North American wetland birds have been subject to significant habitat loss through direct reductions in wetland area and degradation of wetland quality (Brinson and Malvárez 2002). While these threats continue in the form of wetland drainage, water pollution, invasive species, and climate change (Mitsch and Hernandez 2013), targeted restoration efforts combined with favorable precipitation and high productivity in key regions have helped conserve populations of waterfowl (Anderson et al. 2018). These changes have resulted in an overall improvement in the status of North American wetland birds (North American Bird Conservation Initiative 2019, North American Bird Conservation Initiative Canada 2019); however, veiled within this trend are some populations of obligate marsh-breeding birds (e.g. grebes, rails, and bitterns; hereafter "marsh birds") which are in decline or even at risk of extinction (Tozer 2016).
This discrepancy in conservation outcomes for waterfowl and marsh birds is at least partly attributed to limited knowledge of marsh bird ecology and population status (Orr et al. 2020). Protocols typically used to study bird populations (e.g. the Breeding Bird Survey; USGS Patuxent Wildlife Research Center 2018, Environment Canada 2019) yield extremely infrequent detections of many marsh birds due to their elusive behavior, low numbers, and relatively inaccessible habitat (Conway and Gibbs 2011, Steidl et al. 2012). Due to these challenges, many wetland conservation entities have historically developed and implemented strategies focused only on waterfowl or migratory shorebirds (Williams 2018); however, habitat preferences for these other groups may not be equivalent to those of marsh birds.
Fortunately, the last two decades have seen the development of specialized marsh bird monitoring protocols to increase detection probabilities and statistical power (Conway 2011, Steidl et al. 2012), as well as new analytical approaches accounting for imperfect detection (e.g. Fiske and Chandler 2011). These advancements have enabled a sharp increase in research informing the protection, restoration, and creation of habitat specifically for marsh birds (EHJV 2017, Saunders et al. 2019, Grand et al. 2020, Ladin et al. 2020). Notably, diversity in emergent vegetation, surface water inundation, and water:vegetation interspersion have been identified as important site characteristics for marsh birds (Alexander and Hepp 2014, Bradshaw et al. 2020). Landscape composition can be important to marsh birds at a surprisingly broad range of spatial scales (up to 6400m; Tozer et al., 2020) and habitat associations can be region-specific (Roach et al. 2017). While some studies suggest that habitat associations for marsh birds likely differ from those of waterfowl (e.g. Valente et al., 2011), to our knowledge few have explicitly made this comparison using spatiotemporally matched data from both guilds (Connor and Gabor 2006, Baschuk et al. 2012). More common are studies assessing whether wetland conservation actions designed for waterfowl also benefit other species. Wetlands managed for waterfowl, particularly at intermediate levels of management intensity, do have greater marsh bird occupancy than unmanaged wetlands (Tozer et al. 2018, Bradshaw et al. 2020); however, these shared benefits are clearly insufficient to stabilize many populations of marsh birds. Differences in the nuances of site habitat preferences and/or differing effects of landscape composition may be important drivers of this discrepancy.
Although Eastern Habitat Joint Venture (EHJV) partners have been working to protect and restore wetlands and associated upland habitats in eastern Canada for over 30 years (EHJV 2020), region-specific habitat requirements for marsh birds and waterfowl are not well described and the effectiveness of these conservation actions remain largely unstudied (but see Connor and Gabor, 2006). This makes it particularly challenging to know whether provincial requirements for the maintenance of significant wetlands and "no net loss" of function for all other wetlands are being met. In addition, many of the Migratory Bird Joint Ventures responsible for key actions driving waterfowl conservation under the North American Waterfowl Management Plan (NAWMP Canada 2013, NAWMP 2018) are currently expanding their mandates to include the protection of other avian species (“all-bird conservation”) under the North American Bird Conservation Initiative (NABCI 2020). Understanding the habitat requirements of multiple guilds of birds, therefore, provides important feedback for evaluating the effectiveness of conservation action.
Using data collected by the Maritimes Marsh Monitoring Program of Birds Canada (Birds Canada 2020), we investigate spatiotemporally matched occurrence data for waterfowl and marsh birds. Our objectives were to 1) assess land cover habitat associations for breeding marsh birds and priority waterfowl in eastern Canada at a broad range of spatial scales and 2) compare results between guilds using occupancy modelling, and an information theoretic approach to model selection. This study is timely and highlights differences in habitat associations between waterfowl and marsh birds, enabling regional EHJV partners to optimize formerly waterfowl-based site-selection criteria and management plans for all-bird conservation.
The study was conducted from 2015 to 2018 at fresh and saltwater herbaceous emergent wetlands in New Brunswick, Prince Edward Island, and Nova Scotia, Canada (Fig. 1). The majority of the survey sites were in the New Brunswick portion of Bird Conservation Region 14 (Atlantic Northern Forest), within the inland Aukpaque and Peticodiac ecodistricts, and along the coastlines of the Fundy and Kouchibouguac ecodistricts. The study area was dominated by a variety of coniferous and deciduous forest types of diverse ages, followed by open and forested wetlands, and agricultural and developed land uses (Environment Canada 2013).
During the initial years of the Maritimes Marsh Monitoring Program (est. 2012), stratified random sampling was used to establish the first survey routes, with routes being added over time based on wetland size, proximity to volunteers, and location of partner lands (e.g. Ducks Unlimited, Nature Trust of New Brunswick). Routes are comprised of 1-10, 100 m radius survey sites located >250 m apart. Optimally, each route is surveyed twice per year, between 25 May and 15 July, with a minimum of 14 days between visits; however, the realized distribution of effort was variable across sites and years (Table 1). Such variation in effort is accounted for within the occupancy modelling framework (see below).
All surveys took place in the morning, between 30 min before sunrise and 10 AM for each route. For each survey, time of day, air temperature (°C), wind (Beaufort Scale, 0-6), cloud cover (10ths of the sky), and background noise level (0-4) were recorded. Surveys were conducted only when there was no precipitation. At each survey site, observers conducted a 15-minute point count comprised of three consecutive 5-minute periods: passive observation, call broadcast, and passive observation. The call broadcast period was used to increase detection probability of elusive marsh birds (c.f. Conway, 2011) and consisted of 1-minute periods (30 seconds of calls, 30 seconds of silence) for each of five species: Least Bittern (Ixobrychus exilis), Nelson’s Sparrow (Ammodramus nelsoni), Virginia Rail (Rallus limicola), Sora (Porzana carolina), and Pied-billed Grebe (Podilymbus podiceps). Audio or visual detections at any time during the 15-minute point count were included in our analyses, provided they fell within the 100 m site radius and were not classified as fly-throughs (birds not interacting with vegetation or water surface within the site). Black Tern (Chlidonias niger) was the only species for which observations of flying birds were included in our analyses since site use and true fly-through behavior were difficult to distinguish. Marsh type (saline or fresh), the presence of water level control structures (e.g. dike, variable control structure), and land protection status were also recorded for each site.
We selected the four-year period with the greatest sampling effort that was also well-aligned with available landscape data (see "Land Cover and Climate Covariates") to include in our analyses: 2015-2018. We then subset these data to include only surveys with complete observation and site metadata, for a total of 2368 surveys at 338 sites, representing 53 routes (Table 1, Fig. 1). Marsh type was classified as "inland" or "coastal" based on whether the site’s route was comprised of only freshwater marshes or both freshwater and saltwater marshes, respectively. This route-level classification system accounts for both salinity and landscape context.
Land cover data were obtained primarily from the Nature Conservancy’s Northeast Terrestrial Habitat Map (Ferree and Anderson 2013, The Nature Conservancy 2015): a high-resolution 30 m x 30 m raster dataset with a standardized habitat classification scheme, updated in 2015. We chose seven relevant land cover categories to extract for our analyses: marsh, other wetland, open water, urban/residential/built, agriculture, dry forest, and grass/shrub. These categories were defined based on the "macrogroup" classification level in the dataset or, where limited data gaps occurred, data were substituted from the regional land cover dataset compiled for the analysis of the Second Maritimes Breeding Bird Atlas (Stewart et al. 2015). A crosswalk of land cover categories from each dataset is provided in Appendix Table 1.
We used ArcGIS 10.3 (ESRI, Redlands, CA, USA) to generate a seamless habitat layer. Following the approach of Tozer et al. (2020), we extracted percent land cover for each category at a broad range of spatial scales: 200 m, 400 m, 800 m, 1600 m, 3200 m, and 6400 m buffers around each survey site. Regions where these buffers overlapped ocean environments (defined as open water beyond state and provincial boundaries) were removed prior to these calculations. In order to determine which type of wetland cover most influenced the occurrence of each species, we used the original categories of marsh, other wetland, and open water to generate five final wetland cover covariates: Marsh, Other Wetland, All Wetland (Marsh + Other Wetland), Marsh & Water, and All Wetland & Water.
Each survey site (100 m radius) was assigned a "Protection" status based on boundary files obtained for National Parks, Provincial Parks, Wildlife Management Areas, land trusts, and land conservation easements, as well as for all parcels managed by Ducks Unlimited Canada. If the site boundary overlapped at least one of these parcels, the site was considered influenced by protection. Protection status was also confirmed using observer data. Finally, each survey site was assigned a "Water Control" status using similar methodology. If the site overlapped at least one Ducks Unlimited Canada parcel influenced by a water level control structure (e.g. dike, dam, impoundment, fishway, variable control structure), or observer-recorded site metadata indicated the presence of such structures, the site was considered to have water level control.
Historical climate data for each survey site were obtained from Natural Resources Canada, based on a thin plate smoothing spline interpolation model for weather station data developed by McKenney et al. (2013). To represent average conditions at each site, we used the most recent 5 years of data (2014-2018) to calculate mean temperature and precipitation values for May, June, and May-June, as well as mean growing season characteristics. Growing season covariates were defined based on temperature rules outlined in MacKey et al. (1996).
Prior to analysis, all continuous covariates were standardized by z-transformation (mean = 0, SD = 1) to enable direct comparison of resulting model coefficients (Schielzeth 2010). A complete list of covariates considered for inclusion in our occupancy models is provided in Table 2.
Species were selected for analysis based on the Bird Conservation Region 14 priority species list (ACJV 2020), and by requiring a minimum naïve site occupancy of 0.14 (number of sites occupied/number of sites surveyed) to help ensure model convergence and improve model inference (Tozer et al. 2020). This selection procedure resulted in five species of marsh birds and five species of priority waterfowl for modelling (Table 3). All waterfowl included, except Wood Duck, are also priority species of the Eastern Habitat Joint Venture (EHJV 2017); however, Wood Duck is a provincial priority breeding waterfowl in New Brunswick.
To identify the land cover and climate characteristics that best explain the occurrence of species at marsh sites, we used single-season site occupancy models (function "occu", package "unmarked"; MacKenzie et al. 2002, Fiske & Chandler 2011) within R Environment version 3.6.1 (R Core Team 2019). These models consist of two hierarchical logistic regressions - the "detection process" describing true site occupancy as a function of covariates affecting detectability and the "state process" describing observed occupancy during each survey, conditional on true occupancy, as a function of covariates describing site characteristics. Together these processes describe the probability of occurrence, or occupancy, while accounting for uncertainty in detection.
In order to increase survey sample size and enable more robust analyses, we assumed occupancy to be constant over a four-year period (2015-2018) rather than a single season, resulting in 1-11 surveys per site. The response variable was the presence or absence of ≥1 individual of the species in question on each of these surveys. Single-season site occupancy models can be used across multiple years (e.g., Glisson et al., 2017; Tozer et al., 2020), although we recognize that such a broad closure assumption is imperfect (Rota et al. 2009), and that it may differ between waterfowl and marsh birds in ways that are unclear, potentially influencing our results in unknown ways. Given that movement between sites is possible (Connor and Gabor 2006, McClure and Hill 2012), the most appropriate interpretation of model results is the probability that sites were ever used during the 4-year period, which is in keeping with our goal to identify and compare the habitat characteristics which most influence site use by our species.
Following an information theoretic approach, we selected covariates to include in the detection and state processes of a comprehensive global model for each species and then selected the single best overall model (best combination of global model variables) for each species using Akaike’s Information Criterion for small sample sizes (AICc; Burnham & Anderson 2002). Covariate selection proceeded first for the detection process and then for the state process while holding the best detection process constant. To reduce the number of variables included in global models, potential covariates were binned into closely related groups (e.g. different buffer distances of the same land cover type; Table 2) and the most influential univariate predictors were identified from each group using the function "aictab" from the package "AICcmodavg" (Mazerolle 2019). A null model was run within each group and if this model had ∆AICc ≤ 2, then no influential univariate predictors were identified. In addition, univariate predictors were not included in global models if their 85% confidence intervals overlapped zero (Arnold 2010). While the single best covariate (or no covariate) was selected from each land cover group (Wetland, Agricultural, Urban & Built, Dry Forest, Grass & Shrub), selection of multiple covariates with ∆AICc ≤ 2 was permitted from within the detection and site-level groups. Finally, correlations between all selected covariates were examined, and if |r| ≥ 0.6, then only the most ecologically relevant covariate was retained in the global model (r assessed using Pearson or Spearman correlation for continuous pairs, corrected Cramer's V for continuous:categorical pairs, and ANOVA for categorical pairs). Ecologically relevant covariates were those we hypothesized were most likely to influence the occupancy of the species based on our own knowledge and the literature. When ecological relevance was not clear to us, we instead retained the covariate that explained the most variance.
The best overall model for each process was identified by comparing AICc for all possible combinations of global model covariates using the "dredge" function in package "MuMIn" (Bartón 2020), with the added criterion that 85% confidence intervals for all covariates not overlap zero (Arnold 2010). Model fits were assessed using the MacKenzie and Bailey (2004) goodness-of-fit bootstrap test in package "AICcmodavg".
The Maritimes Marsh Monitoring Program produced sufficient metadata for occupancy modelling at 338 sites (22,368 visits, 53 routes) between 2015 and 2018. Of all target marsh bird species, Sora occupied the highest proportion of sites based on naive occupancy (42%), followed by American Bittern, Pied-billed Grebe, Virginia Rail (all ~25%), and Black Tern (14%; Fig. 2). Waterfowl occupied between 37% and 22% of sites (Canada Goose and Wood Duck, respectively).We present results from both the global model stage (univariate selection within covariate groups, screened for acceptable confidence intervals and correlation) and from the best overall model stage (multivariate selection across covariate groups, based on the global model). This approach enables us to compare habitat associations among species within each covariate group even if those covariates were not retained in the single best predictive model for some species. Since assessing detection probability was not a goal of this study, results for this subprocess are not discussed (but see Appendix Table 2). Final model coefficients are presented in Table 4, and model fit in Table 5.
Wetland cover was important to marsh birds at a much broader range of spatial scales than for waterfowl (Fig. 3). The best supported spatial scale for positive effects of wetland cover occurred within only 400 m of the survey site for waterfowl, while for four of the five marsh bird species these effects were strongest within 800 - 6400 m.
Including larger bodies of open water in the calculation of wetland cover improved the explanatory power of this covariate for three marsh bird species and three waterfowl species (Fig. 4). This improvement was substantial in all cases, with only wetland covariates including open water having ∆AICc < 2. The best univariate wetland cover predictors for marsh birds were highly variable across species, with all four wetland definitions represented: Marsh, All Wetlands, Marsh and Water, All Wetlands and Water (Fig. 4). While All Wetland and Water was the most common predictor retained for waterfowl, the relationship was negative for American Black Duck and no wetland covariate was retained for Canada Goose (covariate dropped due to correlation with the urban and built covariate, which explained more variation), making it unclear whether general preferences in wetland type exist across this guild (Fig. 4).
Differences in the importance of wetland cover for marsh birds and waterfowl were also apparent at the final model selection stage (Table 4, Fig. 5). Positive effects of wetland cover were included in the best supported occupancy model for 3/5 marsh bird species (all but Pied-billed Grebe and American Bittern), but for only 1/5 waterfowl species (Wood Duck). Effect sizes were also larger for the marsh bird species than for the Wood Duck (Table 4).
Urban and built land cover had positive effects on many waterfowl species but had either negative effects or no effect on marsh birds. As a univariate predictor, urban and built land cover in the surrounding landscape had negative effects on 2/5 marsh birds (Black Tern, American Bittern), and little to no effect for Pied-billed Grebe (∆AICc < 2 for null model, but ∆AICc = 0 for Urban within 6400 m) and Virginia Rail (∆AICc < 2 for null model, but ∆AICc = 0 for Urban within 200 m), and no effect on Sora. For waterfowl, positive effects of urban land cover were observed for the most common 3/5 waterfowl species and negative effects were observed for the less common 2/5 species (Fig. 3). Urban land cover effects occurred at much broader spatial scales for waterfowl than for marsh birds, with positive effects ranging from 1600-6400 m for waterfowl and negative effects occurring within just 200-800 m for marsh birds. Negative effects also occurred at broad spatial scales for waterfowl, from 800-6400 m. At the final model stage, negative effects of urban land cover were retained for American Bittern and positive effects were retained for Canada Goose, American Black Duck, and Mallard (Table 4; Fig. 6).
Agricultural land cover had more positive effects on occupancy by waterfowl than marsh birds. As a univariate predictor, agricultural land cover helped explain variation in site occupancy by two marsh birds, negatively influencing occupancy by Black Tern, and positively influencing occupancy for Pied-billed Grebe (Fig. 3). In contrast, occupancy by 3/5 waterfowl species was positively related to agricultural cover (Mallard, Ring-necked Duck, Canada Goose). At the final model stage, agricultural land cover was only retained for Ring-necked Duck (Table 4).
As a univariate predictor, grass and shrub land cover helped explain occupancy by 3/5 species of marsh bird (Pied-billed Grebe, American Bittern, Black Tern) and 2/5 species of waterfowl (Ring-necked Duck, Canada Goose; Fig. 3). There was no clear pattern by guild, with all effects being positive at the 6400 m spatial scale except for Black Tern, which showed a negative effect within 800 m. Species with positive effects of grass and shrub cover at the univariate stage all retained this variable at the final model stage (Table 4). Ring-necked Duck had larger effect sizes for agricultural cover than for grass and shrub cover, while Canada Goose and Pied-billed Grebe showed the largest effect sizes for grass and shrub cover.
As a univariate predictor, dry forest cover was a negative predictor of occupancy for Sora and Mallard, and a positive predictor of occupancy for Ring-necked Duck (Fig. 3). Only the negative relationships were retained at the final model stage (Table 4). There was no clear difference between guilds.
As a univariate predictor, marsh type was important for all species except American Black Duck and Black Tern, for which the univariate 85% CIs overlapped zero (Table 6). Most species were more likely to occur at inland, freshwater marsh sites than at coastal marsh sites, except for Canada Goose. Marsh type was retained in the best occupancy models of three marsh birds (Sora, Pied-billed Grebe, Virginia Rail) and two waterfowl (Mallard, Wood Duck; Table 4). In general, the positive association with inland, freshwater marshes was strong (effect size > 1, except Sora), with no clear difference between species groups.
Water level control was beneficial for both marsh birds and waterfowl but may be more important for marsh birds. As a univariate predictor, water-level control was important for 3/5 species from each guild, with the univariate 85% CI overlapping zero for Pied-billed Grebe, Black Tern, American Black Duck, and Ring-necked Duck (Table 6). However, at the final model stage the presence of a water control structure was retained for all three marsh birds, but only for Wood Duck from the waterfowl group (Table 4; Fig. 7). Water control was particularly important (effect sizes > 1) for Sora and Wood Duck.
Site protection status was beneficial for both marsh birds and waterfowl. As a univariate predictor, site protection was important for all species except American Black Duck, where the univariate 85% CI overlapped zero (Table 6). All other species were more likely to occur at sites overlapping protected areas than at unprotected sites. Site protection was retained in the best overall model for 4/5 species in each guild with large effect sizes (Table 4). Effect sizes for marsh birds were slightly larger on average. Site protection was especially important for Pied-billed Grebe and Ring-necked Duck.
Waterfowl are one of the few avian guilds showing population increases over the past few decades, a fact rightfully attributed, at least in part, to international wetland conservation efforts and good waterfowl and hunting management (NAWCA - North American Wetlands Conservation Act, NAWMP, Joint Ventures). Favorable precipitation in key regions since the 1990s, particularly in the Prairies, has also helped boost productivity, and in turn, populations of many waterfowl species (US Fish & Wildlife Service 2020). By contrast, population trends and range for many marsh birds are declining (Correll et al. 2016, Tozer 2016, Stevens and Conway 2020, McGowan et al. 2020), prompting the question of why wetland conservation efforts have not had an equally positive impact on all wetland-dependent birds. The answer likely lies in differences in key habitat associations between these guilds. Using detections of marsh-obligate birds and priority waterfowl sampled concurrently in a suite of wetlands in eastern Canada, we found that differences in land cover preferences at a broad range of spatial scales may be important factors contributing to these inconsistent conservation outcomes.
We found that protected wetlands and those with water-level control structures were more likely to be occupied by the species in our study. While wetland protection seemed to benefit marsh birds and waterfowl equally, water-level control was more strongly associated with increased occupancy by marsh birds. This is likely because hydrologic stability is important to marsh birds during the breeding season, leading to increased structural complexity in emergent vegetation and water depths (Bradshaw et al. 2020). Extreme fluctuations in wetland water levels during the breeding season occur in the St. John River floodplains of New Brunswick, and there is evidence impoundments may be particularly beneficial to marsh birds in this region (Connor and Gabor 2006). Although water-level control structures generally had weaker effects on the occupancy of priority waterfowl, there may be stronger effects within this guild at other times of year (staging, wintering).
One of the most notable differences we observed in the habitat associations of the marsh birds and waterfowl we analyzed was that the marsh birds were influenced by wetland cover at a much broader range of spatial scales, and a larger average spatial scale, than the waterfowl. Positive associations between marsh birds and wetland cover occurred within 800 to 6400 m while those for waterfowl occurred within 400 m. Our results are similar to those for marsh birds in the Great Lakes region, where some marsh birds were also associated with wetland cover at broad scales (e.g., Saunders et al. 2019, Tozer et al. 2020). In comparing types of wetland cover (marsh, marsh and water, all wetland types, all wetlands and water), we found that all wetlands and water was the best supported covariate for a majority of the waterfowl, while the best supported covariates for marsh birds included all four types. Along with variable spatial scale results, this suggests the marsh bird species we analyzed have specialized wetland cover preferences within the landscape that cannot readily be generalized at the guild level. At the final model selection stage, a positive effect of wetland cover was only retained for one species of waterfowl compared to three of the five marsh birds, suggesting that wetland cover in the landscape may be more important for marsh birds than for waterfowl. This means that in order to determine or maintain highly preferred sites for the marsh birds that we analyzed, managers should consider wetland composition much farther into the surrounding landscape than previously thought (e.g., up to 6400 m; see also Tozer et al. 2020). The mechanism(s) behind these patterns remain unclear and would be a productive area for further research.
The marsh birds and waterfowl in our study also differed in their response to human development in the landscape. Very generally speaking, waterfowl were more likely to benefit from agricultural and urban-built land cover, while marsh birds were less tolerant of these land cover types. While probability of occupancy for marsh birds was unrelated or negatively related to increasing urban and built land cover across a broad range of spatial scales, occupancy by three of the five waterfowl species was positively related to this same factor. Only one species, the American Bittern, retained a negative relationship with urban and built land cover at the final model stage, with occupancy gradually decreasing as urban land cover increased within 800 m. Other studies have documented similar detrimental effects of local-scale urban land cover, with marsh birds preferring sites with more rural than urban cover within 1000 m (Smith and Chow-Fraser 2010), and marsh bird community integrity exhibiting a negative threshold response to urban cover within 500 m and 1000 m (at 14% and 25% cover respectively, DeLuca et al., 2004). In contrast, only positive effects of urban land cover at broad spatial scales (1600 m - 6400 m) were retained in the final models for waterfowl, including those for American Black Duck, Mallard, and Canada Goose. Occupancy by these species increased with urban and residential development up to the maximum percent cover present in our study region (~70% for 1600 m, ~50% for 6400 m). A positive association between road density and the occurrence of American Black Duck and Mallard has been found previously in our study area (Lieske et al. 2012, 2018), providing further support that differences in habitat preferences exist between the two guilds. It is likely that road density and urban/suburban land cover are correlated with other landscape features favored during nesting, such as soil productivity and associated wetland productivity. While Canada Geese are known to have a high level of overlap with human habitation resulting in human-wildlife conflict (Smith et al. 1999, Fox 2019), some studies have found that developed land cover was not a significant predictor of occupancy or abundance in this species (North Carolina, Southern Ontario; McAlister et al., 2017; Messmer et al., 2015), with the presence of open water and pasture being the best predictors. Nevertheless, in the Maritime Provinces, Canada Geese showed a strong association with urban and built land cover, possibly reflecting lower development density and agricultural land cover in our region which could lead to more attractive urban forage opportunities.
We observed a similar contrast in the effects of agricultural development where occupancy by marsh birds was either positively (Pied-billed Grebe, 1600 m), negatively (Black Tern, 200 m) or unrelated to agricultural cover, whereas occupancy of three of five waterfowl species were all positively influenced by increasing agricultural cover at broad spatial scales (maximum cover ~80% for 1600 m, ~70% for 3200 m). Agricultural cover was not retained in the final models for marsh birds but was retained for Ring-necked Duck. A positive association between breeding waterfowl abundance and increasing agricultural land cover up to about 50% of the surrounding landscape has been observed previously for dabbling ducks in the Maritime Provinces, after which numbers decline (Lieske et al. 2012, 2018). These species probably benefit by nesting in the vegetation available on low intensity farms in the region, but once farming becomes more intensive the costs for species begin to outweigh the benefits (Lieske et al. 2018). Across Canada, landscape characteristics (patch density, edge density) of cropland were positively associated with Mallard and Canada Goose breeding pair abundance (Adde et al. 2020). Along with our results for urban land cover, these findings further suggest differences in habitat use between marsh birds and waterfowl which, in turn, may be contributing to differences in conservation status for the two guilds.
There are likely important covariates that predict site occupancy by these ten species which were not accounted for in our study, and which may reveal further differences in habitat preferences among guilds. Our understanding could be improved by incorporating data on-site hydrology, land cover configuration (e.g. patch density, edge density, water-vegetation interspersion), and 100 m "site cover" by various species of plants. The absence of such covariates may have been a driver of poor model fit for Pied-billed Grebe and Black Tern in this study; however, poor fit is not uncommon for marsh bird occupancy models and covariate selection for these species is likely informative - though unsuitable for predictive mapping.
While differences in habitat associations by guild were evident in this study, variability within guilds was also notable for the majority of land cover types considered. These differences occurred primarily in the scale of wetland cover effects, the type of wetland cover, and the nature of urban and agricultural cover effects, as well as differences in explanatory power (variable retention in final models). Such within-guild variation suggests that a species-specific, or at minimum life-history-based, approach to wetland conservation and landscape management is best - particularly for marsh birds - rather than applying generalized strategies. Nonetheless, to aid conservation and management, we make generalizations that capture the majority of the effects on occupancy for the marsh bird and waterfowl species we analyzed. We caution that these generalizations may not extend well beyond our study area or beyond the species we analyzed; however, given the prevalence of low marsh bird occupancy (low numbers of species suitable for robust modelling) and support for our findings in the literature, we feel such generalizations will be valuable.
For instance, we suggest that new protected areas intended for waterfowl can be selected in a way that includes added value for marsh birds. For the species strongly impacted by wetland cover (most marsh birds and Wood Duck, for which wetland covariates were retained in the final model), the "All Wetlands and Water" covariate type was among the best univariate predictors in the wetland group (∆AICc < 2). This suggests all wetlands and water would perform nearly as well in many final models where a different wetland covariate was ultimately retained. Considering also the differing spatial scale of effects, managers may be able to optimize site selection by maximizing percent cover by all wetlands and open water at both the local scale and the landscape scale (400 - 6400 m), prioritizing sites with high wetland and water cover located within large wetland complexes. Similarly, we recommend selecting sites with minimal urbanization within 800 m, the largest spatial scale within which marsh birds showed negative effects of such land cover. Waterfowl occupancy was promoted by urban-built cover well beyond 800 m, making these opposing habitat preferences potentially compatible. Finally, at current levels in the region, agricultural cover did not strongly affect the occurrence of marsh birds (only potential negative effect within 200 m for Black Tern, covariate absent from all final models for this guild), thus landscape-scale benefits of agricultural cover to Ring-necked Duck, and potentially Canada Goose and Mallard, currently don’t conflict with the habitat needs of marsh birds. However, we caution that if agricultural cover increases beyond current intermediate amounts, then these benefits will likely disappear due to wetland loss and other influences.
Our results are relevant to the selection of new wetland conservation sites, landscape management surrounding existing conservation sites, and the identification of wetlands where management actions for marsh birds would be most beneficial based on landscape context. We also provide support from eastern Canada that land cover habitat associations for marsh birds (and waterfowl) can depend on landscape composition as far as 6400 m from the site of interest. Future analyses should focus on the effects of site characteristics (hydrology, habitat complexity, food availability) to more fully understand the potential differences in habitat requirements of marsh birds and waterfowl. Where possible, future work should consider wetland size and isolation, as well as site cover and configuration. Finally, there is an opportunity to work with conservation partners to experimentally compare the effects of different management techniques on marsh birds and waterfowl. By increasing our understanding of habitat use within and between these two groups of species, we should be able to better construct sustainable working landscapes that benefit both birds and people.
Financial support for this study, including fieldwork and data analysis, was from Birds Canada, Environment and Climate Change Canada, New Brunswick Wildlife Trust Fund, New Brunswick Department of Natural Energy and Resource Development, Ducks Unlimited Canada, Nature Trust of New Brunswick, the Nature Conservancy of Canada, Canada Summer Jobs, Prince Edward Island Conservation Fund, Wildlife Habitat Canada, and National Wetland Conservation Fund. Furthermore, Birds Canada thanks many field technicians and especially our citizen science volunteers that have contributed hundreds of hours of survey effort to the Maritimes Marsh Monitoring Program.
Adde, A., M. Darveau, N. Barker, and S. Cumming. 2020. Predicting spatiotemporal abundance of breeding waterfowl across Canada: a Bayesian hierarchical modelling approach. Diversity and Distributions 26:1248-1263. https://doi.org/10.1111/ddi.13129
Alexander, B. W., and G. R. Hepp. 2014. Estimating effects of habitat characteristics on abundances of three species of secretive marsh birds in central Florida. Waterbirds 37:274-285. https://doi.org/10.1675/063.037.0306
Atlantic Coast Joint Venture (ACJV). 2020. BCR 14 Priority Species List. https://acjv.org/planning/bird-conservation-regions/bcr-14/.
Anderson, M. G., R. T. Alisauskas, B. D. J. Batt, R. J. Blohm, K. F. Higgins, M. C. Perry, J. K. Ringelman, J. S. Sedinger, J. R. Serie, D. E. Sharp, D. L. Trauger, and C. K. Williams. 2018. The migratory bird treaty and a century of waterfowl conservation. The Journal of Wildlife Management 82:247-259. https://doi.org/10.1002/jwmg.21326
Arnold, T. W. 2010. Uninformative parameters and model selection using Akaike’s Information Criterion. Journal of Wildlife Management 74:1175-1178. https://doi.org/10.1111/j.1937-2817.2010.tb01236.x
Bartón, K. 2020. MuMIn: multi-model inference. R package version 1.43.17.
Baschuk, M. S., N. Koper, D. A. Wrubleski, and G. Goldsborough. 2012. Effects of water depth, cover and food resources on habitat use of marsh birds and waterfowl in boreal wetlands of Manitoba, Canada. Waterbirds 35(1):44-55. https://doi.org/10.1675/063.035.0105
Birds Canada. 2020. Maritimes Marsh Monitoring Program. https://www.birdscanada.org/bird-science/marsh-monitoring-program/.
Bradshaw, T. M., A. G. Blake-Bradshaw, A. M. V. Fournier, J. D. Lancaster, J. O’Connell, C. N. Jacques, M. W. Eichholz, and H. M. Hagy. 2020. Marsh bird occupancy of wetlands managed for waterfowl in the Midwestern USA. PLoS ONE 15.e.0228980. https://doi.org/10.1371/journal.pone.0228980
Brinson, M. M., and A. I. Malvárez. 2002. Temperate freshwater wetlands: types, status, and threats. Conservation 29:115-133. https://doi.org/ https://doi.org/10.1017/S0376892902000085
Burnham, K., and D. Anderson. 2002. Model selection and multimodel inference: a practical information-theoretic approach, 2nd edition. Springer-Verlag, New York.
Connor, K. J., and S. Gabor. 2006. Breeding waterbird wetland habitat availability and response to water-level management in Saint John River floodplain wetlands, New Brunswick. Hydrobiologia 567:169-181. https://doi.org/10.1007/s10750-006-0051-1
Conway, C. J. 2011. Standardized North American marsh bird monitoring protocol. Waterbirds 34:319-346. https://doi.org/10.1675/063.034.0307
Conway, C. J., and J. P. Gibbs. 2011. Summary of intrinsic and extrinsic factors affecting detection probability of Marsh Birds. Wetlands 31:403-411. https://doi.org/10.1007/s13157-011-0155-x
Correll, M. D., W. A. Wiest, T. P. Hodgman, W. G. Shriver, C. S. Elphick, B. J. McGill, K. M. O’Brien, and B. J. Olsen. 2016. Predictors of specialist avifaunal decline in coastal marshes. Conservation Biology 31:172-182. https://doi.org/10.1111/cobi.12797
DeLuca, W. V., C. E. Studds, L. L. Rockwood, and P. P. Marra. 2004. Influence of land use on the integrity of marsh bird communities of Chesapeake Bay, USA. Society of Wetland Scientists. https://doi.org/10.1672/0277-5212(2004)024[0837:IOLUOT]2.0.CO;2
Eastern Habitat Joint Venture (EHJV). 2017. Eastern Habitat Joint Venture North American Waterfowl Management Plan Implementation Plan 2015-2020. Report of the Eastern Habitat Joint Venture. Environment and Climate Change Canada, Sackville, New Brunswick, Canada.
Eastern Habitat Joint Venture (EHJV). 2020. Eastern Habitat Joint Venture. Environment and Climate Change Canada, Sackville, New Brunswick, Canada. https://ehjv.ca/
Environment Canada. 2013. Bird Conservation Strategy for Bird Conservation Region 14 and Marine Biogeographic Units 11 and 12 in New Brunswick: Atlantic Northern Forest, Bay of Fundy and Gulf of St. Lawrence. Environment Canada, Ottawa. https://www.canada.ca/en/environment-climate-change/services/migratory-bird-conservation/regions-strategies/description-region-14/new-brunswick-select-marine-units.html
Environment Canada. 2019. North American Breeding Bird Survey - results for Canada. Government of Canada. https://wildlife-species.canada.ca/breeding-bird-survey-results.
Environmental Systems Research Institute (ESRI). 2014. ArcGIS Desktop: Release 10.3. Environmental Systems Research Institute, Redlands, California.
Ferree, C., and M. Anderson. 2013. A map of terrestrial habitats of the northeastern United States: methods and approach. The Nature Conservancy, Eastern Conservation Science, Eastern Regional Office. Boston, MA.
Fiske, I. J., and R. B. Chandler. 2011. Unmarked: An R package for fitting hierarchical models of wildlife occurrence and abundance. Journal of Statistical Software 43:1-23. https://doi.org/10.18637/jss.v043.i10
Fox, A. 2019. Urban geese - looking to North America for experiences to guide management in Europe? Wildfowl 69:3-27. https://wildfowl.wwt.org.uk/index.php/wildfowl/article/view/2688
Glisson, W. J., C. J. Conway, C. P. Nadeau, and K. L. Borgmann. 2017. Habitat models to predict wetland bird occupancy influenced by scale, anthropogenic disturbance, and imperfect detection. Ecosphere 8(6):e01837. https://doi.org/10.1002/ecs2.1837
Grand, J., S. P. Saunders, N. L. Michel, L. Elliott, S. Beilke, A. Bracey, T. M. Gehring, E. E. Gnass Giese, R. W. Howe, B. Kasberg, N. Miller, G. J. Niemi, C. J. Norment, D. C. Tozer, J. Wu, and C. Wilsey. 2020. Prioritizing coastal wetlands for marsh bird conservation in the U.S. Great Lakes. Biological Conservation 249:108708. https://doi.org/https://doi.org/10.1016/j.biocon.2020.108708
Ladin, Z. S., W. A. Wiest, M. D. Correll, E. L. Tymkiw, M. Conway, B. J. Olsen, C. S. Elphick, W. L. Thompson, and W. G. Shriver. 2020. Detection of local-scale population declines through optimized tidal marsh bird monitoring design. Global Ecology and Conservation 23:e01128. https://doi.org/10.1016/j.gecco.2020.e01128
Lieske, D. J., M. MacIntosh, L. Millet, S. Bondrup-Nielsen, J. B. Pollard, G. Parsons, N. R. McLellan, G. R. Milton, F. MacKinnon, K. Connor, and L. K. Banks. 2018. Modelling the impacts of agriculture in mixed-use landscapes: a review and case study involving two species of dabbling ducks. Landscape Ecology 33:35-57. https://doi.org/10.1007/s10980-017-0579-7
Lieske, D. J., B. Pollard, M. Gloutney, R. Milton, K. Connor, R. Dibblee, G. Parsons, and D. Howerter. 2012. The Importance of Agricultural Landscapes as Key Nesting Habitats for the American Black Duck in Maritime Canada. Source: Waterbirds 35:525-534. https://doi.org/10.1675/063.035.0403
MacKenzie, D. I., J. D. Nichols, G. B. Lachman, S. Droege, J. A. Royle, and C. A. Langtimm. 2002. Estimating site occupancy rates when detection probabilities are less than one. Ecology 83:2248-2255. https://doi.org/10.1890/0012-9658(2002)083[2248:ESORWD]2.0.CO;2
MacKey, B. G., D. W. McKenney, Y.-Q. Yang, J. P. McMahon, and M. F. Hutchinson. 1996. Site regions revisited: a climatic analysis of Hills’ site regions for the province of Ontario using a parametric method. Canadian Journal of Forest Research 26:333-354. https://doi.org/10.1139/x26-038
Mazerolle, M. J. 2019. AICcmodavg: model selection and multimodel inference based on (Q)AIC(c). R package version 2.2-2. https://cran.r-project.org/web/packages/AICcmodavg/AICcmodavg.pdf.
McAlister, M. A., C. E. Moorman, R. K. Meentemeyer, J. C. Fuller, D. L. Howell, and C. S. DePerno. 2017. Using landscape characteristics to predict distribution of temperate-breeding Canada Geese. Southeastern Naturalist 16:127-139. https://doi.org/10.1656/058.016.0201
McClure, C. J. W., and G. E. Hill. 2012. Dynamic versus static occupancy: How stable are habitat associations through a breeding season? Ecosphere 3:60. https://doi.org/10.1890/ES12-00034.1
McGowan, C., N. Angeli, W. Beisler, C. Snyder, N. Rankin, J. Woodrow, J. Wilson, E. Rivenbark, A. Schwarzer, C. Hand, R. Anthony, R. Griffin, K. Barrett, A. Haverland, N. Roach, T. Schnieder, A. Smith, F. Smith, J. Tolliver, and B. Watts. 2020. Linking monitoring and data analysis to predictions and decisions for the range-wide eastern black rail status assessment. Endangered Species Research 43:209-222. https://doi.org/10.3354/esr01063
McKenney, D., J. Pedlar, M. Hutchinson, P. Papadopol, K. Lawrence, K. Campbell, E. Milewska, R. F. Hopkinson, and D. Price. 2013. Spatial climate models for Canada’s forestry community. Forestry Chronicle 89:659-663. https://doi.org/10.5558/tfc2013-118
Messmer, D. J., S. A. Petrie, S. S. Badzinski, M. L. Gloutney, and M. L. Schummer. 2015. Habitat associations of breeding mallards and Canada geese in southern Ontario, Canada. Wildlife Society Bulletin 39:543-552. https://doi.org/10.1002/wsb.550
Mitsch, W. J., and M. E. Hernandez. 2013. Landscape and climate change threats to wetlands of North and Central America. Aquatic Sciences 75:133-149. https://doi.org/10.1007/s00027-012-0262-7
North American Bird Conservation Initiative Canada (NABCI). 2020. North American Bird Conservation Initiative Canada. Canadian Wildlife Service, Environment and Climate Change Canada http://nabci.net/
NAWMP. 2018. North American Waterfowl Management Plan. https://nawmp.org/
NAWMP Canada. 2013. North American Waterfowl Management Plan Canada. http://nawmp.wetlandnetwork.ca/
North American Bird Conservation Initiative. 2019. The State of the Birds - United States. Cornell Univeristy, Ithica, NY.
North American Bird Conservation Initiative Canada. 2019. The State of Canada’s Birds. Canadian Wildlife Service, Environment and Climate Change Canada, Gatineau, QC.
Orr, J. T., C. A. Duquette, T. J. Hovick, B. A. Geaumont, and T. M. Harms. 2020. Secretive marsh bird densities and habitat associations in the prairie pothole region. Wetlands 40:1529-1538. https://doi.org/10.1007/s13157-019-01251-x
R Core Team. 2019. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/
Roach, N. S., E. A. Hunter, N. P. Nibbelink, and K. Barrett. 2017. Poor transferability of a distribution model for a widespread coastal marsh bird in the southeastern United States. Ecosphere 8:e01715. https://doi.org/10.1002/ecs2.1715
Rota, C.T., R.J. Fletcher, R.M. Dorazio, and M.G. Betts. Occupancy estimation and the closure assumption. Journal of Applied Ecology 46:1173-1181. https://doi.org/https://doi.org/10.1111/j.1365-2664.2009.01734.x
Saunders, S. P., K. A. L. Hall, N. Hill, and N. L. Michel. 2019. Multiscale effects of wetland availability and matrix composition on wetland breeding birds in Minnesota, USA. AmericanOrnithology.org 121:1-15. https://doi.org/https://doi.org/10.1093/condor/duz024
Schielzeth, H. 2010. Simple means to improve the interpretability of regression coefficients. Methods in Ecology and Evolution 1:103-113. https://doi.org/10.1111/j.2041-210X.2010.00012.x
Smith, A. E., S. R. Craven, and P. D. Curtis. 1999. Managing Canada Geese in urban environments. Jack Berryman Institute Publication 16, and Cornell University Cooperative Extension, Ithica, NY.
Smith, L. A., and P. Chow-Fraser. 2010. Impacts of adjacent land use and isolation on marsh bird communities. Environmental Management 45:1040-1051. https://doi.org/10.1007/s00267-010-9475-5
Steidl, R. J., C. J. Conway, and A. R. Litt. 2012. Power to detect trends in abundance of secretive marsh birds: effects of species traits and sampling effort. The Journal of Wildlife Management 77:445-453. https://doi.org/10.1002/jwmg.505
Stevens, B. S., and C. J. Conway. 2020. Predictive multi‐scale occupancy models at range‐wide extents: effects of habitat and human disturbance on distributions of wetland birds. Diversity and Distributions 26:34-48. https://doi.org/10.1111/ddi.12995
Stewart, R. L., K. A. Bredin, J. W. Chardine, A. R. Couturier, A. G. Horn, D. Lepage, S. Makepeace, P. D. Taylor, M.-A. Villard, and R. M. Whittam. 2015. Second Atlas of Breeding Birds of the Maritime Provinces. Birds Canada, Ontario, Canada
The Nature Conservancy. 2015. Terrestrial habitat map for the northeast US and Atlantic Canada: Habitat map dataset. https://www.conservationgateway.org/ConservationByGeography/NorthAmerica/UnitedStates/edc/reportsdata/terrestrial/habitatmap/Pages/default.aspx
Tozer, D. C. 2016. Marsh bird occupancy dynamics, trends, and conservation in the southern Great Lakes basin: 1996 to 2013. Journal of Great Lakes Research 42:136-145. https://doi.org/10.1016/j.jglr.2015.10.015
Tozer, D. C., O. Steele, and M. Gloutney. 2018. Multispecies benefits of wetland conservation for marsh birds, frogs, and species at risk. Journal of Environmental Management 212:160-168. https://doi.org/10.1016/j.jenvman.2018.01.055
Tozer, D. C., R. L. M. Stewart, O. Steele, and M. Gloutney. 2020. Species‐habitat relationships and priority areas for marsh‐breeding birds in Ontario. The Journal of Wildlife Management 84:786-801. https://doi.org/10.1002/jwmg.21840
US Fish & Wildlife Service. 2020. Waterfowl population status, 2020. Washington, DC, USA.
US Geological Survey (USGS) Patuxent Wildlife Research Center. 2018. North American Breeding Bird Survey. U.S. Department of the Interior | U.S. Geological Survey. https://www.pwrc.usgs.gov/bbs/
Valente, J. J., S. L. King, and R. R. Wilson. 2011. Distribution and Habitat Associations of Breeding Secretive Marsh Birds in Louisiana’s Mississippi Alluvial Valley. Wetlands 31:1-10. https://doi.org/10.1007/s13157-010-0138-3
Williams, C. K. 2018. Special Section: Celebrating Waterfowl Conservation. The Journal of Wildlife Management 82:245-246. https://doi.org/10.1002/jwmg.21418