*Correspondence: Richard A. Fuller, Biodiversity and Macroecology Group, Department of Animal and Plant Sciences, University of Sheffield, Sheffield S10 2TN, UK. E-mail: firstname.lastname@example.org
Aim Urban environments are often characterized as supporting a few abundant, generalist species best adapted to living alongside humans, and as such, cities are seen as agents of biotic homogenization. However, there are surprisingly few descriptions of biological populations found in cities. Here, we provide the first complete citywide population estimates of birds for any UK city, and examine the conservation status of the assemblage in comparison with the country's avifauna at large.
Location Sheffield city, central UK
Methods We surveyed birds in every 500 m × 500 m square across the 160 km2 of the city. Using a Distance sampling protocol, we estimated bird population sizes for the city and compared these with the size of the human population. We also compared the conservation status of the city's avian population with that of birds across the UK as a whole.
Results Aggregation of population estimates for the 77 species observed during the surveys produced a total estimate of 602,995 (95% confidence interval (CI): 404,565–942,573) breeding birds, equating to 1.18 birds per person. The size of the non-breeding population was similar at 578,603 (464,396–728,574) individuals, or 1.13 birds per person. Surveys revealed only three non-native species, but relatively few species of national conservation concern. However, some species of conservation concern achieved very high population densities within the city, and the overall density of birds was more than six times that of the nation at large.
Main conclusions If declines in some species are to be arrested or reversed, conservation effort will need to focus much more strongly on understanding and managing urban populations, because these might buffer some species against wholesale regional population depletion, particularly where intensive agriculture in the surrounding hinterland has led to declines in bird populations at large. Such a focus will require a significant increase in the priority and resources devoted to conservation activities in urban areas.
The human enterprise has transformed global ecology, precipitating an extinction crisis (Lawton & May, 1995). Many anthropogenic impacts arise through changes in land cover (Meyer & Turner, 1992), and humans have now fundamentally altered between a third and a half of the Earth's terrestrial surface (Vitousek et al., 1997). Concomitantly, the human population has become increasingly concentrated into urban centres, such that about half of the world's people currently live in cities (United Nations, 2004). Although urbanization represents one of the most spectacular ways in which humans have altered landscapes, broad-scale quantitative studies on urban assemblages are scarce, and the conservation tools are often lacking to identify, understand and manage ecological change in such areas (Collins et al., 2000; Marzluff et al., 2001; Miller & Hobbs, 2002; Adams et al., 2006).
Despite these examples, studies that quantify entire urban assemblages are largely wanting, and little is known about how urban assemblages compare with the regional species pool from which they are drawn. This is important for at least three reasons.
First, many common and widespread species are in decline across much of Europe and North America, and such declines are perhaps underemphasized in modern conservation biology (Gaston & Fuller, 2007, 2008; Voříšek et al., 2007). Urban habitats in Europe support significant populations of several native common and widespread bird species, some of which are declining rapidly and hence listed as nationally threatened (Gregory & Baillie, 1998; Mason, 2000; Bland et al., 2004; Cannon et al., 2005; Newson et al., 2005; Robinson et al., 2005; Brichetti et al., 2008). Despite covering only about 7% of the land surface area of the UK, urban and suburban habitats support about 40% of the British population of common starlings Sturnus vulgaris, 12% of song thrushes Turdus philomelos (Gregory & Baillie, 1998) and 49% of house sparrows Passer domesticus (Robinson et al., 2005). Because of national declines in these native species of over 50% in the last 25 years, all three were red-listed in the most recent review of the conservation status of UK birds (Gregory et al., 2002). The Europe-wide conservation status of common starling and house sparrow has been listed as unfavourable (BirdLife International, 2004), and if such populations of declining species form a significant part of urban avifaunas, we may need to reappraise the fundamental conservation importance of such habitats.
Second, a relatively limited set of synanthropic species appears particularly well adapted to urban environments, and gross similarities in the design and function of cities across the world in combination with human-assisted dispersal might promote global and regional biotic homogenization (Johnston, 2001; Kühn & Klotz, 2006; McKinney, 2006). This has led to concerns that urban faunas may appear rich, but simply reflect local subsets of a homogeneous group of largely non-native species particularly well adapted to urban living at the expense of populations of native and/or rare species (McKinney & Lockwood, 1999). However, without comprehensive data on urban assemblages, it is difficult to assess the conservation significance of urban faunas.
Here, we present population estimates for birds across a UK city of half a million people. We then examine the conservation significance of the assemblage by comparing the richness and abundance of species of national conservation concern with those of the national avifauna.
With a human population of c. 513,000, Sheffield is the fifth largest municipality in the UK, and the ninth largest urban area (Office for National Statistics, 2001; Beer, 2005). For the purposes of this study, the urban area of Sheffield (53°22′N, 1°20′W) was defined as the set of 1 km squares within the administrative city boundary with more than 25% of their area comprising urban built form as assessed by eye using 1 : 25,000 Ordnance Survey maps (see Gaston et al., 2005). This resulted in a set of squares totalling 160 km2. Each square was split into four 500 m × 500 m cells, and a sampling point was randomly located within each cell, yielding a final set of 640 sampling locations distributed across the city. While this approach will inevitably include some cells in predominantly non-urban local environments, the aim of our study was to generate citywide bird population estimates rather than restrict our sample to cells with intense local urbanization.
Each sampling point was visited twice, once in winter (4 November 2004 to 28 February 2005) and once in summer (24 May to 1 July 2005). Sampling points were located in the field using a hand-held GPS receiver. A single observer (R. A. Fuller) conducted all survey work. Five-minute point transects (Buckland et al., 2001) were conducted between 0535 h and 1313 h in summer and between 0830 h and 1300 h in winter. Observations were begun immediately on arrival at the point location, and any birds disturbed by the observer prior to arrival at the point were recorded, together with an estimate of their distance from the point location. As might be expected in an urban setting, many of the birds seemed well habituated to human presence, and movement in response to the observer was minimal. Solitary birds were recorded individually, but those clearly associated with other individuals were recorded as clusters. For each detected solitary individual or cluster of birds, species identity, group size, radial distance from the observer and activity (perched or in flight) were recorded. Distances were estimated in the field in 14 bands (0–4.9 m, 5–9.9 m, 10–14.9 m, 15–19.9 m, 20–24.9 m, 25–29.9 m, 30–39.9 m, 40–49.9 m, 50–59.9 m, 60–69.9 m, 70–79.9 m, 80–89.9 m, 90–99.9 m, 100 m +). Distances were estimated with reference to features of the surrounding urban environment and after intensive practice in estimating distances in an urban setting involving measuring exact distances to birds after their estimation. Distances close to the observer (and hence less prone to error) are disproportionately important in calculating the detection function.
Bird density estimates were calculated using Distance software (version 5; Thomas et al., 2005). Where sample sizes permitted, analyses were conducted separately by season (summer, winter) and by species. First, four plausible candidate key functions (uniform with cosine expansion, uniform with simple polynomial expansion, half-normal with hermite polynomial expansion, and hazard-rate with cosine expansion) were used to model how detection declined with distance, and the model with the lowest value for Akaike's Information Criterion (AIC) was chosen. The distance data were truncated beyond the point where the probability of detection in this initial model fell below 0.1 (usually 40–50 m). The analyses were then re-run using the same four candidate models, and options for grouping the distance bands were explored until a good fit between model and data was obtained, as judged by comparing the modelled detection function against the observed distance data, the goodness of fit statistics and AIC values for alternative grouping and detection function options.
Detectability of birds showed very little variation with the degree of urbanization, and resulting density estimates were not biased by urban form (there was no correlation between detection distance and urban form as measured by the proportion of impervious surface in the 100 m radius around a survey point; for details of these analyses see Fuller et al., 2008). Detection functions were therefore estimated globally for each species/season combination. In cases where there were too few (typically < 30) observations of a species to allow adequate independent modelling of the detection function, the probability of detection of a more frequently observed ‘surrogate’ species of similar size with apparently comparable detection characteristics was used to correct the distance data using a multiplier (Buckland et al., 2001; see Appendix S1 in Supporting Information for sample sizes and details of surrogate species assignment). The best fitting model for the detection function of each species/season combination was used to generate density and population estimates, together with their associated confidence intervals.
National population estimates for breeding birds across the UK were taken from Baker et al. (2006). A consortium of national conservation agencies recently reviewed the conservation status of all bird species regularly occurring in the UK (Gregory et al., 2002). This assessment used a three-level system to classify species as green, amber or red according to seven criteria including international threat status, and various aspects of historical and recent declines in population or geographical range size within the UK. Species that were red or amber listed were considered to be of national conservation concern for the purpose of our analyses.
A total of 77 species was observed during the survey work (61 in summer, 64 in winter). Only three non-native species were observed (ring-necked pheasant Phasianus colchicus, feral pigeon Columba livia and little owl Athene noctua; Table 1). Species detected only in flight or only beyond the truncation distance of the candidate surrogate species were excluded from analyses, leaving 55 species in summer and 53 in winter. Resulting Sheffield-wide population estimates for these species are presented in Table 1. The species-abundance distribution was strongly bimodal, characterized by many rare species, yet with an unusually large group of highly abundant species dominating the right-hand side of the plot (Fig. 1). The overall number of individuals estimated to be present in Sheffield was 602,995 (95% confidence interval (CI): 404,565–942,573) during the breeding season and 578,603 (464,396–728,574) during the winter. Using a human population estimate of 513,000 for the city (Office for National Statistics, 2001), these numbers equate to 1.18 birds per person in the breeding season and 1.13 in the winter. Populations of the three non-native species comprised only 2% of the total number of breeding birds in the city (Table 1). To assess the robustness of the values arising from the Distance sampling analysis, we constructed a ‘retrospective’ point-count data set with a fixed radius of 20 m, a distance at which we can be confident that a high proportion of birds was detected. This resulted in total population estimates for the city's birds of 517,453 birds in summer (compared with 602,995 derived from Distance sampling) and 423,750 birds in winter (compared with 578,603 derived from Distance sampling). The smaller population estimates are consistent with some individuals within 20 m eluding detection, but could also result from overestimation of the true values by the Distance sampling method. However, species population estimates derived in the two ways are extremely highly correlated (summer, rs = 0.955; winter, rs = 0.910), and thus our conclusions would be qualitatively identical whichever method is used.
Table 1. Population estimates for breeding and wintering birds in Sheffield, UK. Values in parentheses are 95% confidence intervals for the Sheffield data. Species codes are shown in parentheses after the English name, superscript symbols indicate current national conservation status (Gregory et al., 2002): *Green listed, †Amber listed, ‡Red listed. Non-native species are indicated by an asterisk, although the historical distribution of feral pigeon is uncertain (Cramp, 1985). An additional nine species were observed for which density estimates could not be calculated because detections were beyond the truncation distance employed in the analysis, or were exclusively of flying birds (great cormorant Phalacrocorax carbo, greylag goose Anser anser, northern lapwing Vanellus vanellus, European golden plover Pluvialis apricaria, glaucous gull Larus hyperboreus, lesser black-backed gull Larus fuscus, great black-backed gull Larus marinus, brambling Fringilla montifringilla and lesser redpoll Carduelis cabaret).
The 61 species observed during the summer surveys comprise about 76% of the total of 84 species known to breed regularly within the boundaries of urban Sheffield (Hornbuckle & Herringshaw, 1985). The five commonest breeding species in Sheffield were house sparrow, blue tit, common starling, common blackbird and common swift, which between them comprised 64% of all individual birds (see Table 1 for scientific names). Twenty-seven species occurred at higher breeding densities in Sheffield than in the country as a whole, 10 by more than an order of magnitude (feral pigeon, Eurasian collared dove, common swift, house martin, blue tit, black-billed magpie, common starling, house sparrow, European greenfinch, European goldfinch; Fig. 2). The mean density of bird species breeding in Sheffield was much higher than the mean density across the country as a whole (t = 6.47, d.f. = 272, P < 0.001). The total bird density in Sheffield was 3769 individuals per km2, more than six times the national average of 551 individuals per km2 (obtained by dividing total area of the UK by summed population estimates for all breeding species). If calculated based on birds detected within a fixed radius of 20 m, estimated total bird density drops to 3234 individuals per km2, still 5.9 times the national average. Very few species with a national density of < 1 individual per km2 occurred above that density in Sheffield (Fig. 2), indicating that the majority of the city's avifauna comprised nationally common species, something reflected in the bimodal species abundance distribution (Fig. 1).
Nationally, of 223 breeding species assessed by Gregory et al. (2002), 76 (34%) were green listed, and the remaining 147 (66%) amber or red listed. Proportionately fewer species of national conservation concern bred in Sheffield compared with these national figures. This held using our survey data (32 (58%) green listed, 23 (42%) amber or red listed; Gadj = 10.43, P = 0.001) and using data from the intensive atlas-based work reported by Hornbuckle & Herringshaw (1985) (43 (50%) green listed, 43 (50%) amber or red listed; Gadj = 6.5, P = 0.011). Summing the relevant population estimates for Sheffield from Table 1 gives a total of 298,270 (224,568–413,622) individuals of green listed species and 304,725 (179,997–528,951) individuals of amber or red listed species (i.e. 50.5% of individuals are of national conservation concern). The equivalent national figures are 90,719,062 (89,315,916–92,122,208) individuals from green listed species and 43,833,640 (41,535,272–46,232,008) individuals from amber or red listed species (i.e. 32.6% of individuals are of national conservation concern). Nationally threatened breeding bird species occurred at a significantly greater frequency in Sheffield than across the country as a whole (Gadj = 82,426, P < 0.0001).
As might be expected for birds resident in the UK as a whole, there was a strong positive relationship between the density of Sheffield birds in summer and winter (Fig. 3). However, of these resident species, significantly more common blackbirds, blue tits and carrion crows but fewer European goldfinches and song thrushes were present in the city in winter than in summer (Fig. 3). This increase in numbers of green-listed species resulted in a significantly lower frequency of individuals of threatened species in Sheffield in winter than in summer (Gadj = 3893, P < 0.0001).
Birds occur at very high densities within Sheffield city, much higher than across the UK at large. Moreover, some of the commonest species within the city's avifauna are nationally (Table 1) and continentally threatened, having experienced large population declines across much of their native ranges. At least some cities, therefore, as well as supporting very high densities of birds per se could play a significant part in overall conservation effort in urbanized nations.
In terms of simple species richness, the surveyed Sheffield avifauna (61 species) is comparable with data from France, Finland and Italy summarized in Clergeau et al. (2006), where an average of 43 species was observed in surveys restricted to suburban areas. Atlas-derived data, while more comprehensive will include species only occasionally breeding within city limits. Such data indicate an average of 115 species in the breeding avifauna of 16 large cities (mean population of 1.9 million people) across Europe (Kelcey & Rheinwald, 2005). A citywide survey of Washington, D.C. detected 91 bird species, with the total avifauna estimated at 115 (Hadidian et al., 1997). Sheffield, therefore, does not appear to be particularly unusual in terms of overall bird species richness.
There are some 134.6 million breeding birds in the UK (from data in Baker et al., 2006), equivalent to c. 2.24 birds per person, while within the city of Sheffield, this ratio falls to 1.18. That it does not fall further is remarkable given the extremely densely aggregated human population. The species-abundance distribution was an unusual shape, with distinct peaks of rare and common species, and rather few species of intermediate abundance in the assemblage (Fig. 1). This strongly bimodal pattern differs markedly from the more obviously unimodal and closer to (albeit not actually) log-normal species-abundance distribution of UK farmland birds and British birds as a whole (Preston, 1948; Gregory, 1994; Williamson & Gaston, 2005). Heavily urbanized sites have been associated with low levels of evenness in abundances for several taxonomic groups including lizards (Germaine & Wakeling, 2001), birds (Clergeau et al., 1998; Marzluff, 2001) and bats (Kurta & Teramino, 1992), and this is usually attributed to local extinctions of rare native species (Donnelly & Marzluff, 2004). However, the peaks in rare and common species shown by Sheffield's birds suggest that a significant proportion of species are well adapted to urban environments, while many others achieve only very low densities. This is consistent with predictions that high food density and low predation risk in urban environments will result in a few species becoming superabundant (Shochat et al., 2004; Shochat, 2004). However, the magnitude of the peak on the right-hand side of the species-abundance distribution (Fig. 1) suggests that the proportion of such ‘winners’ in an urban assemblage might be unexpectedly high.
The rural/urban contrast in population densities reported here is probably quite common in areas with intensive agriculture, and likely to spread as regions become more developed, and with recent trends toward increasing production as grain prices have soared. Historical declines in house sparrows and common starlings, two of the commonest species in Sheffield's avifauna, have been most dramatic outside urban habitats and particularly in farmland, where both were formerly abundant. The breeding population of common starlings has declined by at least 92% since 1965 in woodland habitats, and by 66% since 1962 in farmland (Crick et al., 2002). Declines in urban habitats have been less severe, although comparable long-term data are lacking because urban habitats were poorly covered before the advent of a national UK Breeding Bird Survey in 1994 (Newson et al., 2005). In the 4 years between 1979 and 1983, rural populations of house sparrows declined by 38%, although more recently regional differences in decline rates have become apparent. The decline of urban house sparrow populations commenced later, around 1983, and appears to be continuing such that by the late 1990s populations in suburban gardens had declined by c. 60% (Cannon et al., 2005; Robinson et al., 2005). Recent declines in these species exemplify the wider point that cities can function as refugia for some species in the face of widespread intensive agriculture, and we believe that significant conservation effort should now be directed towards urban bird populations. This is true not only for birds. Many rare plants in the USA and Europe are concentrated in and near urban centres (Schwartz et al., 2002; Kühn et al., 2004), highlighting the importance of urban green spaces for plant conservation (Stalter et al., 1996).
Most species nationally resident in the UK showed consistent breeding and non-breeding densities in Sheffield (Fig. 3), suggesting that our results are valid year-round for resident species. Exceptions were carrion crow, common blackbird, blue tit (disproportionately abundant in the non-breeding season), song thrush and European goldfinch (disproportionately abundant in the breeding season). No trait obviously unites the species in these two groups and the results highlight the need to understand in more detail how individual species utilize urban habitats over the course of the annual cycle. Many resident species are more mobile outside the breeding season, perhaps as a result of resource fluxes, and there may be significant movements into (and out of) cities during the winter (Wernham et al., 2002). For example, urban robins are sedentary in urban habitats yet only breeding visitors to nearby woodlands (Adriaensen & Dhondt, 1990), and movement into gardens of passerine birds occurs in years where levels of beech Fagus sylvatica fruiting are low (Chamberlain et al., 2007).
Given that urban areas can support high densities of nationally threatened bird species, our results suggest that greater priority needs to be given to conservation of urban populations. This is particularly crucial given the trend toward building new urban developments at high density in the UK (Department of Communities and Local Government, 2006), which could adversely affect urban bird richness and abundance (Tratalos et al., 2007). Urban bird declines show strong regional variation (Sanderson, 1996; Dott & Brown, 2000; Prowse, 2002; Robinson et al., 2005), and an investigation of within- and between-city factors leading to such changes will help predict the trajectory of such changes in the future.
This work was supported by the Engineering and Physical Sciences Research Council (through the CityForm research consortium). R.A. Fuller is additionally supported by the Applied Environmental Decision Analysis research hub, funded through the Commonwealth Environment Research Facilities programme, Australia, and K.J. Gaston holds a Royal Society-Wolfson Research Merit Award. Ordnance Survey supplied MasterMap data under license to the CityForm consortium. We are grateful to J. Booth, R.G. Davies, K.L. Evans, B. Goettsch, S.E. Newson and P.H. Warren for comments and discussion. Two anonymous referees provided useful comments on an earlier draft of this manuscript.