Habitat specialization explains avian persistence in tidal marshes Maureen D. Correll,1,† Whitney A. Wiest,2 Brian J. Olsen,1 W. Gregory Shriver,2 Chris S. Elphick,3 and Thomas P. Hodgman4 1School
of Biology and Ecology, The University of Maine, Orono, Maine 04469 USA of Entomology and Wildlife Ecology, The University of Delaware, Newark, Delaware 19716 USA 3Department of Ecology and Evolutionary Biology, Center for Conservation and Biodiversity, University of Connecticut, Storrs, Connecticut 06269 USA 4Maine Department of Inland Fisheries and Wildlife, Bangor, Maine 04401 USA
2Department
Citation: Correll, M. D., W. A. Wiest, B. J. Olsen, W. G. Shriver, C. S. Elphick, and T. P. Hodgman. 2016. Habitat specialization explains avian persistence in tidal marshes. Ecosphere 7(11):e01506. 10.1002/ecs2.1506
Abstract. Habitat specialists are declining at alarming rates worldwide, driving biodiversity loss of the
earth’s next mass extinction. Specialist organisms maintain smaller functional niches than their generalist counterparts, and tradeoffs exist between these contrasting life history strategies, creating conservation challenges for specialist taxa. There is little work, however, explicitly quantifying “specialization”; such information is necessary for the development of focused conservation strategies in light of the rapidly changing landscapes of the modern world. In this study, we tested whether habitat specialism explains the persistence of breeding bird populations in tidal marshes of the northeastern United States. We used the North American Breeding Bird Survey (BBS) together with contemporary marsh bird surveys to develop a Marsh Specialization Index (MSI) for 106 bird species that regularly use tidal marshes during the breeding season. We produced four metrics of species persistence (occupancy, abundance, total biomass supported, and 14-yr population trends) and compared them to MSI values in one of the first community-scale demonstrations of specialist loss in disturbed landscapes. Our results confirm that tidal marsh specialism has short-term benefits but long-term consequences for bird persistence in coastal marsh systems, results that are generalizable across many changing landscapes. We then use this robust support of niche theory to recommend MSI as a tool for quantitatively identifying species of conservation concern in disturbed and rapidly changing landscapes such as tidal marsh.
Key words: climate change; niche; specialism; species conservation; tidal marsh. Received 15 December 2015; revised 2 July 2016; accepted 22 July 2016. Corresponding Editor: W. A. Boyle. Copyright: © 2016 Correll et al. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. † E-mail:
[email protected]
Introduction
environment (Shea and Chesson 2002, Chase and Leibold 2003). Any particular point in ecological niche space will describe (1) the environmental requirements of a species to persist (birth rate ≥ death rate) and (2) the impacts that a species has upon its external environment at a particular place and time (Leibold 1995, Chesson 2000, Shea and Chesson 2002). Within this theoretical context, “specialism” is a necessarily relative term where specialist taxa are those with a smaller requirement niche breadth along at least one requirement gradient when compared to
Generalist and specialist life history strategies are fundamental concepts in ecology and can be explained most efficiently through the lens of the ecological niche (Grinnell 1917, Elton 1927, Hutchinson 1957, 1978, Macarthur and Levins 1967, Leibold 1995, Holt 2009). The modern niche concept is defined by the relationship between a group of organisms and their physical (e.g., temperature, precipitation) and biological (e.g., predator interactions, food availability) v www.esajournals.org
1
November 2016 v Volume 7(11) v Article e01506
Correll et al.
more generalist taxa (MacArthur 1972, Julliard et al. 2006). The generalist, while able to endure use of a wider breadth of physical environments and biological interactions, cannot exploit any one combination of environmental settings as effectively as their specialist counterparts. In a static environment, natural selection should faithfully favor the path of the specialist, whose competitive advantage (Levins 1968, MacArthur 1972) benefits species persistence in heterogeneous landscapes (Levins 1968, Kawecki 1994). However, there are known negative consequences to being a specialist. Specialists do not exploit novel environments well, and while they are at a competitive advantage at the center of their most specialized requirement axis, they are at a disadvantage outside of this zone compared with both other specialists and generalists. Specialists thus exhibit smaller range and population sizes, and more limited dispersal capabilities (Gaston et al. 1997, Colles et al. 2009) than generalists in the same landscape (Wilson and Yoshimura 1994, Devictor et al. 2008). Environmental setting is therefore integral to determining the fate of generalist vs. specialist species; predictable, unchanging landscapes favor specialists, while fluctuating landscapes favor the persistence of more generalist species (Levins 1968, Devictor et al. 2008). While theoretical tradeoffs between specialism and generalism are well documented, these concepts have not been rigorously quantified across taxa (Holt 2009). Degree of specialism may refer to variation across individuals, species, or functional groups (Bolnick et al. 2003, Blonder et al. 2014), or to different forms of adaptation (e.g., diet vs. habitat specialization). Quantification measures for specialism also vary, are often limited to a few species (Devictor et al. 2010), and are usually applied across multiple habitat types (Jonsen and Fahrig 1997, Julliard et al. 2006, Devictor et al. 2008). Defining and quantifying specialism in the context of particular habitat or ecosystem requirements, with the end goal of using these findings as conservation mechanisms, is the next logical step in the application of these principles to biodiversity conservation. Given the current mass extinction crisis (Barnosky et al. 2011) and fragmentation of global resources through direct and indirect anthropogenic effects (Fischer and Lindenmayer v www.esajournals.org
2007), the outlook is dire for specialists globally (Futuyma and Moreno 1988, Devictor et al. 2010). Worldwide habitat fragmentation may further decrease the persistence of specialist species. Rising global temperature, sea levels, and altered storm frequency and intensity can also create environmental conditions that fluctuate beyond the constrained niches maintained by specialist species. As a result, specialists have been referred to as the “great losers of past and current global changes” (Devictor et al. 2010), and specialism is now considered one of the dominant factors determining extinction of species (Dennis et al. 2011). Quantification of specialism within a particular habitat type therefore may not only provide a strategy for confirming long-standing theoretical concepts, but also be a potential indicator of overall conservation concern for a species. We explore the persistence of tidal marsh bird species in the northeastern United States as (1) a test of the costs of specialism in degraded, fragmenting landscapes as predicted by niche theory and (2) as a potential rapid assessment mechanism for conservation concern in tidal marsh bird species. These marshes have been used and modified heavily by humans since European colonization for agriculture, mosquito abatement, and ready access to the ocean (Bertness et al. 2002, Silliman and Bertness 2004, Gedan et al. 2009). Further, tidal marshes across the northeastern United States experience rates of sea- level rise roughly twice the global average, with even higher rates recorded within the past 5 yr (Sallenger et al. 2012). This marsh degradation may alter the landscape to the point of deviation from specialist niches, and make them well suited to test hypotheses about species persistence in this scenario. Answering these theoretical questions is also particularly important for conservation planning, as the results will have certain implications for the valuable ecosystem services tidal marshes provide to coastal communities (including biodiversity, Shepard et al. 2011). In this study, we define a measure of specialization to tidal marsh, the Marsh Specialization Index (MSI) for commonly detected species of tidal marsh birds. This index is akin to the Spe cies Specialism Index (SSI) developed by Devictor et al. (2006, 2008) but quantifies specialism relative to a single habitat type. Development of the MSI is intended to both advance the standardized 2
November 2016 v Volume 7(11) v Article e01506
Correll et al.
quantification of habitat specialization and serve as a tool for identifying species of conservation concern. To accomplish these objectives, we test for tradeoffs across a gradient of specialism between short-term success within the habitat (occupancy, abundance, and biomass) and persistence in the face of habitat change (14-yr population trends). Further, we use existing regional bird conservation assessments (Partners in Flight, or PIF, scores) in northeastern tidal marshes to assess the utility of the MSI as a future rapid assessment tool for the conservation status of tidal marsh species. We expect to see increased short-term success and decreased long-term success with increasing MSI value. Additionally, we expect to see a positive relationship between regional PIF score and MSI value.
We identified the most commonly detected species in northeastern U.S. tidal marshes by scree plot (n = 106) of species relative abundance during the 2012 survey season. To quantify tidal marsh specialization for these species, we then compared these relative abundance estimates in tidal marsh from 2012 to relative abundance in terrestrial systems as measured by the North American Breeding Bird Survey (BBS, Sauer et al. 2015). The BBS is a long-running, continental monitoring program comprised of 3-min point counts within a 400-m detection radius conducted along a series of predetermined, roadside survey routes. For each species, we summed count data across all BBS routes where the route center point was within 100 km of the coastline across our survey region. We corrected for effort by dividing the BBS sum by the number of routes (n = 170) and number of count stops on each route (n = 50). Likewise, we summed our count data from tidal marshes for each species, using detections recorded during the first 3 min of each survey at an unlimited detection radius at each survey point. Again, we corrected for effort by dividing the sum of all individuals counted by the number of total visits across all point counts in 2012. To produce our index of specialization for each species, we divided tidal marsh relative abundance by the sum of tidal marsh and terrestrial (BBS) relative abundance. This produces an index for each species (MSI) quantifying its degree of habitat specialization to tidal marsh, with values ranging from 0 (terrestrial specialist) to 1 (tidal marsh specialist) with habitat generalists occurring at intermediate values between these two extremes. This index assumes equal detection probability for each species across habitats and equates 400 m radius counts (BBS data) with unlimited radius counts (tidal marsh data). These detection distances are likely equivalent, as detection and identification of birds to species >400 m from an observer are extremely rare (Emlen 1971).
Methods Developing the Marsh Specialization Index (MSI)
We conducted contemporary bird surveys at 1770 locations during the summers of 2011–2012 between coastal Maine and Virginia (Appendix S1). These survey sites were selected using a Gen eralized Random Tessellation Stratified (GRTS) sampling scheme as described in Wiest et al. (2016). Each survey location was visited two to three times each year between 15 April and 31 July, with smaller, shifting survey windows within each state to account for differences in local phenology. All surveys were completed between sunrise and 11 a.m. by observers proficient in visual and aural identification of tidal marsh birds; technicians were trained through a standardized process at the beginning of the field season and supervised throughout to maintain consistency. When possible, we revisited locations from historical bird surveys during data collection (n = 457). These historical locations were spread across all states in the survey region, although more were located in New England (Maine to Connecticut, n = 265) than the Mid-Atlantic (New York to Virginia, n = 192). Surveys consisted of a 5-min passive point count during which we recorded all birds observed using marsh habitat. We also recorded time of detection as well as the distance of each bird from the observer using distance band categories (0–50 m, 50–100 m, >100 m). Our sampling scheme is fully described in the study by Wiest et al. (2016). v www.esajournals.org
Species metrics
Once we calculated MSI for all 106 species, we selected a subset of species for all further analyses that (1) used northeastern tidal marshes during their breeding season, (2) occurred with enough evenness and regularity across our study
3
November 2016 v Volume 7(11) v Article e01506
Correll et al.
area to withstand a robust trend analysis, and (3) had MSI values of >0.5 to explore the gradient of habitat generalism (MSI ~0.5) to tidal marsh specialism (MSI ~1.0). We excluded beach- and platform-nesting species from this analysis because their abundance in tidal marshes is likely tied to proximity of adequate breeding habitat, and not quality of the tidal marsh habitat that they were using when detected.
et al. (2016) to generate 14-yr population trends for each species. The resulting database contains records of birds observed using tidal marsh during a passive 5-min point count conducted between sunrise and 11:00 h between 1 April and 1 August of the survey year. The vast majority of historical data have records for both 50 and 100 m radii (n = 2782 points); however, due to differing distance sampling methodologies, a small number of observations were limited to 100-m (n = 93) distance bands, making the databases with which we generated population trends at the 50 and 100-m scales slightly different from one another, although both databases were similar in spatial spread across the survey region. Both historical databases contained observational data from many more survey locations in New England states than states in the Mid-Atlantic. Thus, the population trends reported here could amplify patterns in the north and underplay existing patterns in the south. While other studies using this historical data set find similar trends in tidal marsh species when spatially stratifying their analysis by region (Correll et al. 2016), our findings should still be considered with a northern data bias in mind. See Appendix S2 for additional detail. We modeled population change using generalized fixed-effect models (GLM) in a likelihood framework in R (R Core Team 2015). In all analyses, we only used survey points that overlapped the Estuarine Intertidal Emergent Wetland layer of the National Wetlands Inventory (NWI) and were within the published breeding range for each species (Cornell Lab of Ornithology 2015). We modeled regional population trends for each species following model structure and fit assessment described in Correll et al. (2016). We used 50 or 100-m distance band detections to produce population trends depending upon the natural history of each species. See Appendix S3 for additional detail.
Occupancy, abundance and biomass
We modeled probability of occupancy and abundance of tidal marsh birds in northeastern U.S. coastal marshes using N-mixture models (Royle 2004) in a likelihood framework using the package unmarked (Fiske and Chandler 2011) in R (R Core Team 2015). We used our regional survey data from 2012 to produce these estimates. We used the function “occu” to estimate mean probability of occupancy and “pcount” to estimate mean abundance across all surveyed points. We used observation-level covariates of Julian day, time of day, and tidal stage to account for differences in detection probability across visits. For each species, we only included survey sites within a species’ published breeding range (Cornell Lab of Ornithology 2015). We calculated confidence intervals for these estimates using the Wald approximation function. Estimates of occupancy and abundance apply to the area of marsh contained within a 100 m radius circle (31,416 m2). To estimate average biomass supported, we recorded average adult biomass for each species using Cornell Lab of Ornithology’s (2015) estimates for each species and used the mean when multiple mass estimates were given for a species (i.e., across sexes or subspecies). For Nelson’s sparrow (Ammodramus nelsoni) and saltmarsh sparrow (A. caudacutus), we used estimates from more recent work on these two species along the Atlantic coast (Ruskin 2015). We then took the product of the average biomass and the point abundance estimate for each species to produce a value for average biomass supported.
Conservation status
We investigated conservation status information for each of the 22 species through review of the Partners in Flight (PIF) combined concern score for landbirds in Bird Conservation Region (BCR) 30 (Partners in Flight Science Committee 2012). This physiographic region covers from coastal Maine to Virginia, nearly equivalent to
Population trends
We combined our regional survey data from 2011 to 2012 with a historical database of point counts (n = 1550 additional survey points) conducted in tidal marshes from Maine to Virginia, spanning the years 1998–2012 following Correll v www.esajournals.org
4
November 2016 v Volume 7(11) v Article e01506
Correll et al.
our focal area for this study. These scores were produced in 2012 and are the most up-to-date assessment of regional conservation need for coastal landbirds in the northeastern United States. Of the 22 species, 13 are landbirds and were assigned scores in the 2012 BCR 30 database. As the 2012 PIF scores exclude waterbird species, we also reviewed the Priority Species Pools generated for the physiographic areas of Northern New England (Area 27: Hodgman and Rosenberg 2000), Southern New England (Area 09: Dettmers and Rosenberg 2000), and the Mid-Atlantic (Area 44: Watts 1999), which together make up our focal area for this study. These species scores were generated in the mid- 1990s, but are the most recent quantitative assessment of species conservation concern inclusive of waterbird species. Of the 22 species meeting our criteria, 11 were assigned PIF prioritization scores in at least one of the physiographic region plans. When a species was listed and ranked in a priority species pool for more than one physiographic area, we used the mean of the scores across all areas as a combined score for that species. In both the 2012 and earlier assessments, the PIF prioritization plan (Carter et al. 2000) combines assessments of breeding and nonbreeding distributions, relative abundance, and population trends to assign a single conservation score for each species in relation to that physiographic area.
similarities between species but still allows for some variation across these groups. We tested for the effect of specialization by comparing all models to the intercept-only model, and models with a difference in Akaike’s information criterion for small sample sizes (ΔAICc) ≤2.0 were considered equivalent (Burnham and Anderson 2002). We used both fixed and random effects to calculate degrees of freedom for AICc values. We further assessed GLMM model fit using marginal R2 values using the R package MuMIn (Barton 2015). We used linear mixed-effects quantile regressions in a likelihood framework using the lqmm package (Geraci 2014) to further explore the relationship between degree of specialism and biomass supported. We compared models with τ ranging from 0.1 to 0.9 in 0.1 increments to models with τ = 0.5 (equivalent to linear regression). We again assessed relative model performance with AICc values.
Results MSI values
Across the 106 species, MSI values ranged from 0.01 (tufted titmouse, Baeolophus bicolor) to 1.00 (saltmarsh sparrow and others, see Appendix S4). We identified 22 species that fit species selection criteria (Fig. 1). Probability of occupancy point estimates (reported with 95% CIs) ranged from 0.02 (0.01, 0.03) for the alder flycatcher (Empidonax alnorum, Table 1) to 0.70 (0.68, 0.72) for the red-winged blackbird (Agelaius phoeniceus). Point abundance estimates for each of the selected species ranged from 0.3 individuals (0.01, 6.4) for the alder flycatcher to 19.97 individuals (18.25, 21.85) for the red-winged blackbird. The mean biomass supported at each survey point ranged from 4 g (0, 9) for the alder flycatcher to 1593 g (1246, 2035) for the clapper rail. Population trend parameter estimates ranged from −0.43 (−0.56, −0.31) for the saltmarsh sparrow to 0.61 (0.47, 0.75) for the yellow warbler (Setophaga petechia).
Analyses
We tested for relationships individually between the degree of specialization (MSI) and each of the four metrics of species success and persistence (probability of occupancy, abundance, biomass supported, and population trend) using linear mixed-effects models in a likelihood framework using the package lme4 (Bates et al. 2015). We also explored the relationship between MSI and the combined PIF score in a similar model framework to compare assessments of conservation need. To meet assumptions of normality, we first log-transformed two of the five dependent variables (abundance, and biomass supported) and used a logit transformation on probability of occupancy values. To control for the effects of phylogeny, we included taxonomic family as a random effect in all models. This taxonomic grouping accounts for phylogenetic v www.esajournals.org
Analysis of species persistence
We found a negative linear relationship bet ween long-term success in tidal marshes (population trend parameter estimate) and marsh habitat specialism (MSI, Table 2, Fig. 2). Negative parameter estimates occurred, on average, when MSI ≥
5
November 2016 v Volume 7(11) v Article e01506
Correll et al.
Fig. 1. Sliding scale of tidal marsh specialization represented through a Marsh Specialization Index (MSI), a quotient of the amount of tidal marsh detections vs. total species detections in a combined database of North American Breeding Bird Survey records and tidal marsh bird surveys conducted in 2012.
0.93. We found a positive linear relationship between one metric of short-term success (biomass supported) and MSI (Fig. 2c, dotted line). We found no relationship between either point occupancy or abundance and MSI value. MSI and regional PIF score were also positively related using 2012 landbird and physiographic region PIF scores (Fig. 3; Appendix S5). We found improved fit between biomass supported and MSI value using quantile regression where τ ≥ 0.8 (Fig. 2c, dashed line). Model fit was not improved for population trend by varying τ from 0.5.
landscapes as predicted by niche theory. In our analyses, we found a negative relationship bet ween population trend and MSI (Fig. 2d), indicating the more specialized a species is to tidal marsh habitat, the less likely it is to persist in this ecosystem over time. When we examined this pattern across each avian family individually (Appendix S6), we found that species with higher MSI values had lower population trends in five (Anatidae, Rallidae, Hirundinidae, Emberizidae, and Icteridae) of the seven avian families examined (not Ardeidae or Parulidae), suggesting that no one family was driving the larger pattern. Additionally, the biomass supported at a survey point for a given species was positively related to MSI value (Fig. 2c), indicating that specialized species have the ability to support larger amounts of biomass per unit of marsh than of their generalist counterparts. This relationship was quantile rather than linear in nature, indicating specialism was a constraining factor instead of a linear predictor of the overall biomass supported by a particular species. Simply put, specialists in tidal marsh had the option of maintaining low or high amounts of biomass, while generalists were limited in the biomass they can support. These findings are consistent with the hypothesis that generalists are limited by their
Discussion Habitat specialists in a rapidly changing landscape
Tradeoffs exist between specialist and generalist life history strategies. One result of these tradeoffs is that specialists are predicted to reach higher densities than generalists within their defined niche space (Dennis et al. 2011), but habitat generalists are predicted to outperform specialists when these landscapes are degraded or fragmented to the point of divergence from specialist environmental requirements. In our study of tidal marsh birds in the northeastern United States, we found empirical support for the negative consequences of specialism in changing v www.esajournals.org
6
November 2016 v Volume 7(11) v Article e01506
Correll et al.
Table 1. Specialization and persistence metrics for 22 bird species occurring in tidal marsh between Maine and Virginia (for the remaining 84 species calculated, see Appendix S4).
Common name
Family
Individual biomass (g)
Saltmarsh sparrow
Emberizidae
19.7
1.00
Nelson’s sparrow
Emberizidae
17.3
1.00
Seaside sparrow
Emberizidae
24.2
0.99
Marsh wren
Troglodytidae
12.2
0.99
Clapper rail
Rallidae
280
0.99
Scolopacidae
237.2
0.98
Snowy egret
Ardeidae
369
0.98
American black duck Great egret
Anatidae
1203.5
0.98
Ardeidae
473.5
0.97
Virginia rail
Rallidae
82.5
0.96
Boat-tailed grackle
Icteridae
152.6
0.95
Threskiornithidae
650
0.93
Willet
Glossy ibis
MSI
Red-winged blackbird Mallard
Icteridae
54
0.86
Anatidae
1150
0.85
Tree swallow
Hirundinidae
19.5
0.83
Ardeidae
212.4
0.76
Savannah sparrow
Emberizidae
19.9
0.76
Song sparrow
Emberizidae
24.2
0.72
Parulidae
9.8
0.69
Hirundinidae
18.7
0.65
Alder flycatcher
Tyrannidae
13.5
0.64
Common yellowthroat
Parulidae
10.1
0.61
Great blue heron
Yellow warbler Barn swallow
Point abundance
Probability of occupancy
Biomass supported (g)
Trend
4.25 (3.24, 5.58) 1.93 (1.16, 3.2) 6.13 (4.95, 7.58) 5.78 (4.46, 7.49) 5.69 (4.45, 7.27) 5.93 (5.08, 6.91) 1.45 (1.18, 1.79) 0.76 (0.43, 1.35) 1.77 (1.41, 2.21) 0.54 (0.09, 3.38) 1.43 (1.06, 1.93) 1.25 (0.9, 1.73) 19.97 (18.25, 21.85) 0.89 (0.7, 1.13) 4.36 (3.73, 5.09) 1.57 (0.84, 2.93) 0.72 (0.25, 2.07) 4.21 (2.76, 6.41) 4.26 (2.46, 7.4) 7.46 (6.52, 8.54) 0.3 (0.01, 6.4) 7.45 (4.66, 11.92)
0.22 (0.2, 0.25) 0.09 (0.08, 0.1) 0.28 (0.26, 0.3) 0.26 (0.24, 0.29) 0.26 (0.24, 0.28) 0.36 (0.33, 0.38) 0.22 (0.18, 0.27) 0.06 (0.04, 0.08) 0.24 (0.21, 0.27) 0.03 (0.02, 0.05) 0.1 (0.09, 0.12) 0.08 (0.07, 0.11) 0.7 (0.68, 0.72) 0.13 (0.11, 0.16) 0.33 (0.3, 0.36) 0.17 (0.09, 0.28) 0.04 (0.03, 0.05) 0.49 (0.46, 0.51) 0.2 (0.18, 0.22) 0.43 (0.41, 0.46) 0.02 (0.01, 0.03) 0.37 (0.34, 0.39)
83 (64, 110) 33 (20, 55) 148 (120, 183) 71 (54, 91) 1592 (1246, 2035) 1406 (1206, 1638) 535 (434, 659) 916 (519, 1619) 836 (668, 1047) 45 (7, 278) 218 (162, 294) 813 (586, 1126) 1078 (985, 1180) 1023 (805, 1298) 85 (73, 99) 334 (179, 622) 14 (5, 41) 102 (67, 155) 42 (24, 73) 140 (122, 160) 4 (0, 86) 75 (47, 120)
−0.43 (−0.56, −0.31) −0.21 (−0.33, −0.08) 0.05 (−0.21, 0.3) −0.07 (−0.33, 0.17) −0.34 (−0.61, −0.06) 0.13 (−0.01, 0.27) −0.11 (−0.25, 0.02) 0.04 (−0.21, 0.29) 0.26 (0.12, 0.41) 0.23 (−0.32, 0.88) −0.21 (−0.68, 0.25) 0.49 (0.22, 0.76) 0.34 (0.27, 0.41) 0.39 (0.19, 0.59) 0.02 (−0.11, 0.15) −0.09 (−0.25, 0.06) 0.02 (−0.27, 0.31) 0.1 (0.03, 0.18) 0.61 (0.47, 0.75) 0.24 (0.13, 0.34) 0.35 (−0.13, 0.92) 0.38 (0.29, 0.47)
Note: Parentheses contain 95% CIs.
ability to efficiently exploit any one particular environment, while specialists are not. These two main findings quantitatively support the dark future generally predicted for habitat specialists worldwide. As ecosystems are fragmented and our global climate changes at rates unprecedented within recent geological history (Urban 2015), habitat specialists reliant v www.esajournals.org
upon predictability of a single habitat type will be outcompeted by generalists. This pattern has previously been demonstrated in single-species studies quantifying the declines of specialists worldwide (Clavel et al. 2011); however, our findings provide robust empirical support for these theoretical predictions at the community level across a suite of tidal marsh bird species. 7
November 2016 v Volume 7(11) v Article e01506
Correll et al.
Table 2. Model results comparing metrics of species persistence in tidal marshes to Marsh Specialization Index (MSI) using linear mixed-effects model (LMM) and linear quantile mixed-effects model (LQMM). Metric Abundance Occupancy Biomass supported Biomass supported Trend Partners in Flight (PIF) score
Model type
MSI β estimate (CI)
ΔAICc
Marginal R2
LMM LMM LMM LQMM LMM LMM
0.00 (−1.57, 1.56) −0.03 (−1.57, 1.56) 2.37 (0.4, 4.34) 2.92 (1.2, 4.64) −0.94 (−1.69, −0.19) 3.7 (0.51, 6.9)
−0.64 1.19 4.41 5.25 3.32 5.37
