1. Introduction
Decisions surrounding the future of dams can be informed by scientific inquiry into functions and values linked to their removal, alterations or maintenance [
1,
2]. There are numerous tradeoffs to be weighed with every dam decision. For example, in dam removal decisions, there are the potential benefits of improved migratory fish passage and reduced hazards due to dam failures versus the potential costs of removing water supply storage, flood control, recreational opportunities, pollutant retention, or economic opportunities with hydropower.
Dams stymie migratory fish passage, eliminating or degrading vast expanses of aquatic habitats in coastal watersheds [
3]. This loss of connectivity between estuaries and watersheds affects biota across multiple trophic levels—resulting in negative consequences for the economics, sustainability and biodiversity of fisheries [
4,
5,
6]. Even when fish passage across dams is promoted through technical structures like Denil fish ladders, poor design, changes in the annual flow regime and water temperatures present challenges to migratory species [
7].
Dam failure is a constant concern—and is a clear threat to property, ecosystems and human well-being. Poor construction, aging infrastructure, insufficient maintenance and changes in upstream flow regimes can compromise dam safety. Dam failures can also unleash substantial and sometimes contaminated sediment loads to downstream habitats.
Dams provide a number of important economic and societal benefits including hydropower and flood control [
2]. Some reservoirs associated with dams store water for irrigation and drinking water supplies. Dams—and their associated reservoirs—also provide historic and cultural values to some communities [
8].
The ecological and societal impacts of dam removal can be complex, varied and confounded by a scarcity of systematic data-driven science. To advance the science of dam removal, analyses of functions and values need to be linked to the specific attributes of a given dam and its role within the larger watershed context [
9]. Poff and Hart [
7] argue for the development of ecological classification approaches based on characteristics available in governmental databases.
Here, we propose—and illustrate—an ecological classification based on a potential function of dams—the retention and removal of nitrate-nitrogen from coastal watersheds. Elevated nitrogen (N) loads to coastal estuaries can enhance primary productivity accelerating eutrophication that results in degradation of estuarine habitats (e.g., seagrasses replaced by macroalgae) and hypoxia [
10,
11,
12]. Coastal watersheds—from local to regional scales—retain or remove a substantial portion of N inputs, dramatically reducing the ratio of N delivery to N inputs to estuaries [
13,
14]. This N retention and removal occurs in soils, wetlands, lower-order streams, lakes and reservoirs. Denitrification, the microbial process that reduces nitrate-N to nitrous oxide and N gas is a major removal mechanism in aquatic systems [
13]. Plant uptake and immobilization in organic sediments are processes that serve to retain N [
15]. Following prior conventions in the literature, we use the word “removal” to refer to all N retention and removal processes within reservoirs [
16].
Seitzinger et al. [
17], using empirical data from multiple studies, found that a substantial amount of the variability associated with watershed N removal in aquatic ecosystems could be described by the ratio of depth to hydraulic retention time. Shallow rivers, wetlands, lakes and reservoirs are associated with higher N removal. A number of other empirical and process-level studies have demonstrated that, with sufficient retention times and appropriate average depths, reservoirs of varying scales, ranging from <1 to 4400 ha, can be important locations for N removal [
18,
19,
20,
21]. However, many reservoirs with large contributing watersheds have low retention times and thus relatively low capacity to substantially change N flux.
We explore the extent of potential N removal by reservoirs associated with dams through a case study focused on all reported dams (>14,000) in the New England region (area approximately 200,000 km
2) of the Northeast USA. The region has a high density of dams—many constructed decades or centuries ago—on coastal watersheds that drain to estuaries threatened or degraded by N loading. These coastal watersheds are essential to the life cycle of fish that migrate between marine and freshwater. Interest in dam removal is accelerating, motivated by concerns for both improved fish passage and safety. These dams are located on streams of varying sizes and many are run-of-river dams that do not have impoundments (i.e., reservoirs). We focused our study on situations where dam removal may create a marked change in the watershed N to downstream waters. Modeling [
17,
22] and empirical [
23,
24] studies have demonstrated that N can be removed from free flowing rivers, but here we examined dams with reservoirs that altered the retention time within a given stream reach and thus create a localized hotspot of N removal [
18,
19,
20,
21] that should exceed the N removal associated with the unaltered, free flowing river reach.
Where reservoirs exist, the ponded area can range from several hectares to several thousand hectares. The many combinations of stream order, watershed area, land use and ponded area suggest a widely varied set of N loading, depths and hydraulic retention times—thus a large range in N removal associated with reservoirs. Following the suggestion of Poff and Hart [
7] on the need for regional-scale ecological classifications of dams, we used widely available geospatial databases to examine the extent and locations of dams with potential for N removal. Finally, we link our results to state and regional data that identify specific dams as safety risks or severe impediments to migratory fish passage to signal tradeoff situations where additional site specific studies, including additional N abatement approaches or alternatives to removal, may be warranted.
3. Results
A total of 14,291 of dams were compiled from the combined state-based datasets in watersheds of the six New England states. Within the New England wide dataset 7578 dams were associated with the NHDPlusV2 hydrographic river networks. Hereafter we use the value of 7578 as a basis of comparison when we examine the proportion of “New England dams” with different attributes.
Of the dams on NHDPlusV2 rivers, 2921 dams were associated with reservoirs identified on the NHDPlusV2 database. Of these, 2915 dams were within watersheds that drained to estuaries, bays and sounds of New England (i.e., were not tributaries to Lake Champlain). These 2915 locations will be referred to as “dammed reservoirs” throughout most of the paper.
When we took into account the % N removal attributed to these dammed reservoirs based on Equation (6), we identified 2206 of the dammed reservoirs draining to the Atlantic with the potential to serve as N removal sites (e.g., >2.5% N removal, based on
Figure 1). These 2206 locations will be referred to as “N removal dams” throughout the paper. This represents approximately 29% of all New England dams.
These N removal dams are primarily located on lower-order streams (
Figure 2); approximately 91% are located on either first or second order streams. As expected for lower-order streams, the cumulative watershed areas upstream of most of the N removal dams are relatively small (median: 3.3 km
2; interquartile range: 6.3 km
2) relative to the size of the watershed areas to the major New England estuaries (range: 1000 to 20,000 km
2). The % N removal in N removal dams did not show a pronounced pattern with stream order (
Figure 3).
Table 3 compares our results of reservoir N loading (kg N ha
−1·year
−1) normalized for watershed area to the results from 74 small, coastal watersheds in southern New England [
43]. The watershed loading rates to the reservoirs in our study are substantially lower, due to the many reservoirs located in forested, undeveloped watersheds. Across all N removal dams, undeveloped land cover dominated (median: 86%) the watersheds (
Figure 4)—these land covers generate N loading rates that are orders of magnitude lower than developed and agricultural lands found within the highly settled watersheds of southern New England. Nitrogen loading, normalized for watershed area, was highest in first order streams and declined with increasing stream order (
Figure 5).
Based on the NCAT [
40], 138 of N removal dams in watersheds draining to New England estuaries were rated in the upper decile of all Northeast dams based on their ecological value to anadromous fish restoration that would result from elimination of their barrier (
Table 4). These 130 dams represent just 1.8% of New England dams. Dams deemed to be high priorities for barrier elimination, based on value to anadromous fish, constituted a larger proportion (e.g., >29%) of the dams on higher order streams (4th–5th order) than lower-order streams (
Table 4). When examined for safety hazards, 662 N removal dams (8.7% of New England dams are classified by the New England States as high or significant safety hazards. Across a range of stream orders (small to large rivers) safety concerns ranged between 28% and 44%. The majority of N tradeoff dams (>65%) were located on relatively small 1st order streams, but only 3% of those dams are classified as high value for anadromous fish passage (
Table 4).
Combining those data sets, a total of 763 N removal dams could be high priorities for removal or alterations due to their potential for improving anadromous fish values or addressing safety issues (note that removal of some dams will generate benefits for both improved safety and anadromous fish restoration), though dam removal could represent a loss of their capacity for N removal. These 763 N removal dams with high priorities for anadromous fish or safety, “N removal tradeoff dams,” represent 10% of New England dams.
Only 21 of the N removal dams are active hydropower sites (0.3% of New England dams) (
Table 4). Where hydropower is extant, tradeoff analyses might consider the potential losses of both N removal and hydropower versus gains from anadromous fish or safety associated with dam removal.
Table 5 provides the number of N removal dams with potential tradeoffs for anadromous fish restoration, safety or hydropower in the watershed areas of major estuaries of New England. We have grouped these dams into the watersheds of different Atlantic estuaries in New England to foster attention to the receiving waters that may be influenced by changes in N load associated with any dam removal. The watersheds of Maine Coastal estuaries have the highest numbers of N removal tradeoff dams with values for anadromous fish restoration. The watersheds draining to Long Island Sound contain the largest numbers of N removal tradeoff dams where improved safety is a concern.