Avian community composition, but not richness, differs between urban and exurban parks

Urban development alters landscape structure and available resources, potentially threatening avian diversity worldwide. However, it is unclear how bird communities respond in areas currently undergoing urban development, particularly in the non-breeding season. We examined avian communities at 8 parks in urban (within established urban matrix; >50% built cover) and 9 parks in exurban (within adjacent grassland; 5–20% built cover) areas in Sacramento County, CA. We measured bird species and abundance, conducting five line-transect surveys per park (85 total). We investigated factors influencing avian assemblage, including local habitat features (land cover, number of trees, fruiting trees and tree species, tree height and diameter at breast height, park size, park age), and landscape features (land cover within 500 m and distance to riparian habitat). Fifty bird species, including 15 migrants, were observed. Total species richness and abundance at urban and exurban parks was not significantly different, but community assemblages differed significantly. Park area positively predicted species richness. Abundance was negatively associated with inpark percent built cover, average number of fruiting trees and landscape-scale percent water cover. Species composition changes were associated with distance to riparian habitat and landscape percent grass cover for all but one exurban park; and with tree height, DBH and park and landscape percent tree cover for urban parks. The expansion of exurban areas in many parts of the world poses a significant risk for natural habitat loss. Parks in such areas should be planned to harbor some of the displaced biodiversity.


Introduction
In densely populated areas, urbanization is a primary cause of species decline (Czech, Krausman, and Devers 2000;McKinney 2002). Urban development fragments, isolates and degrades natural habitats (Alberti 2005;Chamberlain et al. 2009) while simplifying and homogenizing species composition (Lockwood, Brooks, and Mckinney 2000;Aronson et al. 2014). Over the last 50 years, the USA has experienced a 29% decrease in bird abundance across species (Rosenberg et al. 2019). About one-fifth of all bird species occur within cities (Aronson et al. 2014); however, species vary in their response to urbanization (Rebolo-Ifrá n et al. 2015;Isaksson 2018;Callaghan et al. 2020). Most are unable to occupy new habitats created by urban development (Rebolo-Ifrá n et al. 2015) and the species that disappear quickly after urbanization can be identified with relative ease (Isaksson 2018). Others persist or even increase in density (Rebolo-Ifrá n et al. 2015), often responding positively to increased habitat features for perching and nesting (Emlen 1974;Lancaster and Rees 1979;Isaksson 2018). More difficult to detect are gradual changes in populations. Some species are attracted to cities by available resources and warmer temperatures, but their populations decline slowly due to dangers such as high nest predation rates and incidental collisions (Isaksson 2018).
While not all species can be supported within cities (Shochat et al. 2010), habitat heterogeneity may allow for greater coexistence between prolific urban species and rarer native species (Shochat et al. 2004). The complex mosaic of disturbance and land transformation caused by urban development (Blair 2004;Pautasso 2007) can generate habitat heterogeneity (McKinney 2008). Whether heterogeneity within and between parks can help increase species richness is not yet clear, and questions remain as to the role of green spaces in supporting biodiversity (Lepczyk et al. 2017). However, the argument that parks act as refuges for bird species is well-supported (Ferná ndez-Juricic and Jokimä ki 2001). Remnant vegetation within urban areas, especially native cover, facilitates persistence of native species and higher diversity, even at small scales such as along streetscapes (White et al. 2005). Parks can also increase habitat connectivity, acting as stepping-stones between habitats (Ferná ndez-Juricic and Jokimä ki 2001). Parks commonly have higher bird species diversity and richness than other urban areas (Jokimä ki 1999; Nielsen et al. 2014), highlighting their potentially integral role in maintaining (Savard, Clergeau, and Mennechez 2000) and promoting (Nielsen et al. 2014) biodiversity in urban landscapes.
Avian biodiversity in urban parks has been well-studied within cities and city edges (Jokimä ki 1999;Ferná ndez-Juricic 2000;Clergeau, Jokimä ki, and Savard 2001;Murgui 2007) and during the breeding season (Blair 1996(Blair , 2004Marzluff, Bowman, and Donnelly 2001;Crooks, Suarez, and Bolger 2004). However, exurban areas [city-adjacent areas that have $5% to 20% built area and are surrounded by a natural matrix (Marzluff, Bowman, and Donnelly 2001)] and winter bird use of parks by fall migrant and overwintering species is not thoroughly understood (Smith 2007).
In the USA, exurban development has outpaced other development types, and takes up five times more space than urban and suburban development combined (Lenth, Knight, and Gilgert 2006). Despite this phenomenon, parks within exurban areas currently undergoing such development have been largely overlooked in the scientific literature. In a study comparing forest bird communities in forests versus exurban developments in Virginia, USA, Suarez-Rubio, Leimgruber, and Renner (2011) found similar species richness, but significantly different species composition; and a reduction of forest specialists in exurban areas. The study reinforced a differential impact of exurban development on a species-by-species basis, and the need for further investigation of such areas.
The majority of avian studies are conducted during the breeding season (or use data sets such as the American Breeding Bird Survey; e.g. Chocron, Flather, and Kadmon 2015), and the same is true for studies within parks (Blair 1996, Blair 2004Marzluff, Bowman, and Donnelly 2001;Crooks, Suarez, and Bolger 2004; but see Carbó -Ramírez and Zuria 2011 for their study of greenspaces in Mexico). Less is known about the habitats used by resident (overwintering) and migrant birds during the non-breeding season (Tryjanowski et al. 2015). In areas such as CA, USA, species assemblages shift throughout the year as migrants move along major migration routes, or make more local, seasonal movements. Understanding species richness and abundance during the non-breeding season, which represents the majority of the year and therefore the majority of the life cycle for most species, could provide a more complete picture of the value of urban greenspaces to avian communities.
The Central Valley of California, a 47 000 km 2 landscape dominated by intensive agriculture and growing cities, has a rapidly expanding human population (Department of Finance 2017). The associated intensification and expansion of agriculture and urban development have impacted wildlife habitats (Matchett and Fleskes 2017). For instance, over 90% of historic wetlands have been lost (Matchett and Fleskes 2017). Sacramento County alone experienced a net increase in over 202 km 2 of urban and built-up land between 1988 and 2016 (Department of Conservation 2016). The region contains critical overwintering habitat for many bird species (Matchett and Fleskes 2017) and a diversity of resident species, making this a model system for understanding how expanding urban areas may influence the composition of avian communities. Here, we present the results of a study to investigate species richness and abundance of birds in parks. We predicted that urban parks would have higher richness and abundance. In addition, we expected species richness to be correlated with available foraging substrates within the parks (e.g. trees and grass) and avian abundance would be correlated with a more urban-developed landscape surrounding the parks. To address these hypotheses for avian communities in the non-breeding season, we used avian counts, data on local (within-park), and data on landscape (park surroundings) habitat features.

Study sites
We conducted this study in Sacramento County, CA, USA. Historically, numerous native-American tribes lived along watercourses in the area, followed by an influx of European miners in the early 1800s (US Forest Service 2003). The four rivers that cross the county (Sacramento, American, Consumnes and Mokelumne) contributed to the development of agriculture and transportation (US Forest Service 2003). The county experienced major population growth in the 1950s, due to local aerospace industry, government services growth and military expansion (County of Sacramento 1996). By 1950, $245 km 2 of built landscape existed in the Sacramento region (Sacramento County and adjacent counties with associated urban sprawl) (Wheeler and Beebe 2011). By 2010, developed landscape had increased to 1978 km 2 . The growth pattern includes houses and commercial buildings historically aggregated near the river that expanded into the exurban grassland; this pattern of sprawl continues. Under current land-use patterns, it is estimated that 3691 km 2 of new development will be required to support an influx of 1.7 million additional residents by 2050 (Wheeler and Beebe 2011).
We collected field data at 17 parks in Sacramento County, CA, within the City of Rancho Cordova or adjacent censusdesignated areas (Fig. 1). The City of Rancho Cordova, incorporated in 2002, has a population of 70 087 (U.S. Census Bureau 2019). The elevation is $27 m, mean temperature is 16.22 C with an average annual precipitation of $500 mm (Arguez et al. 2012). The surrounding biome is the Mediterranean climate region of the Central Valley. Within the urban environment, parks are older, were built near the American River, and tend to have more mature trees and complex vegetation. The natural habitats surrounding the urban core of Rancho Cordova are primarily annual and vernal pool grasslands. An $15.5 km 2 area within this grassland is currently in the process of being developed for single family houses and encompasses in some of the newest parks in this study. Parks in exurban areas undergoing residential development tend to be more recently built, and have fewer, smaller trees and less overall vegetation, and are located further from the river. The locations of the urban and exurban parks share many similarities due to proximity that helped control for spatial aggregation between the two groups, including similar elevation (0-76 m), temperature and rainfall. The youngest exurban park was built in 2016, 1 year before the study took place, and construction of housing developments in the study area is ongoing as of 2020. Because development is current, five of the parks in this study were <10 years old, allowing us to study bird communities after recent disturbance. The remaining undeveloped exurban area adjacent to Rancho Cordova is habitat for several bird species of conservation interest, including Swainson's hawk (Buteo swainsoni, California-Threatened and US Fish and Wildlife Service bird of conservation concern [BCC]), tricolored blackbird (Agelaius tricolor, California Threatened, a California species of special concern [SSC], and USFWS BCC), short-eared owl (Asio flammeus, SSC), burrowing owl (Athene cunicularia, SSC) and loggerhead shrike (Lanius ludovicianus, SSC and BCC) (CDFW 2019).

Species surveys
We performed five surveys at each park (e.g. Clergeau, Jokimä ki, and Savard 2001;Crooks, Suarez, and Bolger 2004) for a total of 85 surveys. Surveys were conducted once or twice weekly from September to November 2017, between sunrise and 11:00 A.M. to coincide with peak bird activity (Ferná ndez-Juricic 2000). We randomized the order in which parks were visited. Surveys were not conducted during rain or strong winds (Ralph, Sauer, and Droege 1995) or when parks were being irrigated or mowed.
Single-year and single-visit surveys have been deemed appropriate for predicting bird abundances and species richness in urban spaces ( Jokimä ki and Suhonen 1998 ; Jokimä ki and Kaisanlahti-Jokimä ki 2003).
Each survey consisted of a 50 m line-transect (e.g. White et al. 2005) that was walked for 5 min. All birds seen or heard within 50 m of the transect line, and any birds seen or heard beyond the 50 m truncation (Ralph, Sauer, and Droege 1995) were identified to species using binoculars (Monarch 5 8 Â 42). All transects started midway along one edge of the park (the edge chosen for each park was determined through random number generation) and a single observer (ARH) walked into the park perpendicular to that edge. All five surveys in each park were repeated along the same transect. Care was taken to avoid counting the same individual twice on each transect. Birds that flew over but not within the habitat were labeled as flyovers. Behavior was opportunistically recorded. For each transect, elevation and geographic coordinates were collected (UTM's; Garmin GPSMAP 64).

Habitat measurements
Parks were labeled as urban (located in the established urban matrix, >25 years of age) or exurban (recently built within the exurban matrix of grassland habitat, <20 years of age). In studying avian communities, the proportions of vegetated open space versus human made structures (e.g. percentage of built area) have been used as surrogates for urbanization (Blair 1996;Clergeau et al. 1998;Lee et al. 2004). Area covered by pavement and area covered by buildings have been shown to be redundant (Blair 1996) and can be combined into a single measure of 'built environment'. According to Stein, Gerstner, and Kreft (2014), plant diversity and land cover are probably suitable for capturing the habitat requirements of many species.
We collected habitat measurements within 50 m on either side of the transect within the park, and at the broader landscape scale of 500 m from the park (e.g. Murgui 2007;Smith, Francis, and Fahrig 2014) since this is an appropriate scale for meaningfully measuring landscape heterogeneity (Litteral and Shochat 2017).
Park-scale, 'local' habitat variables.. In each park, we measured a number of ecologically relevant variables shown to predict bird richness and or abundance in previous studies (Blair 1996;Bolger, Scott, and Rotenberry 1997;Jokimä ki 1999 To capture vegetation structure heterogeneity we measured park percent land cover, tree DBH and tree height. We used aerial photographs and Google Earth Pro software to calculate the park area and percent land covers within each park. To gauge anthropogenic disturbance, we recorded the number of pedestrians that entered the survey area per minute, as well as number of cats and dogs because these variables can affect avian species richness (Clergeau, Jokimä ki, and Savard. 2001). These data were collected within the survey timeframe (i.e. sunrise to 11:00 A.M.) to coincide with maximum bird activity. Park age was determined by land acquisition dates as provided by the Cordova Recreation and Park District. For a full list of variables measured at the local habitat scale, see Table 1.
Landscape-scale habitat variables.. We used Google Earth Pro to measure the distance from each transect to the American River, the large river that runs through Rancho Cordova, to account for the fact that many of the urban parks are clustered closer to the river than exurban parks, and riparian habitat could introduce different bird species (Sabo et al. 2005) to adjacent parks. We used iTree Canopy (Version 6.1, United States Forest Service 2006), a peer-reviewed software program (Jacobs, Mikhailovich, and Delaney 2014), to generate cover percentages for five relevant habitat characteristics within a 500 m radius of each park (Table 1). We used 400 randomly generated points for our characterization in iTree Canopy, and all standard errors were <3.0%. For a full list of variables measured at the landscape habitat scale, see Table 1.

Statistical analysis
Species richness and abundance Analyses were carried out on bird species richness (cumulative number of species seen during five surveys) and on avian abundance (the average number of individuals of each species seen in each park during five surveys, summed across all species observed in the park) (e.g. Crooks, Suarez, and Bolger 2004;White et al. 2005;Biadun and Zmihorski 2011). All avian species were assigned resident or migrant status based on their winter ranges. Flyovers were excluded from analyses (Jokimä ki 1999) as were unidentified species.
Because species richness and park area were highly correlated, we used Analysis of Covariance (ANCOVA) to test if species richness varied between urban and exurban parks, controlling for the effect of park size. We used a two-sample ttest to investigate whether maximum abundance differed between the park groups. We used multiple linear regression to investigate what local and landscape-scale habitat features were associated with species richness and abundance.
Dependent variables were tested for normality using the Shapiro-Wilk test (e.g. Husté and Boulinier 2011). We used frequency histograms and quantile-quantile plots to assess whether the data fit the assumptions of normality and homoscedasticity of residuals. Abundance and park area were logtransformed to meet these assumptions (e.g. Jokimä ki 1999; Ferná ndez-Juricic 2000). We examined the Cook's distance to check for outliers. Hagan Community Park has a larger area than all parks and had the highest avian species richness. However, no outliers were found (all Cook's distances were <3) so Hagan was included in all analyses.
To assess the assumptions of ANCOVA, we checked for a linear relationship between species richness and area, the covariate. We checked for homogeneity of regression slopes by verifying that the interaction term was not statistically significant. We used a Shapiro-Wilk test to assess the normality of residuals. We checked for homogeneity of the residual variances using the Levene's Test.
To meet the assumptions of multiple linear regression, and avoid issues of multicollinearity, we used Pearson's coefficient to test for collinearity amongst variables. Several variables were correlated using Pearson's coefficient (see Table 1). In each case, the variable that seemed more biologically relevant and/or had a higher correlation with the dependent variables was selected (Table 1). We then formally checked the remaining independent variables for multicollinearity, and any with variance inflation factors (VIFs) >3 were sequentially removed starting with the covariate with the highest VIF. VIFs were recalculated, and this process was repeated until all VIFs were smaller than 3, a stringent pre-selected threshold (Zuur, Ieno, and Elphick 2010) (Table 1). We included a total of eight variables in our models.
We ran stepwise multiple regression (e.g. Pino et al. 2000;Crooks, Suarez, and Bolger 2004; Carbó -Ramírez and Zuria 2011) using as predictor variables: log park area, park percent built cover, park percent water cover, landscape percent built cover, landscape percent water cover, number of fruiting trees, average tree DBH and average number of pedestrians at parks. Stepwise regression accounts for the contribution each of the independent variables makes to explaining total variance, and thus demonstrates the relative importance of these variables in the regression (Pino et al. 2000). We used a non-automatic variable search (e.g. Carbó -Ramírez and Zuria 2011) using backwards elimination techniques and the Akaike information criterion (AIC) to select the most informative model (e.g. Biadun and Zmihorski 2011; Supplementary Appendix A). The model residuals were examined for deviations from assumptions and none were observed. All analyses were run in R version 3.4.3 (R Core Team 2017), packages vegan (Oksanen et al. 2018), car (Fox and Weisberg 2019) and broom (Robinson 2014).

Community composition
We conducted nonmetric multidimensional scaling (NMDS) to examine differences in avian composition between parks (e.g. Leveau and Leveau 2016). The dissimilarity matrix was calculated using the Bray-Curtis distance metric (e.g. Kneitel 2014; Leveau and Leveau 2016) with species abundance data. Two dimensions were specified. Environmental variables were fitted onto ordination using the envfit function in the R vegan package, and only significant variables (p < 0.05) were kept and plotted. This analysis was conducted in R, version 3.4.3 (R Core Team 2017), packages vegan (Oksanen et al. 2018) and ggplot2 (Wickham 2016).
Species densities were used to conduct a one-way analysis of similarity (ANOSIM) to test for differences in community composition between urban and exurban parks. Density was calculated as individual birds per hectare (e.g. Zhou and Chu 2012): using the abundance of birds across the five surveys within 50 m of the transect. Significant ANOSIM results were followed by a Similarity of percentages (SIMPER) test, using Bray-Curtis dissimilarity to determine the relative contribution of each species to differences in community assemblages (e.g. Kneitel 2014) between park types. ANOSIM and SIMPER were conducted using PAST, version 3.19 (Hammer, Harper, and Ryan 2001).

Results
A total of 437 observations of 50 bird species were made during five surveys at 17 parks (Table 2). Thirty-five species were observed at exurban parks, and 36 at urban parks. Exurban parks had an average of 11.56 6 1.18 (all results are presented 6standard deviation) species per park, and urban parks averaged 12.75 6 2.82 species per park. Species richness and area had a linear relationship (Fig. 2). After adjusting for the effect of park size, park types did not differ significantly in total number of species observed (ANCOVA, F ¼ 0.866, df ¼ 1, P ¼ 0.368). Exurban (x ¼ 22.56 6 11.05) and urban (x ¼ 24.4 6 17.75) parks did not differ significantly in the abundance of birds present (two sample Student's t-test, t¼À0.043, df ¼ 15, P ¼ 0.967). The average number of species observed per survey across all parks varied between 3.4 and 6.6 species.
Exurban parks tended to have been be smaller and have lower tree height and DBH; while urban parks tended to be larger and have higher tree height and DBH (Table 3). Exurban parks had more fruiting trees on average and tended to have more grass cover and less tree cover within the parks than their urban counterparts. On a landscape scale, exurban parks tended to have more grass cover and less tree cover in the surrounding environment (Table 3). Age and area of parks were not correlated (r ¼ 0.35, P ¼ 0.16).
The best model for species richness explained 81% of the variance in the data and included average DBH, landscape percent built cover, landscape percent water cover and park area (R 2 ¼ 0.81, P < 0.001; Table 4). Park area was a significant, positive predictor of species richness (t ¼ 6.84, P < 0.001). Although non-significant, landscape percent built cover was retained in the best fit model and was a positive predictor of species richness (t ¼ 1.33, P ¼ 0.207). Average tree DBH (t¼À1.3, P ¼ 0.218) and landscape percent water cover (t¼À1.69, P ¼ 0.117) were nonsignificant, negative predictors of species richness. The best linear model for avian abundance explained 75% of the variance in the data and included park percent built cover, landscape percent built cover and landscape percent water cover, average number of fruiting trees at park and park area (R 2 ¼ 0.75, P ¼ 0.004; Table 4). Park percent built cover (t¼À3.67, P ¼ 0.003), landscape percent water cover (t¼À4.74, P < 0.001), and average number of fruiting trees at park (t¼À2.71, P ¼ 0.02) were all significant and negatively associated with abundance. Landscape percent built cover was also negatively associated with abundance but was non-significant (t¼À2.09, P ¼ 0.06). Park area was the only positive, though non-significant, predictor of abundance (t ¼ 1.98, P ¼ 0.073).
Communities segregated in NMDS space according to park location type (Stress ¼ 0.15; Fig. 3). When parks were ordinated into multidimensional space, urban and exurban parks segregated along dimension 1. The environmental factors that significantly drove dimension 1 were distance to river (r 2 ¼ 0.67, P ¼ 0.003) and landscape percent grass cover (r 2 ¼ 0.55, P ¼ 0.002), which were associated with exurban parks; and tree DBH (r 2 ¼ 0.57, P ¼ 0.007), tree height (r 2 ¼ 0.76, P < 0.001), park Avian community composition | 5 percent tree cover (r 2 ¼ 0.51, P ¼ 0.005) and landscape percent tree cover (r 2 ¼ 0.62, P ¼ 0.002), which were associated with urban parks. The exception (an exurban park that segregated with the urban parks, Fig. 3) was the oldest exurban park, which exhibited similar characteristics to urban parks, such as more mature trees with higher DBH.

Species richness and abundance
Total bird species richness and abundance was not significantly different at urban and exurban parks during the nonbreeding season, suggesting that divergent park structure and landscape characteristics can support similar numbers of birds and species. This was contrary to our hypothesis that urban parks would have higher species richness and abundance. The exurban parks in this study could be experiencing a temporary uptick in both richness and abundance. Bird assemblages contain more species in the initial stages of urbanization (Jokimä ki and Suhonen 1993) because moderate disturbance increases bird abundance and richness (Blair 1996(Blair , 2004. As expansion into habitats such as grassland occurs, urban adapter species from the regional species pool are seen (Blair 1996;(Suarez-Rubio, Leimgruber, and Renner (2011). However, as the developed landscape spreads, abundance remains high while homogenization of bird communities takes place (Aronson et al. 2014;Leveau, Jokimä ki, and Kaisanlahti-Jokimä ki 2017) and richness decreases. Whether richness and abundance at the parks in this study will remain similar over time is unclear.

Variable
Exurban parks Urban parks reinforces the importance of size when planning to conserve or restore habitat patches within urban areas (Lepczyk et al. 2017).

Community composition
Although species richness and abundance did not differ between the urban and exurban parks, considering these findings in isolation could lead to misleading interpretations about the similarity of their avian assemblages. Species composition between the parks in urban versus exurban areas in the nonbreeding season was significantly different. This is consistent with other studies, where significant differences in community assemblage were driven by urbanization (Trollope, White, and Cooke 2009;Minor and Urban 2010). The majority of species seen across parks (48 out of 50 species) were native to California, suggesting that during the nonbreeding season these parks are important habitat for native birds. Urban parks had only one nonnative species, the European Starling, yet it was the second most abundant species at those parks. This may be due to a higher abundance of mature trees and urban structures, which can provide cavities for nesting. The high abundance of starlings, an urban adapter species (Zhou and Chu 2012) that can outcompete native species for nest cavities, suggests that summary statistics such as species richness and abundance should be considered along with species composition data.
Species overwintering in temperate regions in the USA are declining in abundance (Rosenberg et al. 2019). A total of 15 out of the 50 species observed at all parks in our study were migrants, underscoring the potential importance of parks as refuges for migrant species. Because similar abundances of migrants were observed at both park types, this appears to be true in both urban and exurban habitats. Monitoring exurban areas over time could help establish whether this remains true as exurban parks and surrounding areas mature.
Grassland species have shown the greatest declines in abundance since 1970 of any North American bird species, and 74% of grassland species are declining (Rosenberg et al. 2019). The exurban parks, characterized in part by more grassland habitat in the surrounding landscape, tended to be dominated by grassland species, including Brewer's blackbirds and Western meadowlarks. The Western meadowlark has a habitat requirement of grassland greater than five hectares (Johnson and Igl 2001). One critique of park planning is the tendency to maximize a diversity of habitats within a single park (Aronson et al. 2017). Instead, the management of large, contiguous areas of habitat such as grassland, when multiplied across different habitat types at the broader scale, may be more beneficial to species (Aronson et al. 2017; also see Lepczyk et al. 2017). Recently built parks with young saplings might not support as many migrants or species that forage or roost in dense tree canopies, but they can provide refuge to displaced grassland species whose native habitat was transformed by recent development. Their role as a refuge in this context could be enhanced by increasing park size, perhaps by opting for a single large park within a development complex in lieu of several smaller parks. There are some examples of this in the Central Valley, in the form of Specific or Habitat Conservation Plans, which can promote a common preserve area for multiple development projects. However, as these young parks mature and the adjacent grassland is converted to urban cover, the bird communities will continue to respond and change, and it is possible grassland species will be displaced.

Conclusion
In the study area, bird species showed a spatial organization due to local and landscape habitat features of parks. To support a variety of bird species, no clear patterns emerged on a parkby-park basis. However, parks with divergent features supported comparable species richness and abundances, supporting the assertion that increased environmental heterogeneity between parks can increase bird diversity overall (Lepczyk et al. 2017). The communities of avian species utilizing each habitat type varied, therefore contributing to greater species diversity across the entire study area. Future research could parse these species into their functional groups to explore the relationship between functional diversity and park habitat characteristics. While urban environments pose great challenges to many bird species, they also present new niche opportunities (Shanahan et al. 2014). A diversity of parks or other refuges within urban landscapes, strategically located and prioritized for large, contiguous size, could slow the spread of biotic homogenization and help support native species. Importantly, the development  of well-designed parks cannot replace preservation of threatened habitats as a strategy to conserve species, so we caution that our results do not support the conversion of natural grassland habitats in favor of exurban housing development; however, our results do suggest that exurban parks can support some native species.

Supplementary data
Supplementary data are available at JUECOL online. Avian community composition | 9