Effects of Sediment Pulses on Bed Relief in Bar-Pool Channels

Home , Canal, Riffle

EARTH SURFACE PROCESSES AND LANDFORMS Earth Surf. Process. Landforms (2015) Copyright © 2014 John Wiley & Sons, Ltd. Published online in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/esp.3697

Effects of sediment pulses on bed relief in bar-pool channels

John P. P. Zunka,1* Desiree D. Tullos2 and Stephen T. Lancaster3 1 Department of Geosciences, Oregon State University, Corvallis, OR, USA 2 Department of Biological and Ecological Engineering, Oregon State University, Corvallis, OR, USA 3 Department of Geosciences and Institute for Water and Watersheds, Oregon State University, Corvallis, OR, USA

Received 19 December 2013; Revised 5 December 2014; Accepted 15 December 2014

*Correspondence to: John P. P. Zunka, Department of Geosciences, Oregon State University, Corvallis, OR, USA. E-mail: [email protected]

ABSTRACT: To further develop prediction of the range of morphological adjustments associated with sediment pulses in bar- pool channels, we analyze channel bed topographic data collected prior to and following the removal of two dams in Oregon: Marmot Dam on the Sandy River and Brownsville Dam on the Calapooia River. We hypothesize that, in gravel-bed, bar-pool channels, the response of bed relief to sand and gravel sediment pulses is a function of initial relief and pulse magnitude. Modest increases in sediment supply to initially low-relief, sediment-poor cross-sections will increase bed relief and variance of bed relief via bar deposition. Modest increases in sediment supply to initially high-relief cross-sections, characteristic of alternate bar morphology, will result in decreased bed relief and variance of relief via deposition in bar-adjacent pools. These hypothesized adjustments are measured in terms of bed relief, which we define as the difference in elevation between the pool-bottom and bar-top. We evaluate how relief varies with sediment thickness, where both relief and mean sediment thickness at a cross- section are normalized by the 90th percentile of observed relief values within a reach prior to a sediment pulse. Field measurements generally supported the stated hypotheses, demonstrating how introduction of a sediment pulse to low-relief reaches can increase mean and variance of relief, while introduction to high-relief reaches can decrease the mean and variance of bed relief, at least temporarily. In general, at both sites, the degree of impact increased with the thickness of sediment delivered to the cross-section. Results thus suggest that the analysis is a useful step for understanding the morphological effects of sediment pulses introduced to gravel-bed, bar-pool channels. Copyright © 2014 John Wiley & Sons, Ltd.

KEYWORDS: sediment pulses; dam removal; bed relief; bar-pool channels

Introduction The post-pulse reduction of bed topography is supported by theory (Madej, 2001) that establishes a possible response tra- While gravel supply is necessary for the formation of riffles jectory, which includes the burial or diminishment of channel and pools that form aquatic habitats (Lisle, 1982; Mossop structure and the subsequent channel reorganization and re- and Bradford, 2006), episodes of atypically large sediment covery of bed topography. However, other field and experi- supply (i.e. sediment pulses [Lisle, 2008]) can produce mor- mental data (Sutherland et al., 2002; Downs et al., 2009; phological responses that may negatively impact habitat Kibler et al., 2011; Gaeuman, 2013) report increases in the (Pizzuto, 2002) and increase flood risks (Downs et al., range of channel morphologies, which include pools and bars 2009). Whether from natural sources, such as landslides of varying size and shape, in a reach. The increases result from (Sutherland et al., 2002), or anthropogenic sources, such as a combination of scouring and filling pools and bar building. It dam removal (Shuman, 1995; Snyder et al., 2004), sediment is not yet clear which of these relief trajectories a channel may pulses can be of concern for water resource managers take in response to the introduction of a sediment pulse. (Shuman, 1995; Downs et al., 2009) through their potential However, it appears that the initial relief, defined herein as for filling of pools, homogenization of habitats, forcing in- the elevation difference between the deepest part of a cross- creases in width, or transitioning channels from single-thread section and the highest feature (e.g. top of sediment bar), of a to braided channels. bar-pool channel may be an important control on its response Past studies have found various and sometimes contradictory to a sediment pulse. Several examples suggest that initially high morphological responses to transient increases in sediment relief may decrease in response to a sediment pulse. For exam- supply associated with sediment pulses. Some field and exper- ple, a field-scale flume experiment (Podolak and Wilcock, imental data suggest that sediment pulses generally lead to re- 2013) showed that the introduction of a sediment pulse to a sta- duction of bed topography, whether through pool filling ble set of alternate bars (i.e. high initial relief) in a gravel-bed (Lisle, 1982; Madej, 1999, 2001) or development of braiding channel resulted in smaller, mobile, low-relief bars either conditions (Nicholas et al., 1995; Hoffman and Gabet, 2007). superimposing on or replacing the initial alternate bar structure. J. P. P. ZUNKA ET AL.

Similarly, flume experiments by Madej et al. (2009) demon- = cross-sectional area, A, above datum strated how the introduction of a moderate-sized sediment pulse to an alternate bar sequence (i.e. high initial relief) re- 15% 15% sulted in a decrease in the cross-channel relief as pools were B filled and the introduction of an even larger sediment pulse re- elevation sulted in a decrease in relief via the formation of low-relief, r mid-channel bars. In both experiments, the relief in the pools z recovered to the pre-pulse values with the subsequent passing of the sediment pulse (Madej et al., 2009). Following the re- b lease of a sediment pulse from a dam, Wohl and Cenderelli cross-channel distance (2000) measured a decrease in the initially high alternate bar relief as the pools filled and a subsequent recovery of relief as Figure 1. Definition diagram of terms at a cross-section used in anal- the pools were later scoured out. Lisle (1982) also observed de- ysis: A is the cross-sectional area above a datum; the minimum thalweg elevation of all survey years for each cross-section is used as the vertical creases in initially high relief to ‘riffle-like’ low relief due to the datum in this study; B is the bankfull width of the channel and b is the preferential pool filling during sedimentation associated with a width of the central 70% of the bankfull width; r is the elevation differ- large flood in California. ence between the thalweg and the highest point within b. z = A/B. In contrast to the responses of high-relief initial conditions, low initial relief may increase with the introduction of a sediment pulse. For instance, Ikeda (1983) noted how the introduc- indicators (e.g. ledges) in the pre-removal cross-section data tion of sand-gravel supply to a plane-bed flume (i.e. near zero (Dunne and Leopold, 1978), and the same width is used for initial relief) resulted in the development of small, mobile, each survey year for a given cross-section. For this study, where low-relief bedforms that grew over time into a stable set of depth to bedrock is not known for most cross-sections, the min- higher-relief alternate bars. Similar patterns of formation and imum thalweg elevation for each cross-section across all survey evolution of high relief alternate bars following introduction years is used as the vertical datum. We do not distinguish in the of a sediment mixture to a steep plane-bed flume have been re- analysis between types of bank and bed materials and refer to ported elsewhere (Lisle et al., 1991; Lanzoni, 2000b). Further- everything above the vertical datum as ‘sediment.’ Therefore, more, Gaeuman (2013) chronicled the development of a we also do not distinguish between sediment derived from high-relief alternate bar sequence following the injection of a the banks or pre-removal bed and sediment derived from up- slug of gravel into a low-relief bar-pool reach. stream of the dams. In this study, we evaluate whether initial relief conditions can For each study site, we normalize z and r by the 90th per- be used to predict the response of a bar-pool channel to the re- centile of the observed relief (R), measured in the respective lease of a pulse of sediment. To evaluate the proposed hypoth- study reaches prior to each dam removal, to form the non- eses (see later) for channel response, we compare the theorized dimensional variables: z*=z/R, r*=r/R. The values of and observed changes in bed relief as width-averaged sediment R = 1.2 m and 4.9 m are calculated from the pre-removal sur- thickness varied following the release of sediment pulses with veys for Brownsville and Marmot, respectively. The 90th per- the removal of Marmot and Brownsville Dams on gravel-bed, centile is chosen to scale values by dimensions of larger bars bar-pool channels in Oregon. that a channel is typically capable of forming, while avoiding bias by atypically large peak relief values, such as at a cross- section with considerable bank bias in the calculation of r. Conceptual Development We somewhat arbitrarily use the terms ‘high’ (r* > 0.5) and ‘low’ (r* < 0.5) to describe relief in a reach relative to other Analytical methods values in that reach. For each field site, z* and r* are calculated for each surveyed cross-section from each survey pe- In order to represent changes in pool depths and bar heights in riod. In addition, the mean, standard deviation, and 95% response to deposition or erosion, we operationally define confidence limits of each are calculated for each reach for cross-channel bed relief (r) as the elevation difference between each survey year (Tables I–III). The confidence limits are cal- the deepest part of a cross-section and the highest feature (e.g. culated using bootstrapping, whereby the populations of top of a sediment bar) (Figure 1). Hereafter referred to as ‘relief,’ mean and standard deviation values are resampled this difference is calculated within the central 70% (b) of the (n = 1000) at random to establish the 95% limits of the entire bankfull width (B). Fifteen percent of each cross-section’s resampled mean and standard deviation distributions. For width is removed from each boundary (Figure 1) in order to each cross-section, the pre-removal z* values are subtracted capture maximum bed elevations associated with in-channel from the post-removal values at successive times to yield bed sediment and exclude elevations associated with channel Δz* values for each post-removal survey year. banks. These exclusions do not apply to the minimum bed elevations, because banks often abut thalwegs, particularly at steep outer banks of channel bends. Excluding the outermost Hypothesis development 30% of the bankfull channel width will underestimate relief in situations where a sediment bar top connects smoothly with When composed of coarse material (Lisle et al., 1997; Cui the top of a channel bank and may cause a bank bias towards et al., 2003; Greimann et al., 2006) and introduced to gravel higher relief in channels where parts of banks are included in bed channels (Lisle et al., 2001; Lisle, 2008), sediment pulses b. As defined, relief is independent of stage, and changes in re- are typically modeled and observed in the field as a one- lief are dependent on evolving distributions of sediment depths dimensional stationary (i.e. not translating) dispersive wave in a cross-section. (but see Sklar et al., 2009). This view has been incorporated We calculate mean sediment thickness at a cross-section (z) into models of sediment pulse dynamics (e.g. Nicholas et al., as z=A/B, where A is the sum of the trapezoidally-integrated 1995) and provides a reasonable description of the spatial areas across the bankfull channel width (B) for each survey year and temporal trends that drive the changes in sediment thick- (Figure 1). Bankfull width is determined from topographic ness and relief posited by our hypothesis. If the maximum

Table I. Mean and standard deviation of r* and z* for Brownsville reaches

Reach, Location (m), Mean width (m) Year Mean r* Standard deviation r* Mean z* Standard deviation z*

I, 10 to 150 m, 30 m 2007 0.48 (0.35,0.69) 0.21 (0.04, 0.26) 0.21 (0.13,0.35) 0.13 (0.03, 0.17) 2008 0.75 (0.37,1.01) 0.37 (0.15, 0.49) 0.54 (0.31,0.73) 0.23 (0.16, 0.31) 2009 0.67 (0.38,1.21) 0.51 (0.20, 0.64) 0.52 (0.37,0.67) 0.18 (0.11, 0.25) II, 150 to 700 m, 33 m 2007 0.73 (0.64,0.81) 0.21 (0.18, 0.26) 0.57 (0.51,0.62) 0.13 (0.10, 0.19) 2008 0.71 (0.64,0.78) 0.17 (0.12, 0.23) 0.49 (0.43,0.54) 0.13 (0.10, 0.18) 2009 0.67 (0.57,0.77) 0.24 (0.19, 0.30) 0.64 (0.58,0.72) 0.16 (0.12, 0.23) Note: 95% confidence limits shown in parentheses. Downstream reach location and mean width for the reach are provided.

Table II. Mean and standard deviation of r* and z* for Marmot near-downstream (NDS) and farther-downstream (FDS) reaches

Reach, Location (m), Mean width (m) Year Mean r* Standard deviation r* Mean z* Standard deviation z*

NDS, 0 to 660 m, 58 m 2007 0.64 (0.59,0.72) 0.23 (0.17, 0.36) 0.38 (0.35,0.41) 0.11 (0.08, 0.15) 2008 0.43 (0.39,0.47) 0.13 (0.11, 0.15) 0.80 (0.77,0.83) 0.10 (0.09, 0.12) 2009 0.42 (0.35,0.48) 0.23 (0.20, 0.26) 0.74 (0.71,0.78) 0.11 (0.09, 0.13) 2010 0.40 (0.33,0.46) 0.21 (0.18, 0.24) 0.74 (0.71,0.78) 0.12 (0.11, 0.15) FDS, 1000 to 1600 m, 115 m 2007 0.79 (0.67,0.91) 0.23 (0.18, 0.29) 0.50 (0.44,0.57) 0.13 (0.10, 0.17) 2008 0.72 (0.62,0.82) 0.19 (0.14, 0.26) 0.54 (0.48,0.60) 0.12 (0.10, 0.16) 2009 0.77 (0.64,0.88) 0.22 (0.17, 0.29) 0.52 (0.47,0.59) 0.12 (0.09, 0.15) 2010 0.79 (0.68,0.90) 0.21 (0.16, 0.28) 0.53 (0.47,0.59) 0.12 (0.09, 0.16) Note: 95% confidence limits shown in parentheses. Downstream reach location and mean width for the reach are provided.

Table III. Mean and standard deviation of r* and z* for Marmot sub-reaches

Sub-reach, Location (m), Mean width (m) Year Mean r* Standard deviation r* Mean z* Standard deviation z*

I, 0 to 250 m, 61 m 2007 0.54 (0.47,0.58) 0.11 (0.07,0.15) 0.33 (0.30,0.36) 0.06 (0.04,0.09) 2008 0.43 (0.37,0.50) 0.14 (0.11,0.17) 0.82 (0.80,0.84) 0.04 (0.03,0.04) 2009 0.59 (0.53,0.66) 0.13 (0.09,0.18) 0.65 (0.63,0.67) 0.04 (0.03,0.05) 2010 0.58 (0.52,0.65) 0.13 (0.11,0.17) 0.64 (0.61,0.66) 0.05 (0.04,0.07) II, 250 to 360 m, 60 m 2007 0.99 (0.87,1.18) 0.24 (0.11,0.40) 0.53 (0.46,0.61) 0.12 (0.07,0.20) 2008 0.41 (0.30,0.49) 0.15 (0.10,0.19) 0.94 (0.91,0.97) 0.05 (0.03,0.08) 2009 0.43 (0.34,0.50) 0.13 (0.08,0.20) 0.90 (0.86,0.94) 0.06 (0.04,0.10) 2010 0.41 (0.32,0.48) 0.13 (0.08,0.19) 0.92 (0.88,0.95) 0.06 (0.04,0.11) III, 360 to 570 m, 53 m 2007 0.57 (0.51,0.63) 0.11 (0.09,0.14) 0.36 (0.32,0.40) 0.07 (0.05,0.10) 2008 0.45 (0.40,0.51) 0.10 (0.08,0.13) 0.75 (0.72,0.80) 0.08 (0.06,0.11) 2009 0.12 (0.10,0.14) 0.04 (0.03,0.06) 0.78 (0.74,0.82) 0.07 (0.05,0.09) 2010 0.13 (0.10,0.17) 0.06 (0.04,0.09) 0.79 (0.74,0.83) 0.08 (0.06,0.11) IV, 570 to 660 m, 60 m 2007 0.54 (0.45,0.62) 0.14 (0.10,0.17) 0.32 (0.29,0.36) 0.06 (0.03,0.07) 2008 0.43 (0.350,0.503) 0.13 (0.08,0.20) 0.67 (0.65,0.71) 0.05 (0.03,0.07) 2009 0.49 (0.36,0.62) 0.20 (0.14,0.30) 0.68 (0.66,0.72) 0.05 (0.03,0.07) 2010 0.44 (0.36,0.50) 0.11 (0.08,0.16) 0.68 (0.65,0.72) 0.05 (0.03,0.07) Note: 95% confidence limits shown in parentheses. Downstream reach location and mean width for the reach are provided. thickness of a wave is assumed to be located at a former dam or relief will respond by filling pools and reducing bed relief. Both landslide deposition site, and the pulse evolves by spreading scenarios may impact the variance in relief among the cross- downstream, the sediment thickness at any location down- sections of a reach. Increases in variance are projected for stream increases with the arrival of the wave’s leading edge reaches with low pre-pulse relief, and decreases in variance and subsequently declines as the pulse spreads out. A are projected for reaches with high pre-pulse relief. dispersion-dominated pulse is characterized by decreasing sed- The response of a cross-section with minimal pre-pulse relief, iment thickness downstream of its initial location at the dam low r*, to increasing sediment thickness depends on the nature and zero effect downstream of the pulse toe at a given time of the reach. In alternate bar reaches, which are characterized (Lisle, 2008). Our hypotheses extend this previous work on sed- by low r* cross-sections at riffles and cross-overs between alter- iment dispersion by proposing how spreading sediment will nate bars but high variance in relief at the reach scale, we hy- distribute across the width of channels as a function of initial pothesize that low relief will remain low in response to bed relief and thickness of sediment delivered to a location. increasing sediment thickness. The aggradation in this low r* We hypothesize that the change in relief due to the changes case would be due to uniform deposition across the width of in sediment thickness at a cross-section depends on the initial their characteristically planar form, and/or on the lateral mar- value of r*. More specifically, we hypothesize that, in response gins in small amounts (Wohl and Cenderelli, 2000). In contrast, to increases in sediment thickness, reaches with small pre- in plane-bed reaches that lack bars, pools, and riffles and are pulse relief will respond by forming and growing bars and in- characterized by low r* at cross-sections and low variance in creasing bed relief, whereas cross-sections with high pre-pulse relief at reach scale, we predict the introduction of a sediment

Copyright © 2014 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms, (2015) J. P. P. ZUNKA ET AL. pulse to build bars and associated riffles and to increase the both dams, so that they allowed passage of bedload sediment mean and variance of relief along the reach. The magnitude for most of the time since construction. Relative to the Brownsville of the increase in relief will vary with the sediment thickness Dam, Marmot Dam was taller and impounded 50× the sediment delivered to the cross-section, generally as a function of the dis- volume, which took 10 to 20× longer to accumulate. For both tance from the dam and the morphology of the reach. sites, we refer to the ‘water year’ (WY) as the time period from In a plane-bed channel, relief may be built, first, by the de- October 01 of one year to September 30 of the next year, with velopment of low-relief, mobile mid-channel bars upon intro- the labeled year (e.g. 2008 WY) being the latter year that marks duction of a sediment pulse, and, second, by subsequent the end of the water year (e.g. September 2008) rather than the channel reorganization to form higher-relief, stable alternate year in which it begins (e.g. October 2007). For the sake of bars (Ikeda, 1983). This transition from plane-bed to mid- brevity, ‘water year’ is implied after all numerical years channel bars to alternating bars has been extensively studied referred to hereafter. (see Venditti et al. [2012] for a summary of previous work) and may be explained with instability theory (Lanzoni, 2000a, and references cited therein). This relief-building trajec- Brownsville Dam on the Calapooia River, Oregon tory is likely influenced by many geomorphic processes, including selective transport and sorting of varying grain sizes The Brownsville Dam, near the town of Brownsville, Oregon, (Lanzoni, 2000b), secondary flow (Nelson, 1990), bed reorga- impounded the Calapooia River within its existing banks for di- nization during flood events (Rodrigues et al., 2012), and vari- version of flow to a canal (Figure 2). This small dam (1.8 m ations in sediment supply (Church, 2006; Venditti et al., height × 33.5 m width × 4 m breadth) impounded approxi- 2012). In short, the relief-building sequence appears to follow mately 14 000 m3 of sediment with a median grain diameter some variation on the following trajectory: migrating bars with in the gravel size range (D50 = 0.059 m), based on excavator- short wavelengths develop, then elongate, then stabilize at extracted bulk samples. According to local lore, the some longer wavelength into alternate bars (Ikeda, 1983; Nel- impounded sediment surface was flush with the top of the son, 1990). dam after two years. The dam was removed with an excavator Alternately, if a reach has an initially high relief due to and without dredging in August 2007. Dam removal and sedi- existing alternate bars prior to the arrival of a sediment pulse, ment evacuation revealed an older, smaller log-crib dam (1 m even small increases in the sediment thickness generated from height × 10 m width) at the same location, and this dam was left a sediment pulse may decrease the relief mean and variance by in place. filling bar-adjacent pools. From a process-based perspective, Pre- and post-removal data were collected over a reach of this pool filling may be associated with greater bed mobility, 1.8 km in length (0.4 upstream and 1.4 downstream of the even at relatively low flows (Lisle, 1982; Lisle and Church, dam); based on the predominance of erosion upstream and 2002), as a result of decreased bed armoring and sediment cal- the lack of significant deposition more than 720 m down- iber accompanying greater sediment supply (Buffington et al., stream, only that first 720 m downstream of the dam are in- 2002). The presence and geometry of a pool are reliant on cluded in the analysis (‘the study reach’). Within that 720 m, the dynamic balance between local hydraulic scour and the in- the gradient is 0.3%, and the channel has a single thread, a coming sediment load. An increase in the sediment load, such mixed ‘bedrock’ (resistant silt hardpan) and gravel bed, and as by the release of a sediment pulse, would tend to disrupt this low sinuosity (=1.04). In the first 400 m downstream of the balance, resulting in new pool geometries, likely with smaller dam, the bed is largely underlain by hardpan (partially lithified volumes (Buffington et al., 2002). The subsequent evacuation Missoula flood deposits), and the channel has a plane bed. In of sediment from a reach as the pulse disperses could again in- the remainder of the study reach, hardpan is exposed in some crease the mean and variance of relief as the previously- pools and alternate-bar morphology predominates. The flood- aggraded pools are scoured (Wohl and Cenderelli, 2000; plain and channel banks are composed of silt and gravel. The Madej et al., 2009). In response to sufficiently large increases dam was built into a basalt intrusion outcropped on the right in sediment thickness, such as those large enough to bank only at and immediately upstream of the dam site. completely fill existing pools and even aggrade the bed over The Calapooia River watershed comprises headwaters in the entire bankfull channel width (Madej et al., 2009), a similar the Western Cascades, the predominant source of basalt and sequence of building bar relief on a plane-bed channel (Ikeda, andesite gravel bedload (Sherrod and Smith, 2000), and low- 1983) may occur. For example, the flume experiments by Pryor lands of the Willamette Valley, and is tributary to the Willam- et al. (2011) and Madej et al. (2009) both documented the for- ette River. At the dam site, upstream contributing area is mation of a multi-thread channel with low-relief, mid-channel 394 km2, and annual peak flows range from 68 m3/s (recur- bars in response to an increase in sediment supply. Both studies rence interval, RI, of one year) to 515 m3/s for the flood of re- noted the subsequent development of a high relief alternate bar cord in December 1964. The annual peak flow record is sequence if the channel was given time to equilibrate or degra- based on regressions between the Brownsville Dam site, dation occurred following the reduction in supply. where a gage has been in operation since 2006, the historical flow record at the former US Geological Survey (USGS) gage at Holley (#14172000) located 15.3 km upstream of the dam Field Study of Oregon Dam Removals and draining 272 km2, and the active gages on the nearby South Santiam and Mohawk Rivers (see Walter and Tullos The data collected before and after the removal of the Browns- [2010], for further details). Peak flows generally occur during ville Dam on the Calapooia River and the Marmot Dam on the the rainy season in October through May. Low base flows oc- Sandy River provide the opportunity to investigate the relation- cur in the summer and are typically less than 2 m3/s (see Sup- ship between bed relief and sediment thickness changes associ- plementary Material, Figure A). The study reach has one small ated with a sediment pulse. Both rivers, located in western tributary, Wiley Creek, which enters the former reservoir from Oregon, are predominantly gravel-bed, single-thread channels river right at 90 m upstream of the former dam. with alternate bar morphologies. Both run-of-river dams formed Topographic data comprised cross-sections, longitudinal reservoirs that provided head for flow through diversions. thalweg profiles, and bar surfaces surveyed at summer Impounded sediment filled both reservoirs to meet the tops of baseflow in 2007, 2008, and 2009, or one year before and

Figure 2. Brownsville study site on the Calapooia River, Oregon (A) and Marmot study site on the Sandy River, Oregon (B). (A) Post-removal (ac- quired during the 2009 water year but prior to the 2009 survey) digital elevation model (DEM) of study reach on the Calapooia River shows locations of surveyed cross-sections, sub-reaches, and former Brownsville Dam (44°23′16″ N, 122°55′56″ W). (B) Pre-removal DEM of study reach on the Sandy River shows locations of former Marmot Dam (45°23′59″ N, 122°07′56″ W) and sub-reaches, but for clarity, densely-spaced cross-section locations are not shown. Inset: Map of Oregon with locations of each study site (black dots, labeled) and the corresponding drainage basins (gray).

two years after dam removal (Walter and Tullos, 2010; Zunka, terrace gravels (Major et al., 2012). There are no tributaries in 2011). Prior to the removal, Walter and Tullos (2010) divided the study reach. the study reach into habitat units (i.e. pool, riffle, run, or glide) The Sandy River watershed comprises 1300 km2 of the of varying lengths, and cross-sections divided each unit into Western and High Cascades of Oregon (Cui and Wilcox, three equal lengths of 0.25 to 2× the bankfull channel width, 2008) (Figure 2, inset) and is tributary to the Columbia River. so that, through the whole study reach, cross-section spacing The Sandy River sediment load consists of sand and coarse is non-uniform. volcanic gravel derived from the river’s origins at the base of Reid Glacier on the western slopes of Mt Hood and Pleis- tocene terraces bordering the river (Major et al., 2012). Among various bedrock controls along its length, the most no- Marmot Dam on the Sandy River, Oregon table is the steep (1%) and narrow (average width 30 m) Sandy River Gorge, which begins 2 km below the Marmot The Marmot Dam was a privately-owned concrete dam located Dam and continues for a distance of 7 km (Cui and Wilcox, on the Sandy River between the confluences with the Salmon 2008). Peak flows in the Sandy River basin occur during River and Bull Run. The dam (15 m height × 50 m width) was long-duration, low-intensity precipitation in fall and winter constructed in 1913, renovated in 1989, and diverted flow into and snowmelt and glacial melt in spring and summer. At the a canal and aqueduct system for hydropower generation (Major USGS gage site near Marmot Dam (#14137000), approxi- et al., 2012). The reservoir sediment, which had accumulated mately 48 km upstream of the confluence with the Columbia over an estimated period of 30 years, consisted of 730 000 m3 River, the Sandy River has a drainage area of 684 km2,a of sand and gravel, 328 500 m3 of which was gravel (Cui and mean annual flow of 38 m3/s, and a mean annual flood of Wilcox, 2008). A temporary cofferdam was built to retain the 460 m3/s, based on the 93-year gaging record (1912–2007) reservoir sediment as the dam was removed, and the cofferdam (Major et al., 2012). was breached on October 19, 2007. Topographic cross-section surveys were conducted annually Pre- and post-removal data were collected over a 1.6 km by David Evans and Associates at summer baseflow in 2007, reach (the ‘study reach’) downstream of the dam. In the study 2008, 2009, and 2010, or one year prior to and three years fol- reach, the gradient is 0.6% and the channel has a single thread, lowing removal. The cross-section spacing was 14 m, approxi- a boulder-cobble bed, and low sinuosity (=1.05). The channel mately 20% of the 65 m average bankfull width, in the 660-m banks and floodplain are composed of a combination of volca- reach downstream of the dam and 30 m, approximately 20% nic and sedimentary rocks, lahar deposits, and Pleistocene of the 130 m average bankfull width, in the reach from 1.0 to

1.6 km downstream of the dam site. Cross-sections were not year RI floods but no two-year RI events, and the low volume of surveyed from 660 to 1000 m. See Major et al. (2012) for a de- sediment stored behind the dam resulted in only small morpho- tailed description of data collection at the Marmot Dam site. logical changes downstream. The Calapooia River evacuated 29% of the reservoir’s 14 000 m3 of sediment in 2008 and an additional 10% of the original sediment volume in 2009. By Results 2009, the released sediment covered the previously exposed hardpan in the first 400 m downstream of the dam site with a Brownsville Dam removal high-relief alternate bar sequence, although changes in sediment thickness were modest in magnitude and limited in spa- The moderate flows following the removal of Brownsville Dam tial extent, with Δz* values exceeding 0.3 (Figure 3) at only (see Supplementary Material, Figure A), including several one- the first and third cross-sections (XS1, XS3) at 10 and 100 m downstream, respectively (Figure 2). 0.6 2008−2007 The cross-sections were grouped into two reaches (Figure 3, III 2009−2007 0.4 Table I) based on the observed spatial and temporal patterns in 0.2 net erosion and deposition following the removal. Reach I, 0 comprising the first four cross-sections (Figure 2A), had rela- XS2 XS3 tively low initial relief (mean r* < 0.5, Figure 4, Table I), due −0.2 to sparse sediment cover over the relatively planar silt-hardpan −0.4 0 100 200 300 400 500 600 700 bed. The largest Δz* and most dramatic adjustments in channel downstream distance (m) morphology were in this near-dam Reach I over the first year following dam removal (Figure 3). Reach II, comprising cross- Figure 3. Change in normalized mean sediment thickness (Δz*) ver- sections 5–24 (XS5–XS24), had a much wider range of relief sus distance downstream of Brownsville Dam. Calculated from cross- prior to dam removal than Reach I (Figure 4). Following the re- section data, lines represent difference in normalized mean sediment moval of Brownsville Dam, Reach II was generally degraded in thickness between each post-removal survey and the pre-removal sur- 2008 (Δz* < 0) relative to the pre-removal surface, and vey. The study area is divided into two reaches, I and II. Locations of se- aggraded (Δz* > 0) in 2009 (Figure 3). lect cross-sections are noted. The increases in sediment thickness for Reach I 1.4 corresponded to the development, and subsequent modifica- 2007 tion, of sediment bars that increased initially low relief and de- 2008 1.2 2009 creased initially high relief. A mid-channel bar, which developed at XS1–XS4 (e.g. Figures 5A and 5C), increased mean relief from low to high and increased variance of relief 1 values nearly 150% relative to pre-removal (Table I). This increase from initially low relief (r* < 0.5) was especially evident 0.8 at XS3 (Figures 5C and 5D); XS2 exemplifies the decrease from r initially high relief (r* > 0.5) (Figures 5A and 5B). 0.6 The development of an alternate bar sequence from 2008 to 2009 illustrated the local impacts of channel reorganization. 0.4 The mid-channel bar deposited in Reach I in 2008 attached to the left bank at its lower extent and formed a river-left 0.2 bar in Reach II in 2009. Despite the negligible to small changes in sediment thickness at the upstream end of Reach 0 II in 2008 (Figure 3), the bar deposition, and accompanying -0.4 -0.2 0 0.2 0.4 0.6 0.8 scour of the bar-adjacent pool, led to greater r*. Farther downstream (>200 m distance), an alternate bar ~100 m in Figure 4. Plot of change in normalized mean sediment thickness length formed at river right in 2009. With bar formation, rel- (Δz*) and normalized relief (r*) for pre- and post-removal Brownsville atively small increases in sediment thickness (e.g. Δz* = 0.05 data from both reaches. Data from the four Reach I cross-sections are at 258 m distance) led to relatively large increases in relief displayed with large symbols. (e.g. r* = 0.64 in 2007 to r* = 0.89 in 2009 at 258 m distance).

114 A B 2007 1.2 113 2008 2009 1 0.8 112 0.6 r 111 0.4 vertical exaggeration = 3 0.2 110 0 114 C 2007 D 1.2 113 2008 2009 1 0.8 112 r 0.6 111 0.4

elevation (m) 0.2 110 0 0 5 10 15 20 25 30 35 40 45 -0.4 0 0.4 cross-channel distance (m)

Figure 5. Example pre- (2007) and post-removal (2008, 2009) channel cross-sections for Brownsville with corresponding plots of change in normalized bed sediment thickness (Δz*) and normalized bed relief (r*). View is facing downstream, i.e. river right is on the right side of the figure. The vertical exaggeration for each cross-section is three. Data points in bed relief plots correspond to survey years for each cross-section. XS2 (A, B) is located 54 m downstream of Brownsville Dam in Reach I, and XS3 (C, D) at 98 m downstream in Reach I.

The growth of this bar-pool morphology was accompanied aggradation in 2009 after initial deposition; Reach IV experi- by riffle formation (at 150 to 180 m distance, Figure 3), where enced only small changes in sediment thickness after initial de- r* decreased by 0.25 or more relative to initial conditions of position. In the FDS reach, there was limited impact on relief, r*, which ranged 0.55 to 0.93. Thus, despite only minor primarily through the modification of existing bars and pool changes in Δz* in the upstream portion of Reach II, bar filling. growth increased relief and the creation of cross-over mor- The evolution of the NDS sub-reaches at Marmot demon- phology decreased relief. strated several morphologic variations of decreasing initially Downstream of 300 m distance, changes were characterized high relief in response to the large increases of sediment thick- by relatively uniform degradation and aggradation across the ness (Figure 7). In Reach I where pre-existing relief was high, width of the cross-sections in 2008 and 2009, respectively, with the large increases in z* in the year following removal de- little alteration of bar morphology, so that changes in relief from creased relief (Figure 7A, Table III). The mean relief then in- initially high mean r* values were generally small (Figure 4) and not statistically significant (Table I).

1 A Marmot Dam removal 0.8 The removal of the earthen cofferdam upstream of Marmot Dam caused substantial and rapid channel responses in the 0.6 Sandy River. The breach occurred during the rising limb of 0.4 a moderate (<100 m3/s) flow event (see Supplementary Mate- rial, Figure A), and the reach immediately downstream of the 0.2 dam transitioned, literally overnight, from a single-thread, boulder-bed channel into a multi-thread channel with mobile 0 sandy bars (Major et al., 2012). In 2008, the first year follow- 1.6 ing the removal, more than 50% of the sediment left the for- B 2007 3 mer reservoir, and 110 000 m was deposited in the first 2 km 1.4 2008 downstream of the former dam (Major et al., 2012). This de- 2009 position corresponded to a maximum Δz* of 0.6 near the 1.2 2010 dam site to a minimum of nearly zero at the distal end of the 2 km reach (Figure 6). Sediment thickness near the dam 1 site declined in 2009 (smaller Δz* in Figure 6) but remained nearly constant downstream. Moreover, the entirety of the r 0.8 near-downstream (NDS) reach remained aggraded (Δz* > 0) through 2010. 0.6 The cross-sections downstream of the Marmot Dam site were grouped into two reaches, a NDS reach, with large, statistically 0.4 significant increases in sediment thickness, and a farther- downstream (FDS) reach, with small, insignificant changes in 0.2 sediment thickness (Figures 2 and 6, Table II). In addition, we further subdivided the NDS reach into four sub-reaches 0 – 1 (Reaches I IV in Figure 6, Table II) based on the following pat- C terns in Δz*: Reach I had the largest increases in sediment 0.8 thickness in 2008 and, relative to that ‘high stand,’ degradation in 2009; in Reach II, the sediment thickness changed little after 0.6 initial deposition in 2008; Reach III experienced continued

0.4

2008−2007 0.2 0.5 2009−2007 2010−2007 0.4 0 1 0.3 D 0.2 I II III IV FDS 0.8

0.1 NDS 0.6 0 XS12 XS18 XS28 −0.1 0.4 0 200 400 600 800 1000 1200 1400 1600

downstream distance (m) 0.2

Longitudinal profile depicting normalized change in mean Figure 6. 0 sediment thickness (Δz*) with distance downstream of Marmot Dam 0 0.2 0.4 0.6 on the Sandy River. Profiles are divided into near-downstream (NDS) and farther-downstream (FDS) reaches, and NDS reach is subdivided into four sub-reaches: I, II, III, and IV. See text for explanation of sub- Figure 7. Normalized change in mean sediment thickness (Δz*) and reach divisions. Select cross-section locations are shown. No survey normalized relief (r*) plots for sub-reaches Reach I, Reach II, Reach data exist for downstream distances between 660 and 1000 m. III, Reach IV (A–D, respectively) of the NDS reach for Marmot.

Copyright © 2014 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms, (2015) J. P. P. ZUNKA ET AL. creased in 2009 and 2010 when erosion subsequently reduced 1.6 z* values. For example, the downstream 100 m of Reach I 2007 1.4 1 (e.g. XS12, Figures 8A and 8B) was characterized by high relief 2008 2009 and a deep single-thread channel prior to dam removal. In 1.2 2008, this section underwent a decrease in r*(TableIII)dueto over 3 m of width-averaged deposition, which was manifested 1 0.4 as multiple shallow channels (Figure 8A). The degradation in 00.5 2009 then excavated a single channel on river left and increased r 0.8 mean relief relative to 2007 and 2008 values (Figure 7A). Reach II transitioned from the highest initial mean relief at Marmot to low 0.6 mean r* post-removal (Figure 7B, Table III). In this reach, all post- removal cross-sections registered lower r* values than the lowest 0.4 pre-removal value (Figure 7B). The key morphological response 0.2 to the substantial deposition in Reach II was the creation of a low-relief, mid-channel bar in 2008 that completely filled the 0 U-shaped pre-removal channel and deposited sediment on 0 0.1 0.2 0.3 0.4 0.5 0.6 the tops of the banks (Figure 8C). This bar became attached to the right bank in 2009 (Figure 8C). The initially high relief Figure 9. Plot of normalized change in mean sediment thickness of Reach III was reduced during each of two phases of aggrada- (Δz*) and normalized relief (r*) for Marmot NDS reach and FDS reach tion in 2008 and 2009 (Figures 6 and 7C, Table III). The (inset). Data from 2010 are excluded in this figure for clarity because morphological responses included the deposition of a low- there is little channel change between 2009 and 2010. See Figure 7 relief, river-left bar in 2008 and the creation of low-relief, for 2010 data. riffle-like morphology in 2009 (Figures 7C and 8E, Table III). The additional increase in z* in 2009 resulted in a planar channel (Figure 8E) with the smallest r* values measured 7D, and 9). As a result, the mean r* values in the NDS for any year (Figure 8F). Reach IV experienced a slight reach decreased substantially pre- to post-removal and did decrease from initially high mean relief to low relief not recover during the survey period (Figures 7 and 9, following deposition of a river right bar in 2008 and limited Table II). Thus we observed an overall and persistent de- subsequent modification of sediment thickness, morphology, crease in relief with large increases in z* for the initially or relief (Figure 7D, Table III). Thus in the initially high-relief high relief NDS reach. sub-reaches, pre-removal morphology was completely bur- The sequence of decrease and subsequent increase in relief ied by the deposition of new bars, which had lower relief variance across the NDS reach was consistent with shallow than the pre-removal morphology. bar deposition in 2008 and channel reorganization in 2009, The responses of bed relief in the NDS and FDS reaches respectively. In general, no statistically significant changes demonstrated the substantial effects of large magnitudes of in relief variance occurred in any individual NDS sub-reach, sediment deposition and limited effects of small magnitudes except for the notable decrease in Reach III in 2009 to a of deposition, respectively (Figure 9, Table II). In the FDS near-zero value with the development of riffle-like morphol- reach, reach-averaged Δz* values were less than 0.04 for ogy (e.g. Figure 8E) across all Reach III cross-sections all survey years (Figure 6) and produced only small changes, (Table III). However, the NDS reach underwent a statistically both positive and negative, in mean relief (Figure 9, inset), significant decrease in the relief variance (Table II) associated but the changes were not statistically significant (Table II). with the aggradation and widespread development of shal- In the Marmot NDS reach, over 90% of the post-removal low, mobile bars in the sub-reaches in 2008. Following chan- cross-sections and all cross-sections from sub-reaches III nel reorganization in 2009, the sub-reaches spanned separate and IV had Δz* values greater than 0.3 (Figures 6, 7C, channel units, with Reaches I and II and Reach IV

1 A B 0.8 213 0.6 r 211 0.4 0.2 209 0 1 C D 0.8 212 0.6 r 210 0.4 0.2

elevation (m) 208 0 213 1 E 2007 F 0.8 211 2008 2009 0.6 2010 2007 r 209 2008 0.4 2009 0.2 vertical exaggeration = 2.4 2010 207 0 0 10203040506000.20.4 0.6 cross-channel distance (m)

Figure 8. Select pre- (2007) and post- (2008, 2009, 2010) removal cross-section surveys for Marmot with corresponding plots of normalized change in bed sediment thickness (Δz*) and normalized bed relief (r*). Cross-section view is oriented looking downstream. Vertical exaggeration for each cross-section is 2.4. Datapoints in bed relief plots correspond to survey years of each cross-section. XS12 (A, B) is located 180 m downstream of Mar- mot Dam, within Reach I, XS18 (C, D) is located 265 m downstream, within Reach II, and XS28 (E, F) is located 420 m downstream, within Reach III. To supplement the limited coverage of the 2007 cross-section for XS18, the cross-section points from 40 to 50 m were taken from the 2005 survey.

Copyright © 2014 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms, (2015) EFFECTS OF SEDIMENT PULSES ON BED RELIEF IN BAR-POOL CHANNELS corresponding to two river-right bars, and Reach III corre- reaches of both sites (Tables I and II), the larger Δz* at Marmot sponding to ‘riffle-like’ morphology. The development of this modified the morphology of the cross-sections more pro- bar–riffle–bar sequence increased relief variance for the reach foundly than the smaller Δz* at Brownsville, but the relief met- to pre-removal levels (Table II). Therefore the morphologic ric and comparative statistics did not capture these qualitative variability in the NDS reach decreased as only a single mor- differences in morphological modification. phology (i.e. low-relief bars) was represented in each sub- The depth of sediment delivered to a feature impacted the reach, but increased as both bar and riffle morphologies ap- nature and degree of morphological response. Madej et al. peared in the sub-reaches in 2009. (2009) showed that, when introduced to an alternate bar sequence, small sediment pulses tend to deposit in the bottom of pools and leave pre-existing bar structure intact, but larger Discussion sediment pulses tend to drown out pre-existing morphology and form, albeit temporarily, low-relief mid-channel bars. The The two case studies presented herein indicate that some- analogous behaviors occurred at the Marmot FDS and NDS what predictable behaviors of building or reducing relief reaches, respectively, demonstrating the relative impacts of may occur following the introduction of a sediment pulse. small versus large sediment thicknesses. The high relief pre- The specific behavior of pool filling or bar building appears removal morphology of the NDS reach was thickly buried in to depend on the initial relief and the amount of sediment 2008 by mobile, mid-channel bars that greatly reduced bed re- delivered to individual features. Like other studies (Sklar lief. In contrast, in the FDS reach, changes in sediment thick- et al., 2009), this work confirms the importance of initial con- ness were small and mean relief was minimally affected by ditions and the magnitude of the sediment pulse on channel the subtle bar growth and pool filling. At Brownsville, channel response. The Brownsville and Marmot Dam removal sites reorganization only occurred immediately downstream of the and respective sub-reach divisions provided useful contrasts dam, where deposition of the largest sediment thicknesses re- in thicknesses of sediment introduced to the downstream sulted in the formation of alternate bars. The normalized in- channel, ranges of initial relief, and resulting morphological creases in sediment thickness at Brownsville were not as great impacts with which the hypothesized responses were as the NDS reach at Marmot, and mobile, low-relief, mid- evaluated. channel bars did not develop. The results suggest the hypothesis that the development of mobile mid-channel bars, effec- tively as a lesser degree of braiding, is related to the Effect of initial relief on channel response introduction of larger sediment thicknesses on the continuum of channel responses to sediment pulses. Results from both study sites generally confirmed the hypothe- Whereas the sediment thickness is the cross-sectional area sis that initial relief is an effective predictor of the impacts of a divided by the channel width, the fact that widths in the FDS sediment pulse on channel morphology. Increasing sediment reach are nearly twice those in the NDS reach may account thickness tended to increase relief in locations with initially for some of the differences between the reaches. However, low relief and decrease relief in locations where it was initially we would also expect that the velocity of the sediment flux high. Several studies (Wohl and Cenderelli, 2000; Madej et al., would be lower in a wider channel, and the cross-sectional 2009; Podolak and Wilcock, 2013) documented decreases in area of sediment is the volume flux divided by the velocity. initially high relief with the introduction of sediment pulses These two effects on the sediment thickness would tend to can- through either pool-filling or development of low-relief mid- cel. Average Δz* values in the NDS reach are an order of mag- channel bars. These behaviors were both visible at Marmot, nitude greater than in the FDS reach, where that average is, where the pools of the FDS reach were filled and where initially statistically, indistinguishable from zero. It appears, then, that high relief of the NDS reach was decreased with the formation the sediment flux in the FDS reach is still much smaller than of mobile, sandy bars. In flume studies where sediment pulses in the NDS reach. were introduced to low relief plane bed reaches (Ikeda, 1983; Lisle et al., 1991; Lanzoni, 2000b), relief increased with the development of sediment bars. This behavior was evident in the first 300 m downstream of the Brownsville Dam, where low re- Role of alternate bars in maintaining relief variance lief was increased with the development of an alternate bar sequence. The results from Marmot and Brownsville indicated that development of alternate bars during the processing of a sediment pulse can maintain, and even increase, the variance in relief Effect of variable sediment thickness on channel at the reach scale. While the mean relief increased with bar de- response position in the plane-bed Reach I at Brownsville, variance in relief increased as well with the creation of both high-relief bar- The magnitude of the impacts of the sediment pulses on chan- pool and low-relief riffle-like morphology of the alternate bar nel morphology and relief at Brownsville and Marmot varied sequence. The prevalence of mobile, low-relief bars in most with the thickness of sediment delivered to each reach in pat- of the NDS reach at Marmot in 2008 represented a lack of mor- terns consistent with the literature. The impacts of sediment phologic variability and resulted in the decreased variance of pulses are usually largest at the site of the former dam and de- relief observed. Then, in 2009, as alternate bars stabilized in crease downstream (Lisle et al., 1997; Cui et al., 2003; Reaches I, II, and IV and the cross-over morphology appeared Greimann et al., 2006; Lisle, 2008). Indeed, at both study sites, in Reach III, the relief variance in the NDS reach increased, the largest changes in relief were nearest the dam site, where consistent with our hypothesized models of channel change. the largest increases in sediment thickness occurred, and im- The pattern of reorganization of mid-channel bar sediments to pacts decreased with distance downstream. The contrasting rel- alternate bar morphology is consistent with the responses to in- ative sediment pulse size between Brownsville and Marmot creased sediment supply observed by Ikeda (1983) and is tied resulted in the differences in the scaled responses in z*. Al- to measurable increases in relief variance and morphological though magnitudes of change in r* were comparable in the diversity.

Study scope and further work insignificant to, or overwhelmed by, the substantial deposition that occurred. This study contributes to the understanding of potential mor- Finally, the hypothetical at-a-cross-section responses in this phological responses to sediment pulses in bar-pool morphol- study, when examined collectively, generated a rough three- ogy. Bar-pool systems are common throughout many rivers dimensional portrait of reach response to sediment pulses. with gradients less than 2% (Grant et al., 1990) but are rare in These responses corresponded to the hypotheses that bar-pool steeper gradient streams (Buffington et al., 2002). Thus, the hy- channels will build bars where bars are lacking and fill pools pothesized trajectories are limited to bar-pool systems in rivers where bars are developed. However, a more complete model not characterized by steep gradients (<1% in sites examined based on three-dimensional topography may provide addi- herein) and should be evaluated in other initial conditions (e. tional insight regarding channel responses and the sediment g. morphological, hydrological, pulse characteristics) where processes driving them. The hypotheses and simple heuristic the dominant sediment processes are different. As the hypothe- model in this study are intended to establish expectations for ses are further examined, several issues should be considered. how channels evolve in response to sediment pulses; they are First, the potential for decoupling the patterns of mean and var- not intended to supplant more detailed morphodynamic iance of relief exists if a given study reach does not span more models or to provide an exhaustive description of the dominant than a single alternate bar sequence. The hypotheses predicted sediment transport processes and mechanisms driving morpho- that the development of alternate bars from a plane bed mor- logical responses. phology would increase both mean and variance of relief and While the body of literature on evolving fluvial topography is that the filling of pools would decrease both mean and variance growing richer from flume studies (e.g. Ikeda, 1983; Lanzoni, of relief. However, if a study reach spans only a single bar, in- 2000a, 2000b; Venditti et al., 2012) and field studies (e.g. creasing relief uniformly for all cross-sections in the reach Welford, 1994), some critical questions remain regarding the may actually decrease the variance in relief while increasing impacts of a sediment pulse on the sediment dynamics that the mean. Given that variance in relief is highly dependent build bars (Church, 2006; Venditti et al., 2012) and that are re- on the spatial scale (e.g. a single bar or an alternate bar se- sponsible for the filling and scouring of pools (Lisle and Hilton, quence) over which it is measured, the morphological context 1992; Sear, 1996; Buffington et al., 2002; MacWilliams et al., of a set of cross-sections within the channel planform should 2006). For example, the importance of the hydrologic sequence be considered when evaluating these hypotheses. post dam removal is increasingly reported (Pearson et al., 2011; Second, while small pulses are likely to follow the hypothe- Sawaske and Freyberg, 2012) to be of limited relevance in the ses outlined earlier, the introduction of a sediment pulse larger rates and sequences of sediment dynamics, with process-based than those presented herein may fill the channel to generate erosion of the reservoir sediments initially dominating, while major and unpredictable reorganization of the bed. The mor- initial bed slopes are large and grain sizes small, until event- phologic responses may be more variable in response to a based dynamics take over (Pizzuto, 2002). And yet, within-year larger pulse and could include channel widening, pool filling hydrologic variability has been shown to drive variability in the and/or deepening, the development of low-relief braided mor- caliber and quantity of sediment being transported to pools and phology (Buffington et al., 2002), streamwise shifting of riffle the strength of the flow field that is responsible for scouring sed- crests, or avulsion. iment from pools (Sear, 1996). The continued evolution of the Third, rather than a retrospective analysis, it may be desir- understanding of sediment transport in gravel-bed streams and able to modify the approach outlined herein to predict poten- the mechanisms that drive the building and elimination of relief tial channel responses prior to future dam removals. Such an with a transient sediment pulse will facilitate understanding of analysis should modify the existing approach by: (1) using the where the hypotheses developed herein apply and when other pre-removal thalweg elevation, as opposed to minimum eleva- trajectories are likely to occur, as well as advance the fields of tion across all survey years, to establish the vertical datum for geomorphology and river management. each cross-section, and (2) discretizing sub-reaches by initial relief or morphologies apparent from two-dimensional deposi- tional patterns, such as alternate bars. Conclusion Fourth, bank bias in the calculation of relief values can over- state decreases in relief. Where deposition results in decreased Reasoning that, in so far as bed relief is a reasonable proxy for relief, inclusion of banks in calculations of initial relief may the kind of ‘channel complexity’ often associated with high- simply exaggerate that apparent result. Where bars, either quality aquatic habitat, we addressed the importance of initial mid-channel or alternate, are initially absent, however, the re- bed relief and changes in sediment thickness on channel results due to bank bias may be misleading. For example, many sponse to sediment pulses. Specifically, we examined the hy- pre-removal Marmot cross-sections, particularly in Reaches II potheses that deposition from a sediment pulse would and III, exhibited a U-shaped morphology that lacked sediment increase relief by building bars in channels with initially low re- bars (e.g. XS28; Figure 8E) but, nevertheless, produced high lief, whereas pool filling would reduce relief in channels with mean r* values (Table II) because significant bank relief was in- initially high relief. Additionally, we hypothesized concomitant cluded in the central 70% of the cross-section. increases and decreases in the variance of relief in response to Fifth, the effect of the hardpan in Reach I at Brownsville on the growth of alternate bar sequences and filling of pools, channel morphology in response to increased sediment load respectively. is uncertain. This resistant surface would limit the erosion of The results generally supported these hypotheses. Following the bed and therefore may stunt potential pool growth. Indeed, the introduction of a sediment pulse, high relief generally de- in cross-sections where hardpan was known to be exposed in creased and low relief increased, and greater changes in sedi- the bed (e.g. Brownsville Reach I), there was no evidence deg- ment thickness produced greater changes in the mean and radation below the hardpan. Alternatively, the smooth hardpan variance of relief. Whereas relief variance either changed little surface may inhibit pool-filling by efficiently passing bedload to (as in the Sandy River) or increased (as in the Calapooia) fol- rougher, sediment-covered patches of bed downstream (Iseya lowing dam removal, the results suggest that dam removal and Ikeda, 1987). However, provided the volume of sediment may have limited negative impacts on, or even increase hetero- introduced to Reach I, it is likely that this effect would be geneity in, channels over the long-term. The range of

Copyright © 2014 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms, (2015) EFFECTS OF SEDIMENT PULSES ON BED RELIEF IN BAR-POOL CHANNELS responses, including eventually increasing relief following Greimann B, Randle T, Huang J. 2006. Movement of finite amplitude short-term reductions, suggest that, even where sediment sediment accumulations. Journal of Hydraulic Engineering 132(7): pulses initially reduce bed relief by filling pools, diversity of 731–736. DOI. 10.1061/(ASCE)0733-9429(2006)132:7(731) aquatic habitat may soon recover with the development of al- Hoffman DF, Gabet EJ. 2007. Effects of sediment pulses on channel ternate bar morphology. While the hypotheses did not explain morphology in a gravel-bed river. Geological Society of America Bulletin 119:116–125. DOI. 10.1130/B25982.1 the full range of responses observed at the two field sites, this Ikeda H. 1983. Experiments on Bedload Transport, Bedforms, and work does provide a starting point for estimating the range of Sedimentary Structures using Fine Gravel in the 4-meter-wide responses to dam removal of alluvial channels that exhibit Flume. Environmental Research Center Paper 2. University of bar-pool morphology and for developing more physically de- Tskuba: Tskuba. tailed and site-specific hypotheses. Iseya F, Ikeda H. 1987. Pulsations in bedload transport rates induced by The results advance the understanding of the range and types a longitudinal sediment sorting: A flume study using sand and gravel of morphological responses possible in bar-pool channels to mixtures. Geografiska Annaler. Series A. Physical Geography 69(1): the introduction of small- and medium-sized pulses and have 15–27. several implications for managers concerned about homogeni- Kibler KM, Tullos DD, Kondolf GM. 2011. Learning from dam removal zation of aquatic habitats. First, the potential for new deposition monitoring: challenges to selecting experimental design and estab- to fill pre-existing pools is linked to initial relief and expected lishing significance of outcomes. River Research and Applications (8): 967–975. DOI. 10.1111/j.1752-1688.2011.00523.x changes in sediment thickness. In the cases examined here, 27 Lanzoni S. 2000a. Experiments on bar formation in a straight flume: 1. pools that filled typically had high initial relief, as hypothe- Uniform sediment. Water Resources Research 36(11): 3337–3349. sized. Second, bank erosion and channel widening are not DOI. 10.1029/2000WR900160 necessary conditions for decreased bed relief. Third, where Lanzoni S. 2000b. Experiments on bar formation in a straight flume: 2. bars are initially lacking, sediment additions may increase or Graded sediment. Water Resources Research 36(11): 3351–3363. sustain relief through the formation of alternate bars, as was ob- DOI. 10.1029/2000WR900161 served downstream of both the Brownsville and Marmot Dam Lisle TE. 1982. Effects of aggradation and degradation in riffle-pool sites. Fourth, the development of low-relief, mobile, mid- morphology in natural gravel channels, northwestern California. Wa- channel bars as a first response to a moderate increase in sedi- ter Resources Research 18(6): 1643–1651. DOI. 10.1029/ ment thickness, as was observed in the Sandy River down- WR018i006p01643 stream of the Marmot Dam site (in the NDS reach), is a Lisle TE. 2008. The evolution of sediment waves influenced by varying transport capacity in heterogeneous rivers. In Gravel-bed Rivers VI: temporary feature on the path towards the formation of alter- From Process Understanding to River Restoration, Habersack H, nate bars, consistent with flume and field studies. Piegay H, Rinaldi M (eds). Elsevier: Amsterdam; 213–230. DOI. 10.1016/S0928-2025(07)11136-6 — Acknowledgements Jon Major of the US Geological Survey Cascades Lisle TE, Church M. 2002. Sediment transport-storage relations for Volcano Observatory generously provided the data from the Marmot degrading, gravel bed channels. Water Resources Research 38(11): Dam site. Tom Lisle, Manny Gabet, and an anonymous reviewer pro- 1219. DOI. 10.1029/2001WR001086 vided helpful reviews. This work was funded in part by the Oregon Lisle TE, Cui Y, Parker G, Pizzuto JE, Dodd AM. 2001. The dominance – Watershed Enhancement Board (OWEB), project number 207 091, of dispersion in the evolution of bed material waves in gravel-bed riv- and by the National Oceanic and Atmospheric Association (NOAA), ers. Earth Surface Processes and Landforms 26: 1409–1420. DOI. National Marine Fisheries Service, award number NA07NMF4630201. 10.1002/esp.300 Lisle TE, Ikeda H, Iseya F. 1991. Formation of stationary alternate bars in a steep channel with mixed-size sediment: a flume experiment. References Earth Surface Processes and Landforms 16(5): 463–469. DOI. Buffington JM, Lisle TE, Woodsmith RD, Hilton S. 2002. Controls on the 10.1002/esp.3290160507 size and occurrence of pools in coarse-grained forest rivers. River Re- Lisle TE, Hilton S. 1992. The volume of fine sediment supply in pools: search and Applications 18(6): 507–531. DOI. 10.1002/rra.693 an index of 594 sediment supply in gravel bed streams. Water Re- Church M. 2006. Bed material transport and the morphology of alluvial sources Bulletin 28(2): 371–383, 595. DOI. 10.1111/j.1752- river channels. Annual Review of Earth and Planetary Sciences 34: 1688.1992.tb04003.x 325–354. DOI. 10.1146/annurev.earth.33.092203.122721 Lisle TE, Pizzuto JE, Ikeda H, Iseya F, Kodama Y. 1997. Evolution of a Cui Y, Parker G, Lisle TE, Gott J, Hansler-Ball MA, Pizzuto JE, sediment wave in an experimental channel. Water Resources Re- Allmendinger NE, Reed JM. 2003. Sediment pulses in mountain riv- search 33: 1971–1981. DOI. 10.1029/97WR01180 ers 1: experiments. Water Resources Research 39(9). DOI. 10.1029/ MacWilliams ML Jr, Wheaton JM, Pasternack GB, Street RJ, Kitanidis 2002WR001803 PK. 2006. Flow convergence routing hypothesis for pool-riffle main- Cui Y, Wilcox A. 2008. Development and application of numerical tenance in alluvial rivers. Water Resources Research 42:1–21. DOI. models of sediment transport associated with dam removal. In 10.1029/2005WR004391 Sedimentation Engineering: Theory, Measurements, Modeling, and Madej MA. 1999. Temporal and spatial variability in thalweg profiles Practice, ASCE Manual 110, Garcia MH (ed.). ASCE: Reston, VA; of a gravel-bed river. Earth Surface Processes and Landforms 24: 995–1020. 1153–1169. DOI. 10.1002/(SICI)1096-9837(199911)24:12<1153:: Downs PW, Cui Y, Wooster JK, Dusterhoff SR, Booth DB, Dietrich WE, AID-ESP41>3.0.CO;2–8 Sklar LS. 2009. Managing reservoir sediment release in dam removal Madej MA. 2001. Development of channel organization and roughness projects: an approach informed by physical and numerical modelling following sediment pulses in single-thread, gravel bed rivers. Water of non-cohesive sediment. International Journal of River Basin Man- Resources Research 37(8): 2259–2272. DOI. 10.1029/ agement 7(4): 433–452. DOI. 10.1080/15715124.2009.9635401 2001WR000229 Dunne T, Leopold LB. 1978. Water in Environmental Planning.WH Madej MA, Sutherland DG, Lisle TE, Pryor B. 2009. Channel responses Freeman: New York. to varying sediment input: a flume experiment modeled after Gaeuman D 2013. High-flow gravel injection for construction de- Redwood Creek, California. Geomorphology 103(4): 507–519. signed in-channel features. River Research and Applications 30(6): DOI. 10.1016/j.geomorph.2008.07.017 685–706. DOI. 10.1002/rra.2662 Major JJ, O’Connor JE, Podolak CJ, Keith MK, Grant GE, Spicer KR, Grant GE, Swanson FJ, Wolman MG. 1990. Pattern and origin of Pittman S, Bragg HM, Wallick JR, Tanner DQ, Rhode A, Wilcock stepped-bed morphology in high-gradient streams, Western PR. 2012. Geomorphic Response of the Sandy River, Oregon, to Re- Cascades, Oregon. Geological Society of America Bulletin moval of Marmot Dam, US Geological Survey Professional Paper 102(3): 340–352. DOI. 10.1130/0016-7606(1990)102<0340: 1792. US Geological Survey: Reston, VA; 64 pp. and data ta- PAOOSB>2.3.CO;2 bles. http://pubs.usgs.gov/pp/1792/

Mossop B, Bradford MJ. 2006. Using thalweg profiling to assess and Shuman JR. 1995. Environmental considerations for assessing dam monitor juvenile salmon (Oncorhynchus spp.) habitat in small removal alternatives for river restoration. Regulated Rivers: streams. Canadian Journal of Fisheries and Aquatic Sciences 63(7): Research and Management 11(3–4): 249–261. DOI. 10.1002/ 1515–1525. DOI. 10.1139/f06-060 rrr.3450110302 Nelson JM. 1990. The initial instability and finite-amplitude stability of Sklar LS, Fadde J, Venditti JG, Nelson P, Wydzga MA, Cui Y, Dietrich alternate bars in straight channels. Earth-Science Reviews 29(1): WE. 2009. Translation and dispersion of sediment pulses in flume ex- 97–115. DOI. 10.1016/0012-8252(0)90030-Y periments simulating gravel augmentation below dams. Water Re- Nicholas AP, Ashworth PJ, Kirkby MJ, Mackin MG, Murray T. 1995. sources Research 45(8). DOI. 10.1029/2008WR007346 Sediment slugs: large-scale fluctuations in fluvial sediment transport Snyder NP, Rubin DM, Alpers CN, Childs JR, Curtis JA, Flint LE, Wright rates and storage volumes. Progress in Physical Geography 19(4): SA. 2004. Estimating accumulation rates and physical properties of 500–519. DOI. 10.1177/030913339501900404 sediment behind a dam: Englebright Lake, Yuba River, northern Pearson AJ, Snyder NP, Collins MJ. 2011. Rates and processes of chan- California. Water Resources Research 40(11). DOI. 10.1029/ nel response to dam removal with a sand-filled impoundment. Water 2004WR003279 Resources Research 47(8): 1–15. DOI. 10.1029/2010WR009733 Sutherland DG, Ball MH, Hilton SJ, Lisle TE. 2002. Evolution of a Pizzuto J. 2002. Effects of dam removal on river form and process. Bio- landslide-induced sediment wave in the Navarro River, California. Science 52(8): 683–691. DOI. 10.1641/0006-3568(2002)052[0683: Geological Society of America Bulletin 114(8): 1036–1048. DOI. EODROR]2.0.CO;2 10.1130/0016-7606(2002)114<1036:EOALIS>2.0.CO;2 Podolak CJP, Wilcock PR. 2013. Experimental study of the response of a Venditti JG, Nelson PA, Minear JT, Wooster J, Dietrich WE. 2012. Alter- gravel streambed to increased sediment supply. Earth Surface Pro- nate bar response to sediment supply termination. Journal of cesses and Landforms 38(14): 1748–1764. DOI. 10.1002/esp.3468 Geophysical Research, Earth Surface 117(F2). DOI. 10.1029/ Pryor BS, Lisle T, Montoya DS, Hilton S. 2011. Transport and storage of 2011JF002254 bed material in a gravel-bed channel during episodes of aggradation Walter C, Tullos DD. 2010. Downstream channel changes after a small and degradation: a field and flume study. Earth Surface Processes and dam removal: using aerial photos and measurement error for context; Landforms 36(15): 2028–2041. DOI. 10.1002/esp.2224 Calapooia River, Oregon. River Research and Applications 26(10): Rodrigues S, Claude N, Juge P, Breheret JG. 2012. An opportunity 1220–1245. DOI. 10.1002/rra.1323 to connect the morphodynamics of alternate bars with their sedi- Welford MR. 1994. A field test of Tubino’s (1991) model of alternate bar mentary products. Earth Surface Processes and Landforms 37(2): formation. Earth Surface Processes and Landforms 19(4): 287–297. 240–248. DOI. 10.1002/esp.2255 DOI. 10.1002/esp.3290190402 Sawaske SR, Freyberg DL. 2012. A comparison of past small dam re- Wohl EE, Cenderelli DA. 2000. Sediment deposition and transport movals in highly sediment-impacted systems in the U.S. Geomor- patterns following a reservoir sediment release. Water Resources phology 151–152:50–58. DOI. 10.1016/j.geomorph.2012.01.013 Research 36(1): 319–333. DOI. 10.1029/1999WR900272 Sear DA. 1996. Sediment transport processes in pool–riffle sequences. Zunka JPP.2011. Dam Removals and Downstream Bar-pool Morphology, Earth Surface Processes and Landforms 21(3): 241–262. DOI. Master’s Thesis. Oregon State University, Corvallis, OR. 10.1002/(SICI)1096-9837(199603)21:3<241::AID-ESP623>3.0.CO;2–1 Sherrod DR, Smith JG. 2000. Geologic map of upper Eocene to Holo- Supporting Information cene volcanic and related rocks of the Cascade Range, Oregon: US Geological Survey Geologic Investigations Series Map I-2569.US Additional supporting information may be found in the online Geological Survey: Reston, VA. version of this article at the publisher's web site.