The importance of isolated patches for maintaining local bird biodiversity and ecosystem function : a case study from the Pernambuco Center of Endemism , Northeast Brazil

The Atlantic Forest has been highly fragmented, with the Pernambuco Center of Endemism (PCE) one of the priority areas for conservation. The Mata do Cedro forest, located in Alagoas state, northeastern Brazil, is a forest fragment within the PCE surrounded by a matrix of sugarcane that acts as a refuge for several threatened bird species, some of which are endemic to the region. Here, we characterize the bird community in Mata do Cedro using measures of species abundance, frequency of occurrence, habitat use and sensitivity to human disturbance. The functional role of species was investigated with a functional dendrogram. We registered 111 species, most resident and forest dependent. The most representative trophic categories were insectivores followed by frugivores. Of the species found, 11 are highly sensitive to human disturbances and 11 are endemic to the PCE. The bird community of the fragment is highly diversified, with endemic taxa and balanced trophic categories typical of preserved tropical forests. This community structure together with the occurrence of threatened species reinforces the importance of Mata do Cedro for the maintenance of local biodiversity and ecosystem functions.

Urban ecosystems are complex, heterogenic and dynamic, characterized mainly by dense agglomerations of people living in the same place.The urbanization process involves changes in the landscape, soil modifications, climate changes, and biodiversity loss, resulting in a new, distinct ecosystem (Pickett et al., 2011).City growth changes the landscape -destroying natural habitats and creating new ones -and native species are replaced by a pool of a few species adapted to the urban environment (urban exploiters), promoting biotic homogenization (Blair, 1996(Blair, , 2004;;Rolando et al., 1997;Clergeau et al., 1998;Tait et al., 2005;Chace & Walsh, 2006;McKinney, 2006;Evans et al., 2009;Pickett et al., 2011;Aronson et al., 2014;Puga-Caballero et al., 2014;Beninde et al., 2015;Dallimer et al., 2015).However, cities are not homogeneous environments, but rather have zoning according to the type of activity in or usage given to certain areas (parks, industrial zone, residential zone).Thus, in urban areas, bird species distribution is both related to the local habitat features (tree and shrub cover, density of houses and other buildings) and the degree of urbanization of the city (Rolando et al., 1997;Evans et al., 2009;Pickett et al., 2011;Fontana et al., 2011;Ortega-Álvarez & MacGregor-Fors, 2011;Aronson et al., 2014).
The availability of green areas and the level of noise are two of the most important features affecting urban avian species assemblages (Chace & Walsh, 2006;Evans et al., 2009;Fontana et al., 2011;Toledo et al., 2012;Njoroge et al., 2013, Beninde et al., 2015;Sacco et al., 2015).High bird diversity in urban landscapes has been associated with high densities of trees and the presence of large green spaces connected or near to each other (i.e., not fragmented, but connected by corridors or acting as stepping stones) (Evans et al., 2009;Aronson et al., 2014, Beninde et al., 2015).High densities of human dwellings -and peopleand high levels of noise are associated with lower levels of bird diversity (Evans et al., 2009;Fontana et al., 2011), but higher bird abundances (Evans et al., 2009).This pattern of continuous decline of diversity and increase in abundance is exhibited along the rural-urban gradient (Blair, 1996;Chace & Walsh, 2006;McKinney, 2006;Puga-Caballero et al., 2014;Bino et al., 2008;Ortega-Álvarez & MacGregor-Fors, 2011), although diversity could peak at intermediate levels of disturbance, as in peri-urban areas (Blair, 1996(Blair, , 2004;;Tratalos et al., 2007).
Birds' responses to living in urban centers and the effects of disturbance in these areas have been studied for decades in the northern hemisphere (e.g., Marzluff et al., 2001;Chace & Walsh, 2006;Evans et al., 2009;Pickett et al., 2011;Davis et al., 2012;Taylor et al., 2013;Aronson et al., 2014;Sol et al., 2014;Beninde et al., 2015), but are a relatively new research focus in South America (e.g., Fontana et al., 2011;Ortega-Álvarez & MacGregor-Fors, 2011;Toledo et al., 2012;Njoroge et al., 2013;Puga-Caballero et al., 2014;Leveau et al., 2015;Sacco et al., 2015).Although urban species assemblages in South and North America show similar patterns, still there are several gaps to be filled (e.g., demographic patterns, physiological responses, behavioral ecology, biotic homogenization; Ortega-Álvarez & MacGregor-Fors, 2011).Our goal was to evaluate how urbanization affects the bird species assemblage (species richness and abundance) of a mediumsized city in southern Brazil.We use features of the urban landscape to test the predictions that (1) the degree of urbanization affects bird species distribution, and intensely urbanized areas have lower species richness and higher species abundance than less urbanized areas; (2) both species richness and abundance diminished in proportion to noise level; (3) vegetation is an important component of the urban landscape for birds, and bird diversity will increase in proportion to the area of green space in the city (parks, gardens, orchards).Köppen, 1918), with a hot and humid summer.In the Bioma Pampa, which spreads over 63% of Rio Grande do Sul political territory, the municipality territory is classified as region of ecological tension, where grass, shrub and wetlands are predominant in surrounding areas of the urban zone.Currently Canoas has no rural areas and the population of 323,827 inhabitants is settled in an area of 131.1 km² (demography of 2,470.13inhabitants/km²; IBGE, 2010).The city grew in a disordered way, scattered in patches of neighborhoods and villages that were settled in marshy and flooded areas.Industry had large impact on the city demography, as well as the local economy (Mayer, 2009).Despite the level of urbanization, Canoas has 16.2 m² of green areas per capita, making a total of 5.49 km² in the city (Estado da Cidade, 2014).

Study
Sampling design.We randomly selected, 120 sites based on 60 maps of the municipality of Canoas (www.geo.canoas.rs.gov.br).Each map covers a 1.1 km² area, divided in 20 quadrants of 0.4 km².We randomly selected two quadrants on each map and established one sample unit in the center of each quadrant.Sample units were required to be a public area (i.e., street, sidewalk, square) and were moved to the public location nearest the chosen center if necessary.Sample units were at least 200 m apart to guarantee independence between sampling units (Ralph et al., 1993).We used this sampling design to facilitate a more homogeneous evaluation of the study area and to avoid a concentration of points in few regions of the municipality.We could not access two sites located on private properties, and so removed them because we did not find any accessible location nearby.Thus, we sampled a total of 118 sites (Fig. 1).
In each sample unit (50 m radius from a central point), we measured the following variables related to the degree of urbanization (descriptions of each variable are in Tab.I): 1) Noise; 2) density of 'Trees'; 3) density of 'People'; 4) density of 'Pets'.We measured the percentage of 'Vegetation cover' and 'Grass' (open areas: grassland, gardens; Tab.I) -using satellite images provided by Google Pro (for Canoas, the images with the best available resolution were dated from January/2009 to December/2013).We also measured the abundance of buildings in each sample (number of 'Houses', 'Buildings', 'Pavilions', and 'Other structures'; all but houses later transformed into categorical variables; Tab.I).
We carried out bird surveys in September 2013, at the beginning of the breeding season.We conducted 10-min point-count surveys using a 50m fixed radius (Ralph et al., 1993), starting at dawn and lasting for 4 h (until 10:00 AM).We recorded all birds seen or heard, except birds flying above 20 m over the area, which we ignored in order to avoid double-counting during the census (annotated as an occasional record to compose the list of birds of Canoas).Data analysis.We first constructed three matrices: 1) species abundance (number of individuals per pointcount); 2) variables indicating the urban gradient (eight non-collinear variables); and 3) a spatial matrix with the geographical coordinates of each point-count (latitude and longitude).To avoid multi-collinearity, we selected only variables with Spearman correlation index below |0.6|: we excluded 'Traffic' and 'Pets' from the analyses and instead used the correlated variables 'Houses' (rho = 0.63) and 'Noise' (rho = 0.75).After the first investigation of data we eliminated two samples.These samples were outliers because they were located in an open, not urbanized area, where the major source of disturbance was traffic noise from the highway BR-386.
We tested the spatial correlation between species distribution and urban variables using a Mantel test (Legendre & Legendre, 1998), performed with R (R Development Core Team, 2015) using vegan package (Oksanen et al., 2015).The Spearman rho was used as the correlation coefficient.We used 9999 iterations with permutations of the matrix elements to calculate the P value for the test statistic, assuming no correlation between matrices as null hypothesis.The species distribution was spatially correlated (P = 0.01), so we performed a partial Mantel test to evaluate the correlation between species abundance and urban variables, weighting the spatial correlation.
To analyze the relationships between bird species assemblage and the urban variables, we performed a canonical correspondence analysis (CCA) (Legendre & Legendre, 1998) using on CANOCO v4.5 (Ter Braak & Šmilauer, 2002).Variables were centralized and standardized and the rare species were down-weighted to minimize their individual effects.To test the correlation between species abundance and urban variables, we used a Monte Carlo test with 9999 unrestricted permutations, assuming no correlation as the null hypothesis (Ter Braak & Šmilauer, 2002).
Finally, we used generalized linear models (GLM) to evaluate how urbanization affected bird species richness and abundance.To model species richness, we used the residuals of the linear regression of these variables as the response variable and also used Gaussian error and identity function.Because species richness is correlated with abundance (because more birds are detected as richness increased), we used the residuals as surrogate of species richness.To model the abundance (logarithmically transformed) we used the number of individuals in each point-count as the response variable, and used Gaussian error and identity function.We started by including eight variables as predictors in each model, and searched for the best subset using a backward stepwise procedure.We used second-order Akaike's Information Criterion (AICc) to select competing models, assuming that models with ∆AICc ≤ 2 explain the data equally well.We performed the GLMs with R (R Development Core Team, 2015), using the MuMIn package (Barton, 2014) to build and select models.
Bird species distribution were spatially correlated (Mantel test, p < 0.01), while urban variables were not (Mantel test, p = 0.98).Species assemblages, in turn, were not correlated with urban variables (partial Mantel test, p = 0.25).The canonical axes of CCA, despite significance (p < 0.01) and the average correlation between species and environmental variables (Pearson correlation, axis I = 0.75, axis II = 0.66), explained only 13% of variability in the data (axis I = 7.1%, axis II= 1.9%).Notwithstanding, CCA separated more-urbanized areas from those that were less urbanized, the latter having a larger proportion of open vegetation cover (grass and wetlands) (left to right on axis I) and wooded areas, with greater vegetation cover (distinguished from other variables on axis II) (Fig. 2).Synanthropic and/or exotic species [e.g.House sparrow Passer domesticus (Linnaeus, 1758), Rock dove (Columba livia Gmelin, 1789), Blue-and-white swallow Pygochelidon cyanoleuca (Vieillot, 1817)] were recorded more frequently in more urbanized sites, characterized by higher densities of people and houses; the presence of buildings, pavilions and other structures; and higher levels of noise.Sites with greater proportions of open vegetation, lower levels of noise, and an absence or lower frequency of urban structures (house, buildings) were dominated by wetland species [e.g., Whitebrowed meadowlark Sturnella superciliaris (Bonaparte, 1850), White-faced ibis Plegadis chihi (Vieillot, 1817) and Chestnut-capped blackbird Chrysomus ruficapillus (Vieillot, 1819)], and grass/shrubland species [e.g., White monjita Xolmis irupero (Vieillot, 1823), Shiny cowbird].Forested species, such as Variable antshrike (Thamnophilus caerulescens Vieillot, 1816) and Golden-crowned warble [Basileuterus culicivorus (Deppe, 1830)] were present and more abundant in sites with a greater abundance of vegetation cover (percentage cover and number of trees).
We selected four competing models for analysis of species richness (Tab.II).The most important variables, retained in all models, were abundance of houses, noise and percentage of vegetation cover.Abundance of people, presence of buildings and other structures were the other variables selected in our GLMs (each only in one model).Species richness was positively related to the proportion of arboreal vegetation ('Trees'), abundance of people and presence of buildings (Tab.II), while areas with higher noise  levels, abundance of houses and presence of other structures have low species richness.We selected six models for analysis of species abundance (Tab.III).Species abundance decreased as noise levels increased (noise level was selected in all models), and also decreased with the abundance of houses and presence of buildings.The proportion of vegetation cover (both open and arboreal) in a site had a positive effect on bird abundance (as we can see in Fig. 1), as did the presence of pavilions.
Although some effects of urbanization on the bird species assemblage were not so clear, the regression models nonetheless showed well known effects of urbanization.The proportion of arboreal vegetation cover was the most important variable predicting an increase in species richness in Canoas, corroborating the known role of green areas as biodiversity enhancers in urban centers (Chace & Walsh, 2006;Evans et al., 2009;Fontana et al., 2011;Ortega-Álvarez & MacGregor-Fors, 2011;Toledo et al., 2012;Njoroge et al., 2013;Aronson et al., 2014;Beninde et al., 2015;Sacco et al., 2015).Open landscapes prevail in the Canoas region, and the arboreal component is characterized by gardens, squares and parks in the urban area; this contrasts with cities with large forested areas on the borders, for instance around Porto Alegre (Fontana et al., 2011).As observed in other studies, plots of open vegetation (squares, gardens, golf courses) usually have higher biodiversity (Evans et al., 2009;Aronson et al., 2014;Beninde et al., 2015).Therefore, these areas are home to particular bird species assemblages that are distinct from rural and peri-urban avifauna.
On the other hand, the negative effects of urbanization upon bird diversity are clearly showed in the relationships between species richness and the level of noise and the density of houses.Noise is a striking feature of urban centers, usually related to low species richness and abundance (Fontana et al., 2011;Pickett et al., 2011;Sacco et al., 2015).Together with other human activities, noise may constrain species' ability settle in urban centers, which leads to loss of diversity and differentiation in bird species assemblages (Francis et al., 2009;Fontana et al., 2011;Rolando et al., 1997;Slabbekoorn & Peet, 2003;Bisson et al., 2011;Pickett et al., 2011;Chávez-Zichinelli et al., 2013).
In addition to noise, the density of houses and presence of buildings affected both species richness and abundance in our study.Although both variables are related to urbanization, they are not good indicators of urbanization, as their effects are dependent on the scale analyzed and the particular process of urbanization for each city (Evans et al., 2009).In Canoas, the density of houses appears to indicate densely urbanized areas, and adds to other urban features (noise, buildings) in negatively affecting the avifauna.Although negatively related to bird abundance, presence of buildings was positively related to species richness, probably due to the occurrence of aerial foragers (e.g., Hirundinidae) and species that nest or live in rocky habitats (e.g., Apodidade) (Blair, 1996(Blair, , 2004;;Aronson et al., 2014).
We found a high abundance of urban exploiters, like House sparrows, Rock doves and Eared doves [Zenaida auriculata (Des Murs, 1847)], in more urbanized areas, in agreement with the pattern found in other studies (Blair, 2004;Chace &Walsh, 2006;Bino et al., 2008;Fontana et al., 2011;Ortega-Álvarez & MacGregor-Fors, 2011;Aronson et al., 2014;Puga-Cabalero et al., 2014).However, his pattern of abundant urban exploiters does not reflect a positive overall effect of urbanization on bird abundance.In fact, pavilions were the only urban feature positively related to abundance, probably because they offer local resources for a few common species.On the other hand, peri-urban areas of the city, which were used for agriculture in past decades, showed high abundance of a few species (e.g., Bare-and White-faced ibises, Shiny cowbird).This concentration of birds in peri-urban areas (Fig. 1) masks the expected pattern seen along a rural-urban gradient: increase in abundance from rural areas to urban centers (Blair, 1996;Chace & Walsh, 2006;McKinney, 2006;Bino et al., 2008;Ortega-Álvarez & MacGregor-Fors, 2011;Puga-Caballero et al., 2014).
The city of Canoas does not have a clear pattern of urbanization, reflecting the disorganized growth of the city (Mayer, 2009).Despite the existence of a core area, with buildings and commercial areas, there are still plots of Iheringia, Série Zoologia, 108: e2018017 habitats patchily distributed inside the city, buffering bird species distribution from urban effects.The presence of wetlands, grasslands (vacant lots, lawns and squares) and woodlots (urban parks, gardens) offers habitats to species less adapted to the urban environment (urban avoiders), reducing the biotic homogenization and maintaining part of the pre-urbanization pool of species.We agree with Fontana et al. (2011), that the level of noise seems to be the best variable indicating the degree of urbanization.Along with vegetation cover or related parameters (density of trees, presence of green areas) (Chace & Walsh, 2006;Evans et al., 2009;Aronson et al., 2014;Beninde et al., 2015), the level of noise would be the best variable to use in order to recognize levels of urbanization.
Finally, given the increasing concern with sustainable urban development, seeking environment-friendly urban growth that preserves cities' biodiversity (Pickett et al., 2011;Aronson et al., 2014;Beninde et al., 2015), urban planners need to take into account how the city works ecologically (Pickett et al., 2011), for example, how the biota responds to an urban environment.This is essential for conservation and management purposes.The effects of the level of noise and presence of green areas on urban-dwelling birds, for instance, are well known (e.g., Chace & Walsh, 2006;Evans et al., 2009;Aronson et al., 2014;Beninde et al., 2015), and should be considered in urban planning policies.Supporting citizens in maintaining residential vegetation (e.g., private yards), and, hence, keeping areas of native vegetation inside the urban area, is a simple example of how to increase a city's green areas and promote biological conservation (Smith et al., 2014).
Fig. 1.Bird species richness and overall abundance recorded in point counts (surveyed on September 2013) in the municipality of Canoas, Rio Grande do Sul, Brazil.

Fig. 2 .
Fig. 2. Ordination diagram presenting the first two axes of the Canonical Correspondence Analysis (CCA) (percent of explained variability: axis I = 7.1%, axis II = 1.9%) based on the distribution of species abundance in 118 sample units (dots) in the urban area of Canoas, Rio Grande do Sul, Brazil, and its correlation with seven explanatory variables (arrows).The first axis shows the urbanization gradient (negatives values on left = more urbanized regions; positive values on right = less urbanized regions).All axes were significant (Monte Carlo test with 9,999 permutations: P < 0.001).Species names are given in full in Appendix 1. Variables are described in Tab.I.
area.Canoas (29°55'12"S, 51°10'48"W) is part of the metropolitan area of Porto Alegre (the capital and largest city of the state of Rio Grande do Sul, Brazil) known as Greater Porto Alegre.Built in the Depressão Central region (Central valley) on the Guaiba river basin, Canoas is bordered by the dos Sinos and Gravataí rivers, and is in the transition zone between Planície Costeira (Coastal Plain) and the Planalto Meridional (Meridional Plateau) (www.canoas.rs.gov.br).Climate is temperate (Cfa; ). Competing models (∆AICc < 2) for the influence of environmental variables in bird species abundances.Variables with negative coefficients are indicated by a minus sign within the brackets.Variables retained in the best candidate model are in bold (df, degrees of freedom; AICc, corrected Akaike's Information Criterion; ∆AICc, difference on AICc between the current model and the best model; w, Akaike weights).