Ecological correlates of vulnerability to fragmentation in forest birds on inundated subtropical land-bridge islands

Ecological correlates of vulnerability to fragmentation in forest birds on inundated subtropical land-bridge islands

Biological Conservation 191 (2015) 251–257 Contents lists available at ScienceDirect Biological Conservation journal homepage: www.elsevier.com/loca...

679KB Sizes 0 Downloads 40 Views

Biological Conservation 191 (2015) 251–257

Contents lists available at ScienceDirect

Biological Conservation journal homepage: www.elsevier.com/locate/bioc

Ecological correlates of vulnerability to fragmentation in forest birds on inundated subtropical land-bridge islands Yanping Wang a, Daniel H. Thornton b, Dapeng Ge a, Siyu Wang c, Ping Ding a,⁎ a b c

College of Life Sciences, Zhejiang University, Hangzhou 310058, China School of Environment, Washington State University, P.O. Box 642812, Pullman WA 99164-2812, United States of America Zhejiang Museum of Natural History, Hangzhou 310012, China

a r t i c l e

i n f o

Article history: Received 18 March 2015 Received in revised form 22 June 2015 Accepted 30 June 2015 Available online xxxx Keywords: Detectability Ecological traits Habitat fragmentation Occupancy Synergistic effect Thousand Island Lake

a b s t r a c t Identifying the ecological and life-history traits that render species vulnerable to fragmentation is an important prerequisite for the development of effective conservation strategies to minimize future biodiversity losses. When determining how species traits influence vulnerability to fragmentation, however, several important confounding factors such as detectability and synergistic effects among traits are rarely considered. In this study, after controlling for these methodological shortcomings, we determined how species traits influenced fragmentation vulnerability using bird data collected from islands created by the inundation of the Thousand Island Lake, China. We obtained eight species traits from field surveys and from the literature: natural abundance, geographical range size, habitat specificity, body size, trophic level, mobility, fecundity, and nest type. After phylogenetic correction, these traits were used separately and in combination to assess their associations with the index of fragmentation vulnerability, the proportion of islands occupied. Inclusion of detectability in analysis resulted in considerable increases in overall island occupancy for all species in general and for cryptic species in particular. Accounting for detectability altered the rank of best models and thus influenced the identification of the relationships between species traits and fragmentation vulnerability. We found synergistic interactions between natural abundance and habitat specificity. Our findings highlight the importance of incorporating detectability and synergistic effects among traits into future studies. From a conservation perspective, our results suggest that we should give priority conservation efforts to rare species with low natural abundance and high habitat specificity. © 2015 Elsevier B.V. All rights reserved.

1. Introduction Habitat loss and fragmentation are widely considered to be one of the primary threats to biological diversity throughout the world (Wilcove et al., 1998; Fahrig, 2003). Many studies have shown that following fragmentation, the resident faunal community will undergo a period of species loss or relaxation before a new equilibrium community is achieved (MacArthur and Wilson, 1967; Diamond, 1972; Terborgh, 1974; Wilcox, 1978). During the process of relaxation, species in isolated fragments usually would disappear in a predictable order on the basis of differential extinction vulnerabilities (Feeley et al., 2007; Meyer et al., 2008; Wang et al., 2009). Identifying the ecological and life-history traits that render species vulnerable to fragmentation is an important prerequisite for the development of effective conservation strategies to minimize future biodiversity losses (Laurance, 1991; Henle et al., 2004; Meyer et al., 2008). Theory suggests that species with particular traits are more vulnerable to fragmentation than others (McKinney, 1997; Purvis et al., 2000; ⁎ Corresponding author. E-mail addresses: [email protected] (Y. Wang), [email protected] (D.H. Thornton), [email protected] (D. Ge), [email protected] (S. Wang), [email protected] (P. Ding).

http://dx.doi.org/10.1016/j.biocon.2015.06.041 0006-3207/© 2015 Elsevier B.V. All rights reserved.

Henle et al., 2004). First, rarity in the form of small population size, high habitat specificity and small geographic distribution have been widely recognized as good predictors of fragmentation vulnerability (Rabinowitz et al., 1986; Gaston, 1994; Purvis et al., 2000). Second, large body size, high trophic level, low mobility and low fecundity (small clutch size) are commonly hypothesized to increase a species' susceptibility to fragmentation (Bennett and Owens, 1997; Purvis et al., 2000; Henle et al., 2004; Van Houtan et al., 2007). Finally, species with exposed nests or ground nesting are also predicted to be more vulnerable to fragmentation because of higher nest predation rates (Terborgh, 1974; Wilcove, 1985). Although identifying the ecological traits that predispose species to fragmentation has long been a focus of research in conservation biology (Meyer et al., 2008), several confounding factors may hinder the development of this research field. First, the issues of imperfect detection are rarely considered in most studies (Thornton et al., 2011a,b). Species could be vulnerable to fragmentation merely because they are hard to detect (MacKenzie et al., 2002). Including these cryptic species in analysis without controlling for detectability will underestimate the actual number of fragments occupied by such species (MacKenzie et al., 2002), which in turn would bias the identification of the relationships between species traits and fragmentation vulnerability. In addition, synergistic effects

252

Y. Wang et al. / Biological Conservation 191 (2015) 251–257

among traits are seldom considered in analyses (Henle et al., 2004). The synergistic effect indicates that the traits involved act nonadditively, rendering species more vulnerable to fragmentation than the additive effect predicted by single traits (Davies et al., 2004). The synergistic effect among traits is a pressing challenge for conservation biologists because it makes predictions about vulnerability to fragmentation from a single trait unreliable. The land-bridge islands created by dam construction and the inundation of hydroelectric reservoirs are ideal models or “experimental” systems for studying habitat fragmentation (Diamond, 2001; Wu et al., 2003). First, these land-bridge islands are surrounded by a homogeneous and relatively inhospitable matrix (water), which largely eliminates the confounding matrix effects commonly found in mainland fragments (Laurance, 1991; Newmark et al., 2014). Second, all the islands are formed simultaneously within a short period of time, which controls for the confounding effects of fragmentation history on community composition (Bolger et al., 2000; Fernandez-Juricic, 2000). Finally, many such land-bridge islands in Lake Guri, Venezuela, Sinnamary River, French Guiana and Three-Gorges Dam, China have been inundated very recently (Cosson et al., 1999; Wu et al., 2003; Feeley et al., 2007) and may be still in the process of faunal relaxation (Diamond, 1972; Brooks et al., 1999), providing invaluable opportunities to examine the ongoing processes and mechanisms driving local species extinction. In this study, we determined how imperfect detection and synergistic effects would influence the relationships between species traits and fragmentation vulnerability using bird data collected from islands created by the inundation of the Thousand Island Lake, China. We selected a priori eight well-defined and commonly used species traits that are often hypothesized to influence fragmentation vulnerability: natural abundance, geographical range size, habitat specificity, body size, trophic level, mobility, fecundity, and nest type (Wilcove, 1985; McKinney, 1997; Purvis et al., 2000; Henle et al., 2004). We made the following three hypotheses. First, we hypothesized that imperfect detection would influence the identification of the relationships between species traits and fragmentation vulnerability (MacKenzie et al., 2002; Thornton et al., 2011a). Second, we hypothesized that we would find nonadditive relationships between at least some of the eight traits, given the prevalence of the idea of synergistic interactions in the literature (Lawton, 1994; Davies et al., 2004). Finally, we hypothesized that two or more traits in combination would be more important than a single trait, as single traits alone usually have very limited predictive powers for vulnerability to fragmentation (Henle et al., 2004). Understanding the relationships between species traits and fragmentation vulnerability has important implications for proactive conservation and can be used to help direct management efforts. 2. Material and methods

−7.6 °C in January to 41.8 °C in July. Annual precipitation in the region is 1430 mm, with an average of 155 days of precipitation per year (Wang et al., 2015). As land-bridge islands, the faunas of the archipelago and mainland have the same origin (Wang and Feng, 2009; Wang et al., 2010). 2.2. Bird surveys The study was carried out across a set of 42 islands and four neighboring mainland sites. These islands were selected to represent a range of area size and degree of isolation from the mainland (Table A1) (Wang et al., 2011, 2012). To facilitate surveys, we cut transects (c. 20 cm wide) that traversed the small islands entirely and the mountain ridges of large islands (Wang et al., 2010, 2011). The maximum elevations of the study islands ranged from 106.7 m to 298.5 m (Table A1). To account for the greater habitat variability associated with larger areas, sampling effort was roughly proportional to log (island area) (Wang et al., 2011, 2012). Accordingly, eight transects were sampled on the largest study island 1 (area N 1000 ha), four on islands 2–3 (1000 N area N 100 ha), two on islands 4–7 (100 N area N 10 ha), and one on each of the remaining small islands (area ≈ 1 ha for most islands; Table A1) (Wang et al., 2011, 2012). We used the line-transect method (Bibby et al., 2000) to determine bird occupancy and abundance on the study islands during four breeding seasons (April–June) and four winter seasons (November–January) from 2007 to 2011. During the survey, an individual observer walked each transect at a steady pace (c. 2.0 km/h). We recorded all bird species seen or heard within 50 m of the transect lines, but not high-flying species that just passed over. Surveys were usually conducted from 0.5 h after dawn through to 11:00 h in the mornings and from 15:00 to 0.5 h before sunset in the afternoons. However, surveys were not conducted in the mid-day period because bird activity is low in that interval (Wang et al., 2011, 2012). Surveys were also not conducted during inclement weather such as heavy rain, strong wind or high temperatures (Robbins, 1981). We used GPS receivers to record the length of each transect (Table A1). Each transect on the islands was surveyed 120 times. A similar sampling method was used on the mainland sites as on the islands. To determine the distribution and abundance of bird species in the unfragmented habitats, we censused four plots that were located on a peninsula that juts into the lake. These plots were chosen to match the islands appropriately in elevation, slope, distance from the coast, and vegetation type (Wang et al., 2009). Any birds detected within the plots were recorded using the line-transect method. As on the islands, each transect on mainland sites was surveyed 120 times. To eliminate potential biases owing to observer fatigue or weather conditions, the order in which the sites were surveyed and the direction in which the transect lines were walked were randomized and rotated each new census day (Wang et al., 2010, 2011, 2012).

2.1. Study sites 2.3. Ecological traits We conducted the research in the Thousand Island Lake (29°22′– 29°50′N, 118° 34′–119° 15′E) and adjacent mainland in Zhejiang Province, China. The Thousand Island Lake is a large hydroelectric reservoir that was created in 1959 by the damming of the Xinanjiang River in western Zhejiang Province (Wang et al., 2009). Construction of the Xinanjiang dam inundated an area of 573 km2, creating 1078 islands larger than 0.25 ha out of former hilltops when the water reached its highest level (108 m) (Wang et al., 2009, 2010). The total land area of the archipelago is 409 km2 (Wang et al., 2012). The archipelago and the adjacent mainland have similar vegetation, climate, topography and faunas (Wang and Feng, 2009). The major vegetation type on the islands is a successional forest dominated by the Masson pine (Pinus massoniana) (Wang et al., 2011, 2012). The climate is typical of the subtropical monsoon zone and is highly seasonal, with hot summers and cold winters. The average annual temperature is 17.0 °C, ranging from

We collected data on eight ecological traits for each species using field surveys and published literature (Table A2). We used body length (mm) to represent body size and used clutch size as an index of reproductive potential (Wang et al., 2009). Feeding habits were defined according to Wang et al. (2011) and were quantified as omnivores (1), granivores (2), insectivores (3), and carnivores (4). Wang et al. (2010) identified seven habitat types in the study region: (1) conifer forest, (2) broadleaf forest, (3) coniferous–broadleaf mixed forests, (4) bamboo groves, (5) shrubs, (6) grassland, and (7) farmland (Table A3). Habitat specificity was based on the incidence or number of habitats used by a given species (Feeley et al., 2007; Wang et al., 2010). Its value was the number of habitats used by the species (Table A2) (Feeley et al., 2007; Sodhi et al., 2010). That is, if a species used or occurred only in one habitat type, its habitat specificity was attributed a value of 1 and it was

Y. Wang et al. / Biological Conservation 191 (2015) 251–257

considered highly specific, and vice verse (Wang et al., 2010). Following Jones et al. (2003), the geographic range size (km2) was obtained from the most recent available published species range maps by digitizing the area into a Geographic Information System (ArcView 10.2). Where no range maps were available, the area of the minimum convex polygon of published point data was calculated excluding areas of water (Wang et al., 2010). To obtain an index of a species' mobility, we calculated a dispersal ratio (dp) for each species by dividing its mean wing length (mm) by the cube root of its mean mass (g) (Woinarski, 1989; Fischer and Lindenmayer, 2005). Nest type was classified as cavity (tree or others) (1), open in tree (2), open in shrub (3), and open on ground (4) (Barbaro and van Halder, 2009). All the above data were obtained from Zhuge (1990), and when not accessible were supplemented with Zhao (2001) and Zheng (2005). For each of the species traits, if a range instead of the mean was given, we used the arithmetic mean of the limits (Gaston and Blackburn, 1995). In addition to the traits collected from the literature, we also calculated natural abundance for each species based on the results of the transect surveys at mainland sites (Davies et al., 2000, 2004; Meyer et al., 2008).

2.4. Statistical analyses 2.4.1. Measuring vulnerability to fragmentation Overall patch occupancy, or species prevalence, is widely used as one of the most common measures of vulnerability to fragmentation (e.g. Soulé et al., 1988; Swihart et al., 2003; Viveiros de Castro and Fernandez, 2004; Meyer et al., 2008; Wang et al., 2009). Following Thornton et al. (2011a), we used occupancy modeling as implemented in the program PRESENCE (Hines, 2014) to compute the proportion of islands occupied by each species that controlled for imperfect detection as the measure of vulnerability to fragmentation. The results of the proportion of islands occupied corrected for detectability were then compared with that of naïve proportion of islands occupied by bird species to test our first hypothesis that imperfect detection would influence the identification of the relationships between species traits and fragmentation vulnerability (MacKenzie et al., 2002). We used the program PRESENCE (Hines, 2014) to model occupancy and detection probabilities using the detection/nondetection data collected from 42 islands for 60 species (see below). We modeled occupancy (ψ) as a function of island size, isolation, and habitat richness, which are the three most important variables influencing the distribution of birds in our region (Wang et al., 2010, 2011). We modeled detection probability (p) as a function of survey effort (number of transects), which is the most obvious covariate for modeling detection probability (Ferraz et al., 2007). For each species, we built a combination of 16 occupancy and detection models (Table A4). We then identified the best-fit model using Akaike information criterion corrected for small sample size (AICc) for each species (Table A5). Based on the final best-fit model, we calculated our measure of vulnerability to fragmentation, i.e. proportion of islands occupied, by summing the individual occupancy estimates for each island and dividing by the total number of islands (n = 42) for each species (Thornton et al., 2011a). We used the single-season model (MacKenzie et al., 2002), rather than multiple-season model, mainly for three reasons. First, we were focusing on overall vulnerability to fragmentation (i.e. proportion of islands occupied) (Thornton et al., 2011a), not the extinction–colonization dynamics of study species (Ferraz et al., 2007). Second, some species with high dispersal abilities that often visit islands but are not true residents or breeding individuals were excluded from analysis. Finally, our previous study indicated that the overall species turnover between sampling years was low (Wang et al., 2010). Thus, although many species could move on and off islands during sampling, the observed species distribution represents primarily patterns of occupancy (Wang et al., 2010).

253

2.4.2. Influence of species traits on fragmentation vulnerability To control for phylogenetic non-independence between species, we fitted phylogenetic generalized least squares (PGLS) models to the data using the pgls function of the caper package in R (Orme et al., 2012; R Development Core Team, 2014). Pagel's λ, a branch length transformation indicating the strength of the phylogenetic signal (Pagel, 1999), was optimized in each model by the maximum likelihood method (Orme et al., 2012). The other two branch length transformations, κ and σ, were set as constant (equal to 1) which assumed a Brownian motion model of evolution in each model (Orme et al., 2012). To test our hypotheses, we built a set of relevant PGLS models. We began our analysis by examining the significance of each of the eight ecological predictors of fragmentation vulnerability separately. We then tested the importance of the combinations of species traits and their synergistic interactions as predictors of fragmentation vulnerability. We limited our models to plausible a priori co-varying ecological traits (Table A6), rather than conducting exploratory analyses with all combinations of the eight ecological traits. We only considered the first order of combinations of ecological traits and their synergistic interactions because of limited sample size. We also conducted analyses on data not corrected for phylogeny, and for comparison provided those qualitatively similar results in the supplementary Tables A7 and A8. We calculated AICc and adjusted R2 for each fitted model. The difference in AICc values (Δi) between models can be used to calculate Akaike weights (ωi), which is the probability that the model is the best in the set of candidate models, given the data (Burnham and Anderson, 2002). Only models with Δi ≤ 2 are considered to have substantial support (Burnham and Anderson, 2002). Prior to analyses, logarithmic transformations were performed on body length, geographic range size and natural abundance to achieve normality. Due to correlations between species traits (Laurance, 1991), we assessed the potential effect of collinearity on the results of multivariate analyses by calculating variance inflation factors (VIFs) using R (R Development Core Team, 2014). In all cases, the VIFs were b 4.5 and thus well below the commonly accepted critical thresholds for significant collinearity of 10.0 (Neter et al., 1996). 2.4.3. Data used in the analysis We omitted some groups of species from the present analysis for several reasons. First, aquatic species were excluded because their distributions were mainly driven by the presence of suitable water bodies (Wang et al., 2010, 2011). Second, some nocturnal and crepuscular species, such as the owls and nightjars, were excluded because they could not be reliably detected (Wang et al., 2012, 2013). Third, swifts and swallows were excluded because their use of islands was difficult to determine (Wang et al., 2011). Finally, winter migratory species were also excluded because they only occupied islands during migratory passage periods and thus are not true residents or breeding individuals (Wang et al., 2013). Accordingly, a total of 60 forest breeding species were analyzed (Table A2). We then built a bifurcating phylogeny for these 60 species (Fig. 1) based upon the molecular phylogeny developed by Sibley and Ahlquist (1990). 3. Results 3.1. Vulnerability to fragmentation The 60 bird species included in the analyses exhibited considerable variation in vulnerability to fragmentation as measured by the proportion of islands occupied (Table A2). Occupancy of islands in our study area ranged from 0.0238 (e.g. Gracupica nigricollis) to 1.0 (e.g. Parus major, Aegithalos concinnus, Pycnonotus sinensis) (Table A2). Inclusion of detectability in the analyses resulted in considerable increases in overall island occupancy for all species in general (mean ± SE = 0.093 ± 0.017) and for cryptic species in particular. For cryptic species, the estimated percentage of islands occupied increased from 11.0%

254

Y. Wang et al. / Biological Conservation 191 (2015) 251–257 Table 1 Performance of 36 phylogenetic generalized least squares (PGLS) models relating naïve proportion of islands occupied by bird species as a measure of fragmentation vulnerability to plausible combinations of eight ecological traits.

Fig. 1. Phylogenetic tree of the 60 bird species used in the comparative analysis. The phylogeny is based on Sibley and Ahlquist (1990). We made the assumption that all branches in the phylogeny were of equal length.

(Leiothrix lutea) to 35.7% (Falco tinnunculus) from the naïve estimate that did not control for detectability (Table A2).

ΔAICc

ωi

10.62 33.36 49.78 51.73 51.95 53.15 53.25 53.25

11.58 34.32 50.74 52.70 52.91 54.11 54.21 54.22

0.0017 1.95 × 10−8 5.31 × 10−12 1.99 × 10−12 1.79 × 10−12 9.82 × 10−13 9.35 × 10−13 9.32 × 10−13

4 4 4 4 4 4 4 4 4 4 4 4 4 4

−0.51 12.32 12.57 12.73 34.05 35.35 35.59 50.59 51.42 51.97 53.87 54.06 54.25 55.45

0.45 13.28 13.54 13.69 35.02 36.31 36.55 51.56 52.38 52.93 54.84 55.02 55.21 56.41

0.4405 0.0007 0.0006 0.0006 1.38 × 10−8 7.21 × 10−9 6.39 × 10−9 3.53 × 10−12 2.33 × 10−12 1.77 × 10−12 6.84 × 10−13 6.25 × 10−13 5.68 × 10−13 3.11 × 10−13

3 3 3 3 3 3 3 3 3 3 3 3 3 3

−0.96 10.47 11.42 34.53 34.76 37.90 49.62 51.13 51.36 51.79 51.92 51.97 52.92 53.18

0.00 11.43 12.38 35.50 35.72 38.86 50.59 52.09 52.32 52.75 52.88 52.93 53.89 54.14

0.5529 0.0018 0.0011 1.08 × 10−8 9.68 × 10−9 2.01 × 10−9 5.73 × 10−12 2.70 × 10−12 2.40 × 10−12 1.94 × 10−12 1.82 × 10−12 1.77 × 10−12 1.10 × 10−12 9.69 × 10−13

Model description

K

Single-process models Natural abundance Habitat specificity Dispersal ratio Trophic level Body size Clutch size Nest type Geographic range size

3 3 3 3 3 3 3 3

Additive models Natural abundance + habitat specificity Body size + natural abundance Geographic range size + Natural abundance Trophic level + natural abundance Body size + habitat specificity Trophic level + habitat specificity Habitat specificity + geographic range size Body size + trophic level Clutch size + dispersal ratio Body size + dispersal ratio Trophic level + geographic range size Body size + geographic range size Body size + clutch size Clutch size + nest type Interactive models Natural abundance × habitat specificity Geographic range size × natural abundance Body size × natural abundance Habitat specificity × geographic range size Trophic level × natural abundance Body size × habitat specificity Body size × dispersal ratio Clutch size × dispersal ratio Body size × trophic level Body size × geographic range size Trophic level × habitat specificity Trophic level × geographic range size Body size × clutch size Clutch size × nest type

AICc

3.2. Influence of species traits on fragmentation vulnerability

4. Discussion

For our measure of fragmentation vulnerability that was uncorrected for detectability (naïve proportion of islands occupied), model selection based on AICc identified the synergistic interaction between natural abundance and habitat specificity (ΔAICc = 0, ωi = 0.5529) as the best correlate of vulnerability to fragmentation (Table 1). The additive model of natural abundance and habitat specificity was ranked second (ΔAICc = 0.45, ωi = 0.4405) (Table 1). The interactive model (adjusted R2 = 0.63) and the additive model (adjusted R2 = 0.62) both explained a large amount of variance, which was much larger than that predicted separately from a single model of habitat specificity (adjusted R2 = 0.29) or natural abundance (adjusted R2 = 0.54). Accounting for imperfect detection altered the rank of best models, although not the combination of species traits (Table 2). For the proportion of islands occupied corrected for detectability as a measure of fragmentation vulnerability, model selection based on AICc identified the additive model of natural abundance and habitat specificity (ΔAICc = 0, ωi = 0.7189) as the best correlate of vulnerability to fragmentation (Table 2). The synergistic interaction between natural abundance and habitat specificity (ΔAICc = 1.92, ωi = 0.2750) was the second parsimonious model (Table 2). The additive model (adjusted R2 = 0.58) and the interactive model (adjusted R2 = 0.56) both explained a large amount of variance, which was much larger than that predicted separately from a single model of habitat specificity (adjusted R2 = 0.30) or natural abundance (adjusted R2 = 0.47).

4.1. Influence of species traits on fragmentation vulnerability In this study we examined how species traits influenced fragmentation vulnerability in forest birds on inundated subtropical land-bridge islands in the Thousand Island Lake, China. Accounting for imperfect detection influenced the identification of the relationships between species traits and fragmentation vulnerability. We found synergistic interactions between natural abundance and habitat specificity. In contrast, any single ecological trait received much lower support as a predictor of species vulnerability to fragmentation. Consistent with our first hypothesis, we found that imperfect detection altered the rank of best models and thus influenced the identification of the relationships between species traits and fragmentation vulnerability (MacKenzie et al., 2002). Imperfect detection is a widely acknowledged problem in ecological studies (MacKenzie et al., 2006). However, the issue of detection error is rarely considered in most studies that correlate species traits with fragmentation vulnerability (Thornton et al., 2011a,b). Our study provided an ideal long detection history data set (n = 120) to correct detection error. Accounting for imperfect detection for cryptic species avoids the error of underestimating the actual number of islands occupied by such species in our system (MacKenzie et al., 2002, 2006), which in turn help us to identify the true and unbiased relationships between species traits and fragmentation vulnerability.

Y. Wang et al. / Biological Conservation 191 (2015) 251–257

255

Table 2 Performance of 36 phylogenetic generalized least squares (PGLS) models relating the proportion of islands occupied corrected for detectability as a measure of fragmentation vulnerability to plausible combinations of eight avian ecological traits. Model description

K

AICc

ΔAICc

ωi

Single-process models Natural abundance Habitat specificity Dispersal ratio Body size Trophic level Geographic range size Clutch size Nest type

3 3 3 3 3 3 3 3

15.57 30.37 48.56 49.75 49.78 51.59 51.63 51.65

12.56 27.36 45.55 46.75 46.77 48.58 48.62 48.64

0.0013 8.23 × 10−7 9.23 × 10−11 5.08 × 10−11 5.02 × 10−11 2.03 × 10−11 1.99 × 10−11 1.97 × 10−11

Additive models Natural abundance + habitat specificity Geographic range size + natural abundance Body size + natural abundance Trophic level + natural abundance Body size + habitat specificity Trophic level + habitat specificity Habitat specificity + geographic range size Body size + trophic level Clutch size + dispersal ratio Body size + dispersal ratio Body size + geographic range size Body size + clutch size Trophic level + geographic range size Clutch size + nest type

4 4 4 4 4 4 4 4 4 4 4 4 4 4

3.01 17.27 17.60 17.72 30.34 32.45 32.62 49.00 50.37 50.69 52.00 52.05 52.08 53.89

0.00 14.26 14.59 14.71 27.33 29.44 29.61 45.99 47.37 47.69 48.99 49.04 49.07 50.88

0.7189 0.0006 0.0005 0.0005 8.33 × 10−7 2.91 × 10−7 2.68 × 10−7 7.43 × 10−11 3.73 × 10−11 3.17 × 10−11 1.66 × 10−11 1.61 × 10−11 1.59 × 10−11 6.42 × 10−12

Interactive models Natural abundance × habitat specificity Geographic range size × natural abundance Body size × natural abundance Habitat specificity × geographic range size Body size × habitat specificity Trophic level × natural abundance Body size × dispersal ratio Body size × trophic level Body size × geographic range size Trophic level × geographic range size Trophic level × habitat specificity Clutch size × dispersal ratio Body size × clutch size Clutch size × nest type

3 3 3 3 3 3 3 3 3 3 3 3 3 3

4.93 15.15 15.35 31.52 35.20 37.43 48.32 48.93 49.80 49.86 49.96 49.98 51.45 51.49

1.92 12.14 12.34 28.51 32.19 34.42 45.32 45.92 46.79 46.86 46.95 46.97 48.44 48.49

0.2750 0.0017 0.0015 4.63 × 10−7 7.36 × 10−8 2.41 × 10−8 1.04 × 10−10 7.66 × 10−11 4.98 × 10−11 4.81 × 10−11 4.59 × 10−11 4.54 × 10−11 2.18 × 10−11 2.13 × 10−11

In accordance with our second hypothesis, we found synergistic interactions between natural abundance and habitat specificity. We found that for habitat specialists, there was a strong significant association between natural abundance and fragmentation vulnerability; but for habitat generalists, there was no relationship between natural abundance and fragmentation vulnerability (Fig. 2). In our study system, habitat specialists tended to have low natural abundance (r = 0.306, p = 0.017), which in turn were more vulnerable to fragmentation. Theories and empirical evidence have long suggested that some species are extremely vulnerable to fragmentation because they have combinations of traits that promote extinction (Lawton, 1994; McKinney, 1997; Henle et al., 2004). So far, however, only Davies et al. (2004) has explicitly examined synergistic interactions between species traits. Our study provides further evidence that synergistic interactions between traits can put species at greater risk of extinction than the additive effect predicted by single traits. The results also support our third hypothesis that two or more traits in combination are more important than a single trait to predict vulnerability to fragmentation. We found that any single ecological trait was not sufficient to predict species vulnerability to fragmentation. Had we not analyzed the combined effects of species traits on fragmentation vulnerability, we would have incorrectly concluded that natural abundance was the single best predictor of fragmentation vulnerability in our system (Tables 1–2, A7-A8). To date, few studies have considered

Fig. 2. Interactive models of the responses of bird species to fragmentation based on their natural abundance and whether they were habitat specialists (occurring in 1–3 habitats) or generalists (occurring in 4–6 habitats). The upper and lower lines represent predictions from the regression models for habitat generalists and habitat specialists, respectively. Circles are habitat generalists, and triangles are habitat specialists.

the combined effects of species traits on fragmentation vulnerability (e.g. Swihart et al., 2003; Davies et al., 2004; Meyer et al., 2008). In most studies, however, species traits were analyzed in isolation (e.g. Soulé et al., 1988; Feeley et al., 2007; Wang et al., 2009), probably due to the small sample size involved or other reasons. For example, only 5 bird species were analyzed in Bolger et al. (1991), and only 12 small mammals were included in Viveiros de Castro and Fernandez (2004). Our data support the idea that single traits alone usually have limited predictive powers for vulnerability to fragmentation (Henle et al., 2004), and suggest that future studies, if possible, should analyze the combined effects of species traits on fragmentation vulnerability. Other ecological traits such as population fluctuation, annual survival rates, home range size, social flocking propensity, edge or disturbance sensitivity, have also been proposed as potential predictors of fragmentation vulnerability in other systems (Pimm et al., 1988; Henle et al., 2004; Van Houtan et al., 2006; Meyer et al., 2008; Benchimol and Peres, 2015). The role of these traits in determining fragmentation vulnerability cannot be excluded from the present study. These traits alone or in combination probably may account for some of the remaining variation in species vulnerability to fragmentation in our system. As we currently have no data on these traits, however, the idea warrants further study. We used the proportion of islands occupied by species as a measure of fragmentation vulnerability. However, a species could possibly be absent from an island simply because the island lacks suitable habitats, even before the creation of the dam. This problem could be addressed properly in a pre- vs. post-fragmentation comparison. However, like many other fragmented systems (e.g. Davies et al., 2000; Meyer et al., 2008), historical data on bird occurrences and habitat composition are unavailable because the dam was built about 60 years ago. Despite the lack of historical data, several lines of indirect evidences suggest that the habitat composition on the study islands was similar before dam construction, and thus not a confounding factor in our study. First, the archipelago and the adjacent mainland belong to the same mountain system, and thus have similar vegetation composition (Wang and Feng, 2009). Second, the habitat composition on the study islands are similar to those on unfragmented mainland control sites (Wang et al., 2009), which largely represent pre-fragmentation conditions (Meyer et al., 2008; Benchimol and Peres, 2015). Finally, except for several large islands, all the small islands had very limited variation in habitats and elevations (Tables A1, A3), probably due to the relatively small scale (573 km2) of our study system.

256

Y. Wang et al. / Biological Conservation 191 (2015) 251–257

4.2. Conservation implications Our results highlight the importance of imperfect detection, synergistic effects and combined effects of species traits in determining fragmentation sensitivity. These findings have several general and specific implications with regard to conservation practices in our system. First, the issues of imperfect detection should be taken into consideration in devising management plans as it influenced the identification of the relationships between species traits and fragmentation vulnerability. Second, due to the synergistic effect of species traits, it may be risky to make predictions about fragmentation vulnerability from additive models of ecological traits (Davies et al., 2004). In contrast, we should consider the synergistic interaction between species traits for effective conservation. Furthermore, as single traits alone are poor predictors of fragmentation vulnerability, it would be inefficient to allocate conservation resources based on any single ecological trait. Specifically, our results show that the combination of natural abundance and habitat specificity and their synergistic interactions are key predictors of fragmentation vulnerability. Therefore, conservation efforts giving priority to rare species with low natural abundance and high habitat specificity may prove effective for the preservation of bird species in this system. It is important to determine whether our findings are applicable to other fragmented systems. We found that the combination of natural abundance and habitat specificity and their synergistic interactions were key drivers of fragmentation vulnerability in our system. Natural abundance and habitat specificity are among the traits commonly identified as important predictors of fragmentation vulnerability in many kinds of fragmented systems such as urbanized fragments, inundated islands, and the Wog Wog experimental fragments (Soulé et al., 1988; Bolger et al., 1991; Cosson et al., 1999; Renjifo, 1999; Davies et al., 2004; Feeley et al., 2007). It thus seems that ecological traits identified by us as important are useful as general predictors of fragmentation vulnerability in multiple systems (but see Anjos, 2006; Van Houtan et al., 2006, 2007). However, the important confounding factors such as detectability and synergistic effects among traits are rarely considered in these systems. Further studies should control for these confounding factors and use other taxa in a wide range of systems with different fragment matrices to effectively determine the generality of our findings. From a conservation perspective, such comparative analyses will be especially valuable for developing predictive models of species vulnerability to fragmentation and devising efficient management strategies. Acknowledgements We thank Lenore Fahrig, Richard Corlett and two anonymous referees for valuable comments on the manuscript. We are grateful to Jingcheng Zhang, Peng Li, Meng Zhang, Zhifeng Ding and Jiji Sun for field assistance, and Chunan Forestry Bureau and the Thousand Island Lake National Forest Park for permits necessary to conduct the research. This study was supported by the National Natural Science Foundation of China (31100394, 31471981, 31210103908) and the Project-sponsored by SRF for ROCS, SEM (J20130585). Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.biocon.2015.06.041. References Anjos, L. dos, 2006. Bird species sensitivity in a fragmented landscape of the Atlantic forest in southern Brazil. Biotropica 38, 229–234. Barbaro, L., van Halder, I., 2009. Linking bird, carabid beetle and butterfly life-history traits to habitat fragmentation in mosaic landscapes. Ecography 32, 321–333. Benchimol, M., Peres, C.A., 2015. Predicting local extinctions of Amazonian vertebrates in forest islands created by a mega dam. Biol. Conserv. 187, 61–72.

Bennett, P.M., Owens, I.P.F., 1997. Variation in extinction risk among birds: chance or evolutionary predisposition? Proc. R. Soc. B Biol. Sci. 264, 401–408. Bibby, C.J., Burgess, N.D., Hill, D.A., Mustoe, S., 2000. Bird Census Techniques. Academic Press, London. Bolger, D.T., Alberts, A.C., Soule, M.E., 1991. Occurrence patterns of bird species in habitat fragments: sampling, extinction, and nested species subsets. Am. Nat. 137, 155–166. Bolger, D.T., Suarez, A.V., Crooks, K.R., Morrison, S.A., Case, T.J., 2000. Arthropods in urban habitat fragments in southern California: area, age, and edge effects. Ecol. Appl. 10, 1230–1248. Brooks, T.M., Pimm, S.L., Oyugi, J.O., 1999. Time lag between deforestation and bird extinction in tropical forest fragments. Conserv. Biol. 13, 1140–1150. Burnham, K.P., Anderson, D.R., 2002. Model Selection and Multimodel Inference: A Practical Information–Theoretic Approach. Springer, New York. Cosson, J.F., Ringuet, S., Claessens, O., de Massary, J.C., Dalecky, A., Villiers, J.F., Granjon, L., Pons, J.M., 1999. Ecological changes in recent land-bridge islands in French Guiana, with emphasis on vertebrate communities. Biol. Conserv. 91, 213–222. Davies, K.F., Margules, C.R., Lawrence, J.F., 2000. Which traits of species predict population declines in experimental forest fragments? Ecology 81, 1450–1461. Davies, K.F., Margules, C.R., Lawrence, J.F., 2004. A synergistic effect puts rare, specialized species at greater risk of extinction. Ecology 85, 265–271. Diamond, J.M., 1972. Biogeographic kinetics: estimation of relaxation times for avifaunas of Southwest Pacific islands. Proc. Natl. Acad. Sci. U. S. A. 69, 3199–3203. Diamond, J.M., 2001. Dammed experiments. Science 294, 1847–1848. Fahrig, L., 2003. Effects of habitat fragmentation on biodiversity. Annu. Rev. Ecol. Evol. Syst. 34, 487–515. Feeley, K.J., Gillespie, T.W., Lebbin, D.J., Walter, H.S., 2007. Species characteristics associated with extinction vulnerability and nestedness rankings of birds in tropical forest fragments. Anim. Conserv. 10, 493–501. Fernandez-Juricic, E., 2000. Bird community composition patterns in urban parks of Madrid: the role of age, size and isolation. Ecol. Res. 15, 373–383. Ferraz, G., Nichols, J.D., Hines, J.E., Stouffer, P.C., Bierregaard Jr., R.O., Lovejoy, T.E., 2007. A large-scale deforestation experiment: effects of patch area and isolation on Amazon birds. Science 315, 238–241. Fischer, J., Lindenmayer, D.B., 2005. Nestedness in fragmented landscapes: a case study on birds, arboreal marsupials and lizards. J. Biogeogr. 32, 1737–1750. Gaston, K.J., 1994. Rarity. Chapman & Hall, London. Gaston, K.J., Blackburn, T.M., 1995. Birds, body size and the threat of extinction. Philos. Trans. R. Soc. B. Biol. Sci. 347, 205–212. Henle, K., Davies, K., Kleyer, M., Margules, C., Settele, J., 2004. Predictors of species sensitivity to fragmentation. Biodivers. Conserv. 13, 207–251. Hines, J.E., 2014. PRESENCE 6.9: Software to Estimate Patch Occupancy and Related Parameters. USGS-PWRC, Patuxent Wildlife Research Center, Laurel, Maryland (http://www.mbr-pwrc.usgs.gov/software/presence.html (accessed October 2014)). Jones, K.E., Purvis, A., Gittleman, J.L., 2003. Biological correlates of extinction risk in bats. Am. Nat. 161, 601–614. Laurance, W.F., 1991. Ecological correlates of extinction proneness in Australian tropical rain forest mammals. Conserv. Biol. 5, 79–89. Lawton, J.H., 1994. Population dynamic principles. Philos. Trans. R. Soc. B. Biol. Sci. 344, 61–68. MacArthur, R.H., Wilson, E.O., 1967. The Theory of Island Biogeography. Princeton University Press, Princeton, New Jersey. MacKenzie, D.I., Nichols, J.D., Lachman, G.B., Droege, S., Royle, J.A., Langtimm, C.A., 2002. Estimating site occupancy rates when detection probabilities are less than one. Ecology 83, 2248–2255. MacKenzie, D.I., Nichols, J.D., Royle, J.A., Pollock, K.H., Bailey, L.L., Hines, J.E., 2006. Occupancy Estimation and Modelling: Inferring Patterns and Dynamics of Species Occurrence. Elsevier, Boston, Massachusetts. McKinney, M.L., 1997. Extinction vulnerability and selectivity: combining ecological and paleontological views. Annu. Rev. Ecol. Evol. Syst. 28, 495–516. Meyer, C.F.J., Fründ, J., Lizano, W.P., Kalko, E.K.V., 2008. Ecological correlates of vulnerability to fragmentation on neotropical bats. J. Appl. Ecol. 45, 381–391. Neter, J., Kutner, M.H., Nachtsheim, C.J., Wasserman, W., 1996. Applied Linear Statistical Model: Regression, Analysis of Variance, and Experimental Design. Irwin Professional Publishing, Chicago. Newmark, W.D., Stanley, W.T., Goodman, S.M., 2014. Ecological correlates of vulnerability to fragmentation among Afrotropical terrestrial small mammals in northeast Tanzania. J. Mammal. 95, 269–275. Orme, D., Freckleton, R., Thomas, G., Petzoldt, T., Fritz, S., Isaac, N., Pearse, W., 2012. CAPER: comparative analyses of phylogenetics and evolution in R. R package version 0.5. http://CRAN.R-project.org/package=caper (accessed October 2014). Pagel, M., 1999. Inferring the historical patterns of biological evolution. Nature 401, 877–884. Pimm, S.L., Jones, H.L., Diamond, J., 1988. On the risk of extinction. Am. Nat. 132, 757–785. Purvis, A., Gittleman, J.L., Cowlishaw, G., Mace, G.M., 2000. Predicting extinction risk in declining species. Proc. R. Soc. B Biol. Sci. 267, 1947–1952. R Core Team, 2014. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria (http://www.R-project.org/ (accessed October 2014)). Rabinowitz, D., Cairns, S., Dillon, T., 1986. Seven form of rarity and their frequency in the flora of the British Isles. In: Soulé, M.E. (Ed.), Conservation Biology: The Science of Scarcity and Diversity. Sinauer, Sunderland, Massachusetts, pp. 182–204. Renjifo, R.M., 1999. Composition changes in a subandean avifauna after long-term forest fragmentation. Conserv. Biol. 13, 1124–1135. Robbins, C.S., 1981. Bird activity levels related to weather. Stud. Avian Biol. 6, 301–310.

Y. Wang et al. / Biological Conservation 191 (2015) 251–257 Sibley, C.G., Ahlquist, J.E., 1990. Phylogeny and Classification of Birds: A Study in Molecular Evolution. Yale University Press, New Haven. Sodhi, N.S., Wilcove, D.S., Lee, T.M., Sekercioglu, C.H., Subaraj, R., Bernard, H., Yong, D.L., Lim, S.L.H., Prawiradilaga, D.M., Brook, B.W., 2010. Deforestation and avian extinction on tropical landbridge islands. Conserv. Biol. 24, 1290–1298. Soulé, M.E., Bolger, D.T., Alberts, A.C., Wrights, J., Sorice, M., Hill, S., 1988. Reconstructed dynamics of rapid extinctions of chaparral-requiring birds in urban habitat islands. Conserv. Biol. 2, 75–92. Swihart, R.K., Gehring, T.M., Kolozsvary, M.B., Nupp, T.E., 2003. Responses of ‘resistant’ vertebrates to habitat loss and fragmentation: the importance of niche breadth and range boundaries. Divers. Distrib. 9, 1–18. Terborgh, J., 1974. Preservation of natural diversity: the problem of extinction prone species. Bioscience 24, 715–722. Thornton, D., Branch, L., Sunquist, M., 2011a. Passive sampling effects and landscape location alter associations between species traits and response to fragmentation. Ecol. Appl. 21, 817–829. Thornton, D., Branch, L., Sunquist, M., 2011b. The relative influence of habitat loss and fragmentation: do tropical mammals meet the temperate paradigm? Ecol. Appl. 21, 2324–2333. Van Houtan, K.S., Pimm, S.L., Bierregaard Jr., R.O., Lovejoy, T.E., Stouffer, P.C., 2006. Local extinctions in flocking birds in Amazonian forest fragments. Evol. Ecol. Res. 8, 129–148. Van Houtan, K.S., Pimm, S.L., Halley, J.M., Bierregaard Jr., R.O., Lovejoy, T.E., 2007. Dispersal of Amazonian birds in continuous and fragmented forest. Ecol. Lett. 10, 219–229. Viveiros de Castro, E.V.B., Fernandez, F.A.S., 2004. Determinants of differential extinction vulnerabilities of small mammals in Atlantic forest fragments in Brazil. Biol. Conserv. 119, 73–80. Wang, J.M., Feng, B.C., 2009. The Development Chorography of Xinanjiang River. Zhejiang People's Publishing House, Hangzhou. Wang, Y., Zhang, J., Feeley, K.J., Jiang, P., Ding, P., 2009. Life-history traits associated with fragmentation vulnerability of lizards in the Thousand Island Lake, China. Anim. Conserv. 12, 329–337.

257

Wang, Y., Bao, Y., Yu, M., Xu, G., Ding, P., 2010. Nestedness for different reasons: the distributions of birds, lizards and small mammals on island of an inundated lake. Divers. Distrib. 16, 862–873. Wang, Y., Chen, S., Ding, P., 2011. Testing multiple assembly rule models in avian communities on islands of an inundated lake, Zhejiang Province, China. J. Biogeogr. 38, 1330–1344. Wang, Y., Zhang, M., Wang, S., Ding, Z., Zhang, J., Sun, J., Li, P., Ding, P., 2012. No evidence for the small-island effect in avian communities on islands of an inundated lake. Oikos 121, 1945–1952. Wang, Y., Ding, P., Chen, S., Zheng, G., 2013. Nestedness of bird assemblages on urban woodlots: implications for conservation. Landsc. Urban Plan. 111, 59–67. Wang, Y., Wu, Q., Wang, X., Liu, C., Wu, L., Chen, C., Ge, D., Song, X., Chen, C., Xu, A., Ding, P., 2015. Small-island effect in snake communities on islands of an inundated lake: the need to include zeroes. Basic Appl. Ecol. 16, 19–27. Wilcove, D.S., 1985. Nest predation in forest tracts and the decline of migratory songbirds. Ecology 66, 1211–1214. Wilcove, D.S., Rothstein, D., Dubow, J., Phillips, A., Losos, E., 1998. Quantifying threats to imperiled species in the United States. Bioscience 48, 607–615. Wilcox, B.A., 1978. Supersaturated island faunas: a species-age relationship for lizards on post-Pleistocene land-bridge islands. Science 199, 996–998. Woinarski, J.C.Z., 1989. Some life history comparisons of small leaf-gleaning bird species of south-eastern Australia. Corella 13, 73–80. Wu, J., Huang, J., Han, X., Xie, Z., Gao, X., 2003. Three-Gorge Dam — experiment in habitat fragmentation. Science 300, 1239–1240. Zhao, Z., 2001. A Handbook of the Birds of China. Jilin Science and Technology Publishing House, Changchun. Zheng, G., 2005. A Checklist on the Classification and Distribution of the Birds of China. Science Press, Beijing. Zhuge, Y., 1990. Fauna of Zhejiang: Aves. Zhejiang Science and Technology Publishing House, Hangzhou.