Portal de Periódicos da CAPES

Factors influencing occupancy and density of salt marsh songbirds in northeast Florida

2020; Association of Field Ornithologists; Volume: 91; Issue: 2 Linguagem: Inglês

10.1111/jofo.12336

1557-9263

Amy C. Schwarzer, W. Andrew Cox, Brett Tornwall,

Coastal wetland ecosystem dynamics

Salt marshes and the organisms that depend on them are subject to a variety of anthropogenic threats. In Florida, Worthington’s Marsh Wrens (Cistothorus palustris griseus) and MacGillivray’s Seaside Sparrows (Ammospiza maritima macgillivraii) are species of concern that inhabit a small, narrow range of salt marsh in the northeastern corner of the state, an area of increasing human development. The historic ranges of these subspecies encompassed salt marshes in five counties, but their ranges had contracted to just two counties by the early 2000s and their populations declined. We surveyed the historic ranges of the two subspecies during the breeding seasons of 2014 and 2015 to document their distributions, identify habitat features that influenced occupancy and density, and assess whether any recolonization had occurred in areas previously abandoned. We found that the ranges of both subspecies remained relatively stable compared to the early 2000s, with no signs of either further contraction or recolonization. Both Marsh Wrens and Seaside Sparrows were more likely to occupy areas farther from uplands. Marsh Wren occupancy was positively associated with marshes dominated by smooth cordgrass (Spartina alterniflora) and negatively associated with marshes dominated by black needlerush (Juncus roemerianus). Seaside Sparrows were more likely to occur at sites of moderate elevation. We found greater densities of both subspecies in areas farther from uplands, with moderate elevations, and dense vegetation. Marsh Wren density also increased in smooth cordgrass marshes, whereas sparrow numbers increased in areas of moderate vegetation height. Despite these differences between subspecies, the need for dense vegetation away from uplands highlights the importance of smooth cordgrass marshes in the region. Factores que influencian la ocupación y densidad de aves en pantanos salobres en el Noreste de Florida Los pantanos salobres y los organismos que dependen de ellos están sujetos a una variedad de amenazas antropogénicas. En Florida, Cistothorus palustris griseus y Ammospiza maritima macgillivraii son especies sobre las que existe preocupación, pues habitan un pequeño y estrecho rango de pantanos salobres en la esquina noreste del estado, un área en donde ha incrementado el desarrollo humano. Los rangos históricos de estas subespecies incluían pantanos salobres en cinco condados, pero sus rangos se contrajeron a solo dos condados hacia principios de los 2000 y la población ha disminuido en el estado. Realizamos monitoreos de los rangos históricos de las dos subespecies durante las temporadas de reproducción de 2014 y 2015 para documentar sus distribuciones, identificar características del hábitat que influencian la ocupación y la densidad y determinar si ha ocurrido alguna recolonización de las áreas que habían sido previamente abandonadas. Encontramos que los rangos de las dos subespecies se mantienen relativamente estables comparado con los primeros años de los 2000, sin ninguna señal de contracción o recolonización. Ambos, Cistothorus palustris griseus y Ammospiza maritima macgillivraii ocuparon, con mayor probabilidad, áreas alejadas de tierras altas. La ocupación de Cistothorus palustris griseus estuvo positivamente asociada con pantanos dominados por Spartina alterniflora y negativamente asociada con pantanos dominados por Juncus roemerianus. Ammospiza maritima macgillivraii ocurrió con mayor probabilidad en sitios de elevación moderada. Encontramos densidades mayores de las dos subespecies en áreas apartadas de las tierras altas, con elevaciones moderadas y vegetación densa. La densidad de Cistothorus palustris griseus también incremento en pantanos de Spartina alterniflora, mientras que los números de Ammospiza maritima macgillivraii incrementaron en áreas de vegetación con mediana altura. A pesar de las diferencias entre las subespecies, la necesidad por la vegetación densa y estar lejos de tierras altas resalta la importancia de pantanos de Spartina alterniflora en la región. Salt marshes are highly productive ecosystems that have been heavily degraded around the globe, with an estimated 13–31% loss worldwide (Valiela et al. 2009). In the United States, more than 50% of salt marshes have been either lost or degraded (Kennish 2001). Salt marshes in Florida are considered “poor and declining” (FWC 2012) and among the most threatened habitats in the state. Salt marshes are destroyed or fragmented by many human activities such as dredging and filling, construction of boardwalks and piers, shoreline hardening, and the building and improper management of impoundments (Montague and Wiegert 1990, Kennish 2001, Greenberg et al. 2006, FWC 2012). Climate change also may impact salt marshes through sea level rise, increased frequency or severity of storms and storm surges, and intrusion of mangroves (Donnelly and Bertness 2001, Walton 2007, Cavanaugh et al. 2013, Williams et al. 2014). As sea level rises, urban development may also prevent marshes from migrating due to hardened edges from infrastructure, changes in elevational gradient, and alterations in sedimentation patterns, although the ability of marshes to migrate is dependent on local conditions and the level of rise (e.g., Feagin et al. 2010). Many salt-marsh-dependent species are now vulnerable due to the collective effects of these human-related threats. In Florida, two salt-marsh passerine subspecies, MacGillivray’s Seaside Sparrows (Ammospiza maritima macgillivraii; hereafter, sparrows) and Worthington’s Marsh Wrens (Cistothorus palustris griseus; hereafter, wrens), occur in a narrow strip of salt marshes in the northeastern corner of the state. Both subspecies are non-migratory and live solely in salt marsh (Post and Greenlaw 2009, Kroodsma and Verner 2013). Comparison of surveys conducted from 1975 to 2001 indicate that both subspecies experienced > 40% range contractions and reductions in population (McDonald 1988, Kale 1996, NeSmith and Jue 2003). The historical ranges of these subspecies in Florida encompassed the salt marshes in Nassau, Duval, St. Johns, Flagler, and Volusia counties (Fig. 1). During the most recent survey in 2000–2001, all but four observations of the sparrows and all observations of wrens were north of the St. Johns River in Nassau and Duval counties (NeSmith and Jue 2003). Due to their historic declines and restricted range in the state, both subspecies are of conservation concern in Florida. The sparrows were recently evaluated for federal listing as either threatened or endangered (USFWS 2018), and the wrens were listed as a threatened species in Florida in 2016 (FWC 2016). The results of previous studies revealed a variety of marsh features that may be important for Seaside Sparrows and Marsh Wrens. In Georgia, Seaside Sparrows selected higher-elevation marshes to reduce the likelihood of nests flooding (Hunter et al. 2016, 2017). Also in Georgia, occupancy (i.e., the likelihood that the subspecies would be present in a given area) by both subspecies and abundance of sparrows was greater with increased distance from uplands (Nuse et al. 2015, Hunter et al. 2017). Uplands may be a source of, or may provide, cover for mammalian predators (Picman et al. 1993). Perch availability for avian predators has been found to have a negative influence on habitat use by shorebirds (Yasue 2006, Brush et al. 2016) and may also be important for species in open environments like salt marshes. Vegetation type, density, and average height and amount of foraging habitat in the form of open ground and low succulents may influence nest site selection and success (Post 1981, Post and Greenlaw 2009, Hunter et al. 2016), which in turn may influence sparrow occupancy and density (Hunter et al. 2017). Vegetation height and density were found to influence nest survival and occupancy of a freshwater subspecies of Marsh Wren (Leonard and Picman 1987), and we predicted that similar factors may influence saltwater subspecies. Greater water depth around nesting areas may also positively influence nest success, occupancy, and density of freshwater Marsh Wrens (Leonard and Picman 1987) and, therefore, we suspected it might also do so for a saltwater subspecies. Sparrows, conversely, might avoid areas with extensive water coverage because these areas may flood more frequently, which is a particular risk for sparrows (Greenberg et al. 2006) because they nest lower in vegetation than wrens. Given the overall concern for these subspecies and the vulnerability of their habitat, our objectives were to determine their distribution in Florida and the habitat features that influenced occupancy and density (i.e., the number of birds in a given area). We did this using repeated counts at points throughout the historic range of the subspecies in Florida during 2014 and 2015 breeding seasons. We then developed models to test how different drivers (predation, flooding, and foraging) affected occupancy and density. We generally expected that predator avoidance would be the main driver for both species, but that flood avoidance might also be important for sparrows that must balance predation risk with flooding risk. We also used stepwise regression to explore any relationships that the hypothesis-testing approach may not have detected. We conducted point-count surveys in salt marshes in Nassau, Duval, and St. Johns counties in northeast Florida (Fig. 1). We excluded Flagler and Volusia counties because most salt marshes in those counties within the historical ranges of the subspecies are now mangroves, a habitat not suitable for breeding Seaside Sparrows and most subspecies of Marsh Wrens, including Worthington’s Marsh Wren (Wheeler 1931, Post and Greenlaw 2009, Kroodsma and Verner 2013). Potentially suitable marshes in the study area included salt marshes dominated by either smooth cordgrass (Spartina alterniflora; hereafter, cordgrass) mixed with smaller patches of salt marsh succulents (e.g., Salicornia spp. and Batis spp.) or black needlerush (Juncus roemerianus; hereafter, needlerush), and brackish marshes with mixed vegetation that could include Typha spp., Spartina spp. other than S. alterniflora, and Cyperus spp. The study area is at the southern end of the South Atlantic Bight and has an average tidal range of 0–1.5 m and extreme high tides > 2 m. We identified salt marsh habitat within the historical ranges of the subspecies using ArcMap (version 10.0, ESRI Inc., Redlands, CA) and the most recent salt marsh spatial extent recorded by the state (St. Johns River Water Management District 2009). Based on detection rates published for both species (e.g., NeSmith and Jue 2003), we determined that we needed at least 60 points each north and south of the St. Johns River. We wanted to ensure adequate coverage in both regions despite differences in salt marsh area because the area north of the river represented the known occupied area and the area south of the river represented the recently unoccupied area that may have been recolonized. In ArcMap, we generated 100 random points in salt marshes north and south of the river for a total of 200 points. We then moved points to the nearest navigable water edge to ensure boat access using the National Hydrography Dataset (USGS 2013). We do not believe this approach skewed our counts because wrens and sparrows also typically occupy edges near water in this system, so it is unlikely that we missed any significant populations in the interior of the marsh. We calculated straight-line distance between points and adjusted them as needed to ensure they were at least 500 m apart. Prior to sampling, we ground-truthed all points to determine accessibility and existence of appropriate habitat. South of the St. Johns River, > 70% of the first set of points had to be discarded, so we generated an additional 100 points using the process above and conducted site visits until we identified enough usable survey points. We followed point-count survey protocols established by NeSmith and Jue (2003) and conducted surveys during the 2014 and 2015 breeding seasons (1 May–17 July). We visited each point three times per season, with at least one week between surveys. We organized points into survey routes including at least eight points. We conducted surveys at or near high tide via johnboat. Sparrows and wrens are detectable all day (NeSmith and Jue 2003), but we sampled each survey route at least once during the morning and once during the evening to minimize bias. Our data collection protocol did not include a waiting period following arrival at each point because sparrows and wrens do not flush easily and continue singing when approached (NeSmith and Jue 2003). We recorded auditory observations of individuals of each subspecies during a 5-min passive-listening interval at each point. We conducted surveys in weather conducive to detecting birds and excluded periods of moderate to heavy rain, average wind speed > 20 kph, or intense noise (e.g., dock construction). We visually estimated distance to detected birds and recorded them in one of two distance classes (0–50 m and 51–100 m). We also recorded survey covariates including observer, date, time of day, temperature, cloud cover (0–25%, 26–75%, and > 75%), wind speed (measured using a Kestrel wind meter), and noise level. Noise levels were categorized as 0 = no noise, 1 = faint noise, 2 = moderate noise (unlikely to detect birds beyond 100 m), and 3 = loud noise (unlikely to detect birds beyond 50 m). Once during each season, we estimated percent cover (in 5% increments) within a 50-m radius of each point of bare ground, water, cordgrass, needlerush, salt marsh succulents, shrubs, trees, and mangroves. Despite including birds up to 100 m away in the analysis, we assumed a 50-m radius represented marsh conditions at greater distances because of the homogeneity of tidal marshes in this system. We counted the number of trees, mangroves, and potential perches (e.g., snags and posts) within the 50-m radius. We also collected data on vegetation structure at three random points within 10 m of each survey point. We recorded vegetation height and stem count of the top three dominant species within a 0.25-m2 quadrat and categorized qualitative stem density (sparse = < 30% vegetation, moderate = > 30% to < 60% vegetation, or dense = > 60% vegetation) within the quadrat. For each survey point, we used the average of the three stem counts for analysis. We measured distance to upland edge using a range finder for distances ≤ 300 m and GIS for distances > 300 m. We modeled detection and occupancy probabilities with standard multi-season occupancy models (MacKenzie et al. 2002) in program R (version 3.2.3, 2015) using the occu function in package unmarked (version 0.11-0). We modeled abundance (Fiske and Chandler 2011) using the pcount function in package unmarked. Typically, for both taxa, only males sing during the breeding season (Post and Greenlaw 2009, Kroodsma and Verner 2013), so estimates were for breeding males. We used data from only those points surveyed during both years. We tested all variables for collinearity before including them in the model, including examining correlations between process and state variables. No variables were collinear at the 0.5 level. We also assessed covariates for multicollinearity by calculating the variance inflation factor (VIF) for each covariate (Fox and Monette 1992), with the assumption that a VIF of > 10 suggested that a variable should be removed from the analysis (Hair et al. 1998). No covariates had a VIF > 10 and, therefore, we included all variables in the analysis. We took a two-step approach to modeling occupancy and density. First, we modeled detection as a function of survey covariates (ordinal date, time of day, temperature, wind speed, and observer) with an intercept-only model for the state portion of the model, considering all possible combinations to identify the model that best explained variation in detection. Models including observer would not converge, possibly because some observers conducted only a small number of surveys whereas others sampled a large number of surveys. Subsequently, we substituted year for observer to account for any variability in the detection models that could not be explained by the explicitly tested covariates, including observer effects (MacKenzie et al. 2003). We used Akaike’s information criterion (AIC) to select the most parsimonious model (Burnham and Anderson 2002), i.e., the model with the lowest AIC value. In the second step, we then modeled occupancy and abundance for each species based on habitat features using the covariates from the best-performing detection model. We first converted the percent cover measurements and numerical counts to binary covariates (i.e., presence/absence) because of a large number of zeros. Other covariates included type of marsh (based on dominant vegetation, either cordgrass or a combined needlerush/brackish category), distance from upland, average height of vegetation, vegetation density (categorized as sparse, moderate or dense), and average marsh elevation within a 100-m radius (obtained remotely, University of Florida GeoPlan Center 2013, 5-m raster, vertical error ± 9.25 cm). We also tested for effects of quadratic terms for elevation and vegetation height because both taxa, particularly sparrows, seem to select higher areas of low marsh and areas of moderate vegetation height (Hunter et al. 2016, 2017). We examined models in two ways, using a priori hypotheses and using a backward stepwise approach. We chose to do both because we sampled the full ranges of both subspecies in Florida and were interested in making inferences about the factors that influenced abundance and occupancy specifically within our sampling extent. Traditional hypothesis-based model selection processes can identify driving factors such as predation that influence bird distribution, but may miss important secondary relationships with individual covariates. Conversely, stepwise regression risks identifying specious relationships, but we decided that missing an important relationship was more detrimental than identifying a specious relationship, given the conservation status of the subspecies. For the hypotheses-testing scenario, we constructed three models to test how different factors (predation, flooding, and foraging) might affect occupancy and density, including (1) a predation-avoidance model that included type of marsh, amount of water around points, distance from upland, presence/absence of perches, and vegetation density, (2) a flood-avoidance model that included vegetation height and a quadratic term for elevation, and (3) a foraging model that included the amount of open ground and the presence/absence of succulents. We ranked models using Akaike’s information criterion (AIC; Burnham and Anderson 2002). In the backward stepwise method, we began with the full model. We then removed variables having an alpha level greater than P = 0.10, starting with the least significant variable. We used AIC to determine whether reduced models were better than the full model, stopping only when further reduction in parameters produced larger AIC values. To verify that we had not become trapped in local minima, we then tested the set with the forward regression procedure described in Schuster and Arcese (2013). We calculated an average density for each subspecies by dividing the model estimate of number of males for each point by area (in ha) of saltmarsh within a 100-m radius of each point and then averaging the densities across the points. We sampled 122 points in 2014, including 63 north and 59 south of the St. Johns River. We detected no wrens south of the river and one sparrow late in the season at a point near the river. Song quality indicated this individual may have been a dispersed juvenile, not a breeding adult (A. Schwarzer, pers. obs.). We re-surveyed 82 points in 2015, including 62 north of the river and 20 south of the river. North of the St. Johns River, sparrows were detected within the count radius at 25 and 31 of 62 points in 2014 and 2015, respectively. Wrens were detected at 30 and 34 points in 2014 and 2015, respectively. The maximum number of sparrows detected at a point was seven; the maximum number of wrens detected at a point was 10. Mean model-estimated density was 1.33 ± 1.01 (SE) males/ha (range = 0–6.79) for sparrows and 1.76 ± 0.63 (SE) males/ha (range = 0.01–10.49) for wrens. General habitat characteristics of points north and south of the St. Johns River are summarized in Table 1. We tested 28 detection models for sparrow and wren occupancy analyses (all possible combinations). For sparrows, detection decreased with increasing ordinal date (β = −0.019, 95% CI: −0.035 to −0.003), temperature (β = −0.050, 95% CI: −0.112 to 0.013), and wind speed (β = −0.060, 95% CI: −0.114– −0.005). For wrens, detection was influenced by year (more wrens were detected in 2015 than 2014, β = 1.016, 95% CI: 0.263–1.768) and negatively affected by higher wind speeds (β = −0.048, 95% CI: −0.106 to 0.009). Mean detection rates were 66.3% (95% CI: 53.3–77.2) for sparrows and 81.4% (95% CI: 71.3–88.4) for wrens. We also tested 28 detection models for sparrow and wren density analyses. For sparrows, detection was influenced by year (more sparrows were detected in 2015 than 2014, β = 0.465, 95% CI: 0.142–0.788) and negatively affected by increasing temperature (β = −0.036, 95% CI: −0.068 to −0.005), time of day (β = −0.017, 95% CI: −0.040 to 0.006), and wind speed (β = −0.065, 95% CI: −0.097 to −0.033). For wrens, detection was influenced by year (more wrens were detected in 2015 than 2014, β = 1.166, 95% CI: 0.860–1.471), positively affected by increasing ordinal date (β = 0.008, 95% CI: 0.001–0.014), and negatively affected by increasing wind speed (β = −0.049, 95% CI: −0.072 to −0.026). Year was a proxy for observer in all our detection models, suggesting that observers were better at detecting birds in 2015 than 2014. In the hypothesis-testing framework, the predation-avoidance model significantly outperformed all other models for occupancy and abundance of both species (Table 2). However, only a small number of covariates were significant and had beta estimates that did not overlap zero. Occupancy and abundance increased with distance from upland for both species. Wrens were more likely to occupy and have higher densities in cordgrass marshes. Both wrens and sparrows were more abundant in dense vegetation. In the best-performing stepwise regression occupancy model, sparrows were more likely to occupy areas farther from uplands, at moderate marsh elevations (Fig. 2), and in areas without succulents, although the beta coefficient for succulents was not significant and the 95% CIs overlapped with zero. In the best-performing stepwise regression density model, more sparrows occurred farther from uplands, in dense vegetation, at moderate marsh elevations, at moderate vegetation heights (Fig. 3), and in areas with more open water, although the beta coefficient for open water was not significant and the 95% CIs overlapped with zero. The full model would not converge and was removed from the candidate set. Once the presence of perches was removed, the model converged. In the best-performing stepwise regression occupancy model, wrens were more likely to occur farther from uplands, in cordgrass versus needlerush marshes (Fig. 4), and in areas without succulents, although the beta coefficient for succulents was not significant and the 95% CIs overlapped with zero. In the best-performing stepwise regression density model, more wrens occurred in cordgrass marshes than needlerush marshes, farther from uplands, in dense vegetation, at moderate marsh elevations (Fig. 5), and in areas with more open water and no succulents, although the beta coefficient for succulents was not significant and the 95% CIs for succulents and open water overlapped with zero. The results of our a priori hypothesis testing suggest that predation avoidance is the primary driver of occupancy and abundance for wrens and sparrows (Table 2), with predation-avoidance models often ranking much higher than flood-avoidance or foraging models. This suggests that predation avoidance is a more important driver than either flood avoidance or foraging availability in determining the distribution of these birds across the landscape. Most of this seems to be driven by one covariate, distance from uplands, which was also confirmed by the backward stepwise process. Other studies involving these subspecies have also found that distance from uplands was important (Nuse et al. 2015, Hunter et al. 2017). Selecting habitat farther from uplands may relate to predator avoidance because uplands can be a source of generalist predators that would not normally venture far into marshes, especially if predators are also subsidized by human activity (Greenberg et al. 2006). Neither of the remaining two a priori models had >0.01 of the overall weight, but a number of covariates in each model were included in the stepwise model, which is perhaps unsurprising given the known trade-offs between avoiding flooding and avoiding predation for Seaside Sparrows (Hunter et al. 2016). Occupancy rates and densities of sparrows were greatest at moderate elevations. In contrast, wren occupancy rates did not appear to be influenced by elevation, but, like sparrows, their densities were greatest at moderate elevations. Wrens build their nests an average of ~ 0.4 m higher in the vegetation than sparrows (A. Schwarzer, unpubl. data) so likely do not face the same degree of flood risk as sparrows and can occupy the lower reaches of the marsh (albeit in small numbers), whereas sparrows cannot. The highest densities of both subspecies were constrained largely to higher-elevation patches of low marsh (i.e., areas with moderate marsh elevations). The low marshes are dominated by cordgrass that both subspecies use at greater frequencies for nesting than needlerush in the high marsh (A. Schwarzer, unpubl. data). In addition, predation risk, particularly from rice rats (Oryzomys palustris), may be higher in needlerush (Post 1981). Conversely, at lower elevations of the low marsh, flood risk for both species, but particularly sparrows, is greater than at higher elevations. Hunter et al. (2017) also found that sparrow densities were positively associated with elevation in Georgia, although that relationship was linear. It is unclear whether those researchers chose not to test for a quadratic relationship or this result was a function of true differences in marsh structure between northeast Florida and Georgia. The study area in Georgia apparently consisted largely of only low cordgrass marshes and no needlerush marshes, which tend to be higher in elevation. Thus, testing for a quadratic relationship may not have been relevant. Greater numbers of sparrows occurred at points with moderate vegetation height in our study, whereas wrens showed no preference for certain vegetation heights. Sparrows nest in shorter vegetation than wrens, but do need vegetation tall enough for nests to avoid being flooded. Additionally, the aggressive behavior of wrens may sometimes exclude sparrows from the tallest vegetation, as suggested by observations of sparrow eggs that had been punctured by wrens (Hunter 2016, A. Schwarzer, pers. obs.). Selecting nest sites with moderate vegetation height likely balances predation risk and possibly conspecific competition against flood risk (Greenberg et al. 2006, Hunter et al. 2016). Wren density does not appear to be as constrained as sparrow density by vegetation height, despite wrens nesting in the tallest vegetation in any given area. This may be because wrens do not face the same conspecific pressures as sparrows and because tall clumps of cordgrass are available throughout the study area, whereas vegetation of moderate height is not as readily available (i.e., patches of tall cordgrass often immediately transitioned to areas of short cordgrass). Data from a concurrent demographic study indicate that nest survival is low for wrens and even lower for sparrows (10% and 3%, respectively), with predation being the primary cause of nest failure (Cox et al. 2020). Wrens may be able to nest in taller vegetation than sparrows because their domed nests (versus the open cup nests of sparrows) provide additional protection from nest predators, but additional study, probably via a manipulative experiment, would be needed to test this. The two subspecies appeared to differ in how they used different types of marshes. We found that wren occupancy and density were greater in cordgrass-dominated marshes than in needlerush-dominated marshes. However, sparrow occupancy and density were not influenced by marsh type, despite data showing that nesting was constrained to only cordgrass (Cox et al. 2020). We found that wrens nesting in cordgrass had greater nest success than those in needlerush (Cox et al. 2020). Wrens may avoid needlerush marshes because nest predators such as marsh rice rats may be more abundant there than in other vegetation types (Post 1981). Although we did not quantify the abundance of predators in the system or use nest-monitoring techniques such as game cameras, anecdotal evidence suggests that rice rats are the primary nest predator for both subspecies, including frequent observer encounters with rice rats in marsh wren nest balls (A. Schwarzer, pers. obs.). In addition, needlerush may be inferior for building nests because we observed that wren nests built in needlerush were more likely to slide downward or collapse in severe weather than those in cordgrass (A. Schwarzer, pers. obs.). However, we found no sparrow nests in needlerush and behavioral observations indicated that there were a relatively large number of unpaired male sparrows in the needlerush marshes (Schwarzer, unpubl. data). The presence of singing males in needlerush may mask habitat selection by nesting females for cordgrass marshes. As such, count data may be misleading about the true habitat needs and population status of the sparrows (sensu Van Horne 1983). Both subspecies were more abundant in areas with dense vegetation, and demographic work in the study area revealed that they selected nest sites with higher stem counts than random locations (Schwarzer, unpubl. data), but stem count was not correlated with either nest or fledgling survival (Cox et al. 2019, 2020). Dense vegetation may better conceal nests, reducing the risk of predation, but it may also be a structural prerequisite for nesting females and have no effect on nest success or failure. Areas with moderate or sparse vegetation tended to be occupied by fewer singing males, and we often found no nests in those areas. Interestingly, occupancy for both subspecies was also greater in areas with few or no succulents. However, the beta coefficients for all models were non-significant at the P = 0.10 level. Additionally, few points had succulents present (8 points occupied and 13 points unoccupied) and this small sample size may have skewed the results. Confidence intervals were large for these estimates. Additional study is needed to understand what, if any, relationship exists between succulents and occupancy for these two subspecies. Similarly, the relationship between sparrow and wren densities and open water was weak; beta coefficient CIs overlapped zero, but the coefficient was significant. In our demographic study, wren territories tended to be located along creek edges, whereas sparrow territories were often farther from the water’s edge, but still within 10–20 m of creeks (Cox et al. 2020). The models, however, indicated that wren densities were negatively associated with open water, whereas sparrow densities were positively associated with open water. We do not know whether this relationship for sparrows reflected a real habitat preference or was a function of the landscape features in our study area. More sparrows were found in salt-marsh patches of moderate elevation that tended to be adjacent to rivers or large creeks (i.e., > 50 m wide). Under these circumstances, survey points with the most sparrows would automatically have at least 50% water within a 50-m radius. Thus, we believe these results may be a product of the landscape rather than an indication of true selection. Wrens, however, in addition to these sites, were found at numerous sites along small interior creeks, and they may select areas with many smaller creeks over areas with one or two large creeks. One hypothesis is that smaller creeks provide creek banks and vegetation that is better protected from erosion and disturbance by tidal surge and storms. Another is that multiple smaller creeks simply have a greater edge-to-area ratio and provide more wren habitat than a patch of similar size along the edge of a larger river. More study is needed to elucidate any benefits of such habitat. We were not surprised that marshes south of the St. Johns River were not occupied, given that the suite of characteristics we found associated with greater densities and occupancy rates for both subspecies in Florida was, for the most part, lacking south of the river. These southern-side marshes are much narrower and closer to uplands than marshes north of the river (Table 1). The density of human development is also much higher south of the river, so predator subsidization and other detrimental anthropogenic activities (e.g., docks and piers; Banning et al. 2009) as well as channelization and erosion (Hackney and Cleary 1987, Kennish 2001) are more likely to increase rates of nest predation or degrade habitat. We found that many of the marsh patches were low in elevation and had sparse vegetation, making them unsuitable for either subspecies (Table 1). In addition, marshes in much of Flagler and Volusia counties have been fragmented by or converted to mangroves, and the somewhat suitable marshes in St. Johns and southern Duval counties also had evidence of mangrove and shrub encroachment (Table 1). Surveys were conducted at or near high tide and likely had the effect of concentrating birds at the highest elevations of their territories generally on or near the edges of creeks where we were surveying. We do not believe this influenced our findings about habitat associations because territories are already on or near creek edges in this region (Kale 1965, A. Schwarzer, unpubl. data). Detection and numbers of birds counted may have been higher during high tide given that high water limits foraging opportunities and males may engage in other activities such as singing, but, because we chose to standardize this across all surveys, we are unable to assess this possibility. Both taxa were present south of the St. Johns River into the 1950s, and wintering sparrows still use some of the protected portions of the southern marshes. However, neither species has been present south of the St. Johns River during the breeding season since the 1980s and the current habitat is unsuitable for successful breeding. The restoration and management needed to attract these subspecies to breed in the marshes south of the St. Johns River would require substantial time, energy, and resources, not only to improve, but also to expand, existing habitat. Therefore, we recommend that management activities focus on preserving dense stands of cordgrass marshes north of the St. John’s River. The optimal elevation will likely change with sea level rise, and we therefore refrain from addressing specific elevation recommendations. This system is largely maintained by large tidal variations and periodic storms and currently has little direct human management. Maintenance through fire is not appropriate for all but the highest patches of marsh and those patches tend to be close to uplands and, therefore, unsuitable for these species. In light of the growing threat of sea level rise, steps that might be taken to protect vulnerable habitat are unclear because this system is not necessarily ideal for marsh migration due to development in the uplands. However, modeling should be used to determine whether there are potential corridors for migration (sensu Enwright et al. 2016). Interventions such as living shorelines and thin-layer deposition may also aid in protecting existing marsh, at least in the short term. More exploration of climate-adaptive management options will be required if these two subspecies are to persist in the area. First, we thank the Florida Fish and Wildlife Conservation Commission (FWC), the FWC Wildlife Legacy Initiative, and the U.S. Fish and Wildlife Service for supporting work on Worthington’s Marsh Wrens through the State Wildlife Grant program (F14AF00892 [T-35]). Thanks also to FWC’s Species Conservation Planning Section and Fish and Wildlife Research Institute for providing additional funds through the Threatened and Non-game Species program and the Non-game Trust Fund, respectively. Additional thanks to the U.S. Fish and Wildlife Service for providing monies through the Coordinated Candidate Species Assessment for MacGillivray’s Seaside Sparrows. Many thanks to the technicians who conducted the field work day in and day out despite hot, muddy conditions and malfunctioning boats: Roxan Chicalo, Sean Jeffreys, Lara Mengak, and Laura Evans. The Doris Duke Foundation also provided an intern, Jaclyn Selden, for the 2015 field season. Thanks also go to Katie Malachowski and Carolyn Enloe for providing field support. Erin Leone, Brittany Bankovich, and Ryan Butryn provided essential GIS and study design support. Elizabeth Hunter and Todd Schneider provided data and field knowledge about Seaside Sparrows in Georgia, and Pat Leary provided natural history information about the birds in northeast Florida. We would also like to thank Andrea Alden, Robyn Mcdole, and Ginger Gornto who answered our many questions about administrative and purchasing issues related to the project. We conducted research on the Timucuan Ecological and Historic Preserve under National Park Service permit #84300.