Evaluating the expansion of African species into Europe driven by climate change

Ongoing climate change is presently influencing the distribution ranges of numerous species, with both range expansions and latitudinal shifts being observed. In southern Europe, a biogeographical border that separates African and European biota, while at the same time acting as a migration bridge for many species, these changes are of particular relevance. This study aimed to analyse the responses of nine typically African birds to climate change to provide information on the ongoing and future occupation of Europe by these species.


| INTRODUC TI ON
In recent decades, the global climate has become warmer (Rohde & Hausfather, 2020), with widespread effects on biological systems (DeMars et al., 2021;Karl & Trenberth, 2003;Root et al., 2003;Walther et al., 2002).Several studies have shown how species have responded to this recent climate change through shifts that ultimately transform their ranges (Borzée et al., 2019;Chamorro et al., 2017;Lazo-Cancino et al., 2020;Soultan et al., 2022;Wallingford et al., 2020).For species with potentially high dispersal abilities, such as birds, movements tend to translate into altitudinal or latitudinal ascents (Freeman et al., 2018;Huntley et al., 2006;López-Ramírez et al., 2024;Maclean et al., 2008;Massimino et al., 2015;Stiels et al., 2021) in response to the warming trend in climate (IPCC, 2022).While this contributes to migratory species finding wintering quarters closer to their breeding areas (Gordo, 2007), it can also conversely result in non-migratory species finding new favourable territories in areas outside their normal distribution range (Chamorro et al., 2020;Crawford et al., 2008).Predicting these range shifts is a current challenge for biogeography because they are relevant both for the species that undergo the changes and for the other species that reside in the same areas, as community compositions are ultimately altered (Wallingford et al., 2020).This adds uncertainty to the future status of natural populations and communities and forces conservation programmes to be adapted to new species distributions (Real et al., 2010;Root & Schneider, 2006).
In the context of ongoing climate change, distribution changes are occurring worldwide (Soultan et al., 2022;Stiels et al., 2021;Zhu et al., 2022).This topic is of particular relevance in southern Europe and northern Africa, where an important biogeographical barrier separates their flora and fauna.The Strait of Gibraltar, only 14 km across at its narrowest point, is a migratory bridge for many species (Evans & Lathbury, 1973;Hahn et al., 2009) while at the same time acting, together with the Mediterranean sea, as an effective biogeographical barrier (Carranza et al., 2006;Gantenbein & Largiadèr, 2003;Gil-López et al., 2022).This area delimits the northern distribution for many African species and the southern distribution for a number of European species.In the case of birds, a short-distance expansion of African species' range towards the north represents a major step in biogeographical terms, as a new continent would be reached and colonised.In recent years, the south of the Iberian Peninsula has been acting as a focal point in the process of African birds colonising Europe (Chamorro et al., 2017;Elorriaga & Muñoz, 2010;Ramírez et al., 2011), which makes it an area of great value for studying the effect of climate change on species distribution (López-Ramírez et al., 2023).This is not exclusive to the Mediterranean basin, as a number of species typically distributed in southern Europe have also been expanding their distribution ranges towards the north (Knaus et al., 2018;Stiels et al., 2021).
The number of records of African species in southern Europe has been increasing, aided by a growing number of observers sharing their reports through citizen science platforms.The growth in popularity and scale of these initiatives and their acceptance by the scientific community have made observational data more readily and quickly available to researchers (Feldman et al., 2021;Sullivan et al., 2014).eBird is the most important bird-related platform that collects information about the distribution and abundance of species worldwide, taking advantage of the enormous popularity of birdwatching to create a global network of volunteers who submit their observations to a central data repository via the Internet.
Through a combination of broad-based community engagement and global partnerships, the volume of data being submitted to eBird has increased exponentially (Johnston et al., 2021).
Climate change could further increase the number of African species in Europe, increasing the possibility that self-sustaining populations may become established in southern Europe, as has already occurred with the Little Swift (Apus affinis) (Prieta, 2022;Ramírez et al., 2002), and the Atlas Long-legged Buzzard (Buteo rufinus cirtensis) (Elorriaga & Muñoz, 2010) in the last decade.Other species that are relatively regularly observed and have already been confirmed to breed in southern Spain include the Common Bulbul (Pycnonotus barbatus) (Navarrete, 2022), the Cream-coloured Courser (Cursorius cursor) (Cabrera, 2022) and more recently the Rüppell's Vulture (Gyps rueppelli) (Muñoz et al., 2024), as well as others remaining in northern Africa that have not yet breed (Figure 1).
In the present study, we have selected nine typically African bird species from five different orders: two species from Accipitriformes, one from Apodiformes, one from Charadriiformes, one from Columbiformes and four from Passeriformes.All nine have been observed in Spain within the last decade and, in some cases, are consistently breeding or have sporadically bred.We aimed to analyse the response to climate change of this set of African species in order to provide information on the ongoing and future occupation of the European continent by these birds, as well as to identify European areas susceptible to receiving African species.This information will be of interest in the monitoring of the expansion of African species into Europe as a result of climate change and should also facilitate the environmental management of potentially affected areas and species.

| Study area
The study area comprised the land region between 20°00′ W to 60°00′ E and 09°30′ N to 70° 00′ N, thus covering the Western Palearctic and surrounding areas (Figure 2).We considered the entire Western Palearctic because it is a relevant biogeographical unit for studying the northward movement of different bird species as a consequence of a warmer climate.This area fully covers the current breeding territories of the nine species studied and has a high climatic heterogeneity, involving sub-tropical, desert, Mediterranean, Atlantic, tundra and boreal climates (Font, 2000;Udvardy, 1975).
The study area was divided using a cell grid of 1-degree latitude × 1-degree longitude to obtain operational geographic units (OGUs, n = 4177), using the Create Fishnet and Intersect tools from ArcGIS 10.4.1 software (ESRI, 2016).

| Selected species and species distribution data
In the last decade, citizen science platforms have become key sources of global biodiversity data for certain taxa, including birds (Feldman et al., 2021).For this study, distribution data for determining the breeding areas of the nine selected species were obtained from eBird (https:// ebird.org/ home), an international citizen science platform specialising in birds, with nearly one million users around the world.The nine African bird species selected for this study, the number of records and the number of breeding OGUs in Africa and Europe of each of them are shown in Table 1.
We identified the OGUs where each of the selected species had been reported to breed until the end of 2020 and then used only these breeding presences to model the distribution of the species in the study area (see Figures A.1-A.9).

| Predictor variables and future scenarios
We used a set of 21 environmental variables in the biogeographical modelling procedure, two of them related to topography and the remaining 19 related to climate between 1950 and 2000 (Table 2).We included topography together with climate in the modelling approach because otherwise their differentiated influences may be confounded and mistakenly attributed only to climate (Márquez et al., 2011).These variables were downloaded in raster format at a resolution of 1-km 2 pixels.Values of these variables at each OGU were obtained by averaging the values of the 1-km 2 pixels within them using the ZONAL function of ArcGIS 10.4.1 software.
Expected future values of the climatic variables were obtained for the period 2041 to 2060 (https:// world clim.org/ ).Following the method of Chamorro et al. (2020), four different Representative Concentration Pathways (RCPs) were used to project future CO 2 emissions: 2.6, 4.5, 6.0 and 8.5 (Pachauri & Meyer, 2014).To consider other sources of uncertainty in relation to the future climate of the study area, two different Global Circulation Models (GCMs) were also used: HadGEM2-ES and NorESM1-M (Collins et al., 2011;Real et al., 2010).We chose these two GCMs because they are good predictors of future climate in both Europe and Africa (McSweeney et al., 2015).This process resulted in eight sets of expected values of the climatic variables.We assessed whether the future values for the period 2041-2060 of the climatic variables that entered the models of the nine studied species were within the range of values that these variables had for the present.

| Models for the present
To model the distributions of the different species, we first tested the response of each species to each explanatory variable (see Table A .1).This response can be linear or unimodal.A linear response implies that as the environmental gradient for an explanatory variable increases, the relationship with the presence of the species can be positive or negative.In contrast, an unimodal response implies that when a limiting value in the environmental gradient for an explanatory variable is reached, the linear positive relationship becomes negative.To this end, binary stepwise logistic regression models were performed for each of the explanatory variables separately, using their original (x) and quadratic (x 2 ) forms.A response was considered to be unimodal when the best logistic regression model included the original form in positive and the quadratic form in negative (x − x 2 ), as this was the only biologically plausible unimodal response.For all variables that showed this type of unimodal response by each species individually, the result of the unimodal model (y or logit function) was used as an explanatory variable in the final modelling process.In all other cases, the original form of the variable was used.
To avoid problems related to multicollinearity, whenever two variables were highly correlated within a model (r > .8), the one with the lower contribution was excluded (Zanolla et al., 2018).Based on the set of pre-selected variables in the previous step, the false discovery rate (FDR; Benjamini & Hochberg, 1995) was evaluated to control the increase in type I errors and thereby the likelihood of obtaining false significant results when a large number of variables are used in the modelling process (García, 2003).Following the procedure of Benjamini and Yekutieli (2001), only the variables whose significance in the score test was less than an FDR value of 0.05 were accepted in subsequent modelling procedures.
We performed a multivariate forward-backward stepwise logistic regression of the distributions of the species in the study area on the remaining subset of variables to obtain a comprehensive model for the current probability of breeding at every OGU, according to its climatic conditions.This procedure started with a null model that had no explanatory variables included.A multivariate model was built by adding a variable at each step if the resulting new regression was significantly improved by the new variable, until the step in which any variable significantly increased the predictive capacity of the model was reached (Legendre & Legendre, 1998).By using a forward-backward stepwise variable selection procedure, before adding a new variable to the model, the possibility of improving its F I G U R E 2 Study area divided into operational geographic units of 1-degree latitude × 1-degree longitude.Map projection: WGS_1984_World_Mercator (ArcGIS 10.4.1).

TA B L E 1
African species selected in this study with their respective number of breeding operational geographic units (OGUs) within and outside Western Europe.Note: The number of records from eBird throughout this century until the end of 2020 for each of them is also shown.
predictive capacity was evaluated by eliminating any of the variables introduced in the previous step.Finally, a significant combination of predictors was obtained (y or logit), where the coefficients of the predictor variables were estimated using a machine learning algorithm based on a likelihood ascent gradient.The relative weight of each variable in the final model was assessed using the Wald test (Wald, 1943).The variance inflation factor (VIF) of each variable was used to quantify collinearity between them in the models.VIFs were calculated for each variable as the inverse of the coefficient of nondetermination for a regression of that variable on all others.VIF is a positive value representing the overall correlation of each variable with all others in a model (Muñoz et al., 2015;Zuur et al., 2010).
The result of the multivariate logistic regression was a probability of breeding based on the set of environmental variables, which was affected by the prevalence of the species that was modelled in the dataset.In this study, we combined different species' distributions, so it was mandatory to use commensurate methods unaffected by different prevalences.The effect of prevalence on the probability values was therefore removed, thus obtaining favourability values by applying the favourability function (Real et al., 2006).More detailed discussions of the procedure have been published previously (Chamorro et al., 2020;García-Carrasco et al., 2021;López-Ramírez et al., 2023).The term 'favourability' refers to the degree, ranging from 0 (minimum favourability) to 1 (maximum favourability), to which the environmental conditions are propitious for a species to breed (Acevedo & Real, 2012;Muñoz et al., 2015), with F = 0.5 being the threshold separating favourable from unfavourable areas.A local favourability value of 0.5 indicates that the local probability of a species breeding is the same as its prevalence in the study area-in other words, the probability expected by a null model unaffected by environmental predictors, where breeding is neither favoured nor unfavoured by the environment (Real et al., 2006).Those areas with F > 0.5 therefore favour the species breeding, whereas F < 0.5 indicates areas with conditions that disfavour breeding.Nevertheless, favourability is a continuous and fuzzy concept (Acevedo & Real, 2012), making it unrealistic to use a favourability value of 0.5 as a cut-off to neatly distinguish favourable from unfavourable areas (Hosmer & Lemeshow, 2005).Thus, although in the maps we showed the favourability values of each OGU in 10 classes, from 0 to 1, we classi- TA B L E 2 Variables selected to model the distribution of the nine species studied, grouped by environmental factor.
The favourability models conducted in this study were topoclimatic as we included topography together with climate in the modelling approach.Nevertheless, we did not distinguish the effect of climate and topography on species distributions since we considered topography as an inherent part of climate.Thus, we assumed that topography would remain constant in our future projections and studied how climate changes.For this reason, hereafter we will only refer to climatic favourability instead of topo-climatic.

| Model assessment
The resulting nine climatic favourability models were assessed according to their discrimination and classification capacities.The discrimination capacity was evaluated using the area under the receiver operating characteristic (ROC) curve, known as the AUC (Lobo et al., 2008;Romero et al., 2013).The classification capacity, using the value of F = 0.5 as classification threshold, was assessed through the following classification measures: sensitivity (the conditional probability of OGUs with reported breeding being classified as favourable), specificity (the conditional probability of OGUs with no reported breeding being classified as unfavourable), correct classification rate (CCR: the conditional probability of correctly classified OGUs), the over-prediction rate (OPR: the proportion of OGUs with no reported breeding in the area with favourability higher than 0.5) and the under-prediction rate (UPR: the proportion of OGUs with reported breeding in the area with favourability lower than 0.5).All these measures are widely used, with values ranging from 0 to 1 (Barbosa et al., 2013;Fielding & Bell, 1997;Muñoz & Real, 2006).We also used Cohen's Kappa index (Cohen, 1960), whose values range from −1 to +1, to measure the degree to which the favourability of the OGUs with reported breeding or no reported breeding in the dataset was higher or lower than 0.5, respectively.et al., 2013;Real et al., 2010;Romo et al., 2014).This process resulted in eight expected climatic favourability models for each of the nine studied species.

| Fuzzy logic
We used fuzzy sets to represent our results because they have no clearly defined limits and therefore better reflect the continuous character of nature (Salski, 2006).A fuzzy set is a class of objects with a continuum of degrees of membership, such that the set is defined by a membership function that assigns to each object a value ranging from zero to one (Zadeh, 1965).The size of a fuzzy set is the sum of the degrees of membership of all the objects, which is called the cardinality of the fuzzy set.The height of a fuzzy set is defined as the largest membership value of the elements contained in that set.We defined for each species the fuzzy set of OGUs that are climatically favourable for it, where the degree of membership is the climatic favourability of the OGU for the species.We determined the fuzzy union of these fuzzy sets for the nine birds studied, both for the present and for the period 2041-2060, to obtain a joint vision of the trend that these species are expected to follow.The degree of membership of each OGU to the fuzzy union is here defined as the maximum value of favourability of the nine species in each OGU (Estrada & Real, 2021), that is the degree to which each OGU is favourable for any of the species.
We also defined for each OGU the fuzzy set of species for which the OGU is climatically favourable.Then, we determined the cardinality and the height of these fuzzy sets in each OGU,

| Uncertainty assessment using fuzzy logic
The predicted impact of climate change on species favourability would be informative for policy planning if the coincidence between predictions for different RCPs using the same GCM was lower than the consistence between predictions for the same RCP when applying different GCMs (Real et al., 2010).Coincidence is here defined as the concurrence between predictions according to two RCPs for a given GCM and the period 2041-2060.In Equation 1 below, two RCPs (2.6 and 4.5) have been used as an example, although all six possible pairwise RCP combinations were performed.
Coincidence was computed as follows: (1) where c(X) is the cardinality of the X fuzzy set-that is the sum of all cells' degrees of membership in the fuzzy set X. F 2.6 is the predicted future favourability according to the GCM and the scenario 2.6, and F 4.5 is the predicted future favourability according to the GCM and the scenario 4.5.F 2.6 ∩ F 4.5 is the intersection between the favourabilities of scenario 2.6 and scenario 4.5, and the degree of membership of each cell to F 2.6 ∩ F 4.5 is defined by the minimum of the two favourability values for the species in the cell.F 2.6 ∪ F 4.5 is the union between the favourabilities for scenario 2.6 and scenario 4.5, and the degree of membership of each cell to F 2.6 ∪ F 4.5 is defined by the maximum of the two favourability values for the species in the cell.
Consistence is defined here as the agreement between predictions for a given RCP applying different GCMs and is computed as follows: where F he is the predicted future favourability according to the circulation model HadGEM2-ES and F no is the predicted future favourability according to the circulation model NorESM1-M.
A Wilcoxon test was used to compare the mean values of coincidence and consistence.The significance (p) threshold that indicated if there were significant differences between the compared means was .05.
The GCM that showed the highest coincidence for each pair of RCPs and the RCP that showed the highest consistence for both GCMs were selected to represent the results for the period 2041-2060.

| RE SULTS
The mathematical models and the current climatic favourability models for each of the species as well as their assessment accord-

| Selecting the expected climatic favourability for the period 2041-2060
The coincidences between predictions for different pairs of RCPs using the same GCM and the consistencies of results derived from different GCMs assuming the same RCP are shown in Table 3. (2) TA B L E 3 Coincidence values between predictions using different pairs of Representative Concentration Pathways (RCPs) for each Global Circulation Model (GCM), as well as consistency values between predictions for each RCP using different GCMs.Coincidences (mean = 0.888, n = 108) are higher than consistencies (mean = 0.843, n = 36), differing statistically significant (Z = −2.666,p = .008).Coincidence values are statistically significantly higher (Z = −3.758,p = .0002)when using the Global Circulation Model

Species
HadGEM2-ES (mean = 0.895, n = 54) than when using NorESM1-M (mean = 0.881, n = 54).Furthermore, when comparing between GCMs for each pair of RCPs, coincidence values are higher in all cases when using HadGEM2-ES, with the exception of pair 2.6-8.5 (see A.28).These differences are statistically significant for two pairs of RCPs: 2.6-6.0 (Z = −2.666,p = .008)and 4.5-6.0(Z = −2.310,p = .021).Consistence values do not differ significantly when using RCPs of 2.6, 4.5 and 6.0, although there is a nonsignificant trend towards being higher when using RCP 2.6.Consistence values are significantly lower when using an RCP of 8.5, in comparison with using RCPs of 2.6 (Z = −2.666,p = .008),4.5 (Z = −2.666,p = .008)or  Favourability is also projected to increase in Great Britain, although it will continue to be an unfavourable area for the establishment of our set of species in the near future as values remain below 0.

| DISCUSS ION
Over the last few decades, the number of typically African bird species observed in Europe has sharply increased.Climate change appears to be one of the main causes promoting the northward expansion of these species' ranges, a hypothesis that is reinforced by several studies conducted in Europe and Africa (Chamorro et al., 2017;Huntley et al., 2006;Maclean et al., 2008;Massimino et al., 2015;Real et al., 2013;Stiels et al., 2021;Thomas & Lennon, 1999).Although the number of observers and new technologies has also increased, the recent settlement of some of these African species, and the sporadic breeding events of some others in southern Europe, confirm the regular arrival of individuals.Recent occurrences of arrival and breeding in Europe include the Little Swift (Prieta, 2022), the Atlas Long-legged Buzzard (Elorriaga & Muñoz, 2010), the Common Bulbul (Navarrete, 2022) and the Cream-coloured Courser (Cabrera, 2022).
Additionally, certain species, such as the House Bunting (Emberiza sahari; López- Ramírez et al., 2023) and the Rüppell's Vulture (Muñoz et al., 2024), initiated breeding in 2023, with other species expected to start breeding soon.Climate change is expected to increase both temperature and the frequency and duration of heat waves, as well as reduce the number of cold nights and the annual mean precipitation in many mid-latitude regions (IPCC, 2022).This should result in a warmer and drier environment in southern Europe, increasing its suitability for African species.These climatic alterations make the south of Europe a gateway for African species, allowing them to settle and reproduce in favourable areas and facilitating their subsequent expansion northwards.
Various approaches to species distribution modelling assume an equilibrium of distribution with the environment, and this may hinder their applicability for analysing the effects of climate change (Guisan & Thuiller, 2005;Guisan & Zimmermann, 2000).One of the main drivers of changes in species distribution is the disequilibrium between species' ranges and climatic favourability for those species.This climatic disequilibrium may attract the breeding range of the species to adjacent favourable areas, although there is also a temporal disequilibrium between the actual and the potential range of the species (Chamorro et al., 2020).Griffon Vulture (Gyps fulvus) (Muñoz et al., 2024).This unique pattern may warrant specific research.Furthermore, our models predicted a large availability of highly favourable areas in the southern half of the Iberian Peninsula, southern Italy, Greece and Turkey, together with large Mediterranean islands such as the Balearics, Sardinia, Sicily, Crete and Cyprus.Most of these highly favourable areas remain unoccupied by African birds, indicating that their distributions are not in equilibrium with the favourable environment north of their current ranges (Chamorro et al., 2020) (Balbontín et al., 2008;Ferrero, 1996;Keller et al., 2020;Logeais, 2015).In 2014, there were 150 breeding pairs of the species in France, a number that increased to 200-250 pairs in 2017 (Keller et al., 2020).This example shows how a typically African species has previously become established in the same areas that our models have detected as highly favour- Italy (Corso, 2009), and in the Strait of Gibraltar, Spain (Elorriaga & Muñoz, 2013), as well as between the Rüppell's Vulture and the Griffon Vulture in Malaga province (southern Spain) (Muñoz et al., 2024).These areas represent new contact zones where closely related African and European species currently meet.Hybridisation in these areas is favoured because African species are currently rare and, therefore, their choice of mates is restricted.If hybrids were infertile, hybridisation could slow or even constitute a barrier against the expansion of these species into the European continent (Elorriaga & Muñoz, 2013;Muñoz et al., 2024;Väli et al., 2010).Other factors could be the potential role of habitat (e.g.land use) or species' intrinsic factors (e.g.site-fidelity, conspecific attraction) and their interaction with climate distribution changes (Estrada et al., 2016).These could be some of the reasons why some new climatically favourable European

F
I G U R E 1 A representation of the current ranges of the nine African birds used in this study, showing the species that have already bred in southern Europe (indicated with an asterisk).Map projection: WGS_1984_World_Mercator (ArcGIS 10.4.1).
All modelling processes were run with the IBM SPSS Statistics 25 software package and our modelling approach conforms to modelling protocols proposed by Zurell et al. (2020) and Sillero et al. (2021).Maps were created using ArcMap software (ArcGIS 10.4.1; https:// deskt op.arcgis.com/ es/ arcmap/ ).

Future
climatic favourability values (F f ) were obtained by replacing the present values of the climatic variables in the logit (y) of the favourability function's equation with the expected future values according to each RCP and GCM for the period 2041-2060 (Muñoz both for the present and for the period 2041-2060.Cardinality is here defined as the sum of the climatic favourability values of the nine species in each OGU, and height is the maximum value of favourability of any of the nine species in each OGU(Real et al., 2010).The term 'cardinality' used in this study is similar to the term 'accumulated favourability' used byFa et al. (2014) andEstrada et al. (2008) to detect diversity hot spots.The height of the fuzzy set of species for which the OGU is climatically favourable coincides with the degree of membership of the OGU to the fuzzy union of OGUs favourable for any species.We calculated the increment (I) in the degree of membership to the fuzzy union and cardinality of each OGU by subtracting present favourability values from future favourability values.Positive values of I indicate a gain in favourability in that OGU, whereas negative values of I mean a loss in favourability(Real et al., 2010).
ing to their discrimination and classification capacities are shown in the supplementary material (mathematical models: Tables A.2-A.10; current climatic favourability models: Figures A.10-A.18; and model assessment: Table A.11).All variables included in the models have a VIF lower than 6.The eight expected climatic favourability models for each of the nine studied species according to each RCP and GCM for the period 2041-2060 are also shown in the supplementary material (Figures A.19-A.27).The future values for the period 2041-2060 of the climatic variables that entered the models of the nine studied species are within the range of values for the present in more than 95.5% of grid cells.

6
.0 (Z = −2.310,p = .021)(see Figure A.29).Based on these results, only the favourability values from the Global Circulation Model HadGEM2-ES and an RCP of 2.6 are used to update our results for the period 2041-2060 because it is the future model with the least uncertainty.The HadGEM2-ES Global Circulation Model's values are used because it is the GCM thatshows the highest coincidence for each pair of RCPs, and 2.6 is used because it is the RCP that shows the highest consistence for both GCMs.

3. 2 |
Combining the climatic favourability for the nine species Under current climatic conditions, there are already areas in the southern half of the Iberian Peninsula that are favourable for our set of African birds, with values as high as those found in Northern Africa for the same set of species (Figure 3, first row).There are also favourable areas in southern Italy, Greece, Turkey and in the large Mediterranean islands, such as the Balearics, Sardinia, Sicily, Crete and Cyprus.In the rest of Western Europe, favourability is low and concentrated mainly in southern France, northern Italy and some areas in the United Kingdom, where favourability values do not exceed 0.5.Based on climate change predictions, favourability will increase in Western Europe, especially in the northern half of the Iberian Peninsula, making the whole peninsula highly favourable for our set of African birds.It is also worth noting the increase in favourability in the northern half of Italy, as well as in central and northern France, with some areas reaching high values of favourability.

2 (
Figure 3, first row).The only exception among the species studied is Rüppell's Vulture, which exhibits low favourability values for both present and future conditions(Figures A.15 and A.24, respectively).According to cardinality, the southern half of the IberianPeninsula and the islands of Sardinia, Sicily, Crete and Cyprus, as well as southern Greece and Western Turkey, are currently the climatically favourable areas in Europe for the establishment of the largest number of our set of African birds (Figure3, second row).In line with future forecasts, the number of African species is expected to increase in the northern half of the Iberian Peninsula and throughout Turkey.The fuzzy union and cardinality maps of the remaining seven projected models for the period 2041-2060 can be found in the supplementary material(Figures A.30 and A.31).A gain in climatic favourability for the nine African birds is expected for the period 2041-2060 in southern Europe, especially in France, northern Italy and Greece, while a loss in climatic favourability is expected in Africa (Figure4, left).Furthermore, the number of African species with favourable climatic conditions is expected to increase in southern Europe, especially in the northern half of the Iberian Peninsula and throughout Turkey, while it is expected to decrease in North Africa (Figure4, right).
Favourability models have proven to be useful tools for identifying new potential areas for the arrival and establishment of new individuals in the near future(López-Ramírez et al., 2023;Muñoz & Real, 2006;Pulido-Pastor et al., 2018).In the present study, we constructed climatic favourability models to analyse the responses of species to climate change by identifying potentially favourable breeding areas in the European continent for nine species of African birds that are moving northwards.Our results showed that the environmental requirements of these African species fit with current climatic F I G U R E 3 First row: Degree of membership of each operational geographic unit (OGU) to the union of the fuzzy sets of OGUs that are climatically favourable for each species (meaning the degree to which the OGU is, or is expected to be, climatically favourable for any of them).Second row: Cardinality of the fuzzy sets of species for which each OGU is climatically favourable (meaning the number of species for which the OGU is, or is expected to be, climatically favourable).For the period 2041-2060, only the results of the Global Circulation Model HadGEM2-ES and the Representative Concentration Pathway 2.6 are shown.conditions in southern Europe, as shown by the concentration of highly favourable areas for breeding in the Mediterranean basin, including the European shore, and the generally low climatic favourability values in the rest of Europe.It is noteworthy that, among the nine species under examination, Rüppell's Vulture does not inhabit climatically favourable regions within Europe, either at present and in projected future scenarios.Nevertheless, this species is frequently sighted, with documented breeding evidence, albeit consistently in mixed pairs with able for the establishment of African birds.A similar pattern is now occurring with some typically Mediterranean birds that are moving northwards and reaching central Europe, where they are breeding for the first time.This is happening, for example, in Germany and F I G U R E 4 Left: Increment (I) in the degree of membership of each operational geographic unit (OGU) to the union of the fuzzy sets of OGUs that are climatically favourable for each species (meaning the degree to which the OGU is expected to increase or decrease in climatic favourability for any of them between the present and the period 2041-2060).Right: Increment in the cardinality of the fuzzy sets of species for which each OGU is climatically favourable (meaning the increase or decrease in the number of species for which the OGU is expected to be climatically favourable between the present and the period 2041-2060).Warm colours indicate positive values of I, which means a gain in favourability in that OGU, whereas cold colours indicate negative values of I, which means a loss in favourability.Switzerland, where Mediterranean species, such as the European Bee-eater (Merops apiaster), have increased breeding pairs and founded new colonies(Knaus et al., 2018;Stiels et al., 2021).Another example is the Short-toed Snake-Eagle (Circaetus gallicus), which has recently settled and experienced population growth in Switzerland(Knaus et al., 2018).These observations indicate that a change in species composition is already occurring in Western Europe due to climate change, where the south is being occupied by typically African species and the centre by typically Mediterranean species.Apart from the difficulty of crossing the Mediterranean barrier, other factors could interfere with the northward colonisation process by delaying or even stopping the spread of these African species.One of these factors is hybridisation with European species, as has already been detected between the Atlas Long-legged Buzzard and the Common Buzzard (B.buteo buteo) on Pantelleria Island, areas could remain unoccupied by African species in the future or, conversely, why some hybrid individuals are beginning to be observed, as is happening in the Strait of Gibraltar with the Gibraltar Buzzard (Buteo buteo × Buteo rufinus cirtensis;Elorriaga & Muñoz, 2013).It is important to closely monitor the northward expansion of these and other new potentially colonising African species arriving in Europe because community compositions are being altered.Our results could be useful in direct sampling and monitoring studies of these species when planning field surveys, in such a way that top priority could be given to those areas detected as highly favourable in this study.The arrival of African species in southern Europe appears to be focused on the south of the Iberian Peninsula because yet this is the European area where confirmed new breeding records and observations are concentrated.This is likely since the climatic conditions in the south of Europe are similar to that found in North Africa.If the climate continues to warm, we can expect further arrivals of new potentially colonising African species in southern Europe, with the possibility of an Africanisation of the European fauna.

Common name Scientific name Number of records Breeding OGUs in Western Europe Breeding OGUs in Africa and Eastern Europe
. Our results support the hypothesis that cli- (López-Ramírez et al., 2023)uñoz, 2010;Navarrete, 2022;Prieta, 2022) increasing the possibility that self-sustaining populations may become established in this area.It is necessary to consider that these species face a biogeographical barrier, the Mediterranean, which may slow the process of colonisation.Regardless, several African species have recently started to breed in southern Europe(Cabrera, 2022;Elorriaga & Muñoz, 2010;Navarrete, 2022;Prieta, 2022), and there has been an increasing and regular number of observations of other species that may become established soon(López-Ramírez et al., 2023).The results for the period 2041-2060 are based on the future model with the lowest uncertainty and this future projection turns out to be one of the most optimistic, as it is based on the results of the Representative Concentration Pathway 2.6.If another less optimistic future model was to be fulfilled, the climatic favourability for African birds in southern Europe would be even higher than expected from the results of this study (seeFigures A.30 and A.31).