (figs. B9–B10A).

These observations support a geomorphic framework in which the processes shaping reservoir sedimentation and erosion vary between the different zones of the reservoir according to their proximity to the main reservoir channel and the varying depositional and erosional processes active at different dam management and streamflow scenarios. In areas immediately upstream from the dam where sand and finer-grained sediment accumulation is thick and readily accessible to fluvial erosion and transport, the direct controls on erosion are dam operations (predominantly lake level and changes in lake level) and streamflow (unregulated streamflow entering the reservoir). Sediment eroded from the area proximal to the dam and main streambed drawdown channel is more likely to be transported downstream from Fall Creek Dam during streambed drawdowns than sediment recently mobilized from reservoir margins because of its proximity to the regulating outlets. Reservoir sediments stored farther upstream in the reservoir, and on higher elevation margins of the reservoir floor are thinner than deposits proximal to the dam and reflect underlying pre-dam floodplain and terrace surfaces; these sediment deposits must be eroded and routed farther through the reservoir prior to export. Channel stability imposed by localized occurrences of bedrock limits lateral migration, channel widening, and incision that might otherwise be anticipated from rapid lowering of lake levels and erosive conditions potentially created by this operation (for example, Morris and Fan, 1998; Major and others, 2012; Kondolf and others, 2014).

Downstream Responses to Reservoir Sediments Released During Streambed Drawdowns–Patterns of Sediment Transport and Effects on Channel Morphology

Sediment loads quantified from the sediment transport monitoring program provide insights regarding reservoir sediment erosion and export from the Fall Creek Lake during streambed drawdowns of WY 2013–18, including the amount and character of sediment exiting the reservoir, and temporal trends in the downstream transport of reservoir sediment (Schenk and Bragg, 2014, 2021; USGS, 2021). Sediment transport measured at the Fall Creek streamgage is primarily sourced from sediment evacuation from the reservoir during the streambed drawdowns because there is limited sediment sources or storage in the narrow, steep, and short 1.4-km-segment between Fall Creek Dam and the streamgage site. Measurements of sediment transport in WY 2013–18 reveal the distribution of reservoir sediments was primarily sand and finer-grained particles (less than 2 mm) that traveled in suspension. Some coarser grains (largest measured was 32 mm) were also mobilized and transported downstream during the streambed drawdowns, although the coarse sediment likely comprise a very small portion of total sediment load (Chapter C. Sediment Delivery from Fall Creek Lake and Transport through Downstream Reaches).

The total amount of reservoir sediment eroded from Fall Creek Lake and delivered to downstream reaches as a direct result of the annual streambed drawdowns from WY 2013 to 2018 was more than 80,000 t (Chapter C. Sediment Delivery from Fall Creek Lake and Transport through Downstream Reaches; table C1). Although the exact amount of sediment delivery to downstream reaches due to streambed drawdowns is challenging to quantify without data from WY 2012 and year-round sediment transport monitoring in subsequent years, the available sediment transport measurements, together with net changes in reservoir topography, reveal insights to the total net volumes of sediment delivered from the WY 2013–18 operations and annual variation in reservoir sediment evacuation. The total net reservoir erosion (146,000–221,000 t; Chapter B. Reservoir Morphology and Evolution Related to Dam Operations at Fall Creek Lake; table B1) represented in the WY 2012 lidar and WY 2017 structure from motion (SfM) datasets spans nearly five full streambed drawdowns and part of the WY 2012 drawdown. This results in annual average rate of erosion of 29,200–44,200 t/year over 5 years, which is similar in magnitude to the SSL calculated for WY 2013 (40,200 t) but greater than SSL calculated for later years (4,500–11,300 t/year; Chapter C. Sediment Delivery from Fall Creek Lake and Transport through Downstream Reaches; table C1).

The patterns of sediment transport measured at Fall Creek streamgage also reflect aspects of the streambed drawdown operations and subsequent sequence of regulated streamflows released from Fall Creek Dam. Drawdown operations in WY 2013–18 coinciding with sediment transport monitoring typically lasted 1–2 weeks, occurred in late autumn or early winter, and were followed by several months of high streamflows (albeit regulated; Schenk and Bragg, 2021). Reservoir sediment evacuated from Fall Creek Lake can be temporarily stored in the short reach between Fall Creek Dam and Fall Creek streamgage, but this supply appears to be exhausted annually based on temporal patterns of sediment transport (decreasing transport for a given discharge), field observations, and repeat geomorphic mapping (described in “Results of Repeat Geomorphic Mapping from Aerial Photography” section). The length of time required to deplete the annual reservoir sediment supply delivered downstream from the dam varied during WY 2013–18 from a few weeks to several months depending on the magnitude of sediment delivery and the magnitude and duration of streamflows in the ensuing months after the streambed drawdown (Schenk and Bragg, 2021). By spring of each year during the suspended-sediment monitoring period, SSL typically were small, even during moderate to high streamflows. Likewise, SSL measured at the Fall Creek streamgage vary substantially over time with sequence of drawdown operations, length of time monitored in each year, operational aspects in how the drawdown and near-drawdown conditions were implemented at Fall Creek Dam, reservoir inflow and outflow, climatic conditions (for example, during freezing conditions in WY 2014), and length of time Fall Creek was fully at streambed in the reservoir. Despite year-to year variability, an overall decrease in suspended-sediment loads from the reservoir has been observed since the largest measured loads in WY 2013, which is about 4–9 times greater than loads measured in WY 2014–18.

In gravel-bed channels, sediments carried as bedload form important morphological and habitat features (such as gravel bars and riffles). At Fall Creek, questions arose regarding whether sediment releases from Fall Creek Lake could supplement downstream bedload and potentially offset reductions in coarse sediment resulting from trapping by the dam. Bedload samples at the Fall Creek streamgage revealed that sediment traveling as bedload was predominantly sand, as opposed to the coarser gravels and cobbles that compose the channel bed and bar features. Also, bedload discharge rates were low (less than 140 kg/min) for WY 2013 and 2017, even for measurements collected during free-flowing conditions in the reservoir (no residual pool) during streambed drawdown operations, indicating minimal transfer of coarse gravel substrate through the dam. An inverse pattern of decreasing bedload transport with increasing streamflow over time likely reflects an exhausted, finite supply of bedload sediment released from the reservoir but may also suggest a change in streamflows where fine sand is converted to suspended load with increasing discharge. Altogether, the bedload measurements (together with field observations and geomorphic mapping downstream from Fall Creek Dam) show that (1) the mixed bedrock and coarse alluvial channel between the dam and the Fall Creek streamgage lacks coarse-grained sediment that is transportable during typical regulated winter streamflows, (2) the bedload released from the reservoir during streambed drawdowns is primarily sand, (3) bedload is rapidly exhausted in the weeks and months following the streambed drawdown, and (4) Fall Creek sediment transport is supply-limited following nearly six decades of upstream trapping of coarse sediment by Fall Creek Dam despite substantial concomitant reductions in the magnitude of peak streamflows through dam regulation and the relatively small additions of sand and fine-grained sediment supplied during streambed drawdowns.

In the days, weeks, and months following streambed drawdown operations at Fall Creek Lake, reaches downstream from the dam displayed a variety of geomorphic responses to reservoir sediment delivery. Immediately downstream from Fall Creek Dam, the upper segments of the main channel of Fall Creek became initially blanketed with the large supply of reservoir sediment during streambed drawdown operations delivered in late autumn (typically November-December for WY 2012–18). In addition to high SSL computed from the sediment monitoring program, sands and finer-grained deposits of sediment have been observed within the channel (fig. D2), and temporary increases in bed elevation have been observed (fig. D1). But over the course of the ensuing winter high-streamflow season, transport capacity was sufficiently high enough to clear these sands and finer-grained sediment from the main channel areas of Fall Creek and any minor increases in bed elevation near the Fall Creek streamgage that resulted from drawdown sediment dissipated by spring. The temporary pattern of sedimentation and redistribution in the weeks to months following streambed drawdown operation was systematically measured in WY 2016–17 for this study and with anecdotal observation of previous drawdowns (WY 2012–15) and qualitative observations from more recent drawdowns (WY 2018–19). On the Middle Fork Willamette River downstream from the Fall Creek confluence, transport capacity was also sufficiently high to transport streambed drawdown sediment out of the main channel following initial periods of relatively high turbidity and suspended-sediment conditions. However, sand, silt, and organic matter did repeatedly accumulate in low-velocity areas like side channels, alcoves, and along channel margins in the months and years following implementation of streambed drawdown operations.

Larger and thicker deposits of sand and finer-grained sediments were likely established during the earliest drawdowns (WY 2012–13) in off-channel areas nearer to the dam on Fall Creek than in subsequent years (figs. D10, D15, D16; fig. E3). During WY 2014–18, with reduction of sediment released from the reservoir (fig. C3) and as some off-channel areas have lost accommodation space (figs. D10, D15), relatively less sediment was deposited in reaches of Fall Creek immediately downstream of the dam compared with WY 2011–12; although, the accumulation of new sand and finer-grained sediment deposition on bars and low floodplain surfaces in any particular year is dependent on streamflows and local inundation patterns along with upstream supply (fig. D9). Along the Middle Fork Willamette River, deposition of sediment likely sourced from streambed drawdowns, in addition to sediment potentially derived from other sources, appeared to be more variable longitudinally than was observed along Fall Creek (figs. D9, D19, D20); deposits were still mostly limited to low velocity areas along high bars and low floodplains.

In the spring and summer, newly deposited patches of sand and finer-grained sediment became a focal area for the establishment of riparian vegetation. Canary grass and willows (fig. D6; fig. E3) have been observed to rapidly colonize some monitoring sites during this study; however, the degree to which vegetation preexisted prior to this study in areas of deposition is unknown. Existing grasses and young willow may have been buried by reservoir sediments and grown vertically through deposits or colonized newly created sand bars created from reservoir sediments. Many types of riparian vegetation are adapted to disturbance and can rapidly adjust to burial by or colonization of fluvial sediments (for example, Shafroth and others, 2002; Politti and others, 2017), so the vegetative responses to the Fall Creek deposition are not unexpected, especially considering the temperate Mediterranean climate (Köppen-Geiger climate classification) of the study area and long growing season conducive to vegetation establishment. Once established on unconsolidated, sandy deposits from the Fall Creek streambed drawdowns, grasses and woody shrubs reinforce the stability of the deposit and increase local roughness, and absent large-scale flooding or other disturbance, reduces the likelihood of subsequent scour of those deposits by typical regulated streamflows.

Low-velocity, off-channel areas are prime locations for sand and finer-grained reservoir sediment to accumulate. These features are relatively small and intermittently distributed along Fall Creek both historically (prior to dam construction) and presently (during the period of this study), owing largely to geologic and physiographic controls on channel morphology (Chapter D. Geomorphic Responses to Fall Creek Lake Streambed Drawdowns Downstream from Fall Creek Dam; Keith and others, 2023a; figs. D14–D16; app. 3). Partly due to their relatively small size and shallow depths, off-channel areas on Fall Creek that underwent sand and finer-grained sediment deposition in WY 2012–18 likely only have a limited amount of space to store new sediment from future drawdowns. Because the total area of off-channel features along the Fall Creek reaches is small due to lateral channel stability and lack of coarse sediment supply from upstream, further losses in secondary water features would account for large portions of the total reach area of off-channel features in these reaches. Ultimately, a reduction in off-channel features could signify a loss of low-velocity habitats used by native fish (such as Oregon chub [Oregonichthys crameri]; for example, Bangs and Meeuwig, 2017) and other aquatic life year-round or as slow-water refuges during high streamflows. The relative size of secondary water features and unvegetated bars on the Middle Fork Willamette River is much greater than those along Fall Creek (fig. D14; app. 3), and changes in area of these habitat-related features might comprise a smaller portion of channel than do changes to these features on Fall Creek. However, within all study area reaches, the reach-scale responses of bars, secondary water features, and other landforms to streambed drawdowns was challenging to detect given the mapping approaches, variability in discharge, and relatively small number (five) of streambed drawdowns assessed within the geomorphic mapping periods (figs. D13, D14).

Reservoir Evolution

Similar to Fall Creek, a study of drawdown operations in France found that reservoir evolution is a function of management and sediment supply. Guertault and others (2014) evaluated changes in reservoir volume sediment budgets over nearly 3 decades during flush and inter-flush periods (1984–2012) at the Génissat Dam on the Rhone River in France (though other dams on the Swiss–French Rhone and tributaries contributed to sediment and hydrologic conditions at the site). They identified different areas of erosional and depositional processes within the reservoir related to the flush (drawdown) and inter-flush (not drawdown) periods and found that sediment transport within the reservoir is highly dependent on pool level, so that overall reservoir evolution was mainly a function of reservoir management and sediment supply. Over about a 15-year period of annual sediment flushing events, they also found no significant change in the longitudinal distribution of grain sizes along the reservoir channel bed though coarser sediments could have been buried underneath finer sediments.

Geomorphic assessment of reservoir processes related to moderate fluctuations in lake levels (1.1–6.6 m annually)—though not specifically drawdown operations—of the Bratsk Reservoir in Eastern Siberia by Kaczmarek and others (2016) revealed that existing geomorphic processes were intensified by the construction and operation of the reservoir, that those processes interact with one another, and that geomorphic evolution of the shore area is closely linked to water-level changes. Though the Bratsk Reservoir is larger and local geology and site-specific processes are much different than those at Fall Creek Lake, the reservoir hillslopes and shorelines of Fall Creek Lake display erosional and depositional processes within overprinting of landforms and those processes arealso linked to fluctuation in pool levels (fig. B11).

Downstream Sediment Transport

Suspended-sediment transport resulting from reservoir sediment releases including drawdowns, reservoir flushing, and dam removals are typically characterized in terms of suspended-sediment concentrations (SSC) and suspended-sediment loads (SSL) as measured at sites downstream from the dams because fine-grained sediment dominated by silts and clays though sometimes also sands evacuated during these operations are primarily carried as suspended sediment rather than bedload sediment. Though becoming more common, as of 2021, there are few studies documenting reservoir sediment flushing or drawdown operations that were implemented in conjunction with sediment transport gaging programs. Available studies indicate substantially lower peak values of SSC at Fall Creek compared with other reservoirs during controlled sediment flushing events, but comparable values of SSC and SSL at other reservoirs with partial drawdowns where sediment flushing is not the primary goal. In their synthesis of reservoir sedimentation, Morris and Fan (1998) stated, “The initial period of flushing, with a duration ranging from a few hours to several days, depending on the size of the reservoir, is characterized by extremely high sediment concentrations, typically exceeding 100 grams per liter (g/L) at the dam site and in some cases even exceeding 1,000 g/L.” Likewise, SSC measured 6.7 km downstream from the Cancano Reservoir (Adda River, Italy) from 2010 to 2012, where controlled sediment flushing operations spanned 40–50 days per year had maximum values of SSC exceeding 99 g/L, although average concentrations were much lower (less than 8 g/L) and the implementation of multiple measures to minimize sediment concentration, such as increasing regulated tributary streamflows to reaches downstream from the dam and utilizing an instream settling basin, appeared to reduce SSC at a site 22.9 km downstream from the dam (Espa and others, 2016). At the Valgrosina Reservoir, Valtellina, northern Italy, Crosa and others (2009) compared SSC levels to the impacts to fish and macroinvertebrates during two streambed drawdowns in 2006–07 to inform a regulatory framework for sediment releases resulting from controlled reservoir flushing operations. Maximum SSC from these events ranged up to 70–80 g/L during the initial drawdown phases but average concentrations of 4–5 g/L were more typical during the operations. After flushing operations ceased, geomorphic observations of reaches downstream from the dam revealed few areas with substantial sediment deposits, suggesting much of the reservoir sediments had been transported to downstream reaches. In contrast, at Fall Creek Lake, the peak sampled SSC 1.4 km downstream from Fall Creek Dam during the USGS sampling program from WY 2012–18 was 15.3 g/L, during WY 2015 (November 21, 2014). Although measured SSC from the Fall Creek drawdowns were nearly an order of magnitude lower than the “typical” conditions of 100 g/L (Morris and Fan, 1998), suspended-sediment loads from Fall Creek are comparable with other sites. For example, at the Cancano Reservoir, total annual sediment loads ranged from 14,620 to 70,920, t over the 3 years of controlled flushes (2010–12; Espa and others, 2016), which are within an order of magnitude of SSL computed for the full monitoring periods at the Fall Creek streamgage (ranging from 13,900 to 54,700 t; table C1). At the USACE’s Cougar Dam on the South Fork McKenzie River (also located in the Willamette River Basin in Oregon; fig. A2), Anderson (2007) measured sediment transport during a partial drawdown spanning WY 2002–04 during which lake levels were maintained at 60–90 m below normal low pool to facilitate construction of a temperature control tower. Downstream from Cougar Dam, the peak SSC of the 3-year (WY 2002–04) monitoring period was 0.266 g/L (1/30/2003, 14159500 streamgage) and partial-year SSL ranged from 4,100 to 17,000 t, which is substantially less than the partial-year SSL calculated at Fall Creek WY 2013–18 (fig. C3; table C1). Despite impounding a substantially larger basin, the much smaller SSC and SSL measured in the South Fork McKenzie River during the Cougar Dam partial drawdown compared with the Fall Creek annual streambed drawdowns likely reflect the substantial residual pool at Cougar Dam (typically more than 40 m deep at the dam) that diminished transport capacity of sediment passing through the reservoir and trapped sediments eroded from upper areas of the reservoir. Likewise, a partial drawdown of Lake Mills on the Elwha River in 1994 resulted in temporarily lowering the lake by about 5.5 m (leaving a residual pool behind the 64-m-high dam) to evaluate reservoir sediment deposits and inform the 2014 dam removal (Childers and others, 2000). Although the Lake Mills drawdown resulted in substantial erosion and reconfiguration of the reservoir delta, much of the mobilized suspended sediment was trapped in the residual pool, and maximum SSC measured downstream from the dam was 20 mg/L (Childers and others, 2000).

Previous studies of sediment transport resulting from reservoir drawdowns indicate that coarse sediment (greater than 2 mm) transport in downstream reaches can increase as reservoirs rapidly drain during dewatering periods and streamflow increases as water is evacuated from the reservoir. Reservoir dewatering and increased streamflow can mobilize bed-material sediment in reaches downstream from the dam prior to the period of reservoir sediment erosion and downstream export that occurs during full streambed drawdown conditions. This concept of a two-part sediment pulse (mobilization of downstream sediment during reservoir draining, followed by export of reservoir sediment) stems from studies of Cachí Reservoir on the Reventazón River, Costa Rica, together with an extensive review of downstream sediment transport conditions produced by reservoir flushing (Brandt, 1999, 2005). Bed-material transport during the draining of Fall Creek Lake prior to the full streambed drawdown was not measured at Fall Creek between WY 2013 and 2018. However, substantial increases in coarse bed-material mobilization were probably not likely during those periods because extensive in-channel bedrock and a limited amount of exposed, readily erodible sands and finer gravels mantle the channel bed downstream from Fall Creek Dam and create a vertically stable channel with coarse sediment unlikely to mobilize during streamflows that are typically observed in reservoir dewatering period (less than 2,000 ft³/s). Additionally, any finer sediments from the prior year drawdown that could be mobilized at low streamflows have likely already been transported farther downstream during the prior winter (Chapter C. Sediment Delivery from Fall Creek Lake and Transport through Downstream Reaches). During free-flowing streambed conditions, streamflows are typically much lower (39–777 ft³/s in WY 2013–18; Schenk and Bragg, 2021) than during the reservoir draining period, further limiting coarser gravel transport in downstream reaches. If the annual streambed drawdowns at Fall Creek Lake continue to be operationally similar to those that occurred during WY 2013–18, a substantial increase in sediment transport prior to full streambed drawdown (such as that described by Brandt [2005]) is not likely to occur. However, if the operations were to change so that reservoir dewatering produced much greater streamflows capable of mobilizing bed-material sediment in the reaches downstream from the dam, either by inducing a rapid drawdown rate or with prolonged streambed drawdown that would be more susceptible to stormflows passing through the reservoir, increased mobilization and transport of bed material could result, with coarser sediment than has been seen in recent drawdowns.

Downstream Changes in Channel Morphology and Habitats

Deposition of reservoir sediment in the reaches downstream from dams on Fall Creek and Middle Fork Willamette River in WY 2013–18 primarily occurred in low-velocity areas along channel margins, which is similar to sediment releases from drawdowns elsewhere. At Halligan Reservoir on the North Fork Poudre River, Colorado, most deposition occurred in low velocity pools and those deposits were reworked during subsequent streamflows (Wohl and Cenderelli, 2000). Accordingly, bedload depletion from the pools was strongly related to transport rates downstream (Wohl and Cenderelli, 2000). Similarly, Espa and others (2016) found that the downstream deposition of reservoir sediments from the Cancano Reservoir (Italy) drawdowns were largely confined to low velocity areas of the Adda River with minimal deposition in the main channel except in some very low gradient reaches. In another study, monitoring hydrogeomorphic responses to controlled drawdown on the Inn River at the Gepatsch Reservoir (Austria) showed no significant changes in particle-size distributions along gravel bars downstream from the reservoir (Hauer and others, 2020). These findings are comparable with other examples of sediment releases into high-energy streams (from dam removals and landslides) where transport capacity of finer sediments is sufficient to transport those finer sediments to downstream reaches as suspended-sediment load with net deposition primarily occurring locally along pools and channel margin areas (for example, Major and others, 2012; Wilcox and others, 2014; Anderson and others, 2017).

Infiltration of fine-grained sediment into coarser substrate can negatively affect certain aquatic habitats such as spawning redds for native resident and anadromous fish or benthic invertebrates like mussels. Although the impacts of fine sediment to salmon spawning habitats are well-documented elsewhere (for example, Tappel and Bjornn, 1983; Everest and others, 1987; Roy and others, 2003), impacts to spawning habitat for spring Chinook salmon on Fall Creek were not evaluated in this study, but could be a focus of future monitoring. Although fine-sediment burial and intrusion into spawning beds was not evaluated, the bed-material grain-size measurements at gravel bars downstream from Fall Creek Lake for this study, together with anecdotal observations do not indicate wide-spread deposition and persistence of fine-grained reservoir sediments on coarse bed substrates for the main channel reaches downstream from the Fall Creek streamgage. Although the majority of historical and present-day spawning of spring Chinook salmon occurs upstream from Fall Creek Dam (McIntosh and others, 2009), redd burial could be more problematic in the upstream segment between the Fall Creek streamgage and Fall Creek Dam, where sediment delivery during the streambed drawdown can result in short-term deposition until subsequent streamflows in following months transports this sediment downstream (Chapter C. Sediment Delivery from Fall Creek Lake and Transport through Downstream Reaches), the timing of which overlaps with incubation period for spring Chinook salmon (NMFS, 2008). Elsewhere in the Willamette River Basin, fine-grained sediment intrusion into gravel-bed substrate was monitored downstream from Cougar Dam on the South Fork McKenzie River, where spring Chinook salmon spawning is concentrated and where concerns about impacts to spawning habitats were prompted by drawdowns for construction of a reservoir temperature control tower. Monitoring of the WY 2002–04 Cougar Dam drawdowns showed an abundance of fine-grained sediment in transport downstream from the dam (Anderson, 2007) and infiltration to gravels (Anderson, 2007; Grant and others, 2015) that almost certainly resulted at least in part to the drawdown operations.

Implications for Reservoir Morphology and Future Erosion

Reservoir evolution observed over the streambed drawdowns of WY 2012–18 informed delineation of Fall Creek Lake whereby reservoir zones and landforms are classified according to geomorphic processes including the varying depositional and erosional processes active at different dam management and streamflow scenarios as well as proximity to the main reservoir channel (which partly implies susceptibility to erosion; see “Fall Creek Dam Terminology Used in this Study,” “Hydrology, and Streamflow Regulation by Dams,” and “Dam Infrastructure and Streambed Drawdown Operations at Fall Creek Dam” sections of Chapter A. Introduction). For example, in areas immediately upstream from the dam where fine-grained sediment accumulation is thick and readily accessible to fluvial erosion and transport, the direct controls on reservoir sediment erosion are dam operations (predominantly lake level) and streamflow (a function of unregulated streamflow entering the reservoir). In comparison, sediment stored along reservoir margins at higher elevations of the reservoir floor (including the pre-dam floodplain and terraces) is less susceptible to erosion and downstream export because it is generally inaccessible to erosive streamflows during the streambed drawdowns, and any eroded sediment must be routed through the reservoir prior to export.

Multiple aspects of Fall Creek Dam infrastructure and operations exert first-order controls on the magnitudes of reservoir erosion that occur during the streambed drawdowns and ultimately dictate the sediment delivery to downstream reaches. Key aspects of the dam and its operations that are most relevant to assessing geomorphic responses to streambed drawdowns include the (1) dam infrastructure, including configuration and size of regulating outlets and their proximity to the streambed, which dictates the amount and type of sediment that can be transported to downstream reaches and mode of sediment transport; (2) frequency of historical streambed drawdowns and the overall pattern of long-term, year-round dam operations and lake level management, which partly dictate reservoir morphology and locations and magnitudes of readily erodible materials; (3) dam operations during the streambed drawdown, which directly control the timing and duration of reservoir erosion and sediment evacuation; and (4) dam operations following the streambed drawdown operation that regulate streamflows (and thereby sediment transport conditions) downstream of Fall Creek Dam which primarily reflect interactions between hydroclimatic conditions and flood control.

Variability in sediment yield to lower the reservoir area, streamflows entering the reservoir during streambed drawdowns, and other hydroclimatic conditions likely also exert controls on the magnitude and character of sediment mobilized to reaches downstream from Fall Creek Dam. The historical and annual volumes and types of sediment delivered to the lower parts of Fall Creek Lake directly determines the amount of readily erodible sediment that can be eroded and delivered to downstream reaches during streambed drawdowns. Varying sediment yield due to changes in land use, climate, or occurrences like forest fire and landslides could change the magnitude of sediment or suite of grain sizes that are delivered to the lower reservoir and later eroded and passed downstream. For example, the recent Jones Fire (summer 2017) burned about 42 km² (9 percent) of the watershed upstream from Fall Creek Dam, and the Bedrock Fire (summer 2023) burned about 120 km² (24 percent) of the watershed upstream from Fall Creek Dam, including areas adjacent to the Fall Creek channel. Similarly, streamflows entering Fall Creek Lake during the streambed drawdown operation directly control magnitudes and styles of reservoir erosion, sediment entrainment, and export from the reservoir. The role of streamflow as a first-order control on reservoir erosion and sediment mobilization has been previously documented in dam removal and reservoir flushing studies (for example, Major and others, 2012) and corroborate the WY 2012–18 findings from Fall Creek. Generally, greater discharge magnitudes through the reservoir during free-flowing conditions at streambed drawdown will increase transport capacity of sediment through the reservoir, hence increasing erosivity along reservoir channel banks and bed and leading to greater amounts of erosion and downstream transport. Storm events that increased streamflow during low reservoir pool elevations at Cougar Dam in WY 2004 were responsible for lateral erosion and channel migration of the channel through reservoir deposits (Anderson, 2007). In addition to reservoir discharge, the findings from Fall Creek drawdowns WY 2012–18 indicate that overall hydroclimatic conditions during the streambed drawdown, particularly precipitation and freezing air temperatures, can also influence the magnitudes of reservoir erosion and sediment export.

Patterns of sediment erosion and evacuation observed in this study at Fall Creek Lake from WY 2012–18 suggest that reservoir erosion during annual streambed drawdowns may remain similar or decrease in future years assuming (1) annual streambed drawdown operations are implemented in similar manner as the WY 2012–18 drawdowns (in terms of duration, late autumn or early winter implementation, rate of pool-level lowering to reach streambed, and other factors), (2) streambed drawdowns coincide with similar conditions as were observed WY 2012–18 (similar sediment yield into reservoir, low reservoir inflows, limited precipitation, moderate air temperature), and (3) no major geomorphic changes in the main reservoir channels of Fall and Winberry Creeks occur (for example, channel avulsion). Under such conditions, it is hypothesized that the stream channel within the reservoir would achieve a quasi-equilibrium state with respect to annual influx and export of sediment and aided by the substantial amount of in-channel bedrock, will remain laterally stable without erosion across reservoir deposits. This scenario is supported by the conceptual model proposed by Morris and Fan (1998, p. 15.6), which described the evolution of repeat reservoir flushing.

If the annual streambed drawdowns were to continue for many years or decades into the future, some of the sediment stored in the upstream portion of the reservoir could shift downstream, triggering morphological adjustments in the Fall Creek stream channel within the reservoir and possibly altering the trajectory of annual reservoir erosion and sediment export. Though Guertault and others (2014) found no significant changes in the reservoir channel of the Verbois Reservoir over multiple years of flushing, observations at other sites (Morris and Fan, 1998; Kondolf and others, 2014) suggested persisting progradation of coarse sediment through the reservoir. Morris and Fan (1998, p. 15.6) also suggested the possible occurrence of later or additional channel migration through floodplain deposits related to bed aggradation from coarse delta deposits.

Although Fall Creek Lake lacks a well-defined coarse sediment delta at the head of reservoir, the main reservoir channel above minimum conservation pool (728 ft) includes exposed gravel bars (fig. B9; table 3.1). It is possible that coarse sediments could be mobilized downstream, resulting in aggradation of the main reservoir channel. Aggradation of the main reservoir channel could, in turn, increase channel and streamflow connectivity to adjacent to reservoir floor deposits that are currently inaccessible to fluvial erosion during typical low streamflows that occur during streambed drawdowns. However, there is considerable uncertainty regarding the amount and character of coarse sediment stored in the upper reservoir, the mobility of this sediment under different streamflow scenarios, and potential implications for reservoir erosion.

Implications for Future Sediment Transport Downstream from Fall Creek Dam

Analyses of suspended sediment measured in WY 2013–18 shows a stark reduction in suspended-sediment loads between WY 2013 and later years and suggests future suspended-sediment loads may remain variable but similar to those observed WY 2015–18 if factors such as operations, sediment yield, and reservoir inflows remain similar. Suspended-sediment loads downstream from Fall Creek Dam will likely remain highest during the official drawdown period when regulating outlets are fully open and Fall Creek is free flowing or near streambed with little to no residual pool. Over time, the suspended-sediment load sourced from the reservoir would increasingly comprise a larger portion of sediment deposited the previous year rather than sediments that have been deposited historically and remained for several years on the reservoir floor near the dam. Bedload is likely to remain a small fraction of the total sediment load evacuated from the reservoir and is also likely relatively modest compared with historical (pre-dam) bedload transport rates because nearly all coarse sediment is trapped by the dam (Schenk and Bragg, 2014; O’Connor and others, 2014). An alternative, though highly uncertain longer-term response, is that the coarse gravel deposited within the reservoir channel may migrate farther through the reservoir (see the “Implications for Reservoir Morphology and Future Erosion” subsection of this chapter), and downstream from Fall Creek Dam resulting in greater bedload transport than has been observed for this study.

Implications for Future Geomorphic Changes Downstream from Fall Creek Dam

Despite substantial reductions in peak streamflow magnitude due to regulation from Fall Creek Dam, streamflows released from the dam in the months following each drawdown of WY 2012–18 had ample transport capacity to entrain and transport reservoir sediment delivered to downstream reaches during the streambed drawdowns, resulting in modest net geomorphic changes to the downstream reaches. Multiple controlling factors influence processes, patterns, and magnitudes of geomorphic responses to streambed drawdowns and fate of sediment in the reaches downstream from Fall Creek Dam. Interactions among channel morphology, streamflows, hydraulics, and vegetation were observed to partially control patterns of sediment deposition and overall changes in channel morphology that resulted from streambed drawdowns of WY 2012–18. If sediment releases and ensuing streamflow conditions follow a similar pattern in the future (after WY 2018), then the near-term reach-scale geomorphic adjustments from future drawdowns (perhaps next 2–10 years) are expected to be modest, based on the lack of substantial, reach-scale changes in channel planform or landforms detected from aerial photographs (figs. D13, D14; app. 3). However, reach-scale geomorphic changes may become more apparent if (1) streambed drawdowns continued for several decades and geomorphic changes were evaluated at sub-decadal timesteps to provide a longer record from which to assess net change or (2) the magnitude of sediment supply introduced to downstream reaches substantially increases and (or) sediment transport capacity decreases in the months immediately following streambed drawdown operations. For example, a streambed drawdown and major sediment release in late spring, followed by several months of low streamflows would likely result in different sediment transport and deposition patterns compared with the late autumn or early winter drawdowns of WY 2012–18 evaluated in this study, which were each followed by several months of relatively high streamflows. At the landform-scale (for example, the scale of an individual alcove), low velocity areas nearest to Fall Creek Dam will likely continue to undergo rapid deposition immediately after a streambed drawdown event, similar to patterns observed for WY 2012–18 (figs. D6, D10). Similar to observations from WY 2012–18, some of the sediment entering these off-channel features and margin areas may be temporarily stored, then later remobilized and dispersed farther downstream from post-depositional streamflows are large enough, but if deposits persist through the following spring, there is a greater likelihood that local vegetation establishment (as observed, primarily reed canary grass and willows) will reinforce deposited material and serve as a sediment trap following later drawdowns. Additionally, as the limited area of alcoves and side channels along Fall Creek continue to accumulate sediment, increased amounts of released reservoir sediment may be bypassed directly to the Middle Fork Willamette River. Regardless of how much or what types of sediment enter reaches downstream from Fall Creek Dam, the presence of the dam provides an opportunity to strategically manage streamflows during and after the streambed drawdowns to minimize downstream sediment impacts while also ensuring other operational thresholds are satisfied (Loire and others, 2021; fig. A11).

Implications for Monitoring Geomorphic and Sediment Transport Responses to Future Drawdowns

Although annual streambed drawdowns such as those implemented at Fall Creek Lake are uncommon, even more infrequent is the comprehensive and coupled investigation of reservoir evolution, downstream sediment transport, and geomorphic responses to channel and overbank areas documented in this study. Nonetheless, the datasets from this study provide a baseline for future monitoring, and lessons learned from the multi-faceted monitoring approach employed at Fall Creek provides a foundation for identifying useful monitoring activities at various temporal and spatial scales at Fall Creek or other reservoirs where major lake level fluctuations or deep drawdowns occur. Future monitoring could be tailored to address specific management questions or concerns (such as impacts to habitats, flood hazards, or cultural resources) and could include status and trends monitoring to track changes in reservoir conditions, downstream sediment transport, or downstream channel morphology (app. 4). Drawing upon the methods used in this study, as well as those from dam removal studies, future monitoring could focus on the three geomorphic process categories affected by the streambed drawdown operation: (1) reservoir erosion and net sediment evacuation; (2) sediment delivery to downstream reaches, including magnitude and temporal pattern of sediment transport; and (3) geomorphic responses of downstream reaches to sediment delivery. Outstanding questions and priority monitoring to evaluate streambed drawdown responses at Fall Creek Lake, Fall Creek, and the Middle Fork Willamette River are described in appendix 4, and priority monitoring activities are summarized here: