The effect of habitat changes along the urbanization gradient for breeding birds: an example from the Xiong’an New Area

Study sites

The total size of the Xiong’an New Area is 1,566 km² (38°43′–39°10′N, 115°38′–116°20′E). The elevation of our study area ranges from 7 m to 19 m. Historically, this area was part of the Central Plains region of ancient China, and the landscape has experienced intensive agricultural exploitation for more than 2,000 years. In addition to land being used for construction, the other main land-use types are agricultural, with the main crop being wheat (Triticum aestivum L.), and rows of poplar (Populus X canadensis Moench) planted at the boundary of some farmland that act as shelterbelts. The total vegetation coverage is 63.75% (Xu et al., 2017), and Baiyangdian, the largest lake within Hebei Province, is located between AX and XX (Fig. 1). A recent survey showed that there were 72 bird species (among which 51 were forest birds and 21 were water birds) that occurred in the summer and winter seasons in this area (Zhou et al., 2018). The area has a warm temperate continental monsoon climate with distinctive seasons and an average temperature of 12.1 °C. By the end of 2015, there were 1.13 million permanent residents within the area, and the urbanization rate was 42.74% (Baoding Statistical Bureau, 2016).

Field surveys

Based on the relative spatial position and configuration of the three constructed areas of the Xiong’an New Area (Fig. 1), we first determined the geometric center points of the three cities and connected each pair of the three points with three straight lines. Sample points were then selected along the lines originating from the three city centers, and representative sample points were selected every 1–4 km outward from the three centers (Bibby, Jones & Marsden, 1998) that is, 1 km for urban and urban fringe points to ensure the independency of each field count, 1 or 4 km for rural points, we didn’t chose rural points by 1 km separation for all of the survey because of the unanimous configuration of the crisscross rural fields. Within the three constructed areas, densely constructed buildings and roads dominate the landscape, and each area is surrounded by an obvious edge that extends between the constructed area and agricultural land. Given these landscape patterns, we classified our sampled points into a specific urbanization gradient based on landscape construction and spatial position; that is, points that were surrounded by more than 50% of artificial land cover within a 200 m radius buffer region and that were located inside the city edge were defined as the urban points, points that were surrounded by 20–50% artificial land cover and located within 100 meter from the city edge were urban fringe points, and points with artificial land cover totaling less than 20% and agriculture land cover totaling more than 50% within the 200 m buffer region were classified as rural points (Verma & Das Murmu, 2015). To ensure spatial independence, the minimum distance between each pair of sample points was 1 km (Legendre et al., 2002). A total of 30 sample points were selected (Fig. 1 and Table 1).

Bird survey and anthropogenic disturbance records

The fixed-point count method (Bibby et al., 2000) was adopted for our bird survey in the breeding season (April 2017 and June 2018), and every point was sampled at least two times. The field survey was conducted only on days with good weather (no rain or strong winds with speeds of more than 30 km/h), from 6:30∼10:30 and 16:00∼18:00 on each day when birds were active and thus easy to detect. Bird census began at the time the surveyor arrived at the center of the sample point, and bird species and individual birds heard or seen within a 50 m radius were recorded for 5 min. Individuals flying through the sample points that did not stop were excluded from our count. As a typical representation of anthropogenic disturbance, the number of pedestrians within the bird census circles during the bird count period was recorded, and environmental noise was measured just after the bird census finished. Environmental noise level was measured with a high precision noise meter (SW-524) in four directions, 25 m from the center of each sample point. The highest and lowest values within 20 s were recorded for each measurement, and the average value of the census within each sample point was used as its environmental noise level. To reduce the survey error among different investigators, all of the surveys of the birds in this study were completed by one investigator (the first author of this paper, with six years of experience of bird watching and field surveys). The field survey was approved by the Research Council of Research Center for Eco-environmental Science, Chinese Academy of Sciences.

Vegetation survey

Vegetation surveys were conducted at the same time as the bird count, and plant species (including arbor, shrub and herb) and the individual number of trees (more than 1-meter height) were recorded within the 50 m radius circle. Multiple layers of coverage were evaluated based on visual estimation by an experienced plant surveyor. The vertical structure of the vegetation was measured as the foliage height diversity (FHD). From the center of each sample point, four observing line (50 m long) which are perpendicular to each other were decided with a north arrow, observation points were chosen at 10-m intervals along each line. The presence of leaves at different vertical height levels (0–1, 1–2, 2–5, 5–10, 10–20, and 20–30 m) was observed and recorded at each point, using a 5-m-long pole as visual reference. For details, please see Lee & Rotenberry (2015).

Data analysis

Avian species richness (Spe) was measured as the number of species recorded at each sample point. Tree species diversity and FHD were calculated using the Shannon-Wiener index, and the formula is H′ = ∑PilogPi, P_i is the ratio of the individual number of species I and total individual number. The individual density (Den) of each urbanization gradient was calculated using $\frac{\mathrm{Den}=\sum {\mathrm{N}}_k}{\pi {\mathrm{r}}^2}$ , where N is the individual number of sample point k, and r is the radius of the sample point (50 m).

Based on resident types, all birds recorded in the field survey were classified as residents, passengers, summer visitors or winter visitors (Zheng, 2017). Based on feeding types, all birds were defined as insectivores (I), granivores (G), insectivore-frugivores (IF), omnivores (O) or carnivores (C) (Kwok & Corlett, 2000). The three city areas of the Xiong’an New Area were spatially independent, and the minimum distance between each pair of cities exceeded 10 km, separated by agricultural land and villages. A comparative analysis for avian communities between independently constructed areas could provide critical information for the master plan of the newly established special zone. To compare avian community structure between cities and urbanization gradients, we defined bird species as exclusive species when they only appeared in a single city.

To compare the differences between bird guilds of each urbanization gradient and constructed area, one-way ANOVA was used to check whether the differences between avian species richness and individual abundance within each gradient or city had reached significant levels. To further understand overall bird species diversity within the study area and to describe the ecological driver of current species distributions, beta-diversity of avian community between city pairs was calculated as the multisite Sorensen dissimilarity index (β_SOR). To understand the sources of species compositional differences, β_SOR was divided into two additive components, β_NES(nestedness) and β_SIM(turnover) (Baselga, 2010), and the ratio between them (β_NES/β_SIM) represented the relative contribution of the two components (Quan et al., 2018). Considering only the overall diversity and ignoring its respective components may lead to deviations in the formulation of conservation strategies (Angeler, 2013). Thus, pairwise dissimilarity and the corresponding additive components (βsor, βsim, βnes, and βsim/βnes) were also calculated. In this study, only breeding bird species were utilized in beta-diversity calculation and the subsequent avian-environment correlation analysis, which could help avoid errors resulting from the occasional occurrence of passenger birds.

Unfortunately, only 25% of the total sample points contained shrub layers, and thus, shrub indexes were excluded from further analysis. In our further analysis, pedestrian number, environment noise, tree species richness, diversity and coverage of trees, and FHD are predictors, which have all been demonstrated to be important for urban birds in other studies, and avian species richness and individual abundance are the response variable. To ensure the normality of the data, all variables were log-transformed and conformed to a normal distribution. Multicollinearity impairs the reliability of regression analysis (Mac Nally, 2000), and to resolve this problem, we first utilized Spearman’s correlation to determine the environmental factors that showed significant correlation with avian community indexes. Then, we used principal component analysis (PCA) to extract the principle components, and further, stepwise regression was conducted, with the principal component scores being the predictors and avian species richness and individual abundance are the response variables. Beta-diversity was calculated using the “betapart” package of R 3.5.1 (Baselga & Cdl, 2012), and other analyses were conducted using SPSS 22.0.

Avian community characteristics recorded in the field survey

A total of 37 bird species (from nine orders and 26 families) were recorded in the field survey. For resident types, residents and passengers accounted for the majority of these species and accounted for the same proportion (40.54%); breeding birds (including residents and summer visitors) accounted for 78.38% of the total species number. In terms of feeding type, insectivores accounted for the largest proportion of the birds (48.65%), followed by insectivore-frugivores (18.92%) (Fig. 2). Three national class II protected bird species (all raptors) were recorded. In this study, taxonomy and residence type were based on A Checklist on the Classification and Distribution of the Birds of China (Third Edition) (Zheng, 2017). The species list and the supplemental information are provided in Appendix S1.

Community characteristics within different urbanization gradients

Significant differences between species richness (F = 13.42, P < 0.001) and individual abundance (F = 10.94, P < 0.001) for the three urbanization gradients were detected by one-way ANOVA (Figs. 3 and 4). A total of 22 bird species were recorded in the rural area, three of which were exclusive species (including two breeding birds); 34 bird species were detected within the urban fringe, with eight exclusive species (including five breeding birds); only 18 bird species were found within the urban area, with two exclusive species (only one breeding bird). Nine occurred in all urbanization gradients, and all nine were breeding birds; 15 species were shared by two of the three gradients, of which 12 were breeding birds; and 13 species existed within only one gradient, of which eight were breeding birds (Table 2). Different guilds of bird species showed similar distribution patterns in the urbanization gradient (urban fringes contained all peaks), but the pattern was more obvious for species richness of residents, passengers, insectivores (Fig. 5), and omnivore individual density (Fig. 6).

Avian community characteristics of the three constructed areas

Statistical analysis results showed that 13 of the 37 bird species (37.14%) were exclusive species of one specific city; six species (17.14%) were shared by two cities, and only 16 species (45.71%) were shared by all cities. The results for breeding bird species are as follows: eight species (27.59%) were specific to one city, five species (17.24%) were specific to two cities, and 16 species (55.17%) were shared by all species. The difference among the breeding bird species richness of the three cities was not significant (F = 0.623, P > 0.05), the value of multisite beta-diversity (β_SOR) was 0.260, and species spatial turnover processes seemed to be more important (β_NES/β_SIM = 0.634). Table 3 shows that of the sites, pairwise beta-diversity between RC and XX was completely decided by spatial turnover (DISnes/sim = 0.000) and had the lowest dissimilarity value (β_sor = 0.167). Dissimilarity between RC and AX was mainly decided by nestedness (DISnes/sim = 2.429), while dissimilarity between XX and AX was decided by turnover and nestedness simultaneously (DISnes/sim = 1.143).

Avian-environment relationships along the urbanization gradient

Using Spearman correlation analysis, we obtained five environmental indexes that were significantly correlated with both breeding bird species richness and individual abundance. Substantial multiple collinearity existed among them (Table 4), and thus, PCA was conducted to extract the principle components. The ratio of the sample size (30) to the number of independent variables (five) analyzed was 6:1, which was fit for PCA; the Kaiser-Meyer-Olkin (KMO) value was 0.652, passing the Bartlett sphericity test (P < 0.001) (Zhang & Dong, 2004). Two principle components with eigenvalues >1 were extracted by PCA, PC 1 and PC 2, which contained 86.34% of the information of all the environmental variables to be separated. Based on the matrix in Table 5, we suggested that PC 1 is largely the reflection of tree species diversity and PC 2 represent for the effect of FHD. The expression of PC 1 is y₁ = 0.269*Pedestrians + 0.522 ∗H′_tree + 0.386 ∗S_tree − 0.071 ∗C_tree − 0.238*FHD; and for PC 2 it is y₂ = 0.071*Pedestrians − 0.274 ∗H′_tree − .042 ∗S_tree + 0.485 ∗C_tree + 0.640*FHD. Principle component regression produced two models, that is, Y₁ = 1.543 + 0.206 ∗y₁ + 0.347 ∗y₂; Y₂ = 3.025 + .0321 ∗y₁ + 0.421 ∗y₂. They described 60.9% and 54.3% of species richness and individual abundance, respectively (Table 6). After the transformation of coefficients, we got the final models, for breeding bird species richness it is ${\hat{Y}}_1=1.543+0.080\text{*Pedestrians}+0.012\mathrm{\ast }{\mathrm{H}}^{\prime }\text{\_tree}+0.065\mathrm{\ast }\mathrm{S}\text{\_tree}+0.154\mathrm{\ast }\mathrm{C}\text{\_tree}+0.173\text{*FHD}$; for breeding bird individual abundance it is ${\hat{Y}}_2=3.025+0.116\text{*Pedestrians}+0.052\mathrm{\ast }{\mathrm{H}}^{\prime }\text{\_tree}+0.106\mathrm{\ast }\mathrm{S}\text{\_tree}+0.181\mathrm{\ast }\mathrm{C}\text{\_tree}+0.193\text{*FHD}$. It should be noticed that all the variables in the above models have been log transformed.