Landscape context and the biophysical response of rivers to dam removal in the United States

Abstract

Dams have been a fundamental part of the U.S. national agenda over the past two hundred years. Recently, however, dam removal has emerged as a strategy for addressing aging, obsolete infrastructure and more than 1,100 dams have been removed since the 1970s. However, only 130 of these removals had any ecological or geomorphic assessments, and fewer than half of those included before- and after-removal (BAR) studies. In addition, this growing, but limited collection of dam-removal studies is limited to distinct landscape settings. We conducted a meta-analysis to compare the landscape context of existing and removed dams and assessed the biophysical responses to dam removal for 63 BAR studies. The highest concentration of removed dams was in the Northeast and Upper Midwest, and most have been removed from 3rd and 4th order streams, in low-elevation (< 500 m) and low-slope (< 5%) watersheds that have small to moderate upstream watershed areas (10–1000 km2) with a low risk of habitat degradation. Many of the BAR-studied removals also have these characteristics, suggesting that our understanding of responses to dam removals is based on a limited range of landscape settings, which limits predictive capacity in other environmental settings. Biophysical responses to dam removal varied by landscape cluster, indicating that landscape features are likely to affect biophysical responses to dam removal. However, biophysical data were not equally distributed across variables or clusters, making it difficult to determine which landscape features have the strongest effect on dam-removal response. To address the inconsistencies across dam-removal studies, we provide suggestions for prioritizing and standardizing data collection associated with dam removal activities.

About the Authors:

Melissa M. Foley

Roles Conceptualization, Formal analysis, Methodology, Writing – original draft, Writing – review & editing

* E-mail: [email protected]

Current address: Institute of Marine Sciences, University of California Santa Cruz, Santa Cruz, California, United States of America

Affiliation: Pacific Coastal and Marine Science Center, United States Geological Survey, Santa Cruz, California, United States of America

ORCID http://orcid.org/0000-0002-5832-6404

Francis J. Magilligan

Roles Conceptualization, Methodology, Writing – original draft, Writing – review & editing

Affiliation: Department of Geography, Dartmouth College, Hanover, New Hampshire, United States of America

Christian E. Torgersen

Roles Conceptualization, Formal analysis, Writing – original draft, Writing – review & editing

Affiliation: Forest and Rangeland Ecosystem Science Center, United States Geological Survey, Seattle, Washington, United States of America

Jon J. Major

Roles Conceptualization, Methodology, Writing – original draft, Writing – review & editing

Affiliation: Cascades Volcano Observatory, Volcano Science Center, United States Geological Survey, Vancouver, Washington, United States of America

Chauncey W. Anderson

Roles Conceptualization, Methodology, Writing – original draft, Writing – review & editing

Affiliation: Oregon Water Science Center, United States Geological Survey Portland, Oregon, United States of America

Patrick J. Connolly

Roles Conceptualization, Methodology, Writing – original draft, Writing – review & editing

Affiliation: Columbia River Research Laboratory, Western Fisheries Research Center, United States Geological Survey, Cook, Washington, United States of America

Daniel Wieferich

Roles Data curation, Formal analysis, Writing – original draft, Writing – review & editing

Affiliation: Denver Federal Center, United States Geological Survey, Lakewood, Colorado United States of America

Patrick B. Shafroth

Roles Methodology, Writing – original draft, Writing – review & editing

Affiliation: Fort Collins Science Center, United States Geological Survey, Fort Collins, Colorado, United States of America

James E. Evans

Roles Methodology, Writing – original draft, Writing – review & editing

Affiliation: Department of Geology, Bowling Green State University, Bowling Green, Ohio, United States of America

Dana Infante

Roles Data curation, Writing – original draft, Writing – review & editing

Affiliation: Department of Fisheries and Wildlife, Michigan State University, East Lansing, Michigan, United States of America

Laura S. Craig

Roles Data curation, Writing – original draft, Writing – review & editing

Affiliation: American Rivers, Washington, D.C., United States of America

Introduction

Dams have been a fundamental part of the U.S. national agenda and economic-development ideology over the past two hundred years because of their essential role in flood control, municipal water supply, power generation, and irrigation. In the past several decades, however, there has been a paradigm shift in dam and watershed management—driven by environmental, economic, and engineering concerns—leading to the removal of obsolete, unsafe, and economically non-viable dams emerging as a significant management and restoration strategy. This new agenda has led to the removal of > 1,000 dams in the past few decades [1, 2], yet scientific assessment of the effects of dam removal lags the rate of removal [2], a theme typical of other river restoration efforts nationally [3–5]. Though the scientific community has been studying various aspects of dam removal in limited capacity for the last few decades, there is still a need to provide resource managers with basic information about the likely effects of dam removal that could affect the cost, planning process, permit requirements, and monitoring components of dam removal projects. Moreover, dam removal studies have not been conducted across a broad enough range of landscapes to establish a predictive framework linking the context of the dam location to anticipated outcomes affecting river hydrology, channel morphology, sediment budgets, water quality, and ecological trajectories.

General lessons regarding river response to dam removal, however, are slowly emerging. These lessons can help identify fundamental operative processes and biophysical responses to dam removal and further enlighten management decisions [6–8]. Multiple factors drive the variability in geomorphic responses to dam removal, including dam size; removal method; reservoir size and shape; sediment volume, cohesiveness, and grain size; and released sediment volume relative to background sediment flux [1, 6, 9, 10]. Contrary to some perceptions, Major et al. [11] and East et al. [12] found that river channels can stabilize relatively quickly after dam removal—within months or years, not decades—approaching pre-dam emplacement morphology.

Ecological response trajectories after dam removal are difficult to generalize because response rates can be highly variable across taxa [13] and can be affected by past and current conditions [14–16]. In addition, most dam removal studies are short in duration and focus on a single response metric [2]. Despite these limitations, some patterns have emerged from the literature: there may be a lag between geomorphic and ecological responses [17, but see 18]; aquatic species typical of flowing rivers (lotic habitats) tend to replace stillwater (lentic) communities in the reservoir after dam removal [15]; and upstream fish migration that was formerly impeded by the dam may occur swiftly after dam removal in some cases [19–22].

Biophysical river responses to dam removal are affected by the surrounding landscape, but these effects are poorly understood [15, 23] because the literature consists mainly of specific case studies focused on short-term responses with limited comparison across regions. Without understanding a site’s landscape context (i.e., location within a watershed or regional and local patterns of climate, geology, and vegetation), it is difficult to interpret the broad applicability or local limitations of the biophysical responses to dam removal [24]. As a result, our fundamental understanding of long-term trajectories and broad-scale patterns of ecological, geomorphic, and hydrologic responses to dam removal is lacking. Furthermore, the breadth (or lack thereof) of published studies is directly tied to the expertise of researchers in each case study; truly interdisciplinary studies are rarely conducted for the same dam removal [2], and studies that integrate the trajectories of physical and biological responses are even more rare [25, but see 26].

To assess the state of science for understanding outcomes to dam removal and to determine the representativeness of dam removal case studies compared to the national dam population, we conducted a meta-analysis examining the landscape context—including natural and anthropogenic factors—of more than 50,000 existing dams and nearly 900 removed dams in the conterminous U.S. We also reviewed 104 published studies [27] with before- and after-removal (BAR) data from 63 dam removals to analyze the influence of landscape context in driving the biophysical response to dam removal. Because removals have occurred in settings where the human footprint may influence the response trajectory, we characterized landscape context as a combination of natural (e.g., ecoregion, watershed size) and anthropogenic (e.g., population density, transportation infrastructure) attributes. We used this approach to assess the landscape context of dams and removed dams; examine the biophysical response of river systems in different landscape settings; and highlight landscape contexts where additional research is needed. In doing so, we attempt to unpack “environmental context” into more specifically defined statistical associations but with the full knowledge that we are working with limited and geographically biased data. Finally, propose approaches for standardizing elements of dam-removal research that could increase our understanding of biophysical responses and help guide watershed management and restoration efforts. This type of comprehensive review has not been reported and our study is the first to formally examine the landscape context of dam removals with linked geospatial data at a national scale in the U.S. We also had access to a unique dam removal database compiled by American Rivers that allowed us to examine the geographic context for a larger population of dam removals than has been previously publicly available.

Methods

We compiled geographic information for 50,772 existing dams listed in the National Anthropogenic Barrier Dataset (NABD) [28], a subset of dams from the 2009 National Inventory of Dams (U.S. Army Corps of Engineers– http://nid.usace.army.mil/cm_apex/f?p=838:12, accessed July 2010). We gathered the same information for the 874 removed dams included in the USGS Dam Removal Information Portal (DRIP– https://www.sciencebase.gov/drip/; accessed 1 July 2016). All existing and removed dams were linked to the National Hydrography Dataset Plus Version 1 (NHDPlusV1), allowing us to gather additional information from the National Fish Habitat Partnership’s (NFHP) 2015 National Assessment of Fish Habitat Condition Database [29, 30] and Anthropogenic Disturbance Database [31], as well as land-cover data summaries from the National Land Cover Database (NLCD– http://www.mrlc.gov/). From these sources, we identified natural and anthropogenic landscape-context factors in the river segment for each existing and removed dam (Table 1). We also examined the distribution of dams and dam removals in relation to Environmental Protection Agency (EPA) Level III Ecoregions, which characterize nation-wide landscape characteristics based on geology, landforms, soils, vegetation, climate, land use, wildlife, and hydrology [32] (https://www.epa.gov/eco-research/ecoregions; accessed 30 January 2017).

[Figure omitted. See PDF.]

Table 1. Landscape variables.

https://doi.org/10.1371/journal.pone.0180107.t001

We generated summary statistics using these landscape characteristics for existing and removed dams to determine how representative removals have been of the overall dam population. We reduced the number of individual landscape factors used in our analyses because some of the variables were used to derive a habitat condition index (HCI) [29, 30], an index based on regionally specific responses of stream fishes to anthropogenic landscape factors. We excluded the variables used to create that index—including land cover, population density, road density, and the number of dams, mines, and point-source pollution sites—from subsequent analyses to avoid overrepresentation. The HCI uses a ranking of 1 through 5, with low scores corresponding to low risk of fish-habitat degradation and high scores to high risk of fish-habitat degradation.

We used seven variables—mean watershed slope, elevation, and area; precipitation; air temperature; ground water input; and HCI—in a Principal Components Analysis (PCA) to determine which landscape variables best explained the variability in landscape context among dam removals. Because landscape variables were measured using a variety of units, all variables were normalized prior to analysis (for each variable, the variable mean was subtracted from the value and then divided by the standard deviation). We also conducted a cluster analysis (using a resemblance matrix based on Euclidean distance) to determine whether dam removals formed significantly distinct clusters based on landscape context. We used similarity profile analysis (SIMPROF) to assign groupings for all dam removals and a subset of dam removals with BAR studies that had statistically different (p ≤ 0.01) landscape characteristics [33].

To determine if biophysical response to dam removal varied with landscape context, we selected a subset of dam removals from the DRIP database that had BAR data upstream of the reservoir, within the reservoir, and/or downstream of the dam site. When we accessed the database (November 2016), it contained information from 104 BAR studies from 63 dam removals (S1 Table). For each study, we classified the response to dam removal categorically for each biophysical variable as “increased,” “no change,” or “decreased.” We did not control for time frame of response (e.g., weeks to years after dam removal) following dam removal because the duration of studies was highly variable. For each variable (e.g., turbidity), we tallied the number of each response type, irrespective of methodological differences. We recognize that this way of analyzing the data resulted in a loss of resolution and somewhat limits our ability to compare across dam removals, but vagaries among studies required a level of generalization to assemble data coherently.

Based on the clusters determined in our SIMPROF analysis, we examined biophysical responses to dam removal within each geographic cluster having more than three dams and qualitatively compared the responses across geographic clusters. We could not conduct robust quantitative analyses on landscape context and biophysical responses because we were limited by the number of removals within each cluster, as well as by the overlap of data types among studies (S1 Table). We used PRIMER (v. 7, Primer Ltd.) and QGIS (v. 2.8.1) for all our analyses [34, 35].

Results

The densities of existing and removed dams, and studied dam removals varied greatly across the U.S. (Fig 1) [2]. The highest concentrations of existing dams were in the Southeastern Plains, Central Great Plains, Piedmont, and Northwestern Great Plains EPA Level III Ecoregions (Fig 2, S2 Table), predominantly on headwater streams (stream order = 1) that had small upstream watershed areas (< 10 km2) and low mean catchment slope (< 5 degrees) (Fig 3A–3C). Many were also located in areas where the HCI of the upstream catchment was very high, indicating an anthropogenic stressor(s) causes significant fish habitat degradation in those areas (Fig 3D). In contrast, the highest concentrations of dam removals have occurred in the Ridge and Valley, Northern Piedmont, Northeastern Highlands, and Northeastern Coastal Zone Ecoregions (Fig 2, S2 Table). Unlike existing dams, dam removals have occurred in a range of stream sizes, with a nearly equal number coming out of stream orders 1–4; and in river systems where the upstream catchment area is large, up to two orders of magnitude greater than that of existing dams (Fig 3B). Similar to existing dams, removed dams were located in watersheds with a low mean catchment slope (Fig 3C). Nearly 40% of removed dams were in watersheds with a low or very low risk of upstream habitat degradation (i.e., very low or low HCI score), and just over 30% of existing dams that were located in areas with a very high risk of habitat degradation (Fig 3D).

[Figure omitted. See PDF.]

Fig 1. Geography of dams.

U.S. distribution of (a) existing dams listed in the National Anthropogenic Barrier Dataset (n = 50,772); (b) removed dams from the Dam Removal Inventory Project (n = 874); (c) removed dams with before-after studies (n = 63).

https://doi.org/10.1371/journal.pone.0180107.g001

[Figure omitted. See PDF.]

Fig 2. Ecoregions.

EPA Level III Ecoregions for existing and removed dams in the U.S., and before-after-removal studies.

https://doi.org/10.1371/journal.pone.0180107.g002

[Figure omitted. See PDF.]

Fig 3. Landscape context.

Comparison of (a) stream order; (b) watershed area (km2); (c) watershed slope (degrees); and (d) habitat condition index (risk of degradation) for existing (n = 50,772) and removed dams (n = 874), and dam removals with before- and after-removal studies (n = 63).

https://doi.org/10.1371/journal.pone.0180107.g003

The landscape context of dam removals with BAR studies also differed in many respects from that of existing or removed dams (Figs 2 and 3). BAR studies were most numerous in the Eastern Corn Belt Plains (Ohio), Driftless Area (southern Wisconsin), Cascades (western U.S.), Southeastern Wisconsin Till Plains, and Piedmont Ecoregions (Fig 2, S2 Table). Except for the studies from the Eastern Corn Belt Plains and Cascades Ecoregions, these studies represented fewer than 13% of the removals in those areas (S2 Table). Studied dam removals have occurred predominantly on larger streams and in watersheds with low mean catchment slope and moderate to low risk of habitat degradation (HCI > 3) (Fig 3D).

We identified several notable spatial patterns in the landscape context of dam removals with respect to their distribution throughout the U.S. (Fig 4). The clusters of removals in the upper Midwest and upper New England generally occurred in large, low elevation, low slope watersheds, many with degraded fish habitat (Fig 4A–4D). In contrast, removals in the western U.S. were in high elevation, steep, small watersheds with predominantly moderate- to low-risk of habitat degradation (Fig 4A–4D). We identified a dearth of dam removals in central and south-central areas of the continental U.S., despite this area having one of the highest concentrations of existing dams (i.e., Central Great Plains).

[Figure omitted. See PDF.]

Fig 4. Spatial distribution of landscape characteristics.

Spatial distribution of landscape characteristics for all removed dams: (a) upstream watershed elevation (m), (b) upstream watershed area (km2), (c) upstream watershed slope (degrees), and (d) habitat condition index (risk of habitat degradation).

https://doi.org/10.1371/journal.pone.0180107.g004

For the seven variables we analyzed for all dam removals, the cluster analysis revealed 57 unique clusters of dam removals based on their landscape characteristics (Fig 5A). Although many of these clusters were concentrated in specific geographic regions, some removals that had similar landscape characteristics were widely distributed across the U.S. (Fig 6A). The PCA suggested the main factors differentiating the dam removal clusters were watershed elevation, groundwater input, and air temperature on principal component axis 1 (PC1); watershed area, watershed slope, and precipitation on PC2; and groundwater input, watershed elevation, and watershed slope on PC3 (Table 2); these three PCA axes explained 66% of the variation in landscape characteristics. Studied dam removals were represented in 36 of the 57 geographic clusters (Fig 5B); and BAR studies were conducted in 32 of the 57 geographic clusters (Fig 5C). Although over half of the geographic clusters had at least one BAR study, clusters in the lower right quadrant of the PCA axes had the greatest number of studies, representing large, low-elevation, and low-slope watersheds. Few BAR studies were conducted in small, high-elevation, high-slope watersheds (Fig 5C).

[Figure omitted. See PDF.]

Fig 5. Principal Components Analysis results.

Principal Components Analysis results for (a) all dam removals (57 clusters); (b) all studied dam removals (36 clusters); and (c) before-after studied dam removals (32 clusters). The number of clusters in (a) was determined from a cluster analysis; clusters in (b) and (c) show how many original clusters were represented in those subsets.

https://doi.org/10.1371/journal.pone.0180107.g005

[Figure omitted. See PDF.]

Fig 6. Spatial distribution of landscape clusters.

(a) Spatial distribution of clusters based on landscape characteristics for all dam removals. Upper Midwest and Northeast sections magnified to show details. (b) Spatial distribution of clusters based on landscape characteristics of before-after studied dam removals.

https://doi.org/10.1371/journal.pone.0180107.g006

[Figure omitted. See PDF.]

Table 2. Principal component loadings.

https://doi.org/10.1371/journal.pone.0180107.t002

We conducted a separate cluster analysis based only on the landscape variables for BAR studies. This reduced analysis revealed eight significantly distinct clusters of removals, some with a wide geographic distribution (Fig 6B). Seven removals could not be categorized because data were not available for all landscape variables (Table 3). We used these new clusters to look at patterns of biophysical responses to dam removal.

[Figure omitted. See PDF.]

Table 3. Landscape clusters for before and after-removal studies.

https://doi.org/10.1371/journal.pone.0180107.t003

Parameters reported most frequently across all BAR studies included sediment grain size, water temperature, aquatic invertebrates, and fish (Table 4). However, not all of these parameters were reported above the dam, within the reservoir, or downstream of the dam. For nearly all of the biophysical parameters we characterized, measurements were most frequently reported downstream of the removed dam (Fig 7). For example, approximately half of the BAR studies reported grain size measurements downstream of the former dam, while only 33% reported measurements within reservoir reaches and 25% in upstream reaches. Biotic variables were more consistently reported for all three river reaches than physical variables. Some variables were quantified using different metrics, particularly nutrients, aquatic invertebrates, and fish (Table 5).

[Figure omitted. See PDF.]

Fig 7. Before- and after-removal biophysical study parameters.

Number of dam removals with upstream, reservoir, and downstream studies that reported the before- and after-removal responses of physical, water quality, and biological parameters.

https://doi.org/10.1371/journal.pone.0180107.g007

[Figure omitted. See PDF.]

Table 4. Reported biophysical metrics.

https://doi.org/10.1371/journal.pone.0180107.t004

[Figure omitted. See PDF.]

Table 5. Measurement metrics.

https://doi.org/10.1371/journal.pone.0180107.t005

We could not formally test for differences in biophysical responses to dam removal because variables were not consistently reported across all dam removals in the eight geographic clusters (Fig 8). Water quality variables, including phosphate, nitrate, water temperature, and dissolved oxygen, were measured in only three of eight geographic clusters. In contrast, sediment grain size and fish species data were reported in all clusters (Fig 8). Physical responses to dam removal tended to be more consistent across geographic clusters than either water quality or ecological parameters. Sediment grain size tended to remain mostly unchanged in upstream reaches, coarsened in reservoir reaches, and fined downstream after dam removal; turbidity did not change upstream or in the reservoir reach but increased downstream; and suspended-sediment concentration increased downstream after dam removal. On the basis of limited measurements, water quality responses varied by geographic cluster, but many locations showed no change in water quality parameters in any of the three river reaches. In the Upper Midwest and New England clusters, however, phosphorus concentration (inclusive of all reported phosphate metrics listed in Table 5) increased downstream after dam removal, while nitrate concentration increased in the reservoir reach and downstream after dam removal in the Southeast cluster. Water temperature did not change in any river section after dam removal in a majority of studies, but after three removals—including two large removals in the Pacific Northwest—water temperature decreased downstream of the removed dams. A decrease in water temperature in the reservoir reach after dam removal was observed in only two of ten studies (Fig 8), yet a decrease in water temperature in the reservoir reach is assumed to be a typical response following dam removal [36].

[Figure omitted. See PDF.]

Fig 8. Biophysical responses to dam removal.

Biophysical response for all before- and after-removal studies (top row) and within each distinct geographic cluster.

https://doi.org/10.1371/journal.pone.0180107.g008

Biological responses to dam removal were more variable than physical and water quality responses, particularly downstream of the dam. Aquatic invertebrate and fish (single species or community) responses varied among geographic clusters, and were also highly variable within a single geographic cluster (Fig 8). This variability was particularly evident at sites downstream of the dam in the two geographic clusters with the highest number of biological BAR studies, the Upper Midwest and New England, where there were nearly equal numbers of studies showing an increase, decrease, or no change in aquatic invertebrate and fish responses (Fig 8). In contrast, a majority of the BAR studies from reservoir and upstream reaches reported an increase in aquatic invertebrates or fish. BAR studies reporting the response of fish community composition were entirely absent from dam removals in the West and Pacific Northwest clusters—the clusters containing some of the largest dam removals.

Discussion

Dam removals have occurred throughout the United States but have been concentrated in watersheds that represent a relatively narrow range of landscape characteristics compared to the characteristics of the existing dams throughout the U.S. Most dam removals have occurred in low-elevation watersheds with low catchment slope and large upstream areas, and most BAR studies were conducted in watersheds with similar characteristics. Watersheds in wet climates (high precipitation) with steep slopes, high mean elevations, and good fish habitat conditions (low chance of degradation) were poorly represented in BAR studies. Apparent geographic discrepancies between existing-dam density and removed dams may be due to factors related to economics, historical context, and dam function (e.g., irrigation, flood control, hydropower), but that information is rarely reported, and the discussion of those factors is beyond the scope of our analyses.

Biophysical responses to dam removal varied by geographic region, and not all biophysical variables were consistently reported after dam removals. Inconsistencies in the metrics reported, measurement timing, and study duration made it difficult to quantitatively assess biophysical responses in geographic regions with different landscape characteristics and predict how a system might respond to a dam removal based on its landscape context. We identified distinct differences in landscape context among existing and removed dams, and BAR studies. Our analysis comparing existing and removed dams, however, was limited to dams that were either ≥ 8 m tall with an 18,500-m3 or larger impoundment or ≥ 2 m tall with an impoundment at least 62,000 m3. Many removed dams did not meet those height or impoundment size requirements; of the 874 removed dams included in our analysis, only 165 of them met the criteria for being listed in the NID. Some states have more comprehensive inventories of existing dams, including small dams, but the NABD—which draws data from the NID—is the only publicly available list of existing dams throughout the country that is spatially linked to the NHDPlusV1.

Fish habitat condition index scores for existing dams were nearly the opposite of HCI scores for removed dams. Very low-quality fish habitat with high risk of degradation characterized many landscapes around existing dams, but dam removals have occurred in landscapes with moderate to high quality fish habitat with moderate to low risk of degradation. Dam removals may have been more common in areas with high quality fish habitat to enhance the probability of a successful outcome, particularly if ecosystem restoration was a goal of the dam removal. The HCI was calculated based on landscape characteristics of the watershed above a dam, and is an important metric to consider when planning dam removals because habitat quality within the watershed influences the biophysical responses to dam removal and the potential for habitat condition improvements [37]. River ecosystems may be more likely to recover to pre-dam conditions if dam emplacement is a main source of anthropogenic stressors on the landscape, contributing to an increased risk of habitat degradation (Table 6).

[Figure omitted. See PDF.]

Table 6. Anthropogenic landscape context.

https://doi.org/10.1371/journal.pone.0180107.t006

Watershed area, elevation, and precipitation were the dominant landscape variables separating the geographical clusters of removed dams. Of the 57 unique clusters identified in the cluster analysis, 43 were concentrated in the lower-right quadrant of the PC plot, representing large-area, low-elevation, and low-precipitation watersheds (Fig 5A). This pattern held for all studied dam removals (Fig 5B) and BAR studies (Fig 5C). This predominant clustering of removals suggests that our frame of reference for understanding the biophysical response to dam removal is quite limited. This was especially true for BAR studies, which rarely examined physical and biological responses.

For the eight clusters of dam removals with BAR data, we were unable to quantitatively compare the biophysical responses with respect to landscape context because reported variables were neither consistent nor standardized. With the exception of BAR studies in the New England cluster, each cluster had missing data for at least one of the ten metrics examined, and many variables were only reported in one dam removal in the cluster. For dam removals with landscape characteristics outside the main groupings, whole classes of response variables were missing, including water quality and fish response data (Fig 8).

A number of factors likely contributed to the variation in biophysical response among geographic clusters. Firstly, the metrics used to measure responses varied among dam removals. For example, in the papers we reviewed, investigators used five different metrics to measure changes in phosphorous concentration and six metrics to measure changes in aquatic invertebrates (Table 5). Secondly, metrics were reported over different temporal scales [2], thus potentially obscuring differences between short-term and long-term changes after dam removal. For instance, after the removal of the Boulder Creek dams in South Carolina, soluble reactive phosphorus concentration increased within hours [38], but concentrations decreased after two weeks. As a result, there was no significant long-term change in phosphate concentration before and after dam removal. Similarly, dissolved phosphate concentration increased when the Good Hope Mill Dam in Pennsylvania was breached, but it returned to pre-removal levels within hours [39]. Aquatic invertebrate and fish responses to dam removal, particularly downstream, were strongly dependent on study timing [40]. Studies conducted immediately after dam removal commonly showed a decrease in invertebrate and fish metrics downstream of a dam removal, especially if sediment grain size changed [41, 42]. In contrast, studies that were conducted after the initial pulse of sediment moved through the system following dam removal showed either no change or a positive effect of dam removal [43, 44]. Finally, the reported biological response to dam removal can be influenced by the species monitored. For instance, in our analysis, some aquatic invertebrate studies reported responses of species that were present before dam removal [45, 46] and others reported the responses of species that were expected to colonize after dam removal (i.e., EPT taxa) [47–49]. In the first case, species tended to decrease after dam removal, while in the latter case they tended to increase. Other studies reported species diversity or richness [40, 50–53], which can be difficult to interpret without species-specific information because those metrics may not change if an equal number of pre-removal species are replaced with post-removal species.

Some of the biophysical responses we found in the literature were unexpected. For instance, water temperature decreased in reservoir and downstream reaches in only 10% of studies we examined [47, 54, 55] and nutrients increased downstream in only 30% of studies [39, 45, 46, 48, 56–58]. We expected these percentages to be much higher based on assumptions of biophysical response to dam removal [15, 17, 59, 60]. Individual fish species and fish communities upstream of a dam removal did not change in 40% of dam removals examined, nor did they change in more than 25% of downstream sites examined. The scientific community needs more data to understand ecosystem response in order to inform management decisions and create realistic expectations for post-removal recovery.

Our analysis did not consider all of the possible variables that could contribute to biophysical response to dam removal, and we recognize that landscape context does not affect all responses. For example, at Fossil Creek Dam, Arizona, native fish density increased after flow was restored to the downstream system, but only in stream segments where invasive species had been removed [61].

To date, most dams have been removed for economic, safety, or liability reasons rather than to restore ecosystem function [62]. As with many restoration efforts, removal rationales are not always available, nor is there a singularly agreed upon reason for removing a dam. Therefore, the objectives (or lack thereof) for removing dams may affect the types of variables monitored before and after dam removal. However, every dam removal, no matter the rationale, is an opportunity to gain further insight into how ecosystems respond and how physical and biological responses are connected. Recognizing the need for studies to remain focused on their objectives, we suggest that variables sampled before and after dam removal be prioritized and protocols developed in an attempt to coordinate and standardize the type of data that are collected. Many studies on dam removals were not comparable because of differences in metrics measured, methodologies employed, and study interval, including whether or not pre- and post-removal data were collected. Standardization of the type, frequency, and duration of data collection can help the scientific community better understand how the responses of river ecosystems vary as a function of landscape context. The following approaches provide examples of potential standardized procedures for evaluating biophysical responses of river systems to dam removal:

1. Sample before and after dam removal and at temporal and spatial scales that are meaningful for the metrics sampled and the magnitude of anticipated change.

2. Sample upstream, within the reservoir, and downstream of the proposed dam removal site. Studies that focus solely on the downstream response to dam removal do not provide a comprehensive view of biophysical response.

3. Sample a broad range of metrics for comparative purposes. If that is not possible, prioritize response measurements for indicator species that have known relationships to other variables.

4. Use technology and citizen science to expand the duration of the sampling or monitoring program. Satellite images are becoming increasingly available (e.g., Digital Globe) and can be used to assess landform and vegetation changes. Enlist citizen scientists to take spatially aligned repeat photographs, measure stream temperature, or record other parameters from fixed locations.

5. Compile, preserve and publically release data. Add data to public databases, including the Dam Removal Information Portal (https://www.sciencebase.gov/drip/).

Conclusion

Dam removal has become an increasingly common restoration strategy, and new efforts are underway to help prioritize removals and guide removal decision-making [63]. Management decisions to remove dams are beginning to adopt a range of strategies, from a “hot spot” approach [64] to more overt economic strategies for prioritizing barrier removal [65, 66], to those aimed at achieving broader ecological gains [67, 68]. Our results indicate that landscape context may inform possible biophysical responses to removal, but a broader geographic range of removals is required. Thus, along with other management priorities, decisions about dam removal might consider where the proposed removal is located and how its removal can help advance our understanding of biophysical responses of river systems. Dam removals are large-scale experiments that offer tremendous opportunities to understand fluvial systems and the influence of humans on watershed processes and ecosystem dynamics. Knowledge of biophysical responses to dam removal in a regional context can be leveraged to anticipate the effects of pending dam removals and to help coordinate management efforts to meet conservation and restoration goals.

Supporting information

[Figure omitted. See PDF.]

S1 Table. Before- and after-removal studies.

Before-after-removal studies used in our statistical analysis; biophysical parameters measured in each study are indicated with grey shading.

https://doi.org/10.1371/journal.pone.0180107.s001

(DOCX)

S2 Table. EPA Level I, II, and III Ecoregions.

EPA Level I, II, and III Ecoregion classifications (https://www.epa.gov/eco-research/ecoregions-north-america) for dams listed in the National Anthropogenic Barrier Dataset (NABD), removed dams in the USGS Dam Removal Information Portal (DRIP), and removed dams with before- and after-removal studies (BAR).

https://doi.org/10.1371/journal.pone.0180107.s002

(DOCX)

Acknowledgments

We gratefully acknowledge funding from the U.S. Geological Survey’s John Wesley Powell Center for Analysis and Synthesis, which supported our efforts to synthesize dam removal science. In particular, we thank Jill Baron and Leah Colasuonno at the Powell Center for their help and encouragement. We also thank Amy East and one anonymous reviewer for their insightful comments. Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government. Data used in this synthesis are publicly available and are listed in the references.

Citation: Foley MM, Magilligan FJ, Torgersen CE, Major JJ, Anderson CW, Connolly PJ, et al. (2017) Landscape context and the biophysical response of rivers to dam removal in the United States. PLoS ONE 12(7): e0180107. https://doi.org/10.1371/journal.pone.0180107

References

1. O'Connor JE, Duda JJ, Grant GE. 1000 dams down and counting. Science. 2015;348(6234):496–7. pmid:25931537

2. Bellmore JR, Duda JJ, Craig LS, Greene SL, Torgersen CE, Collins MJ, et al. Status and trends of dam removal in the United States. Wiley Interdisciplinary Reviews: Water. 2017;

3. Bernhardt ES, Palmer MA, Allan JD, Alexander G, Barnas K, Brooks S, et al. Synthesizing U.S. River Restoration Efforts. Science. 2005;308(5722):636–7. pmid:15860611

4. Bernhardt ES, Sudduth EB, Palmer MA, Allan JD, Meyer JL, Alexander G, et al. Restoring rivers one reach at a time: results from a survey of US river restoration practitioners. Restoration Ecology. 2007;15(3):482–93.

5. Palmer M, Bernhardt E, Allan J, Lake P, Alexander G, Brooks S, et al. Standards for ecologically successful river restoration. Journal of applied ecology. 2005;42(2):208–17.

6. Major JJ, East AE, O'Connor JE, Grant GE, Wilcox AC, Magirl CS, et al. Geomorphic responses to dam removal in teh United States—a two-decade perspective. In: Tsutsumi D, Laronne J, editors. Gravel Bed Rivers and Disasters: Wiley and Sons; 2017. p. 355–83.

7. Pess G, Bellmore JR, Duda JJ, O'Connor JE, East AE, Foley MM, et al. Ecosystem response to dam removal: a synthesis. BioScience. In review.

8. Foley MM, Bellmore JR, O'Connor JE, Duda JJ, East AE, Grant GE, et al. Dam removal—Listening in. Water Resources Research. In review.

9. Grant GE, Lewis SL. The remains of the dam: what have we learned from 15 years of US dam removals? In: Lollino G, Arattano M, Rinaldi M, Giustolisi O, Marechal J-C, Grant GE, editors. Engineering Geology for Society and Territory. 3: Springer; 2015. p. 31–6.

10. Sawaske SR, Freyberg DL. A comparison of past small dam removals in highly sediment-impacted systems in the US. Geomorphology. 2012;151:50–8.

11. Major JJ, O’Connor JE, Podolak CJ, Keith MK, Grant GE, Spicer KR, et al. Geomorphic response of the Sandy River, Oregon, to removal of Marmot Dam: US Geological Survey; 2012.

12. East AE, Pess GR, Bountry JA, Magirl CS, Ritchie AC, Logan JB, et al. Large-scale dam removal on the Elwha River, Washington, USA: River channel and floodplain geomorphic change. Geomorphology. 2015;228:765–86.

13. Hansen JF, Hayes DB. Long‐term implications of dam removal for macroinvertebrate communities in Michigan and Wisconsin rivers, United States. River Research and Applications. 2012;28(9):1540–50.

14. Shafroth PB, Friedman JM, Auble GT, Scott ML, Braatne JH. Potential responses of riparian vegetation to dam removal. BioScience. 2002;52(8):703–12.

15. Stanley EH, Doyle MW. Trading off: the ecological effects of dam removal. Frontiers in Ecology and the Environment. 2003;1(1):15–22.

16. Tullos DD, Collins MJ, Bellmore JR, Bountry JA, Connolly PJ, Shafroth PB, et al. Synthesis of common management concerns associated with dam removal. JAWRA Journal of the American Water Resources Association. 2016;52(5):1179–206.

17. Doyle MW, Stanley EH, Orr CH, Selle AR, Sethi SA, Harbor JM. Stream ecosystem response to small dam removal: lessons from the Heartland. Geomorphology. 2005;71(1):227–44.

18. Tullos DD, Finn DS, Walter C. Geomorphic and ecological disturbance and recovery from two small dams and their removal. PLoS One. 2014;9(9):e108091. pmid:25233231

19. Hitt NP, Eyler S, Wofford JE. Dam removal increases American eel abundance in distant headwater streams. Transactions of the American Fisheries Society. 2012;141(5):1171–9.

20. Hogg RS, Coghlan SM Jr, Zydlewski J, Gardner C. Fish community response to a small-stream dam removal in a maine coastal river tributary. Transactions of the American Fisheries Society. 2015;144(3):467–79.

21. McHenry ML, Pess G, Moses R, Brenkman S, Crain P, Anderson J. Spawning distribution of Chinook salmon in the Elwha River, Washington State during dam removal from 2012 to 2014. Lower Elwha Klallam Tribe, Port Angeles, WA: 2015.

22. Jolley JC, Silver GS, Whitesel TA. Occurrence, detection, and habitat use of larval Lamprey in the Lower White Salmon River and mouth: Post-Condit Dam removal. FY 2012 Annual Report. U.S. Fish and Wildlife Service: 2013.

23. Church M. Geomorphic response to river flow regulation: Case studies and time‐scales. River Research and Applications. 1995;11(1):3–22.

24. Swanson FJ, Sparks RE. Long-term ecological research and the invisible place. BioScience. 1990;40(7):502–8.

25. Palmer MA. Reforming Watershed Restoration: Science in Need of Application and Applications in Need of Science. Estuaries and Coasts. 2009;32(1):1–17.

26. Foley MM, Warrick JA, Ritchie A, Stevens AW, Shafroth PB, Duda JJ, et al. Coastal habitat and biological community response to dam removal on the Elwha River Ecological Monographs. In press.

27. Wieferich DJ, Bristol RS, Bellmore JR, Vittum KM, Duda JJ, Craig L. National Dam Removal Science Database. U.S. Geological Survey Data Release 2016.

28. Ostroff S, Wieferich DJ, Cooper A, Infante D. 2012 National Anthropogenic Barrier Dataset (NABD). Denver, CO: U.S. Geological Survey—Aquatic GAP Program; 2013.

29. Crawford S, Whelan G, Infante DM, Blackhart K, Daniel WM, Fuller PL, et al. Through a Fish's Eye: The Status of Fish Habitats in the United States 2015. National Fish Habitat Partnership. Accessed on 10 October 2016. http://assessment.fishhabitat.org/: 2016.

30. Daniel WM, Infante DM, Herreman K, Cooper A, Ross J. National Fish Habitat Partnership (NFHP) 2015 Cumulative Habitat Condition Indices and Limiting Disturbances for the Conterminous United States linked to NHDPlusV1, https://doi.org/10.5066/F73R0R1P. 2017.

31. Herreman K, Cooper A, Infante DM, Daniel WM. National Fish Habitat Partnership (NFHP) 2015 Anthropogenic Disturbance Data for the Conterminous United States linked to the NHDPLUSV1. U.S. Geological Survey data release, https://doi.org/10.5066/F7000086;2017.

32. Omernik JM. Ecoregions of the conterminous United States. Annals of the Association of American geographers. 1987;77(1):118–25.

33. Somerfield P, Burton M, Sanderson W. Analyses of sublittoral macrobenthic community change in a marine nature reserve using similarity profiles (SIMPROF). Marine environmental research. 2014;102:51–8. pmid:25012227

34. QGIS. QGIS Geographic Information System. http://www.qgis.org/: Open Source Geospatial Foundation Project; 2016.

35. Clarke K, Gorley R. PRIMER v7: User Manual/Tutorial. PRIMER-E, Plymouth, UK2015.

36. Bednarek AT. Undamming rivers: a review of the ecological impacts of dam removal. Environmental Management. 2001;27(6):803–14. pmid:11393315

37. Ward JV, Stanford J. The serial discontinuity concept of lotic ecosystems. Dynamics of Lotic Ecosystems. 1983;10:29–42.

38. Orr CH, Rogers KL, Stanley EH. Channel morphology and P uptake following removal of a small dam. Journal of the North American Benthological Society. 2006;25(3):556–68.

39. Chaplin JJ, Brightbill RA, Bilger MD. Effects of removing Good Hope Mill Dam on selected physical, chemical, and biological characteristics of Conodoguinet Creek, Cumberland County, Pennsylvania. Report. U.S. Geological Survey Scientific Investigations Report 2005–5226: 2005 2005–5226.

40. Maloney KO, Dodd HR, Butler SE, Wahl DH. Changes in macroinvertebrate and fish assemblages in a medium‐sized river following a breach of a low‐head dam. Freshwater Biology. 2008;53(5):1055–68.

41. Bulak J, Bettinger J, Kubach K, Leitner J, Marion C, Poly W, et al. Twelvemile Creek Dam removal monitoring: South Carolina. Department of Natural Resources: 2011.

42. Thomson JR, Hart DD, Charles DF, Nightengale TL, Winter DM. Effects of removal of a small dam on downstream macroinvertebrate and algal assemblages in a Pennsylvania stream. Journal of the North American Benthological Society. 2005;24(1):192–207.

43. Burdick SM, Hightower JE. Distribution of spawning activity by anadromous fishes in an Atlantic slope drainage after removal of a low-head dam. Transactions of the American Fisheries Society. 2006;135(5):1290–300.

44. Burroughs BA, Hayes DB, Klomp KD, Hansen JF, Mistak J. The effects of the Stronach Dam removal on fish in the Pine River, Manistee County, Michigan. Transactions of the American Fisheries Society. 2010;139(5):1595–613.

45. Harbold W, Stranko S, Kilian J, Ashton M, Graves P. Patapsco River Dam removal study: Assessing changes in American Eel distribution and aquatic communities. Maryland Department of Natural Resources: 2013.

46. Sherman M. Potential impacts of small dam removal on fish and mussel communities in North Carolina:: Duke University; 2013.

47. Howard G, Cusack M, Savidge T. Carbonton Dam—Deep River Watershed restoration site. Ecosystem Enhancement Program: 2007.

48. Gibson AM, Villalobos LM, Snyder EB. Ecosystem Responses to Low-Head Dam Removal: Assessment of Physical Habitat, Water Chemistry, and Macroinvertebrates. Grand Valley State University: 2011.

49. Claeson S, Coffin B. Physical and biological responses to an alternative removal strategy of a moderate‐sized dam in Washington, USA. River Research and Applications. 2016;32:1143–52.

50. Knittel MJ. The impact of a small dam removal on benthic macroinvertebrate Ccommunities of Mill Creek (Washtenaw Co., MI.): University of Michigan; 2010.

51. Hutchison B. Fish and macroinvertebrate assemblages following the removal of low-head dams in two Pennsylvania streams: Pennsylvania State University; 2008.

52. Pollard AI, Reed T. Benthic invertebrate assemblage change following dam removal in a Wisconsin stream. Hydrobiologia. 2004;513(1):51–8.

53. Whitener P. Twelve Mile Creek Channel Morphology, Substrate Composition, and the Macroinvertebrate Community Changes Resulting from Multiple Dam Removals: Clemson University; 2013.

54. Kornis MS, Weidel BC, Powers SM, Diebel MW, Cline TJ, Fox JM, et al. Fish community dynamics following dam removal in a fragmented agricultural stream. Aquatic Sciences. 2015;77(3):465–80.

55. Stewart GB. Patterns and processes of sediment transport following sediment-filled dam removal in gravel bed rivers: Oregon State University; 2006.

56. Catalano MJ. Evaluating fish-habitat relations, fish distribution, and effects of dam removal in the Baraboo River, Wisconsin: University of Wisconsin-Stevens Point; 2002.

57. Doyle MW, Stanley EH, Harbor JM. Channel adjustments following two dam removals in Wisconsin. Water Resources Research. 2003;39(1).

58. Ahearn DS, Dahlgren RA. Sediment and nutrient dynamics following a low-head dam removal at Murphy Creek, California. Limnology and Oceanography. 2005;50(6):1752–62.

59. Hart DD, Johnson TE, Bushaw-Newton KL, Horwitz RJ, Bednarek AT, Charles DF, et al. Dam removal: challenges and opportunities for ecological research and river restoration: we develop a risk assessment framework for understanding how potential responses to dam removal vary with dam and watershed characteristics, which can lead to more effective use of this restoration method. BioScience. 2002;52(8):669–82.

60. Stanley EH, Doyle MW. A Geomorphic Perspective on Nutrient Retention Following Dam Removal. BioScience. 2002;52(8):693–701.

61. Marks JC, Haden GA, O’Neill M, Pace C. Effects of flow restoration and exotic species removal on recovery of native fish: lessons from a dam decommissioning. Restoration Ecology. 2010;18(6):934–43.

62. Pohl MM. Bringing down our dams: trends in American dam removal rationales. JAWRA Journal of the American Water Resources Association. 2002;38(6):1511–9.

63. Magilligan F, Graber B, Nislow K, Chipman J, Sneddon C, Fox C. River restoration by dam removal: Enhancing connectivity at watershed scales. Elementa. 2016;4.

64. Hoenke KM, Kumar M, Batt L. A GIS based approach for prioritizing dams for potential removal. Ecological Engineering. 2014;64:27–36.

65. Kemp P, O'Hanley J. Procedures for evaluating and prioritising the removal of fish passage barriers: a synthesis. Fisheries Management and Ecology. 2010;17(4):297–322.

66. Neeson TM, Ferris MC, Diebel MW, Doran PJ, O’Hanley JR, McIntyre PB. Enhancing ecosystem restoration efficiency through spatial and temporal coordination. Proceedings of the National Academy of Sciences. 2015;112(19):6236–41.

67. McKay SK, Schramski JR, Conyngham JN, Fischenich JC. Assessing upstream fish passage connectivity with network analysis. Ecological applications. 2013;23(6):1396–409. pmid:24147411

68. Tonitto C, Riha SJ. Planning and implementing small dam removals: lessons learned from dam removals across the eastern United States. Sustainable Water Resources Management. 2016;2(4):489–507.