River restoration by dam removal: Enhancing connectivity at watershed scalesRiver restoration by dam removal

The prolonged history of industrialization, flood control, and hydropower production has led to the construction of 80,000 dams across the U.S. generating significant hydrologic, ecological, and social adjustments. With the increased ecological attention on re-establishing riverine connectivity, dam removal is becoming an important part of large-scale river restoration nationally, especially in New England, due to its early European settlement and history of waterpower-based industry. To capture the broader dimensions of dam removal, we constructed a GIS database of all inventoried dams in New England irrespective of size and reservoir volume to document the magnitude of fragmentation. We compared the characteristics of these existing dams to the attributes of all removed dams over the last ∼25 years. Our results reveal that the National Inventory of Dams significantly underestimates the actual number of dams (4,000 compared to >14,000). To combat the effects of these ecological barriers, dam removal in New England has been robust with 127 dams having been removed between ca. 1990–2013. These removed dams range in size, with the largest number (30%) ranging between 2–4 m high, but 22% of the removed dams were between 4–6 m. They are not isolated to small drainage basins: most drained watersheds between 100–1,000 km2. Regionally, dam removal has re-connected ∼3% (3,770 river km) of the regional river network although primarily through a few select dams where abundant barrier-free river lengths occur, suggesting that a more strategic removal approach has the opportunity to enhance the magnitude and rate of river re-connection. Given the regional-scale restoration of forest cover and water quality over the past century, dam removal offers a significant opportunity to capitalize on these efforts, providing watershed scale restoration and enhancing watershed resilience in the face of significant regional and global anthropogenic changes.


Introduction
One of the pressing challenges facing biophysical scientists, policy makers, environmental managers, and environmental advocates is how to rehabilitate ecological systems that are increasingly characterized by longterm, significant, and complex anthropogenic changes. There is a growing consensus-representing fields as diverse as restoration ecology, conservation biology, sustainability science, political ecology, and a host of others-that research seeking to understand and enhance the sustainability of human-environment systems within the context of the Anthropocene must embrace trans-disciplinary perspectives ( Jerneck et al., 2011;Steffen et al., 2011;Van Andel and Aronson, 2012;Seidl et al., 2013;Olsson et al., 2014) and think and act across multiple social-ecological scales from the local to the transboundary (Ogden et al., 2013). Nowhere are these challenges of the Anthropocene more important than in the case of efforts to understand, govern, and manage water systems (Sivapalan et al., 2014), wherein decades of human alteration through dams and other infrastructure have profoundly affected a host of hydrological and ecological processes. Elementa: Science of the Anthropocene • 4: 000108 • doi: 10.12952/journal.elementa.000108 removal does not occur within an institutional and social vacuum. For example, because dams are nested within watersheds, their removal may have effects extending well beyond the removal site (Grant and Lewis, 2015) or may generate unintended consequences (Doyle et al., 2005;Sethi et al., 2004). Dam removal -if done "right" -thus presumes a coordinated regulatory and institutional vision and/or an extremely well-organized and well-funded grassroots mobilization to meet desired aims. In contrast, removal may simply reflect an ad hoc process lacking formalized top-down or bottom-up structures (Fox et al., 2016). We are thus mindful of how the goals of restoration advocates are shaped and modified by the institutional structures for regulating the social and ecological use of watersheds. Our specific research questions address: (1) what is the spatial distribution of removed dams and how does this pattern relate to stated management goals of restoring critical habitat for native and diadromous fish; (2) what are the structural or management commonalities in dam types that have been removed; (3) what has been the incremental addition of free-flowing river length accomplished in terms of ecological changes; and (4) what policy or management lessons can be derived from the expected and unexpected biophysical benefits of dam removal? Our results present the ecological achievements at the watershed scale associated with dam removal as an on-going and future river restoration mechanism -a management option that may further serve to enhance the resilience of watersheds. Given recent discussions in the literature regarding watershed resilience (McCluney et al., 2014;Nemec et al., 2014;Waldman et al., 2016), our results point to the ancillary benefits of dam removal as a means of increasing the resilience of social-ecological systems (Walker et al., 2004).

Methods
To best document the geomorphic, hydrologic, and potential ecological effects of dam removal at a regional level, we have compiled a database from state and federal agencies of all inventoried dams in each state in New England and compared the attributes of these existing dams to the population of removed dams compiled by NGOs (e.g. American Rivers) and state agencies, where available. Unlike the NID, there are minimal to no height or storage restrictions, and each state, whether for liability or environmental reasons, maintains a record of its dams, often including information about its type, function, and characteristics (e.g. height, length, etc.). For each existing and removed dam having geospatial information, we "snapped" its specific location directly to the river in ArcMap using the 1:100,000 hydrography layer. To ensure that we snapped correctly, we visually inspected the snapped location for each of the removed dams to be certain it was associated with the correct waterbody. Snapping thus permitted calculating watershed attributes such as basin size, distance to next upstream barrier, and number of free-flowing river km opened up by the removed dam. Most, if not all, of the removals lacked geomorphic attributes, such as sediment characteristics (bedload vs. suspended load, local or reach slope, etc.). To best represent these geomorphic parameters, we measured the distance and elevation change between the former dam and the headwater divide and used this basin-derived slope (also known as relief ratio) as a proxy of sediment type, based on established relationships between grain size and slope (Dade and Friend, 1998;Snyder et al., 2013).
For each removed dam we also calculated watershed attributes (e.g. percent of watershed developed, urbanized, or forested) using the National Land Cover Database ( Jin et al., 2013). Moreover, the EPA (https://www.epa.gov/wed/pages/ecoregions/na_eco.htm) divides New England into five major ecoregions (Acadian Plains and Hills, Atlantic Coastal Pine Barrens, Eastern Great Lakes Lowlands, Northeastern Coastal Zone, and the Northeastern Highlands) differing in channel habitat, riparian and watershed vegetation, aquatic biodiversity and fish assemblages and potential for river restoration. Our GIS assessment groups the existing and removed dams occurring in each of the ecoregions. This linking to ecoregions provides an opportunity to ascertain which ecological types and settings are currently under-or overrepresented by dam removal efforts and the extent to which dam removal currently contributes to the ecological integrity of river systems at a regional scale. We also compare watershed and structural attributes of removed dams to the general population of dams, to ascertain the extent to which removals reflect the broad array of dam types and settings in the region. Our institutional analysis is based on the findings of an ongoing assessment of the social dimensions of dam removal in New England (Fox et al., 2016) consisting of semi-structured interviews with state and non-state stakeholders, participant observation at public meetings regarding dam removal, and textual analysis of hundreds of documents.

Results
Our compilation from states agencies and NGOs indicate that the number of dams documented in New England by the NID (∼4,000) significantly underestimates the actual number of dams as more than 14,000 inventoried dams are currently peppered throughout the New England landscape (Table 1; Figure 1A) generating a density of ∼8 dams per 100 km 2 . Most of these inventoried dams are in Connecticut, Maine, Massachusetts and New Hampshire, with the fewest in Rhode Island and Vermont (Table 1). These existing structures obstruct an array of watershed types that possess orphaned mill dams, to small headwater water Elementa: Science of the Anthropocene • 4: 000108 • doi: 10.12952/journal.elementa.000108 supply dams, to larger hydropower facilities ( Figure 2). Based on our detailed compilation, 127 dams have been removed in New England over the past several decades (up to 2013), presently averaging ∼12 yr -1 . Most of the removed dams were small with ∼45% of them being less than 4 m in height ( Figure 3); for comparison, across New England 35% all existing dams are less than 4 m ( Figure 3). The frequency of removed small dams (< 4 m) is slightly larger than the binned values for existing dams; however, it is important to keep in mind the two orders of magnitude difference between the number of removed dams (10 2 ) relative to the sheer number of existing dams (10 4 ). For example, although the frequency of removed dams in the 6-8 m height category seems high (8.8%), this occurrence only corresponds to 8 dams, while there still remains more than 500 dams of this size regionally. Even with the prevalence of small dams among those eliminated, removals have not been exclusively restricted to small dams; over 40 dams > 4 m have been removed, including five > 8 m. These removed dams occurred widely across and within drainages. Moreover, despite the predominance of small dams among those removed, these dams were not restricted to headwater locations; most (38%) occurred in medium-sized watersheds having upstream drainage areas between 100-1,000 km 2 (Figure 4), with 8% formerly impounding watersheds between 1,000-10,000 km 2 .  Within New England, most of the removed dams were located in Massachusetts (31) with Maine and New Hampshire accounting for 28 and 26 removed dams, respectively. Connecticut, which has over 3,600 documented dams, has only removed 21 dams and Rhode Island, an important coastal state, has only removed 4 dams. Vermont has removed 17 dams, many of them being very small run-of-river facilities. At the broader regional scale, the two dominant ecoregions of New England are the Northeastern Coastal Zone (NCZ) and the Northern Highlands, which account for 40% and 38% respectively of all dam removals in New England ( Figure 1B). Yet, despite the NCZ having the most removals, most of the removed dams were generally far inland ( Figure 1B), and this ecoregion also houses a large number of existing dams, with many of the removal sites possessing multiple dams downstream further fragmenting the watersheds and disconnecting them from the ocean. Most of the removed dams occurred at elevations below 100 m with a peak between Dam height (m) for existing and removed dams in New England.
The specific location of the dam within a watershed relative to the location of the next upstream barrier, in combination with upstream watershed structure (drainage density, number of tributaries, etc.), best represents the acquired benefit of dam removal in opening access to upstream river reach. Our comparison of the removed and existing dams databases shows that ∼3,770 river km have been re-connected in New England waterways by dam removal (Figure 7). With New England possessing ∼104,000 km of total river length, this liberated space of 3,770 km by dam removal represents 3.61% of all river lengths. Areas upstream of the former dams are well forested with little development or agriculture ( Figure 8). Most of the liberated free-flowing river space has occurred in Maine. Removal of the Sandy River, Fort Halifax, Pleasant River and Bangor hydroelectric dams represent ∼1500 km of the reconnected river lengths in Figure 7. With Maine accounting for the few viable remaining runs of the federally-endangered Atlantic salmon, these dam removals have provided significant access to upstream habitat (Hogg et al., 2013(Hogg et al., , 2015Pess et al., 2014). The shape of the cumulative river length curve also reveals how most of the gain in reconnected river length is due to the removal of a small number of key barriers. However, even in instances where considerable re-connectivity has occurred, as for example the gain of 130 river km during the removal of the Homestead Woolen Mills Dam on the Ashuelot River in NH, several large dams occur both upstream and downstream (Gartner et al., 2015), ultimately blocking diadromous fish from accessing this newly available habitat.

Discussion
Besides documenting the general characteristics of removed dams at a regional assessment, our results provide, for the first time, the magnitude and extent of watershed scale re-connectivity resulting from dam removal. This, in turn, provides a basis for assessing how dam removal may enhance watershed resilience in the face of multi-scalar human impacts. As discussed below, we perceive dam removal as a potentially critical tool for not only re-establishing connectivity at watershed scales, but for enhancing the capacity of important upstream catchments to respond and adapt to regional and global anthropogenic changes (e.g., climate change) in ecologically desirable ways.

Regional benefits from dam removal
In terms of the general features of the removed dams, our data indicate that dam removal in the northeastern U.S. is not restricted to dams isolated in small headwater locations. Although well below the height of most flow regulation structures, most (30%) of the removed dams were between 2-4 m with 22% of them between Watershed slope.

Figure 7
Liberated river kilometers. 4-6 m -a height generally sufficient to meet the requirement to be registered in the NID. Additionally, they formerly dammed moderate size watersheds, generally between 100-1,000 km 2 , corresponding in size to a HUC-4 or HUC-5 USGS Hydrologic Unit Code. Moreover, the gain in longitudinal re-connectivity is further augmented by the associated quality of the now liberated upstream watershed. In a sense, dam removal will allow the region to capitalize on over a century of water quality improvements. These new liberated sections of the watershed are well forested with minimal development (Figure 8), thus providing access to high quality habitat. Besides providing important fish passage, these newly liberated watershed sections may thus serve to be important refugia as they are in some of the least disturbed watershed sections. The lack of development and broader forest cover may further serve to maintain the resilience of these ecosystems at a time of increasing concern of shifting freshwater thermal regimes regionally (Hayhoe et al., 2007). For example, distributions of native coldwater-dependent species (such as salmon and trout) are predicted to move upstream in response to downstream warming. Removing barriers such as dams is critical to this response, and in New England our analysis indicates that the high levels of forest cover and low development pressure of these upstream sites will increase the likelihood that they will serve as refugia (Kanno et al., 2014). Further, given that the scope for mitigating or ameliorating change via forest or water quality restoration may be limited (due to existing high quality and extensive cover), dam and barrier removal may be the most effective remaining adaptation strategy. Although much of the public attention has been focused on large dam removal (Lovett, 2014;O'Connor et al., 2015;Service, 2011aService, , 2011b, our results elucidate the important singular and cumulative benefits of removing run-of-river dams. In their national assessment of river restoration projects Bernhardt et al. (2005) document the associated expense and limitations of the various approaches for restoration, and dam removal, on average, is cheaper than channel reconfiguration or other restoration measures. Not all of our removals provided firm costs for removal, but we estimate from our regional compilations that, in general, dam removal costs ∼$80k per vertical meter of dam height, corresponding to inflation-adjusted values reported by Born et al. (1998). This base removal cost, however, becomes progressively more expensive for any sediment remediation requirements. As dam removal becomes more prevalent in New England and other regions, such cost estimates provide an important additional consideration for environmental managers and other concerned actors. In addition, not only are restoration efforts (e.g. bank stabilization, channel reconfiguration, etc.) more costly, they often are site-specific generating, at best, reach scale improvements; whereas dam removal offers a greater spatial scale of watershed rehabilitation.

Geomorphic and ecological implications
We were not able to assess any geomorphic adjustments or specific ecological benefits with each dam removal as post removal monitoring in New England, as elsewhere, has generally been absent. In fact, in a recent national assessment of dam removal monitoring, Bellmore et al. (2015) document that of the > 1,100 dams Box and whisker plots of land use and land cover types for portions of the watershed now made accessible by dam removal (from National Land Cover Database ( Jin et al., 2013). The solid line in the box is the 50 th percentile while the box edges define the 25 th and 75 th percentiles. The circle is the mean and the whiskers identify the minimum and maximum values. Elementa: Science of the Anthropocene • 4: 000108 • doi: 10.12952/journal.elementa.000108 removed nationally, only 139 had any monitoring whatsoever, and only 35 had both geomorphic and ecologic assessments. Detailed post-removal geomorphic assessments have been done on just a few rivers in New England (Pearson et al., 2011;Magilligan et al., 2016), and these studies show that most of the geomorphic adjustments occur within the first year driven primarily by the initial base level adjustment -a trend typical of dam removals elsewhere (Grant and Lewis, 2015;Sawaske and Freyberg, 2012). Because of the thin alluvial cover over bedrock in New England, upstream knickpoint migration is commonly limited to dam proximal reaches (Pearson et al., 2011;Gartner et al., 2015) suggesting that post-removal prolonged sediment production may not occur or impair downstream geomorphic and ecological stability. Although no published sediment data exist for these removals, the modal basin slope of 0.5 -1% ( Figure 6) corresponds generally to channel bed sedimentological environments typical of a sand to sandy gravelly matrix -typical of the coarse-grained Pleistocene deposits that mantle most of New England. The lack of fines regionally (Rainwater, 1962) and at the sites of removed dams further serves to assuage environmental concerns that pollutants sorbed to finegrained sediment (silts and clays) may be flushed downstream following post-removal sediment evacuation. Although New England has experienced significant historical industrialization, downstream dispersal of contaminated sediments following dam removal has not been generally documented, in part due to the coarse (sand to gravel) nature of the released sediment where few binding sites -relative to finer grained material -exist for sorbing pollutants and because of the limited monitoring of these effects. Additionally, the length scale for sediment dispersal is grain size dependent; in a detailed analysis of downstream travel distance, Grant and Lewis (2015) document the spatially limited transport distances for the coarse fraction following dam removal. Contaminated sediment may potentially be a more significant issue in southern New England where lower slopes and finer-grained sediment exist.
From an ecological standpoint, the pattern revealed in Figure 1B depicts, perhaps, an operational disconnect between management goals to restore diadromous fish populations and the dam removal process. Although most of the dams have been removed in Massachusetts and in the Northeastern Coastal Zone, very few of them were along the most important reaches for marine-freshwater exchange -i.e. the most downstream watershed positions. Our analysis does not include efforts to restore fish passage (through ladders, lifts, and bypasses) at existing structures, which have been implemented widely throughout the region. However, in contrast to dam removal, the efficacy of these passage structures is highly variable across sites, years, and species , and even when they do facilitate adequate passage, they do not ameliorate the effects of dams on habitats and sediment regimes (Bunt et al., 2012). Further, the presence of downstream dams does not mean that dam removal has been an ineffective agent in the restoring aquatic ecosystems. Instead, with more than 3,770 river km now made accessible, resident fish and at-risk species have greater access to previously unavailable habitat, and re-establishing the continuity of sediment transport has likely facilitated the development of important fluvial habitats (bars, banks and floodplains) downstream.

Landscape of strategic opportunism
From an environmental management and policy perspective, the pattern evident for dam removals in New England ( Figure 1B) evokes a landscape of strategic opportunism more than a well-articulated management scenario planned out and implemented by local and regional stakeholders. This opportunistic strategy reflects the ad hoc nature of dam removals whereby NGOs and state agencies essentially react and respond to willing dam owners who -either by environmental awareness, business/personal foresight, FERC requirements, or economic liability -decide that removal may the best economic or personal option. The lack of top-down or bottom-up driven processes may appear to be an effective strategy as it has generated over 125 removals (∼10% of national removals), but the ecological/geomorphic effectiveness of removals as an intervention would almost certainly be enhanced by more programmatic approaches. There is an emerging literature engaging with various management scenarios, ranging from a "hit list" approach (Hoenke et al., 2014) to more market based strategies for prioritizing barrier removal (Kemp and O'Hanley, 2010;Neeson et al., 2015). With the advent of geospatial databases and associated spatial algorithms, these approaches offer salient management strategies to prioritize dam removal as it is now possible to quantify which dam, once removed, may liberate the greatest number of free-flowing river km or offer the greatest opportunity for watershed or river restoration. For example, the US Army Corps of Engineers (USACE) is proposing to investigate dam removals as part of an overall "portfolio" approach to ecosystem restoration at the watershed scale as evidenced in their recent comprehensive plans for fish passage along the multi-dammed Blackstone River (USACE, 2015).
Both the "strategic opportunism" and the "priority list" approaches, however, assume a social arena free of political confrontation, an institutional structure or arrangement that has the political and economic will and resources to make environmentally informed decisions, and/or a general agreement that dam removal is the most effective restoration strategy (Fox et al., 2016). Indeed, current research on dam removals in New England highlights the vagaries of efforts to promote watershed resilience through dam removal and outlines the numerous pathways that local political processes can derail even the most thoughtful restoration efforts (Fox et al., 2016). Our results point to perhaps, by default, a "happy medium" wherein dam removal -and its associated environmental gains -has progressed incrementally through political, economic, and Elementa: Science of the Anthropocene • 4: 000108 • doi: 10.12952/journal.elementa.000108 environmental expediency to achieve some of the stated management and river restoration goals. Yet, lacking any significant sustained monitoring or post-removal assessment, it is difficult to specifically determine the actual gains ecologically from removal.
Assessments are further complicated by the lack of common ecological metrics or stated management goals to evaluate the success of a given restoration strategy or outcome Palmer et al., 2005;Bernhardt et al., 2007), a condition common to management efforts in the Anthropocene (Seidl et al., 2013). NOAA does provide baseline metrics for assessment (Wildman, 2013), but in most instances, the benchmark may be merely the re-introduction of migratory and resident fish species to previously unattainable watershed locales (Hogg et al., 2013(Hogg et al., , 2015Pess et al., 2014) or the dam removal may generate more re-connected coupled geomorphic and ecological attributes . The lack of a singular metric of restoration success for dam removals also manifests as dams are removed for numerous socio-economic reasons (e.g. safety, liability, aesthetics, etc.) that often lack a clearly articulated ecological goal. Even when environmental reasons are stated, dam removals rarely identify clearly stated ecological standards. In some instances, the implicit assumption by many pro-removal NGOs is that a river with a removed dam is in a more "natural" state and therefore the restoration goal is self-evident. More often, the stated primary goal, like at the recent large dam removals on the Elwha (WA) and Penobscot (ME) Rivers, is re-connecting migratory fish runs, yet the metric of success is unclear: is it fish presence, fish abundance, or long-term viability of populations? Because these demographic metrics are difficult to measure or assess scientifically, and, in the case of diadromous fish, are also determined by coastal and marine influences, NGOs (e.g. American Rivers) and federal agencies (e.g. NOAA) that fund dam removals will commonly just present the number of river kilometers opened up by the removal -again signaling that a watershed with more kilometers of a "re-connected" free-flowing access is in a more natural state. Although still limited in regional or temporal scope, recent literature in New England suggests that "if you take it down, they [fish] will come" (Hogg et al., 2013(Hogg et al., , 2015Pess et al., 2014). Our results highlight some of these ambiguities in river restoration, but also show that measurable "gains" can result from dam removal -at least from a watershed resilience perspective based on achieved gains in accessible river habitat. This is particularly salient when considering the Anthropocene context where baseline ecological knowledge is difficult if not impossible to determine and human-altered systems are the new ecological reality.

A forever dam(n)ed landscape?
From a management and restoration perspective, New England remains a dammed landscape ( Figure 1A). "Hit list" or basin-oriented perspectives may initially appear attractive but the optimal coalescence of cost, ecological gain, and political/institutional arrangements may not manifest in strategic and meaningful ways. For example, the much heralded removals of two dams on the Penobscot River required sustained negotiations and political maneuvering (Day, 2006;Opperman et al., 2011), yet our GIS-based results indicate that due to the watershed structure -where few major tributaries enter the mainstem and the remaining presence of the upstream dam -very few river kms were made available by these removals. Fish passage along the mainstem may have been achieved, but it remains to be seen whether successful access to a greater extent of upstream spawning habitat can or will occur. Management strategies are slowly becoming more coordinated in New England and there has been some movement away from the singular removal to think in a more "portfolio" approach to removals. For example, the near collapse of the ∼200 year old Whittendon Dam on the Mill River near Taunton MA in 2005 generated an initial political frenzy to "save the dam", but the response slowly shifted to a more holistic strategy to remove each of the three abandoned mill dams over a period of 3-5 years. Rather than merely removing the decaying Whittendon Dam -that would have offered minimal gains as the Hopewell Mills Dam still existed downstream -a partnership of nonprofit groups and state and federal agencies coalesced to initiate the removal of three closely spaced dams that will now liberate more than 50 km of free flowing river length. The aforementioned plan, spearheaded by the Corps of Engineers, to systematically assess the viability of multiple dam removals and other rehabilitation measures for the entire Blackstone River watershed in Massachusetts and Rhode Island is another example of the more coordinated approach. Although still politically charged and institutionally fragmented, these broad-scale prioritizations can offer greater opportunities for river restoration regionally and can help guide the removal process.

Toward watershed resilience in the Anthropocene
The concept of resilience as applied within ecology and the management of complex social-ecological systems (SES) has been the subject of decades of debate and refinement (see Curtin and Parker (2014) for an overview). We understand resilience as "the capacity of a system to absorb disturbance and reorganize while undergoing change so as to still retain essentially the same function, structure, identity and feedbacks" (Walker et al., 2004). While there have been some efforts to apply resilience to watershed systems (see McCluney et al., 2014;Nemec et al., 2014), these efforts have failed to clarify what watershed resilience might look like and Elementa: Science of the Anthropocene • 4: 000108 • doi: 10.12952/journal.elementa.000108 how it might be measured. Most research quantifies the loss of resilience due to acute and chronic human interventions (Nemec et al., 2014;Waldman et al., 2016), but less effort has addressed the gain in, or metric of, resilience following some restoration mitigation effort such as dam removal. Some metrics used in a SES approach to assess a watershed's improved resilience post-removal are difficult if not impossible to determine with certainty. Other metrics (e.g., presence of formerly absent diadromous fish species), while superficially appealing and potentially important, may not reveal much about the overall integrity of a watershed but, instead, reveal more about fish community resilience following dam removal (Waldman et al., 2016). These limitations are especially significant in regions such as New England with a relatively long history of intensive human modification. Instead, watershed resilience is perhaps best characterized by the capacity of a watershed, as a complex social-ecological system, to absorb an on-going or future stressor in a way that limits degradation of not only the habitats of aquatic organisms but also an array of biophysical functions, structures, and processes.
Our results underscore the capacity of dam removal to certainly enhance watershed resilience in the sense of river stretches opened up and greater connectivity. These are the expected benefits of ecological interventions such as dam removal. However, one of the less expected benefits of removal has been improvement to critical habitats in the upstream catchments of New England's watersheds. A brief example illustrates this point. As climate change progresses and average annual temperatures continue to climb over the coming decades, viable habitat for cold-water fish species becomes even more crucial. The removal of dams in upper catchments of New England's rivers capitalizes on decades-long improvements in forest cover ( Figure 8) and water quality, implying that the resilience of these systems in the face of broader-scale and long-term anthropogenic changes may be enhanced. In a very real sense, dam removal links the scale of catchment and reach level biophysical dynamics (e.g., connectivity, fish migrations) to broader scales of environmental change (e.g., climate change). While impossible to predict the longer-term impacts of dam removal given the rapidly changing socialecological contexts of the Anthropocene, there is some room for cautious optimism that dam removal and other interventions may actually enhance watershed resilience in important and unexpected ways.

Conclusion
There has been considerable effort over the past several decades to re-establish longitudinal and lateral watershed connectivity and to, more broadly, "restore" rivers in the United States and globally. River restoration is a broad (and often contested) term that may take multiple forms and is often scale dependent: these strategies are often localized and occur primarily at the reach scale (Bernhardt et al., , 2007. At somewhat larger scales, river restoration may refer to the re-establishment of river connectivity by environmental flow releases (Arthington et al., 2006;Hughes and Mallory, 2008;Poff and Zimmerman, 2010), but these efforts can be fraught with design, implementation, and management issues surrounding sediment flux and ascertaining the correct magnitude and timing of flow releases (Mahoney and Rood, 1998;Richter and Thomas, 2007;Schmidt et al., 2001). Moreover, environmental flow releases may serve to enhance downstream lateral connectivity, but because the dam remains, they do not un-fragment watersheds or promote longitudinal connectivity. For these reasons, perhaps a turn towards thinking about dam removals as a kind of ecological intervention designed to enhance watershed resilience offers an opportunity to more carefully situate dam removal within the broader goals and activities of ecological restoration.
Our region-wide analysis points to the greater scale of restoration associated with dam removal, and its ability to regenerate a suite of riverine processes including enhanced sediment connectivity, unfragmenting watersheds to allow fish passage, and the opening up significant river length and important habitat for resident and diadromous fish. Dam removal is progressively becoming part of the management toolkit nationally, and our results point to the greater potential for re-connectivity at the watershed scale and, perhaps more importantly, for enhanced watershed resilience. Accordingly, our results point to some unexpected biophysical benefits of undamming New England rivers. Dam removal is at best presented by restoration advocates as a means of enhancing fish passage and returning watersheds to some previous state that is virtually impossible to determine with precision. Some of these claims are accurate, but there is a value added to dam removal that is rarely voiced. This value is related to the capacity of dam removal to increase watershed resilience-as evidenced by the opening up of critical upstream habitats for certain fish species-in the context of large-scale and enduring anthropogenic changes (e.g., climate change). To be certain, additional research is necessary to specify how the multiple-scale dynamics associated with dam removal function. The results presented here represent a step in that direction, and bolster the notion that dam removal is a potentially critical tool not only for restoration activities but also for thinking about and assessing watershed resilience.