1. Introduction

Urbanization is known to alter the species richness and diversity of many communities and/or assemblages [1, 2], and many researchers have chosen to focus on the response of bird species richness to urbanization [1, 3–5]. Birds have been the focus of so many studies because they are sensitive to changes in habitat structure [1] and may serve as indicators of the ecological consequences that can accompany urban development. Additionally, many years of data on species richness and abundance exist through standardized bird surveys such as the North American breeding bird survey (BBS) in the United States, by which long-term changes in bird assemblages can be monitored.

Studies of how anthropogenic factors impact bird species richness have offered several contradictory perspectives. First, many studies that assess richness along an urban-to-rural-gradient have found that richness is generally negatively [1, 4], or unimodally (i.e., widely distributed species are added at the expense of native species with moderate urbanization) [6] related to urban development. Second, several recent studies have suggested that many areas of high human population density (HHPD) are also areas with high avian species richness [5]. This positive correlation is generally thought to result from the positive impact that high energy inputs (and thus high primary productivity) have on both human population density and avian richness. Finally, many organisms have demonstrated a pattern of decreasing global species richness, while local richness increases via species invasions [7]. A more complete description of the impact HHPD has on bird assemblages can be achieved by examining assemblage dynamics over long time periods.

We studied bird assemblage dynamics in counties with high and low human population densities (LHPDs) across the U.S. to determine if assemblage dynamics differ over a 25-year period as a function of human population density.Our analyses use statistical techniques that account for the fact that not all species present are necessarily detected, which is an important consideration in long-term studies [8].

2. Methods

We evaluated all U.S. ecoregions [9] and selected those for study containing pairs of BBS routes with one route in a county of LHPD (65 persons/) and one route in a county with HHPD (700 persons/) based on data from the 2000 United States census. We only evaluated routes with at least 20 of 25 years of available BBS data. Twelve of the 68 ecoregions within the United States contained LHPD and HHPD routes that met all of our criteria (Figure 1).

Figure 1: To illustrate our route selection process, the main map shows portions of two ecoregions containing breeding bird survey (BBS) routes used in this study. Within each ecoregion, we selected one BBS route in an LHPD area and one in an HHPD area (see Methods for definitions). Shaded polygons on main map indicate county population density. Inset shows all 12 ecoregions used for the study.

We compared assemblage change over the past 25 years between HHPD and LHPD sites using measures of the rate of species increase, local extinction rate, turnover, and the number of colonizing species (mathematical definitions and derivations are provided in [8, 10]). The rate of species increase is a measure of an overall change in species richness between two points in time within a site. Local extinction rate is the probability that a given species becomes locally extinct between the two sampling periods. Species turnover is the probability that a given species colonized the focal area since the previous sampling period.The number of colonizing species is an estimate of the number of species new to the assemblage relative to a previous sampling period.Collectively, these four measures are hereafter referred to as assemblage properties. We calculated assemblage properties using Program COMDYN [10]. This software allows for the calculation of species’ detection probabilities, rather than relying only on the species observed to generate assemblage properties.

To test for an effect of HHPD on bird assemblages, we compared mean year-to-year assemblage properties using Hotelling’s multivariate -test [11]. By calculating the mean year-to-year values for our assemblage properties, we estimated the persistence of an assemblage, where persistence was defined as the maintenance of assemblage composition over time. We first used two one-sample -tests to examine the null hypotheses that bird assemblages in HHPD and LHPD areas experienced no assemblage change over the past 25 years. Next, we used paired -tests to examine the prediction that average year-to-year assemblage parameters were higher in HHPD habitats relative to LHPD habitats. When an overall test was significant, we tested for pairwise significance between each pair of variables in the overall test using Bonferroni simultaneous confidence intervals. To examine the species composition that contributed to any observed changes in assemblage parameters, we compared the relative numbers of area-sensitive and nonarea-sensitive species [12] present between past (species list from earliest 5 years of data) and present (species list from latest 5 years of data) using a paired t-test.

We used least-squares linear regression to test for an effect of human population density increase on bird assemblage properties. We plotted the difference between past (mean from earliest 5 years of data) and present (mean from latest 5 years of data) bird assemblage properties against the difference in past (1975) and present (2000) human population density. This approach tested the assumption that areas with greater increases in human population density over time will have greater impacts on bird assemblages. The test for each COMDYN-estimated property was performed using the pooled dataset of rural and urban BBS routes. All analyses were performed using Minitab 13.1. We considered P-values .05 to be significant for all tests.

3. Results

We found evidence for assemblage change over the past 25 years in both HHPD and LHPD areas ( and , resp.). In each area, the probability of local extinction, turnover, rate of species increase, and the number of colonizing species were significantly greater than the null hypothesis of no change (Table 1). Mean turnover at HHPD and LHPD sites was slightly higher than extinction rates (although confidence intervals overlapped), which suggests that on average the colonization of a route by a species absent in a previous year is more likely than the local extinction of a species.Rate of species increase from one year to the next was significantly greater than 1, which suggests a slow increase in species richness over the span of years we examined. Finally, the number of colonizing species was 8.3 in HHPD sites and 7.9 in LHPD sites. Both of these values (and associated confidence intervals) indicated that along a route in an average year, approximately eight species appeared that were not present in the previous year. Our examination of species composition revealed 75% of both LHPD and HHPD routes examined had increases in species richness. The mean number of area-sensitive species added along a route was not significantly different from the number of nonarea-sensitive species (). We found no evidence to support the hypothesis that bird assemblages in HHPD habitats had higher assemblage parameters (i.e., were more dynamic) than bird assemblages in LHPD habits ().

Table 1: Estimated mean year-to-year properties for bird assemblages in HHPD and LHPD areas over a 25-year period throughout the United States. All assemblage properties were first compared to the null expectation of no change, and then each individual assemblage property was compared between HHPD and LHPD areas. Rate of species increase is defined as a ratio between species richness in year 1 and species richness in year 2, thus if there is no change in assemblage structure (the null expectation), this value equals 1.

The increase in human population density within an area did not predict the magnitude of change in any of the assemblage properties over the 25-year period (all P-values .30). This was true for both HHPD and LHPD regions, despite the fact that the mean increase in human population density between 1975 and 2000 for HHPD areas was 477 people/, but only 7 people/ for LHPD habitats.

4. Discussion

For both LHPD and HHPD areas, the mean values for rate of species increase and number of colonizing species indicated that on average species richness increased over time (Table 1). A similar pattern of increasing species richness was reported by La Sorte and Boecklen [13]. While their study tracked several diversity measures overall BBS routes within the United States, they did not directly compare HHPD and LHPD habitats or incorporate species detection probability into their analysis. No mechanism has been identified empirically to explain thepattern of increasing species richness described here by La Sorte and Boecklen [13]. Because only two nonindigenous bird species colonized our routes during the dates evaluated and since observer bias is accounted for by our analysis, we suggest that biotic homogenization (i.e., increasing assemblage or community similarity between two sites) may be acting across native bird assemblages in the United States. As land continues to undergo development, bird species may be driven from preferred habitat into other areas, where they were previously absent. This process could lead to increased local diversity in either urban or rural assemblages [14]. In addition, common species that benefit from anthropogenic activity are colonizing new habitats [15], while species particularly sensitive to human development are likely to go locally or globally extinct. Along the routes, we examined both area- sensitive and nonarea-sensitive species were added to species lists in more recent years; however, we caution that only a small percentage of the total species richness along our routes has been classified into one of these two categories. As a result, a more thorough classification of all species will be necessary before a definitive statement can be made regarding the expansion (on a continental scale) of nonarea-sensitive species.

The absence of a significant difference in assemblage properties between areas of high and low human population density was surprising, as several previous studies documented a negative relationship between urban development and avian species richness [4, 16]. Differences between our findings and others may result because we included all observed bird species rather than a selected group such as neotropical migrants [4]. It is also possible that we did not detect a difference between assemblage dynamics in high and low areas of human population density because land use effects such as fragmentation from agricultural or urban development [17] (rather than human population density per se) drive bird assemblage dynamics. More data are needed to test this idea directly.

Future studies modeling large-scale assemblage or community dynamics with the mark-recapture approach used here and by others [8, 16–18] should consider two as of yet unaddressed aspects of the analysis. First, it is important to explicitly consider the relationship between mean year-to-year assemblage dynamics (e.g., local extinction rate) and ecological concepts such as community persistence or stability. An increase in the fluctuations of assemblage or community dynamics over time can signal a decrease in the overall stability of the assemblage or community [19]. Thus, we believe that measures of mean year-to-year assemblage dynamics offer a suitable, easily measured proxy to stability and persistence. A second analytical issue is the assumption that the null expectation for change within a reference assemblage (i.e., LHPD areas) is zero. Most assemblages and communities are not static with respect to species composition over time [20], but in the absence of good information on what degree of species change should be expected, we used no change as our null expectation. We hope the study we present here provides a baseline from which such null expectations might be developed, enabling one to distinguish assemblage composition shifts due to natural fluctuations from those caused by anthropogenic factors.

Acknowledgments

The authors thank J. E. Hines for help with Program COMDYN, T. Boulinier for helpful discussions on the analysis, and The Nature Conservancy for information on North American Ecoregions. Earlier versions of the manuscript were improved by comments from three anonymous reviewers, C. Guyer, J. A. Stratford, and S. M. Boback. J. B. Grand provided the impetus for this project.