Introduction

Climate is one of the most important factors determining the distribution of avian species richness (e.g. Rahbek & Graves 2001) by its effect on primary production (van Rensburg et al. 2002, Chown et al. 2003, Huribert & Haskell 2003), and by its interaction with physiological requirements and tolerances of species (Turner et al. 1988, Woodward & Kelly 2003). Factors determining species richness are not universal, but vary among taxa (e.g. Miller et al. 2003). This is because the distribution of species is mediated by their ecological niche (Pulliam 2000, Wiens & Donoghue 2004), and groups of species differentiated by their taxonomy, behaviour, or physiology can have a shared or convergent evolution, responding differentially to certain factors which interact with the necessities imposed by their niche. Knowledge on the way how climate affects the distribution of avian species richness is of prime importance as the Earth is under a climatic warming (IPCC 2001). This climatic change is already affecting bird distribution (Thomas & Lennon 1999), and dramatic changes in the distribution of birds are predicted (Huntley et al. 2006).

Figure 1.

Distribution map for the percentage of southern species in Spain in 10 × 10 km squares.

Species in different climate zones of the Earth must be adapted to different environmental conditions, which spawned the distinction between biogeographic types (Voous 1960, Newton 2003). As species of different biogeographic types are adapted to different climates, climate might differentially affect species richness of birds of different biogeographic types. We tested this hypothesis on the basis of the distribution of birds in Spain. We predict that the richness of northern species should be negatively correlated with temperature, and positively with precipitation, while the reverse should occur in the richness of southern species (also see Santos & Tellería 1995).

Methods

The study area was peninsular Spain, which has a variety of environments ranging from Mediterranean aspects to an oceanic climate along the Cantabrian coast. The study area was divided in 5331 UTM squares of about 10 × 10 km (Fig. 1). Cartographic distortions caused some squares to be less than 100 km2, and these were removed from the analyses. Squares without environmental information were also dropped from analyses, resulting in a final sample size of 5070 squares.

Species richness was defined as the number of bird species in each cell. We used all species of breeding birds listed in the national atlas (Marti & del Moral 2003, Ministerio de Medio Ambiente 2003). We assigned each species to the category ‘northern’ or ‘southern’, according to its biogeographic type (listed in Online appendix 1). Biogeographic types for which distribution was ubiquitous (e.g. Cosmopolitan or Old World) were discarded (Online appendix 1). With these data, we calculated the percentage of southern species in each square. The percentage of southern species was positively correlated with the richness of southern species (r = 0.66; P < 0.001; n = 5070 squares) and negatively with the richness of northern species (r = -0.64; P < 0.001).

Independent variables were acquired from the European Environment Agency ( www.eea.europa.eu), using a geographic information system (SAGA; Conrad 2005). To test effects of climate, we considered two variables for each of the squares: (1) mean annual temperature, and (2) total annual precipitation. Mean annual temperature was strongly correlated with mean temperature in the coldest and hottest months (r > 0.88). Moreover, we considered (3) altitude range, (4) habitat diversity, as the sum of types of land use, taken from Corine Land Cover ( www.eea.europa.eu), (5) human population density (log-transformed), (6) humanized surface area, as the percentage of area used by humans (croplands and urban zones; arcsine-transformed), which served as a negative indicator of natural land available. Lastly, in order to minimize possible effects of spatial autocorrelation, we introduced the geographic variables longitude (Lon) and latitude (Lat) of the centre of the squares, as well as the composite variables Lon², Lat², Lat³, Lon²×Lat and Lon×Lat², according to Legendre (1993). We did not use Lon³ and Latitude×Longitude because this destabilized the matrix and least squares could not be calculated. The inclusion of these terms successfully removed most of the spatial autocorrelation, indicated by Moran's I of the residual models, always lower than 0.15 (Diniz-Filho et al. 2003).

Variables had an almost normal distribution, otherwise they were transformed to fit a normal distribution. Variables were standardized with a mean of 0 ± 1 SD (Sokal & Rohlf 1995). To test the relationship between independent variables and species richness, we used a Generalized Linear Model (GLM) of Ordinal Least Squares (OLS). This analysis statistically controls for the effects of other independent variables. Multicollinearity was not high, as absolute values of correlations between independent variables were 0.66 or less, and tolerance was always higher than 0.3 (Quinn & Keough 2002). To test for curvilinear relationships, we introduced polynomial terms of variables 1–6 into the model. Variables in the final models were selected by a stepwise backward process.

Results and discussion

Figure 1 shows the distribution of the percentage of southern species in Spain. As expected, the percentage of southern species per square correlated with temperature in a positive way (r = 0.65; P < 0.001; n = 5070 squares; Fig. 2). When controlling for effects of other variables, the GLM showed a significant positive relationship between percentage of southern species and temperature (Table 1). This model explained 76% of variation in percentage of the southern species. We repeated the GLMs in order to examine the relationship between temperature and species richness of southern and northern species. After controlling for the other independent variables in this study, species richness of southern and northern species were related to temperature in a quadratic way (Table 1). In both cases, species richness dropped with increasing temperature, but less so in southern species, thus explaining the results found for the percentage of southern species. Other studies have shown that the distribution or abundance of avian species, in general, is affected by temperature (Root 1988, Turner er al. 1988, Lennon et al. 2000), our study extending on this by showing that the effect of temperature depends on the biogeographic type considered.

Figure 2.

Relationship (without controlling for other variables) between temperature and number of southern species (y=-7+0.24x-0.18x²), number of northern species (y=66-0.66x-0.12x²), and percentage of southern species (y=0.65x). Data points are not shown for clarity.

Table 1.

Results of GLM's analysing dependence of percentage of southern species, richness of southern species, and richness of northern species on various variables. The b coefficients of the multiple-regression model are given, as well as values of R2 and F-statistic. In bold, slopes that significantly differ from zero at P < 0.0025 (corrected by Bonferroni). - indicates that variable was removed by backward stepwise selection.

While, percentage of southern species correlated negatively with precipitation (r = -0.64; P < 0.001). However, when the relationship with precipitation was controlled for other variables, a quadratic relationship emerged, with higher percentages of southern species for intermediate values of precipitation. Analysing richness of northern and southern species separately, the two biogeographic types showed a similar quadratic relationship with precipitation (Table 1). Probably, precipitation favours species richness through primary productivity (Waide er al. 1999, Hawkins et al. 2003, Whittaker et al. 2007), but high levels of precipitation harm species survival and breeding, affecting plumage impermeability and foraging opportunities (Lennon et al. 2000).

The GLM also showed a significant quadratic relationship between percentage of southern species and human population density, indicating a decline of percentage for high values of this variable (Table 1). Many southern species are associated with farmland (Suáres-Seoane et al. 2002), which could explain this relationship. On the other hand, the relationship between human population density and species richness of the two southern and northern species differed, being quadratic for southern species, while linear for northern species (Table 1). Therefore, although avian species richness usually correlates with human population (e.g. Araújo 2003), this study shows differences in this relationship according the biogeographic type considered. This may have implications for avian conservation, as southern species seem to be more sensitive to high levels of human disturbance.

The percentage of southern species showed a significant quadratic relationship with altitude range, with a decline of percentage for highest values of altitude range (Table 1). The concave-up relationship between richness of northern species and altitude range (Table 1) is probably caused by the inclusion of different faunas in the ‘northern species’ category, with some species inhabiting mountains, while other dwelling in plains.

In sum, this study shows that ecological factors differentially correlate with richness of species of different biogeographic types, resulting in which factors such as temperature mediate differences in the composition of avian species throughout Iberian Peninsula. Similarly, other studies have shown that ecological factors differentially affect avian species richness according to the distribution range of species considered (Jetz & Rahbek 2002); species richness of birds of different foraging guilds is also affected by different environmental variables (Miller et al. 2003). The ecological determinants of avian species richness also vary geographically (Davies et al. 2007), which might be a consequence of different avian communities in which species richness is affected by different factors.

Lastly, temperature has increased in Spain in the last century (Hulme & Sheard 1999), and it is predicted to continue increasing in the coming years (IPCC 2001). According to the findings of this study, this will provoke a change in the composition of avian communities in the Mediterranean region by increasing the percentage of southern species. Apparently, northern species are more threatened by climatic change in the region, while in relatively cold zones southern species would be favoured by a rise of temperatures. However, our analysis provides evidence that in relatively hot zones southern species would be harmed as well. Thus, considerable increases of temperature may cause a decline in avian species richness in Mediterranean regions.

Comments by anonymous referees, Juan Manuel Pleguezuelos, David Nesbitt and Jouke Prop improved the manuscript.