On humans and wildlife in Mediterranean islands

Authors


*Jacques Blondel, CEFE-CNRS, 1919 Route de Mende, 34293 Montpellier Cedex 5, France.
E-mail: jacques.blondel@cefe.cnrs.fr

Abstract

Aim  To investigate the effects of human-induced landscape changes in Mediterranean islands on the ecological and evolutionary responses of bird communities and populations. The combination of mass extinction of large mammals and massive deforestation by humans was hypothesized to produce new selection regimes to which organisms were likely to respond. Habitat selection and niche breadth have been investigated at the scale of species, and phenotypic variation at the scale of local populations.

Location  The study was carried out along habitat gradients and in habitat mosaics at different spatial scales on the island of Corsica and in areas of similar size and structure in continental France.

Methods  Two sets of gradients have been used for investigating habitat selection and niche breadth: gradients of altitude, and gradients of vegetation structure. Population studies focused on the blue tit, Cyanistes caeruleus. Large samples of breeding attempts by this species in 10 habitats provided detailed data on phenotypic variation of fitness-related traits both on Corsica and on the mainland.

Results  The extent of niche space used by birds differed substantially depending on which habitat gradient was considered. Many species have been found to contract their habitat niche along the elevation gradient on Corsica compared with the mainland, whereas all species in the vegetation gradient broadened their niche on the island. Breeding patterns of the blue tit differed considerably depending on whether they settle in deciduous oaks (Quercus humilis) or in evergreen sclerophyllous oaks (Quercus ilex). Phenotypic variation of breeding traits was much higher on the island, where more populations were correctly timed for the best breeding period than on the mainland, a pattern that is likely to result from lower dispersal of organisms on the island.

Main conclusions  The differences in observed niche breadth between the two series of habitat gradients is explained both by the species-specific ecology of the species and the human-induced environmental history of Corsica. Large-scale landscape changes provided new opportunities for island colonization by non-forest species, which are isolated as small, ‘fugitive’ local populations. In both gradients, forest species that are typical components of the Corsican bird fauna definitely expanded their niche and occupied a wider range of habitats on Corsica than on the mainland. At the population scale, landscapes included habitat patches with contrasted selection regimes, which resulted in high phenotypic variation for many fitness-related traits. Reduced dispersal of birds on the island resulted in a much higher degree of local differentiation on Corsica than on the mainland.

Introduction

Mediterranean islands provide a fascinating framework for studying the impact of past human activities on the distribution, dynamics and evolution of vertebrate faunas. Large series of comprehensive palaeobotanical, palaeontological and archaeological studies provide a wealth of data for the past 10 millennia, the Holocene period, especially in Corsica and Sardinia (Reille, 1992; Vigne, 1992; Reille et al., 1996; Vigne et al., 1997). For many issues concerning environmental changes, islands provide more insight than mainland areas because drastic transformations in the composition of animal communities and the structure of landscapes have been more important and better studied on islands (Vigne & Valladas, 1996). In addition, islands are more vulnerable and less resilient to environmental changes than are mainland areas, which makes these changes easier to analyse (Walter, 2004). As pointed out by Butzer (2005), understanding cause-and-effect relationships in ecological change is difficult without an integrative and interdisciplinary methodology combining natural sciences and human sciences. Corsica is one of the few examples of a successful integration of the two approaches.

Colonization of most of the larger Mediterranean islands by modern humans dates back to the 11th–10th millennia bp (Cherry, 1990). During this long period, the islands have been deeply modified by human activity through two types of change: an almost complete turnover of mammal faunas, and active deforestation by fast-growing human societies that needed open, grassy spaces for developing agriculture and animal husbandry. These changes deeply modified the structure, composition and design of landscapes. For example, thousands of teeth and bones from Corsican archaeological sites comprise an almost complete succession of mammal species from the end of the Pleistocene (c. 10,000 years bp) to the present (Vigne & Valladas, 1996). Human activities brought about the extinction of the entire autochthonous mammalian fauna and the gradual introduction of more than 25 taxa of mammals, which constitute the present wild and domestic fauna of these islands. Of the endemic mammals, only four small-sized species survived until the Iron Age: Prolagus sardus (Lagomorpha), Rhagamys orthodon (Muridae), Tyrrhenicola henseli (Microtidae) and Episoriculus corsicanus (Soricidae). Human influence on vegetation, mostly through deforestation, was apparent as early as the end of the Neolithic (Reille, 1992) and has not really stopped since then, despite many ups and downs resulting from political, demographic and economic pulses (Thirgood, 1981; Vigne & Valladas, 1996; Blondel & Aronson, 1999; Butzer, 2005). Pollen analyses indicated that the most severe deforestation period occurred after the conquest of the island by the Romans, and did not stop through classical antiquity up to the 13th century. The mechanisms of landscape change are fairly well known through various techniques, including palynological, pedoanthracological, palaeontological and archaeological analyses (Thinon, 1978; Reille, 1984, 1992; Butzer, 2005). Fossil assemblages of small mammals determined from the contents of fossil pellets of the barn owl Tyto alba proved good indicators of changes in vegetation cover and landscape structure over several millennia (Libois, 1984; Vigne & Valladas, 1996). For example, a global decrease of the small mammals Rhagamys and Apodemus, associated with a sudden increase of Mus some 2500 years ago, indicates a strong increase of human impact with large-scale opening up of vegetation. Signatures of vegetation change from various cues have been used by Vigne & Valladas (1996) as indicators of agricultural cycles affecting the landscape structure at different periods of the Middle Ages. The most dramatic vegetation changes were the partial replacement of deciduous tree species (e.g. downy oak, Quercus humilis) by sclerophyllous tree species (e.g. holm oak, Quercus ilex). When humans first colonized Corsica, most of the lowlands of the island were covered by downy oaks, and human action progressively resulted in the replacement of this species by the evergreen sclerophyllous holm oak (Reille, 1992). This replacement has been associated, as almost everywhere in the Mediterranean Basin, with a general desiccation of the region (Blondel & Aronson, 1999; Butzer, 2005).

This paper aims first to set the scene of environmental transformation by documenting the role of humans in biodiversity decline, changes and faunal turnover, then to show that the human-induced transformation of landscapes has resulted in two different wildlife responses (only birds are studied here): (1) changes in immigration processes and habitat selection by the new colonizers (niche breadth), and (2) ecological and evolutionary responses of populations to habitat heterogeneity and fragmentation (phenotypic variation). A nested scaling from landscapes to communities and populations was used.

Changes in biodiversity

Landscapes have been designed and redesigned by humans for almost 10,000 years in the eastern part of the Basin, and for 8000 years in the western part (Le Houérou, 1981; Braudel, 1985; Pons & Quézel, 1985; Butzer, 2005). Discussing whether human impact has been beneficial or detrimental to biodiversity in the Mediterranean would be a challenging issue, and is not the aim of this paper, but two contrasting theories are worth mentioning briefly. The first is the ‘ruined landscape’ or ‘lost Eden’ theory advocated by Attenborough (1987), which argues that human action (deforestation and overgrazing) resulted in a progressive and cumulative degradation and desertification of Mediterranean landscapes. Challenging this pessimistic view, the second approach dismisses the supposedly detrimental effects of humans, arguing that the imaginary past idealized by artists and scientists does not reflect reality. According to this school of thought, humans actually contributed to maintaining Mediterranean landscapes as they progressively established since the last glacial episode, and favoured biodiversity by shaping a variety of cultural landscapes (Grove & Rackham, 2001; see Butzer, 2005; Blondel, 2006). Blondel & Aronson (1995, 1999) argued that many traditional land-use practices act as surrogates of natural disturbance regimes, with the consequence that, according to the intermediate disturbance hypothesis (Huston, 1994), several components of biodiversity have actually been higher in landscapes shaped by humans than in primitive plant communities such as oak woodlands.

The ups and downs in biodiversity on Mediterranean islands are not discussed in detail here; only events with the most relevant consequences for the ecology and evolution of ecological systems are considered. If the current biodiversity crisis is indisputable, with many plant and animal species being at risk in the Mediterranean Basin (Blondel & Médail, in press), especially on islands, the most important events occurred in the past, especially at the end of glacial times and the first millennia of the Holocene, so the current level of extinction in the Mediterranean is perhaps not significantly higher than during geological times. The fact that drastic human-induced changes occurred long ago is interesting in that it resulted in significant ecological and evolutionary responses that would not have happened if these changes had been more recent.

Perhaps the most dramatic, and still controversial, event of the Late Pleistocene, with many consequences for habitats and landscapes, is the tempo and mode of the mass extinction of large mammals. The question of whether this mass extinction was caused mainly by humans (as advocated by the overkill hypothesis; Martin, 1984), or resulted from natural environmental changes, or a combination of both, is still open and hotly debated, but recent studies emphasize the importance of human impact (e.g. Brook & Bowman, 2004; Burney & Flannery, 2005). The decimation of large endemic mammals continued well into the Holocene, as sadly illustrated by the human-induced extinction of all the large ‘mega-nano-mammals’ of Mediterranean islands, for example the dwarf hippos, elephants and deer of Cyprus, Malta, Sicily and other islands following the colonization of these islands by humans (Simmons, 1988, 1991). No fewer than 12 species of dwarf descendants of the ancestor elephant Palaeoloxodon antiquus inhabited Mediterranean islands (Lister, 1996). These species were very variable in size, and are an interesting example of evolutionary convergence because the different species evolved independently from P. antiquus. Even the small island of Tilos (64 km2) in the Aegean Sea had its own species of dwarf elephant. The smallest species, Palaeoloxodon falconeri, less than 1 m high, from Sicily, probably gave rise to the myth of the Cyclops Polyphem in the Odyssey of Homer because of the large frontal hollow of the nostrils, which looks like an enormous single eye. In addition to these mammals, tortoises, giant rodents and flightless owls, which populated most Mediterranean islands, were also decimated by humans as soon as they invaded them (Blondel & Vigne, 1993; Vigne et al., 1997). Recent studies showed that human population pressures, which increased abruptly during the late Palaeolithic and again during the Upper and Epi-Palaeolithic periods, prompted people to expand into new territories and colonize islands, as exemplified by the colonization of Cyprus (Davis, 2001). Hunting pressures became so high that they resulted in a gradual shift in culinary remains from very large animals, which became less and less common, to small animals, especially fish, birds and shellfish (Stiner et al., 1999). These demographic pulses are evidenced by increasing reliance on agile, fast-reproducing, small animals at the expense of slowly reproducing but easily caught tortoises and marine shellfish, which became less common as a result of overharvesting. Demographic pressure, with more and more people to feed, ultimately stimulated both animal husbandry for food, and the search for new territory overseas, which is clearly the case for Cyprus (Morales et al., 1998; Stiner et al., 1999; Davis, 2001). Such high human pressure must have had a great impact on habitats and landscapes. Although we still lack clear evidence for the co-existence of man and the endemic Cypriot hippos and elephants, Simmons (1988) suggested that their bones, found with cultural remains at Akrotiri, indicate the possibility of a pre-Neolithic ‘kill-site’ of these animals. It is reasonable to assume that soon after the arrival of man, these pygmy animals became extinct through overharvesting or through competition with the feral pig, which had been imported from the nearby mainland (Simmons, 1991).

All the endemic mammals of Mediterranean islands were doomed to extinction, and a complete turnover of the mammal fauna took place over the course of the past 10 millennia or so. More than 20 mammal species, both domestic and wild, immigrated from the nearby mainland, voluntarily or involuntarily introduced by humans (Vigne, 1992). Few species of birds have been introduced by humans on Mediterranean islands, but one interesting example is the partridges of the genus Alectoris. A striking characteristic of the fossil avifauna of the Palaeolithic and Holocene on Mediterranean islands is the absence of galliform species, except the quail (Coturnix coturnix), perhaps because galliforms are mostly sedentary and do not cross the sea channels that separate the islands from the continent (Alcover et al., 1992; Blondel & Vigne, 1993). The four partridge species that occur today around the Basin (Alectoris rufa, A. graeca, A. chukar, A. barbara) have been introduced as game species several times on each of the larger islands. However, only one species of partridge occurs today on each island, which is that currently occurring on the nearest mainland. This is an interesting example of competitive exclusion of allospecies – species that are differentiated enough to be considered as full species, but are ecologically too similar to live in sympatry.

Landscape changes and habitat selection: niche breadth of birds in Corsica

A classical component of the insular syndrome is niche enlargement, which has repeatedly been shown to occur on islands (MacArthur et al., 1972; Williamson, 1981; Blondel, 2000; Whittaker & Fernández-Palacios, 2007). Contradictory results have recently been reported on Corsica, depending on which components of the niche are considered. This discrepancy is probably due to secondary human-induced changes in habitats and landscapes, which have modified patterns of habitat use by birds and provided new opportunities for several species to colonize the island. In a recent study, Prodon et al. (2002) found that most species contract their habitat niche on Corsica, a result that runs counter to the classical tenet of island biology, as well as with previous results that also analysed the habitat niche of land birds on this island (Blondel et al., 1988; Pavoine et al., 2007). Prodon and colleagues compared the altitudinal extension of bird species along the slopes of mountain ranges from sea level to 2784 and 2710 m a.s.l. in the Pyrenees (southern France) and in Corsica, respectively. The study of Blondel and colleagues focused on the distribution of birds along seven stages of two forested habitat gradients, starting from very low matorral and ending with an old mature forest of the Mediterranean holm oak. In both studies, the mainland and island systems matched each other reasonably well, with a similar zonation of vegetation belts (for details and methods see Blondel et al., 1988; Prodon et al., 2002). Prodon and colleagues found that a majority of bird species contracted their habitat niche on the elevation gradient, with only a small proportion of them enlarging it, whereas Blondel and colleagues found the opposite on the vegetation structure gradient. The reasons why the two studies differ are interesting to analyse.

Following the taxon-cycle rationale (Ricklefs & Cox, 1978), species have been assigned by Prodon and colleagues to one of six categories according to their degree of morphological differentiation compared with their mainland counterparts. This scale, referred to as the ‘immigration-differentiation’ scale, was assumed to correspond to the time since the species colonized the island. Categories I and II included irregular breeders or species recorded on Corsica only within the previous 20 years; category III included undifferentiated regular breeders recorded on Corsica over more than a century; and categories IV and V encompassed species that are well differentiated at the subspecific level. Category VI corresponds to the two insular endemic species that have no mainland counterparts (the Corsican nuthatch Sitta whiteheadi and the Marmora’s warbler Sylvia sarda). Fossil remains have also been considered by Prodon et al. (2002) for assigning species to these categories, but this does not make sense in an island that is visited twice a year and in winter by tens of millions of migrants and wintering birds in the course of their migration between Africa and Eurasia.

The number of species in common in Corsica and on the mainland was 85 in the elevation gradient and 26 in the vegetation structure gradient. This large difference is due to two factors in the elevation gradient: (1) a larger diversity of species as a result of a larger variety of habitats; and (2) a large number of species with small groups of individuals that are highly localized and do not constitute established permanent self-sustainable populations. In contrast, all species in the vegetation structure gradient are members of well established communities that have evolved traits of the insular syndrome and are widespread everywhere in forested habitats of Corsica. The number of habitats per species, which is a proxy for measuring niche breadth (Cox & Ricklefs, 1977), is reported in Table 1. In Table 2 the niche space occupied by the species is expressed as the ratio of the size of this space on the island to that on the mainland (island/mainland amplitude ratio, IMAR; Prodon et al., 2002). Species enlarge their niche for a ratio > 1 and reduce it for a ratio < 1. Mean niche breadth for these two estimates of habitat selection differed considerably between the two studies, and was clearly much narrower along the altitudinal gradient than along the vegetation gradient. On average, species extended over 6.33 intervals of 200 m altitude in Corsica, compared with 8.43 on the mainland (Prodon et al., 2002), whereas species occupied 4.07 habitats on Corsica compared with 2.22 on the mainland (Blondel et al., 1988). With 66 species (78%) exhibiting an IMAR < 1, Prodon et al. (2002, p. 1294) concluded that ‘there was an overall trend in favour of range compression on Corsica’, which runs counter to the classical tenet of niche enlarging on islands. It could be argued that species respond differently to these two components of niche space (altitude and habitat structure), but the diversity of habitats along the elevation gradient is greater than that in the vegetation structure gradient, so species should be expected to exhibit an even larger niche breadth along this gradient.

Table 1.   Average number of habitats occupied by bird species along the two gradients ± 1 SD.
 Elevation gradient (number of 200-m elevation stripes per species)Habitat gradient (number of habitats per species)
  1. For the elevation gradient, the number of elevation stripes and bird species were 12 and 85, respectively (Prodon et al., 2002; original data in Ecological Archives EO83-021-A1). For the habitat gradient, the number of habitats and bird species were seven and 26, respectively (original data in Blondel et al., 1988).

Mainland8.43 ± 2.712.88 ± 1.58
Corsica6.33 ± 3.244.54 ± 1.79
Table 2.   Island/mainland ratios of niche breadth for species that occur in both Corsican and mainland habitat gradients.
CategoriesAltitude gradient*Vegetation gradient†
  1. Results ± 1 SD; numbers of species in parentheses. Categories refer to the degree of morphological differentiation of Corsican populations, which is assumed to be proportional to the time since the species colonized the island (see text; Prodon et al., 2002, p. 1297).

  2. *Data from Prodon et al. (2002).

  3. †Data from Blondel et al. (1988). Categories of the ‘immigration-differentiation’ scale are: I and II, irregular and/or recent breeding birds; III, undifferentiated regular breeders; IV and V, species with well differentiated subspecies; see text.

I0.29 ± 0.26 (11)1.60 ± 1.12 (15)
II0.67 ± 0.32 (7)
III0.82 ± 0.40 (49)
IV0.84 ± 0.13 (8)2.79 ± 1.68 (10)
V1.18 ± 0.25 (10)

Niche changes on islands have traditionally been considered as a proximate plastic response to the number of potential competitors: a population contracts its niche when meeting more competitors, and enlarges it when meeting fewer competitors. The competition paradigm has been hotly debated, but the basic rationale of niche changes on islands has not been fundamentally questioned. Whether the causal mechanisms that lead to niche shifts on islands result from changes in interspecific interactions such as predation, competition or parasite loads (e.g. density compensation; MacArthur et al., 1972), or another causal mechanism related to population-specific responses to new environments (e.g. population spillover resulting from density inflation; Crowell, 1983; Blondel et al., 1988), these shifts are evolutionary responses that require time to become established as population-specific life-history traits. In the framework of the taxon-cycle hypothesis, it has been argued that niche expansion is common in new colonizers, and that the new selection regimes prevailing on islands lead to local differentiation and ultimately to niche contraction as a result of local specialization. This suggests that niche breadth should be inversely proportional to the time since a population colonized an island (Ricklefs & Bermingham, 1999). Niche shifts have been discussed in the framework of a taxon cycle by Prodon et al. (2002), as components of evolutionary changes that characterize populations after they colonized an island, but their interpretation differed from the taxon cycle rationale. Steps of a taxon cycle do not involve niche enlarging, but instead a progressive reduction of range, specialization and eventually extinction (Ricklefs & Cox, 1978; Ricklefs & Bermingham, 1999). Following the rationale of insular evolution, one may predict that species that colonized an island a long time ago have a larger niche than recent colonizers. In both studies (habitat gradients and elevation gradients), species of categories IV and V (which have been pooled together in the vegetation structure gradient because of low sample size) exhibit a larger habitat niche than those of lower categories on the immigration–differentiation scale (categories I–III). Thus the studies of Prodon et al. (2002) and Blondel et al. (1988) came to the same conclusion for those (mostly forest) species that colonized the island a long time ago. The increase in mean IMAR along the five categories of increasing morphological differentiation suggests that species tend to enlarge their habitat niche with increasing time since colonizing Corsica (Prodon et al., 2002, fig. 4), which is the expected trend on an island.

Species that have the narrowest altitudinal niche in Prodon et al.’s (2002) study, and which occur as small localized populations, with many not breeding regularly each year, raise two crucial questions in island biogeography: What is an island population? and To what extent does immigration differ from colonization? These points have been addressed by Diamond & May (1977) in analyses of species turnover on the islands of Farnes and Skokholm off the coast of the British Isles. These authors showed that most ‘insular populations’ on these islands were actually fugitive members of mainland populations that failed to establish themselves on the island. As pointed out by Haila & Hanski (1993), isolated breeding pairs or small groups of birds that breed sporadically in small habitat patches cannot be considered as demographically functional populations belonging to an insular fauna sensuEbenhard (1991). Local groups of birds that breed on islands not far from the mainland are often parts of larger populations that occur on much larger areas. These erratic and ephemeral individuals do not contribute to insular faunal turnover, and cannot be considered as colonizers. Williamson (1981) and Haila et al. (1993) gave several similar examples of pseudo-extinction in continental habitat islands in the British Isles (Eastern wood, Surrey, England) and in the Finnish archipelago, respectively. In Corsica, no fewer than 15 species have been reported to immigrate to the island and to breed in one or two breeding attempts, but to have failed to become true colonizers (Thibault & Bonaccorsi, 1999; personal observation). These species, included in the list of Prodon et al. (2002), strongly contribute to ‘contracting’ the average niche of the species. Examples include Asio otus, Phoenicurus ochruros, Monticola saxatilis, Hippolais polyglotta, Sylvia hortensis, Phylloscopus bonelli, Phylloscopus collybita, Pica pica and Sturnus vulgaris, to cite just a few. All are recent ephemeral colonizers or vagrants (the so-called ‘tramp’ strategists of Diamond, 1974). They belong to categories I and II, and are localized or patchily distributed in the lower parts of the island. The presence on Corsica of these ‘tramp’ species is presumably a secondary feature mostly due to human-induced habitat changes that have made many new habitats suitable for them.

On the other hand, forest species have occurred on the island for a long time and many have differentiated as subspecies, involving a significant enlargement of their habitat niche, as noted in the two studies. They repeatedly followed the vegetation belts as the alternation of glacial and interglacial cycles moved up and down the limits of forests during the Pleistocene. Prodon et al. (2002) rightly pointed out that they evolved traits that allowed them to survive these climatic vicissitudes, as shown previously by Blondel & Farré (1988); Blondel & Vigne (1993), and confirmed by Pavoine et al. (2007). They extended over the whole range of forest belts, which were much narrower during full-glacial than during inter-glacial periods such as the present one (Beug, 1967; Reille et al., 1997, 1999) and enlarged their altitudinal range accordingly (see fig. 12 in Blondel, 1988). Most Corsican endemic subspecies or well differentiated forms are forest birds that experienced these shifts in altitudinal range in response to Pleistocene climatic changes.

Habitat heterogeneity and population changes: a reaction norm approach

Scaling down the effects of human action on living systems, this section examines the consequences of habitat heterogeneity and patchiness on life-history evolution at the population level. Landscape fragmentation is a consequence of ancient practices throughout the Mediterranean Basin, adding to the natural diversity of landscapes and habitats, and potentially producing new selection regimes to which populations may show evolutionary responses (Blondel, 2006; Blondel et al., 2006). If habitat fragmentation often threatens biodiversity by increasing inbreeding and extinction risks of small, isolated populations, the process is not necessarily detrimental to all aspects and metrics of biodiversity, as shown by the consequences of traditional land-use practices such as the ancient agro–sylvo–pastoral systems on alpha, beta and gamma diversity of plant and animal communities in man-made habitat mosaics (Blondel & Aronson, 1999).

Human impact on Mediterranean ecosystems has two consequences of great importance for wildlife. The first is the fragmentation of forests into patches of various sizes and degrees of isolation. The second is that most broad-leaved forests are dominated either by deciduous trees (e.g. downy oak, Q. humilis) or by evergreen sclerophyllous trees (e.g. holm oak, Q. ilex). This makes forest landscapes into mosaics of habitat patches dominated by one or the other of these two oak species, with more than 95% of the trees of each forest patch usually being of the same species. A long-term programme designed for investigating the response of blue tit (Cyanistes caeruleus) populations to these two features gave an insight into the huge phenotypic variation of this small passerine (Blondel et al., 2006). This man-made landscape design is a unique natural experiment, because the different physiology of the two oak species has many consequences for several components of food chains that depend on the oaks. A crucial element of the story is that the spring flush of leaves, measured as leaves developing from dormant buds to fully developed leaves, starts 1 month earlier in deciduous oaks than in evergreen oaks (Figs 1 & 2). Bud burst is a key event because it triggers the hatching of leaf-eating caterpillars, the preferred prey of blue tits, which will start to feed on young leaves. This event activates food chains, which cascade from oak leaves through caterpillars to insectivorous blue tits, and follow different temporal trajectories in the two types of oak forest. In addition to this large qualitative difference in timing, there is also a large quantitative difference. Caterpillars are much more abundant in the deciduous downy oaks because the whole foliage is renewed each year, as compared to only 30% renewal in evergreen oaks (Blondel et al., 1993). Only young leaves are edible by caterpillars, so the window of food availability is short, 2 weeks at most, which makes the best time for tit breeding rather short (Blondel et al., 1999). The 1-month difference in bud burst and leaf production between the two oak morphotypes results in an early- and high-amplitude peak in caterpillar production in downy oak forest, which contrasts with a late- and low-amplitude peak in holm oak forest (Fig. 2). This combination of large differences in phenology and abundance of caterpillars is crucial because food supply has repeatedly been shown proximately and ultimately to determine the values of breeding traits such as laying date, clutch size and breeding success in income breeders such as small passerines (Lack, 1968; Drent & Daan, 1980). The question arises: how do blue tits respond to these large differences in timing and abundance of food resources? To answer this question, we examined the phenotypic variation and reaction norms of a series of populations from an experimental design that included two landscapes with two deciduous habitat patches and three evergreen habitat patches each. One landscape was near Montpellier on the mainland, southern France, and the other on the island of Corsica. Each habitat patch was equipped with nest boxes for monitoring the breeding process of tits, and trays under the canopy of trees were used to collect the droppings of caterpillars and thus to quantify the amount of food available for tits (for details see Blondel et al., 1993, 2006). An important point that must be kept in mind is that deciduous oakwoods are more common than evergreen oakwoods on the mainland, whereas the reverse is true on the island.

Figure 1.

 Variation of laying date (± 1 SD) for 10 blue tit (Cyanistes caeruleus) populations in landscapes including habitat patches dominated either by deciduous downy oak (Quercus humilis) or by evergreen holm oak (Quercus ilex). Horizontal dotted lines denote the best date for starting to breed relative to food availability in deciduous downy oaks and in evergreen holm oaks, respectively. Laying dates, 1 = 1 March, 32 = 1 April, etc.). Black symbols, deciduous habitats; hatched symbols, evergreen habitats. Note the much higher variation of laying dates in Corsica than on the mainland (modified from Blondel et al., 2006).

Figure 2.

 Caterpillar abundance (curves) estimated by caterpillar droppings falling from oak leaves and number of blue tit (Cyanistes caeruleus) fledglings (histograms) in deciduous downy oaks (Quercus humilis) and evergreen holm oaks (Quercus ilex) on the mainland and on Corsica. Scale not given on left vertical axis because of the huge between-habitat variation of caterpillar abundance. (a) Between-landscape scale; (b) within-landscape scale on the mainland; (c) within-landscape scale on Corsica. See text (simplified from Blondel et al., 2006).

The following briefly summarizes some results of this study that are relevant to discussion of the evolutionary consequences for wildlife of human impact on landscapes. The dotted lines in Fig. 1 (laying dates, 1 = 1 March; 32 = 1 April, etc.) represent the best date for starting to breed in each type of oakwood. These lines represent the time when the leaf buds start to burst, triggering the hatching of caterpillars that become available as food for tits. On the mainland, only the two populations breeding in deciduous oaks can be considered as more-or-less ‘correctly’ timed on the caterpillar peak, but the three populations breeding in evergreen oaks obviously miss the best time to breed because they started too early in relation to caterpillars. On Corsica, three populations were correctly timed, two in evergreen and one in deciduous oaks, which made the phenotypic variation much higher on the island than on the mainland despite the fact that the two landscapes match each other fairly well (Fig. 1).

In a first step, we examined breeding patterns of tits in a deciduous oakwood on the mainland, where deciduous oaks dominate landscapes, and in an evergreen oakwood on the island, where evergreen oaks dominate landscapes. In both habitats, blue tits nicely synchronized their breeding time in such a way that the supply/demand ratio of food was optimal in the two habitats (Fig. 2a). On average, tits started to lay 3–4 weeks earlier in deciduous mainland oaks than in evergreen island oaks (6 April vs. 12 May); produced larger clutches (9.8 vs. 6.4 eggs); and raised on average seven fledglings in the food-rich deciduous oakwood compared with only four to five fledglings in food-poor evergreen oakwood. Common garden experiments have shown that the > 3-week difference in breeding time was genetically determined (Lambrechts et al., 1997).

In a second step, we examined the breeding patterns of blue tits in the two types of oakwood, but at the within-landscape scale on the mainland. We found that tits started to breed at least 2 weeks too early in the evergreen oakwood, so they missed the best period for raising their young (Fig. 2b). As a consequence, they produced offspring of poor quality, which had low prospects of being recruited into the breeding population. Several lines of evidence from molecular genetics (Dias et al., 1996) and quantitative genetics (Charmantier et al., 2004) have shown that this mistiming is a consequence of gene flow across habitat patches. Density-dependent mechanisms have resulted in birds that are specialized in deciduous oaks, where they developed habitat-specific adaptations, migrating into evergreen oaks. Such patterns resulted in a source–sink population structure that involved asymmetrical dispersal from the former (source) to the latter (sink). In the mainland landscape, where birds were assumed to disperse freely across habitat patches, little local differentiation of both breeding and morphometric traits has been found within a range of c. 40 km, supporting the hypothesis of gene swamping between populations (Blondel et al., 2001).

In a third step, a null hypothesis predicted that a similar maladaptation would occur in Corsica, but the other way round, with blue tits being nicely synchronised to the peak of caterpillars in the evergreen oakwoods (which are more abundant on the island) and maladapted to this peak in deciduous oakwoods. Interestingly, blue tits proved to be equally well synchronized to the caterpillar peak in the two habitat types, which were no more than 25 km apart (Fig. 2c). In addition, the two populations differed in a number of traits, including demographic and morphometric traits, suggesting a genetically based local adaptation (Blondel et al., 1999). This case study shows that local specialization may occur between populations at a scale that is much smaller than the scale of dispersal of the organism and potential gene flow. This difference between the mainland and Corsica has been interpreted as resulting from reduced dispersal ranges of birds on islands, and supports the divergence-with-gene-flow model of speciation (Blondel et al., 1999) that had been already proposed long ago by Maynard Smith (1966) and Felsenstein (1976). What explains the large difference between the mainland and the island is a greatly different gene flow/selection tension, with the effects of gene flow being counteracted by local selection on the island, but not on the mainland. Lower dispersal on the island, which is a component of the insular syndrome, explains that populations may respond to strong directional selection even if there is some gene flow. Reduced dispersal combined with assortative mating and oak-specific habitat selection makes birds more tightly linked to their local habitat on the island than on the mainland.

To summarize, from a reaction-norm approach (Kawecki & Stearns, 1993), mainland and Corsican populations of blue tits belong, so to speak, to two genotypes, a ‘deciduous genotype’ on the mainland and an ‘evergreen genotype’ on the island of Corsica. The reaction norms of these genotypes do not cross, and correspond to a genetically determined local specialization to local selection regimes. In addition to this local specialization, there is some phenotypic plasticity both in time and in space. Plasticity in time is expressed by the proximate year-to-year variation of breeding time depending on yearly variation of temperatures and food (standard deviation around the mean laying date). Plasticity in space results from the fact that tits always breed earlier in deciduous than in evergreen oakwoods, whatever the region, and always earlier in the wild than in aviaries, whatever the dominant tree species.

Conclusion

Eight millennia of human impact on ecosystems have resulted in dramatic changes in the composition and structure of landscapes in Mediterranean islands. These changes were probably more important (and easier to demonstrate) on islands than on the mainland because islands are less resistant and resilient to disturbance than mainland areas. They deeply modified the distribution and structure of habitats and opened the route to invasions by new colonizers, which may lead to misleading conclusions about habitat selection and niche shifts on islands.

After so many centuries of co-habitation between humans and nature, most Mediterranean ecosystems are so inextricably linked to human interventions that evolutionary responses to human-induced changes are difficult to decipher. There could have been a myriad of case studies exemplifying evolutionary processes such as those described in this paper with the blue tit model. Such studies are badly needed because a detailed analysis of potential evolutionary responses of organisms to human-induced changes could help evaluate the tempo and mode of the forthcoming response of organisms to the many changes that will occur in the Mediterranean in the future. According to scenarios of the IPCC (2001), the Mediterranean region will be particularly affected by global warming, with a strong decrease in rainfall.

Studies on habitat niche breadth in Corsica showed that species do not all respond in the same way to environmental changes. This means that sorting processes may shape new species assemblages as a result of changes in probabilities of extinction and colonization as well as changes in the migratory behaviour of many species. On the other hand, the blue tit studies provide good evidence that micro-evolutionary changes can occur rapidly in fitness-related traits with high heritability values, such as those related to breeding processes in birds (Visser et al., 2003; Charmantier et al., 2004) or flowering time in plants (Peñuelas et al., 2002). Palynological studies have shown that most lowland forested areas in Corsica were covered by deciduous trees before humans destroyed them. Blue tits responded within a few generations to the substitution of tree species, and succeeded in adapting to the new human-made design of landscapes. Such rapid evolutionary responses suggest that birds will be able to cope with the new environmental conditions associated with global warming.

Once again, islands shed an interesting light on several aspects of evolutionary responses to environmental heterogeneity. In particular, the much higher phenotypic plasticity of blue tits observed on Corsica than on the mainland demonstrates that species impoverishment in islands may, to some extent, be compensated by an increase in intraspecific diversity, as shown by the high differentiation of populations at small spatial scales. Comparing the responses of less-dispersive island birds and highly dispersive mainland birds to the same spatial diversity of habitat gives us a practical demonstration of the relationships between dispersal, spatially variable selection and local adaptation.

Acknowledgements

I express my warmest thanks to the organisers of the third Biennial Conference of The International Biogeography Society and especially to Robert J. Whittaker and José María Fernández-Palacios for inviting me to give the talk on which this paper was based. I warmly thank José María Fernández-Palacios and two referees for their very useful comments, suggestions and corrections on a first draft of the manuscript.

Biosketch

Jacques Blondel is Director of Research emeritus at the Centre d’Ecologie Fonctionnelle et Evolutive, CNRS, Montpellier, France. His main interest is the biogeography, evolutionary ecology and population biology of birds at different spatial scales, from the whole Mediterranean Basin to local habitats, focusing on island–mainland comparisons.

Editor: José María Fernández-Palacios.

Special Issue: This paper arose from a talk presented at the third biennial meeting of the International Biogeography Society, held in Puerto de la Cruz, Tenerife, Canary Islands, 9–13 January 2007.

Ancillary