1. Urbanization is a landscape process affecting biodiversity world-wide. Despite many urban–rural studies of bird assemblages, it is still unclear whether more species-rich communities have more individuals, regardless of the level of urbanization. The more-individuals hypothesis assumes that species-rich communities have larger populations, thus reducing the chance of local extinctions.
2. Using newly collated avian distribution data for 1 km2 grid cells across Florence, Italy, we show a significantly positive relationship between species richness and assemblage abundance for the whole urban area. This richness–abundance relationship persists for the 1 km2 grid cells with less than 50% of urbanized territory, as well as for the remaining grid cells, with no significant difference in the slope of the relationship. These results support the more-individuals hypothesis as an explanation of patterns in species richness, also in human modified and fragmented habitats.
3. However, the intercept of the species richness–abundance relationship is significantly lower for highly urbanized grid cells. Our study confirms that urban communities have lower species richness but counters the common notion that assemblages in densely urbanized ecosystems have more individuals. In Florence, highly inhabited areas show fewer species and lower assemblage abundance.
4. Urbanized ecosystems are an ongoing large-scale natural experiment which can be used to test ecological theories empirically.
Using newly collected data from the whole administrative area of Florence, Italy, we investigated breeding bird species richness and abundance patterns in relation to landscape composition. Florence is a town of about 365 000 inhabitants over an area of slightly more than 100 km2, for a total density of c. 3600 inhabitants per km2. This human density is lower than the value for other Italian cities of similar size but with more inhabitants [e.g. Naples (c. 8500 n km−2), Palermo (c. 4000 n km−2), Turin (c. 6500 n km−2)], but is higher than for cities whose administrative area includes also large agricultural and semi-natural areas [e.g. Rome (c. 2000 n km−2, but more than 10 times larger than Florence)]. However, the municipality of Florence still contains much semi-natural habitat, as nearly 60% of its 1 km2 grid cells are urbanized in less than 50% of their area.
The aim of this study was to test one condition required for the more-individuals hypothesis to be a potential explanation of patterns in bird species richness across an urbanized landscape. We investigated whether the hypothesis that more species-rich communities have more individuals applies for the urbanized ecosystems of Florence, and whether there are differences in the form of the species richness–assemblage abundance relationship for highly vs. less urbanized parts of a completely sampled urban area.
Materials and methods
Bird species distribution data for 124 grid cells (1 km2 each) of the whole administrative area of Florence were collected during the years 2007–2008. This study follows the previous censuses at the end of the 1980s and of the 1990s (Dinetti & Ascani 1990; Dinetti & Romano 2002). The census followed the rules of the working group Urban Avifauna (Dinetti et al. 1995; Dinetti 2005). These are based in turn on the recommendations of the International Bird Census Committee (Stastny & Bejcek 1990). The census was carried out by 23 surveyors funded by the Italian Society for the Protection of Birds (LIPU) in collaboration with the Environment Department of the Municipality of Florence.
The census was both qualitative [mapping of distribution areas, with indication of nesting (possible, probable and confirmed)] and quantitative (estimation of number of pairs for breeding species in each grid cell). Each grid cell was surveyed by experienced ornithologists or birdwatchers at least two times during the breeding season (start of April–end of June), with at least 1 month between the two visits. Each visit was about 3 h long and started at dawn. Each grid cell was visited either during 2007 or during 2008. The species for which there was no evidence of nesting and which were observed during their migration period were not recorded.
Florence (43°47′N, 11°15′E) was founded in Roman times (59 BC, by Julius Caesar) and is situated in the valley of the river Arno, on the South-West of the Apennine mountain range, at about 80 km from the Mediterranean sea and at an altitude of about 75 m. The climate is sub-Mediterranean (hot, dry summers and cool winters). Mean annual precipitation is about 800 mm and mean annual temperature is about 14 °C. The urban area has a hilly topography and a range of ecosystems, from a densely built town centre to woodland patches, villas and olive groves in the outskirts, from public and private gardens to vineyards, waste land and fields.
The relationships between breeding bird species richness and assemblage abundance, as well as between these two variables and the proportion of (i) urbanized territory, (ii) open habitats (not reported) and (iii) land use with trees were analysed in sas 9.1. Urbanized territory was defined by the continuous presence of buildings. Open habitat was the sum of the areas of shrub land, grassland, cropland, orchards, allotments, pastures, vineyards and plant nurseries. Land use with trees was the sum of the areas of broadleaved, coniferous, mixed and riparian woodland, as well as poplar and olive plantations, private and public green areas, tree avenues and agro-forestry.
We also ran models of bird species richness as a function of assemblage abundance (i) inserting urbanization level as a continuous variable and (ii) for two subsets of the grid cells differing in urbanization level (cells with urbanized territory ≥50% vs. <50% of the cell). Breeding bird species richness and assemblage abundance were log transformed prior to analysis to better approach a normal distribution. At first, we ran linear regressions. We also tested whether quadratic terms could improve model fit and we controlled for spatial autocorrelation using mixed models with exponential co-variance structure (as e.g. in Pautasso & Fontaneto 2008). Results from non-spatial and spatial models were qualitatively consistent, but for simplicity only results which take into account a potential spatial non-independence of data are presented. Spatial non-independence of data over short distances can lead to misleading parameter estimates as it can inflate the effectively available degrees of freedom (e.g. Dormann 2007; but see Hawkins et al. 2007). We found significant Moran’s I at short distances for the residuals of all models. Significant Moran’s I at short distances justify the use of models which take into account spatial autocorrelation.
Breeding bird species richness varied in the 124 grid cells of the administrative area of Florence between 6 and 47 (average: 31, median: 31, SD: 9). Overall, 86 bird species were recorded. Assemblage abundance (the number of individuals of all bird species reported in a grid cell) varied between 13 and 707 individuals (average: 244, median: 219, SD: 143). Overall, 30 000 individual contacts were recorded. The proportion of urbanized territory was between 0% and 98% (average: 44, median: 40, SD: 30%). The proportion of land use with trees ranged between 0% and 100% (average: 32, median: 25, SD: 30%), whilst the range of the proportion of open habitats was 0–84% (average: 22, median: 19, SD: 19%).
Breeding bird species richness significantly increased with increasing assemblage abundance throughout the 124 grid cells (Fig. 1a) and for grid cells (i) with less and (ii) with more than 50% of urbanized territory (Fig. 1b). For these two subsets of data, the species richness–assemblage abundance relationship had no significantly different slopes, allowing us to test whether the intercepts were different. The intercept of the species richness–assemblage abundance relationship was significantly lower for the grid cells with more than 50% of urbanized territory compared to those with less than 50% of urbanized territory (P < 0·001; Fig. 1b). When inserting urbanization as a continuous variable in the model of species richness as a function of assemblage abundance (all cells), species richness still increased significantly with increasing assemblage abundance. In this model, urbanization had a significant negative influence on species richness.
There was a significant quadratic relationship between breeding bird species richness and the proportion of land use with trees (Fig. 2a), so that species richness first increased with increasing presence of trees (up to c. 50%) and then declined. This quadratic relationship was found also for the grid cells with proportion of urbanized territory lower than 50% (Fig. 2b). For the grid cells with proportion of urbanized territory greater than 50%, the relationship between species richness and proportion of land use with trees was instead linear and significantly positive (Fig. 2b). There was a significant quadratic relationship between breeding bird species richness and the proportion of urbanized territory for the whole of the data set (Fig. 3).
For the whole administrative area of Florence, there is a significantly positive relationship between bird species richness and assemblage abundance, as predicted by the more-individuals hypothesis. This positive relationship persists for the 1 km2 grid cells with less than 50% of urbanized territory, as well as for those with urbanization covering more than the majority of the grid cell, with no significant difference in the slope of the relationship for the two subsets of data. The data available for the administrative area of Florence confirm that the more-individuals hypothesis can be an explanation of patterns in species richness also for highly modified ecosystems such as urban areas. Given that the positive species richness–assemblage abundance relationship is present both for highly and less urbanized cells, this condition of the more-individuals hypothesis (at least in this case study) appears to apply regardless of the level of urbanization.
However, the intercept of the species richness–assemblage abundance relationship is significantly lower for highly urbanized grid cells than for those with a substantial presence of semi-natural ecosystems. This result runs counter the common assumption that urban centres have higher assemblage abundance but confirms the lower species richness of the centre of urban areas, which follows from the insufficient availability of habitat patches for the large majority of species which could potentially be present (McKinney 2008) and from the increased disturbance because of the higher number of human beings (Schlesinger, Manley & Holyoak 2008). The significantly lower intercept of the species richness–assemblage abundance relationship in highly urbanized grid cells also supports the more-individuals hypothesis, as it links the lower presence of species with a lower presence of individuals. Florence is thus not only a town of history, art and tourists, but also an example of how urbanized ecosystems can be used as an open-air, large-scale, natural experiment to test ecological hypotheses. It is possible that the different result (lower richness and lower abundance at higher urbanization levels) obtained here in comparison with previous studies along rural–urban gradients (lower richness but higher abundance towards town centres) may be due to the census in Florence having been carried out over the whole administrative area.
The bird distribution data available for Florence show that species richness peaks at intermediate availability of trees in the landscape. This is because if grid cells only contain wooded habitat, then species typical of open habitats are less likely to be present. However, the data available confirm the importance of the presence of trees for bird species richness in highly urbanized areas (Fernandez-Juricic 2000; Marzluff 2005; White et al. 2005). For the grid cells with proportion of urbanized territory higher than 50%, there was a significantly positive linear relationship between presence of land use with trees and species richness. Because of the topographical location between hills, Florence has a town centre with a very dense housing pattern so that the presence of even single trees in squares or alleys can make a difference for bird biodiversity (Pauleit et al. 2005).
This result has implications for the practical conservation of birds in this and other urban areas of similar human population density (Fernandez-Juricic & Jokimäki 2001; Chamberlain et al. 2007; MacGregor-Fors 2008; Palmer et al. 2008; Vallejo, Aloy & Ong 2009). In Florence, there is a wide range of land use with presence of trees, from urban woodland to private and public gardens, from poplar and olive tree plantations to vegetation along watercourses (Pauleit et al. 2005). Our study shows that these remnant habitats with presence of trees are fundamental for the persistence of bird biodiversity throughout the whole urban area, not only in the areas with less anthropogenic impact.
Many thanks to the many people involved in the collection of bird distribution data in Florence, to F. Cooke, K. Evans, K. Gaston, O. Holdenrieder, M. Jeger, M. McKinney, L. Vazquez, P. Weisberg for information, insight and discussion and to V. Bahn, S. Chown, J. Marzluff, T. Matoni and M. McKinney for helpful comments on a previous draft.