The centre–periphery hypothesis predicts that habitat suitability will decrease at the edge of a species' range, a pattern often questioned by empirical data. Here we explore if habitat suitability decreases southwards and shapes the abundance distribution of rear edge populations of forest birds within the restricted geographical setting of the south-western Palaearctic. We also test if birds endemic to the area fit more poorly to the latitudinal decrease of habitat suitability due to the putative effect of adaptations to regional conditions. Location: North-western Africa (Morocco) Time period: Present day Major taxa studied: Passerines (11 species) Methods: Bird occurrences were used to model species distribution and line transects were used to estimate bird abundance. Occurrence probabilities provided by species distribution models were used to display the spatial patterning of habitat suitability. Habitat suitability was employed to predict abundance after controlling for the effect of the distance to some regional source areas of forest birds (tree covered large areas). The species were classified as North African endemic according to an updated review of their taxonomic status. Results: Habitat suitability decreased southwards, supporting the predicted relationship between ecological and geographical marginality in most species. Abundance was positively correlated to habitat suitability and negatively correlated to distance to source areas. The taxonomic status of birds did not affect the patterns. Main conclusions: The southward decrease of habitat suitability predicted by the centre–periphery hypothesis shapes the distribution of rear edge populations of forest birds within the south-western Palaearctic. As most of these populations are endemic, the results suggest that they track the gradients in isolation within the geographical setting of north-western Africa. These results support the vulnerability of these isolated, peripheral populations of forest birds to large-scale environmental changes in a region under the effect of increasing drought and temperature.

The first set of data ("Bird occurrences") provide geo-referenced observations of the study species, which result from extensive bird sampling carried out by the authors within the study area. These data were used to model species distribution (Maxent), and the occurrence probabilities provided by species distribution models were used to assess the spatial patterning of habitat suitability.

The second set of data ("Transects") were used to estimate bird abundance. We counted all birds within a belt 500 m long and 25 + 25 m wide. To control the effect of local habitat structure, we set two 25-m-radius circles distributed at 150 and 350 m of the starting point of each itinerary. In each point we visually assessed (%): grass layer (GRASS), shrub (vegetation between 0.5 and 2 m height; SHRUB) and tree (vegetation >2 m height; TREE) covers, and tree and shrub (>0.5 m) species numbers (SPP-SHRUB-TREE). In this data set, the acronyms of the species is: CERBRA: Certhia brachydactyla; CYATEN: Cyanistes teneriffae; ERIRUB: Erithacus rubecula; FRICOE: Fringilla colebs; PERATE: Periparus ater; PARMAJ: Parus major; PHYBON: Phylloscopus bonelli; REGIGN: Regulus ignicapillus; SYLATR: Sylvia atricapilla; TROTRO: Troglodytes troglodytes; TUERME: Turdus merula).

These last data were used to perform a multivariate approach for a comprehensive view on the interacting effects of local forest cover, geographical location, and habitat suitability on bird abundance. After testing for multicollinearity, we controlled for the effect of spatial autocorrelation and, because of the scarcity of most forest birds within the study area, we coped with the effect of a zero-inflated structure in our data. To solve both problems, we performed a Zero-Inflated Poisson model with an intrinsic Conditional Autoregressive model (iCAR) process in which we accounted for the effect of the zero-inflated structure on abundance and controlled for the potential effects of spatial-autocorrelation.

The files ("Bird occurrences" and "Transects") contain the data on which all analyses have been carried out.