Predicting the type, location and magnitude of geomorphic responses to dam removal: Role of hydrologic and geomorphic constraints

    Predicting the Type, Location and Magnitude of Geomorphic Responses to Dam Removal: Role of Hydrologic and Geomorphic Constraints John D. Gartner, Francis J. Magilligan, Carl E. Renshaw PII: DOI: Reference:

S0169-555X(15)00106-3 doi: 10.1016/j.geomorph.2015.02.023 GEOMOR 5112

To appear in:

Geomorphology

Received date: Revised date: Accepted date:

1 April 2014 15 January 2015 4 February 2015

Please cite this article as: Gartner, John D., Magilligan, Francis J., Renshaw, Carl E., Predicting the Type, Location and Magnitude of Geomorphic Responses to Dam Removal: Role of Hydrologic and Geomorphic Constraints, Geomorphology (2015), doi: 10.1016/j.geomorph.2015.02.023

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT Predicting the Type, Location and Magnitude of Geomorphic Responses to Dam Removal:

Authors

b

SC

Department of Earth Sciences, Dartmouth College, Hanover, NH 03755 Department of Geography, Dartmouth College, Hanover, NH 03755

NU

a

RI

John D. Gartnera*, Francis J. Magilliganb, Carl E. Renshawa

PT

Role of Hydrologic and Geomorphic Constraints

D

MA

*corresponding author

TE

Keywords

Abstract

AC CE P

Dam removal; Sediment transport gradients; Erosion; Knickpoint; Channel evolution

Using a dam removal on the Ashuelot River in southern New Hampshire, we test how a sudden, spatially non-uniform increase in river slope alters sediment transport dynamics and riparian sediment connectivity. Site conditions were characterized by detailed pre- and post-removal field surveys and high-resolution aerial lidar data, and locations of erosion and deposition were predicted through one-dimensional hydrodynamic modeling. The Homestead Dam was a ~200 year old, 4 m high, 50 m wide crib dam that created a 9.5 km long, relatively narrow reservoir. Following removal, an exhumed resistant bed feature of glaciofluvial boulders located 400 m upstream and ~2.5 m lower than the crest of the dam imposed a new boundary condition in the drained reservoir, acting as a grade control that maintained a backwater effect upstream. During the 15 months following removal, non-uniform erosion in the former reservoir totaled ~60,000

1

ACCEPTED MANUSCRIPT m3 (equivalent to ~9.3 cm when averaged across the reservoir). Net deposition of ~10,700 m3 was measured downstream of the dam, indicating most sediment from the reservoir was carried

PT

more than 8 km downstream beyond the study area. The most pronounced bed erosion occurred

RI

where modeled sediment transport increased in the downstream direction, and deposition occurred both within and downstream of the former reservoir where modeled sediment transport

SC

decreased in the downstream direction. We thus demonstrate that spatial gradients in sediment

NU

transport can be used to predict locations of erosion and deposition on the stream bed. We further observed that bed incision was not a necessary condition for bank erosion in the former reservoir.

MA

In this characteristically narrow and shallow reservoir lacking abundant dam-induced sedimentation, the variable resistance of the bed and banks acted as geomorphic constraints.

TE

D

Overall, the response deviated from the common conceptual model of knickpoint erosion and channel widening due to dam removal. With thousands of dams likely to be considered for

AC CE P

removal or repair in the coming decades, this study helps to advance predictions of the geomorphic response to dam removal and contributes to a broader understanding of the variability in both style and timing of fluvial responses to disturbances. 1. Introduction

Following an era of dam building that peaked in the 1960s, the U.S. is now entering an era of increasing dam removal or repair as the majority of the ~80,000 large and medium sized dams (Graf, 1999) are reaching the end of their ~50 year design lifespan (Shuman, 1995; Maclin and Sicchio, 1999). Many of these dams are no longer needed for the original uses, and the benefits of undammed rivers for aquatic resources and riparian habitat are becoming increasingly understood and valued (Nilsson and Berggen, 2000; Bednarek, 2001; Bushaw-Newton et al., 2002; Poff and Hart, 2002; Stanley and Doyle, 2003; Petts and Gurnell, 2005). For example, dam

2

ACCEPTED MANUSCRIPT removals can restore connectivity of sediment in fluvial systems, both in the longitudinal direction downstream and the lateral direction between channel and floodplains (Kondolf et al.,

PT

2006). To determine the costs and benefits of repairing or removing a dam, specific predictions

RI

of channel response to dam removal and changes in connectivity are important to dam owners and agencies that oversee dam safety, waterway engineering, and riparian habitats. However, to

SC

date, predicting the locations of erosion and sedimentation has been elusive in these transient

NU

systems (Draut and Ritchie, 2012), especially when local and watershed scale constraints play a

MA

role (Major et al., 2012).

Beyond the management imperatives, dam removals merit investigation to improve our basic

D

understanding of river processes. Dam removals present compelling natural-scale experiments on

TE

rivers—the dramatic increase in water surface slope due to lowering of the local base level creates a disturbance at a known place where the influence of slope on sediment transport and

AC CE P

channel evolution can be tested. It is well established that water surface slope is a fundamental parameter in sediment transport and that a threshold for particle motion must be exceeded for a stream channel to erode (e.g. Shields, 1936; Meyer-Peter and Müller, 1948; Ferguson, 2012). Yet the interplay of temporal and spatial changes in slope and sediment transport and deposition is poorly understood in transient systems adjusting to disturbance. In particular, few studies have highlighted the importance of downstream spatial gradients in sediment transport in controlling the spatial pattern of erosion and deposition in recently disturbed fluvial systems (Paola and Voller, 2005; Bizzi and Lerner, 2013). In the current predominant conceptual model, channel response to dam removal begins with the upstream migration of a knickpoint (e.g. Womack and Schumm, 1977), resulting in channel incision through reservoir sediments followed by channel widening (Simon, 1989; Doyle et al.,

3

ACCEPTED MANUSCRIPT 2002; Pizzuto, 2002; Pizzuto and O’Neal, 2009; Major et al., 2012). The steepness and height of the knickpoint front attenuates as it migrates upstream in the absence of resistant bedrock or

PT

other constraining features. Thus the greatest channel changes are expected generally to occur at

RI

dam proximal locations where the steep water slopes result in the greatest increases in shear stress (Doyle et al., 2002; Williams, 1977). This knickpoint migration and channel evolution

SC

conceptual model is most applicable to transitional channels forming in approximately

NU

homogenous material, a situation that can arise when relatively deep or wide reservoirs with

MA

pronounced sedimentation are drained.

However, dams and their impoundments occupy a variety of geomorphic settings, and this

D

variability in settings cautions against expecting a similar response at every dam removal (Graf,

TE

1999; Poff and Hart, 2002). Two recent dam removal studies highlight that reservoir response to dam removal is a) related to the width of the reservoir relative to the upstream channel (Sawaske

AC CE P

and Freyberg, 2012) and b) influenced by local boundary conditions, such as bedrock outcrops (Major et al., 2012; Pearson et al., 2011). Here, we contend that narrow, shallow reservoirs tend to have thin deposits from sedimentation due to reduced accommodation space compared to deep, wide reservoirs. As these narrow reservoirs are drained, the variable integrity of the emerging bed and banks is likely to influence the spatial distribution of resistance to erosion and thus the spatial distribution of changes in slope and erosion potential within the drained reservoir. In such conditions, the geomorphic response to the disturbance of a dam removal may be highly controlled by boundary conditions, namely the spatial variability in resistance to erosion and erosive forces. The purpose of this research is to test conceptual models of geomorphic response to dam removal in low gradient settings where contemporary (e.g., upstream flow regulation,

4

ACCEPTED MANUSCRIPT urbanization) and pre-historical (locally thin alluvial cover over resistant bedrock) boundary conditions exits. We use the findings from the recent removal of the Homestead Dam and

PT

draining of its relatively narrow reservoir on the Ashuelot River in southeastern New Hampshire

RI

to test the prediction that downstream changes in erosive factors, such as bed resistance and spatial gradients in sediment transport, dictate the channel bed and bank response. Predicted and

SC

observed responses are documented through a combination of field surveys, innovative aerial

NU

lidar analysis, and one-dimensional hydrodynamic modeling. We characterize conditions before dam removal, use these conditions to predict locations of ensuing erosion and deposition, and

MA

monitor conditions for 15 months post removal. We document a channel response that deviates from the knickpoint migration and channel evolution conceptual model in significant ways

slope in the drained reservoir.

TE

D

because boundary conditions in the narrow, shallow reservoir impose variability on water surface

AC CE P

We investigate the locations of erosion and deposition as a function of downstream changes in sediment transport and thresholds for sediment transport, which are modeled and predicted using information from the pre-removal surveys. Our analysis provides a framework for understanding the wide variety of river responses to dam removals and, more broadly, in transient systems.

2. Spatial patterns in sediment transport, erosion and deposition Building on the Exner equation, which states that changes in stream bed elevation are a function of the negative divergence of sediment flux (Exner, 1920, 1925; Paola and Voller, 2005), it follows that, in the absence of significant tributary inputs, erosion should occur in reaches where sediment flux increases in the longitudinal direction. Conversely, aggradation should occur

5

ACCEPTED MANUSCRIPT where sediment discharge decreases in the downstream direction. A necessary condition for either erosion or deposition is that flows are sufficient to mobilize sediment, which occurs when

PT

the Shields parameter, θ, exceeds a critical threshold for bed material entrainment, θcrit, either at

RI

a location of erosion or upstream of a location of deposition. This concept can be expressed as

D

MA

NU

SC

follows:

TE

where Qs is volumetric sediment discharge across the entire river width (in units volume per time

AC CE P

in this study, but also could be expressed as mass per time), and x is distance downstream. We use the dam removal as a natural experiment to predict two potential outcomes: (a) reaches of increasing downstream sediment transport are prone to erosion, and (b) reaches of decreasing downstream sediment transport are prone to deposition. A combination of aerial lidar, ground surveys, and HEC-RAS modeling are used to determine Qs in a design storm and compare zones of predicted and observed erosion and deposition.

3. Site description The Homestead Dam was an approximately 200 year old, 4 m high, 50 m wide, failing timber and rock crib structure located on the Ashuelot River in West Swanzey, New Hampshire that provided water storage for a now-defunct woolen mill (Figs. 1 and 2). It was a run-of-river dam

6

ACCEPTED MANUSCRIPT lacking flow regulation. The ~9.5 km long reservoir behind the dam was relatively narrow, averaging ~40 m wide, and apparently contained within the banks of the pre-dam river, although

PT

pre-dam surveys are unavailable. The ratio of the average width of the reservoir to the width of

RI

the upstream channel is ~1.5.

SC

The deteriorating, unused mill dam was breached on August 23, 2010 (Fig. 2). Over the next two months the entire structure was removed. Three rock vanes were constructed at the grade of the

NU

river over a 60 m reach centered at the former dam to mitigate potential bed erosion at the site of

MA

the former dam to protect an immediately upstream historical covered bridge. These rock vanes were >2 m lower than the crest of the ―boulder reach,‖ described below, and thus likely had little

TE

AC CE P



D

effect on stabilizing the boulder reach.

Figure 1. Ashuelot River, showing the study reach and related dams.

7

ACCEPTED MANUSCRIPT

AC CE P

TE

D

MA

NU

SC

RI

PT



Figure 2. Locations of cross sections, Homestead Dam, reservoir reach, and upstream control reach. Base image is lidar-derived topography. Inset shows breaching of the Homestead Dam. The 820 km2 watershed area upstream of the former dam contains two flood control dams, the Surry Mountain Dam and Otter Brook Dam, built in 1941 and 1958, respectively (Fig. 1). The Surry Mountain Dam, located 30 km upstream of the Homestead Dam, was completed in 1941

8

ACCEPTED MANUSCRIPT with a reservoir capacity of 0.039 km3. Otter Brook Dam is located 18 km upstream on Otter Brook and was completed in 1958 with a reservoir capacity of 0.022 km3. Downstream of the

PT

former Homestead Dam, the Ashuelot flows 25 km to the Algonquin Dam, and then an

RI

additional 7 km to the Connecticut River. Watershed land cover is ~ 84 % rural forest, ~10 % farmland, and ~6 % urban/suburban, mostly in the City of Keene, population 22,000, situated 12

SC

km upstream of the former dam. .

NU

The Ashuelot River flows through mixed alluvial and bedrock-boulder reaches of a glaciated

MA

landscape (Goldthwait, 1951). Sand and gravel from Pleistocene glaciolacustrine deposits dominate the lowlands of the roughly 1.2 km wide valley and constitute most of the river bed.

D

Occasional boulder- and cobble-dominated reaches with riffles and rapids exist up- and

TE

downstream of the former dam. Compact basal till is exposed at the bed of a few deep pools.

AC CE P

An erosion-resistant, boulder-dominated deposit, likely of glaciofluvial origin, extends for 0.4 km upstream of the dam (Fig. 3b). For brevity, we call this the ―boulder reach.‖ When the dam was intact, the crest of this deposit was ~2 m lower than the crest of the dam. It exhibited little to no sedimentation of the sandy material observed throughout the rest of the reservoir. After the dam was removed, this feature was an immobile grade control, forming a riffle and creating a backwater (Fig. 3c). Although this feature is specific to this site, many dam removal projects have encountered similar geologic features, such as bedrock outcrops, in the vicinity of the removed dam (Pearson et al., 2011; Major et al., 2012). Indeed the choice of location for the Homestead and other dams may be related to these pre-existing, natural grade controls.

9

TE

D

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

AC CE P

Figure 3 a) Looking upstream at the reservoir prior to dam removal from 100 to 400 m upstream of dam. This is the location of the Boulder Reach. Flow is 15 m3s-1. b) Similar view from a slightly different angle three weeks after the dam removal showing the size and extent of boulders that had been inundated in the Boulder Reach. Flow is 0.7 m3s-1. c) Similar view of post-removal conditions with flow of 13 m3s-1, for comparison with panel (b) to show that reservoir was relatively narrow. River banks had dense woody and herbaceous vegetation, especially in the former reservoir. Soils are mapped as fine sandy loam and loamy fine sand throughout the study area (Soil Survey Staff, NRCS, 2013), although visual observations indicate that some banks have higher sand content and lower cohesion than other banks.

10

ACCEPTED MANUSCRIPT We divide the study areas into 5 segments (Fig. 4), listed here from up- to downstream (negative distance correspond to locations upstream of the dam): a) the control reach upstream of the

PT

reservoir, from -12.5 to -9.5 km, b) the sand-dominated portions of the reservoir, from -9.5 to -

RI

0.5 km, c) the boulder reach of the reservoir, from -0.4 to 0 km, d) the dam proximal riffle, from 0 to 0.4 km and e) the sand-dominated portions downstream, from 0.4 to 9 km. The control reach

SC

is situated upstream of the backwater of the former reservoir. Two relatively steep, cobble- and

NU

boulder-dominated riffles with a slope of ~0.01 extended for ~150 and 30 m centered at river stations 5.2 and 6.8 km, respectively. These riffles are not considered in detail in the figures and

MA

analysis because no cross section measurements were made there, but they are indicative of the

D

mixed alluvial and bedrock-boulder reaches of this glaciated landscape.

AC CE P

TE



Figure 4. Long profile of river bed and water elevations at time of lidar flights in July 2010 and July 2011. Segments of the study areas are differentiated.

11

ACCEPTED MANUSCRIPT 4. Methods 4.1 Aerial lidar: Differencing of lidar-derived digital elevation models (DEMs) was used to

PT

quantify topographic change in the upper bank and near-river floodplain, which we define

RI

functionally here as the region from 1 to 15 m upslope of the pre-removal, summer 2010 water edge. DEM differencing is performed by subtracting the pre-from the post-removal ground

NU

do not characterize below water surfaces accurately.

SC

surface. We limit this analysis to the upper bank and near floodplain areas because our lidar data

MA

A 22 river km swath centered on the dam was measured with near-infrared, airborne lidar on July 19, 2010 and July 20, 2011, just prior to, and one year following the dam removal. The

D

native pulse density was 8 pulses/m2. Flows were 3.6 and 3.0 m3 s-1 on these dates, respectively,

TE

allowing nearly direct comparisons between these flight dates. Data were collected and classified by the National Center for Airborne Laser Mapping (NCALM), and point clouds were processed

AC CE P

to 1 m gridded bare-ground DEMs. The DEMs were registered in the vertical direction based on the registration values between years of the ground-classified point clouds in CloudCompare, which uses an iterative closest point algorithm (Besl and McKay, 1992). The vertical offset was 0.02618 m upstream of the dam and 0.04709 m downstream of the dam. DEM differencing is highly sensitive to the vertical registration, since a signal of erosion or deposition could be an artifact of accumulation of error in poorly registered DEMs. These values used for vertical registration were corroborated by analysis of paved surfaces throughout the study area. DEM differencing on a cell by cell basis and zonal statistics were performed in ArcGIS. Lidar-derived topography was imprecise in some patches, because data gaps, misclassified points, and unreliable returns on water can create noise on the DEMs at locations under dense vegetation and at steep, curving stream banks. For example, a return from a tree branch may be

12

ACCEPTED MANUSCRIPT classified as a ground point in one year and not the other, creating a false hill in one year but not the other on the bare-ground DEMs. This can create a false signal of erosion or deposition in

PT

DEM differencing. Likewise, if the constellation of ground-classified points over rough terrain differs from one year to the next, slight differences in the modeled terrain will exist even in the

RI

absence of true topographic change. The patches of false signals (―noise‖) were on the order of

AC CE P

TE

D

MA

NU



SC

20 m2 or smaller in this lidar dataset (Fig. 5).

Figure 5. Results of differencing lidar-derived DEMs on upper river banks and near-channel floodplain from 1 to 15 m from water edge at 3km upstream of the dam, a typical location of the reservoir. Erosion (positive values) and deposition (negative values) from before to after removal (July 2010 to July 2011) embed true signal and noise. Most areas show little change, but arrows point to characteristic locations of a) falsely represented erosion and b) verified erosion

13

ACCEPTED MANUSCRIPT based on field surveys and observations. Hillshade shows topography from 2010 lidar data, including the unrealistically rough lidar-derived water surface topography.

PT

To circumvent the difficulty in distinguishing the true signal from the noise in the lidar-derived

RI

topographic change, we a) computed cumulative erosion and deposition from DEM differencing

SC

as a function of distance downstream, and b) limited the lidar analysis to above water regions from 1 to 15 m from the water edge in 2010 before the reservoir was drained. We contend that

NU

the areas of false deposition counterbalance the areas of false erosion when aggregating net

MA

change in topography over areas that are much larger than the patch size of the false signals. The general lack of net change in the control reach supports this idea of a counterbalance of false

D

signals over large areas. By examining cumulative erosion as a function of distance downstream,

TE

locations of abundant erosion or deposition can then be compared with modeled sediment

AC CE P

transport and geomorphic constraints based on distance downstream. 4.2 Field surveys: To characterize conditions below water on the bed and lower banks where the lidar data are not accurate, we established 30 cross sections over the 22 river km study reach (Fig. 2). Cross sections were surveyed ~1-2 months prior to the dam removal, ~1-2 months after removal, ~12 months after removal, and ~15 months after removal (Fig. 6). Notable flow events occurred between but not during the latter three time periods of surveys.

14

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

MA

Figure 6. Flow record at former Homestead Dam (solid line) and 2-year recurrence interval flow (dashed line). Arrows indicate times of dam breach, lidar flights, and field surveys.

D

Channel dimensions at cross sections were measured in the field using a real-time kinetic global

TE

positioning system (RTK GPS) when possible, otherwise using a robotic total station, auto-level,

AC CE P

or lead line in dense vegetation or deep water. Channel width was measured at the 2011 high water mark elevation, and cross section area was measured below this elevation. Volumes of erosion and deposition on the bed and lower banks were computed by integrating changes in cross sectional area over reaches assigned to the nearest cross section. Bed sediment grain size was determined at each cross section one month pre-removal and 12 months post-removal by Wolman pebble counts of 60 to 100 particles when median grain size was > 3 mm, or by sieving when median grain size was < 3 mm. Since grain size did not change significantly pre- and post-removal at each cross section (paired t-test, two-tail, α = 0.05, p = 0.48), we use the same characteristic grain size for 2010 and 2011 to avoid the inherent uncertainties in grain size measurements between surveys (Church et al., 1987; Wohl et al., 1996). Also, grain size was averaged within distinct segments because the river was

15

ACCEPTED MANUSCRIPT characterized by homogenous sandy bed material interspersed by coarse-grained riffles. Specifically, D84 used for θ calculations was 2.1 mm, 191.7 mm, 130.4 mm, and 1.1 mm for the

PT

sandy reservoir, boulder reach, dam proximal riffle, and downstream sandy reach, respectively.

RI

Grain size was not averaged across the upstream control segment because bed material ranged from fine sand to cobbles at those 3 cross sections. For additional insight into sediment

SC

thickness, stratigraphy and material properties, we surveyed with ground penetrating radar

NU

(GPR) using a 200 MHz antenna along the centerline of the reservoir for 4 km upstream of the

MA

dam.

4.3 Sediment transport modeling: A primary intent of the modeling was to use the geometry

D

measured prior to the dam removal to predict the transient and spatially variable flow dynamics

TE

and sediment transport post removal. To make this analysis easily applied to other sites, we used input variables that are commonly measured at dam removal sites, such as channel dimensions

AC CE P

and grain size, the widely-used and free HEC-RAS one dimensional hydraulic modeling software to simulate pre-and post-removal hydraulics, and the widely-used Meyer-Peter and Müller (1948) sediment transport law to calculate volumetric sediment flux. We did not seek to simulate sediment transport or routing over the entire study period explicitly. Rather we focused on magnitudes and gradients in sediment transport in a model storm, and used this as an index of transport over the study area given the conditions at this site. Changes in Shields parameter, θ, as a function of distance downstream and of time (pre- and post-dam removal), were computed using HEC-RAS. The Shields parameter, θ, was determined from the HEC-RAS output as

(Eqn. 1)

16

ACCEPTED MANUSCRIPT where ρ is density of water (100 kg m3), ρs is density of sediment (assumed 2.65 g cm-3), g is gravity (9.8 m s-2), h is depth of water, s is energy gradient slope, and D is the characteristic

PT

grain size of the bed surface material (Knighton, 1998). We used D84, the 84-percentile grain

RI

size, rather than D50, the 50-percentile grain size. Given the often bimodal nature of the bed material, D84 seemed to better characterize bed mobility, or lack thereof, in sediment transport

SC

computations. In the boulder reach and other reaches downstream of the reservoir, θ was also

NU

computed using the D84 of the sandy portion of the former reservoir to indicate the mobility of

MA

material delivered from upstream.

Many conventional sediment transport equations are a function of the Shields parameter

(Eqn. 2)

AC CE P

TE

D

in the form

where Qs is modeled volumetric sediment flux, and w is channel width. The values of the exponent, a, and coefficient, b, differ among transport formulae. The value of the exponent, a, varies from 3/2 for strictly bedload transport up to a range of 5/2 to 8 for mixed and suspended load transport (Graf, 1984; Julien, 2010; Dade et. al, 2011). We chose a = 3/2 and b = 8, following the widely used Meyer-Peter and Müller (1948) equation, and used D84 as the characteristic grain size. θcrit was set at 0.03, on the low end of a range of published values (Buffington and Montgomery, 1997). The magnitudes of modeled sediment flux change depending on the values for a, b and θcrit. However, we stress that the spatial patterns of increasing versus decreasing Qs, which are a focal point of the paper, are largely independent of the chosen values for a, b and θcrit.

17

ACCEPTED MANUSCRIPT Qs was computed under three modeled conditions using a 2 year recurrence interval design flow: 1) pre-removal, 2) immediate post-removal, and 3) one year post removal. In the first two

PT

conditions the channel geometries were based on surveys in summer 2010 prior to the dam

RI

removal; they were identical with the exception that the dam is removed from the model in the immediate post-removal condition. In the third condition, the channel geometry was derived

SC

from cross section measurements in summer 2011, a year after the dam removal. Planform

NU

geometry was derived from the lidar data and aerial photographs in HEC-GeoRAS, a GIS interface for HEC-RAS. The HEC-RAS model was calibrated by comparing the water surface

MA

profiles measured by lidar with those generated by the model at the same discharges. Modeled

D

water surface elevation was within 5 cm of the lidar water surface elevation at each cross section.

TE

In addition to computing Qs using the D84 of the bed material at each cross section, Qs was also computed using the D84 of the sandy reservoir material (2.1 mm) for all cross sections

AC CE P

downstream of the sandy portion of reservoir. The aim was to examine how this material would be transported through the study area, especially in the region from the sandy reservoir through the boulder reach.

4.4 Reservoir sedimentation: To determine the fraction of sediment stored by the dam exported after removal, we compared the measured export with the estimates of dam-induced reservoir sedimentation. We bracketed the possible volume of sedimentation with low and high values because direct measurements of the sedimentation in the reservoir were not possible at this site, as is the case at many older dams that have no pre-dam surveys. Here, the dam pre-dates the earliest available topographic maps in 1898. Furthermore, the homogenous, sandy bed material exhibited no distinct layer indicative of a pre-dam bed surface that could be detected by

18

ACCEPTED MANUSCRIPT probing with a metal rod or by ground penetrating radar that imaged the bed stratigraphy to a depth of ~3 m.

PT

The low and high values of possible reservoir sedimentation differed primarily due to the

RI

estimated maximum thickness of sedimentation, tmax. In both cases we assumed the thickest

SC

deposits are located 0.5 km upstream of the dam, just upstream of the ―boulder reach,‖ and the deposits tapered linearly to a thickness of 0 m at the head of the reservoir, 9.5 km upstream of

NU

the dam. We also assumed the deposits were channel-wide throughout the reservoir, which has

MA

an average width of 40 m. We did not observe reach-wide sedimentation in the boulder reach from 0 to 0.4 km upstream of the dam in pebble counts, snorkel surveys, or GPR surveys prior to

D

dam removal or visual observation as the reservoir drained. Thus this reach did not contribute to

TE

the estimates of reservoir sedimentation.

AC CE P

One approach to estimate tmax was to assume that the thalweg eroded to the elevation of the predam thalweg in at least one location of the reservoir over the course of the study. The greatest measured incision of the thalweg was 0.54 m at the cross section located 1.8 km upstream of the reservoir. Extrapolation downstream to 0.5 km upstream of the reservoir yields a tmax of 0.64 m. Another approach was to assume tmax is equivalent to the height that the water surface was raised by the dam which yields an estimate of the maximum possible amount of sedimentation. In this case, tmax was 2.5 m, which is the height that the dam elevated the reservoir surface above the crest of the boulder reach. Based on pre-removal surveys, the boulder reach was likely a grade control for the water surface before the dam was constructed. The GPR profiles of streambed stratigraphy also showed a faint, relatively level layer overlain by onlap and downlap features of sandy material approximately 2.5 m below the surface of the pre-removal bed approximately 0.6

19

ACCEPTED MANUSCRIPT km upstream of the dam and 0.1 km upstream of the boulder reach (Fig. 7). This layer may have been the pre-dam bed surface with 2.5 m of sandy material deposited on top since the dam was

PT

built.

AC CE P

TE

D

MA

NU

SC

RI



Figure 7. Ground penetrating radar profile from dam to 1.2 km upstream prior to dam removal. Symbol (A) denotes water surface, (B) surface of sediment horizon, the channel bed, (C) interpreted multiple of sediment horizon, thus this is not a real horizon, (D) interpreted sub bottom horizons, perhaps indicative of pre-dam channel bed in locations upstream of boulder reach, approximately 2.5 m below surface of sediment horizon at -0.6 km. 4.5 Flow: River discharge and flow recurrence intervals were computed using the USGS gaging station 01161000, located in Hinsdale 30 km downstream of the dam with a record from 1907 to

20

ACCEPTED MANUSCRIPT present. Flows were prorated by area to obtain discharge and flow frequency at cross sections throughout the study area. During the study period, the flows derived from the Hinsdale gage for

PT

the former dam site were within 10 % of flows measured at the former dam site at USGS gaging

RI

station 01160350. Since this gage has a short record, 1994 to present, flow recurrence intervals were computed from the 1958-2011 post-flow-regulation record at the Hinsdale gage, based on

MA

NU

SC

Log-Pearson Type III analysis.

5. Results

TE

D

5.1 Pre-removal conditions: Across the upper 9.1 km of the 9.5 km long reservoir, D50 and D84 measured at cross sections averaged 1.0 and 2.1 mm, respectively. The continuous data of the

AC CE P

GPR survey and visual observations at low flows corroborated that the sand-dominated bed was continuous in this portion of the reservoir. Estimates of impounded sediment in the reservoir range from 57,600 to 225,000 m3, which equates to ~0.17 to 0.66 m average depth of sedimentation.

The boulder reach contrasted dramatically with the other, sandy portions of the former reservoir. D50 and D84 averaged 12.2 and 197 mm, respectively, at the three cross sections here. Several larger boulders (≥ 3 m in diameter) were scattered though the reach. Bed slope averaged 0.003. A boulder- and cobble-dominated riffle extended 0.4 km downstream of the dam with a slope of 0.004. This indicates that the dam (at 0.0 km) was placed amid this relatively steep, coarsegrained bed feature from -0.4 to 0.4 km.

21

ACCEPTED MANUSCRIPT Downstream of this dam-proximal riffle, the channel bed was predominately sandy for ~8 km to the end of the study area and beyond. Bed slope in the sandy reaches averaged 3 x 10-4, and D50

PT

and D84 averaged 0.8 and 1.1 mm, respectively.

RI

5.2 Flows and timing of response: Flows were low and insufficient to transport sediment for ~2

SC

months post removal, but higher flows occurred during the periods from 2-12 and 12-15 months post removal (Fig. 6). Flows capable of transporting sediment were approximately equal during

NU

the latter two time periods, with each period having 3 events greater than 50 m3 s-1. Peak flows of

MA

102.1 m3 s-1 on March 8, 2011 and 96.9 m3 s-1 on September 7, 2011at the former dam were approximately equal in magnitude and duration. These peak flows slightly exceed 95.8 m3 s-1,

D

the 2 year recurrence interval flow at the former dam. Flood-control operations at the upstream

TE

dams attenuated flood flows effectively, as evidenced by the peak flow of 59.4 m3 s-1 during Tropical Storm Irene on August 28-29, 2011. This storm produced the flow of record for many

AC CE P

unregulated rivers in the region.

Despite the approximately similar flows, the magnitude of erosion in these two latter time periods differed dramatically. Of the ~40,000 m3 of material eroded from the bed and lower banks of the reservoir over the 15 month study period, ~91 % was exported in period from 2-12 months post removal (Supplemental Data Table). 5.3 Predictions and observations of bed and lower bank conditions Net erosion from the bed and lower banks measured by cross section surveys totaled ~40,000 m3 in the former reservoir, equivalent to ~10.5 cm average depth, and net deposition over the ~8 km downstream of the former dam on the bed and lower banks totaled ~8,600 m3, equivalent to ~2.5 cm average depth.

22

AC CE P

TE

D

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

23

AC CE P

TE

D

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

24

ACCEPTED MANUSCRIPT Figure 8. a) For time period immediately after dam removal, modeled sediment discharge of 2year recurrence interval storm using bed sediment grain size (solid line: Qs, Bed Sed.) and sandy

PT

reservoir grain size (dashed line: Qs, Res. Sed.). Arrows indicate zones of increasing Qs with

RI

respect to distance downstream. Shading corresponds to the arrows and indicates predicted zones

SC

of erosion.

b) Surveyed erosion and deposition of bed material at cross section intervals from 0 to 12 months

NU

after dam removal, with positive values for erosion and negative values for deposition.

MA

c) Cumulative bed erosion, where a positive slope of the line indicates a zone of net erosion and a negative slope indicates a zone of net deposition.

TE

D

d) Lidar measured erosion of bank material from 1 to 15 m from water edge at 500 m intervals

AC CE P

from time period 0 to 12 months after dam removal. d) Cumulative bank erosion measured by lidar. As above, positive slope of the line indicates erosion and negative slope indicates deposition. f) Modeled sediment discharge (Qs) as in panel (a), except for time period starting at 12 months after dam removal.

g) Surveyed erosion and deposition of bed material from 12 to 15 months after dam removal, expressed as volume change for each reach between cross sections, with positive values for erosion and negative values for deposition. h) Cumulative bed erosion, where a positive slope of the line indicates a zone of net erosion and a negative slope indicates a zone of net deposition for time period 12 to 15 months post removal,

25

ACCEPTED MANUSCRIPT and solid line shows from 0 to 15 months after dam removal. Scale is same as in panel (c) to show decreases in magnitude of channel adjustment between time periods.

PT

Closer inspection reveals a complex pattern of eroding, aggrading, and stable reaches that can be

RI

delineated along similar lines as the modeled spatial patterns in Qs. Figures 8a and 8f

SC

demonstrate downstream gradients in modeled sediment transport. Results indicate that d(Qs)/dx > 0 from -9.5 to -5 km and -3.2 to -1.3 km and d(Qs)/dx < 0 from -5.2 to -3.2 km. Figures 8b and

NU

8g show the amount of erosion or deposition for each reach assigned to each cross section at two

MA

different time intervals. Figure 8c and 8h show the cumulative erosion as a function of distance downstream.

D

Spatial patterns of erosion and deposition on the bed and lower banks of the sandy reaches in the

TE

former reservoir and downstream support the prediction that zones of increasing downstream Qs

AC CE P

are prone to erosion and zones of decreasing Qs are prone to deposition (Table 1). For the time interval 0 to 12 months after dam removal, erosion was measured in 8 of 11 locations that were predicted to have erosion (θ > θcrit and d(Qs)/dx > 0). At the remaining 3 of 11 locations, deposition was observed (1.5, 2.0, and 8 km). However these were the locations within ~1 km downstream of the greatest declines in Qs. Averaging changes in Qs over slightly longer distances would show a strong decline in Qs in the longitudinal direction at the depositional locations. Deposition was observed in 10 of 11 location where d(Qs)/dx < 0 in the time period 012 months post removal. In the time period 12 to 15 months after removal, gradients in Qs were again a strong predictor of erosion and deposition within the sandy reaches of former reservoir, but less so in sandy reaches downstream. In the former reservoir, erosion was observed in 6 of 7 locations from 12 to 15

26

ACCEPTED MANUSCRIPT months post removal where d(Qs)/dx > 0x in the flow modeling of conditions at 12 months. Deposition was observed in 3 of 4 locations predicted to have deposition (d(Qs)/dx < 0 ).

PT

Downstream, notable deposition was observed in three locations predicted to have erosion

RI

(d(Qs)/dx > 0), but two locations (0.7 and 1.5 km) were downstream of the greatest decline in Qs, within 1 km of the dam proximal riffle. Also downstream, three locations (2.0, 2.9, and 6.1 km)

SC

expressed erosion that exceeded channel changes measured in the control reach, but deposition

NU

was predicted (d(Qs)/dx < 0 ). In these locations, the amount of erosion was roughly equivalent to

MA

the amount of deposition in the first year following dam removal. Conditions in the boulder reach also support the predictions of erosion and deposition, but in a

D

more nuanced manner. Erosion of the bed was minimal in the boulder reach, although there was

TE

some erosion of sandy material on the lower banks. Calculated Shields parameter was ~0.003 for the coarse bed material, well below the lower values for thresholds for motion of ~0.03 (Mueller

AC CE P

et al., 2007). Here the coarse bed material was not predicted to erode because θ << θcrit for D84 = 191.7 mm, but sandy material did meet the conditions for erosion (θ > θcrit and d(Qs)/dx > 0 for D84 =2.1 mm). Likewise sandy material delivered from upstream was not prone to deposition because Qs was increasing greatly in the transition from the former reservoir into the boulder reach and dam proximal riffle. These data show that predictions of channel change are not simply a function of Shields parameter or of Qs exceeding a threshold value or increasing over time; instead channel change correlates reasonably well with the spatial gradients in modeled sediment transport. For example, the channel from -4 to – 3 km in the former reservoir was depositional, even though θ values were 0.18 to 0.15, respectively, well above θcrit and higher than the pre-removal values of 0.14 to 0.13. Furthermore, the scalar Qs values in this depositional reach are equal to or greater than that

27

ACCEPTED MANUSCRIPT of some erosional reaches in other locations of the former reservoir. However, this was a region where Qs was decreasing over space (d(Qs)/dx < 0x ).

PT

5.4 Upper bank and floodplain conditions Changes in the upper banks and near-channel

RI

floodplains (1 to 15 m from the summer 2010 water edge) were quantified by repeat lidar

SC

measurements. Results show a general trend of bank erosion in the reservoir and minimal net bank deposition on the downstream bank. However, gradients in Qs were not as predictive for

NU

upper bank erosion and deposition as for the channel bed adjustments described above (Fig. 8d).

MA

In total there was ~20,000 m3 of erosion from the upper banks and floodplains in the reservoir. This averages to ~7.5 cm of erosion (Fig. 8e).

D

With the exception of a few noticeable bank failures, in most locations these changes were subtle

TE

and not obvious in cross section measurements or visual observations. The control reach from -

AC CE P

12.5 to -9.7 km exhibited little to no bank erosion. Within the former reservoir, there was a broad pattern of bank erosion from -10.2 to -7.8 km and from -6.4 to -1.2 km that contrasted with minimal to no bank erosion from -7.8 to -6.8 km and from -1.2 to -0.6 km. Downstream of the removed dam, there was slight deposition on the upper banks and floodplain from 0 to 4.5 km, and little to no net deposition from 4.5 to 8.9 km. There was not a direct correlation between bed erosion measured by repeat cross section surveys and the bank erosion measured by repeat lidar. Some locations had bed and bank erosion (e.g. -10.2 to -7.8 km), some showed one but not the other (e.g. -7.8 to -6.8 km and -5.1 to -3.1 km), and some exhibited neither (e.g. -1.2 to -0.6 km). This indicates that bed erosion is not a prerequisite for bank erosion, which is inconsistent with the conceptual model of incision inducing widening. Furthermore, it indicates that other factors

28

ACCEPTED MANUSCRIPT influence bank erosion beyond the physical requirements for bed erosion proposed in this paper (θ > θcrit and d(Qs)/dx > 0).

PT

It is unclear from our results if bank resistance to erosion varied systematically in correlation

RI

with the spatial patterns of bank erosion in the drained reservoir. The spatial patterns of bank

SC

erosion did not coincide with variations in bank height, sinuosity, mapped soil type, vegetation density or tree assemblages. All the mapped soil types were fine sandy loam or loamy fine sand,

NU

but the soil properties may have varied in a subtle way not captured in the soil survey. Our

MA

observations of bank material indicated that some banks had higher sand content and were less cohesive than other locations, and these patterns did not align with different mapped soil units.

D

5.5 Net effects Net erosion of the bed, banks, and near-channel floodplains measured by surveys

TE

and lidar upstream of the former dam totaled ~60,000 m3, which is equivalent to an average of

AC CE P

~9.3 cm of vertical erosion over this area. Net deposition downstream of the former dam was ~10,700 m3, which equates to an average of ~1.9 cm of vertical deposition. More sediment was exported from the reservoir than was deposited within 8 km downstream of the dam. Most sediment was therefore transported beyond the study site. Moreover, the net measured effects were small, considering the removal of the ~4 m high dam compared with an average vertical change of < 10 cm in the reservoir and < 2 cm downstream.

6. Discussion The results of this study emphasize the importance of spatial gradients in sediment transport in transitional settings, where the river is adjusting following disturbance. Specifically, if the modeled sediment transport is increasing in the downstream direction, then the river has the

29

ACCEPTED MANUSCRIPT ability to transport the sediment delivered from upstream and remove additional material from the bed and banks. In contrast, conditions are preferable for deposition if sediment transport is

PT

decreasing from one location to the next downstream location, because the river cannot transmit

RI

the sediment delivered from upstream. This importance of gradients in sediment transport has roots in the Exner equation (Exner 1920, 1925), but is generally underappreciated in the current

SC

literature. In the limited recent work where gradients have been recognized, Paola and Voller

NU

(2005) derived a generalized Exner equation without supporting observations, and Bizzi and Lerner (2013) use spatial gradients in stream power as part of a classification tree for confined

MA

channel types.

D

Of course, the threshold for sediment transport must be exceeded to mobilize bed and bank

TE

material. The channel response in the boulder reach illustrates how localized thresholds affect erosion in addition to the sediment transport relative to upstream locations. The thresholds for

AC CE P

motion of the bed material in the boulder reach were not exceeded, thus a) it was stable, b) no knickpoint progressed through the reach, and c) the resistant bed created a slight backwater that affected the mobility of sediment further upstream in the sandy portions of the reservoir. However, the thresholds for motion of sandy material were exceeded in the boulder reach and, moreover, the modeled sediment transport of this reach was increasing relative to locations upstream. Thus, sandy material entering this reach from upstream and eroded from the channel margins within this reach was efficiently exported from the reach, and deposition of sandy material did not occur. Although a large knickpoint did not progress from the dam upstream at this site, our findings can be applied to better understand the physical processes in the knickpoint migration model. At a knickpoint, the local increase in slope can be conceptualized as driving a longitudinal increase in

30

ACCEPTED MANUSCRIPT sediment transport, favoring bed erosion. Yet this knickpoint can only erode and migrate upstream if the threshold for erosion is exceeded.

PT

On the Ashuelot River, the resistance to erosion in the boulder reach prevented a knickpoint from

RI

progressing upstream from the dam. Instead incision was concentrated several km upstream of

SC

the dam, where more mobile sediment existed and the longitudinal gradients in sediment transport promoted erosion. If any unobserved knickpoint migration did occur, it likely occurred

NU

several km upstream of the dam where bed erosion was greatest. Furthermore, the decoupling of

MA

bed incision and bank erosion over the first 15 months post removal indicates that incision alone does not induce channel widening immediately. We speculate that subtle differences in the bank

D

erodibility affected the spatial patterns of lidar-detected bank erosion. Other dam removals that

TE

drained relatively narrow reservoirs were similarly influenced by geomorphic constraints expressed by relict bed and bank features. For example, Pearson et al. (2011) found bedrock

AC CE P

outcrops at the head of the reservoir and vegetated islands on the reservoir margins that limited erosion upstream of the Merrimack Village Dam. Major et al. (2012) stressed the importance local bedrock outcrops in influencing the variable export of sediment from the drained reservoir of the Marmot Dam.

Typically, local conditions impose site-specific constraints on the geomorphic response. This study provides a broadly applicable method to understand how these constraints affect the response. In some cases, they can create longitudinal variations in sediment transport. As the data show, gradients in sediment transport are a primary control on whether a location is prone to erosion or deposition, provided thresholds for transport are exceeded. This process lesson improves our understanding of sediment transport dynamics and river response in transient

31

ACCEPTED MANUSCRIPT systems. This understanding can be used to predict river response at the many dam removals expected during the next few decades.

PT

Restoring longitudinal and lateral connectivity is one of the stated benefits of using dam

RI

removals as a river restoration tool (Kondolf et al. 2006); however, it appears that connectivity of

SC

sediment was not greatly enhanced by the removal of the Homestead Dam. Even though our measurements show that sandy material from the reservoir was exported downstream from the

NU

drained reservoir, the modeling results and field observations suggest that sandy material was

MA

generally bypassing the dam prior to its removal. Modeling of the 2 year recurrence interval flow prior to the dam removal indicates that the thresholds for sediment transport of sandy material

D

were exceeded in high flow events throughout the reservoir, and that the d(Qs)/dx > 0 in several

TE

reaches of the reservoir. Thus, there is a physical argument for the continuity of sediment through this narrow, old reservoir prior to the dam removal. This finding is bolstered by field

AC CE P

evidence from the Merrimack Village Dam on the Souhegan River in southern New Hampshire, where Pearson et al. (2011) measured loss of sandy sediment from the reservoir prior to the dam removal. Inefficient trapping of sandy material in relatively narrow reservoirs is perhaps common. The extent of sediment trapping in the reservoir may be a rough indicator of the extent of dam removal effects. It appears the lateral connectivity of material from channels onto floodplains was not enhanced either, although for a different reason. Abundant floodplain sedimentation was not observed downstream of the former dam, despite the available sediment. Upstream flow regulation reduced peak flows during the study period and prevented the river from flooding and accessing the floodplains. While longitudinal connectivity was not enhanced by the dam removal because it was not severely compromised by the placement of the Homestead Dam, lateral connectivity was

32

ACCEPTED MANUSCRIPT not enhanced by the dam removal because the run-of-river Homestead dam was not the factor constraining lateral deposition of sediment. In regions like this, with hydrologic constraints from

PT

flood control by multiple dams, the efficacy of dam removal as a restoration tool must be

RI

considered in the context of other watershed-scale impacts.

SC

Two distinct future directions emerge from this work. First, the analysis presented here provides a tool to predict locations of erosion, sedimentation, and stability, but it does not exhibit how to

NU

predict the magnitudes of erosion or deposition. Second, the exact processes controlling bank

MA

erosion are not fully captured. We contend that the variable response of the banks in the drained reservoir is due in part to the variable integrity and resistance to erosion of the banks and perhaps

D

variable fluid erosional stresses not captured in our one-dimensional Qs modeling. Future work

TE

can further examine the forcing and resistance to erosion on the banks to more precisely depict

AC CE P

the widening in this drained reservoir.

7. Conclusions

This study shows that the observed locations of erosion and deposition on the bed and lower banks of the drained reservoir following dam removal can be predicted by modeling the sediment transport dynamics within the drained reservoir based on pre-removal surveys of channel geometry. These predictions rely on both a) exceeding thresholds for sediment transport and b) the spatial gradients in sediment transport, namely whether transport is increasing or decreasing from one location to the next downstream location. The patterns of bank erosion are not clearly predicted simply by the modeled forcing of sediment transport, such as the temporal and spatial

33

ACCEPTED MANUSCRIPT changes in Shields parameter or Qs. Subtle differences in the bank resistance to erosion likely play a critical role.

PT

This study emphasizes that variable resistance to erosion on the bed and banks due to the

RI

physiogeographic setting can set boundary conditions on river response to dam removal,

SC

especially in reservoirs with relatively little sedimentation. Reservoirs with relatively little sedimentation do not have thick layers of approximately homogenous material that buffer the

NU

influence of, for example, bedrock outcrops or coarse glacial deposits. Thus, channel response to

MA

dam removal is not adequately predicted by common conceptual models, such as knickpoint migration and channel incision followed by bank failures. The wide variety of dam removal

D

settings makes it unlikely that a single conceptual model can be applied universally. However,

TE

the physically-based approach outlined here, that accounts for spatial gradients in sediment transport and variable resistance to erosion, may be more widely applied to a variety of dam

AC CE P

removals and other transient geomorphic settings. Acknowledgements

This work was supported by the U.S. National Science Foundation (BCS-0724348 and BCS1103172), a Seed Grant from the National Center for Airborne Laser Mapping, and Dartmouth College Department of Earth Sciences. Reviews by J.E. Pizzuto, W.B. Dade, G. Whitfield, and an anonymous reviewer greatly improved the manuscript. D. Finnegan and S. Arcone helped with lidar and GPR analysis, respectively. A. Hartman, A. Kasprak, R. Chaudhary, J. Herman, G. Gartner, M. Truehart, B. Bramhall, A. Wearn, E. Anderson, J. Underwood, A. Singler, S. Huda, and E. Buraas helped immensely in the field.

34

ACCEPTED MANUSCRIPT References Bednarek, A.T., 2001. Undamming rivers: A review of the ecological impacts of dam removal.

PT

Environmental Management 27, 803-814.

RI

Besl, P.J., McKay, N.D., 1992. Method for registration of 3-D shapes. Proceedings SPIE 1611,

SC

Sensor Fusion IV: Control Paradigms and Data Structures 586. doi:10.1117/12.57955

NU

Bizzi, S., Lerner, D., 2013. The use of stream power as an indicator of channel sensitivity to erosion and deposition processes. River Research and Applications. doi: 10.1002/rra.2717.

MA

Buffington, J. M., Montgomery, D. R., 1997. A systematic analysis of eight decades of incipient motion studies, with special reference to gravel‐bedded rivers. Water Resources Research 33,

TE

D

1993-2029.

AC CE P

Bushaw‐Newton, K.L., Hart, D.D., Pizzuto, J.E., Thomson, J.R., Egan, J., Ashley, J.T., Johnson, T.E., Horwitz, R.J., Keeley, M., Lawrence, J., 2002. An integrative approach towards understanding ecological responses to dam removal: The Manatawny Creek study. JAWRA Journal of the American Water Resources Association 38, 1581-1599. Church, M.A., McLean, D., Wolcott, J., 1987. River bed gravels: sampling and analysis. In: Thorne C.R., Bathurst J.C., Hey R.D. (Eds.), Sediment Transport in Gravel-Bed Rivers. John Wiley and Sons, Chichester. pp 43-88. Dade, W. B., Renshaw, C. E., Magilligan, F. J., 2011. Sediment transport constraints on river response to regulation. Geomorphology 126, 245-251.

35

ACCEPTED MANUSCRIPT Doyle, M.W., Stanley, E.H., Harbor, J.M., 2002. Geomorphic analogies for assessing probable channel response to dam removal. Journal of the American Water Resources Association 38,

PT

1567-1579.

RI

Draut, A., Ritchie, A., 2013. Sedimentology of new fluvial deposits on the Elwha River,

SC

Washington, USA, formed during large‐scale dam removal. River Research and Applications.

NU

doi:10.1002/rra.2724

Exner, F. M., 1920, Zur physik der dünen, Akad. Wiss. Wien Math. Naturwiss. Klasse 129, 929–

MA

952.

Exner, F. M., 1925, Über die wechselwirkung zwischen wasser und geschiebe in flüssen, Akad.

TE

D

Wiss. Wien Math. Naturwiss. Klasse, 134, 165–204

AC CE P

Ferguson, R .I., 2012. River channel slope, flow resistance, and gravel entrainment thresholds. Water Resources Research. 48, W05517. Goldthwait, J.W. Surficial geology of New Hampshire, The New Hampshire State Planning and Development Commission, Concord, New Hampshire. 83 pp. Graf, W.H., 1984. Hydraulics of sediment transport. Water Resources Publications, Highlands Ranch, Colorado. 513 pp. Graf, W.L., 1999. Dam nation: a geographic census of American dams and their large-scale hydrologic impacts. Water Resources Research 35, 1305-1311. Julien, P.Y., 2010. Erosion and sedimentation, 2nd ed. Cambridge University Press, Cambridge, UK. 371 pp.

36

ACCEPTED MANUSCRIPT Knighton, D., 1998. Fluvial Forms and Processes A New Perspective. Arnold, London, UK. 383 pp.

PT

Kondolf, G.M., Boulton, A.J., O'Daniel, S., Poole, G.C., Rachel, F.J., Stanley, E.H., Wohl, E.,

RI

Bang, A., Carlstrom, J., Cristoni, C., Huber, H., Koljonen, S., Louhi, P., Nakamura, K., 2006.

SC

Process-based ecological river restoration: Visualizing three-dimensional connectivity and

NU

dynamic vectors to recover lost linkages. Ecology and Society 11, 5. Maclin, E., Sicchio, M., 1999. Dam removal success stories. Special. Report by Friends of the

MA

Earth, American Rivers, and Trout Unlimited, Washington, D.C. 175 pp. Major, J.J., O’Connor, J.E., Podolak, C.J., Keith, M.K., Grant, G.E., Spicer, K.R., Pittman, S.,

TE

D

Bragg, H.M., Wallick, J.R., Tanner, D.Q., Rhode, A., Wilcock, P.R., 2012, Geomorphic response of the Sandy River, Oregon, to removal of Marmot Dam. U.S. Geological Survey Professional

AC CE P

Paper 1792, 64 pp.

Meyer-Peter, E., Müller, R., 1948. Formulas for bed-load transport. Proceedings of the 2nd Meeting of the International Association for Hydraulic Structures Research. Stockholm, pp. 3964.

Mueller, E.R., Pitlick, J., Nelson, J.M., 2005. Variation in the reference Shields stress for bed load transport in gravel-bed streams and rivers. Water Resources Research 41, 1-10. Nilsson, C., Berggren, K., 2000. Alterations of riparian ecosystems caused by river regulation ram operations have caused global-scale ecological changes in riparian ecosystems. How to protect river environments and human needs of rivers remains one of the most important questions of our time. Bioscience 50, 783-792.

37

ACCEPTED MANUSCRIPT Paola, C., Voller, V., 2005. A generalized Exner equation for sediment mass balance. Journal of Geophysical Research: Earth Surface 110, F04014. doi:10.1029/2004JF000274.

PT

Pearson, A.J., Snyder, N.P., Collins, M.J., 2011. Rates and processes of channel response to dam

RI

removal with a sand-filled impoundment. Water Resources Research 47, W08504.

SC

doi:10.1029/2010WR009733.

NU

Petts, G.E., Gurnell, A.M., 2005. Dams and geomorphology: research progress and future directions. Geomorphology 71, 27-47.

MA

Pizzuto, J., 2002. Effects of dam removal on river form and process. Bioscience 52, 683-691.

D

Pizzuto, J., O'Neal, M., 2009. Increased mid-twentieth century riverbank erosion rates related to

TE

the demise of mill dams, South River, Virginia. Geology 37, 19-22.

AC CE P

Poff, N.L., Hart, D.D., 2002. How dams vary and why it matters for the emerging science of dam removal. Bioscience 52, 659-668.

Sawaske, S.R., Freyberg, D.L., 2012. A comparison of past small dam removals in highly sediment-impacted systems in the US. Geomorphology 151-152, 50-58. doi: 10.1016/j.geomorph.2012.01.013

Shields, A. 1936. Anwendung der aehnlichkeitsmechanik und der turbulenzforschung auf die geschiebebewegung, Mitt. Preuss. Versuchsanst. Wasserbau Schiffbau 26, 36 pp. Shuman, J.R., 1995. Environmental considerations for assessing dam removal alternatives for river restoration. Regulated Rivers: Research & Management 11, 249-261.

38

ACCEPTED MANUSCRIPT Simon, A., 1989. A model of channel response in disturbed alluvial channels. Earth Surface Processes and Landforms 14, 11-26.

PT

Soil Survey Staff, Natural Resources Conservation Service, United States Department of

RI

Agriculture. Web Soil Survey. Available online at http://websoilsurvey.nrcs.usda.gov/. Accessed

SC

12/5/2013.

NU

Stanley, E.H., Doyle, M.W., 2003. Trading off: the ecological removal effects of dam removal. Frontiers in Ecology and the Environment 1, 15-22.

MA

Williams, D.T., 1977. Effects of dam removal: An approach to sedimentation. Hydraulic

D

Engineering Center HEC-TP-50, U.S. Army Corps of Engineers, Davis, CA. 31 pp.

TE

Wohl, E.E., Anthony, D.J., Madsen, S.W., Thompson, D.M., 1996. A comparison of surface

AC CE P

sampling methods for coarse fluvial sediments. Water Resources Research 32, 3219-3226. Womack, W., Schumm, S., 1977. Terraces of Douglas Creek, northwestern Colorado: an example of episodic erosion. Geology 5, 72-76.

Figure Captions Figure 1. Ashuelot River, showing the study reach and related dams. Figure 2. Locations of cross sections, Homestead Dam, reservoir reach, and upstream control reach. Base image is lidar-derived topography. Inset shows breaching of the Homestead Dam. Figure 3 a) Looking upstream at the reservoir prior to dam removal from 100 to 400 m upstream of dam. This is the location of the Boulder Reach. Flow is 15 m3s-1. b) Similar view from a

39

ACCEPTED MANUSCRIPT slightly different angle three weeks after the dam removal showing the size and extent of boulders that had been inundated in the Boulder Reach. Flow is 0.7 m3s-1. c) Similar view of

PT

post-removal conditions with flow of 13 m3s-1, for comparison with panel (b) to show that

RI

reservoir was relatively narrow.

SC

Figure 4. Long profile of river bed and water elevations at time of lidar flights in July 2010 and

NU

July 2011. Segments of the study areas are differentiated.

Figure 5. Results of differencing lidar-derived DEMs on upper river banks and near-channel

MA

floodplain from 1 to 15 m from water edge at 3km upstream of the dam, a typical location of the reservoir. Erosion (positive values) and deposition (negative values) from before to after

D

removal (July 2010 to July 2011) embed true signal and noise. Most areas show little change, but

TE

arrows point to characteristic locations of a) falsely represented erosion and b) verified erosion

AC CE P

based on field surveys and observations. Hillshade shows topography from 2010 lidar data, including the unrealistically rough lidar-derived water surface topography. Figure 6. Flow record at former Homestead Dam (solid line) and 2-year recurrence interval flow (dashed line). Arrows indicate times of dam breach, lidar flights, and field surveys. Figure 7. Ground penetrating radar profile from dam to 1.2 km upstream prior to dam removal. Symbol (A) denotes water surface, (B) surface of sediment horizon, the channel bed, (C) interpreted multiple of sediment horizon, thus this is not a real horizon, (D) interpreted sub bottom horizons, perhaps indicative of pre-dam channel bed in locations upstream of boulder reach, approximately 2.5 m below surface of sediment horizon at -0.6 km. Figure 8. a) For time period immediately after dam removal, modeled sediment discharge of 2year recurrence interval storm using bed sediment grain size (solid line: Qs, Bed Sed.) and sandy 40

ACCEPTED MANUSCRIPT reservoir grain size (dashed line: Qs, Res. Sed.). Arrows indicate zones of increasing Qs with respect to distance downstream. Shading corresponds to the arrows and indicates predicted zones

PT

of erosion.

RI

b) Surveyed erosion and deposition of bed material at cross section intervals from 0 to 12 months

SC

after dam removal, with positive values for erosion and negative values for deposition.

NU

c) Cumulative bed erosion, where a positive slope of the line indicates a zone of net erosion and a negative slope indicates a zone of net deposition.

MA

d) Lidar measured erosion of bank material from 1 to 15 m from water edge at 500 m intervals

D

from time period 0 to 12 months after dam removal.

TE

d) Cumulative bank erosion measured by lidar. As above, positive slope of the line indicates

AC CE P

erosion and negative slope indicates deposition. f) Modeled sediment discharge (Qs) as in panel (a), except for time period starting at 12 months after dam removal.

g) Surveyed erosion and deposition of bed material from 12 to 15 months after dam removal, expressed as volume change for each reach between cross sections, with positive values for erosion and negative values for deposition. h) Cumulative bed erosion, where a positive slope of the line indicates a zone of net erosion and a negative slope indicates a zone of net deposition for time period 12 to 15 months post removal, and solid line shows from 0 to 15 months after dam removal. Scale is same as in panel (c) to show decreases in magnitude of channel adjustment between time periods.

41

ACCEPTED MANUSCRIPT

PT

Table 1. Number of of locations of predicted and observed erosion in sandy reaches in reservoir and downstream of dam 0-12 months 12-15 months predicted erosion 11 11 observed erosion 8* 7**

AC CE P

TE

D

MA

NU

SC

RI

predicted deposition 11 11 observed deposition 10 4*** * Of the 3 locations where deposition was observed, all were within 1 km downstream of a strong decline in Qs, indicating that deposiiton would be predicted if gradients in Qs was averaged over 1 km distances. ** Of the 4 locations where deposition was observed, 2 were within 1 km downstream of a strong decline in Qs, indicating that deposiiton would be predicted if gradients in Qs were averaged over 1 km distances. *** Of the 7 location where deposition was predicted but not observed, in 6 locations the volume of erosion from 12-15 months offset the volume of deposition from 0-12 months within 10%

42

ACCEPTED MANUSCRIPT Highlights

PT

We examine physical controls on channel response to dam removal Channel response reflects spatial gradients and thresholds in sediment transport (Qs)

RI

Channel response deviates from common conceptual models of dam removal

AC CE P

TE

D

MA

NU

SC

Qs gradients are often overlooked, but important in predicting erosion and deposition

43