Habitat influences on urban avian assemblages


*Corresponding author.
Email: karl.evans@sheffield.ac.uk


Urbanization is increasing across the globe and there is growing interest in urban ecology and a recognition that developed areas may be important for conservation. We review the factors influencing urban avian assemblages, focusing on habitat type and anthropogenic resource provision, and analyse data from a common bird monitoring scheme to assess some of these issues. The review suggests that (1) local factors are more important than regional ones in determining the species richness of urban avian assemblages, raising the potential for the management of urban sites to deliver conservation; (2) habitat fragmentation frequently influences urban avian assemblages, with the effects of patch size being greater than those of isolation, and (3) urban bird assemblages appear to respond positively to increasing the structural complexity, species richness of woody vegetation and supplementary feeding, and negatively to human disturbance. Data from Britain's Breeding Bird Survey, combined with habitat data obtained from aerial photographs, were used to assess a number of these issues at the resolution of 1-km squares. Green-space constituted 45% of these squares, and domestic gardens contributed 50% of this green-space, though their contribution to large continuous patches of green-space was negligible. There was no significant positive correlation between the densities of individual species in urban areas and surrounding rural areas. Rural species richness declined with increasing latitude, but urban species richness was not correlated with latitude. This contrast contributes to slightly higher avian species richness in rural squares in Southern England than urban ones. Occupancy and abundance were strongly positively correlated in urban avian assemblages, and some indicator species of conservation concern occurred in few urban areas and at low densities. Such species will require conservation action to be precisely targeted within urban areas. Of the urban indicators of conservation concern, only the House Sparrow Passer domesticus and Common Starling Sturnus vulgaris were more abundant in urban than rural areas. Moreover, the densities of these two species were strongly and positively correlated, indicating that they may be limited by shared resources, such as nest-sites or supplementary food. There was little evidence that high densities of nest-predating corvids were associated with reduced densities of their prey species. Species richness and the densities of individual species frequently declined with an increasing number of buildings. Current trends for the densification of many British urban areas are thus likely to be detrimental for many bird species.

Globally, urbanization is one of the fastest growing land uses, and over half of the world's human population is now located in urban areas (UN 2008). Urbanization has already consumed large amounts of land in some regions; nearly one tenth of the UK's land area is urbanized (Haines-Young et al. 2000, Fuller et al. 2002). Interest in urban ecology has risen commensurately (Marzluff 2001). Much of this interest has focused on how assemblage composition and structure vary along a gradient of land-use from rural to urban areas. Birds are particularly well studied in this regard. Urbanization is almost invariably associated with a loss of many avian species present before development, particularly ground-nesting species, habitat specialists and those that require large areas of intact habitat (Marzluff 2001, Yeoman & MacNally 2005, Chace & Walsh 2006, Clergeau et al. 2006, McKinney 2006). Avian species richness typically peaks at intermediate levels of urbanization, a pattern associated with higher habitat heterogeneity in such areas, yet total avian density usually peaks in highly developed regions due to high densities of a few synanthropic species (Blair 1996, Marzluff 2001, Chace & Walsh 2006, Tratalos et al. 2007, Grimm et al. 2008).

These synanthropic species can be of conservation importance due to large population declines in the wider landscape. In the UK, 27 bird species have been identified as urban indicators because they occur in urban areas to a greater extent than expected if their distribution was random (DEFRA 2002, 2003). Seven of these are of formal conservation concern (Table 1; Gregory et al. 2002). Three are on the UK red list due to population declines of more than 50% in the last 25 years (House Sparrow Passer domesticus, Common Starling Sturnus vulgaris and Song Thrush Turdus philomelos). Four are amber listed due to population declines of greater than 25% (Eurasian House Martin Delichon urbicum, Dunnock Prunella modularis and Mistle Thrush Turdus viscivorus), or because they are of European conservation concern (Green Woodpecker Picus viridis). An additional urban indicator species, the Common Swift Apus apus, has declined by 41% between 1994, when monitoring began, and 2007 (Risley et al. 2008). It thus now also meets the criteria for a species of conservation concern, although it is not yet formally listed.

Table 1.  The type and amount of green-space in 100 urban 1-km squares, and the contribution of domestic gardens to green-space. Sample sizes vary as some habitat features were absent in the focal squares.
 No. of sampling points that are green-space% green-space in gardens% green-space in patches ≥ 2500 m2% green-space in patches ≥ 2500 m2 in gardens
Total44.8 ± 1.41–74 (n = 100)50.3 ± 1.957–100 (n = 100)26.0 ± 1.60–68.5 (n = 100)4.6 ± 1.90–100 (n = 91)
Grassland25.0 ± 1.00–48 (n = 100)48.1 ± 2.1 0–100 (n = 99)36.3 ± 2.10–82.4 (n = 99)2.7 ± 1.40–100 (n = 90)
Bushes/trees18.9 ± 0.81–37 (n = 100)55.9 ± 2.5 0–100 (n = 100)13.3 ± 1.60–100 (n = 100)0.6 ± 0.40–16.7 (n = 62)

Some important ecological issues relevant to urban conservation are relatively unexplored, particularly regarding the spatial variation in biodiversity within urban areas and the drivers of such patterns. Here, we synthesize the available literature on factors influencing urban avian assemblages, focusing on habitat type and supplementary resource provision. We then use data from the UK's BTO/JNCC/RSPB Breeding Bird Survey (BBS) and information on the habitat characteristics of urban areas, from aerial photographs, to investigate a number of patterns. Specifically, we assess (1) the nature of urban green-space in the UK; (2) the characteristics of avian urban assemblages and variation in their structure and composition; (3) relationships between the densities of each species, including nest-predating corvids; (4) the extent to which regional factors structure urban avian assemblages and thus whether local management is likely to provide significant benefits; (5) the identity of species likely to benefit from conservation action in urban areas; (6) the nature of abundance-occupancy relationships in urban areas; and (7) the nature of urban bird–habitat relationships. We then synthesize our results to formulate recommendations for the conservation management of urban bird populations.


Literature review

Web of Knowledge was searched in February 2008, using the terms ‘urban and bird’ and ‘urban and avian’ with appropriate truncation. This identified papers that assessed the relative influence of regional and local factors on bird assemblages, bird–habitat relationships and supplementary resource provision by humans on urban bird assemblages. Papers on predation of urban birds were not included unless they assessed how predation risk was mediated by habitat type. For most of the issues which we addressed, insufficient studies were available for formal meta-analyses. Many studies were multi-functional, investigating issues other than our focal ones, and some reported both negative and positive results; publication bias is thus unlikely to influence the conclusions of our review.

Breeding Bird Survey (BBS) data

Data on breeding densities were obtained from the 2006 BBS. The BBS collects data on species densities in 1-km squares that are selected from a random stratified sample by BTO organizational regions. In each square, data are collected along two 1-km transects which are divided into 10, 200-m sections. We classified urban squares as those in which all 10 transect sections were coded as either urban or suburban. Of the 152 squares meeting this criterion, we randomly selected 100 for analyses of bird–habitat relationships (Fig. 1). This arbitrary sample size reflected a balance between maximizing the sample size and reducing the time spent collecting additional habitat data (see below). Our assessment of the influence of regional level factors on urban bird assemblages incorporated data from rural areas. We defined rural BBS squares as those in which no more than two of the 10 transect sections comprised human sites, the vast majority of which were small settlements. Species’ densities within each 1-km square were estimated with distance sampling and took variation in detectability functions between nine main habitat types and BTO regions into account (for more details see Evans et al. 2008 and Appendix S1).

Figure 1.

Location of the 100 randomly selected urban 1-km squares used to model bird–habitat relationships.

Habitat data

The BBS collects habitat data but at insufficient resolution to enable estimation of the distribution of features that may influence urban assemblages, such as buildings or trees. Therefore, further habitat data were taken from aerial photographs using Google Earth (accessed in February 2008). Exact dates are rarely available for specific images, but they are updated regularly in highly populated areas (http://en.wikipedia.org/wiki/Google_Earth). The dates of images thus generally corresponded with the timing of the bird surveys. A sampling grid of 100 evenly spaced points, 100 m between each point, was superimposed over aerial images of each BBS square. The habitat type at each point was allocated to one of a number of categories: (1) building, (2) road, (3) other hard surface (i.e. concrete, asphalt etc.), (4) water-body, (5) allotment, (6) garden grass, (7g) non-garden grass, (8) garden bushes/trees, and (9) non-garden bushes/trees; the last four categories were subdivided into areas of continuous habitat type smaller than 2500 m2 and those greater or equal to this area.

Statistical analyses

Analyses at the assemblage level focused on (1) the entire assemblage, (2) the 27 urban indicator species, and (3) the eight urban indicator species of conservation concern (including Common Swift). Analyses of individual species’ densities were confined to the 27 urban indicator species; these data were not normally distributed and so were analysed using non-parametric tests. We followed the recommendations of Nakagawa (2004) and did not apply Bonferroni corrections.

Latitudinal variation in the number of bird species present in urban 1-km squares was compared to that in an equal number of rural 1-km squares using bootstrapping with replacement in R (version 2.6.2). For each of 10 000 random samples of 152 rural 1-km squares, the correlation with latitude was computed and confidence intervals estimated using the quantiles of the parameter distribution.

We calculated the average density of each species across all rural squares within 15 km of each urban square; whilst the chosen distance is arbitrary, it is close to the mean natal dispersal distance of the 27 urban indicator species (mean 10.5 km, range 1.7–41.2 km; Paradis et al. 1998).

Species richness and density of the three species groups were analysed using Generalized Linear Models. The density of species of conservation concern was square-root transformed to meet the requirements of statistical tests, all other response variables were untransformed. We also modelled the densities, rounded to the nearest whole number, of each of the urban indicator species that occurred in 20 or more squares. Poisson errors and a log link were specified, with rescaling to account for overdispersion.

There was strong collinearity between some of the habitat variables so we selected those that recorded the main habitat types, but whose tolerance values were always substantially higher (range 0.48–0.94) than the threshold at which collinearity between predictors becomes a concern (0.1; Quinn & Keough 2002). The resulting predictors included three continuous variables: buildings, grassland and bushes/trees, together with their squared terms, and two factors: presence/absence of allotments and waterbodies. The last two variables were recorded as presence/absence due to their rarity in our focal squares.

We used an information theoretic approach to model building, which was conducted in SAS (vs9.1), and fitted all possible models. To assess the contribution of large habitat patches relative to smaller ones we constructed models (1) using the total number of sampling points that were grassland and bushes/trees, regardless of the size of the habitat patch, and (2) excluding those points that related to areas of grassland and bushes/trees smaller than 2500 m2. This gave a total of 489 models. The Akaike Information Criteria (AIC) values were used to estimate each model's weight, i.e. the probability that it provided the most parsimonious fit to the data. The fewest models whose cumulative model weights summed to 0.95 were included in the 95% confidence set of models. For all response variables, inspection of semi-variograms indicated that there was no significant spatial autocorrelation, and thus that the latter did not unduly bias the results of the independent error models.


There was marked geographical variation in the 72 studies describing the influence of habitat on urban avifaunas, the greatest number being conducted in Europe (38 studies). Eight of these were in the UK, fewer than both the United States (14 studies) and Australia (12 studies). Such studies therefore appear to be under-represented in the UK, one of the most urbanized countries in the world (Fuller et al. 2002, UN 2008) with an abundance of non-urban bird–habitat studies. Breeding assemblages were studied in 84% of papers, and wintering assemblages in 28%. Slightly fewer studies reported assemblage level responses (57%) than focused on individual species (69%), although many of the latter reported results for numerous species. Despite the relatively small number of urban bird–habitat studies, some topics have received sufficient attention to merit an initial synthesis of their findings (see Supporting Information Table S1).

Regional versus local factors

Whilst all studies consistently treat regional level factors as ones operating over a much larger area than local factors, there was much variation in the precise definitions. Some studies categorized features within 1 km of the focal site as the landscape or regional level (Melles et al. 2003, Smith 2007), others viewed habitat characteristics at this scale as local (Blewett & Marzluff 2005, Sims et al. 2008). Such variation hinders comparisons between studies, although a previous review concluded that the species richness of breeding and wintering urban bird assemblages is independent of the bird diversity of adjacent landscapes, and that local features are more important than surrounding landscapes (Clergeau et al. 2001). This conclusion is supported by additional studies. In Madrid, the number of urban parks occupied by a breeding species is not typically related to its regional density, except in newly created habitat patches (Fernández-Juricic 2000). Similarly, in Canada local features exert a greater influence than regional ones on urban breeding season distributions (Melles et al. 2003) and species richness in winter (Smith 2007). In London, the number of breeding and wintering species in green-spaces is influenced more by local factors than by the habitat composition of surrounding 3-km squares (Chamberlain et al. 2007a), and breeding densities in urban areas in Arizona tended to be more closely related to local habitat characteristics than regional ones (Hostetler & Knowles-Yanez 2003).

Although only a few studies have considered whether species’ densities respond more strongly to local or regional factors, the available evidence suggests that the former are more important in determining avian assemblage structure within urban areas. The design and management of particular urban sites thus clearly has the potential to influence their avifaunas.


Both breeding and wintering urban avian assemblages can exhibit strong species–area relationships (Tilghman 1987, Bolger et al. 1991, Jokimäki 1999, Natuhara & Imai 1999, Fernández-Juricic 2000, 2004, Pavlik & Pavlik 2000, Cornelis & Hermy 2004, Drinnan 2005, Watson et al. 2005, Antos et al. 2006, Hustéet al. 2006, Platt & Lill 2006, Chamberlain et al. 2007a, Husté & Boulinier 2007, Murgui 2007a; but see Kim et al. 2007). These relationships are among the most frequently observed ecological patterns and the mechanisms driving them in urban areas appear to be similar to those operating elsewhere, i.e. larger urban habitat patches tend to support greater habitat diversity (Cornelis & Hermy 2004, Chamberlain et al. 2007a), have reduced edge effects and support larger, and thus more stable, populations. For example, in North America a number of breeding bird species did not occur in linear urban habitat patches of less than a critical threshold width (Mason et al. 2007). Similarly, in Madrid the breeding densities of most bird species were higher in the centres of parks than at their edges, which differed little in habitat type (Fernández-Juricic 2001). Smaller habitat patches in urban environments support fewer individuals, thus increasing the risk of stochastic extinction and decreasing species richness (Bolger et al. 1991), and rates of temporal turnover in breeding assemblages are often greater in smaller urban habitat patches than in larger ones (Fernández-Juricic 2004, Husté & Boulinier 2007, Murgui 2007a). Species occurrence patterns tend to be nested with respect to the size of urban habitat patches, such that the species found in small patches are a consistent subset of those found in larger patches (Bolger et al. 1991, Jokimäki 1999, Natuhara & Imai 1999, Fernández-Juricic 2004, Platt & Lill 2006). In the UK, some woodland species, such as Long-tailed Tit Aegithalos caudatus, have a significantly lower probability of occurring in small woods than expected given the size of the wood (Hinsley et al. 1996, Dolman et al. 2007), and may thus be particularly susceptible to fragmentation in urban areas. A number of species occurring in urban environments thus appear to be area-sensitive (Mörtberg 2001, Environment Canada 2007).

Domestic gardens contribute little to large patches of urban grassland as garden lawns are often fragmented by boundary features such as hedges and fences. Even small changes to vegetation height or structure can reduce the rate of predator detection and thus alter avian habitat selection (Whittingham et al. 2004, Butler et al. 2005). Garden boundary features may have a similar effect but this has not been empirically assessed. Anecdotal evidence also suggests that some urban birds that typically forage on grassland, such as gulls Larus spp. and some corvids (e.g. Rook Corvus frugilegus), are observed infrequently on garden lawns. More work is, however, required to investigate the extent to which fragmentation of garden grassland influences birds, particularly compared to other potentially important factors such as sward management.

The degree of isolation from other habitat patches may also influence species distribution patterns. Species may vary in their willingness to cross habitat gaps in urban areas (Hodgson et al. 2007), so some are likely to exhibit a negative response to isolation of habitat patches. Such effects have been reported from Swedish urban forests (Mörtberg & Wallentinus 2000, Mörtberg 2001), and even for species that occur relatively frequently in urban areas, such as Great Tits Parus major (Hashimoto et al. 2005). Few studies have reported a lack of effect of isolation on species’ distribution (but see Mörtberg 2001), but the extent to which this is influenced by bias in the selection of focal species is uncertain. However, although isolation may suppress avian species richness (Natuhara & Imai 1999), studies that have considered isolation and patch size in combination have found that the latter tends to have the greatest effect (Hustéet al. 2006, Murgui 2007a).

When designing new urban developments, attention should be paid to both the size of individual patches of green-space and their spatial arrangement (Marzluff & Ewing 2001, Colding 2007). In the UK, it has been recommended that urban green-space should be at least 10 ha to maximize the number of urban bird species (Chamberlain et al. 2007a), although much larger patches may be required elsewhere (Environment Canada 2007). Increasing the size of existing urban habitat patches is unlikely to be practical, but the effects of area appear to arise, at least partly, from the influence of edge effects and habitat diversity, factors that may prove more amenable to management.

Vegetation structure and diversity

Urban areas are highly dynamic systems, and the successional stage of vegetation can have a marked influence on their bird assemblages. Some species typical of open landscapes only occur in the early stages of urban development, before mature vegetation cover becomes established (Vale & Vale 1976, Kelcey & Rheinwald 2005). The Black Redstart Phoenicurus ochruros, for example, shows a strong preference for urban areas with little tree cover (Sedláček et al. 2004). However, the majority of species occurring in urban environments prefer sites with more mature and structurally diverse vegetation (DeGraaf & Wentworth 1986, Pavlik & Pavlik 2000, White et al. 2005, Murgui 2007b).

Habitat structure can have a marked influence on breeding success through its interaction with predation risk (Evans 2004). Increased vegetation density often increases nest crypsis, at least to visually orientated predators, and more visible artificial nests are more likely to be predated (Jokimäki & Huhta 2000, Jokimäki et al. 2005). However, increasing crypsis may reduce the ability of parents to detect predators and defend the nest (Weidinger 2002), and studies using artificial nests cannot take such effects into account. Urban predation studies of real nests are rare. Wysocki et al. (2004) found that visually orientated predators used patches of dense vegetation as a search cue, leading to higher rates of nest predation in such habitats. In contrast, and at a larger spatial scale, such ecological traps were not detected by Leston and Rodewald (2006).

The influence of grassland structure and diversity on urban birds has not been investigated. However, in farmland the height and density of vegetation can have a marked influence on access to food, its intake rate and perceived predation risk, and thus determine species distributions (Whittingham & Evans 2004). Such effects probably arise in urban areas, but may be more limited due to reduced variation in grassland structure compared to rural environments.

Few studies have assessed the influence of plant species richness on urban bird assemblages. The relative contribution of different trophic guilds to Australian urban bird assemblages is strongly influenced by tree species identity (Young et al. 2007). Tree species richness tends to correlate positively with avian species richness in urban areas (Fernández-Juricic 2004, Hustéet al. 2006, Murgui 2007b). Indeed, the diversity of shrub and tree species was the most important factor, after patch size, in explaining breeding avian species richness in Parisian habitat patches (Hustéet al. 2006). However, breeding avian species richness and tree diversity in urban parks in Finland were marginally negatively correlated (Jokimäki 1999). Even fewer studies have considered the impact of tree species richness on breeding densities, although a positive correlation was present at the assemblage level in Madrid (Murgui 2007b), and for the densities of Pied Flycatcher Ficedula hypoleuca and Spotted Flycatcher Muscicapa striata in Oulu (Jokimäki 1999).

The few available studies suggest that increasing vegetation structure and diversity, at least of trees and shrubs, in urban areas is likely to enhance avian species richness. The responses of individual species are, however, likely to vary. Not all species may benefit from any one particular change, and maintaining habitat diversity will promote greater species richness.

Human disturbance

Cities are ideal sites for investigating the effects of disturbance on avian assemblages. However, surprisingly little research has been conducted on this topic, perhaps because it is assumed that urban birds are habituated to humans. Indeed, urban birds have shorter flight initiation distances, averaging approximately 10 m, than their rural conspecifics (Tomiałojć 1976, Cooke 1980, Campbell 2006, Møller 2008). Moreover, within urban areas flight distances are shorter at sites with more human visitors, demonstrating the ability of birds to adapt at least partly to disturbance (Fernández-Juricic et al. 2001).

Despite the apparent tolerance of urban birds, there is evidence of a disturbance effect. The Blackbird is one of the most urban-adapted birds, yet habitat use in urban areas is influenced by temporal variation in human presence, and breeding densities are lower in parks with more human visitors (Fernández-Juricic & Tellería 2000). Similarly, the density of breeding House Sparrows in urban parks in Madrid initially increased with the number of human visitors, probably in response to food availability, but densities declined when the number of human visitors was high (Fernández-Juricic et al. 2003). The nested pattern of species distribution in urban habitat patches in Madrid can partly be explained by human disturbance (Fernández-Juricic 2004), and densities of many breeding bird species in parks are significantly higher at park centres than at the edges, where human traffic is greatest (Fernández-Juricic 2001). Other studies have, however, failed to find an effect of human disturbance on urban bird distributions (Jokimäki 1999, Platt & Lill 2006).

These few studies suggest that disturbance effects may be significant, even in common urban species. Managing human traffic in urban settings, and that of their commensals such as dogs, is hindered by the lack of a framework for predicting and minimizing disturbance effects (Beale & Monaghan 2004). One possible solution is to increase the amount of cover habitat to reduce alert and flight distances (Fernández-Juricic et al. 2001, Campbell 2006). Attempts to lower human disturbance in urban areas must ensure that people are not displaced to sensitive rural areas on the fringes of urban settlements, where human disturbance can negatively impact species of conservation concern (Banks & Bryant 2007, Langston et al. 2007, Mallord et al. 2007).

Anthropogenic provision of resources

Feeding garden birds is common in many highly urbanized countries. In the UK, approximately 60 000 tonnes of supplementary food are provided annually (Glue 2006) and 48% of households provide such food, more than half of them through stocked feeders (Davies et al. in press). Numerous studies, some experimental, have detected positive effects of food provision on breeding success, survival rates and population size in rural areas (Newton 1998, Hole et al. 2002, Robb et al. 2008), suggesting that supplementary food may also positively influence urban bird assemblages. However, the factors regulating bird populations may be habitat dependent and the effects of supplementary feeding in urban areas are poorly understood.

Supplementary feeding may have negative consequences arising from provision of nutritionally poor supplementary food, increased disease transmission due to high densities of birds at feeding stations, and the attraction of birds to areas with competitors or predators (Parsons et al. 2006, Bradley & Altizer 2007, Jones & Reynolds 2008). Supplementary feeding can also adversely affect other species. For example, provision of supplementary food for hummingbirds can reduce pollination rates and seed production of native plants (Arizmendi et al. 2007).

Providing food increases the probability of occurrence and density of many species within domestic gardens (e.g. Wilson 1994, Chamberlain et al. 2004, Daniels & Kirkpatrick 2006, Parsons et al. 2006). However, it is more difficult to determine the effects of supplementary feeding at larger spatial scales. In Sheffield, UK, total avian density in the breeding season, but not species richness, was strongly positively correlated with the density of bird feeders (Fuller et al. 2008). This response was largely driven by the abundance of species known to take supplementary food. Species richness and total abundance of wintering birds in 30-ha plots increased with the number of feeding stations in Finland (Jokimäki et al. 2002). This pattern was not detectable in France, although some species’ densities were positively correlated with the density of feeders (Jokimäki et al. 2002). Further research is required to determine whether positive correlations are causal, or arise because more food is provided in areas where people see more birds, and to assess whether such patterns are restricted to more seasonal temperate regions where rural food availability is likely to be particularly low in winter.

Humans also frequently provide nest-boxes. In the UK, for example, there is a minimum of 4.7 million nest boxes within gardens, the vast majority in urban areas (Davies et al. in press). In rural areas, population size and trends may be determined by nest-site availability (Newton 1992, 1998), and such factors may thus be important in urban areas. In the UK, newer houses and those with recently repaired roofs provide fewer nesting opportunities for declining species such as House Sparrow and Common Swift (Wotton et al. 2002). There is particular concern that the lack of nest-sites may partly be responsible for declines in the Common Swift population (http://www.concernforswifts.com). However, empirical evidence for a role of nest-site availability in regulating urban bird populations is almost entirely lacking. The one exception is a positive correlation between the number of nestboxes and Pied Flycatcher density in urban parks in Finland, although the same study found that Blue Tit Cyanistes caeruleus densities were not influenced by nestbox availability, perhaps a reflection of nestbox hole size (Jokimäki 1999).

Provision of supplementary and targeted resources, particularly food, appears likely to influence the structure and composition of urban bird assemblages. Although provision of supplementary food is currently generally high, it is significantly less frequent in low-income households and economically deprived areas (Fuller et al. 2008, Davies et al. in press). Conservation expenditure that enables provisioning in such areas to be increased may thus benefit avian assemblages whilst also promoting human engagement with the natural world.


Habitat characteristics of urban BBS squares

Estimates of the amount and composition of urban green-space that used data from a sub-sample of 20 squares were very similar to estimates using data from all 100 sampled squares, suggesting that sampling additional squares would not have significantly altered the results. On average, 45% of the sampling points within BBS squares were green-space (Table 1). Most green-space comprised patches smaller than 2500 m2, patches larger than this comprised grassland more frequently than bushes or trees (Table 1). Domestic gardens contributed about half the total green-space, but a very small proportion of the larger patches (Table 1).

Bird assemblages in urban BBS squares

Urban species richness and density varied by an order of magnitude (see Supporting Information Table S2a), and were strongly correlated (see Supporting Information Table S2b). Similar variation was exhibited by the density and occupancy of the 27 indicator species, of these only the Common Blackbird occurred in all squares and two species, Common Swift and Eurasian House Martin, occurred in fewer than 10% of squares (Table 2).

Table 2.  Occupancy and density in 100 urban 1-km squares of urban indicator species (DEFRA 2002, 2003). †Species of conservation concern; ‡urban specialists. The Common Swift is listed as being of conservation concern due to rapid and large population declines that now meet the criteria for official listing (see text for more details).
SpeciesOccupancy (%)Density
RangeMedianMean ± se
Blackbird Turdus merula1008–322100114.3 ± 6.8
Woodpigeon Columba palumbus980–40299119.2 ± 9.7
Blue Tit Cyanistes caeruleus960–3758497.4 ± 7.5
†‡House Sparrow Passer domesticus920–1258137203.4 ± 20.3
Robin Erithacus rubecula920–1833842.8 ± 3.4
Common Starling Sturnus vulgaris890–88590117.1 ± 11.4
Eurasian Magpie Pica pica 880–1182329.2 ± 2.5
Carrion Crow Corvus corone870–2011522.7 ± 3.3
Great Tit Parus major840–992024.0 ± 2.1
Greenfinch Carduelis chloris840–22125.536.0 ± 4.0
Wren Troglodytes troglodytes800–982025.3 ± 2.4
Collared Dove Streptopelia decaocto710–2052138 ± 4.7
Chaffinch Fringilla coelebs700–1021314.5 ± 1.7
Dunnock Prunella modularis670–130822.5 ± 2.3
Song Thrush Turdus philomelos410–3904.1 ± 0.5
Goldfinch Carduelis carduelis390–112013.8 ± 2.3
Mistle Thrush Turdus viscivorus250–2803.1 ± 0.6
Blackcap Sylvia atricapilla250–3702.7 ± 0.6
Mallard Anas platyrhynchos210–410012.0 ± 4.7
Eurasian Jay Garrulus glandarius200–4802.4 ± 0.6
Long-tailed Tit Aegithalos caudatus190–8805.4 ± 1.4
Jackdaw Corvus monedula180–17306.7 ± 2.6
Pied Wagtail Motacilla alba130–2401.3 ± 0.4
Green Woodpecker Picus viridis100–2000.6 ± 0.2
†‡Common Swift Apus apus70–13502.6 ± 1.4
†‡Eurasian House Martin Delichon urbicum40–3200.7 ± 0.4
Sparrowhawk Accipiter nisus10–300.03 ± 0.03

The densities of indicator species were significantly correlated in 81 cases (23%), all but two of these correlations being positive (see Supporting Information Table S3). Among the species of conservation concern, the strongest correlation occurred between Common Starling and House Sparrow (rs = 0.60, P < 0.0001). The only other significant correlations between the densities of species of conservation concern were between Dunnock and (1) Song Thrush (rs = 0.34, P < 0.001), (2) Mistle Thrush (rs = 0.26, P < 0.01) and (3) House Sparrow (rs = 0.21, P < 0.05).

There was marked variation in the densities of nest-predating corvids (Eurasian Magpie Pica pica, Eurasian Jay Garrulus glandarius and Carrion Crow Corvus corone), each being absent in some squares (see Supporting Information Table S3). However, their densities were rarely significantly correlated with those of small passerines. The only exception was a negative correlation between House Sparrow and Jay; the former nests in cavities which offer protection from avian nest predators and the negative correlation probably reflects contrasting responses to increasing urbanization (Tratalos et al. 2007). Indeed, densities of nest-predating corvids were sometimes significantly positively correlated with those of small passerines (e.g. Carrion Crow with Robin Erithacus rubecula; Magpie with Greenfinch Carduelis chloris, Robin and Winter Wren Troglodytes troglodytes).

Regional influences and comparisons with rural assemblages

Urban species richness was not significantly correlated with latitude (all species: r = –0.03, urban indicators: r = –0.02, urban indicators of conservation concern: r = 0.12; P > 0.05, n = 152). This pattern was consistent with the lack of spatial autocorrelation detected in the structure of urban assemblages, but contrasted with the significantly negative latitudinal gradients in rural species richness (bootstrap analysis of 152 rural squares with 100 000 replicates: total richness r = –0.30, 95% confidence intervals –0.45 to –0.12; urban indicator species richness –0.39, 95% confidence intervals –0.53 to –0.23; urban indicators of conservation concern species richness –0.25, 95% confidence intervals –0.40 to –0.10).

Rural species richness increased towards the south, but urban species richness did not, so southern rural areas had higher species richness (mean 20.3 ± 0.2) than southern urban ones (mean 16.4 ± 0.4). The 84% confidence intervals of these estimates did not overlap, indicating that the average loss of four species from urban areas was statistically significant at P < 0.05 (Payton et al. 2003). Southern Britain was defined as the region south of 52.5° N. Such patterns were not detectable for the number of urban indicator species, or those of conservation concern.

Urban density and the mean density in all rural squares within 15 km of the focal urban cell were positively correlated in just two urban indicator species, Wood Pigeon Columba palumbus (rs = 0.26, P < 0.05) and Great Tit (rs = 0.24, P < 0.05), but these correlations did not remain significant when analyses were restricted to the urban squares with at least five rural squares within 15 km. Urban and rural Jay densities were consistently negatively correlated (all comparisons: rs = –0.22, P < 0.05; restricted data: rs = –0.30, P < 0.05). These results thus provide little evidence that avian densities in urban areas are strongly influenced by the densities in nearby rural areas, or vice versa, but they do provide further evidence that regional factors do not exert a marked influence on urban bird assemblages.

Urban assemblages exhibited strong interspecific abundance occupancy relationships across all species, and urban indicator species exhibited similar patterns. The explanatory power of both of these relationships was higher than their rural equivalents, but the slopes and intercepts of urban and rural patterns were similar (Table 3).

Table 3.  Interspecific abundance–occupancy relationships across 1-km squares for all species and urban indicator species. The proportion of occupied squares was treated as the response variable, and ln(density) as the predictor, with density averaged across occupied squares. Analyses were not conducted for urban indicator species of conservation concern due to the small number of such species. Phylogeny has little influence on abundance–occupancy relationships in British birds (Webb et al. 2007), and we thus did not take it into account. P < 0.0001 in all cases.
 All urban cellsAll rural cells
All speciesn= 58 r2= 52.4%n= 92 r2= 26.3%
intercept =–0.45 ± 0.10intercept =–0.30 ± 0.09
slope = 0.25 ± 0.03slope = 0.17 ± 0.03
Urban indicator speciesn= 27 r2= 51.2%n= 27 r2= 36.2%
intercept =–0.31 ± 0.18intercept =–0.26 ± 0.19
slope = 0.25 ± 0.05slope = 0.21 ± 0.06

Six of the 27 urban indicator species had higher densities in urban BBS squares than in paired rural squares within 15 km (House Sparrow, Common Starling, Common Blackbird, Magpie, Collared Dove and Wood Pigeon), with similar but less strong effects in two additional species (Carrion Crow and Greenfinch; Table 4). Thirteen urban indicator species had significantly lower densities in urban squares than in surrounding rural squares. These included three species of conservation concern, Eurasian House Martin, Song Thrush and Green Woodpecker, the first of which is also an urban specialist (DEFRA 2002, 2003). The remaining species all exhibited a non-significant trend towards higher densities in rural areas (Table 3).

Table 4.  Paired comparisons of urban and rural densities for 27 species identified as urban indicators (DEFRA 2002, 2003); analyses are Wilcoxon signed rank tests as density data are not normally distributed. Presented results use all urban squares with at least one rural square within 15 km. Restricting data to urban cells with five or more rural cells within 15 km does not change the nature of the patterns, but Greenfinch and Carrion Crow no longer show significant differences. †Represents species of conservation concern, and ‡urban specialists (DEFRA 2002, 2003). ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05, ns not significant. Following Nakagawa (2004) Bonferroni corrections have not been applied.
DirectionSpeciesNo. cells with urban > mean rural densityNo. cells with mean rural density > urban densityTiesZ
Urban > rural†‡House Sparrow7910 37.51****
"Common Starling7812 27.47****
"Blackbird7616 07.10****
"Magpie7614 27.05****
"Collared Dove6123 85.93****
"Woodpigeon5735 03.20***
"Greenfinch5537 03.19***
"Carrion Crow5339 02.55*
Rural > urbanChaffinch684 27.95****
"Great Tit1280 07.30****
"Blackcap1081 16.95****
"Wren1973 06.73****
"Song Thrush2267 34.34****
"†‡Eurasian House Martin 333563.88****
"Pied Wagtail1160213.82****
"Sparrowhawk 122693.59****
"Green Woodpecker1358213.38***
"Robin2765 03.24***
"Long-tailed Tit1653232.08*
Not significantDunnock3259 11.40 ns
"Mistle Thrush2345240.46 ns
"†‡Swift 513740.24 ns
"Jay1839350.20 ns
"Goldfinch3054 80.05 ns
"Blue Tit3953 00.93 ns

Urban bird–habitat relationships

For each of the three measures of species richness, a large number of models were retained in the 95% confidence set (Table 5). Explanatory power was low, with model averaged r2 values of 12% for total species richness and the number of urban indicator species, and 4% for the number of urban indicator species of conservation concern (Table 5). Total species richness was positively related to the number of sampling points comprising bushes/trees that were part of an area larger than 2500 m2 (model averaged partial r2 = 5%; Fig. 2a). This positive effect, although small, is supportive of studies demonstrating that patch size influences species richness in urban environments (see above). No other predictors in any model had a partial r2 greater than 3% (Table 4).

Table 5.  Multiple regression models of species richness and density of all species, urban indicator species and urban indicators of conservation concern in relation to habitat type within urban 1-km. All measures of explanatory power are model averaged. Relationships were typically non-linear, and are illustrated in Fig. 2 for all variables with partial r2 greater than 5%.
 No. models in 95% confidence setLargest model weightModel r2 %Partial r2 grassland large %Partial r2 bush/tree large %Partial r2 grassland total %Partial r2 bush/tree total %Partial r2 building %Partial r2 allotment %Partial r2 water%
Species richness – all species420.1612.4%1.84%5.31%0.44% 0.94% 1.24%0.07%2.10%
Species richness – urban indicators500.1111.9%2.30%2.89%0.04% 1.63% 3.54%0.13%0.89%
Species richness – urban indicators of Conservation concern350.12 4.4%1.93%0.03%0.10% 0.15% 1.58%0.01%0.83%
Density – all species170.15 8.6%0.28%0.53%0.09% 5.03% 2.32%0.02%0.54%
Density – urban indicators150.1616.5%1.11%0.58%0.46% 2.79% 5.63%0.13%0.10%
Density – urban indicators of Conservation concern180.4824.9%0.00%0.00%0.04%20.45%10.76%1.59%0.13%
Figure 2.

Relationships between assemblage level metrics of the structure of urban bird assemblages and habitat variables with partial r2 values greater than 5% in multiple regression models (see Table 5). (a) Total species richness and the number of bushes/trees that formed a continuous canopy cover greater than 2500 m2, (b) total density and the total number of bushes/trees, (c) density of urban indicator species and the number of buildings, (d) density of urban indicator species of conservation concern and the total number of bushes/trees, and (e) the number of buildings. Fitted lines indicate model averaged predicted values whilst holding other continuous variables at their mean values, and categorical variables at the most frequent value.

Assemblage level models of species densities also contained a large number of models in the 95% confidence set (Table 5). Explanatory power was low for total bird density (r2 = 8%), most of this variation being explained by the total number of bushes/trees (partial r2 = 5%), which was negatively related to total density (Fig. 2b). Explanatory power was higher for the density of urban indicator species (r2 = 17%), with the number of buildings having the highest partial r2 (6%) and a negative effect on density (Fig. 2c). Explanatory power was higher still (r2 = 25%) for models of the density of urban indicators of conservation concern. The latter was negatively related to both the number of bushes and trees (partial r2 = 20%, Fig. 2d) and the number of buildings (partial r2= 11%, Fig. 2e). These relationships for urban indicators of conservation concern are largely driven by the responses of Common Starling and House Sparrow, which in combination constitute 83% (± 2.5%) of all individuals of species of conservation concern in our urban sites.

Individual species models varied greatly in their explanatory power, from 5% for Chaffinch to 33% for Mallard Anas platyrhynchos (Table 6). Explanatory power tended to be slightly lower for species of conservation concern (mean 12.9 ± 3.8, n = 5) than for other species (mean 16.8 ± 2.1, n = 18), but this difference was not statistically significant (t-test, P > 0.05). Five species of conservation concern occurred in a sufficient number of squares for models of their densities to be constructed (Table 6). Habitat models explained 30% of the variation in House Sparrow density, most of this variation being attributable to the total number of bushes/trees (partial D2 19%), which correlated negatively with density. The number of buildings was also significant (partial D2 10%), with a moderate increase in density from low to intermediate numbers of buildings, and then a decline as the number of buildings increased further (Fig. 3). Habitat models explained 17% of the variation in Common Starling density, most of this explanatory power being attributable to the effect of the total number of bushes/trees (partial D2 13%), which was negative. For the remaining urban indicator species of conservation concern, models had low explanatory power (Dunnock 10%, Song Thrush 9% and Mistle Thrush 7%), and no single predictor in any of these models had a partial D2 greater than 3%.

Table 6.  Multiple regression models of the densities of urban indicator species in relation to habitat type. Models were not constructed for the four species (Green Woodpecker, Swift, House Martin and Sparrowhawk) which occurred in 10 or fewer of the 100 urban BBS squares. All measures of explanatory power are model averaged. †Indicates species of conservation concern, and ‡indicates urban specialists (DEFRA 2002, 2003). Relationships were typically non-linear, and are illustrated in Fig. 3 for all variables with partial D2 greater than 5%.
 No. of models in 95% confidence setLargest model weightModel D2 %Partial D2 grassland large %Partial D2 bush/tree large %Partial D2 grassland total %Partial D2 bush/tree total %Partial D2 building %Partial D2 allotment %Partial D2 water %
Blue Tit30.589.780.000.000.414.920.270.531.58
Carrion Crow90.4213.171.834.350.000.007.751.333.82
Collared Dove40.4219.980.000.004.724.559.970.950.05
Great Tit130.2410.720.000.004.712.441.340.420.00
†‡House Sparrow20.5430.
Mistle Thrush1360.046.750.61< 0.010.461.052.021.340.03
Song Thrush160.358.710.413.< 0.012.17
Figure 3.

Relationships between individual species densities and habitat variables with partial D2 values greater than 5% in multiple regression models (see Table 6). Fitted lines indicate model averaged predicted values whilst holding other continuous variables at their mean values, and categorical variables at the most frequent value. Variation in the fit of the predicted values to the raw data reflects the variation in the explanatory power.

The explanatory power of the density models of the remaining species, i.e. those not of conservation concern, was greater than 10% for seven species (Table 6). In most of these models, only a small number of variables had partial D2 values greater than 5% (Table 6, Fig. 3). One general pattern to emerge from these relationships was that densities tended to be negatively related to the number of buildings, including for the Collared Dove, a species previously identified as an urban specialist on the basis of its occurrence patterns (DEFRA 2002, 2003). The effect of bushes/trees was largely positive (e.g. Jay, Magpie and, somewhat surprisingly, Mallard). The amount of grassland rarely had a large effect on species densities, although it was positively correlated with Magpie density.


Despite an increasing trend of converting British domestic gardens to hard surfaces (Goode 2006, RHS 2007, Perry & Nawaz 2008), gardens contributed about half of the urban green-space. Our estimates that 45% of urban areas constitute green-space and that 50% of this green-space is located in gardens implies that 23% of urban areas are covered by garden green-space; this matches well the estimate that in each of five UK cities, between 22% and 27% of the urban area is covered by domestic gardens (Loram et al. 2007).

Latitudinal gradients in species richness were present in rural areas, but not urban ones. Our results match those of other European studies (Jokimäki et al. 1996, Clergeau et al. 2006, Sorace & Gustin 2008), and support suggestions that urbanization promotes biotic homogenization (McKinney 2006, Devictor et al. 2007). The results also suggested that regional factors that co-vary with latitude to influence species richness, such as climate (Evans et al. 2005), are relatively unimportant in structuring urban bird assemblages in our focal region. The contrasting nature of these latitudinal gradients results in urban squares in southern England supporting, on average, four fewer bird species than rural ones. A direct comparison of suburban species richness with that of adjacent rural squares revealed a similar negative impact of urbanization on species richness (Henderson et al. 2007). These patterns support the finding that in Britain avian species richness in 1-km squares declines at high levels of housing density (Tratalos et al. 2007). They are also compatible with the finding that bird species richness ceases to be positively correlated with human population density at high levels of the latter (Evans et al. 2007). It thus seems likely that any future urbanization in southern Britain, which is projected to receive the bulk of new homes, will result in a reduction in local species richness.

Abundance–occupancy relationships were strong in urban assemblages and similar in their form to those in rural assemblages, confirming the generality of the abundance–occupancy relationship (Gaston et al. 1997, 2000). The slope of abundance–occupancy relationships amongst British farmland and woodland birds has declined markedly over the last few decades, probably in response to environmental degradation (Webb et al. 2007). It appears that this deterioration has been sufficient for some ecological characteristics of the wider countryside to resemble those of urban areas. The strong abundance–occupancy relationship demonstrates that species that occur in few urban sites also occur at low densities. Such species include several of conservation concern (e.g. Common Swift, Eurasian House Martin and Mistle Thrush) and these may be particularly difficult to conserve in urban areas, and will certainly require conservation action to be targeted at precise locations within urban areas.

A different suite of management action may generally be required for each species of conservation concern as their densities were typically not correlated with each other, suggesting that each species is limited by a separate suite of factors. However, Common Starling and House Sparrow densities were strongly positively correlated with each other. Although these species differ markedly in some aspects of their ecology, the former is largely granivorous and the latter is a probing insectivore, they both take supplementary food and frequently nest in holes in buildings. The strength of the correlation between their densities may thus indicate that their urban populations are limited by the availability of these shared resources.

Contrary to expectation, the majority of species identified as urban indicators, and some urban specialists, did not have significantly higher densities in urban areas than rural ones. Whilst urban indicator species were identified on the basis of the spatial distribution along BBS routes, irrespective of spatial patterns in their density (DEFRA 2002, 2003), density and occupancy are strongly correlated and the choice of metric for defining urban indicators is perhaps unlikely to influence markedly which species are selected. Differences between our focal spatial grain and that used to identify urban indicators are probably more important. The analysis of urban indicators (DEFRA 2002, 2003) considered an individual to occur in urban areas if it was detected on a 200-m transect section classified as either urban or sub-urban, irrespective of the surrounding habitat type. In contrast, we defined urban areas as those in which the entire BBS transect route was classified as urban or suburban. Our results strongly indicate that the majority of urban indicator species of conservation concern do not have significantly higher densities in large urban areas than in rural ones. The rural environment thus cannot be neglected in attempts to conserve such species.

The explanatory power of the individual bird–habitat models was low for some species, but certainly not all. Overall, it was equivalent to that found in a study of bird–habitat relationships in rural Britain at a similar spatial scale (Fuller et al. 1997). Moreover, habitat type explained little of the variation in urban bird densities in Sheffield, UK (R.A. Fuller unpubl. data) or Phoenix, AR, USA (Hostetler & Knowles-Yanez 2003) and the magnitude of explanatory power may be general. Whilst our analyses use a relatively large spatial grain, it is unlikely that stronger relationships would have emerged if finer spatial scales had been used, as the Sheffield and Phoenix studies were conducted across 250-m squares and 200-m transect sections, respectively. In addition, an analysis of House Sparrow density in 250-m squares in relation to habitat type (Wilkinson 2006) did not have greater explanatory power than our House Sparrow model. Higher explanatory power may have arisen from a finer definition of habitat types, and inclusion of other measures of resource availability, such as bird feeder density, which can have a strong relationship with urban bird densities (Jokimäki et al. 2002, Fuller et al. 2008). Alternatively, habitat type may explain little variation in bird densities because some species are increasing in urban areas, such as Wood Pigeon, Long-tailed Tit and Goldfinch Carduelis carduelis (Cannon et al. 2005), and thus their populations may not be in equilibrium with habitat availability.

We found no evidence that the densities of passerines that are vulnerable to nest-predating corvids correlate negatively with their density. In some urban areas, the risk of artificial nest predation increases with corvid density (Jokimäki et al. 2005), but such studies may not accurately mimic predation risk of real nests. Jokimäki and Huhta (2000) found that ground-nesting bird species occurred at lower abundances in urban areas with less dense vegetation, where avian predation of artificial nests was high, but it is unclear if changes in abundance were due to direct effects of predation rather than more indirect effects of vegetation structure. In contrast, removal experiments and other studies in urban areas suggested that nest predation by corvids has limited impact on the population size of their avian prey species (Wysocki 2005, Chiron & Julliard 2007, Marzluff et al. 2007), thus concurring with our results and with studies conducted in more rural areas (Gooch et al. 1991).

One of the more consistent patterns in our data was that high numbers of buildings have a negative influence on the densities of many urban indicator species, including the urban specialists Collared Dove and House Sparrow. Intraspecific variation in the nature of these relationships was marked, but declines invariably started when no more than 25% of the sampling points were buildings (Fig. 3). The decline in avian abundance in areas of high density housing has previously been highlighted, and such declines typically start well before housing density reaches that at which the UK government requires new developments to be built (Tratalos et al. 2007).


Although more research is required, especially of wintering assemblages, a number of general trends concerning the influence of habitat type and supplementary resource provision on urban birds are apparent. The structure of urban avian assemblages is more strongly influenced by local factors than ones operating at larger spatial scales, thus yielding the potential for management of urban areas to positively influence their bird assemblages. Species richness is strongly influenced by fragmentation. Increasing patch size, and to a lesser extent reducing isolation, are likely to promote higher species richness in urban environments, but will be difficult to achieve in existing urban areas. Greater habitat diversity, increased structural complexity of vegetation, and higher woody plant species richness can also promote higher avian species richness. However, assuming that additional areas of green-space do not become available in urban environments, increasing habitat diversity will inevitably lead to a reduction in the extent of one or more existing habitat types. The level to which habitat diversity should be increased will thus depend on the relative conservation value of existing habitat types and those intended to replace them.

It has been suggested that supplementary feeding can have negative effects on urban birds, but their densities are often positively associated with supplementary food provision. More research is needed to determine the mechanisms underlying this pattern, but the available evidence suggests that supplementary feeding should be encouraged. The biodiversity and educational gain that could be generated from increasing urban bird feeding in economically deprived areas where it is currently rare, should be considered (Fuller et al. 2008). There is some evidence that human disturbance can reduce breeding bird densities in built-up areas, even of species that appear well adapted to urban environments. Such disturbance could be reduced by the redesign of paths, and increased provision of habitat cover. Finally, high-density housing developments, such as those currently recommended by the UK government, are likely to result in reduced numbers of many urban bird species than could be achieved if developments were constructed at lower densities. Urbanization of currently rural areas can also result in reduced species richness; in the UK this seems particularly likely in the south.

In the UK, few specific management actions can be advocated at this stage for individual species in urban areas. Indeed, our understanding of the factors limiting the densities of urban bird populations falls far short of that for farmland and woodland species (Newton 2004, Fuller et al. 2007, Gill & Fuller 2007). A detailed research programme on the needs of urban indicator species of conservation concern, akin to those conducted in these other habitats, which emphasizes the mechanisms driving urban bird–habitat relationships, is thus urgently needed. It is encouraging to note that this process has recently started for the House Sparrow (Vincent 2006, Wilkinson 2006, Chamberlain et al. 2007b, Shaw et al. 2008).

Conservationists working in urban areas have a range of potential management objectives to aim for. One extreme view would be to focus on maximizing the population densities of the relatively small number of species that do relatively well in urban areas. Alternatively, one may seek to maximize the number of species found in urban areas. There has been little formal discussion of where the most efficient or effective conservation strategies lie along this continuum of management objectives. However, in Canada, it is recognized that whilst a number of declining forest bird species are present in urban areas, maintaining long-term viable populations of these species in urban areas is not possible given the current size of habitat patches (Environment Canada 2007). Similarly, we demonstrate that in Britain, relatively few urban indicator species occur at higher densities in highly urbanized areas than more rural ones, and of the species of conservation concern, only House Sparrow and Common Starling do so. In many urban areas, it may thus be more appropriate to focus management on such species that appear particularly well adapted to urban environments.

We are grateful to the thousands of volunteers who have taken part in the bird surveys which have provided the data on which these analyses are based. The Breeding Bird Survey is jointly funded by the British Trust for Ornithology, the Joint Nature Conservation Committee (on behalf of the Countryside Council for Wales, English Nature, Scottish Natural Heritage and the Department of the Environment for Northern Ireland) and the RSPB. D. Chamberlain, L. Marini and J. Tratalos provided assistance. K.L.E. was supported by the Natural Environment Research Council. K.J.G. holds a Royal Society-Wolfson Research Merit Award.