Imperiled sparrows can exhibit high nest survival despite atypical nest site selection in urban saltmarshes

. Understanding habitat selection and its fitness consequences in remnant populations of birds in fragmented and urbanized habitat may provide guidance to land managers for imperiled species conservation. We studied Saltmarsh Sparrow, Ammospiza caudacuta


INTRODUCTION
The relationship between avian nest survival and habitat selection has important implications for species distributions and for conservation planning (Jones 2001). Nest site selection is expected to be shaped by fitness within a particular ecosystem. When these two factors are no longer positively correlated, an ecological trap may result (Misenhelter and Rotenberry 2000). For altricial birds, fledgling production is often most influenced by availability of food for provisioning nestlings (Martin 1987, Holmes et al. 1992, Marshall and Cooper 2004, Post and Greenlaw 2006, coverings that conceal nests from predators Roper 1988, Liebezeit andGeorge 2002), or nest placement and structures that protect nestlings from abiotic factors Greenlaw 1982, Martin 2001). High quality habitat features are not always available to all individuals because of competition for limited nest sites by native (Dobkin et al. 1995) or introduced species (Schlaepfer et al. 2005, McChesney andAnderson 2015), or because of habitat fragmentation or degradation (Stephens et al. 2004, Berger-Tal andSaltz 2016). Self-sustaining populations may persist in patches of abject habitat if individuals can adjust their breeding strategies to novel habitat arrangements over time (Yeh et al. 2004, Salinas-Melgoza et al. 2013 or densely cluster in space in the highest quality portions of the habitat (Fretwell and Lucas 1969) while reproducing successfully.
Tidal marshes provide habitat for a suite of specialist nesting birds (Correll et al. 2016) and are experiencing rapid ecological change that could affect nest survival. Between the 1950s and 1970s, over 50% of tidal marshes along the Atlantic Coast of the United States were destroyed due to draining and filling for habitation and agriculture or channelized and treated with chemicals for mosquito reduction (Tiner 1984, Dahl 1990). These impacts were especially acute on Western Long Island, in which parts of New York City (NYC) lie. Here, marshes suffered losses of over 75% between 1900 and 1970 and continued to decline at rates of 0.5-3% per year (Hartig et al. 2002). These continued losses are due to a combination of factors including sea level rise (Gornitz et al. 2001, Hartig et al. 2002, increased tidal range (Swanson and Wilson 2008), and urban development, which has hardened shorelines and starved marshes of inorganic sediment making them fragile and prone to fragmentation (Peteet et al. 2018).
For tidal marsh specialist songbird species in northeastern North America, the largest reported influence on breeding failure is tidal inundation of ground nests (Gjerdrum et al. 2005, Shriver et al. 2007, Ruskin et al. 2017a, which often leads to multiple nesting attempts per season Greenlaw 1983, Shriver 2002). Because sea level rise amplifies tides and intensifies flooding, the nest building behavior of the tidal marsh obligate Saltmarsh Sparrow, Ammospiza caudacuta, and is put at great risk of extinction because of flooding-induced nesting failure Elphick 2011, Correll et al. 2017).
Saltmarsh Sparrow nest survival in large, relatively intact saltmarshes has been found to be positively correlated with ground elevation (DeRagon 1988, Shriver et al. 2007 and to benefit from nest initiation immediately following high spring tides (DeRagon 1988, Gjerdrum et al. 2005, Shriver et al. 2007, Walsh et al. 2016. Seaside Sparrows, A. maritima, found along the northeastern Atlantic Coast are also largely restricted to tidal marshes and nest sympatrically with Saltmarsh Sparrows from New Hampshire to Virginia. Seaside Sparrow nest survival has also been found to benefit from rapid re-nesting following loss to spring tides in some studies (DeRagon 1988, Marshall andReinert 1990) and by height and density of the vegetation surrounding the nest in other studies (Gjerdrum et al. 2005, Hunter et al. 2016. Seaside Sparrows are potentially afforded greater variability in nest height selection than Saltmarsh Sparrows because of their propensity to nest among taller, more rigid, low-elevation, Spartina alterniflora, grasses. Seaside Sparrows are also likely to have access to suitable nest vegetation later into the progression of sea level rise than Saltmarsh Sparrows (Field et al. 2016) because low-marsh vegetation is expected to increase in prevalence on the landscape as the climate changes because of its ability to survive tidal flooding. High-marsh zones selected by nesting Saltmarsh Sparrows may not persevere because marsh migration into current uplands may not be able to keep up with the rate of sea level rise in some areas (Watson et al. 2016). Although in many areas, high-marsh vegetation is likely to be prevented from landward migration by urban or forested borders (Field et al. 2016) or steeply sloping uplands (Kirwan et al. 2016). This combination will result in the conversion of much high marsh to low marsh, without concomitant replacement in the uplands.
Previous studies of nesting habitat used for these species have primarily taken place in broad expanses of marsh surrounded by low densities of human development, which is thought to be optimal for these birds (Benoit and Askins 2002). Large contiguous patches of habitat are known to have positive associations with species richness and abundance in forests (Ambuel andTemple 1983, Askins et al. 1990), and this relationship also holds true for tidal marshes (Benoit and Askins 2002). In urban environments, tidal marshes often get caught between sea level rise and urban borders causing marshes to be reduced in overall area. When urban saltmarshes maintain a wide expanse, the benefit of area could be undermined by fragmentation and degradation often associated with human development. This constraint may lead to reduced avian abundance and biodiversity (Friesen et al. 1995) and negative demographic consequences for specialist species (Chace and Walsh 2006).
As small, urbanized marsh patches become more frequent on the landscape, determining whether such patches could be managed to keep species of conservation concern within them will aid in proper use of limited resources (Bayard and Elphick 2010). We aimed to determine if the urban landscape of NYC is indeed suitable for tidal marsh specialist songbirds by studying nest site selection and factors affecting nest survival of Saltmarsh and Seaside Sparrows in NYC. We predicted that to persist in this distinctive environment, these birds may nest in alternate habitat (taller low-marsh habitat or upland habitat) or alter their nest building strategies (use taller vegetation or nest higher off the ground) from individuals living in larger, less urbanized marshes throughout the species' ranges. We also sought to provide new insights into the ecology of saltmarsh obligate birds in small, human-impacted patches and aid in the management of these birds in urban tidal marsh systems.

Study area
We identified tidal marsh sparrows nesting in five locations throughout NYC in 2012 and a sixth location was added just outside NYC in neighboring Nassau County on Long Island, New York in 2013 (Fig. 1). Study sites included: Sawmill Creek (SA1, SA2, 40°36′36.227″N, -74°11′32.982″W) on Staten Island; Four Sparrow Marsh (FS, 40°35′58.772″N, -73°54′24.390″W) in Brooklyn; Idlewild (ID, 40°39′9.140″N, -73°45′7.729″W) in Queens; and Marine Nature Study Area (MN, 40°37′16.050″ N, -73°37′19.542″W) in Oceanside, Nassau County. We designated 1-ha to 4-ha (depending on total patch size) study plots at each site, centered at the densest known sparrow nest site locations. For analysis, two sites were designated at Sawmill Creek, i.e., Sawmill Creek East (SA1) and Sawmill Creek West (SA2), because these sites were separated by train tracks, differed in vegetation composition, and Seaside Sparrows were only observed at the western site. Study marshes ranged from consisting of predominantly high marsh (S. patens interspersed with Juncus effusus, SA2), to predominantly low marsh (S. alterniflora, ID), to a mixture of both (FS, MN, SA1) vegetation types, and no marshes contained Juncus gerardii, a high-marsh grass found to typically be used by nesting Saltmarsh Sparrows (Greenlaw et al. 2020). Sawmill Creek West (SA2) contained a large patch of upland Iva frutescens shrubs in the center of the marsh, but this shrub was only present in the upland periphery at the other study marshes, as is typical. Study marshes were generally small (5 ha to 35 ha) and had a highly urbanized periphery of commercial industry (SA1), urban transportation (highways, railways, and airports; FS, SA1, SA2, and ID), and/ or residential areas (MN, ID). Saltmarsh Sparrows were found nesting at all five sites, and Seaside Sparrows were only observed nesting at SA2 and ID. We included these specific sites in the study because they represented a range of marsh-level characteristics and because they were the only accessible sites within and near the NYC area with confirmed nesting by tidal marsh sparrows.

Field methods
Nest searching was performed twice weekly from mid-May to mid-August at each site according to methods described in Ruskin et al. (2017b). Three to four observers stood two-arm lengths apart and walked transects across the marsh through all possible nesting habitat, watching for birds flushing from a nest and checking under thatch and overhanging vegetation for nests. Checks of known nests were performed every three to four days. The status of the nest (active, failed, or fledged) was recorded, as were the number of eggs or chicks and the cause of failure (flooding, depredation) if applicable (Ruskin et al. 2017b).
Nest-specific data were collected either at the time of discovery, if eggs were present, or once eggs were first observed in the nest to reduce the likelihood of research activity causing nest abandonment. The height of the bottom of the nest to the ground, rim of the nest to the ground, and the depth of the nest cup were measured to the nearest 0.5 cm using a meter stick. Percent vegetative cover over the nest was estimated by placing a round, white disc (the same diameter as the mean nest cup diameter for these species) into the nest and recording the percent of the disc that was visible from above, following Saltmarsh Habitat and Avian Research Program (SHARP) protocols (Roberts et al. 2017). Percent ground cover in the 1 m² around each nest was estimated using a vegetation sampling frame within one week of the failure or fledging of a nest. The PVC frame contained two sets of parallel lengths of twine, perpendicular to each other, to create 64 evenly spaced intersections. We recorded the type of stem emerging from the ground beneath each intersection or other type of ground cover as appropriate (bare ground, water, wrack), for the estimation of percent of basal ground cover. We also visually assessed the proportion of foliar cover within the 1 m². Additionally, we recorded the mean and tallest height of vegetation at five points (one at the mid-point of each side of the frame, and one at the center of the frame) within the vegetation frame.
We generated random points within each site, at least 5 m apart, in ArcMap 10.0 (ESRI 2012) to compare vegetation characteristics at nest sites to unused available habitat. We measured proportion ground cover and vegetation height within 1-m² quadrats as described for nests. Random points were paired temporally but not spatially with nest points. This design was used to ensure a useful distribution of random points throughout the study period for comparison with nest habitat data. In late July to August, 6 to 50 additional random points were sampled at each site for vegetation composition comparisons among marshes. The sample size in 2012 was small (6-14 samples per study site) and was used to plan an optimal sampling design in 2013 (25-50 samples depending on the area of the study site).

Nest site selection
We tested for differences in proportion of basal vegetative cover between nests and random points at each study site using multiresponse permutation procedures (MRPP) in the Blossom Package (Talbert and Cade 2005)  had low nest sample sizes for Seaside Sparrows at some sites, so we only conducted the analysis for sites with six or more nests per year (SA2 in 2012 and ID in 2013).
Collecting habitat data at the time the fate of a nest is determined can bias the interpretation of nest site selection results because the fate of failed nests occurs earlier in the breeding cycle than that of successful nests (McConnell et al. 2017). Although our design did not account for such bias because of frequent renesting of sparrows throughout the study period, the distribution of collection dates for vegetation data was similar between plots.

Nest success
We estimated daily nest survival rates using the logistic-exposure method (Shaffer 2004) that accommodates varying exposure periods and does not require knowledge or guesses of when nest losses occur in an interval between nest checks. We included site, year, and nest stage (incubation or brooding) in all models to account for likely sources of heterogeneity (Mayfield 1961). Nest fate was then modeled as a function of several predictor variables related to nest structure, i.e., nest height (bottom to ground, cm), nest depth (cm), proportion of vegetative cover over nest, species, and all biologically significant interactions with species. All models, including a null intercept-only model, were run using the glm function in R with a custom logit link function to account for exposure days.
We used an information-theoretic approach to evaluate support for the models (Burnham and Anderson 2002). We assessed the candidate models using Akaike's Information Criterion (AIC) corrected for small sample sizes, including ΔAIC c , Akaike weights (w i ), and deviance. We generated estimates of daily survival rates using the best-supported model, or model-averaged predictions if there was not a single best model based on a ΔAIC c < 2 (Burnham and Anderson 2002) or deviance > 1.

Region-wide comparisons
We compared vegetation (composition and structure) and nest structure characteristics between our urban NYC marshes (FS, ID, MN, SA1, and SA2), our Long Island site (MN), 12 SHARP locations outside of New York, i.e., other sites; Maine (4 sites), New Hampshire (2) Table 1). For our vegetation composition comparison, we separately modeled the proportion cover of four common grasses (S. alterniflora, D. spicata, S. patens, and J. effusus), mean vegetation height within 1 m² around the nest and tallest vegetation height within 1 m² around the nest by location (NYC, Long Island, other), and study year, with a random effect of site using linear regression models in R (R Core Team 2013). We separately modeled our nest structure characteristic nest height (rim to ground and bottom to ground; cm) by location (NYC, Long Island, other, i.e., literature, and other, SHARP) using a linear mixed-effects model with site as a random effect in the R package lme4 (Bates et al. 2015).

Nest success
Controlling for differences among sites and between years and nesting stages, we found evidence that nest survival increased with nest height (bottom to the ground, mean estimate 0.072 ± 0.025 SE) above the substrate because it appeared in all of the top models ( Table 2). The 95% confidence intervals for the effects of nest height, stage, year, and some sites in the top three models did not contain zero (Table 3). No model structure was unequivocally best, although there was no support for the null model. Additional covariates (nest depth, percent cover, and species) appeared in some of the top models; however, inclusion of these variables did not lower deviance by more than one (Table 2) and 95% confidence limits on the regression parameters of these covariates contained zero (Table 3), so we considered them to be non-informative parameters (Arnold 2010) and used the top model (containing only nest height and our control variables) for further inference.
Nest survival increased with nest height (bottom to ground) above the ground at all sites, in both years, and at both egg and chick stage (Fig. 3). Nests at the brooding stage had an average trend of approximately 5-10% higher survival than nests at the  (Fig. 3).

Marsh level
Mean proportion cover of D. spicata was higher at NYC sites than all other sites in the species range in both study years (NYC 2012: 0.188 ± 0.035 SE, 2013: 0.189 ± 0.024; other 2012: 0.065 ± 0.013, 2013: 0.058 ± 0.013; Long Island 2013: 0.043 ± 0.053), and no difference in cover by J. effusus was detected between sites. Our Long Island site appeared to have an overall higher mean proportion cover of S. alterniflora and a lower mean proportion cover of S. patens than all other study sites, whereas proportion cover of these two vegetation types at NYC sites did not differ from sites outside of New York (Fig. 4). Tallest vegetation height around the nest was greater at NYC sites, although the difference was not statistically significant based on wide confidence intervals for sites outside of New York (Fig. 5). Tallest vegetation height around the nest at our Long Island site appeared lower than at NYC sites (Fig.  5).

Nest level
We found mean nest heights (both bottom to ground and rim to ground) of both sparrow species in NYC to be taller than nest heights for each species in all other locations and in the popular literature (P < 0.001). Saltmarsh Sparrow nest heights at our Long Island site did not differ from nest heights at other SHARP sites outside of New York (bottom to ground: P = 0.985, rim to ground: P = 0.822) or in the popular literature (Fig. 6).

DISCUSSION
Nest site selection by Saltmarsh and Seaside Sparrows in NYC was different than is typically recorded elsewhere in their ranges. Sparrows used vegetation in atypical marsh zones out of proportion to their availability, and both species nested higher off the ground   (Table 3), and from Saltmarsh Habitat and Avian Research Program (SHARP) study sites throughout the Northeast, i.e., Saltmarsh Sparrows: Maine, New Hampshire, Massachusetts, Connecticut, and New Jersey; Seaside Sparrows: Connecticut, and New Jersey; 2012 and 2013 combined, with standard errors shown. Sample sizes above bars. Within sparrow species, rim to ground means with the same capital letter are not significantly different and bottom to ground means with the same lowercase letter are not significantly different. Heights from literature were not included in the statistical comparative analysis.
in NYC than at other locations throughout the sparrows' northeastern range. Nest height (bottom to ground) was positively associated with nest survival for Saltmarsh Sparrows, which has not previously been documented as a significant predictor of nest survival for this species. New York City marshes appear to contain on average taller vegetation than other tidal marsh sparrow breeding sites in the northeast Atlantic, although the great variation in vegetation height in non-New York marshes affected our ability to detect statistical significance. New York City marshes had higher than average proportions of D. spicata, an indicator species for transition from high marsh to low marsh (Lonard et al. 2013). Thus, atypical nest site selection in our small, urbanized marshes may reflect ecological differences between degraded and pristine habitat or a bias in sampling nest sites in only expected habitat.
Selection of low-elevation marsh plants by Saltmarsh Sparrows in NYC for nesting afforded them the ability to place their nests higher off the ground than has typically been observed for this species. Seaside Sparrows using similar habitat also built nests higher off the ground than has been recorded for the northeastern breeding population. This difference may be due to the greater than average vegetation heights observed in NYC, which would provide more opportunity for sparrows to select taller grasses for nesting. Both species had similar nest survival rates that were higher than the published average (0.38, Table 4) at 0.55 in 2012 but dropped substantially to 0.25 in 2013. It is difficult to interpret the drop in 2013 because it occurred just after Hurricane Sandy altered the habitat in our study area. However, our results demonstrate that in at least some years, atypical nest habitat selection may not have negative nest survival consequences.
The positive relationship we found between Saltmarsh Sparrow nest survival and nest height contrasts with the Gjerdrum et al. (2005) finding of no effect of nest height on nest survival for Saltmarsh Sparrows and the Kern (2015) finding that the hypervolume niche space of Saltmarsh Sparrows in New Jersey covered a region with below-average nest height and aboveaverage high-marsh cover. Our results are consistent with findings by Benvenuti et al. (2018) that the mean nest height for Saltmarsh Sparrow nests that survive to fledging is higher than nests lost to flooding in Maine, Massachusetts, and New Hampshire. Nest height may be more important for nest survival in NYC than at other sites throughout the birds' range due to a combination of limited available habitat and marsh subsidence (in Idlewild in particular) causing birds to be forced to nest in marshes with higher than normal tidal flows (Cook 2019) because they don't have nearby alternate locations to move to and exhibit strong site fidelity. The response of nesting higher off the ground appears to allow some nests to successfully avoid flooding without inordinately increasing detection and loss to predators. This result may be because of the taller than average vegetation present in NYC that likely aided in concealing higher nests.  declines in this species to the point of extinction by mid-century (Field et al. 2017(Field et al. , 2018. Vegetation height may therefore be a valuable consideration in addition to marsh zone when planning restoration, especially at smaller, urbanized, or otherwise degraded sites. We provide evidence that Saltmarsh Sparrows may be able to vary their nest site selection tactics more readily than previously suspected because we found that Saltmarsh Sparrows at all five study marshes selected S. alterniflora (primarily in its taller lowelevation marsh form except at our Long Island site) and did so out of proportion to its occurrence at two of our smallest marshes (Four Sparrow Marsh and Sawmill Creek East). This result contrasts with many prior studies that documented use of predominantly (80% +) high-elevation marsh vegetation, sometimes interspersed with short-form S. alterniflora (DeRagon 1988, DiQuinzio et al. 2002, Shriver 2002, Gjerdrum et al. 2005, Humphreys et al. 2007, Meiman et al. 2012), but has been suggested anecdotally in earlier studies from New York (Greenlaw et al. 2020). Use of tall S. alterniflora grasses may be related to the importance of nest height in determining nest survival in these NYC sparrow populations.
Previous studies in New York found that Seaside Sparrows often used tall enough grasses that most individuals did not lose nests to flooding, except early in the season when tidal marsh grasses are still short and when seasonally high tides are coupled with extreme storm events (Post et al. 1983). Our findings support that Seaside Sparrows in NYC are searching out tall grasses, but perhaps not tall enough grasses because flooding was the primary cause of nest loss for these birds. Seaside Sparrows at Idlewild nested in typical habitat, whereas at Sawmill Creek West, where low-elevation marsh was lacking, Seaside Sparrows primarily nested at the base of tall upland shrubs (I. frutescens). Substrate switching to I. frutescens bushes in the absence of plentiful lowmarsh substrate is uncommon but was also observed by Woolfenden (1956) in New Jersey and also likely occurred because tall grasses were limited.
Our smallest marsh, Four Sparrow Marsh, had the highest nest survival rates in both years, although not significantly so. This result was surprising because only 1.2 ha of the 5.2 ha marsh was used by nesting sparrows, and the site had a highly urbanized fringe including a major highway. Birds at this site predominantly selected low-marsh S. alterniflora for nesting despite the marsh's predominant high-marsh composition. In contrast, Marine Nature Study, our Long Island site, had what appeared to be the lowest nest survival rate in 2013 (it was not surveyed in 2012), and this site was found to be more similar to other nest sites throughout the Northeast than NYC (shorter overall grass heights and nests built lower to the ground). The higher mean nest survival at Four Sparrow Marsh, a small, urbanized patch, in comparison to Marine Nature Study area, a more typical sparrow nesting marsh, suggests that the typical way we characterize habitat requirements for Saltmarsh Sparrows and other tidal marsh birds (i.e., low-marsh or high-marsh obligates) may be too simplistic, and may therefore obscure potential conservation strategies.
We demonstrated variation in nest site selection within populations of two tidal marsh obligate sparrows. Whether or not the responses we observed were due to population-level or individual-level processes, they suggest potential conservation strategies in small, urbanized marshes. Habitat restoration that provides a heterogeneous matrix of high-and low-elevation zones and marsh plant species, while maintaining an above average availability of high-marsh vegetation, would mimic the conditions at our sites where birds were successful while providing continued habitat options as sea level rise progresses.
By focusing study on the largest or densest populations of species of conservation concern, researchers may miss critical phenotypes that may inform management (Hunter and Hutchinson 2002) and introduce sampling bias. Smaller populations on the edge of the realized niche may be more likely to exhibit adaptive plasticity and evolutionary rescue (Ghalambor et al. 2007) and can easily go undetected. These small populations are especially important to not miss in the study of declining species such as the Saltmarsh Sparrow, which require immediate intervention to persist. Focusing on these small populations and understanding how individuals within them are using habitat at their niche's edge will