What Have We Lost? Modeling Dam Impacts on American Shad Populations Through Their Native Range

American shad (Alosa sapidissima) are native to the east coast of North America from the St. Johns River, Florida, to the St. Lawrence River region in Canada. Since the 1800s, dams have reduced access to spawning habitat. To assess the impact of dams, we estimated the historically accessed spawning habitat in coastal rivers (485,618 river segments with 21,113 current dams) based on (i) width, (ii) distance from seawater, and (iii) slope (to exclude natural barriers to migration) combined with local knowledge. Estimated habitat available prior to dam construction (2,752 km2) was 41% greater than current fully accessible habitat (1,639 km2). River-specific population models were developed using habitat estimates and latitudinally appropriate life history parameters (e.g., size at age, maturity, iteroparity). Estimated coast-wide annual production potential was 69.1 million spawners compared with a dammed scenario (41.8 million spawners). Even with optimistic fish passage performance assumed for all dams (even if passage is completely absent), the dam-imposed deficit was alleviated by fewer than 3 million spawners. We estimate that in rivers modeled without dams, 98,000 metric tons of marine sourced biomass and nutrients were annually delivered, 60% of which was retained through carcasses, gametes and metabolic waste. Damming is estimated to have reduced this by more than one third. Based on our results, dams represent a significant and acute constraint to the population and, with other human impacts, reduce the fishery potential and ecological services attributed to the species.


INTRODUCTION
The migration of animals remains one of the most recognizable and ecologically spectacular occurrences in nature. Animals from diverse evolutionary lineages share a behavioral solution to the seasonal and ephemeral nature of habitat suitability for different phases of their life histories (Dingle and Drake, 2007). These may occur across wide expanses of aerial, terrestrial and aquatic habitat, and the flux of these organisms spans and inextricably links disparate and sometimes distant ecosystems (Weaver et al., 2018). Organisms transport biomass and nutrients (Kelt and Van Vuren, 2001), sometimes over great distance and large geographic scales. The timing of movements, the food-web function and life-history strategies may all influence ecological significance (Rosenzweig, 1971). These phenomena fundamentally shift interlinked systems from marine to inland ecosystems with annual regularity (Doughty et al., 2016).
Migration in sea-run or diadromous fishes is illustrative of seasonal movements that links the ocean with fresh water habitat. The global patterns of diadromy have been well described (McDowall, 1987) and the general trends have been linked to productivity differences between inland and marine habitats (Gross, 1987) in complex ways (Dodson et al., 2009). These fish species have been important to humans prior to colonization of North America, through to the present day. Wilcove (2010) describes the four great threats to migratory animals, each of which is a result of human population growth: habitat destruction, overexploitation, climate change and barriers to migration. In coastal regions, all of these threats are evident, but impoundments represent a conservation challenge that results from both complementary and contradictory socioeconomic tradeoffs that directly influence fish populations (Roy et al., 2018;Song et al., 2019;Roy et al., 2020). Migratory fish have declined, somewhat predictably, through the loss of connectivity to habitat critical for the expression of their life history. Damming fundamentally alters the longitudinal connectivity of freshwater ecosystems, particularly for anadromous fish (Hall et al., 2011;Liermann et al., 2012), a fact that has been effectively revealed through the removal of dams in coastal systems (e.g., Watson et al., 2018;Wippelhauser, 2021).
American shad (Alosa sapidissima) are native to the east coast of North America and demonstrate the human toll on migratory organisms. This anadromous clupeid has an extensive native range, from the St. Johns River, Florida, to the St. Lawrence River region in Canada. Adults must enter freshwater to spawn as a critical part of their life history (Zydlewski and McCormick, 1997b;Zydlewski and Wilkie, 2012). While some populations have flourished outside of their native range (e.g., the Columbia River; Petersen et al., 2003), native populations have been, and remain, depleted (Atlantic States Marine Fisheries Commission (ASMFC), 2020). Since the late 1800s, the four great threats to migratory animals have driven precipitous population declines range wide (Bilkovic et al., 2002;Limburg and Waldman, 2009;Hasselman and Limburg, 2012) and remain continuing and persistent threats to this fish (Burdick, 1954;Talbot, 1954;Bradley, 1959;Chittenden, 1969).
The development of American shad fisheries in the 1700s grew to a pattern of over exploitation by the late 19th century that, combined with habitat loss into the 20th century, resulted in a functional collapse of fisheries and fishery closures in the late 20th and early 21st centuries. Historically abundant, this species supported commercial fisheries with coast-wide landings that exceeded 20,000 metric tons (MT) in the late 1890s (Walburg and Nichols, 1967;Hightower et al., 1996;Limburg et al., 2003, Atlantic States Marine Fisheries Commission (ASMFC), 2007. The ability to access upriver habitat permitted great population potential in river systems and provided sustainable human value (Limburg et al., 2003). Early settlers of Cooperstown, New York, were said to have avoided starvation by American shad reaching Otsego Lake, the Susquehanna River's source, more than 1,000 km from the coast (Taylor, 1995). In the Gulf of Saint Lawrence, American shad reached the Ottawa River, a distance of over 1,100 km (Provost, 1987). Modern fishery impacts are mollified by stricter harvest regulations such as moratoria (Olney and Hoenig, 2001; Atlantic States Marine Fisheries Commission (ASMFC), 2007) but much of the historic human value (economic, recreational and cultural) for this species has been lost with its absence or reduction in coastal rivers due to persistent anthropogenic influences.
Most notably, dam construction greatly restricted access of American shad to spawning and rearing habitat, directly limiting the scope for population and spawner abundance (Rulifson, 1994;Limburg et al., 2003). Dams may provide critical societal functions to meet needs for electricity, water supply, and flood control (Roy et al., 2018) but these dams, their operations, and impoundments may also conflict with fish conservation goals (Song et al., 2019). Fish passage is often a requested and implemented strategy through the Federal Energy Regulatory Commission (FERC) in the United States, but implementation levels can vary widely for these multi decade permits (Vogel and Jansujwicz, 2021).
Latitudinal differences in the life histories of American shad have shaped the influence imposed by dams through their range. Early and influential assessments of these clinal variations in life history traits (i.e., Carscadden and Leggett, 1975;Leggett and Carscadden, 1978;Glebe and Leggett, 1981a) remain heavily relied upon by both researchers and managers. American shad are entirely semelparous (and more fecund) in the southern rivers with a lower length at age (Leggett and Carscadden, 1978;Gilligan-Lunda et al., 2021) than in northerly populations. North of the Cape Fear River in North Carolina, some level of repeat spawning is observed, and the proportion increases in rivers to the north. Northern American shad are larger and are observed to spawn in as many as five (Grote et al., 2014;McBride et al., 2016) and even seven seasons (Provost, 1987). The degree to which this clinal life history variation reflects phenotypic plasticity (versus genotypic differences) remains an important management interest.
American shad exhibit spawning fidelity to their natal rivers (Talbot, 1954;Hill, 1959;Nichols, 1966;Carscadden and Leggett, 1975) with moderate levels of straying (Mansueti and Kolb, 1953;Williams and Daborn, 1984;Melvin et al., 1986). While divergence among rivers in Canada is notable (Hasselman et al., 2010), human mediated transfers among rivers are likely causal to diminished apparent divergence among some populations within the United States (Hasselman et al., 2013). Thus, while genetic discrimination at a river level may be most appropriate, categorization of three main eco-regions (Northern Iteroparous [NI], Southern Iteroparous [SI] and Semelparous [SM]) provides a convenient and logistically functional division of American shad rivers (Figure 1; Atlantic States Marine Fisheries Commission (ASMFC), 2020).
Habitat obstruction or inadequate dam passage may influence populations in these three eco-regions differently. For all ecoregions, access to spawning habitat represents a significant outcome of effective upstream fish passage through dams. The survival of both juveniles and adult migrants during downstream migration is important for population dynamics (Castro-Santos and Letcher, 2010;Stich et al., 2019), albeit to varying degrees for each eco-region. In the SM eco-region, downstream adult passage is not required as they will have made their terminal migration. For the SI and NI eco-regions, the importance of downstream passage for adults is of obvious importance. Failure to provide downstream passage results in an ecological trap that truncates the age distribution of the population (Stich et al., 2019). This is important as repeat spawners have increased reproductive potential (higher fecundity) and provide a buffer against poor recruitment years (Carscadden and Leggett, 1975). The progeny of adult spawners that successfully ascend a dam in all eco-regions, also must all pass downstream as juveniles to reach the ocean.
Together, differences in the life histories among eco-regions may differentially influence the delivery of biomass and nutrients to a river. Semelparous Pacific Salmon (Oncorhynchus spp.; Cederholm et al., 1999;Naiman et al., 2002;and Schindler et al., 2003) and sea lamprey (Petromyzon marinus; Weaver et al., 2016) deliver significant nutrients subsidies through the decay of carcasses. With northern populations displaying lower spawning mortality rates and a greater instance of repeat spawning than southern populations, nutrient delivery through excretion and spawning (gametes) may be most important. American shad generally do not feed upon entering rivers, but their mass loss during the run can be considerable (Glebe and Leggett, 1981a,b;Leonard and McCormick, 1999;Walter and Olney, 2003). Regardless of the source, the level to which the historical delivery of marine nutrients, and the seaward return of nutrients by juveniles, has shifted because of dams.
Conservation efforts on the Gulf of Maine's Penobscot River resulted in the removal of two main-stem dams (Opperman et al., 2011) and spurred interest in understanding the restoration potential of American shad for this river which historically had supported an abundant population. Annual landings of over 2 million adults were reported in the 1860s (Foster and Atkins, 1867) and American shad had been a critical fishery for the Penobscot Nation. Trinko Lake et al. (2012) defined the spatial extent of American shad habitat in this system and characterized connectivity. The development of a simple population model provided a means of assessing the influence of dam removals on the production potential of a newly opened river (Bailey and Zydlewski, 2013) but fell short in providing a tool that incorporated fish passage. Stich et al. (2019) developed a generalizable model platform that could be applied to any river system based on management needs. This approach allowed for the assessment of upstream and downstream passage on river-specific productivity. The importance of understanding fish passage in light of both upstream and downstream efficiencies and survival was made evident. This approach was then applied to a series of rivers, range wide, based on informed habitat assessments (Gilligan-Lunda et al., 2021). Similar work by Barber et al. (2018) used this population modeling approach to characterize the delivery of nutrient by another alosine, the alewife (Alosa pseudoharengus), by employing estimates of nutrient loss, delivery, and exodus.
In this paper we aspire to implement these complementary approaches of spatial analysis, river specific population modeling and assessment of nutrient dynamics to address the lament of Limburg et al. (2003): We can only imagine today what full Atlantic coastal ecosystems (rivers, estuaries, and coastal marine areas) looked like, but one thing is clear, shad played a far more important role then than they do today. To plan for a sustainable future for American shad, we should reenvision those systems and strive to balance fisheries demands with their ecological function. To do this will require better modeling, better data, and above all, renewed commitment.
To estimate the opportunity cost realized by this species through dam construction throughout its range, we sought to first estimate habitat that was historically and currently exploited by American shad for spawning in Atlantic coastal rivers. This was accomplished by characterizing 485,618 river reaches and 21,113 dams. We identified potential spawning habitat based on criteria of (i) river width, (ii) distance from seawater intrusion and (iii) slope (to exclude natural barriers to migration) combined with local knowledge. The areas of potential spawning habitat were aggregated to estimate historic habitat available prior to the construction of dams and impoundments for each coastal river. Each river was then assessed using a life history-based population model incorporating latitudinally appropriate life history parameters (e.g., clines in size at age, maturity rates, iteroparity, and maximum age). In aggregate this approach allowed a direct assessment of the theoretical spawning potential lost coast-wide to the construction of dams and allowed us to estimate the historic and current capacities for biomass and nutrient delivery.

Characterization of American Shad Habitat Through Their Native Range
We estimated American shad habitat based on available knowledge of habitat extent, area, and accessibility. These data were collected for the entire historic geographic extent of American shad, spanning eastern United States and Canada using a two-step approach. First, we used the United States National Hydrography Dataset (USGS, 2019) and Canadian National Hydrographic Network (Natural Resources Canada, 2019) to determine the potential freshwater networks available for migration, spawning, and rearing. The data were organized as a series of flowline segments, representing interconnected stream and river reach segments (Figure 2). To simplify this analysis, we assumed initially that American shad would not migrate to reach segments with a mean channel width of less than 15 m, in accordance with pre-existing habitat suitability FIGURE 1 | Estimated American shad habitat through a combination of local knowledge and a priori modeling along the eastern coast of North America (a), the full extent of their historic range. In the full panel (b), gray lines indicate suitable riverine habitat. Only the first dam on each river or tributary is shown (black circles) for clarity. Dashed lines indicate divisions between northern iteroparous (NI), southern iteroparous (SI) and semelparous SM) eco-regions Insets provide greater detail for four rivers; Penobscot River (c), Connecticut River (d), Susquehanna River (e), Cooper River and (f) the Santee River, indicating the numerous impoundments that constrain the migration of American shad in these watersheds.
Frontiers in Marine Science | www.frontiersin.org 4 October 2021 | Volume 8 | Article 734213 models (Stier and Crance, 1985;Harris and Hightower, 2012). Second, we validated our a priori assessment of historic habitat extent of American shad with the help of local experts from each state or province in our study region (see section "ACKNOWLEDGMENTS"). The experts modified historic extents based on the presence of natural barriers (e.g., steep rapids, waterfalls) or environmental conditions (e.g., temperature, salinity, and reach segment width) that are unsuitable for American shad spawning. Empirically determined stream discharge-width relationships were used to calculate potential habitat area from the flowline data (e.g., Leopold and Maddock, 1953). This approach allowed for the estimation of horizontal surface area at any reach segment based on drainage area, geographic location, and nearby stream gage data. We first used the enhanced unit runoff method (EROM;McKay et al., 2012) to estimate mean annual discharge, then estimated mean reach segment width (w) using the power law equation: where Q is discharge, k was a derived width coefficient, and b was a derived exponent. Values for k and b vary by region but are typically close to 10 and 0.5, respectively (Bray, 1982;Sweet and Geratz, 2003;Dudley, 2004;Mohamoud and Parmar, 2006;Bent and Waite, 2013). Horizontal surface area was then calculated as: where l is segment reach length. We assumed that fluctuations in discharge cause 20% of this area to be periodically dry along the shorelines and therefore inadequate for migration and spawning (sensu NOAA, 2009). This approach allowed us to specifically exclude dam impoundments from any analyses. Additionally we excluded lake and pond areas in the watershed assessments as American shad avoid lacustrine habitat (Stier and Crance, 1985). We then calculated cumulative habitat areas segmented by the presence of dams by combining our habitat flowline data with a congruent dam geodatabase compiled from multiple sources (Martin and Apse, 2011;Martin, 2013Martin, , 2019Natural Resources Canada, 2019). We summed habitat area for all reach segments upstream of each dam point, in addition to each coastal outlet point of streams and rivers (Figure 1). These point data were also vetted by local experts. Sums were taken iteratively by starting at the reach segment where the point was located, then adding upstream neighbor reaches until upstream dam points or headwater points were found. This search was recursive to avoid summing overestimations due to downstream flow bifurcations (Figure 2). Finally, we tagged individual streams and rivers with common names used by the Atlantic States Marine Fisheries Commission (Atlantic States Marine Fisheries Commission (ASMFC), 2007) for United States rivers and best available names for Canadian Rivers (Natural Resources Canada, 2019). In the end, this analysis produced a range wide assessment of putative American shad habitat as influenced by each impoundment found within the identified area. Data are compiled for the three eco-regions based on regional spawning strategies (Northern Iteroparous [NI], Southern Iteroparous [SI], and Semelparous [SM]) as delineated by Hasselman et al. (2013) and adopted by Atlantic States Marine Fisheries Commission (ASMFC) (2020).

Population Model
In order to assess the theoretical impact of dams on American shad range wide, we simulated the population potential for each of 164 identified systems. This approach allowed us to compare three broad scale scenarios (i) historical or intact "Undammed" rivers, (ii) "Dammed", with current dams in place with no fish passage, and (iii) dams with favorable upstream and downstream passage to reflect the "Current" condition ( Figure 3). All modeling routines were implemented in the 'anadrofish' package (Stich et al., 2020) for R (R Core Team, 2019) to provide a user interface for reproducibility and further exploration.
The river specific population modeling effort was based on Stich et al. (2019), who applied a stochastic life-history based simulation model to assess the theoretical effects of dam passage and migratory delay on abundance, population demographics, and spatial distribution of spawning adults over time. However, we followed Harris and Hightower (2012) and Bailey and Zydlewski (2013) in adopting an age-structured (rather than individual) approach to migration dynamics to generalize the modeling framework across the known range for American shad. This general model structure facilitated the incorporation of geographically appropriate life history parameters. Because these fish are iteroparous in the northern extent of their range and semelparous in the southern extent, downstream migration of both juveniles and adults was considered important for coastwide population dynamics. Each river was identified as being in the NI, SI or SM region and assigned life history parameters estimated for each eco-region (Atlantic States Marine Fisheries Commission (ASMFC), 2020). As this population model has been described previously, the description and equations are included as Supplementary Material 1 and available as open source code 1 .
Dam passage probabilities were assigned based on each of the passage scenarios. For the Undammed scenario, the dams exerted no influence on passage (upstream and downstream passage probabilities were 1.00), whereas for the Dammed scenario, upstream passage probability was set to zero. For the Current scenario, upstream passage was assigned based on available estimates (Table 1), assuming a passage probability of 0.40 (the unweighted mean of reported values). Separate downstream survival probabilities were used for adults (S DA = 0.80) and juveniles (S DJ = 0.95) and were based on available estimates or recent modeling efforts (e.g., USFWS, 2019). These values reflect reportedly "excellent" upstream passage rates for American shad  and are intended to represent the most optimistic assessment of American shad passage at all dams (even when there are fish ways). Downstream values are averages of reported values from sources in Table 1.
For each scenario, upstream and downstream passage probabilities were fixed across all dams. In all cases, upstream passage probabilities restricted the number of adults reaching spawning areas while downstream passage rates were applied as cumulative, catchment-wide mortality risks for juveniles and FIGURE 2 | Example of our approach to habitat area calculation. (A) Habitat was calculated for all reach segments. We then summed habitat groups that occurred between point features (e.g., habitat between a coastal outlet point and the first upstream dams in that river, habitat between a dam and other dams immediately upstream). (B) Rivers that bifurcated along the coast were summed as a single habitat group due to their upstream connection. Reach segments deemed inaccessible or unsuitable by experts were excluded from the habitat calculation, in addition to pond and lake areas. (C) We included downstream bifurcations in our habitat group calculations. Dams located on a bifurcation obstructed habitat only to the end of the bifurcation when other dominant flow paths were unobstructed. adults based on their locations above dams. The probability of fish reaching each habitat segment (P Access ) was calculated as the cumulative product of P UP through the number of dams (ORDER) downstream of that segment (P ORDER i UP ). The amount of available habitat (km 2 ) in each i th habitat segment (H S ) was pro-rated based on this product, and the resulting "functional" habitat (H F ) was summed throughout the catchment to yield over n habitat units an estimate of total habitat (H T ): The available habitat for returning adults and the simulated population of spawning fish were used to develop a river-and scenario-specific life history-based population model. Survival of adult fish to spawn was randomly drawn for each simulation. We used a Beta distribution (a = 90 and b = 10) to achieve a distribution of pre-spawn survival (mean of 0.90, standard deviation of 0.025, constrained on the interval [0, 1]). Based on the scenario-specific habitat (area) available in each river, we applied carrying capacity (k) to catchment-wide larval production (r) from a vector of age-structured spawners (s) using a Beverton and Holt (1957) recruitment curve with density dependence and a multiplicative error structure: The density dependent parameter (b ) was tuned to impose a k of 24,711 adult fish per square kilometer (100 adult fish/acre), an often used production potential for stock assessment (Atlantic States Marine Fisheries Commission (ASMFC), 2020) based on Stevenson (1899). This resulted in b = 0.340297 when mean fecundity was used for the value of a (density-independent parameter, specified as F t ): Because H T was summed across all units, this approach includes the implicit assumption that spawning fish distribute in the river according to proportional availability of habitat at the reach scale. Likewise, this assumption is also implicit in density-dependent recruitment due to the inclusion of H T in the denominator of the function describing b (density dependence in larval recruitment at the catchment level). FIGURE 3 | Conceptual diagrams of the three scenarios used to assess the impact of lost habitat and reduced connectivity on the coast-wide production potential of American shad due to dams. The Undammed scenario represents habitat accessibility in the historically undammed rivers. The Dammed scenario depicts the access to habitat that is not impacted by extant impoundments (assuming worst case, non-existing passage). The Current scenario depicts the status quo, where the current array of dams is in place, but upstream and downstream passage is applied at levels that represent the "best case" for American shad.
We assigned the larval-to-outmigrant survival (S J ) as a random draw from a normal distribution with a mean of 0.007535 and a standard deviation of 0.04 [70-d survival based on daily mortality rates reported by Crecco et al. (1983)] to estimate juvenile migrants from each river (N juv ) for each habitat unit. This quantity was simulated on the log e -scale and back-transformed to avoid negative values. Adults incurred post-spawn mortality based on the difference between natural mortality (M) and the projected degree of iteroparity as derived from Leggett and Carscadden (1978) and applied by Bailey and Zydlewski (2013). Iteroparity (I) was simulated from riverspecific latitude using the equation developed by Leggett and Carscadden (1978): We assumed that natural annual survival (S = 1 -A) was the joint product of surviving the spawning period (S post ) and surviving the duration of the year. We assumed there was no fishing mortality in all systems. Therefore, we simulated postspawn survival as the quotient of I and S: For populations in which I was predicted to be negative based on the equation from Leggett and Carscadden (1978) we set I = 0, and therefore S post was likewise zero (semelparous populations in extreme southern extent). Likewise, if the value of S post was greater than one, we set S post = 1 (iteroparous populations in extreme northern extent).
For an assumed 2-month residence time in freshwater (based on Leggett, 1972 [40-100 days] and Chittenden, 1976 [ > 60 days]; Table 1), survival in freshwater during the spawning run (S spawn ) was calculated as the product of two components: Where S 2month is 2-month interval survival (S) applied as an instantaneous rate of mortality over 2 months and calculated by using the relationship between annual interval mortality (A) and instantaneous mortality Z 12 (for a 12-month period; Miranda et al., 2007) to calculate Z for a 2-month interval (Z 2 ) as: So that monthly interval survival is calculated as: For semelparous populations, S spawn is determined by S post (which is set to zero), while for the northern iteroparous populations, in river mortality is determined by natural annual survival (S 2month ) that is the same survival as for fish remaining in the ocean. Mortalities were imposed at the end of this period and therefore did not influence spawning.
Downstream mortality of juveniles or adults through dams was based on proportional distribution of available habitat as appropriate to each scenario. The total mortality incurred through downstream dam passage was estimated as a function of the imposed downstream survival probability through dams (for both juvenile [S DJ ] and adults [S DA ]) and the proportional distribution of available habitat with respect to dam order. First,   downstream survival from each i th river segment to the ocean was calculated as the cumulative probability of downstream dam survival (S DC i ) separately for juveniles (S DC J ) and adults (S DC A ) based on the number of dams downstream of each segment. Next, the total proportion of accessible habitat (P H i ) in each i th river segment was calculated as the quotient of prorated habitat in each segment (H S−P ) and the sum of all available habitat: We then calculated the catchment-wide survival during downstream migration for adults, or juveniles as the weighted sum of S DC i (sum of the products of P H i and S DC i ): Age-structured out-migrants were then added back into the non-spawning (ocean) population, the population was projected one time-step based on instantaneous mortality (M in the absence of fishery impacts or Z otherwise) and the simulation continued for a total of 50 years to ensure stabilization of abundance estimates within rivers. We repeated this simulation for each river and each passage scenario by randomly sampling the river, passage scenario (Dammed, Undammed, Current) and each of the stochastic parameters during each simulation. This resulted in approximately 2,000 simulations per passage scenario per river. For each of the 164 river and scenario combinations, the output for the last year (50) was collected for the number of age specific spawners and juvenile emigrants.

Biomass and Nutrient Delivery
To estimate changes in biomass, total N and total P delivery to the freshwater environment (by adults) and to the marine environment (by juveniles), we used the average outputs from the simulations from each river. The average age specific numbers of spawners returning to each river and passage scenario combination were used to estimate the total spawner biomass seasonally entering each river: where P is the proportion of male (M) or females (F), N is the average number of spawners of age t and W t ,F or W t,M is the sex specific mass at age t. Sex-specific mass for an individual fish was calculated by calculating average length at age and mass as a function of length described in Supplementary Equations 4, 5, using region-specific parameters. Juvenile biomass leaving each river was calculated as the summation of the number of juveniles produced from all habitat units in each river (N JUV ) multiplied by the average, weighted survival downstream to the ocean (S DT J , Equation 13).

Juvenile Biomass River
The total river output (JuvenileBiomass River ) was simply the sum of juvenile biomass reaching the ocean from each habitat unit. The estimated mass of a juvenile emigrant, W JUV , was based on reported juvenile masses ranging from 2.0 to 4.5 g (Zydlewski and McCormick, 1997a;Haskell, 2018); we applied an average W JUV of 3.36 g ( Table 2).
The delivery (and retention) of marine derived biomass (Biomass delivered ) and nutrients into each river during the spawning migration was estimated as a function of: (i) the initial biomass of adults arriving in freshwater (SpawnerBiomass River ), (ii) the probability of death during spawning, and (iii) if surviving and returning to the ocean (for rivers in the NI and SI regions), the estimated biomass lost during the spawning migration through spawning and metabolic loss.
Biomass metabolic is the sum of metabolism expenditures in freshwater and shed biomass in the form of gametes for spawners who successfully enter a river as a presumptive spawner and return to the ocean. This biomass lost in river by surviving fish (gonadal and metabolic) is calculated as: All parameters are defined as described in Equation 17. Biomass lost during the spawning run for surviving migrants ( ) is the proportional sex specific loss of mass due to metabolic expenditure and gametic release. These values are based on empirical data reported by Leggett (1972), Chittenden (1976), Raabe and Hightower (2014), and Haskell (2018) that range 18-51% observed mass loss for males and 33-57% for females. We applied average values of 36 % and 46% mass loss for males ( M ) and females ( F ) respectively ( Table 2). We acknowledge this as a simplifying assumption as it is well-known that distance traveled (Leonard and McCormick, 1999), temperature experience (Glebe and Leggett, 1981b;Raabe and Hightower, 2014) and residence time (Raabe and Hightower, 2014) all influence the extent of individual mass lost in freshwater.
Marine derived nutrient transport (phosphorus or nitrogen) was calculated using estimated nutrient density (grams of nutrient per gram of fish wet mass) applied to estimated of Biomass delivered . Several simplifying assumptions were applied. We used nutrient density values 0.670 g/g wet mass for phosphorus and 2.950 g/g wet mass for nitrogen (based on averaged male and female data from Haskell, 2018). We assumed nutrient density was the same for males and females, and across tissues (i.e., nutrient density for gonads, metabolic tissue loss and whole-body densities did not differ). Thus, pre-spawn estimates of nutrient density are assumed to be representative of the nutrients shed through gonadal and metabolic loss. This is generally the case as whole body nutrient densities differ by < 1% from male to female (Haskell, 2018). For the gonads, however, we note that the testes are enriched in both N and P (∼10 % over other tissues) while the ovaries are enriched in N (∼10%) but depleted in P (∼50%), resulting in an over estimate for P. However, because female gonadal mass represents approximately 10 % of the mass of a prespawn female (Leonard and McCormick, 1999), and the applied mass loss based on observations is more than four-fold greater (46% mass loss during spawning), the influence on phosphorus estimates is minimal.

Habitat
Historic American shad habitat was estimated to be 2,287 km 2 for the entire range and comprised 164 rivers that were identified to have suitable habitat based on the identified criteria. This assessment was based on a combination of a priori assumptions and informed by local knowledge through most of the United States and part of Canada. For some of the assessment in Canada, estimates are based solely on the a priori model and our inability to locate appropriate experts. The habitat was divided fairly evenly among eco-regions with 30, 38, and 32% identified in the NI, SI and SM eco-regions, respectively (Figure 4). In the NI eco-region, 18% of the habitat was found in Canada while 12% was found in the United States (see also The presence of dams on 68 rivers throughout the range of American shad fully or partially blocks access to an estimated 41% of their coast-wide habitat (Figure 4). The constraint of access to habitat found upstream of at least one barrier is comparably high for all eco-regions, with 37, 44, and 39% loss of habitat due to first main-stem dams for NI, SI and SM, respectively. It is notable that habitat in the Canadian jurisdiction of the NI eco-region has remained comparably intact with an estimated 19% loss. In contrast, the United States NI region is heavily impounded, with a loss of connectivity to 65% of the habitat.
Throughout their range, ten rivers account for more than half (52%) of riverine habitat of American shad ( Figure 5; Table in Supplementary Material 2). Dams on these rivers have resulted in the reduction of habitat access from a minimum of 14% loss in the Delaware River to complete habitat blockage in the Congaree River. Dam blockage at these ten rivers accounts for the loss FIGURE 4 | Graphic representation of the Undammed (outer) and Dammed (inner) scenario freshwater spawning habitat available to American shad throughout their native range. The size of the circles is proportional to the area of historic unimpounded habitat (2,752 km 2 ) and current habitat access (1,639 km 2 ), indicating that American shad are impeded from reaching nearly 41% of their historic habitat. of 575 km 2 of habitat, nearly 62% of the coast-wide loss in habitat for this species. On the Susquehanna River alone, dams block 243 km 2 of habitat, 10.6% of total historically accessible habitat coast-wide.

Population
Historic American shad population potential was estimated to be 69.1 million adults for the entire range. While historic habitat area is roughly equal for each of the eco-regions, the modeled spawner density is heavily skewed to the NI and SI eco-regions. We calculated an estimated historic potential of 37.5 million (54%) and 27.2 million (39%) spawners for the NI and SI regions, respectively (Figure 6). Historical abundance in the Canadian NI ecoregion was estimated at 24.0 million fish (35% of the coastal run) while the United States portion of the NI eco-region was 13.6 million (20% of the coastal run; See also Supplementary Material 2). The entire SM was estimated at only 4.3 million spawning fish for the entire region, accounting for only 6% of the spawner potential. When considering the potential for all 164 rivers to support spawning runs of adult American shad, estimates of spawner potential are directly linked to estimates of available habitat (Figure 7). However, latitudinal variation in growth, maturation, and post-spawn survival (Leggett and Carscadden, 1978;Gilligan-Lunda et al., 2021) all influence the relative differences observed in population potential. On average, a square km of habitat results in 55,200 spawners in the NI eco-region but only 31,000 in the SI eco-region. Because of the lack of additive age classes in the SM eco-region (i.e., no repeat spawners), a square kilometer of area is modeled to support only 5,900 spawners.
Coast-wide, there is an estimated 39% loss in population potential, directly attributable to lost habitat between the Undammed and Dammed scenarios (Figure 8). Regional population potential is diminished by 35, 46, and 40% reductions in abundance for NI, SI and SM regions, respectively. Population potential loss in the Canadian jurisdiction of the NI eco-region is 18% while the United States NI region is reduced by 65%.
Under the Current scenario, where we applied upstream passage probabilities that represent some of the most favorable conditions reported in the literature, there was limited increase in spawning potential of 9% coast-wide from the Dammed scenario (from 41.8 to 44.6 million). Thus, the most optimistic application of dam passage parameters to each of the hundreds of dams that block access to American spawning habitat provides a theoretical benefit of restoring only 4% of the historic spawning potential that is diminished by 39% due to dams. This limited alleviation of population potential influence is lowest in the SI eco-region (2.7% increase), and is comparable in the NI (4.9%) and SM (5.4%). Passage mediated alleviation of population potential in the Canadian jurisdiction of the NI eco-region is 4.1% while the United States NI region is 6.2%, suggesting a greater potential for recovery.

Biomass and Nutrients
Historic American shad populations supported an estimated average of 69.1 million spawners coast-wide, delivering 98,000 FIGURE 5 | Total habitat available to American shad upstream and downstream of the first dam by river throughout their native range. Dotted lines indicate transitions between eco-regions. Rivers (y-axis) are ordered in descending latitude.
MT of biomass to undammed freshwater coastal systems on an annual basis. We estimate that 59,100 MT (more than 60%) of the delivered biomass remained in the river systems in the form of carcasses, tissues shed during spawning, and metabolic loss. Historic biomass delivered to the NI was 51,500 MT (with 25,800 MT retained) and 41,200 MT delivered to the SI ecoregion (with 35,500 MT retained; Figure 9). For the SM ecoregion, a much smaller 5,200 MT biomass was delivered, but all would be retained as part of the semelparous life history in the region. Differences among eco-regions are driven by competing trends of lower potential spawner numbers in the southern latitudes, greater spawner number due to iteroparity in the north and reduced individual spawner sizes (due to a truncated age distribution as repeat spawning declines in the SI and the SM eco-regions).
Under an Undammed scenario, juvenile migration from fresh water transports an estimated 6,600 MT of biomass to the ocean coast-wide, with comparable transport from each of the eco-regions (1,900, 2,900 and 1,800 MT respectively for NI, SI and SM eco-regions). Estimated biomass export to the ocean is considerably lower than the biomass import into freshwater. We estimate 13-fold greater import in the NI eco-region, 12-fold in the SI, but only 3-fold greater in the SM (Figure 9).
For the Dammed and the Current scenarios, the relative relations between import and export remain similar among eco-regions and coast-wide, however the magnitudes of both import and export reflect the reduced population under habitat access restriction. Coast-wide, retained biomass is reduced from 59,100 MT for the No Dam scenario to 38,900 MT for the No Passage scenario. Under the Current Scenario, biomass delivered FIGURE 6 | Modeled spawner abundance potential of American shad under the no passage ("Dammed"), status quo ("Current"), and no dam ("Undammed") passage scenarios by eco-region through their native range from Canada to Florida.
is slightly increased to 41,400 MT. These represent a 34% and 30% reduction in biomass delivered for the No Passage and Current scenarios respectively, compared to the No Dam scenario. Similar reductions in retained biomass import for the Dammed (37, 45, and 40%) and Current scenarios (32, 42, and 34%) were observed in each of the three eco-regions (NI, SI, and SM). Juvenile nutrient export was reduced from 6,600 to 3,500 MT for the Dammed scenario, and modestly recovered to 3,900 MT under the Current scenario (37 and 42% reductions, respectively).
Because the inputs we used for nitrogen and phosphorus densities are not markedly different between males, females and juveniles (  Figure 9). Parallel reduction in retained P delivery is reduced from 446 to 260 MT for the Dammed scenario (108, 131, and 21 MT for NI, SI and SM), and boosted to 277 MT under the Current scenarios (117, 137, and 23 MT for NI, SI and SM; Figure 9). Juvenile nitrogen export was reduced from 186 to 110 for the Dammed scenario (34, 46, and 30 for NI, SI and SM), and grew to 119 MT under the Current scenario (38, 48, and 33 for NI, SI and SM). Juvenile phosphorus export was reduced from 59 to 35 MT for the Dammed scenario (11,14,and 9 for NI,SI and SM), and grew to 37 MT under the Current scenario (12, 15, 10 for NI, SI and SM).

DISCUSSION
Based on a priori modeling and local knowledge, the historic habitat availability for American shad spanning the east coast of North America approached 2,300 km 2 . While our estimation was derived from 164 coastal rivers, it is likely that American shad may spawn in additional smaller systems, though these would contribute minimally to the overall coastal population. The estimate of 41% loss of habitat connectivity through damming of coastal systems corroborates assertions that loss of spawning habitat access is likely the major cause of population declines in this species (Limburg et al., 2003;Atlantic States Marine Fisheries Commission (ASMFC), 2020). The proportional distribution of accessible habitat is relatively constant among the three eco-regions (when governmental borders are not considered). This indicates that, while impacts are regionally variable, dam construction has negatively influenced habitat access by American shad, similarly, throughout their range.
Because the recruitment function we used incorporates habitat area directly (see Stich et al., 2019), it is unsurprising that theoretical capacity to support runs of adult American shad into rivers mirrors this assumption. Despite the fact that total available habitat was divided fairly evenly among eco-regions, the relative scope for population size is substantially greater in the NI, decreased in the SI and lowest in the SM ecoregion due to differences in life history. While this is intuitive based on the degree of iteroparity, growth rates, and maximum sizes (Gilligan-Lunda et al., 2021), it highlights the difference between relative and absolute losses of spawner potential due to dams. The SM eco-region's lost scope of 1.74 million fish is dwarfed by the loss of more than 13 million in the NI ecoregion, suggesting that parallel differences in fisheries potential also exist. While the historic spawner production potential is higher in the north and lower to the south, it is notable that each of the regions generally lost comparably high proportions of production habitat. Thus, the fish in southern rivers may be viewed as having the highest conservation vulnerability, particularly at the southern end of the range where climate change may impose additional ecological constraints. Based on this modeling exercise, coast-wide production potential is more than 69.1 million spawners per year without dams compared with the Dammed scenario of just under 41.8 million spawners, a reduction of 39 %.
The intent is of this modeling exercise has been to provide a realistic quantification of the lost potential for American shad that has altered coastal ecosystems and reduced commercial and recreational fisheries opportunities over the species' native range. We paint a clear picture of the relative loss of potential coast-wide, but a cautionary note is warranted. Population estimation is a challenge even when the input parameters are well characterized, and that is not the case here. We are aware that American shad select areas for spawning based on temperature, water velocity, depth and substrate FIGURE 7 | Estimated spawning potential of American shad for each river through their native range from Canada to Florida. Dotted lines indicate transitions between Eco-regions regions. Rivers (y-axes) are ordered in descending latitude. The stacked bars depict estimated potential for the No Passage, Current and Undammed scenarios. , evidence that some habitats are more suitable than others. Additionally, juvenile alosine recruitment from reaches of a watershed likely differs based on migratory distance and conditions (Tommasi et al., 2015), yet we have made the implicit assumption of equal habitat quality. Models such as the ones presented here are built with considerable uncertainty, and assumptions in the model parameters. This has been well highlighted here, and elsewhere (Castro-Santos and Letcher, 2010;Bailey and Zydlewski, 2013;Stich et al., 2019). For American shad, there remain many fundamental relations that are poorly characterized (e.g., passage, reproduction, survival and juvenile recruitment), a condition common in such model efforts (Goethel et al., 2011).
It is self-evident that the parameterization of the system we imposed defined the boundaries of our outcomes, and we anticipate that continued research efforts will aid in refining our estimates. However, our estimates are likely to be conservatively low based on model construction, parameter values applied, and assumptions made. Specifically, the Beverton-Holt recruitment curve we have employed is derived from Stevenson (1899), an approach that may grossly under estimate capacity. These cautions withstanding, the values of spawners we generated are realistic in magnitude and represent the best available information. Our estimate of 98,000 MT of spawners returning to all rivers is congruent with historic commercial fisheries landings that exceeded 20,000 MT annually (Walburg and Nichols, 1967;Limburg et al., 2003). Undammed potentials for specific rivers (Supplementary Material 2) are generally consistent with historic annual landings harvest from Walburg and Nichols (1967), the data suggest a high fisheries exploitation rate of FIGURE 8 | Coast-wide spawning potential of American shad under the no passage ("Dammed"), status quo ("Current"), and no dam ("Undammed") passage scenarios. Boxes indicate 25th to 75th interquartile range and whiskers indicate 95% CI.
nearly 20% of the average entire run in the Neuse, James, and Delaware Rivers, 50% in the Potomac and 70% in the York River. A reported harvest of over 2 million adults in the Penobscot River in the 1860s (Foster and Atkins, 1867) would represent a harvest of 65% (of our average estimated run of 3.1 million fish). Harvest data from other rivers indicate that our model may significantly undervalue spawner potential. Reported annual harvest in the Nanticote and Choptank Rivers (Walburg and Nichols, 1967) were two-and three-fold greater than our average undammed population projections. Such difference are undoubtedly due to both population stochasticity and model underestimation. Thus, the most robust value of this exercise is the relative influence of dams which provides an unambiguous relative index of the impact of impaired habitat access.
The inability for current realistic fish passage measures to restore any more than 9% of the estimated spawning potential coast-wide is troubling news for the restoration of these fish. Our results indicate that the application of current upstream and downstream passage rates at all dams affords a remarkably small increase in the theoretical production potential relative to rivers that are wholly inaccessible upstream of the first dam. We estimate that fishway passage coast-wide at dams represents a fixed constraint of about 35% on the spawning run potential of American shad. It is possible that as advances in fish passage engineering, other protective measures, and understanding of fish behavior continue to evolve, passage efficacy may improve beyond our optimistic estimates. The use of fish passage performance criteria (e.g., Stich et al., 2019;CRASC, 2020) may also facilitate fish passage improvements by providing biologically relevant targets. These potential advances withstanding, the low theoretical return on investment of fishways is heavily influenced by the presence of multiple dams on rivers, resulting in a compounding influence on passage and survival. FIGURE 9 | Estimates of annual import (IMP: positive from the zero line) and export (EXP: negative from the zero line) of biomass, nitrogen and phosphorus delivered to each eco-region by American shad through their native range from Canada to Florida under the no passage ("Dammed"), status quo ("Current"), and no dam ("Undammed") passage scenarios.
It is notable that this significant imposition on theoretical spawning potential results from a model that estimates contemporary passage values considered to be "excellent" compared to those typically achieved . True impacts of dams in systems with poor fish passage may be -and are likely to be -greater. Even when significant engineering efforts are made to construct fishways to allow upstream migration, effective connection to upstream habitat can be illusory (Noonan et al., 2012;Bunt et al., 2012Bunt et al., , 2016. It is notoriously challenging to provide safe, effective and timely upstream passage past dams for American shad, and the poor passage performance of this species has long been noted (Stevenson, 1899). Challenges associated with dams (beyond passage and survival) have not been incorporated in our model. Flow disruptions, temperature, and changed community composition could further increase the negative impacts of dams on American shad populations. Loss through coastal commercial and recreational fisheries continues in states where Fisheries Management Plans (or alternatives) are in place (Atlantic States Marine Fisheries Commission (ASMFC), 2020). Lastly, other factors (e.g., habitat degradation) may impose further loss of production potential on river-and region-specific levels.
Fishway design continues to be a critical interface of biology and engineering with significant advances in concepts and application (Silva et al., 2018). However, though fishways have been constructed at some of the numerous dams that many American shad encounter in their migration, many, if not most, are largely or wholly ineffective . Other dams simply have no specific passage accommodations (e.g., Roanoke River; Hightower and Sparks, 2003), yet for this exercise we attributed optimistic values of fishway performance. Because the time course for hydropower relicensing through the FERC process is 40 years or more (Vogel and Jansujwicz, 2021) the current state of fish passage has a long sustained effect. Managers may only experience one opportunity to shape passage at a dam once during their professional careers. For these reasons, poor utilization of fishways is likely to remain a substantial, if not dominant, culprit for diminishing American shad spawning runs (Atlantic States Marine Fisheries Commission (ASMFC), 2020).
Fish passage is often assessed as a binary event, but even successful dam passage may still exact a toll that is not revealed through modeling. Even when fish passage opportunities are available, dams restrict the timing of access to spawning habitat (e.g., Grote et al., 2014). Dams may therefore increase a migrant's vulnerability to predation (Andrews et al., 2018) or a fishery (Atlantic States Marine Fisheries Commission (ASMFC), 2020), as well as depleting energy stores by imposing delay. The sustained delay below a dam may, however, may preclude successfully passing fish from reaching their reproductive potential (Leonard and McCormick, 1999). Among teleosts, American shad display an elevated basal metabolic rate that is likely causal to their upstream migration being more energetically costly than anadromous salmon (Leonard and McCormick, 1999). Metabolism is determined by water temperature through its effects on enzymatic, metabolic, and cardiac processes (Lennox et al., 2018) so that accelerated energetic depletion occurs in the warmed water near dams (Martin et al., 2015).
Reduction in energy reserves may result in reduced individual fitness (Nadeau et al., 2010) by impairing spawning behaviors and gamete production (Brett, 1962;Rand and Hinch, 1998;Martin et al., 2015). Migrating adults passing a dam do experience an advantage in gaining access to habitat with fewer conspecific competitors. This may afford an advantage in cases where the juvenile rearing carrying capacity is reached, resulting in greater per capita recruitment to the juvenile stage (Walters and Korman, 1999). This advantage, if realized, comes at a significant cost for American shad in the iteroparous ecoregions. In order for spawning adults to spawn again, they must retain enough energy to reach the ocean. Fish must not only reach their spawning habitat in time to spawn, but also must maintain energy stores sufficient to return to the marine environment where they will recommence feeding and growth (Doucett et al., 1999). After spawning, adults leave spawning areas and move into the estuary (Grote et al., 2014) to feed and recondition (Walter and Olney, 2003). Delays at dams are commonly observed in downstream migrating adult shad (Weaver et al., 2019) and undoubtedly exacerbate energy loss at a time when stores are at their lowest. Impediments to the resumption of feeding may therefore increase mortality through prolonged energy loss (Castro-Santos and Letcher, 2010) or reduced performance (Leggett, 1972;Chittenden, 1976;Raabe and Hightower, 2014). Based on comparisons of iteroparous and semelparous populations, it has been suggested that a depletion of greater than 60% of energy reserves may serve as a constraint to an iteroparous life history (Glebe and Leggett, 1981b), while the threshold for post-spawn survival may be as low as 30-40% (Leonard and McCormick, 1999). In addition to a physiological constraint, these fish face high risks of mortality passing dams, particularly if passing through a turbine or other unsafe route (Bell and Kynard, 1985;Hogans and Melvin, 1985;Haro and Castro-Santos, 2012).
The population impact of an impoundment may be amplified by a sequence of dams that must be navigated twice by iteroparous adults to reach spawning habitat, thereby imposing compounding mortality risk on spawners. It is useful to differentiate between the latitudinal cline that is observed in American shad (semelparity in the south to iteroparity in the north) from reduced post-spawn adult survival in the southern and northern iteroparous eco-regions. The degree to which these life history differences reflect population level characteristics versus phenotypic plasticity is poorly characterized. Populations accessing suitable spawning habitat through anthropogenic facilities are subject to an adult mortality akin to an intercept fishery on size and age distributions and results in the systematic loss of "big old fat female fish" (Hixon et al., 2014). This has been well characterized in the Connecticut River where the proportion of repeat spawners declined from 49% in the late 1950s (Walburg and Nichols, 1967;Carscadden and Leggett, 1975;Limburg et al., 2003), to a mean of 5 % for the period 2006-2015 (Atlantic States Marine Fisheries Commission (ASMFC), 2020). Failure to recognize the important role of passage has led to the unlikely assertion that upstream passage may be detrimental to a population because of energetic constraints (Leggett et al., 2004) rather than poor downstream passage. In order to repeat spawn, adults in the SI and NI eco-regions need both access to habitat and an effective exit strategy or "forced semelparity" may result.
Whether parents are semelparous or iteroparous, the progeny of American shad spawned upstream of dams suffer the risk of mortality when moving downstream. Young American shad generally remain in freshwater until migrating downstream in the fall associated with declining river temperature (Leggett and Whitney, 1972;O'Leary and Kynard, 1986). Downstream passage at dams may entail delay, confinement and turbulence during a period of time when these fish exhibit a heightened sensitivity to stress (Shrimpton et al., 2001) and an environmentally influenced loss of ion regulatory ability in freshwater (Zydlewski and McCormick, 1997b). Juvenile shad, being far smaller than adults, are generally not effectively excluded from turbines. These fish have less individual risk than entrained adults (Heisey et al., 1992;Mathur et al., 1994) though it is biologically significant. The impacts of dams through injury or delay may shape the disposition of migrants. Their success (or failure) depends on phenology of their physiological development (Zydlewski et al., 2003) in conjunction with biotic and abiotic environmental factors (McCormick et al., 1998;Limburg, 2001).
Prior to the construction of dams range wide, American shad delivered considerable biomass and nutrient loads to freshwater systems (Limburg et al., 2003 ; Figure 9). It is difficult to fully appreciate the ecological significance or magnitude of delivery of 1,963 MT of nitrogen and 446 MT of phosphorus on an annual basis. The seasonal influx driven by the carcasses, gametic release and metabolism would undoubtedly change the freshwater system as has been observed for other species (Weaver et al., 2018). The estimated 34% reduction in nutrient delivery associated with dams is significant. As Haskell (2018) asserts, for American shad, spatial context is of particular importance because shad are broadcast spawners using open water rather than being associated with substrate as is the case for sea lamprey (Weaver et al., 2018) and salmon (Rex and Petticrew, 2008). Spawner distribution in a natural river results in distribution that is punctuated over time and space. Therefore the reduction in spawner potential associated with dams is not only a reduction in the magnitude of nutrient delivered, but a curtailment of the spatial distribution of marine derived nutrient incursions.
In the context of nutrient spiraling theory, in streams and rivers, downstream nutrient changes are driven by delivery, uptake and flushing (Newbold et al., 1982). As such, American shad represent only one broken link in the chain that has historically connected the marine and freshwater ecosystems through the seasons (Limburg et al., 2003). The interruption of ecological connectivity is common for many coastal ecosystems; in the Pacific Northwest nutrient delivery by Pacific salmon is estimated to be only 6-7 % of historical levels (Gresh et al., 2000). The nature of these connections likely changes over the range of the American shad, as anadromous species have a greater scope for influence in freshwater systems with nutrient limitation (nitrogen, phosphorus or co-limited) that influences the degree of nutrient incorporation (Cederholm et al., 1999;Chaloner et al., 2002;Bellmore et al., 2014;Samways and Cunjak, 2015). Increased N and P availability leads to increased primary production, but these effects may be swamped by nutrients generated from anthropogenic land use practices (Twining et al., 2013). The nutrient dynamics of American shad are less comprehensively explored than other alosines (Durbin et al., 1979;Post and Walters, 2009;Walters et al., 2009;West et al., 2010) offering an important direction of future research.
The presence of dams greatly curtails estimated nutrient transport (both import and export) and, in parallel with spawner estimates, fish passage provides a low level of restoration of this ecological function. The dynamics and magnitude of nutrient exchange are strongly influenced by population size and life history differences among the three eco-regions in the natural range. Our models project that in the NI and SI ranges, where iteroparity results in greater spawner potential, delivery is tilted towards import. Haskell (2018) also assessed American shad to be net importers of N and P in the Columbia River where iteroparity levels are 32% (Petersen et al., 2003). In the SM eco-region, despite the higher proportion of carcasses delivered during migration, we estimate that the net flux of nutrients in the southern part of their range to be significantly lower than the northern regions. The nature of delivery among regions differs as well. Carcasses of migrating fish can break down over days and weeks (Garman, 1992;Weaver et al., 2015) so nutrients are not immediately liberated whereas nutrients released via gametes and metabolic processes are more readily available.
Our presentation of average spawner escapement and juvenile transport likely obscures a more nuanced story of nutrient delivery that has been altered by the imposition of dams on the landscape. The complex role of nutrients delivered by alosine species is largely driven by levels of escapement (Barber et al., 2018), allowing locally disparate net nutrient balance. At high levels of spawner escapement, as occurs in our stabilized population models, biomass delivery (and resulting nitrogen and phosphorus delivery) are markedly skewed to import, particularly in the NI and SI regions (Figure 9) where spawner numbers are higher (Figure 6). In cases where adult dam passage is poor, the juvenile recruits produced per spawner can be expected to be higher than average, resulting in a more equitable exchange, or even a net export (Barber et al., 2018). In such cases, with reduced competition, larger migrating juveniles may further shift the net balance in favor of export (Moore and Schindler, 2004;Moore et al., 2011). Such a localized patchwork of nutrient balance may also occur in natural systems with impediments to migration (e.g., natural falls) or simply due to the attrition of spawners over the length of a river (Meixler et al., 2009;Hall et al., 2011;Pess et al., 2014). Even when a river is producing spawners, failure to connect these fish to their habitat can result in a lower import or net export in nutrient limited systems.
As juvenile American shad develop and grow, they are directly embedded into pelagic, littoral, benthic, and terrestrial systems (Limburg et al., 2003). Feeding -or being preyed upon -directly links these fish to food webs. As is the case for many anadromous species, juvenile American shad (and adults) may be directly consumed through aquatic (Willson and Halupka, 1995;Jaecks and Quinn, 2014), terrestrial and avian predators (Dalton et al., 2009;DeBruyne et al., 2012). Aquatic macroinvertebrates also actively feed on carcasses during their freshwater residency, serving as further conduit of nutrients between the terrestrial and aquatic environments (Polis et al., 1997;Vanni, 2002;Hocking and Reimchen, 2009). Both bottom-up and top-down pathways of nutrient incorporation may result from the influence that American shad likely have in the freshwater ecosystem.
The restriction of American shad to lower reaches of the coastal rivers is part of a fundamental shift in the riverine communities due to impoundment (Kiraly et al., 2015;Watson et al., 2018). Fish communities in these altered systems have both winners and losers, often favoring native and non-native "invaders" (sensu Carey et al., 2012) such as black bass (Micropterus spp.). By presenting an obstacle to migrations, dams may make American shad and other alosines increasingly vulnerable to novel predators (such blue catfish and [Ictalurus furcatus] and flathead catfish [Pylodictis olivarisast] in the Chesapeake Bay region; Schmitt et al., 2017).
In addition to the impacts on the freshwater environment, the dam mediated loss of connectivity also affects estuarine and coastal systems. The presence of adult and juvenile American shad in coastal systems benefit estuarine and marine organisms. The exodus of juvenile shad from freshwater is protracted from the summer into the fall (Williams and Bruger, 1972;O'Leary and Kynard, 1986;Zydlewski and McCormick, 1997b), overlapping with the migration of other alosines. Such migrations attract predators (Davis et al., 2012;McDermott et al., 2015), but at the same time reduce individual risk. Many predatory species depend on the seasonal pulses of prey species (Willson and Womble, 2006;Richardson et al., 2014;Furey and Hinch, 2017). As such, patterns in alosine migration may shape both current and historic distributions of marine fish species (Baird, 1883;Ames, 2004). Specialized detection of relatively high frequency sounds by American shad may also indicate evolutionary predatory pressures, suggesting that changes in abundance may influence the foraging success of echolocating marine mammals (Mann et al., 1998).
We have considered the influence of dams on American shad through their coastal range and used the best available data to quantify the dam-mediated impact on habitat loss, spawner production potential and nutrient transport range wide. The ecological potential and human value of these fish has been markedly reduced through the partial and complete occlusion of access to spawning and juvenile rearing habitat. Historically these fish have linked freshwater systems to the marine environment, from the mangrove estuaries of Florida to the boreal forests of Canada. More than 100 years ago Stevenson (1899) asserted that "There is no species of fish more important to residents of the Atlantic seaboard than the shad." The fisheries, and the human connections to the fish, have diminished in spite of fisheries closures and extensive passage efforts (Atlantic States Marine Fisheries Commission (ASMFC), 2020). The data suggest that dams remain the most significant impediment to restoration of the "founding fish" (McPhee, 2003). While contemporary passage rates fail to achieve even modest population recovery, dam removal appears to remain a viable solution (Raabe and Hightower, 2014;Izzo et al., 2016;Moser and Paradis, 2017;Watson et al., 2018). The recolonization of newly accessible habitat by migratory fish is well-documented (e.g., Burdick and Hightower, 2006;Hogg et al., 2013). While decision-making at dams involves a wide range of stakeholders with diverse and sometimes conflicting objectives (Roy et al., 2018), "active restoration" has been framed as a balanced approach that integrates both values and science (Hart et al., 2002). There is a growing appreciation for the biological and economic benefits of restoring coastal connectivity (Dias et al., 2019), thereby regaining that which we have lost.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author/s.

AUTHOR CONTRIBUTIONS
JZ led in the framing of the manuscript and compilation of data and led the writing. DS led the population modeling and was a major contributor to writing. SR led the GIS analysis and was a major contributor to writing. All authors contributed to the framing of the manuscript, interpretation and editorial review.

ACKNOWLEDGMENTS
Logistical support was provided by the United States Geological Survey Maine Cooperative Fish and Wildlife Research Unit. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the United States Government. The findings and conclusions in this article are those of the authors and do not necessarily represent the views of the U.S. Fish and Wildlife Service. All data generated or analyzed during this study are included in the main text or as Supplementary Information for this publication. We thank the Atlantic States Marine Fisheries Commission (ASMFC) and individuals from the ASMFC American Shad Technical Committee and Stock Assessment Subcommittee (SAS) members who contributed data and knowledge that has informed our efforts. We are exceptionally grateful to the dozens of biologists from both the United States and Canada who generously shared their local knowledge of the landscape, habitat use and history of American shad in their region. Only fear of accidental exclusion prevents the authors' attempt at listing the names of those who spoke for the American shad.