Influence of Small Dams on Downstream Channel Characteristics in Pennsylvania and Maryland: Implications for the Long-Term Geomorphic Effects of Dam Removal1

Authors


  • 1

    Paper No. JAWRA-07-0006-P of the Journal of the American Water Resources Association (JAWRA). Discussions are open until June 1, 2009.

(E-Mail/Skalak: kskalak@udel.edu)

Abstract

Abstract:  We evaluate the effects of small dams (11 of 15 sites less than 4 m high) on downstream channels at 15 sites in Maryland and Pennsylvania by using a reach upstream of the reservoir at each site to represent the downstream reach before dam construction. A semi-quantitative geomorphic characterization demonstrates that upstream reaches occupy similar geomorphic settings as downstream reaches. Survey data indicate that dams have had no measurable influence on the water surface slope, width, and the percentages of exposed bedrock or boulders on the streambed. The median grain diameter (D50) is increased slightly by dam construction, but D50 remains within the pebble size class. The percentage of sand and silt and clay on the bed averages about 35% before dam construction, but typically decreases to around 20% after dam construction. The presence of the dam has therefore only influenced the fraction of finer-grained sediment on the bed, and has not caused other measurable changes in fluvial morphology. The absence of measurable geomorphic change from dam impacts is explicable given the extent of geologic control at these study sites. We speculate that potential changes that could have been induced by dam construction have been resisted by inerodible bedrock, relatively immobile boulders, well-vegetated and cohesive banks, and low rates of bed material supply and transport. If the dams of our study are removed, we argue that long-term changes (those that remain after a period of transient adjustment) will be limited to increases in the percentage of sand and silt and clay on the bed. Thus, dam removal in streams similar to those of our study area should not result in significant long-term geomorphic changes.

Introduction

The effects of dams on river morphology and fluvial processes have become increasingly important to watershed management during recent decades. Managers now routinely assess the effects of dam operations on sediment transport processes downstream of dams to aid stream ecological processes, and issues related to dam removal are widely discussed by a variety of stakeholders (Graf, 2003). The geomorphic changes that arise downstream of dams have important effects on endangered species, riverine habitat, engineering structures, and a host of other important issues in water resources management (Petts, 1984).

There is an extensive literature documenting how dams change downstream channels (Williams and Wolman, 1984; Brandt, 2000; Fassnacht et al., 2003; Grant et al., 2003; Graf, 2005). However, these studies have tended to focus on larger dams rather than the much more abundant smaller dams, and field observations are not uniformly distributed geographically (Graf, 2005). Several authors have tried to develop generalized conceptual models to help predict the downstream effects of dams on rivers, but these efforts are in their infancy (Brandt, 2000; Schmidt and Wilcock, 2008). Our present understanding of the geomorphic influence of dams on downstream channels recognizes a welter of important controls that can provide widely varying geomorphic responses in individual cases (Grant et al., 2003).

The geomorphic effects of dams are important in Pennsylvania, Maryland, and surrounding states because this region is a hotbed of activity for dam removal (Bushaw-Newton et al., 2002). Understanding the potential channel changes that can arise after dams are removed often requires an assessment or understanding of existing dam impacts. The influence of dams on sediment budgets and fluvial sediment transport processes is of concern for a variety of efforts to restore channels and estuaries downstream of dams in the region (Walter et al., 2006). Despite the extensive interest, however, few studies have documented the geomorphic effects of dams in this region, though some regional assessments of the hydrologic impacts of dams are available (Magilligan and Nislow, 2001).

This study represents an initial effort to document the geomorphic influence of dams on downstream channels in Pennsylvania and Maryland. Because the majority of dams in the region are small dams (Poff and Hart, 2002), these provide the primary focus for our research.

Study Design

The clearest means of documenting how dams influence channels downstream would be to initiate a monitoring program before dam construction, and to document how the channel responds through time after the dam is built. Because the initial surveys are unavailable (and also because the dams are already in place), we were forced to adopt a different approach.

In addition to field observations of the present channels below dams, we required a reasonable surrogate to represent conditions downstream before the dams were built. More precisely, we needed to locate a channel that had not been influenced by the construction of the dam, but whose underlying bedrock, levels of human disturbance, hydrologic regime, and sediment supply are similar to those of the downstream channel.

As a working hypothesis, we propose that the channel upstream of the impoundment provides the best available means to define geomorphic conditions downstream before dam construction (assuming that we can always carefully locate our sampling reaches above any influence of hydraulic backwater effects or deltaic sedimentation associated with the impounded channel downstream). This approach should be reasonable if the geomorphic character of the channel above the impoundment is similar to that below the dam, and if differences in controlling variables before dam removal can be scaled (if necessary) by simple measures such as drainage basin area. We address these conditions in detail below.

Study Areas

We selected 13 streams in Pennsylvania and two in Maryland (Figure 1). Twelve sites are located in the Piedmont Physiographic Province, while the other three sites are located in the Ridge and Valley Physiographic Province (Sevon, 2000). Pennsylvania has a humid continental climate, with temperatures usually ranging between −18°C and 38°C. Summers are generally warm, averaging between 20°C and 23°C. During the coldest months, temperatures average near the freezing point. Annual precipitation varies from 865 to 1320 mm (http://climate.met.psu.edu/data/IA/).

Figure 1.

 Locations of the 15 Study Sites Used to Assess Impacts of Dams on Downstream Channels. The number for each site corresponds to the site identification in Table 1.

The streams of our study sites are primarily single-thread channels with coarse-grained bed material and forested riparian zones. Aside from the presence of dams, the study reaches have no engineering structures or other evidence of direct anthropogenic controls. The streams vary from first to fourth order (using the Strahler stream ordering scheme) (Strahler, 1957) (Table 1). Drainage areas range from less than 1 to about 300 km2 (Table 1).

Table 1.   Representative Characteristics of Streams, Dams, Impoundments, and Watersheds at Each Study Site.
Dam NameStreamID#Dam Height (m)Drainage Area (km2)Stream OrderHRT (days)Dam Age (years)
  1. Notes: Mean Hydraulic Residence Time (HRT) was estimated by dividing the volume of the impounded water by the stream discharge measured below the impoundment on the day of sampling. This method of estimating mean HRT assumes steady state conditions of impounded water inflow, outflow, and volume, which we could not assess because of the one-time nature of our discharge measurements. To account for this possible source of error, we also estimated mean HRT by dividing the impoundment volume at normal water levels by the long-term average flow estimate derived from regional regression relationships between watershed area and average streamflow (Reed, 1999). Because our two types of estimated mean HRT were highly correlated (Spearman correlation = 0.82), we opted to use the estimate based on one-time downstream flow and volume measurements made in fall, 2002. ID# refers to the number given to the site in Figure 1.

Core Cr.Core Cr.114.324.46354.4431
Fish WeirDismal Run20.90.8020.01NA
Prettyboy ReservoirGunpowder R.357.0198.853382.2674
Haskins DamTohickan Cr.41.5110.1930.20150+
HiberniaBirch Run519.811.982110.8113
Hickory Run DamHickory Run63.07.2411.7470+
HopewellFrench Cr.78.86.70157.3070
Mensch Mill DamW. Br. Perkiomen Cr.82.111.7710.12185
Woodbine DamMuddy Cr.91.5287.1930.01150+
Stametz DamSand Spring Run104.69.5110.2670+
Stimson Run DamStimson Run118.80.72126.8222
UnnamedTributary Middle Cr.120.93.4041.0770+
Dr. Jackson DamSwamp Cr.131.2115.9810.18240
Lake ConewagoConewago Cr.145.53.94114.4388
Atkinson DamWinter’s Run, Otter Pt. Cr.1511.3116.8820.7170+

Dams in our study reaches are predominately small dams that have relatively little storage capacity and hence little influence on flow distribution or flow frequency. For example, six dams in our study are less than 2 m high, two dams are between 2 and 3 m high, three dams are between 3 and 4 m high, and the remaining four dams are higher than 4 m (Table 1). The hydraulic residence times for the impoundments of nine of the 15 sites are two days or less. Most of the dams release water from spillway overflows. The ages of the dams varied: four of the dams are more than 100-years old, seven of the dams are between 50-years and 100-years old, and three of the dams are between 10-years and 50-years old. The age of one dam could not be determined.

Methods

Field Methods

Our study design is based on the assumption that the geomorphic setting of the reach upstream of the impoundment is similar to that below the dam. To verify this assumption, we described the geomorphic setting of each study site using a qualitative stream classification scheme designed for a previous study of streams in southeastern Pennsylvania (Patrick Center for Environmental Research, 1999). The classification scheme involves visual observations of (1) the dominant grain size class of the bed material (based on the Udden-Wentworth grain size classification) (Prothero, 1989); (2) the type of landforms that store bed material in the reach (alternate, mid-channel or point bars, or no bed material storage features); (3) the morphology of the bed (plane-bed, pool/riffle, or step-pool); (4) the presence or absence of an active floodplain; (5) the channel planform (single-thread, braided, or anastomosing); and (6) the presence or absence of bedrock exposures in the study reach. These qualitative observations were supplemented with quantitative measurements of sinuousity obtained from aerial photographs.

We established study reaches that encompass approximately three pool riffle sequences above and below all the dams of Table 1. Study reaches downstream of dams were located as close to the dams as possible. Upstream reaches were located as close as possible to the impoundments, but above any hydraulic influence of ponded water downstream. To ensure this, we followed the following criteria. First, we noted that as the stream channels approached the impoundments, the bed material typically became significantly finer. We always placed at least one riffle between the downstream end of our study sites and areas where finer grained sediment began to accumulate. Second, we performed backwater computations to assess the upstream extent of the hydraulic influence of the downstream impoundment. Our results from the computations indicate that most of the sites (with the potential exception of four sites) are located above any backwater effects.

At each study reach, we surveyed the elevation of the bed and water surface along the center of the channel using a TOPCON Total Station. After analyzing the data, we decided to use the slope of the water surface to represent the slope of the channel at our study reaches rather than the slope of the bed. The two slopes were generally similar, but the correlation coefficients fit to the water surface slopes were much higher than those fit to the bed topography, due to the greater topographic variability of the bed when compared with the water surface. The slope of the water surface also provides a better measure of the frictional energy loss associated with flow in the channel (French, 1985).

We also established 10 equally spaced transects across the channel at each site. We surveyed the morphology of the channel along the third, fifth, and seventh of these transects. Because floodplain deposits were poorly developed at many transects, we did not define the bankfull depth of the stream during our study. The channel width was defined by the extent of the water surface across the channel cross-section (Skalak, 2004). Upstream and downstream reaches were surveyed on the same day to ensure that flow conditions for determining width and water surface slope were always comparable. All surveys were completed during low flow conditions.

The size distribution of the bed material was measured using the Wolman (1954) method by sampling at twenty equally spaced locations along each transect, giving a total of 200 samples. Rather than measuring the diameter of each grain, a template of square openings was used. This method provides a measure of size that is equivalent to conventional sieving (Church et al., 1987). The template openings were sized to place each grain into a class of the Udden-Wentworth grain size scale (silt and clay (<0.0625 mm), sand (0.0625-2 mm), granule (2-4 mm), pebble (4-64 mm), cobble (64-256 mm), and boulder (>256 mm). With a sample size of 200 particles, results obtained by Rice and Church (1996) suggest potential errors from 8% to around 20% for individual grain size fraction percentiles, depending on the size and sorting of the bed material.

All data were collected in the Fall of 2002 and the Spring of 2003 during low flow conditions.

Data Analysis

We used the following methods to assess the magnitude of dam impact on downstream reaches. First, we normalized all water surface width data by the square root of the drainage basin area to account the commonly observed increase in width with increasing discharge and drainage basin area (Knighton, 1998). Then, the “normalized downstream width” for each site was subtracted from the “normalized upstream width” for each site. The resulting dimensionless value was then multiplied by the square root of the downstream drainage basin area to estimate the change in width of the channel at the downstream site before construction of the dam. Additionally, the “normalized downstream width” and “normalized upstream width” were used as down and up, respectively, in Equation (1) (presented below) to determine percent difference (Skalak, 2004).

To generate box plots summarizing the ranges of the percent difference in measured variables, we used the following formula

image(1)

where down is the percentage for a particular variable measured in the downstream reach and up is the corresponding percentage in the upstream reach.

To determine the statistical significance of our results, we used a Wilcoxon signed rank test with a two-tailed distribution in SPSS version 14 (SPSS Inc., 2005). This test was used because it includes information about both the sign of the differences and the magnitude of the differences between pairs. We could not use the more common paired t-test, because our data were not normally distributed and could not be transformed to achieve normality.

Results

Geomorphic Setting of Upstream and Downstream Reaches

Streams of the study reaches are primarily single-thread, sinuous channels lacking in bed storage landforms (i.e., bars) with planar, cobble, or boulder beds (Table 2). A few reaches have sandy beds, pool-riffle, or step-pool morphology, and bed material storage landforms. About half of the reaches (16) are bordered by floodplains, while the remaining 14 reaches have no floodplain; 25 of the reaches have sinuosities of less than 1.5, while five are meandering (sinuosity >1.5). Bedrock is exposed in the bed or the banks at 14 of the reaches.

Table 2.   Geomorphic Characteristics of Upstream and Downstream Reaches at the 15 Study Sites.
SiteReachBed SedimentSediment Storage FeaturesBed MorphologyFloodplainPlanformSinuosityBedrockReach Length (m)Channel Width (m)Depth (m) Discharge (m3/s)
  1. Two examples where upstream (UP) and downstream (DOWN) reaches have different geomorphic characteristics are in bold font.

Core Cr.UPCobbleNonePlane bedPresentSinuous1.15No648.741.330.00886
Core Cr.DOWNCobbleNonePlane bedPresentSinuous1.18No8910.681.480.02985
Dismal RunUPSandNonePool/rifflePresentMeandering2.08No341.650.720.01462
Dismal RunDOWNSandNonePool/rifflePresentMeandering1.76No422.810.750.01532
Gunpowder R.UPBoulderNonePlane bedAbsentSinuous1.30Yes10510.062.10.18976
Gunpowder R.DOWNBoulderNonePlane bedAbsentSinuous1.48Yes8722.682.171.23233
Tohickan Cr.UPBoulderNonePlane bedAbsentSinuous1.26No12019.673.250.06249
Tohickan Cr.DOWNBoulderNonePlane bedAbsentSinuous1.22No10016.491.780.03660
Birch RunUPCobbleGravel barsPlane bedPresentSinuous1.43No514.812.290.00919
Birch RunDOWNCobbleGravel barsPlane bedPresentSinuous1.20No717.443.320.13483
Hickory RunUPBoulderNonePlane bedAbsentSinuous1.38Yes424.851.450.01612
Hickory RunDOWNBoulderNonePlane bedAbsentSinuous1.38Yes368.001.750.01296
French Cr.UPCobbleNonePlane bedPresentSinuous1.46No823.910.920.10758
French Cr.DOWNCobbleNonePlane bedPresentSinuous1.37No1036.490.630.00128
W. Br. Perkiomen Cr.UPBoulderNonePlane bedAbsentSinuous1.33Yes508.862.330.05094
W. Br. Perkiomen Cr.DOWNBoulderNonePlane bedAbsentSinuous1.28Yes789.021.680.03743
Muddy Cr.UPBoulderNonePlane bedAbsentSinuous1.38Yes14824.003.220.39434
Muddy Cr.DOWNBoulderNonePlane bedAbsentMeandering1.54Yes12521.144.170.33208
Sand Spring RunUPBoulderNoneStep poolAbsentSinuous1.22Yes486.100.650.04758
Sand Spring RunDOWNBoulderNoneStep poolAbsentSinuous1.22Yes336.951.310.04302
Stimson RunUPCobbleNonePlane bedPresentSinuous1.30No511.641.230.00509
Stimson RunDOWNCobbleNonePlane bedPresentSinuous1.19No432.280.620.00295
Tributary Middle Cr.UPSandNonePlane bedPresentSinuous1.36No771.020.410.00085
Tributary Middle Cr.DOWNCobblesNonePlane bedPresentSinuous1.40No462.660.720.00056
Swamp Cr.UPCobbleNonePlane bedAbsentSinuous1.37Yes19619.132.040.30104
Swamp Cr.UPCobbleNonePlane bedAbsentSinuous1.06Yes18023.792.290.26774
Conewago Cr.UPCobblePt barPool/rifflePresentMeandering1.81No432.940.740.02643
Conewago Cr.DOWNCobblePt barPool/rifflePresentMeandering1.61No483.420.840.03093
Winters Run, Otter Pt. Cr.UPBoulderNonePlane bedAbsentSinuous1.08Yes10918.362.450.13388
Winters Run, Otter Pt. Cr.DOWNBoulderNonePlane bedAbsentSinuous1.16Yes10924.342.350.15595

Table 2 indicates that 13 of the 15 sites have similar geomorphic characteristics upstream and downstream. Two sites where characteristics are different upstream and downstream are identified in Table 2 by bold font. All sites have the same sediment storage features, bed morphology, extent of bedrock exposure, and floodplain characteristics in upstream and downstream reaches. Channel planform above and below the dam is also the same, except at Muddy Creek, where the upstream reach is sinuous and the downstream reach is meandering. However, the sinuousities for these two reaches differ by only 0.18, which is not highly significant. The dominant bed material grain size is also the same at all sites, except for the Middle Creek Tributary, where the upstream reach is sandy and the downstream reach is composed of cobbles. We demonstrate below, however, that this difference is likely caused by dam construction, so it does not invalidate our study design. These results indicate that the upstream and downstream reaches of our study sites are geomorphically similar and that differences between them cannot be ascribed to differences in geomorphic setting.

Effects of Dams on Channel Slope and Width

Figure 2 compares the slope of the water surface in upstream and downstream reaches. Figure 2A shows the downstream water surface slope plotted vs. the upstream water surface slope, as well as a regression equation fit to these data and a line illustrating perfect agreement between upstream and downstream values. Seven sites had higher water surface slopes upstream, while eight sites had higher water surface slopes downstream. The 95% confidence intervals for the intercept and slope of the regression equation are −0.004 to 0.015 and 0.096 to 1.021. These values include an intercept of 0 and a slope of 1, suggesting that the regression equation cannot be distinguished from the line of perfect agreement, i.e., that the upstream and downstream slopes are not significantly different. The box plot of percent differences illustrates considerable variability, but the median value of 16% and the spread of the data above and below 0% support these interpretations. A Wilcoxon signed ranks test performed on the paired slope data also indicates that the upstream and downstream slopes are not significantly different (p = 0.91). Equivalent results were obtained when the slope values were divided by drainage basin area raised to a variety of different exponents, suggesting that this result is particularly robust.

Figure 2.

 (A) Slope of the Water Surface Upstream of the Impoundment vs. Slope of the Water Surface Downstream of the Dam. A regression equation and a line indicating equal values of upstream and downstream slopes are also illustrated. (B) Box Plot Illustrating the Range of % Differences in Slopes Between Upstream and Downstream Reaches (computed using Equation 1). The box plots indicate values of the following percentiles: 5, 25, 50, 75, and 95%.

Figure 3 compares the width of the water surface (measured at low flow) in upstream and downstream reaches. Figure 3A is a plot of upstream and downstream water surface width normalized by the square root of the drainage area. It indicates that the downstream channel width is wider at four sites, narrower at eight sites, and essentially the same at three sites. Figure 3B illustrates the range of percent differences in normalized widths between upstream and downstream reaches. The range of individual values of percent differences includes both positive and negative values, with a median value of 13%, indicating a weak trend towards increased channel widths below the dam. This trend is not highly significant statistically (p = 0.078, Wilcoxon signed rank test), although the marginal p-value makes it difficult to rule out potentially significant impacts of dams in the downstream reach.

Figure 3.

 (A) Average Channel Width Upstream of the Impoundment vs. Average Channel Width Downstream of the Dam. Widths are scaled by the square root of the drainage basin area. A line indicating equal values of upstream and downstream width values is also illustrated. (B) Box Plot Illustrating the Range of % Differences in Scaled Widths Between Upstream and Downstream Reaches (computed using Equation 1). The box plots indicate values of the following percentiles: 5, 25, 50, 75, and 95%. The open circles indicate the values of outliers above and below these percentiles.

Because the effects of dams on downstream water surface width are ambiguous, we also analyzed our data within the context of variations in water surface widths observed at individual sites. We used our data to estimate the changes in width that could have been caused by dam construction at each site (using the procedure described in the Methods section), and then we compared these changes to an envelope defining the maximum variability in water surface width below the dam (based on three surveyed cross sections at each site) (Figure 4). At nine sites, the predicted changes in width lie inside of this envelope, while differences in width at six of the sites lie outside the envelope. At two of the sites where predicted changes exceed the observed variability, an increase in width is predicted to have resulted from dam construction, while at four sites, a decrease is width is predicted. The average predicted change in width caused by dam construction represents a decrease in width of 0.57 m, which is smaller than the envelope of width variability at every site.

Figure 4.

 Estimated Differences in Water Surface Width Caused by Dam Construction at All 15 Sites. An envelope indicating the variability in width for three transects downstream of dams is also illustrated. The overall mean difference in water surface width is +0.57 m.

Composition of the Bed Material

When evaluating the influence of dams on stream channels, it is important to assess the ability of streams to respond to changes in discharge and sediment supply (Brandt, 2000; Schmidt and Wilcock, 2008). The ability of a channel to adjust its boundaries in response to a disturbance (i.e., “lability” from Grant et al., 2003) will largely determine the magnitude of the dam’s impact on the channel. Consequently, underlying bedrock lithology and bed material composition are significant factors influencing downstream impacts.

Table 3 presents a summary of the percentages of boulders and exposed bedrock at the upstream reaches for all the sites (only upstream reaches are included in Table 3 because we have already established that the upstream and downstream reaches are comparable, and also because the presence of the dams influences the composition of the bed material in downstream reaches); 12 of the 15 sites have bed material that includes boulder-sized sediment, and six of these have bedrock exposed on the bed. On average, boulders and exposed bedrock comprise 11.3 and 4.7% of the bed material at our study sites, suggesting that these channels may have limited ability to adjust their bed elevations through incision or other processes involving active transport of the entire stream bed.

Table 3.   Percentages of Boulders and Exposed Bedrock at the Upstream Reaches of the Study Sites.
Site% Boulders% Bedrock% Boulders  + Bedrock
Core Cr.1.601.6
Dismal Run000
Gunpowder R.14.12.917
Tohickan Cr.46.8046.8
Birch Run3.303.3
Hickory Run14.110.424.5
French Cr.5.305.3
W. Br. Perkiomen Cr.13.73043.7
Muddy Cr.23.421.144.5
Sand Spring Run17.61.619.1
Stimson Run0.900.9
Tributary Middle Cr.0.000.0
Swamp Cr.10.72.313.0
Conewago Cr.000.0
Winter’s Run Otter Pt. Cr.18.42.921.3
Mean11.34.716.1
Median10.70.013.0
SD12.59.017.2

Figure 5 presents data summarizing differences in bed material grain size between upstream and downstream reaches. The box plots presented in Figure 5 demonstrate that fractions of silt and clay, sand, and granules are significantly lower in downstream reaches (= 0.036, 0.006, and 0.363, respectively, Wilcoxon signed ranks test), whereas fractions of pebbles and cobbles are significantly greater (= 0.001 and = 0.112, respectively). In contrast, there were no significant upstream-downstream differences in the fraction of boulders (= 0.865), or bedrock (= 0.612).

Figure 5.

 Box Plots Illustrating the Percent Differences in Bed Material Grain Size Classes for All 15 Sites. Results are presented for six grain size classes and for exposed bedrock. The box plots indicate values of the following percentiles: 5, 25, 50, 75, and 95%. Square boxes indicate the values of outliers above and below these percentiles.

Comparing percentiles of the bed material cumulative grain size distributions from upstream and downstream reaches further supports these results. The 16th percentile grain diameter (16% of the sample is finer than this size) is significantly coarser downstream of the dam than upstream for all but three of the sites (Table 4). According to the Wilcoxon signed rank test, the probability that the 16th percentile grain diameter is the same in upstream and downstream reaches is only 0.036. The 50th percentile grain diameter (the median grain size) is coarser in the downstream reach for all but three sites (Table 4). The Wilcoxon signed rank test indicates a probability of 0.023 that D50 is the same upstream and downstream reaches. The coarser fractions of the bed, as represented by the 84th percentile grain diameter, are unaffected by the presence of the dam (Table 4). The Wilcoxon signed rank test indicates that the probability that the upstream and downstream 84th percentile grain diameters are the same is 0.776.

Table 4.   D16, D50, D84, and Percent Difference (as defined in Equation 1) for Each Site.
Site IDD16 Upstream (mm)D16 Downstream (mm) D16–% DifferenceD50 Upstream (mm)D50 Downstream (mm) D50–% DifferenceD84 Upstream (mm)D84 Downstream (mm) D84–% Difference
Core Cr.0.3760.906141.317.39318.8138.295.05890.961−4.3
Dismal Run0.0170.02446.50.5640.62711.119.12016.935−11.4
Gunpowder R.0.5220.030−94.233.37224.390−26.9153.717142.015−7.6
Tohickan Cr.50.85229.222−42.5155.528187.36420.5473.784486.3162.6
Birch Run0.0896.2966945.29.99031.646216.895.222124.43230.7
Hickory Run0.5481.980261.534.81238.37510.2157.896138.250−12.4
French Cr.0.1940.550183.722.37525.82215.4121.421110.855−8.7
W. Br. Perkiomen Cr.0.39220.7295193.727.21980.605196.1243.960397.91463.1
Muddy Cr.0.7440.697−6.249.62530.830−37.9374.464297.922−20.4
Sand Spring Run0.47727.2415608.077.01986.26012.0207.137274.27632.4
Stimson Run0.3910.54840.29.17614.82261.561.61261.7450.2
Tributary Middle Cr.0.0100.01445.30.0300.6161926.00.79633.5534117.2
Swamp Cr.0.0303.52611,601.225.56334.00033.0138.487134.938−2.6
Conewago Cr.0.0260.03119.81.5889.089472.221.39135.91567.9
Winters Run, Otter Pt Cr.0.40218.9974622.318.26272.835298.8231.943181.880−21.6
Mean3.6717.3862304.432.16843.740214.5159.734168.527281.7
Median0.3910.906141.322.37530.83020.5138.487134.938−2.6
SD13.05410.7783616.439.73647.921494.6131.854137.8231061.4

Figure 6 illustrates the changes in the grain size distribution of the bed material caused by dam construction. Two curves are illustrated in Figure 6, one representing the average cumulative grain size distribution for all the upstream sites and the other representing the corresponding curve for the downstream sites. The finer-grained “tail” of these average grain size distributions is coarser downstream, but the larger particles are not impacted by the presence of the dam. According to Figure 6, the presence of the dam leads to a change in D16 from 0.31 to 1.10 mm, a shift from medium sand to very coarse sand according to the Udden-Wentworth grain size scale. The impact of the dam on the median grain size causes an increase from approximately 36 to 54 mm (both are pebbles). D84 averages 247 mm in upstream reaches, whereas the value for downstream reaches averages 260 mm (both are cobbles).

Figure 6.

 Average Cumulative Grain Size Distributions of Upstream and Downstream Reaches. D16, D50, and D84 are indicated. Arrows indicate the coarsening of the fine-grained tails of the grain size distribution caused by dam construction.

Discussion

Effect of Dams on Channel Morphology and Grain Size

Our results demonstrate that the effects of dams on downstream channel morphology are minor. No significant differences in the water surface slope upstream and downstream of dams were observed. Differences in water surface width at low flow (scaled by drainage basin area) were not statistically significant, though a weak tendency towards larger widths below dams was noted. However, predicted changes in water surface width are relatively small on average (less than 1 m), and at most sites, changes in width are smaller than the observed variability in width, which is typically on the order of several meters (Figure 4). Thus, the effects of dam construction on water surface width are likely to be small enough to be difficult to distinguish over the background variability in width.

The influence of dam construction on bed material differs depending on the size of the sediment being considered. The largest bed material (boulders or, equivalently, D84) and the fraction of exposed bedrock are not influenced by dam construction (Figures 5 and 6). Smaller size fractions, however, show consistent trends: the presence of the dam causes significant coarsening of the bed, with increases in the percentages of cobbles and pebbles, and decreases on the percentages of silt and clay and sand below the dams.

The Importance of Geologic Setting

The streams of the study reaches are primarily single-thread, sinuous channels with planar, cobble, or boulder beds (although a few reaches have sandy beds, pool-riffle, or step-pool morphology). Landforms that store abundant bed material, such as alternate bars, are rare (Table 2). Half of the reaches are not bordered by a floodplain. Bedrock is exposed in the bed or the banks at approximately half of the reaches.

Turowski et al. (2008) suggest that a “bedrock river” should be defined as any channel that “cannot substantially widen, lower, or shift its bed without eroding bedrock.” Indicators of bedrock streams in the field include “bedrock outcrops in the river channel, thin or discontinuous alluvial cover, and/or steep bedrock channel walls.”Turowski et al. (2008) recognized both “gravel-bed bedrock rivers” and “gravel-bed alluvial rivers.” We argue that most, if not all, of the streams in this study fall into the former category: bedrock channels whose beds are mantled with a thin gravel cover, a morphology that is related to the low rates of bed material supply, transport, and storage typical of the mid-Atlantic region (Meade, 1982; Pizzuto, 2006).

Viewing the streams of our study as bedrock channels provides a clear explanation for our observations. For example, the lack of downstream morphologic change in width and slope can be attributed to the difficulty of eroding exposed bedrock and boulders. Stability imparted by geologic controls is enhanced by well-vegetated, cohesive banks (Allmendinger et al., 2005) and low rates of bed material supply and transport (Meade, 1982; Pizzuto, 2006).

The composition of the bed material downstream of the dams in our study areas likely reflects a combination of geologic controls and changes in sediment transport processes induced by dam construction. Because the fraction of boulders on the bed is not affected by the presence of dams, it is reasonable to speculate that boulders are supplied to the channel locally and not from upstream reaches, either by very slow stream incision (Reusser et al., 2004) or by hillslope processes. It is also logical to consider that the boulders are not actively transported by the stream because the supply in both reaches is the same. Cobbles and smaller sediment, however, are actively transported and these sediments are probably trapped in the reservoir upstream of the dam. As a result, the supply of cobbles and smaller sedimentary particles to channels directly below the dams is probably less than the supply to channels above the dams, resulting in the observed coarsening of the bed material.

These hypotheses indicate that the stream channels of our study area reflect the combined influence of geologic setting and alluvial sediment transport processes. The presence of boulders and exposed bedrock over significant fractions of the bed (as documented in Table 3), combined with the absence of continuous floodplains, indicates the potential influence of long-term geologic processes such as valley incision on these stream channels, and the continued influence of these processes on channel morphology. The presence of alluvial gravel, sand, and silt and clay suggests that alluvial sediment transport processes have also partially controlled the composition of the bed material.

These interpretations are compatible with the process-based definition of bedrock channels provided by Turowski et al. (2008) and therefore, the classification of these streams as gravel-bed bedrock channels is not only justified: it provides a means of explaining and interpreting our observations.

These ideas provide a unified working hypothesis for interpreting the results of our study. After dams are constructed in these channels, we predict the following changes (Table 5). The fractions of exposed bedrock and boulders in downstream channels will remain the same. The median grain diameter will increase, but changes will be small, such that the median grain diameter will remain in the pebble size class (Figure 6). Statistics that describe the fractions of finer-grained sediment will change significantly. D16 will increase from medium to very coarse sand, and the percentage of sand and silt and clay on the bed will decrease from about 35% to about 20% (Figure 6). Because the channel is supported by bedrock exposures and relatively immobile boulders, however, changes in channel morphology will be limited. Changes in water surface slope and width will not be detectable (Table 5).

Table 5.   Summary of the Predicted Effects of Dam Construction on Downstream Channels in Our Study Area.
ParameterChangeNotes
D16YesD16 increases from medium sand to very coarse sand after dam construction
D50YesD50 increases after dam construction, but remains in the pebble size class
D84NoRemains in the cobble size class after dam construction
WidthNo 
SlopeNo 
% Sand and MudYesIncreases from ∼20 to ∼35% after dam construction (Figure 6)
% BouldersNo 
% BedrockNo 

These results are also consistent with emerging ideas in the literature on downstream dam impacts. The effects of dams on downstream channels have often been considered within the context of the theory of alluvial channels (Williams and Wolman, 1984; Brandt, 2000; Magilligan et al., 2003). This approach has provided many useful explanations, including the widely observed phenomena of armoring of stream beds below dams (Williams and Wolman, 1984; Schmidt and Wilcock, 2008) and the commonly cited narrowing of stream channels below dams caused by reduced peak discharges (Andrews, 1986; Allred and Schmidt, 1999). However, considerably different channel responses have been documented in watersheds where bedrock exerts a first-order control over channel processes (Fassnacht et al., 2003; Grant et al., 2003). In channels that have low sediment supply or a limited capacity to adjust their boundaries, dams may not significantly influence channel morphology (Petts and Gurnell, 2005). Our results support the notion that predicting the effects of dams on downstream channels requires both an understanding of geologic setting (Grant et al., 2003), and an assessment of the ability of downstream channels to adjust to changes in discharge and sediment supply (Schmidt and Wilcock, 2008).

Estimating the Long-term Effects of Dam Removal

Although monitoring studies of dam removals are becoming more common (Bushaw-Newton et al., 2002; Doyle et al., 2003; Wildman and McBroom, 2005; Cheng and Granata, 2007), empirical knowledge of the effects of dam removal is still limited. Furthermore, most observations and conceptual models tend to focus on the transient effects of dam removal, the shorter-term patterns of upstream sediment mobilization and downstream sediment storage. Very little research has been conducted on the long-term effects of dam removal. Graf (2005) suggests that one of the most important unanswered questions involves the likely course of channel change following dam removal.

We believe that the results of our study can provide some useful estimates of the long-term effects of dam removal on downstream channels. Our reasoning is as follows: we have argued that the reaches upstream of existing dams provide a useful surrogate for the channel downstream before dam construction. If the dam is removed, we envision the following scenario. For an initial period of adjustment, sediment will be eroded from reservoir deposits upstream, and a transient sediment pulse will likely pass into and through the reach below the dam (Pizzuto, 2002). During this period, changes in channel morphology and bed composition may be expected. However, after the new channel within the reservoir reach has stabilized, the supply of sediment and distribution of discharges should approach predam levels, and the channel will slowly stabilize. We argue that the reach upstream of the dam has already adjusted to these conditions, and therefore we use the characteristics of the upstream channels to represent the equilibrium channel that will develop downstream of the dams long after dam removal.

Following this approach, the results summarized in Table 5 can be directly applied as predictions of downstream channel characteristics long after dam removal. Thus, we propose that long after dam removal, the water surface width, slope, percent exposed bedrock, and percent boulders will be unchanged. The median grain diameter will decrease slightly, but it will remain in the pebble size fraction. The percentages of sand and silt and clay on the bed will increase significantly. Apparently, our results indicate that the gross geomorphic characteristics of these channels will remain relatively unaffected by dam removal. However, as the amount of sand increases on the bed, parts of the bed could become increasingly mobile, depending on the extent to which the distribution of flows is influenced by dam removal. This increased mobility and the disturbance it represents could have important impacts on riverine organisms (Minshall, 1984; Waters, 1995; Peterson, 1996; Bond and Downes, 2003).

Conclusions

We have documented the effects of small dams on downstream channels at 15 sites in Maryland and Pennsylvania using a site upstream of the reservoir to represent the downstream reach before dam construction. We first demonstrated that the upstream reaches are geomorphically similar to reaches below the dams. Then, we evaluated differences in geomorphic and bed material characteristics between upstream reaches and downstream reaches to quantify the effects of the dams on downstream channels. Our data indicate that dams have had no measurable influence on water surface slope, water surface width, and the percentages of exposed bedrock or boulders on the streambed. The median grain diameter (D50) of channels below dams is increased slightly by dam construction, but D50 remains within the pebble size class. The percentage of sand and silt and clay on the bed averages about 35% before dam construction, but typically decreases to around 20% after dam construction. These results indicate that dam construction has not significantly modified the geomorphic characteristics of stream channels below the dams at our study sites: measurable changes have only occurred to the finer fractions of the bed material.

Understanding the geologic setting of our study sites is central to an accurate interpretation of our results. The streams of the study reaches are primarily single-thread, sinuous channels with planar, cobble, or boulder beds and little or no sediment storage features. Half of the sites lack a floodplain and half have bedrock exposures in the bed and banks. Boulders are exposed at 12 of 15 sites, and they average 11% of the bed material at our study reaches. We have classified these streams as gravel-bed bedrock channels, streams where fluvial processes are largely controlled by frequent exposures of bedrock. Therefore, we attribute the lack of downstream change in width and slope to the exposed bedrock and boulders, which is further enhanced by well-vegetated, cohesive banks and low rates of bed material supply and transport. The bed material downstream of the dams likely reflects both geologic setting and dam-induced changes in sediment transport processes.

Our results also have important implications for channel changes that could persist long after the removal of the dams at our study sites. We argue that after an initial period of transient adjustment as sediment eroded from upstream reservoirs passes through downstream channels, changes in channel geomorphic characteristics will be limited to increases in the percentages of silt and clay, sand, and granules, and that changes in the other geomorphic characteristics measured during this study will not be detectable. Thus, dam removal projects in streams similar to those of our study area should not cause significant long-term geomorphic changes to stream channels downstream.

Acknowledgments

Funding for this project was provided by a Growing Greener Grant from the Pennsylvania Department of Environmental Protection (PA DEP). The views expressed herein do not necessarily reflect those of PA DEP. We thank colleagues in the Patrick Center for Environmental Research of the Academy of Natural Sciences for their help with study design and data collection, especially Rebecca Brown, Dianne Winter, Timothy Nightengale, and Jamie Carr. Patricia Jenkins assisted with field data collection. We would also like to thank the reviewers of the original version of this manuscript, whose thoughtful comments greatly improved the work.

Ancillary