Numerical Modeling of Sediment Transport in the Sandy River, OR Following Removal of Marmot Dam

Supplement to Environmental Assessment for the Bull Run Hydroelectric Project FERC Project No. 477 Numerical Modeling of Sediment Transport in the Sandy River, OR Following Removal of Marmot Dam Technical Report Prepared by Berkeley, California Prepared for Portland General Electric Company Portland, Oregon

Table of Contents 1. INTRODUCTION...1 2. BACKGROUND INFORMATION ON EXISTING CONDITIONS IN THE SANDY RIVER BASIN...2 2.1 Watershed Description...2 2.2 Geology of the Sandy River Basin...3 2.3 Hydrology...3 2.4 Geomorphic Characteristics of the Sandy River...4 3. NUMERICAL MODELING OF SEDIMENT TRANSPORT FOLLOWING DAM REMOVAL...6 3.1 Gravel Model...7 3.2 Sand Model...8 3.3 Discussion of Modeling of Reservoir Erosion...9 3.4 Uncertainties in the numerical modeling...10 4. GOVERNING EQUATIONS...12 4.1 Governing Equations for the Gravel Model...12 4.2 Governing Equations for the Sand Model...15 5. MODEL INPUT DATA...20 5.1 Gravel Model Input Data...20 5.1.1 Channel gradient and width...20 5.1.2 Discharge data and hydrologic scenarios used in numerical modeling...20 5.1.3 Grain size distribution of the reservoir sediment...23 5.1.4 Surface grain size distribution and abrasion...24 5.1.5 Background gravel transport rate...25 5.2 Sand Model Input Data...26 5.3 Zero Process...28 6. RESULTS...29 6.1 Reference runs of numerical models...30 6.2 Alternative B: Single-season dam removal, minimal sediment removal...31 6.2.1 Gravel Model Results...31 6.2.2 Sand Model Results...34 6.3 Alternative C: Removal of top of dam in Year 1, complete dam removal in Year 2 with sand layer excavation up to a point 830 m upstream of Marmot Dam...36 6.3.1 Gravel model results...37 6.3.2 Sand model results...38 6.4 Alternative D: Removal of sediments to the bottom of the sand layer and to a point 830 m upstream of Marmot Dam...38 6.4.1 Gravel model results...38 6.4.2 Sand model results...39

7. DISCUSSION...39 Acknowledgments...41 8. REFERENCES...42 List of Symbols...46 ATTACHMENT A...48

List of Figures: Figure 1Map of Sandy River basin Figure 2Sandy River Longitudinal Profile (Source: PGE photogrammetry, 1999) Figure 3Grain size distributions of surface gravel from pebble counts by Figure 4Sandy River geomorphic reaches delineated by Figure 5Channel bed slopes in the Sandy River Figure 6Flood frequency, Sandy River near Marmot, OR gaging station Figure 7Predicted and observed annual daily average discharges, Sandy River near Marmot, OR gaging station Figure 8Annual hydrographs used in numerical modeling Figure 9Grain size distributions (average and upper and lower Bounds) of Units 1 and 2 in Marmot Reservoir sediment deposit Figure 10Simplified representation of stratigraphy of Marmot Reservoir sediment deposit, based on Squier Associates coring study Figure 11Thickness of gravel deposition in Sandy River, reference run Figure 12Predicted Total suspended sediment (TSS) at selected locations in the Sandy River, reference run Figure 13a:Thickness of gravel deposition following removal of Marmot Dam (Alternative B, Run 1) (3- D version) Figure 13b:Thickness of gravel deposition following removal of Marmot Dam (Alternative B, Run 1) (Non-3D version) Figure 14:Evolution of long profile in vicinity of reservoir following removal of Marmot Dam (Alternative B, Run 1) Figure 15 Annual change in bed elevation following removal of Marmot Dam (Alternative B, Run 1) Figure 16 Thickness of gravel deposition following removal of Marmot Dam (Alternative B, Run 2) Figure 17 Thickness of gravel deposition following removal of Marmot Dam (Alternative B, Run 3) Figure 18 Thickness of gravel deposition following removal of Marmot Dam (Alternative B, Run 4) Figure 19 Thickness of gravel deposition following removal of Marmot Dam (Alternative B, Run 5) Figure 20 Relationship between slope and multiplier applied to gravel transport rate in order to test sensitivity of gravel model to potential incision of gravel deposition Figure 21 Thickness of gravel deposition following removal of Marmot Dam (Alternative B, Run 1, Sensitivity Test) Figure 22 Thickness of gravel deposition (Alternative B, Sensitivity) Figure 23 Thickness of sand deposition in lower 14 km of Sandy River following removal of Marmot Dam (Alternative B, Run 1) Figure 24 Thickness of sand deposition at selected locations in the first two years following removal of Marmot Dam (Alternative B, Run 1) Figure 25 Daily sand release at Marmot Dam for 10 years following dam removal (Alternative B; Runs 1, 2, and 3) Figure 26 Daily sand release at Marmot Dam for 10 years following dam removal (Alternative B; Runs 1, 4, and 5) Figure 27 Daily sand release at Marmot Dam for 10 years following dam removal (Alternative B and D) Figure 28 Total suspended sediment (TSS) at selected locations in first two years following removal of Marmot Dam (Alternative B, Run 1)

Figure 29a Sensitivity test of sand release from Marmot reservoir, with 10-fold increase over basic model predictions: Thickness of sand deposition Figure 29b Sensitivity test of sand release from Marmot reservoir, with 10-fold increase over basic model predictions: TSS Figure 30a Sensitivity test of sand release from Marmot reservoir, with 5-fold increase over basic model predictions: Thickness of sand deposition Figure 30b Sensitivity test of sand release from Marmot reservoir, with 5-fold increase over basic model predictions: TSS Figure 31 Thickness of gravel deposition following lowering of Marmot Dam by 9 m (Alternative C) Figure 32Evolution of long profile in vicinity of reservoir following lowering of Marmot Dam by 9 m (Alternative C) Figure 33Evolution of long profile in vicinity of reservoir following lowering of Marmot Dam by 8 m (Alternative C) Figure 34Evolution of long profile in vicinity of reservoir following lowering of Marmot Dam by 11 m (Alternative C) Figure 35Evolution of long profile in vicinity of reservoir following lowering of Marmot Dam by 9 m, wet hydrologic conditions (Alternative C) Figure 36Evolution of long profile in vicinity of reservoir following lowering of Marmot Dam by 9 m, dry hydrologic conditions (Alternative C) Figure 37 Thickness of gravel deposition following removal of Marmot Dam and excavation of sediment to 830 m upstream of the dam (Alternative D) Figure 38Evolution of long profile in vicinity of reservoir following removal of Marmot Dam and excavation of sediment to 830 m upstream of the dam (Alternative D) Figure 39 Total suspended sediment (TSS) at selected locations in first two years following removal of Marmot Dam (Alternative D) List of Tables: Table 1 Table 2 Table 3 Table 4 Table 5 Table 6 Table 7 Summary of geomorphic reaches of Sandy River reaches Water year series selected for use in simulation Drainage area and discharge increases at tributary junctions along the Sandy River Summary of Squier Associates, Inc. interpretation of reservoir deposit Summary of input parameters and sources used in gravel model Summary of model runs performed to evaluate Alternatives B, C, and D for removal of Marmot Dam Change in channel bed slope through time near Marmot Dam for Alternative B

Dam EXECUTIVE SUMMARY Marmot Dam, a hydroelectric facility on the Sandy River, OR, is scheduled to be removed by its owner, Portland General Electric. The dam is located on the Sandy River at River Mile (RM) 30 (Rkm 48), has an upstream drainage area of about 680 km 2, and is 14-m high. The dam s impoundment is currently filled to near the dam s crest with approximately 750,000 m 3 of fine and coarse sediment, of which 490,000 m 3 is primarily gravel and 260,000 m 3 is primarily sand (Squier Associates 2000). The Sandy River originates from Mt. Hood in the Cascade Range of northwestern Oregon and drains into the lower Columbia River. Mt. Hood glaciers, the ongoing influence of past laharic events, and the basin s underlying lithology result in naturally high sediment loading in the Sandy River. developed one-dimensional numerical models of coarse and fine sediment transport to quantify the routing of sediment from behind Marmot Dam downstream through the Sandy River. The model results presented here examine three dam removal alternatives, in which varying amounts of the sediment stored behind Marmot Dam would be released downstream: (1) Alternative B, which entails removal of Marmot Dam in one season, with removal of only a minimal amount of sediment as required for construction activities, (2) Alternative C, in which the dam would be lowered in Year 1 and completely removed in Year 2, with sediment excavation up to a point 830 m upstream of Marmot Dam during Year 2, and (3) Alternative D, which entails removal of the dam in one season and concurrent removal of accumulated sediments from the dam to a point about 830 m upstream (this would allow excavation of most of the fine sediment in the reservoir and of the majority of the overall amount of sediment in the reservoir). A fourth alternative, Alternative A, would entail excavation of nearly all sediment in the reservoir before dam removal and was not modeled. We also conducted model runs in which hydrologic conditions, assumed grain size distributions of reservoir sediment, and rates of sediment release from the reservoir were varied to test the sensitivity of model results to these factors. The gravel (coarse sediment) transport model is based on Parker s surface-based bedload equation (Parker et al. 1982, Parker 1990); the sand transport model is based on Brownlie s bed material equation (Brownlie 1981). Key input data for the models include channel gradient, channel width, grain size distribution of the sediment stored behind Marmot Dam, and water discharge. Results of modeling of Alternative B, in which nearly all sediment stored behind Marmot Dam would be released downstream, indicate that in the first year following removal, coarse sediment would quickly form a debris fan immediately downstream of the dam, with small amounts of deposition predicted further downstream. In subsequent years, additional sediment would move out of the reservoir and downstream from the debris fan, resulting in a gradual increase in deposition thickness of coarse sediment about 2 4 km downstream of the dam, reaching a maximum of about 1 m on a reach-averaged basis. Coarse sediment is predicted to travel quickly through most of the Sandy River gorge, which has a steep gradient and high confinement, with aggradation increasing downstream of the gorge (9 13 km downstream of the dam, where the channel widens and decreases in gradient) from Years 1 through 10 following dam removal, where aggradation is predicted to gradually build to a maximum predicted thickness of about 1.5 2 m. Further downstream, the model predicts small amounts of aggradation (<0.5 m), although this aggradation is similar in magnitude to aggradation predicted in a reference run of the model. In most reaches and in most years that were modeled, particularly with increasing time after dam removal, the F:\PGE\YCMODELING\REPORT\CURRENT\ABSTRCT2.WPD i

Dam predicted rate of change in bed elevation is low (<0.5 m/yr); such changes may be indistinguishable from natural scour and fill processes. Under Alternatives C and D, overall temporal and spatial patterns of downstream sediment deposition would be similar to Alternative B, with smaller predicted deposition thicknesses. On a reach-averaged and time-averaged basis, patterns of downstream thickness of gravel deposition show limited sensitivity to variations in hydrologic conditions and assumed grain size distribution of the reservoir deposit. All model runs indicate that the coarse sediment stored behind Marmot Dam would move downstream by both translation and dispersion. The sand transport model indicates that, under all alternatives, sand aggradation is most likely to occur in the lower 10 km of the Sandy River, which has a very low gradient. No sand aggradation is predicted further upstream. The model predicts deposition thicknesses of up to about 0.4 m, with the greatest aggradation expected to occur in the first year following dam removal. Aggradation is predicted to occur primarily in the lower 3 km of the river, which roughly corresponds to the current location of the gravel/sand transition area in the Sandy River. Deposition thickness could be greater than predicted because of the backwater effect of the Columbia River, which was not modeled. Sand aggradation could be lower than model predictions in the lower Sandy River, however, if deposition occurs in upstream reaches, which is not predicted in basic model runs but which is likely to occur as local deposition in wider reaches, as transient deposition between high-flow periods, and as infiltration of sand into the channel bed interstices. Post-dam removal total suspended sediment (TSS) concentrations are not predicted to increase substantially under any of the removal alternatives. The modeling predicts that, downstream of Marmot Dam, peak TSS of about 500 ppm would occur in the first winter following dam removal under Alternative B, and that sand transport out of the reservoir would likely continue for approximately a decade following dam removal. TSS levels associated with dam removal are predicted to be relatively low because the fine sediment in the reservoir deposit is armored by a coarser surface layer and would therefore be released gradually, rather than as one large pulse. Overall, increases in suspended sediment concentrations following dam removal under Alternatives C and D are expected to be lower in magnitude and shorter in duration than under Alternative B. There is considerable uncertainty in the pattern of sand release from the reservoir, however, and sensitivity tests evaluating accelerated rates of sand release suggest that the magnitude of increases in TSS and sand aggradation could be greater than for basic model runs, with deposition thickness potentially reaching nearly 1 m in the lower Sandy River and TSS concentrations reaching a peak of 2,000 to 4,000 ppm in the winter following dam removal. The model results are most applicable on a reach-scale and time-averaged basis. The models do not account for certain depositional processes, such as infiltration of fine sediment into the interstices of the channel bed, bar formation, channel migration, sand transport out of the reservoir during low flows, and the production of sand and silt from gravel abrasion (i.e., suspended load estimates do not include products of gravel abrasion). The models are one-dimensional, providing predictions of sediment deposition thicknesses that are averaged over the width of the channel cross section and that represent deposition over the existing channel bed. Modeled deposition thicknesses do not account for local variations in shear stress caused by features such as deep pools, bedrock outcrops, or large boulders, and the amount of sediment actually deposited may be substantially higher or lower than predicted by the model in localized areas of the channel. Modeling results are particularly well-suited for comparing F:\PGE\YCMODELING\REPORT\CURRENT\ABSTRCT2.WPD ii

Dam impacts expected under the different alternatives. Key areas of uncertainty in this modeling effort include the following: (1) reservoir erosion processes and sediment transport out of the reservoir area (onedimensional modeling does not capture processes such as channel incision in the reservoir reach and assumes laterally uniform erosion); (2) effects of coarse bed materials on sediment transport (particularly sand transport, as there are no equations specifically designed for sand transport over a coarse bed); (2) the separation of sand and gravel components in the modeling of sediment transport, rather than modeling of a mixed sand/gravel mixture (sand and gravel transport are treated separately, although they do likely affect each other); (3) simplifications in channel geometry used in the model (e.g., a rectangular channel is assumed, and widths do not vary with stage); (4) rough assumptions used for several input parameters (e.g., background sediment transport rates in the Sandy River); and (5) uncertainties surrounding sediment transport mechanics. This is a pioneering modeling study using state-of-the art methods to route sand and gravel through the Sandy River following a simulated dam removal. Despite the uncertainties, the model is unique in its ability to provide predictions of sediment transport and deposition over large temporal and spatial scales, to allow comparison of various dam-removal alternatives, and to provide a basis for evaluating the potential downstream impacts of sediment release under these alternatives. Although the models upon which this effort is based (e.g., Cui et al. 2000, Cui and Parker 2000) have been tested in flumes and in the field, the accuracy of the model developed for dam removal in the Sandy River has not been verified, and model results should be interpreted accordingly. We have used professional judgement and field observations of the Sandy River to interpret model results in terms of expected geomorphic effects in the Sandy River (for example, discussing how predicted reach-averaged deposition patterns may be manifested in the river) and potential impacts on anadromous salmonid habitats, as presented in Stillwater Sciences (2000). Data collected following removal of Marmot Dam will permit testing of the accuracy of this modeling effort. F:\PGE\YCMODELING\REPORT\CURRENT\ABSTRCT2.WPD iii

1. INTRODUCTION Portland General Electric (PGE), the holder of the Federal Energy Regulatory Commission (FERC) license for the Bull Run Hydroelectric Project, is planning to decommission the Bull Run Project. This will entail removal of Marmot Dam, on the Sandy River, Oregon and Little Sandy Dam (on the Little Sandy River), as well as decommissioning of associated flumes and canals, Roslyn Lake, and the Bull Run powerhouse. Marmot Dam is located on the Sandy River at River Mile (RM) 30 (Rkm 48), has an upstream drainage area of about 680 km 2, and diverts water to the Bull Run powerhouse. This dam was originally completed in 1913, and the original wood crib, rock-filled structure was replaced in 1989 by a 14-m high, 104-m wide concrete dam. Marmot Dam is equipped with a fish ladder for upstream passage and a juvenile bypass facility for downstream passage. Approximately 750,000 m 3 of sediment are stored behind Marmot Dam, of which 490,000 m 3 is primarily gravel/pebble and 260,000 m 3 is primarily sand (Squier Associates 2000). The Sandy River originates from Mt. Hood and drains into the lower Columbia River. The Sandy River basin is located about 48 km east of Portland, Oregon on the western slopes of the Cascade Range. has evaluated the geomorphic effects of removing Marmot and Little Sandy Dams and associated impacts on anadromous salmonid habitats and populations. A central component of this assessment was numerical modeling to estimate sediment transport and deposition dynamics in the Sandy River downstream of Marmot Dam for various removal alternatives. This report describes the methods and results of the numerical models of sediment transport in the Sandy River. Brief background information on the lithologic, hydrologic, and geomorphic setting of the Sandy River basin is also provided. (2000) includes interpretation of model results in terms of geomorphic and biological effects on the Sandy River, including discussion of effects of predicted sediment transport patterns on channel morphology, spawning and rearing habitats, and fish migration. Four alternative methods have been developed for removal of Marmot Dam, each of which differs in the amount of sediment accumulated behind the dam that would be released downstream. Stillwater Sciences= analysis included numerical modeling for Alternatives B, C, and D. The alternatives are as follows: Alternative A - Remove all sediments to a point 1,700 m upstream of Marmot Dam. This alternative entails excavation of approximately 690,000 m 3 of sediment and concurrent removal of the dam in one season. No numerical modeling of sediment release under Alternative A was performed because this alternative calls for nearly all sediment to be removed prior to dam removal; therefore, very little reservoir sediment would be released downstream under this alternative. Alternative B - Single season dam removal - minimal sediment removal. This alternative calls for removal of Marmot Dam and associated facilities in one season, with removal of sediments only as required for construction activities. Approximately 750,000 m 3 of reservoir sediments would therefore be released downstream following Page 1

removal of a cofferdam built to facilitate construction work (the actual amount released downstream may be somewhat smaller because some sediment would be excavated to facilitate construction work). Alternative C - Removal of top of dam in Year 1, complete dam removal in Year 2 with sand layer excavation. This alternative entails removal of Marmot Dam over a 2- year period, with excavation of sand from the reservoir deposit during the second year. In the first year, the top part of the dam would be removed. In the second year, sediment would be excavated to the bottom of the sand layer and the rest of the dam will be removed. The amount of sediment that would be transported downstream under this option is unknown, as it would depend in part on the flows occurring in the first winter following removal of the top of the dam and before the bottom was removed. Alternative D - Remove sediments to the bottom of the sand layer and to a point 830 m upstream of Marmot Dam. This alternative entails removal of the dam, canal, and fish ladder in one season and concurrent removal of accumulated sediments from the back of the dam to a point about 830 m upstream. Under Alternative D, approximately 75% of the reservoir sediment would be excavated (about 560,000 m 3 ), including most of the fine sediment. The remaining 25% of reservoir sediment (about 190,000 m 3 ), nearly all of which is believed to consist of coarse material, would be available for downstream fluvial transport. Little Sandy Dam, a 5-m high concrete dam completed in 1912 on the Little Sandy River, will also be removed. Because of the small amount of sediment stored behind this dam, no numerical modeling was performed to assess dam removal effects on the Little Sandy River. 2. BACKGROUND INFORMATION ON EXISTING CONDITIONS IN THE SANDY RIVER BASIN 2.1 Watershed Description The Sandy River and its tributaries drain a 1,316-km 2 basin on the western slope of the Cascade Range of northwestern Oregon (Figure 1). The basin extends approximately 89 km from its headwaters to its confluence with the Columbia River near Troutdale (Columbia River RM 120.5). Principal tributaries to the Sandy River include Zigzag, Salmon, and Bull Run rivers and Still, Cedar, Gordon, and Beaver creeks. The Bull Run River is the largest tributary to the Sandy River and has a drainage area of approximately 265 km 2 (USDA Forest Service 1997). Annual precipitation ranges from 102 cm near the Sandy River=s mouth to 279 cm near its source on Mt. Hood (ODFW 1990). Page 2

2.2 Geology of the Sandy River Basin The geology of the Sandy River basin reflects Tertiary (Miocene and Pliocene) and Quaternary (Pleistocene and Holocene) volcanic events and Pleistocene glaciations. The most extensive geologic units are Pliocene volcanics (basaltic andesite lava flows); the Miocene Rhododendron Formation, which is composed of weak andesitic tuffs and breccias (volcanic rock) and is common in the upper basin; and the Miocene Troutdale Formation, which is a sedimentary formation that is fluvially derived from the erosion of local volcanic rocks and is common in the lower part of the Sandy River basin (USDA Forest Service 1996). Volcanic mudflows (lahars) originating from Mt. Hood have had an important geomorphic effect on the Sandy River and have traveled as far as the confluence with the Columbia River (Allen 1988). The last two episodes of eruptive activity occurred 1,500 and 200 years ago (the Timberline and Old Maid episodes, respectively), when numerous pyroclastic flows and lahars occurred. During the most recent (Old Maid) eruptive period, the Sandy River became choked with sediment over 20-m deep that completely buried the pre-eruption valley floor between Sandy and Troutdale. Since then, channels have incised into these deposits, leaving behind several-meter-high terraces with actively eroding banks. The ongoing influence of past laharic events, Mt. Hood glaciers, and the basin=s underlying lithology result in naturally high sediment loading in the Sandy River. 2.3 Hydrology Hydrologic regimes in the Sandy River basin are characterized by low flows in August and September and high flows generated by rainfall and rain-on-snow events in winter and snowmelt in spring. Because the headwaters of the Sandy and Zigzag rivers originate from glaciers on the slopes of Mt. Hood, at an elevation of about 1,890 m, flow and sediment loading are greatly influenced by glacial processes and steep unstable slopes (ODFW 1997). Glacial sediments frequently cause the mainstem Sandy River to remain turbid throughout the summer snowmelt period. Basin hydrology has been altered by flow regulation and diversion. The City of Portland and PGE own major water development facilities in the basin, both of which are called the Bull Run Project. These facilities include dams on the Sandy, Little Sandy, and Bull Run rivers (Figure 1). The City of Portland=s Bull Run Project, located on the Bull Run River, provides water supply to the City of Portland and sells power to PGE. No minimum flows are currently required at the City's Bull Run Project; consequently, the Bull Run River is dewatered under normal conditions downstream of the City=s facilities. The City=s dams also block delivery of coarse sediment from the Bull Run basin to downstream reaches. The U.S. Geological Survey operates streamflow gauges on the Sandy, Little Sandy, and Bull Run rivers. Discharge data from the Sandy River near Marmot gauge (station number 14137000) and the Sandy River below Bull Run gauge (station number 14142500) were used in sediment transport modeling, as discussed in Section 5.1.2 below. Page 3

2.4 Geomorphic Characteristics of the Sandy River The Sandy River exhibits many characteristics typical of alluvial rivers, including a longitudinal profile that decreases in gradient in a downstream direction (Figure 2), and a decrease in channel bed particle sizes in a downstream direction. Wolman pebble counts conducted by at 12 sites upstream and downstream of Marmot Dam indicate median grain sizes (D 50 ) ranging from about 120 mm (RM 31.8) to about 50 mm (RM 2.5) (Figure 3). The Sandy River becomes sand-bedded within about 0.5 km of the Columbia River confluence. Large amounts of sediment are stored in active alluvial features (e.g., bars), especially in the lower river (as described in [2000]), and lateral migration of the river over a decadal time scale is limited. For the purposes of geomorphic analysis, the Sandy River from Marmot Dam downstream to the Columbia River can be delineated into five reaches (Reaches 1Β5), with the reach immediately upstream of Marmot Dam representing a sixth reach that will be affected by dam removal. Geomorphic reaches are shown in Figures 2 and 4, and the characteristics of these reaches are summarized in Table 1. Brief descriptions of these reaches are provided below; detailed reach-by-reach descriptions are provided in (2000). Table 1. Summary of geomorphic characteristics of Sandy River reaches delineated by Stillwater Sciences Reach Length (km) Average width (m) Average gradient Confinement Morphology Upstream of Marmot dam 2.5 50 0.0024 High Pool-riffle Marmot Dam to gorge (Reach 1) Sandy River gorge (Reach 2) Downstream end of Sandy River gorge to Bull Run River (Reach 3) Bull Run River to Dabney Park (Reach 4) Dabney Park to mouth (Reach 5) 2.4 45 0.008 Medium Forced pool riffle/plane bed 6.4 30 0.01 High Step pool 9.6 50 0.006 Medium Forced pool riffle/plane bed 20 70 0.0025 Medium/low Pool riffle/ plane bed 9.6 100 0.0007 Medium/low Pool riffle/dune riffle Upstream of Marmot Dam/Reservoir-influenced reach: The Sandy River upstream of Marmot Dam is affected by the backwater effect of the dam for a distance of approximately 1.9Β4 km. The impoundment formed by the dam has filled to near the dam=s crest with sediment and now functions as an alluvial river reach. Compared to upstream and downstream reaches, this reach has a lower gradient (about 0.002) and Page 4

smaller bed substrates (i.e., a higher percentage of gravels) as a result of the grade control provided by the dam and the backwater effect of the dam=s impoundment. The valley in this reach is about 40Β60 m wide. The length of time required for the reservoir to fill with sediment is not known, but it is generally believed that it was filled in the early years following the dam closure. Results of sediment sampling in the reservoir conducted for this project in 1999 are described in Section 5.1.3 below. Reach 1 extends from Marmot Dam to the upstream end of the Sandy River gorge and is 2.4-km long. This reach is characterized by a 0.008 gradient, moderate confinement at bankfull flow, and moderately pronounced forced pool-riffle morphology with a few small lateral cobble/boulder bars. The bed surface is armored and consists mainly of cobbles and boulders; gravels are limited. Several large woody debris accumulations create potential depositional zones in this reach. Reach 2 is the Sandy River gorge, which is 6.4 km in length and is bounded by 20Β30 m high bedrock strath terraces with steep hillslopes above. This reach is characterized by a 0.01 gradient, high confinement, and step-pool morphology with only patchy cobble/boulder deposits and long, deep bedrock pools that are separated by coarse-bedded riffles and boulder rapids. Large (housesized) boulders are present in the channel and create enhanced frictional losses, thereby reducing the effective shear stress available to transport sediment. In general, few deposition areas are present in this reach. Bedrock exposure is more common in the channel bed in this reach than in other reaches of the Sandy River. The steep gradient and high confinement in this reach create very high shear stresses, resulting in high sediment transport capacity. Reach 3 extends from the downstream end of the gorge (near Revenue Bridge) to the Bull Run River confluence at Dodge Park and is about 10 km in length. This reach widens considerably compared to Reaches 1 and 2 (with an average width of 50 m), has an average gradient of 0.006, and is characterized by forced pool-riffle morphology with many cobble/boulder bars and a cobble/gravel-dominated channel bed. Because the channel and valley bottom widen and gradient decreases downstream of the gorge, sediment transport capacity is reduced and the potential for sediment deposition increases. Shear stresses remain relatively high, however, and although average bed particle sizes in this reach decrease compared to Reaches 1 and 2, gravels suitable as spawning substrate are limited, occurring only in scattered patches and pool tail-outs (ODFW 1990, 1997). Reach 4 extends from the Bull Run confluence (Dodge Park) to Dabney State Park, a length of 20 km. This reach has an average gradient of approximately 0.0025, is bounded by high (mostly alluvial) terraces, and is characterized by pool-riffle morphology with many cobble/gravel bars. The channel bed is a mixture of cobbles, gravel, and sand. In Reach 4, channel confinement, gradient, and bed particle size decrease further compared Page 5

to reaches upstream, with these tendencies particularly evident in the reach from Oxbow Park (RM 11.9) to Dabney Park (RM 6.6). Large bars, side channel, overflow channel, and island features are present in larger magnitude and greater frequency. Sand content in the bed subsurface is generally high, notably increasing at Oxbow Park. Portions of the active bed and bars are covered with and/or saturated with sand. Reach 5 extends 9.6 km from Dabney State Park to the confluence with the Columbia River. This reach is characterized by a 0.0007 gradient, moderate-to-low confinement at bankfull flow, and dune-ripple morphology with large gravel/sand alternate and medial bars. The channel bed is a mixture of sand and gravel, is highly mobile, and has a very high sand content in the bed subsurface. The Sandy River delta forms the downstreammost portion of Reach 5. In the delta, the channel is sand-bedded and depositional dynamics are strongly influenced by the backwater effect of the Columbia River. 3. NUMERICAL MODELING OF SEDIMENT TRANSPORT FOLLOWING DAM REMOVAL developed one-dimensional numerical models of fine and coarse sediment transport to predict the routing of sediment from behind Marmot Dam downstream through the Sandy River. The numerical modeling allows assessment and comparison of sediment transport characteristics under different dam removal methods and river conditions. developed a coarse sediment transport model based on Parker=s surface-based bedload equation (Parker et al. 1982, Parker 1990) and a model of sand transport from reservoir deposits based on Brownlie=s bed material equation (Brownlie 1981). 1 Current sediment transport equations for gravel/sand mixtures are still at a developing stage, and verification of even the best equations in this category (e.g., Wilcock 1997, 1998) has been limited. To avoid using an equation for a gravel/sand mixture, we developed separate models to simulate gravel and sand transport. Separate modeling of gravel and sand was based on our hypotheses that (1) as the sediment is released from the reservoir deposit, gravel particles will be transported as bedload and sand will be transported mostly as suspended load because of the steep slope of the Sandy River, and (2) gravel and sand transport occur in different time scales (years vs. days). For example, a gravel particle may take years to travel 48 km, the distance from Marmot Dam to the Columbia River confluence. A sand particle, however, may travel the same distance, once it is exposed to the surface, in only several days during a flood event. Transport of fine sediment (including sand) was therefore modeled separately from coarse sediment (gravel). We acknowledge that in reality, transport of gravel and sand will each influence the transport rate of the other. Model results include predictions of the time required for sediment to be cleared from the reservoir area, 1 In discussions of the sediment transport modeling, we use Αgravel to refer to all particles with diameters greater than 2 mm (i.e., pebble, gravel, cobble, boulder). ΑSand is used to refer to particles less than 2 mm in diameter. Page 6

time required for sediment to travel out of the Sandy River (including various sub-reaches), thickness of downstream sediment deposits in various reaches (on a reach-averaged and cross-section-averaged basis), changes in deposition thickness through time, and total suspended sediment concentrations through time along the river=s longitudinal profile following dam removal. Questions that were explored with the numerical models for different dam removal alternatives include: How does sediment (both suspended and bedload) travel over the coarse alluvial and bedrock channel bed of the Sandy River? How does transport distance from the dam affect suspended sediment concentrations and coarse and fine sediment accumulations following dam removal? Is there a distance downstream at which no detectable changes are expected? Will substantial bed aggradation occur following dam removal or is the sediment transport capacity in reaches downstream of Marmot Dam high enough that little aggradation would occur? If there is an aggradational effect, how long will it last and in what reaches will it be most prominent? How do discharge conditions during and following dam removal affect downstream sediment transport and deposition characteristics? While numerical modeling provides estimates of sediment transport rates and cross-section and reachaveraged depths of sediment deposits, current state-of-the-art modeling typically cannot predict complex three-dimensional geomorphic responses over long river reaches and time scales, such as depositional patterns in channel cross section, local changes in sediment particle size distribution, infiltration of sand into the channel bed, or changes in the mobility of the existing channel bed downstream of the Marmot Dam. 3.1 Gravel Model The gravel model we developed is essentially the same as those of Cui et al. (2000) and Cui and Parker (2000), with minor adjustments to accommodate the specific conditions of the Sandy River and Marmot Dam. The Sandy River gravel model, as well as Cui et al. (2000) and Cui and Parker (2000), are based on Parker=s surface-based bedload equation (Parker et al. 1982, Parker 1990a,b, 1991a,b), which calculates gravel transport rate and bedload grain size distribution based on the grain size distribution of the surface layer and the boundary shear stress. The Parker equation (1990a, b) was developed to apply to gravel-bed streams and was not intended for application to sand or for suspended material of any size. It is only applicable to particles larger than 2 mm in diameter. Application of the Parker equation (1990a,b, 1991a,b) to a mixture with a relatively large amount of sand, such as the sediment accumulation behind Marmot Dam, may therefore create some error in predictions of the gravel transport rate. The model of Cui et al. (2000) simulated evolution of gravel pulses in a laboratory flume, and the model of Cui and Parker (2000) extended the model to a field scale to include abrasion as a factor affecting evolution of gravel pulses. Hansler et al. (1998) successfully simulated the evolution of a large landslide in the Navarro River, California with an early version of the Cui and Parker (2000) model. In both the laboratory flume and field cases, the model predicted that pulses traveled predominantly by dispersion rather than by translation, matching with flume and field observations. In comparisons of numerical Page 7

model predictions versus flume observations of the evolution of a sediment pulse, the numerical model performed best when the sediment pulse was coarser than the ambient sediment load (i.e., when there was less fine sediment in the pulse). As the sediment in the pulse became finer, with larger amounts of sand and silt, the model underpredicted the rate of pulse deformation (i.e., the pulse moved downstream in the flume and changed shape more quickly than predicted by the model) (Cui et al. 2000). The surface-based bedload equation used in the model (Parker et al. 1982, Parker 1990a, b) was based on field data for gravel and designed for gravel-bed rivers, which may explain their better performance in predicting the evolution of a coarser pulse of sediment. In comparisons of model predictions with the field case of the Navarro landslide, the model overpredicted the rate of pulse deformation, dispersing all the sediment in the landslide area in three years while field observation indicated that some landslide material remained in its original location three years after the landslide. The predicted areas of subsequent downstream deposition of landslide sediment, however, matched those observed in the field (Hansler et al. 1998; T. Lisle, pers. comm., 1999). Parker=s equation has been tested elsewhere in both laboratory flume and natural river situations. Cui et al. (1996) tested the Parker equation in three large-scale flume experiments, and Cui and Parker (1998) successfully applied the equation to predict channel profile. Uses of the equation in simulation of gravelbed rivers are also discussed in Cui and Parker (1997), Andrews (1994, 2000), Chang (1994), and Hoey and Ferguson (1994). The gravel transport model is written in FORTRAN as a DOS application. The gravel model provides predictions of thickness of downstream coarse sediment deposits in various reaches on a reach-averaged and cross-section-averaged basis (i.e., predictions are one-dimensional). Model results also include predictions of the time required for sediment to be cleared from the reservoir area, time required for sediment to travel out of the Sandy River (including various sub-reaches), and changes in deposition thickness through time. The gravel model is also used to generate predictions of sand (sediment finer than 2 mm and coarser than 0.0625 mm) and silt (sediment finer than 0.0625 mm) release from the reservoir, for use as input data to the sand model (as described in Section 3.3). 3.2 Sand Model We developed a one-dimensional model of sand transport from reservoir deposits based on Brownlie=s bed material equation (Brownlie 1981). Brownlie's equation was developed for sand-bedded rivers but is used here because no sediment transport equations exist to calculate sand transport in a bedrock or coarse sediment-dominated river. In applying Brownlie's equation of sediment transport and friction, the roughness height is modified to account for the bedrock, boulders, and gravel present along the bed of the Sandy River. We hypothesize that with these roughness adjustments, application of Brownlie=s equation can provide reasonable estimates of sand transport rates. Calibration and validation of this approach is required, however, and the error associated with applying Brownlie=s equation to a gravel-bed river, even with roughness adjustments, is not known. Some basic assumptions of this model include: sand transport can be represented as transport over a rough bedrock surface (i.e., the existing gravel bed of the Sandy River remains immobile with respect to sand transport); silt is transported as throughput load that is carried in suspension and cannot be deposited in the Page 8

channel bed; reservoir sediment is not cohesive; and sand release from the reservoir is completely dependent on the mobilization and transport of gravel out of the reservoir (i.e., data on the rate of sand transport out of the reservoir is derived from the gravel model, as described further in Section 3.3). With the sediment release information generated by the gravel model as the input (see Section 3.3), the sand model predicts the thickness of downstream sand deposition on a reach-averaged and cross-sectionaveraged basis, including the changes in deposition thickness through time. The model also predicts suspended sediment concentrations through time along the river=s longitudinal profile following dam removal. The sand transport model is written in FORTRAN as a DOS application. 3.3 Discussion of Modeling of Reservoir Erosion A key uncertainty in this modeling relates to the expected pattern of reservoir erosion and release of sediment from the reservoir. The model assumes laterally uniform sediment transport out of the reservoir, with sediment mobilization and transport derived from Parker=s (1990) sediment transport equation. This section describes the methods and assumptions used in modeling of reservoir erosion. The sediment deposit behind Marmot Dam is stratified, with a sandy layer at the bottom and coarser sediment in the upper layer of the deposit. This is because upon completion of the dam, the backwater zone behind the dam enabled the deposition of bedload as well as the settling of suspended particles. Thus, the first necessary element of the model is the capability to handle stratification of the reservoir deposit. Reservoir stratification is accounted for by dividing the reservoir deposit vertically into layers and assigning each layer with a different grain size distribution, based on the results of sediment coring by Squier Associates (see Section 5.1.3 for description of these results). Although the overall stratification of the reservoir results in a larger proportion of fine sediment in the lower layers of the reservoir and more coarse sediment in the upper layers, each layer in the deposit typically contains a range of grain sizes. Because we developed separate models of gravel and sand transport, modeling of erosion of the mixed layers in the reservoir deposit required simplifying assumptions, as described below. In the reservoir area, the model assumes that the erosion of the sediment deposit is exclusively dependent on the erosion of gravel within each layer of the deposit. The model predicts erosion of each layer according to the grain size distribution of the layer (based only on particles > 2 mm within the layer), the local slope, and discharge. As each layer in the reservoir deposit is exposed, the model assumes that the entire layer will be mobilized when shear stresses are sufficient to mobilize the gravel (> 2 mm) component of the layer, as indicated by the Parker equation. Because the Parker equation is not intended for application to particles smaller than 2 mm, it is not used here to directly estimate sand mobilization. Rather, erosion of sand is estimated on a volumetric basis, based on the estimated volume of sand in each layer of the deposit. As the gravels within a layer are mobilized, the sand volume within that layer is also mobilized and transported downstream. Sand is therefore mobilized and available for transport only when the gravel within the same layer as the sand is mobilized. Estimates of the volume of sand release from the reservoir deposit that are generated by the gravel model using this method are given as a daily Page 9

average value at the dam site and are used as the input for the sand model. The upstream boundary used in the gravel model starts about 4 km upstream of Marmot Dam, whereas the upstream boundary of the sand model is Marmot Dam. The model also assumes that transport out of the reservoir would be laterally uniform. Erosion of reservoir sediment would in fact likely result in incision of a channel within the valley walls, although the valley width is relatively narrow (40Β60 m) in the reservoir reach. Incision of a channel in the reservoir deposit could accelerate exposure of the underlying sand layer in the incised area and increase the time (compared to model predictions) required for sediment on the margins of the reservoir deposit to be eroded downstream. Informal observations of the evolution of reservoir sediment deposits following approximately 30 dam removals in Pennsylvania indicate that vertical incision almost always occurs, leaving isolated terraces or Αprotofloodplains (J. Pizzutto, pers. comm., 2000). We acknowledge that modeling of reservoir erosion incorporates major simplifying assumptions and is a key area of model uncertainty. Gravel and sand components in the reservoir will likely interact in manners that are not captured by this modeling. For example, sand transport from a given layer in the reservoir deposit will likely occur more rapidly than transport of the gravel in that layer, and the presence of a large amount of sand in certain layers likely reduces the validity of applying the Parker equation, as observed by Cui et al. (2000). In addition, sand following the gravel leaving the reservoir could smooth the bed and increase the mobility of the leading gravel front downstream (T. Lisle, pers. comm., 2000). In order to address uncertainties in modeling of sediment transport from the reservoir and to qualitatively assess the potential effects of incision, we performed sensitivity tests to evaluate how increased rates of sediment delivery from the reservoir to downstream reaches affect the model results, thereby simulating incision of a channel through the reservoir sediment deposit and more rapid downstream sand transport. These included a sensitivity test of the gravel model, in which we assumed that the accelerated sediment transport by downcutting is related to the cross sectionally averaged channel bed slope, and sensitivity tests on the sand model, in which faster rates of sand release from the reservoir were assumed. The results of these sensitivity tests are described in Section 6 below. 3.4 Uncertainties in the numerical modeling The sediment transport models developed for the Sandy River are state-of-the-art models based on tested sediment transport equations, and the models have been reviewed by experts in the field of sediment transport (Attachment A). Some uncertainty is inherent in numerical modeling, however, and the modeling effort described here contains a number of such uncertainties. Many hypotheses are incorporated in the models, both in terms of theoretical development (i.e., reflecting uncertainties in current scientific understanding about the mechanics of sediment transport) and input data. Key areas of uncertainty in this modeling effort, each of which is discussed in more detail either in previous sections or below, are as follows: reservoir erosion processes and sediment transport out of the reservoir area (Section 3.3); effects of coarse bed materials, especially large boulders, on sediment transport (particularly sand transport, as there are no equations for sand transport over a coarse bed); the separation of sand and gravel components in the modeling of sediment transport, rather than Page 10

modeling of a mixed sand/gravel mixture (sand and gravel transport are treated separately, although they do likely affect each other); simplifications in channel geometry used in the model (e.g., a rectangular channel is assumed, and widths do not vary with stage); rough assumptions used for several input parameters; uncertainties surrounding sediment transport mechanics. The models are one-dimensional, providing predictions of sediment deposition thicknesses that are averaged over the width of the channel cross section and that represent deposition over the existing channel bed. Model predictions do not account for local variations in shear stress caused by features such as deep pools, bedrock outcrops, or large boulders, and the amount of sediment actually deposited may be substantially higher or lower than predicted by the model in localized areas of the channel. The modeling also does not account for certain depositional processes, such as infiltration of fine sediment into the interstices of the channel bed and the production of sand and silt from gravel abrasion (i.e., suspended load estimates do not include products of gravel abrasion). The model results are most applicable on a reach-scale and time-averaged basis. Model results reflect a set of hydrologic conditions used as input, which was based on historical flow data from the Sandy River near Marmot gaging station. We modeled a range of hydrologic scenarios, as discussed in Section 5.1.2, to provide insight into the potential effects of different flow scenarios on sediment transport and deposition in the first year following dam removal. Use of a different set of hydrologic conditions would alter certain model predictions, such as the rate of sediment movement out of the reservoir reach. Although the specific hydrologic conditions that occur after dam removal will invariably be different than those used in the model, and extreme flow events were not modeled, the hydrologic scenarios we modeled represent a range of hydrologic conditions that could reasonably be expected to occur in the Sandy River following dam removal. The gravel and sand models require input parameters on a range of physical characteristics that influence sediment transport in the Sandy River (Sections 5.1 and 5.2), and these input data contain varying levels of uncertainty. For those input parameters to which the model is most sensitive (i.e., channel gradient, channel width, grain size distribution of the reservoir deposit, and water discharge), we used data collected specifically for this project or that were already available (water discharge). Uncertainty in these data should be relatively small, and sensitivity analyses were performed to examine the effects of varying input data on hydrologic conditions and on the grain-size distribution of reservoir sediment. For other input parameters, such as background gravel and sand transport rates, size distribution of bedload, and abrasion rates in the Sandy River, existing data were not available and new data were not collected for this project. For many of these input parameters, only order-of-magnitude estimates are required for the models, and we therefore used rough assumptions based on our observations of the Sandy River and on published data from elsewhere in the region. Many of the assumptions used in the modeling are conservative with respect to predicted impacts. For example, model runs assume a slightly greater amount of sediment would be released from behind Marmot Dam than volumes estimated by PGE and Squier Associates (2000). This is because our review of PGE photogrammetry data suggests a slightly greater upstream extent of reservoir sediment than estimated by Squier Associates (2000) based on their coring study. Page 11