Are urban bird communities influenced by the bird diversity of adjacent landscapes?

Summary

• 1
  The species diversity of adjacent landscapes influences the conservation or restoration of several animal groups in urban areas, but the effect on birds is unclear. To address this question, we compared bird species richness (BSR) and community composition between periurban (area surrounding the town) and urban (suburban and centre areas) landscapes across three spatial scales.
• 2
  At a large biogeographical scale (temperate and boreal climatic zone), relationships between the BSR of urban areas and their surrounding landscapes were examined in a meta-analysis of 18 published studies. In general, BSR was negatively correlated with latitude and urbanization. The BSR of suburban and centre landscapes correlated positively with the BSR of periurban landscapes. However, latitudinal effects were also involved, as BSR in urban and periurban landscapes declined as town latitude increased. Similarity indices were low (50%) between periurban and centre bird communities.
• 3
  At a regional scale, we assessed winter bird data from several towns within three regions of temperate and boreal countries (western France, northern Finland and eastern Canada). The type of periurban landscape, number of inhabitants and town diameter did not affect BSR. BSR was similar between the cities of a given biogeographical area. Bird communities were more similar between similar habitat types of different cities than between different habitats of the same city.
• 4
  At a local scale, we tested the influence of proximity to the periurban landscape on BSR in parks of western French towns of different size. Neither BSR nor community similarity changed in relation to the distance of the park from the periurban landscape.
• 5
  Guild composition according to diet and feeding habitat did not vary between urban and periurban locations at regional or local scales.
• 6
  We conclude that, at regional and local scales, urban bird communities are independent of the bird diversity of adjacent landscapes, and that local features are more important than surrounding landscapes in determining BSR. Whatever the biodiversity quality of the periurban landscape, site-specific actions such as shrub and tree planting, water restoration and increasing vegetation diversity can change bird diversity in towns and improve the quality of human–wildlife contacts.

Introduction

Processes associated with urbanization are one of the major causes of landscape change and represent an important threat to biodiversity (Wilcox & Murphy 1985). Urban planners need better information about the factors affecting the distribution of species and structure of communities in order to create or maintain biodiversity in urban areas. Conservation or restoration efforts related to urban wildlife focus on limiting artificial habitat, developing citizen participation in wildlife conservation, improving the quality of life of urban dwellers, and educating them about ecological concepts (Gilbert 1989; Adams 1994; Niemelä 1999). The interest of urban residents for their immediate environment is becoming increasingly important. Additionally, their decisions on ecological issues at regional or national levels are based largely on their perception of nature at local levels (Michelson 1970; Middleton 1994). Therefore, urban wildlife diversity can influence significantly the management of biodiversity at regional, national and even global scales (Owen 1978; Hadidian et al. 1997).

In general, human activities have produced similar ecological structures in urban areas even in different biogeographical regions. The response of birds to these environmental changes could lead to the dominance of bird communities by a few very abundant species (Bezzel 1985). This, in turn, might lead to the general hypothesis that urbanization causes uniform bird communities in urban areas (Jokimäki et al. 1996). A town can be an original ecosystem with its own characteristics and species (Davis & Glick 1978; Bezzel 1985).

A few ecological studies have been conducted on the relationships between urban (artificial built-up area) and periurban (area adjacent to the town) landscapes and on the conservation of biodiversity in urban ecosystems. Data on dispersion, colonization and settlement of several wildlife groups such as mammals (Harris 1977; Dickman & Doncaster 1989), insects (Owen 1978) and lizards (Germaine 1995) have clearly underlined the effect of proximity to non-urban source populations in urban biological conservation. Could this relationship hold also for birds in urban landscapes, i.e. could periurban diversity influence bird species richness (BSR) in towns?

The composition of urban flora and fauna may be determined by the biotic and abiotic factors associated with the available species pool in the region (Böhning-Gaese 1997; Roy, Hill & Rothery 1999). Surrounding habitat features may affect bird community structures (Andrén 1994; Kubes & Fuchs 1998; Jokimäki 1999), and bird diversity in urban areas has been linked to periurban landscapes (Siegfried 1968; Jones 1983; Munyenyembe, Harris & Hone 1989). However, Erz (1966) suggested that bird species do not colonize newly urbanized areas from surrounding countryside, but immigrate from already urbanized populations. Thompson, Greenwood & Greenway (1993) compared garden birds between several European countries and found that BSR in urban, suburban and rural gardens was similar in countries of northern Europe, but not in western or southern Europe.

These opposite results might depend on the geographical location or size of the town, but also on the spatial scale used in these studies. The landscape and community structures are scale-dependent and different factors are involved at each ecological organization level (Forman & Godron 1986; O’Neill 1989; Böhning-Gaese 1997; Baillie et al. 2000).

Our main goal in this study was to evaluate the relative importance of local and regional landscapes in shaping the structure of urban bird communities. An effect of periurban diversity on urban BSR would stress the need to incorporate landscape consideration in managing communities of urban birds, whereas a lack of strong relationships would emphasize local actions. We used a multi-scale approach to investigate the relationship on bird diversity between urban and periurban landscapes. At a large biogeographical scale, we compared bird community indices of bird studies from towns in temperate and boreal climatic zones. At a regional scale, we assessed the level of variability in the avifauna of several cities within similar geographical areas. At a local scale, we compared the avifauna of different habitats within a single town. As life histories of species also constrain the spatial scales at which resources can be exploited (Hansen & Urban 1992), we analysed our data in more detail by using groups of species (guilds) in our regional and local scale analyses.

Methods

Large biogeographical scale data

To look at relationships between urban areas and their surrounding landscape, we reviewed studies where breeding BSR had been estimated similarly in both urban and periurban landscapes. We excluded some studies where BSR was estimated by numerous observers (Luniak 1990; Konstantinov 1996). We limited our sample to studies conducted in temperate and boreal regions. From about 50 papers on urban birds, we found 18 studies that fulfilled our pre-selection criteria.

In all these studies, description of study sites allowed the coherent definition of three landscape types representative of typical forms of urban development. We characterized the urban area by two landscape types that we called centre and suburb for simplification. (i) The centre includes both the commercial downtown area, i.e. historic centre of the city with less than 15% of vegetated area and buildings with three to five stories, and the residential sector adjacent to the commercial area, where buildings have in general one to three stories and vegetated open areas cover between 20% and 40% of the area. (ii) The suburb consists of residential areas with single-family detached houses, large apartment building complexes, parks and cemeteries, and of large office or supermarket sites with lawns and parking areas. The height and disposition of houses and buildings vary considerably and vegetated areas cover up to 70% of this sector. (iii) The third landscape, the periurban sector, includes the various landscapes adjacent to the town, up to 10 km from the city centre for large cities. It mainly comprises managed rural features such as crops and cattle farming, but also includes some leisure sites such as golf courses and parks, and ‘natural’ sites such as woodlots and lakes. In some cases, towns are surrounded mainly by natural landscapes like forest or heath.

For all selected studies (Table 1), we classified surveyed sites into these three landscape types. For periurban landscapes, we extracted separately, based on the previous remarks, data for managed (dominated by agriculture or leisure sites) and natural (dominated by forest or heath) landscapes when possible. In some studies, we selected only one year of results because other years did not involve the same study sites. Two main types of census were represented: absolute methods (spot-mapping or strip census), used in the majority of publications, and relative methods (fixed point counts on transect lines or not), used in only three papers.

Table 1.  Selected studies in temperate and boreal regions where bird surveys were conducted in urban and periurban areas with the same method and in the same season. Managed landscape = areas where agriculture or leisure sites predominate; Natural landscape = areas where forest or heath predominate. *Absolute method (mapping method, strip census); †relative method (plot method, fixed point method on transect line) (reviewed in Bibby, Burgess & Hill 1992)

Authors
	City (country)	Size agglomerate ×1000 inhab.	Latitude world part‡	Method census/site sample size	Centre data (n sites)	Suburban data (n sites)	Periurban data (n sites)	Centre/ suburban	Suburban/ periurban	Centre/ periurban
‡ World part = old world, OW; new world, NW. § Study plots were not standardized and we have applied rarefaction diversity method (basis of 30 pairs). ML, managed landscape; NL, natural landscape.
Weber (1972)	Vancouver (Canada)	1380	49 – NW	2 years*	–	14 (2)	27·5 (2)	–	0·56	–
				16 census			27·5 (2) ML
				10 ha
Geis (1974)	Columbia (USA)	395	34 – NW	1 year (1972)*	5·1 (7)	11·1 (5)	16·1 (4)	0·68	0·69	0·58
				1 census			16·1 (4) ML
				8–97 ha§
Campbell & Dagg (1976)	Waterloo (USA)	80	43 – NW	1 year*	13 (1)	25 (2)	31 (1)	0·43	0·71	0·36
				7 census			31 (1) NL
				25 ha
Marchetti (1976)	Marseille (France)	900	43 – OW	1 year†	9 (2)	28 (2)	37·3 (3)	0·52	0·78	0·40
				1 census			36·5 (2) ML
				50 points			39 (1) NL
Huhtalo & Jarvinen (1977)	Tornio (Finland)	20	66 – OW	1 year*	10·4 (1)	10·8 (1)	9·4 (1)	0·63	0·63	0·57
				4 census			9·4 (1) ML
				30–95 ha§
Savard (1978)	Toronto (Canada)	3425	44 – NW	1 year*	7·5 (6)	15 (2)	12·2 (5)	0·63	0·75	0·58
				5 census			8·8 (4) ML
				10 ha			26·1 NL
Hohtola (1978)	Kuopio (Finland)	80	63 – OW	1 year*	5 (3)	13·6 (7)	11·1 (1)	0·32	0·58	0·21
				3 census			11·1 (1)ML
				2–17 ha§
Lancaster & Rees (1979)	Vancouver (Canada)	1380	49 – NW	1 year*	6 (1)	18·8 (4)	23 (1)	–	–	–
				> 5 census			23 (1) NL
				8 ha
Tatibouet (1981)	Lyon (France)	455	46 – OW	1 year†	15 (2)	21·5 (2)	31 (3)	0·78	0·85	0·62
				1 census			30·5 (2) ML
				7 points			32 (1) NL
Bessinger & Osborn (1982)	Oxford (USA)	115	52 – NW	1 year*	–	18·5 (2)	27·5 (2)	–	0·56	–
				> 10 census			27·5 (2) NL
				6 ha
Jokimaki (1989)	Rovanieni (Finland)	35	67 – OW	1 year*	5 (1)	10·5 (4)	11·9 (10)	0·44	0·64	0·24
				1 census			11·9 (10) ML
				30 ha
Vauhkonen (1990)	Heinola (Finland)	15	67 – OW	1 year (1990)*	9·9 (1)	12·2 (1)	16·4 (1)	0·67	0·83	0·58
				2 census			16·4 (1) NL
				50–420 ha§
Sasvari (1990)	Budapest (Hungary)	2100	47 – OW	5 years*	–	24 (3)	24·9 (7)	–	–	–
				4 census			23·3 (3) ML	–	–	–
				50 ha			26 (4) NL
Sodhi (1992)	Saskatoon (Canada)	200	52 – NW	3 years*	–	16 (6)	15·7 (6)	–	0·47	–
				6 census			15·7 (6) ML
				9 ha
Rauhala (1994)	Kemi (Finland)	25	66 – OW	1 year	10·6 (1)	9·6 (1)	16·3 (2)	0·65	0·67	0·77
				1 census			16·3 (2) ML
				13–73 ha§
Blair (1996)	Palo Alto (USA)	55	37 – NW	2 years†	7 (1)	14·5 (2)	23·7 (3)	0·30	0·59	0·23
				12 census			25 (2) ML
				16 points			21 (1) NL
Clergeau et al. (1998)	Rennes (France)	210	48 – OW	1 year*	20·3 (3)	24 (3)	27 (2)	0·82	0·81	0·79
				4 census			27 (2) ML
				12 ha
Clergeau et al. (1998)	Quebec (Canada)	605	47 – NW	1 year*	19 (3)	24·3 (4)	25 (2)	0·67	0·64	0·57
				4 census			25 (2) NL
				12 ha

Regional scale data

We analysed winter bird data from the towns of three regions: (i) western France (between 47° and 49° north), with a temperate climate and numerous cities surrounded by agricultural landscapes (P. Clergeau & G. Mennechez, unpublished data); (ii) northern Finland, which belongs to the northern coniferous forest biome (between 65° and 67° north), with a cold climate and small cities surrounded mainly by forests (J. Jokimäki & M. Kaisanlahti, unpublished data); and (iii) eastern Canada (between 45° and 47° north), with a cold climate and big cities surrounded by a large river and forests (Clergeau et al. 1998). We compared BSR and similarity indices (see below) between different types of (i) suburban development in Europe and (ii) urban parks in Canada. Because seasonal differences may modify community composition and structure, our regional scale results from the winter period may not be directly comparable with large and local biogeographical scale results collected during the breeding period.

We compiled data in winter (January and February 1999) on birds from five cities in a small area (40 000 km²) of the eastern part of Brittany (western France) and Lapland (northern Finland). These towns differed in size and types of periurban landscapes (Table 2). In each city, we identified two traditional suburban landscapes typical in their structure: blocks of large apartment buildings and areas of single-family detached houses. All sites were approximately 30 ha in size and were surveyed using a single visit, i.e. a zigzag walk through the site taking 60 min (Jokimäki et al. 1996). The mapping method of Svensson (1974) used in these studies reduced many problems associated with counting birds in urban areas (DeGraaf, Geis & Healy 1991). The validity of this quick technique has been confirmed (Jokimäki & Suhonen 1998). All birds were counted, except individuals flying over that did not stay within the study site.

Table 2.  Principal characteristics of towns studied in France, Finland and Canada

City
	Inhabitants	Diameter (km)	Major component	Secondary component
* Hedgerow network landscape with grasslands and some farming.
Redon	11 000	2	Bocage*	Wetland
Dinan	16 000	2	Bocage	Woodland
Fougères	28 000	3	Open field	Bocage
St Malo	46 000	4	Littoral	Open field
Rennes	200 000	7	Open field	Bocage
Angers	140 000	4·4	Bocage	Woodland
Nantes	260 000	12	Open field	Bocage
Kemijärvi	10 000	2·5	Forest	Lake side
Tornio	20 000	2·5	Littoral	Farming
Kemi	25 000	5	Littoral	Farming
Rovaniemi	35 000	5	Forest	River side
Oulu	105 000	8·5	Littoral	Farming
St Nicolas	7500	1·5	Farming	Forest
Levis	40 000	2·5	Farming	Forest
Quebec	600 000	15	Forest	Farming
Montreal	1010 000	30	Farming	Forest

In addition, we compiled data on wintering birds from the parks of four cities in the St Lawrence Valley, southern Quebec (eastern Canada) (Table 2). We selected data from two types of parks according to their structure (Morneau et al. 1996): managed parks with some trees or wooded portions (< 20%) and large lawns (more than 50%), and more natural parks with forest or wetland biotopes and with lawn area < 50%. To avoid size effects (Opdam, Rijsdijk & Hustings 1985), as parks varied from 0·3 ha to 117 ha (Morneau et al. 1996), we selected only parks between 4 and 10 ha in size.

Local scale data

We tested the neighbourhood effect of periurban landscapes using a similar habitat type located along a distance gradient from periurban areas. In three cities of different sizes in western France, Angers, Rennes and Nantes (Table 2), we chose one park in the town centre, one in the suburbs and one at the edge of town: the distance between town centre parks and potential species sources (periurban area) increased with the size of the town. We selected 4–10-ha parks to avoid size effects; in the large parks of town edges, we chose only an isolated part of about 10 ha. All selected sites presented similar structures, with old trees, lawns and ponds (Clergeau 2000). BSR was determined during the breeding season by the point-count method (Bibby, Burgess & Hill 1992). Two survey stations were located in each park, and all birds seen or heard were recorded during a 20-min count between 07:00 hours and 09:00 hours. Each station was surveyed twice per breeding season (May and June 1998 and 1999). A more detailed description of the study sites is available from Clergeau (2000).

Data analysis

We used the cumulative number of bird species obtained in each landscape type (centre, suburban and periurban areas) as well as the mean number of species per landscape type. The two methods gave similar results, so we present here only the mean number of species. In cases where the size of study plots differed, we used the rarefaction method to estimate species richness (Heck, van Belle & Simberloff 1975). Biogeographical data were first analysed by the general linear model (GLM) univariate procedure (SPPS 1999). BSR was the dependent variable and fixed factors included the number of inhabitants (three groups: > 600 000, n = 16; < 600 000–60 000, n = 19; < 60 000, n = 15), latitude (two groups: > 50°N, n = 19; others, n = 31), urbanization level (three groups: centre, n = 14; suburban, n = 18; periurban, n = 18) and location in the old or new world (two groups). The first model was constructed by using main effects of all independent variables and their interactions. After the second phase, only significant variables (P < 0·05) were entered in the model.

We used the Sörensen similarity index (Jongman, ter Braak & Tongeren 1995) to measure the similarity of two assemblages. It is based on species numbers and does not take species abundance into account:

• similarity index = 2c/(a + b)

where c is the number of species shared by the two sites, and a and b the total number of species at each site. Values of this index vary from 0 to 1; 0 indicates that assemblages differ totally, and 1 that they are identical.

To analyse potential ecological variations, we compared biological traits of bird species. First, we defined a diet guild from the main diet of each species: insect feeder; seed and vegetation feeder; omnivore; carnivore (Clergeau et al. 1998). Secondly, we defined each species according to its main feeding habitat: aquatic; tree dwelling; shrubs; meadows; other. Classification of each species observed in France, Finland and Canada is presented in the Appendix. Data were analysed by comparing the number of species at regional and local scales for each guild and each town using chi-squares.

Non-parametric tests were performed using correction for ties (two-tailed). All mean values are given with standard errors (95% confidence limits).

Results

Relationships at a large biogeographical scale

According to the GLM univariate model using all independent variables and their interactions, only the level of urbanization (F = 14·85, d.f. = 2, 27, P < 0·001) and latitude (F = 6·55, d.f. = 1, 27, P = 0·016) significantly affected BSR. None of the interaction terms was significant (P > 0·05). When we dropped non-significant variables from the model, both urbanization level (F = 13·48, d.f. = 2, 46, P < 0·001) and latitude (F = 16·51, d.f. = 1, 46, P < 0·001) were significant, explaining together 55·5% of the variation in BSR.

BSR of periurban landscapes was positively correlated with the BSR of suburban (Spearman r_s = 0·75, n = 18, P = 0·002) and town centre landscapes (r_s = 0·58, n = 14, P = 0·04), but the correlation between BSR of suburban and centre landscapes was not significant (r_s = 0·46, n = 14, P = 0·10). Relationships between suburban and periurban BSR according to types of periurban landscapes (managed or natural) corroborated the correlation: significant regression coefficients and similar slopes were obtained with managed and natural landscapes (Fig. 1). BSR decreased with latitude both in suburban and periurban landscapes, but not in town centres (Fig. 2). In northern Europe, the number of species in each landscape (centre: 8·2 ± 5·7; suburban: 11·3 ± 3·1; periurban: 13·0 ± 6·2) was half that in southern Europe (14·8 ± 11·1, 24·4 ± 5·3, 30·1 ± 10·7, respectively).

Figure 1. Bird species richness (BSR) in suburban landscape is correlated with BSR observed in natural (y = 5·12 + 0·57x, r = 0·66, t = 4·20, P = 0·002) or in managed (y = 5·53 + 0·54x, r = 0·58, t = 3·72, P = 0·004) periurban landscapes (area adjacent to the town). Data from 18 selected studies in temperate and boreal regions.

Figure 2. Bird species richness (BSR) both in periurban and in suburban landscapes are correlated with latitude, but BSR in town centre is independent of latitude (Spearman r_s).

BSR decreased from periurban (21·5 ± 15·4) and suburban landscapes (16·9 ± 11·4) to centre landscapes (10·2 ± 9·5) (n = 14, 14, 18 pairs, all Z < −2·92, P < 0·003). In four cases in northern countries (Finland and Canada), BSR did not follow this trend, being greater in suburban than periurban landscapes (Table 1).

Highest similarity values were observed between periurban and suburban landscapes (0·67 ± 0·21) and lowest between periurban and centre landscapes (0·50 ± 0·38); similarity values between suburban and centre landscapes (0·58 ± 0·33) were intermediate. Bird communities were more similar between periurban and suburban landscapes than between periurban and centre landscapes (Mann–Whitney U = 45·0, n = 26, P = 0·01). Similarity between centre and suburban landscapes did not differ from others (U = 72·5, n = 26, P = 0·17; U = 57·0, n = 26, P = 0·16, respectively). Similarity indices between communities were not influenced by latitude (n = 13, 13, 16, respectively, all r_s < −0·2 and P > 0·4).

Relationships at a regional scale

In Brittany, similar wintering BSR were found in blocks of apartment building (19·20 ± 3·80) and single-house areas (18·80 ± 3·13; Z = −0·67, n = 10, P = 0·50). BSR did not vary between cities (Table 3) nor was influenced by the number of inhabitants (below and above 26 000 inhabitants), town diameter (below and above 3·5 km) or the type of periurban landscape (bocage/others) (Table 2) (Mann–Whitney, all n = 10, P > 0·24).

Table 3.  Number of wintering bird species in suburban landscapes (two housing types in Europe; two park structures in Canada) of towns varying in terms of size and periurban landscapes (see Table 2). Chi-squares on matrix towns/landscape types per region did not show significant variation according to towns

	Chi-square	d.f.	P
Housing type	Redon	Dinan	Fougères	St Malo	Rennes
Block of apartment buildings	17	22	18	18	21
Single-family detached houses	21	19	17	17	20	0·70	4	0·95
Housing type	Kemijärvi	Tornio	Kemi	Rovaniemi	Oulu
Block of apartment buildings	4	7	7	4	8
Single-family detached houses	9	9	11	9	11	0·94	4	0·92
Park type	St Nicolas	Levis	Quebec	Montreal
With lawn > 50%	9	4	4·5	4
With lawn < 50%	6	8	5	10		2·76	3	0·43

In Lapland, BSR was lower in areas with blocks of apartment building (6·00 ± 3·28) than in single-house areas (9·80 ± 1·92; Z = −2·03, n = 10, P = 0·04; Table 3). BSR did not vary between cities, nor was influenced by the number of inhabitants, town dia-meter or the type of periurban landscape (forest/littoral) (Mann–Whitney, all n = 10, P > 0·07). In pooled data, BSR was lower in Lapland than in Brittany (U = 0·001, n = 20, P = 0·008).

In Quebec, BSR did not differ between parks with lawn occupying > 50% (5·50 ± 4·04) and parks with lawn occupying < 50% (8·05 ± 1·99; Z = −1·09, n = 8, P = 0·27) and BSR did not vary between cities (Table 3). Also in this region, neither the size of town nor its periurban landscape could explain BSR variations (Mann–Whitney, all n = 8, P > 0·22).

In Brittany, bird community similarity between blocks of apartment building and single-house areas within the same town was about the same as that between similar habitats in different towns (Mann–Whitney, n = 15, 15, 25, all P > 0·27; Table 4). In Lapland, within-town similarity in bird communities was lower than that between bird communities of single-house areas of different towns (U = 0·001, n = 15, P = 0·002) and similar to that between areas of blocks of apartment buildings of different towns (U = 21·5, n = 15, P = 0·67; Table 4). Bird communities of single-house landscapes were more similar than those of blocks (U = 20·5, n = 20, P = 0·02). In Quebec, park similarity indices did not differ between different towns and between different parks with lawn < 50% or with lawn > 50% (n = 10, 10, 12, all P > 0·91; Table 4). Therefore, in the three regions, similarity indices were not significantly greater within town than within habitat types. In pairwise comparisons, similarity indices did not vary with size of towns (number of inhabitants, town diameter) or types of periurban landscapes (n total = 16, 25, 25, all P > 0·08).

Table 4.  Similarity indices within towns and between towns for single-family detached house/blocks of apartment building areas (Brittany and Lapland) and between towns for parks more or less managed (Quebec)

	Mean ± SE
	Redon	Dinan	Fougères	St Malo	Rennes
Redon	0·68	0·75	0·84	0·68	0·68
Dinan	0·72	0·59	0·78	0·72	0·72	Single-house: 0·75 ± 0·11
Fougères	0·74	0·70	0·80	0·77	0·81
St Malo	0·63	0·75	0·78	0·80	0·70
Rennes	0·63	0·70	0·62	0·67	0·78	Within town: 0·73 ± 0·16
						Blocks: 0·69 ± 0·11
	Kemijärvi	Tornio	Kemi	Rovaniemi	Oulu
Kemijärvi	0·62	0·78	0·80	1·00	0·80
Tornio	0·73	0·75	0·90	0·78	0·80	Single-house: 0·84 ± 0·15
Kemi	0·73	0·86	0·78	0·80	0·91
Rovaniemi	0·75	0·73	0·55	0·46	0·80
Oulu	0·67	0·80	0·93	0·50	0·74	Within town: 0·67 ± 0·12
						Blocks: 0·72 ± 0·26
	St Nicolas	Levis	Quebec	Montreal
St Nicolas	0·24	0·13	0·12	0·25
Levis	0·50	0·40	0·67	0·57		Park lawn > 50%: 0·40 ± 0·51
Quebec	0·57	0·43	0·57	0·67
Montreal	0·40	0·32	0·33	0·50		Within town: 0·43 ± 0·29
						Park lawn < 50%: 0·43 ± 0·19

The BSR of different diet and feeding habitat guilds did not vary between different towns in Brittany or Lapland (chi-square test, d.f. = 8, 12, all P > 0·79). In Brittany, the number of species per category did not vary with location (Wilcoxon test, all n = 10, P > 0·18) but, in Lapland, seed and vegetation feeders (species noted g) were more abundant in single-house areas (3·40 ± 2·14) than in blocks of apartment buildings (1·20 ± 2·55; Z = −2·02, n = 10, P = 0·04). In the feeding habitat guild, tree-dwelling species were more numerous in Lapland single-house areas (5·20 ± 1·63) than in areas of blocks of apartment buildings (2·40 ± 2·97; Z = −2·02, n = 10, P = 0·04). No other differences were observed (P > 0·05). The Canadian study did not reveal any variation between sites (Wilcoxon test, diet guild: all n = 8, P > 0·07; feeding habitat guild: all n = 8, P > 0·08) and corroborated a lack of difference between towns (chi-square, all d.f. = 6, P > 0·42).

Relationships at a local scale

Using French towns of different sizes, we tested whether BSR in urban parks was related to the distance of the park from the periurban landscape, i.e. the distance that birds would have to travel across the urban matrix. BSR varied between years (more species in 1998, U = 14·50, n = 18, P = 0·02) but BSR of parks located in the centre landscape (1998: 26·33 ± 2·06; 1999: 23·67 ± 2·99), suburban landscape (1998: 25·33 ± 1·13; 1999: 21·33 ± 5·99) or just at the edge of town (1998: 25·67 ± 1·47; 1999: 23·00 ± 5·88) did not vary with the diameter of the town (1998: χ² = 1·35, d.f. = 4, P = 0·86; 1999: χ² = 1·05, d.f. = 4, P = 0·90). Thus the distance from the periurban landscape did not affect park BSR (Fig. 3).

Figure 3. Bird species richness (breeding species) count in similar parks (4–10 ha parks with old trees, lawns and ponds) of three towns of western France (km = distance between the town edge and the town centre). 1998, grey bars; 1999, dark bars. All chi-square tests not significant.

Similarity indices between the different habitats of a town did not differ from those between parks in town centres or between parks just at the edge of different towns (all U > 0·12, n = 6, P > 0·34; Table 5). Similarity indices did not vary between years (U = 26·50, n = 18, P = 0·22). BSR of diet and habitat guilds did not vary between park locations in 1998 (χ², all d.f. = 4, P > 0·63 and all d.f. = 8, P > 0·91, respectively) and in 1999 (all d.f. = 4, P > 0·73 and all d.f. = 8, P > 0·84, respectively).

Table 5.  Similarity indices within French towns and between towns for park in centre/park just in fringe; in first line 1998, in second line 1999. n km = distance between park in centre and park just in fringe

				Mean ± SE
	Angers	Rennes	Nantes
Angers (2·2 km)	0·61	0·79	0·71	Park just in fringe:	0·73 ± 0·09
	0·68	0·69	0·70		0·72 ± 0·07
Rennes (3·5 km)	0·69	0·75	0·68
	0·65	0·76	0·78
Nantes (6 km)	0·69	0·63	0·74	Within town:	0·70 ± 0·13
	0·74	0·82	0·79		0·74 ± 0·09
				Park in centre:	0·67 ± 0·06
					0·74 ± 0·13

At a local scale, distance from the periurban landscape did not affect park BSR, community similarity or guild composition.

Discussion

Our results indicated that BSR in urban areas was independent of the adjacent landscapes of study towns across different spatial scales. In addition, the low similarity values observed between periurban and urban bird communities supported this conclusion. Using several studies from the northern hemisphere, we found positive correlations in BSR between suburban and periurban landscapes. However, this was mostly due to a latitudinal effect on BSR rather than a relationship between birds of urban and periurban landscapes; BSR in urban and periurban landscapes increases or decreases together with town latitude. These results are coherent with classical decreasing trends in BSR from south to north, e.g. from 273 regular breeding bird species in France to 235 in Finland (Yeatman-Berthelot & Jarry 1994; Väisänen, Lammi & Koskimies 1998). This latitudinal relationship did not always persist at the regional scale of the country. Highly urban areas tended to have similar BSR independent of their geographical location, as obtained by Jokimäki et al. (1996). Similarity indices obtained from our selected studies indicate that a large part of the bird community is similar throughout the urbanization gradient, but also that in a town centre about 50% of the avifauna is independent of periurban species (mean similarity index between centre and periurban = 0·50). Interestingly in Finland and Canada, some BSR was higher in suburban than in periurban landscapes. This indicates that intermediate level disturbances caused by urbanization (Blair 1996), and also food supplementation given by humans in some habitats (see our Finland results on seed-eater functional group; Jokimäki & Suhonen 1998), might be beneficial for birds in northern conditions. Our analysis at the regional scale corroborates the hypothesis that BSR does not vary greatly within the city as a result of the types of periurban landscapes. Sites with similar structures had similar BSR, independent of the size of the city or the type of periurban landscapes, as Blondel et al. (1984) have suggested. Bird communities were more similar between cities for a given habitat than within cities between different habitats. Our local scale analysis, which compared similar types of sites, parks with trees and lawns, at different distances from the source, did not reveal any difference in BSR and similarity data. Similarly, guild comparisons between sites did not produce any significant results at either regional or local scales. This lack of modification in community structures supports the relatively small influence of periurban landscapes on urban BSR.

It could be that individuals of some species emigrate to a new town from other urbanized areas, as suggested by Erz (1966). Thus, urban bird communities can be composed of species not common in the surrounding landscapes (Clergeau et al. 1998) and local rather than regional scale factors play a more important role in shaping the structure of urban bird communities. This result agrees with Emlen (1974), who noted, in the recent town of Tucson, that some species absent from the periurban landscape live in the urban area. This sort of independence has also been observed in several population level bird studies (Havlin 1962; Erz 1966; Coombs et al. 1981; Birkhead 1991).

Among studies involving distances from a species source (patches of ‘native vegetation’) to explain bird populations in urban areas, only that of Munyenyembe, Harris & Hone (1989) obtained significant results. However, they worked along a short distance gradient (less than 1·5 km) and only differences within 0·2 km from the woodland area were significant. This represents a classic fringe effect (De Graaf, Geis & Healy 1991; Chou & Soret 1996), where proximity of resources or habitat explains the presence of birds.

Our results suggest that, for a similar number of species in periurban areas, we can expect different BSR in urban areas due mainly to differences in local features. Thus, in an urban biodiversity perspective, the management of BSR would be best served by actions at a local scale. For example, diversity and density of vegetation and habitat heterogeneity in urban sites increase BSR (Davis & Glick 1978; Lancaster & Rees 1979; Goldstein, Gross & DeGraaf 1986; Clergeau et al. 1998), affect bird assemblages (Savard 1978; Jokimäki et al. 1996) and in some cases decrease nest predation rate (Jokimäki & Huhta 2000). Also, the age of a residential area is known to influence directly BSR through the development of its vegetation (Vale & Vale 1976; Hohtola 1978; Munyenyembe, Harris & Hone 1989). The size of each landscape element also influences BSR by fulfilling minimum area requirements of species (Hohtola 1978; Tilghman 1987; Jokimäki 1999). In addition, urban feeders can not only change the density but also the composition of bird assemblages, as well as increase bird diversity (Jokimäki & Suhonen 1998; Morneau et al. 1999).

Our study reinforces a suggestion not always clearly expressed in recent syntheses of urban ecology (Germaine et al. 1998; Jokimäki 1999; Savard, Clergeau & Mennechez 2000): whatever the biodiversity quality of the periurban landscape, site-specific actions such as shrub and tree planting, water restoration and increasing vegetation diversity can change bird diversity in the town and improve the quality of human–wildlife contacts.

Acknowledgements

We thank our research organizations that allowed the development of a new ecological perspective, with studies in urban systems. This work was partially supported by grants from Ministère Français de l’Environnement and the Canadian Wildlife Service. The authors are grateful to F. Gilot, A. Bregeon, G. Falardeau and M.L. Kaisanlahti-Jokimäki for their field works and for two anonymous referees for their comments on the manuscript.

Appendix

Table 6. Appendix. Birds observed in France (spring and winter census), Finland (winter census) and Canada (winter census)

Name	Country*	Season census†	Diet guild‡	Feeding habitat guild§
* Fr = France; Fi = Finland; Ca = Canada. † sp = spring; wi = winter. ‡ i = insect feeder; g = seed and vegetation feeder; o = omnivorous; c = carnivorous. § a = aquatic habitat; t = tree dwelling habitat; s = shrub habitat; m = meadow habitat; o = other habitat.
Accipiter striatus (Viellot)	Ca	wi	c	t
Aegithalos caudatus (Linnaeus)	Fr	sp/wi	i	t
Anas platyrhynchos (Linnaeus)	Fr	sp	g	a
Anas rubripes (Brewster)	Ca	wi	g	a
Apus apus (Linnaeus)	Fr	sp	i	o
Ardea cinerea (Linnaeus)	Fr	sp	c	a
Bombycilla garrulus (Linnaeus)	Fi	wi	g	t
Bonasa umbellus (Linnaeus)	Ca	wi	g	t
Bucephala clangula (Linnaeus)	Ca	wi	i	a
Cardinalis cardinalis (Linnaeus)	Ca	wi	g	s
Carduelis cannabina (Linnaeus)	Fr	sp	g	s
Carduelis carduelis (Linnaeus)	Fr	sp/wi	g	m
Carduelis chloris (Linnaeus)	Fr/Fi	wi	g	o
Carduelis flammea (Linnaeus)	Fi/Ca	wi	g	t
Carduelis pinus (Wilson)	Ca	wi	g	t
Carduelis spinus (Linnaeus)	Fr	wi	g	t
Carduelis tristis (Linnaeus)	Ca	wi	g	t
Carpodacus mexicanus (Miller)	Ca	wi	g	t
Carpodacus purpureus (Gmelin)	Ca	wi	g	t
Certhia brachydactyla (Brehm)	Fr	sp/wi	i	t
Coccothraustes coccothraustes (L.)	Fr	sp	g	t
Colaptes auratus (Linnaeus)	Ca	wi	i	m
Columba livia dom. (Gmelin)	Fr/Fi/Ca	sp/wi	o	o
Columba palumbus (Linnaeus)	Fr	sp/wi	g	o
Corvus brachyrhynchos (Brehm)	Ca	wi	o	m
Corvus corone (Linnaeus)	Fr/Fi	sp/wi	o	m
Corvus frugilegus(Linnaeus)	Fr	sp	o	m
Corvus monedula (Linnaeus)	Fr	sp/wi	o	m
Cyanocitta cristata (Linnaeus)	Ca	wi	o	t
Delichon urbica (Linnaeus)	Fr	sp	i	o
Dendrocopos major (Linnaeus)	Fr/Fi	sp/wi	i	t
Emberiza cirlus (Linnaeus)	Fr	wi	g	o
Emberiza citrinella (Linnaeus)	Fr/Fi	wi	g	o
Erithacus rubecula (Linnaeus)	Fr	sp/wi	i	s
Falco sparverius (Linnaeus)	Ca	wi	c	m
Fringilla coelebs (Linnaeus)	Fr	sp/wi	g	o
Fulica atra (Linnaeus)	Fr	sp	g	a
Gallinula chloropus (Linnaeus)	Fr	sp	g	a
Garrulus glandarius (Linnaeus)	Fr	sp/wi	o	t
Hippolais polyglotta (Vieillot)	Fr	sp	i	s
Hirundo rustica (Linnaeus)	Fr	sp	i	o
Lanius excubitor (Linnaeus)	Ca	wi	c	m
Larus argentatus (Pontoppidan)	Fr/Ca	sp/wi	o	o
Larus marinus (Linnaeus)	Ca	wi	o	a
Larus ridibundus (Linnaeus)	Fr	wi	c	a
Loxia curvirostra (Linnaeus)	Ca	wi	g	t
Loxia leucoptera (Gmelin)	Fi/Ca	wi	g	t
Luscinia megarhynchos (Brehm)	Fr	sp	i	s
Mergus merganser (Linnaeus)	Ca	wi	c	a
Motacilla alba (Linnaeus)	Fr	wi	i	o
Muscicapa sriata (Pallas)	Fr	sp	i	s
Parus ater (Linnaeus)	Fi	wi	i	t
Parus atricapillus (Linnaeus)	Ca	wi	i	t
Parus caeruleus (Linnaeus)	Fr/Fi	sp/wi	i	t
Parus cristatus (Linnaeus)	Fr	sp/wi	i	t
Parus major (Linnaeus)	Fr/Fi	sp/wi	i	t
Parus montanus (Conrad)	Fi	wi	i	t
Parus palustris (Linnaeus)	Fr	sp/wi	i	t
Passer domesticus (Linnaeus)	Fr/Fi/Ca	sp/wi	o	o
Phoenicurus ochruros (Gmelin)	Fr	sp	i	o
Phylloscopus collybita (Vieillot)	Fr	sp/wi	i	s
Phylloscopus trochilus (Linnaeus)	Fr	sp	i	s
Pica pica (Linnaeus)	Fr/Fi	sp/wi	o	m
Picoides pubescens (Linnaeus)	Ca	wi	i	t
Picoides villosus (Linnaeus)	Ca	wi	i	t
Picus viridis (Linnaeus)	Fr	sp/wi	i	m
Podiceps ruficollis (Pallas)	Fr	sp	c	a
Prunella modularis (Linnaeus)	Fr	sp/wi	i	s
Pyrrhula pyrrhula (Linnaeus)	Fr/Fi	sp/wi	g	t
Regulus sp.	Fr	sp/wi	i	t
Serinus serinus (Linnaeus)	Fr	sp/wi	g	t
Sitta canadensis (Linnaeus)	Ca	wi	i	t
Sitta carolinensis (Latham)	Ca	wi	i	t
Sitta europea (Linnaeus)	Fr	sp/wi	i	t
Sterna hirundo (Linnaeus)	Fr	sp	c	a
Streptopelia decaocto (Frivaldsky)	Fr	sp/wi	g	o
Strix aluco (Linnaeus)	Fr	sp	c	t
Sturnus vulgaris (Linnaeus)	Fr/Ca	sp/wi	o	m
Sylvia atricapilla (Linnaeus)	Fr	sp/wi	i	s
Troglodytes troglodytes (Linnaeus)	Fr	sp/wi	i	s
Turdus iliacus (Linnaeus)	Fr	wi	i	m
Turdus merula (Linnaeus)	Fr	sp/wi	i	m
Turdus migratorius (Linnaeus)	Ca	wi	i	s
Turdus philomenos (Brehm)	Fr	sp/wi	i	m
Turdus viscivorus (Linnaeus)	Fr	sp/wi	i	m
Zenaida macroura (Linnaeus)	Ca	wi	g	o