HBW 16 - Foreword on Climate Change and Birds, by Anders Pape Møller

View foreword in PDF format: 



Most ornithologists older than 40 and living in the northern hemisphere will be able to remember a time just a couple of decades ago when spring migrants arrived considerably later, and many species that now breed at high latitudes did not do so. When these ornithologists are polled about their concerns for the future, they identify climate change as one of the most important issues. Climate change is currently occurring at an unprecedented rate, with severe consequences not just for birds but for humans and all other living beings alike. The immensity of the effects of climate change is on a scale that defies imagination, from the extensive current melting of the ice in Greenland and the Antarctic to the retreat of numerous glaciers around the world (Figure 1).

Sea levels rose rapidly by 15–20 cm during the 20th century (Figure 2), with dramatic implications for hundreds of millions of people living in coastal areas; future projections of sea level rise vary from a few tens of centimetres to several metres, depending on the rate of ice melt in Greenland and the Antarctic. Global temperature increased by 0·7°C during the last century, but hardly at all during the previous 900 years (Anon. 2007c). Reconstructed climate variables covering more than 1000 years clearly show that climate change in recent years has been unprecedentedly rapid, with concomitant biological change. Further to climate warming, additional effects of climate change include an increased frequency of droughts, hurricanes and altered seasons. These changes are attributed by the majority of scientists to the effects of man-made changes to the atmosphere due to greenhouse gases (Hurrell & Trenberth 2010), with more than 98% of scientists with the largest scientific impact agreeing on this assessment (Anderegg et al. 2010). Thus, so-called climate sceptics constitute a tiny minority of scientists with little or no standing in the scientific community at large.

Climate change biology is a young science. The first paper that addressed this issue for birds, by Peter Berthold, only appeared 20 years ago (Berthold 1991). Since then there has been a dramatic increase in research on this topic, and almost 300papers on birds and climate change were published in 2010 alone. Most of this research is purely descriptive, which is natural for a phenomenon where the basics still need to be fully elucidated. However, a purely descriptive approach raises concerns about causation versus correlation, and the possible impact of unknown variables on what might at first sight look like causal relationships. Climate change has increased linearly in many parts of the world, but so too have numerous other factors, such as intensification of agriculture, forestry and fisheries, expansion of urban areas, and release of pollutants. Caution needs to be taken in investigating the role of climate change alone, since any of these factors on their own or in combination may also affect living organisms.

Because of its obvious utility to decision-makers and politicians, who want to know the consequences of climatic change, research to date has been increasingly based on predictions or projections of future climate. This may be problematic, because models often only include a single climatic variable as a driver, making it unlikely that the projections will fit any future empirical observations. Especially given the influence of climate change sceptics, it is important to make careful conclusions that can face reality. Climate change research being shown up as ill-substantiated or tendentious will not benefit anyone: scientists, amateurs, decision-makers and certainly not the general public who must cope with a changing world. It is with this perspective in mind that I have written this overview of what we know and do not know about birds and climate change.

In this review, I first describe the major recent changes in climate that have already occurred, including extreme climatic events, to set the stage for the biological phenomena. Second, I review information on the annual cycle of birds and its component parts in relation to climate change, describing its effects on the start of the reproductive season, nest-building, egg-laying date, clutch size, incubation and nestling periods, number of clutches, reproductive failure, temporal mismatch between reproduction and food availability, duration of the breeding season, and dispersal. Third, I present an extensive review of the effects of climate change on bird migration, emphasizing timing and duration of spring and autumn migration, and recent changes in arrival dates (their distribution, consistency and evolution, and how they may be influenced by geographical variation, breeding and winter effects, phenotypic plasticity and sex differences). Fourth, I describe how morphological traits such as secondary sexual characters, colour and body size have changed in response to climate change. Fifth, I assess to what extent the responses of individual birds to climate change may have effects at the population level, including population dynamics, interspecific competition, predation, host-parasite interactions and extinction risk. Sixth, I analyse studies of range expansion including observed and projected changes in distributions and range margins, and dynamics at the edges of populations. Seventh, I discuss changes in bird communities, and how generalist species have become predominant and communities are becoming ever more impoverished. Finally, I relate climate change to conservation issues by identifying which species are particularly susceptible to climate change, and briefly suggest means by which climate change effects can be mitigated and hopefully reversed. I have tried to find examples from a broad geographical range of species and systems. However, long-term time series data on migration, breeding and many other aspects of the life of birds are specialities of the subtropical, temperate and arctic climates of the northern hemisphere, and as yet few data are available from the tropics and the southern hemisphere.


Climate in a changing world

Climate has always changed on the long time-scale of millennia, but rarely so much and so rapidly as during the last century (Figure 3). This is disturbing, because while all living beings should supposedly be able to adapt to gradual climate change, the current speed of change may exceed what is possible for adaptation both by natural selection, and by plastic adjustments of individuals to changing conditions during their lifetime (known as phenotypic plasticity). During the ice ages, climate changed due to changes in precipitation caused by shifts in the Intertropical Convergence Zone, and this sometimes occurred at an abrupt pace, with temperatures increasing by 2–4°C from one year to the next (Steffensen et al. 2008). Current climate change, by contrast, is mainly linked to CO2 and other greenhouse gases such as methane, and has caused an overall climate warming of 0·7°C during the last 100 years. However, temperature increases are unevenly distributed across the globe, and increases are particularly dramatic at high altitudes and latitudes (Figure 4). Hurrell & Trenberth (2010)provide an extensive overview of climate change, and interested readers can find further references to different aspects of climate change in that source.

Climate change has been particularly marked in the Arctic. The diminution of the Arctic ice sheet since the 1970s has been unprecedented, with an average decrease of7·4% per decade during 1978–2005. Since then, several years have set records in terms of ice melt, with a 20% reduction in ice cover in 2007 compared to 2005, which was the warmest year since measurements started in 1850. The effects of warming have been particularly pronounced for species dependent on or associated with sea ice. Likewise, effects of climate change in the Antarctic Peninsula have been marked; interestingly this has not been so for the Antarctic south of 65° S, where temperature has not shown any clear trends to date. In the Antarctic, changes are largely driven byt he Southern Oscillation which governs sea surface temperatures and sea ice extent in he Southern Ocean, which in turn affects abundance of krill. The Southern Oscillation Index has shown a regime shift since the 1970s, resulting in less sea ice, with dramatic consequences for marine birds (Masson-Delmotte et al. 2003).

Climate change in the temperate zone has been almost as dramatic as in the Arctic. Temperature increases per decade have occurred unevenly in different geographical regions, and have been particularly strong in parts of Western Europe, Mongolia and China, but moderate or absent in large parts of North America, Eastern Europe and Russia. Decreases in snow cover in many regions of the northern hemisphere, especially in spring, have advanced spring phenologies. Precipitation has been increasing in eastern parts of North and South America, Northern Europe and Northern and Central Asia, while increasingly dry summers in Russia and North America have increased the frequency of forest fires. In the subtropics, climate change has mainly affected temperature and precipitation: many areas are experiencing increasingly dry conditions and accelerating desertification.

Parts of the Mediterranean basin, the Sahel zone of Africa, southern Africa and southern Asia have all reported rises in desiccation.

In the tropics, climate change has been relatively modest in terms of temperature change. However, changes in sea level and the frequency of tropical storms are likely to have a negative effect on communities of birds in coastal areas, including many oceanic islands which have a disproportionately high fraction of endemic species. Indirect effects of climate change on birds in the tropics will arise as a consequence of displacement of human populations living in coastal areas, in concert with a projected expansion of such human populations.

Climate change may have its most profound effects on islands, where birds and other organisms have no other place to go. This is particularly serious because island bird communities contain disproportionately large fractions of endemic species. Many islands will become inundated completely owing to the rise in ocean levels, although rises in seawater levels will be uneven due to differences in water temperature. The effects of climate change will also be particularly acute on mountain tops, where high-altitude species similarly have no other place to go once low-altitude vegetation and its associated fauna move up. Reductions in glaciers and ice caps were dramatic during the 20th century, and during the next century many glaciers in the tropics will disappear entirely.

Extreme climate change: Is it just more of the same, or is it different?

While there is no doubt that climate has changed dramatically during the last century, and in particular during the last couple of decades, there has been less research emphasis on extreme climate change. Not only will average temperatures and precipitation change, but extremes are predicted to become more frequent and pronounced. Extreme climatic events are defined by climatologists as rare events that occur at a frequency of 5% or less of the time as judged from the expected distribution of the climate variable (Anon. 2011b), and include droughts, heat waves and hurricanes. All of these are predicted to become much more common in most future climate scenarios (Anon. 2007c, Hurrell & Trenberth 2010). Already the frequency of heat waves has increased over the last 50 years, and there has also been a rapid increase in the number of warm nights. These changes have consequences for the tropical cyclones that since 1970 have increased in frequency and duration, although it needs to be borne in mind that older data before satellites became common may be of less rigorous quality (Trenberth 2007).

A consequence of extreme drought is increased risk of forest fires, as seen in large parts of Russia and Ukraine during the summer of 2010. The scientific study of the biology of extreme climatic events is in its infancy, precisely because such events by definition are rare. The obvious question arising is whether extreme events just constitute a continuation of the distribution of more commonly encountered events, or whether extreme events represent qualitatively different conditions that affect birds and other animals in a completely novel way (Møller 2011b).

Extreme climatic events are well known to biologists, as shown by effects of a violent snowstorm on House Sparrows (Passer domesticus) (Bumpus 1899) and powerful El Niño events on Darwin’s finches (Grant, P.R. & Grant 2002a). Both of these events completely changed the phenotypic composition of the bird populations concerned, although they ultimately reverted to their former condition. Another common example is heavy mortality during migration (Newton 2007), as seen in Barn Swallows (Hirundo rustica) when passing mountain ranges in autumn, with dramatic consequences for life history and behaviour (Møller 2011b). Moreno & Møller (2011) recently reviewed the life history consequences of such extreme events and found that under normal conditions, complete reproductive failure occurred on average in 27%of breeding attempts; however, this rose to 79% during extreme events (Figure 5).Likewise, adult survival rate was three times worse during extreme events than under normal conditions. These findings clearly show that the intensity of selection is dramatically increased during extreme events, because only a small fraction of individuals will survive and reproduce and hence contribute to the next generation. Under such extreme natural selection, whether an individual survives or not often depends on the size of a given morphological character (known as “truncation selection”); for example, wing length may dictate survival during severe conditions on migration. This kind of selection has particularly strong effects on the composition of the population after a selective event.

Extreme climatic change may also increase variation in demographic parameters such as fecundity and survival. This has the effect of reducing the overall population rowth rate, even if mean rates of fecundity and survival remain unchanged. For example, European Shags (Phalacrocorax aristotelis) are coastal seabirds with high variation in annual survival rate caused by mortality due to severe winter storm (Frederiksen et al. 2008). A 43-year study revealed that survival rates varied greatly among years, especially in second-year birds and adults which were greatly affected by years with the strongest winds and heaviest rainfall. Shags and other cormorants do not have fully waterproof plumage as do other seabirds. This allows them to dive more efficiently, but also renders them very susceptible to the effects of cold and wet weather. Extreme weather conditions are predicted to increase in frequency, forecasting future population declines.

A particularly illuminating example of an extreme event is the heat wave that affected most of Western Europe, and France in particular, during 2003. This qualifies as an extreme event since temperature anomalies of up to +6°C occurred during April–August. It resulted in severely reduced primary productivity across Europe (Ciais et al. 2005) and an excess mortality of 2600 humans in France alone (Hémon & Jougla2003). Intriguingly, bird species that were declining before 2003 declined particularly strongly in 2003, while increasing species increased particularly strongly in 2003 (Jiguet et al. 2006). It is therefore interesting to ask whether the natural thermal range of different species influenced how they responded to these temperature anomalies. To answer this question each species’ thermal range was first determined from the range of temperatures experienced in the 50 warmest and 50 coldest grid cells within its European breeding distribution. This showed that bird species with the smallest thermal range suffered the largest declines in population size between 2003 and 2004, following the temperature anomalies of the summer of 2003 (Figure 6). These findings imply that the thermal ranges of birds reflect the past temperature fluctuations hey have experienced in their evolutionary history, and thus their thermal adaptation to the environment. These findings also clearly imply that species with narrow thermal niches are particularly likely to be negatively affected by future heat waves. Indeed, modelling exercises suggest that future heat waves may lead to reductions in abundance and perhaps extinctions in many of the warmer parts of the world (McKechnie & Wolf 2009).

Effects of climate change on the annual cycle of birds

Numerous examples illustrate how dramatic changes have already occurred the in timing of reproduction and other annual events. These changes include an earlier start to singing by breeding birds in spring, advancing dates of migration and breeding, longer annual breeding cycles, and altered patterns of departure from the breeding grounds and, hence, autumn migration. Many birds can now be seen at times when and in places where they were previously never to be found. These observations raise questions about the organisation of the annual cycle, which is the subject of the next section. This is followed by a review of how climate change can affect when and where birds sing, as well as other aspects of the breeding cycle. Finally, this section is rounded off by discussing how climate change has caused reproductive timing to interfere with the other two major components of the annual cycle, migration and moult.

How rigid is the annual cycle and what determines the duration of its component parts?

All living organisms must schedule their life histories to fit within the time they have available. In birds this is reflected by the relative duration of reproduction, moult and migration in the annual cycle. Because the duration of each of these components increases with body mass, large-bodied species may come up against a threshold value at one year. If they are unable to fit these activities in less than a year, large species are forced either to reproduce every other year, to moult less than once a year, or partially to overlap migration and breeding with the start of gametogenesis (the production of gametes). The last mentioned means that sperm and egg production occurs during migration or arrival at the breeding grounds, while retaining energy reserves that allow an early start to reproduction (Hedenström 2006, Wingfield 2008). These tradeoffs between different activities suggest that the timing of events in the annual cycle may be constrained, particularly in species with long migrations, moults or reproductive cycles. Species with very slow developmental rates of offspring, species with slow rates of moult (such as aerial foragers) and species with very long migrations may particularly suffer from constraints on the timing and duration of life history events. Migrants are further constrained by typically short breeding seasons imposed by seasonal environments, especially when producing two or more clutches.

Moult and breeding, and moult and migration, rarely overlap, owing to the extreme costs of each activity. At northern latitudes, however, moult and breeding may overlap, and some passerines even become partly or fully flightless while breeding (Haukioja 1971b). On the other hand, some species of raptor, and other large-bodied species with prolonged moult periods, do extend moult across two or more years, apparently because fast moult is incompatible with prey capture (Rohwer, Ricklefs et al. 2009).These examples underscore the severe constraints on the timing of annual events.

The timing of moult determines when subsequent components of the annual cycle can take place. For example, migration with moulting feathers is energetically much more expensive than migration with full plumage. We currently know very little about changes in the timing of moult by migrants when their timing of migration has changed. Superficially, we should expect that earlier migration would allow moult to start earlier. However, Barn Swallows of Eastern European breeding origin wintering in South Africa have delayed their moult during the last eleven years, resulting in a delay in timing of spring migration (Møller, Nuttall et al. 2011). This would prevent an early departure from the South African winter quarters. The timing of the start of moult (as well as the maturation and regression of gonads) is under photoperiodic control, and therefore should not change if the photoperiod remains unchanged in the migrants’ winter quarters. That is the case even when a photoperiodic response is fine-tuned by factors such as temperature at the breeding grounds (Dawson 2005, 2008). Moreover, migratory responses to climate change could be constrained by genetic correlations between timing of egg-laying and timing of autumn migration (Coppack et al. 2001), and between onset of autumn migration and timing of juvenile moult (Pulido & Coppack 2004). This is because these genetic correlations mean that selection for earlier laying would simultaneously result in selection for earlier autumn migration. Therefore, there are multiple mechanistic constraints on whether and to what extent the timing of moult can respond to climate change.

Start of the breeding season and singing

The start of the breeding season is commonly ascribed to increasing levels of hormones and their influence on reproductive behaviour such as singing and other sexual display. However, the empirical evidence suggests that these mechanisms can evolve to change in response to earlier springs. Schmidt & Hüppop (2007) analysed trends in first song dates of 56 resident and migrant bird species for 43 years during a period of rapid climate change. First song dates advanced by 0·3–0·6 days per year, similar to the advances in arrival date of migrants. A key conclusion from this unique dataset is that both migrants and non-migrants responded to climate change by advancing singing date. This suggests that migration appears not to impose a constraint on response to climate change by preventing long-distance migrants from responding to climatic changes at the breeding grounds until after spring arrival. Rubolini et al. (2010) compared changes in onset of singing in spring in short- and long-distance migrants using the same dataset. Species that migrated a distance of less than 4° latitude (450 km) advanced their first song dates more than long-distance migrants (Figure 7). This advancement was explained by increasing spring temperatures. However, species laying two or more clutches per year showed a stronger advance in first singing date than species laying only a single clutch. This result implies that more than a single clutch per year imposes a stronger selection pressure for earlier start of reproduction. Interestingly, the effect of temperature on first song date was unrelated to migration status. Taken together, these findings imply that the difference in the temporal trend between short- and long-distance migrants was mainly due to environmental cues regulating departure of migrants from the wintering areas, rather than to differential responses of first song date to local temperatures.

Climate change may also affect the environment that birds use for singing. Birds typically sing from exposed positions in the vegetation even when such positions make them vulnerable to predators. An important consequence of recent climate change is that bushes and trees are leafing much earlier in spring than just a couple of decades ago. This has consequences for the choice of song-posts, because early-sprouting foliage imposes greater interference to sound transmission, making it more favourable to sing from higher positions (Møller 2011a). Indeed, the average song-post height of various species of birds increased by 18% (or 1·2 m) between 1986–1989 and 2010 in a Danish study. These changes were absent for species singing in the herb layer, moderate for species singing in bushes, and most pronounced for species singing in trees, as expected from the effect of earlier leafing. In addition, species with increasing population trends and species with sexually dichromatic coloration (and hence more intense competition for mates) were those that increased song-post height the most. These findings suggest that climate change can affect vocal behaviour of birds not only in terms of phenology, but also in terms of habitat use.


Nest-building takes up a significant amount of time in the reproductive cycle. A study of 200 European species found that females, males or both sexes spend from two to 105 days on the activity. Nests are sometimes much larger than required for containing and protecting eggs and young; nests of species in which both sexes contribute to building are on average twice as large as those of species where females build singlehandedly (Soler et al. 1998). This implies that in some species nest size and building plays a role in mate choice and sexual selection. In the Barn Swallow, pairs consisting of a male with a short tail that is less attractive to females build larger nests than pairs with a long-tailed male. Interestingly, nest size has decreased over time by a factor of three owing to the negative effect of climatic conditions in North Africa on the survival of short-tailed, poor-condition males, which build larger nests over longer periods than long-tailed, good-condition males (Møller 2006). These findings suggest that not only is nest-building time-consuming and likely to constrain the speed of breeding, but also that it may be influenced by non-random selection imposed by climate change.

Laying date

Timing is everything when it comes to reproduction. Since timing of migration and timing of singing has advanced in response to climate change in many different bird species, it is not surprising that equally many species have started to lay eggs earlier, as in the striking case of the Northern Lapwing (Vanellus vanellus) (Both et al. 2005). Dunn & Winkler (2010) reviewed the literature and showed that 59% (against the random expectation of 5%) of 68 species mainly from Europe and North America have significantly advanced their laying date. Likewise, 79% advanced their laying date during the last decades when temperatures increased. The rate of advance was 0·13 days per year. This can be compared to the rate of advance of 0·39 days per year for phenology in general in the UK (Thackeray et al. 2010) and a rate of advance of 0·37 days per year for first arrival date of migrants (Lehikoinen et al. 2004). Because the rate of change for laying date is only a third of the rate of change for migration and for phenology in general, and because migration phenology is delayed compared to general phenology, these data suggest that the breeding date is lagging severely behind.

While there is general consistency in the advancement of laying date among species, implying that the same species respond similarly at different sites, there is also additional variation between populations of the same species. Several studies have shown heterogeneity in response between populations, with those encountering the largest degree of increase in temperatures also showing the strongest responses (Dunn & Winkler 1999, Sæther et al. 2003, Visser et al. 2003, Both et al. 2004, Both & te Marvelde 2007), even at very local scales of a few kilometres (Møller 2008b). This effect of temperature is causal, since experimental increases in temperature advance the breeding date of captive birds (Visser, Holleman & Caro 2009). Therefore, selection for earlier laying should increase as optimal conditions for breeding advance due to climate change.

Much of the variation in the advancement of laying date lies among species (rather than among different populations of the same species, for example), and some of this can be explained by differences in ecology. The single most important factor explaining advancement in laying date is the number of clutches produced per year (Dunn & Winkler 2010). While species with a single clutch advanced their laying date by 0·19 days per year, species with multiple clutches only advanced theirs by 0·06 days, or less than a third. This has been clearly shown for tits (Visser et al. 2003, Husby et al. 2009). This difference is expected because laying two or more clutches constitutes a more difficult problem, in terms of adjusting laying date to the peak of food abundance, than does laying a single clutch. Another ecological determinant of advancement in laying date is trophic level. Across all organisms, species at higher trophic levels, such as consumers of herbivores (predators) and super-predators (predators of predators), have shown slower phenological responses to climate change than species at lower trophic levels (Thackeray et al. 2010). Indeed, in birds, Dunn & Winkler (2010) showed greater advancement in laying date for herbivores (0·45 days per year) compared to insectivores (0·11 days per year) and carnivores (0·14 days per year), although these differences were not statistically significant (see also van der Jeugd et al. 2009). Surprisingly, there was no difference in advancement of laying date in relation to migratory behaviour. Residents and short-distance migrants advanced their start of laying at a rate that did not differ from that of long-distance migrants. This lack of difference in laying date with respect to migration behaviour may appear to be contrary to what would be expected from the observed delay in arrival for long-distance migrants caused by climate change (Lehikoinen & Sparks 2010), and future research will be needed to resolve the effects of climate change on the simultaneous evolution of migration and breeding phenology.

It is usually advantageous to lay early, because early hatching coincides with the annual food peak while late laying prevents parent birds from benefiting from the annual peak in food abundance. In addition, early reproduction results in more time for offspring to develop independence before migration and/or moult. These benefits imply that there is a greater fitness advantage to early laying, and many studies have found such an affect. However, the pattern is not clear-cut. Dunn & Winkler (2010) reported that selection for early breeding has increased in recent years in only half the studies to date, in species as diverse as European Pied Flycatchers (Ficedula hypoleuca) (Figure 8; Both & Visser 2001), Great Tits (Parus major) (Visser et al. 1998, Charmantier et al. 2008), Barnacle Geese (Branta leucopsis) (van der Jeugd et al. 2009), Common Murres (Uria aalge) (Reed et al. 2009) and Arctic Terns (Sterna paradisaea) (Møller et al. 2006a). However, equally many studies have shown no evidence of change in directional selection (review in Dunn & Winkler 2010).

What evolutionary mechanism underlies such changes? Shifts in laying date could result either from evolutionary change (changes in the genetic composition of the population), or through phenotypic plasticity (individuals changing their behaviour according to environmental conditions). The foregoing studies have generally suggested that despite the fact that natural selection favours an advancement in laying date, there is no evidence of evolutionary change, and responses to selection have been entirely phenotypically plastic (Charmantier et al. 2008).

Egg-laying and clutch size

While the literature on phenology is extensive, there is much less information on change in clutch size. Because early laying and large clutches tend to go together both within and across species, we should expect increasingly earlier laying to be associated with increasingly larger clutch size. However, while studies of European Pied Flycatchers and Barn Swallows have shown such increases (Figure 9; Winkel & Hudde 1997, Møller 2002), studies of Collared Flycatchers (Ficedula albicollis), Great Tits and Common Blue Tits (Cyanistes caeruleus), Tree Swallows (Tachycineta bicolor), Eastern Bluebirds (Sialia sialis) and Red-winged Blackbirds (Agelaius phoeniceus) have not (Winkel & Hudde 1997, Winkler et al. 2002, Sheldon et al. 2003, Torti & Dunn 2005). Dunn & Winkler (2010) suggested three explanations for a lack of increase in clutch size. One possibility is that in warmer years it is easier for individuals to lay early, but this shift is not associated with an increase in clutch size because the association between clutch size and laying date has changed. Second, laying date may advance without clutch size increasing if a larger fraction of females of poor quality (who are unable to achieve large clutches) enter the population under such benign environmental conditions. A final possibility is that laying date has a greater impact on total fitness than clutch size, due to the negative effects of mistiming laying in relation to the food peak. Therefore, laying another or a few more eggs would delay hatching and run the risk of further mistiming. There have been no attempts yet to evaluate these alternative explanations.

An additional consequence of climate change for egg-laying is that the difficulty of finding sufficient resources to produce eggs should ameliorate as spring weather becomes warmer. One measure of this difficulty is the incidence of deviations from the normal rule of one egg being laid daily, since birds may skip a day if insufficient resources for egg production are available. Indeed, there was a decrease in laying interruptions in the Common Blue Tit by 0·58 days per egg during 1979–2007 (Matthysen et al. 2011). By contrast, no significant change was detected in Great Tits, as expected since a Great Tit’s clutch is a smaller proportion of its body weight and therefore less costly to produce.

Duration of incubation and nestling periods

Changes in the duration of developmental periods may also allow birds subtly to adjust the timing of breeding in relation to changing environmental conditions. The incubation period is largely determined by the developmental requirements of the embryo, which should be fixed in relation to climate change. Nonetheless, one study suggested that the incubation period of Great Tits has increased by on average two days in response to climate change; together with clutch size and onset of incubation, this enabled tits to fine-tune hatching date in relation to food peaks (Cresswell & McCleery 2003). Interestingly this effect was not statistically significant in a study of Common Blue Tits and Great Tits in Belgium (Matthysen et al. 2011), showing that different populations do not respond to climate change in the same way.

The only evidence to date on changes in nestling period also comes from tits: Matthysen et al. (2011) showed that in recent years, the duration of the nestling period has on average decreased by 1·5 days for Common Blue Tits, and by 1·3 days for Great Tits. Obviously neither the incubation nor nestling periods can dramatically adjust timing of breeding to altered phenology of food, although they may be contributing factors, as shown for incubation in Great Tits.

Number of clutches

Many species lay more than a single clutch per year, because the season is sufficiently long to permit multiple reproductive events. However, in order to benefit from a single annual food peak, the timing of these clutches may have to be relatively early for the first clutch and relatively late for the second clutch, compared to an optimally timed single clutch (Crick et al. 1993). This should particularly be the case for residents but less so for migrants, which are constrained by the twice-yearly migrations between breeding areas and winter quarters. If the problem of fitting two (or more) clutches into a single breeding season is an important factor determining reproductive success, then increasing spring temperatures should allow birds to space their clutches better, without sacrificing their own survival prospects. Indeed, warming springs have allowed Barn Swallows to start breeding earlier in recent years, thereby increasing the duration of the interval between first and second clutches by more than ten days during a period of only 40 years (Figure 10; Møller 2007a). This less “hectic” breeding schedule, with a longer time between the start of the first and the second clutch, allows females to recover from the large investment in first clutches. Pairs with a longer inter-clutch interval can increase their reproductive output without compromising the survival of reproducing females. This study shows that there is scope for changes in the timing of multiple reproductive events when climate changes.

Can climate change also influence the incidence of second clutches, as well as their timing? There is marked geographical variation in the frequency of second clutches among species, and even among populations within species. For example, the frequency of second clutches in the Great Tit ranges from zero in high-latitude populations, and populations in the UK, to more than a third in some other populations (Sanz 2002). Some studies have shown a decreased incidence of second clutches with increasing temperatures, while others have not. As noted above, populations that lay two clutches have to optimise the timing of laying of these two clutches, causing the first to be too early and the second to be too late for the main single food peak. For example, caterpillars are the main food source for breeding tits. Increasing spring temperatures will advance the development rate of caterpillars, causing an advancement in the date of the food peak, but also a reduction in the width of the food peak, thereby reducing the reproductive value of second broods. In this situation, when the width of the food peak is reduced, females should adjust their laying date of the first clutch to the food peak, delaying the laying date relative to the optimum when two broods are attempted, and in that way counteracting the effect of temperature on advancement of laying date due to advancement of the food peak. Visser et al. (2003) predicted that laying date should advance in populations with either no second clutches or a stable proportion of second clutches, because the first clutch laying date is not a compromise between the laying date of the first and the second clutch. They also predicted that laying date will have advanced to a lesser extent or not at all in populations with a decreasing frequency of second clutches. That was indeed the case for both Great Tits and Common Blue Tits: populations that experienced a decline in the proportion of second clutches experienced greater advancements in laying date (Figure 11). These findings suggest that exactly how birds might optimally adjust laying date to climate change depends on the trade-offs between optimal timing with respect to the food peak, and the frequency of multiple clutches.

However, the patterns just described for tits may not be general. Barn Swallows and Northern House Martins (Delichon urbicum) show large differences in the frequency of second clutches among European populations, declining markedly with increasing latitude (Møller 1984). Although the breeding dates of both species have advanced considerably during the last 20 years associated with climate change, there has been no change in the frequency of second clutches whatsoever (Møller 2002 and unpublished data).

Reproductive success

While there are many data on changes in phenology and few data on changes in clutch size, there are even fewer data relating changes in reproductive success to climate change. An earlier start of reproduction and a longer growing season may reduce costs of reproduction. Thus under certain circumstances climate change may even improve reproductive success: as noted above, warming spring temperatures have enabled an increase in the interval between first and second clutches in Barn Swallows, and Møller (2007a) has shown that reproductive success almost doubled between clutches with the shortest and longest intervals between first and second clutches. Female Barn Swallows that invest more in reproduction than males benefitted from this longer breeding season and longer interval between clutches by no longer paying a viability cost of producing a second clutch, as they did just a couple of decades ago.

Reproductive success may also be affected by climate change in unexpected ways. In the Barn Swallow, the frequency of unmated males has decreased through improved male survival caused by improved environmental conditions in North Africa during spring migration (Møller 2004b). Unmated males can secure reproductive success through infanticide at nests that are not well guarded by the owners, causing a divorce and subsequent remating by the female with the infanticidal perpetrator. Incidence of infanticide has decreased with climate change in North Africa (Møller 2004b). Because fewer males remain unmated now than before climatic changes in North Africa, the frequency of infanticide has changed from affecting 5% of all nests to being virtually absent in recent years, thus improving reproductive success with respect to this parameter (Figure 12; Møller 2004b). Thus, the effects of climate change on behaviour and reproductive success are often idiosyncratic and difficult to predict.

Mismatch between food, breeding date and optimal timing of reproduction

Animals typically breed at times when the probability of successful reproduction reaches a maximum (Lack 1954). Therefore, breeding is timed such that food requirements for offspring are typically maximal when food availability peaks. There are numerous examples of such a match between supply and demand, since selection has favoured those individuals that use cues that ensure breeding at an optimal time, and therefore leave the most descendants to the next generation. A classical example of reproductive match between timing of breeding and food supply is provided by Great Tits and their caterpillar prey, which show a marked peak in abundance in spring well after laying, when chicks are being fed (Figure 13).

Obviously, however, matching reproductive phenology to food availability represents a challenge, mainly because the phenology of more than one organism is involved, providing ample opportunities for errors (Visser et al. 2004, Both 2010b).Many species have responded to increasing spring temperatures by laying earlier,although not early enough to maintain synchrony with their caterpillar food (Buse etal. 1999, Visser et al. 2006). Mismatch in timing of breeding by Great Tits results in areduced probability of offspring being recruited into the population (Visser et al. 1998).The impact of climate change on timing of Great Tit reproduction has differed among populations, with the Oxford (UK) and Antwerp (Belgium) populations having become better synchronised than before (Charmantier et al. 2008, Matthysen et al. 2011),whereas the Hoge Veluwe (Netherlands) population has become less synchronised. Mistiming may arise from a number of different mechanisms, including timing of food changing more than timing of breeding, the food peak becoming more spread in time (thus reducing the effects of mistiming), or the food peak becoming more concentrated in time (thus increasing the effects of mistiming) (Both 2010b). However, caterpillars and tits do not live in a vacuum, but coexist with different food plants, competitors and predators. Indeed, Both et al. (2009) showed that while the caterpillar emergence date has advanced considerably, and the breeding dates of three species of tits and European Pied Flycatchers have all advanced (albeit to a lesser degree than caterpillars and to a different degree in different bird species), the timing of budburst of oaks (Quercus spp.) and of breeding by Eurasian Sparrow hawks (Accipiter nisus),which eat tits and flycatchers, has not changed at all (see also Nielsen & Møller 2006).Thus mismatch is not just a question of optimal individual decisions and optimal timing of life history events to the environment and to climate change, but also of optimal timing to the web of life that surrounds all living beings. Although studies have revealed mismatches in timing of reproduction relative to food abundance in many species, equally many species have shown no such mismatch (Dunn et al. 2011). The reasons for such interspecific differences remain poorly understood. There are limits to optimal behaviour by birds because there are limits to local adaptation. For example, gene flow at a local scale may prevent individuals from behaving optimally, if individuals adapted to timing of food availability in other populations immigrate from elsewhere (Dhondt et al. 1990). Likewise, evolutionary lag may prevent birds from catching up with ever-changing peaks in food abundance. There has always been mistiming of phenological events, and that will remain thecae with or without climate change. So when should we expect mismatch and when should we not? At which level of climate change should we expect mismatch? Visser & Both (2005) suggested that an appropriate yardstick for assessing the level of mistiming is the extent to which a species adjusts its timing of reproduction to the availability of its primary food. However, every individual of every species is predicted to have to time its reproduction to numerous factors such as prey, predators, competitors or parasites. Clearly, these adjustments cannot all be simultaneously maximised; we should rather expect the temporal match to different species to depend on the strength of selection that each species imposes. Moreover, degree of temporal match should also perhaps depend on the availability of cues that can influence decision-making by individuals of any given species.

 A different way of addressing the question about temporal mismatch is to test for changes in phenological response to climate change at different trophic levels, because overall mismatches between trophic levels would increase the possibility of widespread mismatches between particular interacting species. Thackeray et al. (2010) analysed more than 25,000 rates of phenological change for 726 species in the UK, and found consistent advances in phenology by 0·39 days per year. Such widespread and homogeneous advance is suggestive of a common external driver, namely climate change. This advancing phenology also paralleled the rate of change in temperature, as one should expect if climate change was the main driver. Importantly, the change in phenology depended on trophic level: secondary consumers such as Great Tits showed slower response of 0·22 days per year compared to the response of primary consumers such as caterpillars (0·43 days per year) and primary producers such as oaks (0·40days per year). These findings raise significant concern about the synchrony of different trophic levels, and hence the risk of breakdown of interspecific interactions and ecological communities.

Duration of the breeding season

The breeding season is defined as the period during which environmental conditions allow adults to rear offspring successfully. As described above, the timing of breeding may not always be optimal with respect to food because very early breeding may allow more clutches to be produced during a year, hence temporally displacing both first and second clutches from the time when food availability peaks. Because the growing season at lower trophic levels (i.e. plants) has advanced due to climate change(Schwartz et al. 2006), we should also expect the breeding season to advance. That would result in a longer breeding season if more broods were produced, or the interval between broods expanded. Møller, Flensted-Jensen et al. (2010) analysed change in the duration of the breeding season for 20 bird species in Denmark. There was considerable interspecific variation, from a shortening of the breeding season by 36days in the Sandwich Tern (Thalasseus sandvicensis) to an extension by 36 days in the Common Woodpigeon (Columba palumbus). This change in the duration of the breeding season was directly related to mean temperature during the month when breeding started, with species breeding in cooler months experiencing increasingly longer breeding seasons. Long-distance migrants have not responded as strongly interms of arrival phenology as short-distance migrants (Lehikoinen & Sparks 2010),so we might expect a different change in duration of the breeding season between migrants and residents. Surprisingly, there was only a weak effect of migration distance, with small reductions in the duration of the breeding seasons of long-distance migrants over the period that climate change has occurred, rather than large increases as we might expect. While species laying multiple clutches per year increased the duration of the breeding season by 0·43 days per year, species with single clutches actually reduced the duration of their breeding season by 0·44 days per year (Figure14). This effect can be expected because an increase in temperature causes a shortening of the period of full gonadal maturation (Dawson 2005). Therefore, breeding seasons show considerable flexibility in duration depending on the number of clutches produced per year.

Evidence for a role of the number of clutches in response to climate change also comes from studies of number of clutches and migration. Bird species with multiple clutches per year advanced their spring migration in Hungary more than species with just a single clutch (Végvári et al. 2010).


Dispersal may affect adaptation to changing climatic conditions because gene flow from populations at different latitudes that already experience warmer climates may promote local adaptation; conversely, gene flow may erode local adaptation. Recent studies suggest that dispersal propensity and habitat breadth may increase at range margins, thereby increasing the rate of range expansion (Thomas et al. 2001) and perhaps thus hastening responses to climate change. Unfortunately, there are very few studies that have related dispersal propensity of birds to climate change. Arctic Terns are long-distance migrants that breed in temperate and Arctic regions of the northern hemisphere, and winter in Antarctic parts of the Southern Ocean. In this species, natal dispersal distances from the site of hatching to the site of first breeding have increased from a mean of 15 km in the 1930s to almost 90 km in the 1990s in a marginal Danish population (Figure 15; Møller et al. 2006b). However, this study also found that long natal and breeding dispersal distances were costly as they delayed reproduction in given year. Their incidence was related to climatic conditions both on the breeding grounds and in the winter quarters: natal dispersal distance depended on climatic conditions in the year of hatching and the year of breeding, but breeding dispersal distance was only affected by climate conditions in the second year of the dispersal event. Thus, while natal dispersal distance depended on conditions in the breeding areas, breeding dispersal also depended on the weather in the Antarctic winter quarters, showing that via dispersal, the extent of gene flow was affected by climate in two different parts of the annual range.

Indirect estimates of dispersal can also be obtained from stable isotope profiles, since the isotopic fingerprint of a moulting site is recorded in the chemical composition of feathers. American Redstarts (Setophaga ruticilla) grow or moult their feathers at the breeding grounds. Juveniles winter in either favourable mangrove habitat or inferior scrub, with the latter birds departing on spring migration later than the former. Early departure allowed mangrove birds to migrate relatively short distances to their breeding grounds and settle there to benefit from early spring phenology and an early food supply(Studds et al. 2008). In contrast, juvenile American Redstarts wintering in secondary scrub habitat left late on spring migration, forcing them to migrate further north and thereby risk a temporal mismatch between food availability and timing of breeding. Adults showed highly consistent stable isotopic profiles of their feathers among years, suggesting that they bred at the same sites once having dispersed during natal dispersal.

Another consideration in relating dispersal to climate change is that dispersal decisions may differ among populations, resulting in divergent patterns of selection (Balbontín et al. 2009a). Barn Swallows from a Spanish and a Danish population differed in their dispersal propensity because Spanish birds were six times as likely tobe philopatric (to breed in the same place as were they hatched) as Danish birds. Environmental conditions in Africa as reflected by the Southern Oscillation Index differentially increased philopatry in Spain, while decreasing philopatry in Denmark. Thus, population differences in dispersal propensity as mediated by climatic conditions in the winter quarters (Southern Oscillation Index) may help maintain local adaptation and prevent mixing of populations. Furthermore, such differences may prevent homogeneous responses to climate change, even when the same climatic conditions (Southern Oscillation Index) are acting on different breeding populations.

Bird migration under climate change

Studies of bird migration provide some of the clearest examples of the effects of climate change. However, they also illustrate many of the problems encountered when analysing and interpreting biological data relating to climate change.

Timing and duration of spring migration

Starting in 1749, Linnaeus (1757) organised the recording of the first arrival dates of birds in Finland (Figure 16), which have been continued ever since. These long-term phenological data for first arrival dates are particularly illuminating for several reasons. First, they even show patterns corresponding to shorter-term changes in climate, such as the two relatively brief periods of warming during the years 1860–1889 and1930–1940 (Lehikoinen et al. 2004). Second, all species tend to change in unison, so that periods with early arrival are consistent among species. Third, birds advanced their arrival dates in recent years in a manner consistent with current climate change. Even longer phenological time series exist for cherry blossoms in Kyoto, Japan, dating back to the 9th century (Aono & Kazui 2008), showing that timing of flowering now is earlier than it has been during the period 850–2010.

Migration is an adaptation to the exploitation of seasonal environments. Birds may change from migrants to residents over a few generations if environmental conditions during winter allow. Many populations of northern species that were completely migratory just 100 years ago are now almost exclusively resident, for example Hooded Crow (Corvus cornix), Common Blackbird (Turdus merula), European Robin(Erithacus rubecula), Dunnock (Prunella modularis) and Common Chaffinch (Fringilla coelebs), all of which are now residents in central and northern Europe(Berthold 2001, Newton 2008). The optimal timing of arrival depends on its benefits in terms of competition for territories and mates and the possibilities of early breeding, the production of an additional clutch and recruitment of the offspring to the next generation, and its costs in terms of risk of mortality from adverse weather during spring arrival (Jonzén et al. 2007).

Species differing in migration ecology have responded differently to climate change, with long-distance migrants responding less than short-distance migrants (Lehikoinenet al. 2004, Lehikoinen & Sparks 2010). There are two main reasons why long-distance migrants might be expected to arrive later relative to the phenology of their breeding grounds: first, timing of migration may be under the influence of conditions in the winter quarters often thousands of kilometres away from the breeding areas, and second, because endogenous rhythms determining departure for the breeding grounds are responding only weakly or not at all to climate change. For example, European Pied Flycatchers in the Netherlands have shown a steadily advancing breeding date over the past 20 years but no advance at all in their spring arrival date (Figure 17;Both & Visser 2001).

Earlier spring arrival may be permitted by such factors as faster migration, earlier autumn departure, and the use of wintering areas closer to the breeding grounds. For example, numerous North American species have been shown to migrate more quickly across the US in warmer years (Marra et al. 2005). Several studies also suggest that changes in stopover time and fuelling time are possible when climatic conditions allow( Bairlein & Hüppop 2004). Barn Swallows may rapidly change their departure date from their South African winter quarters in response to changes in the annual cycle of breeding, migration and moult (Møller, Nuttall et al. 2011). Examination of the geographical distribution of ringing (banding) recoveries reveals that migratory birds now winter further north than they used to do just a few decades ago (Siriwardena & Wernham 2002, Fiedler et al. 2004, Visser, Perdeck et al. 2009, Ambrosini et al.2011). These findings of rapid change in migration schedules are supported by recent increases in the frequency with which tropical migrants such as Barn Swallows and European Pied Flycatchers now winter in the Mediterranean region (Lehikoinen &Sparks 2010).

Distributions of arrival dates under climate change

Databases on arrival (and departure) dates are extensive. For example, Lehikoinen &Sparks (2010) were able to analyse 3827 sets of time series data of first and mean arrival dates for 455 species, ranging in duration from 18 to 53 years. An astonishing82% showed a tendency for earlier arrival, with significant effects of species and geographical region (Figure 18). The latter effects imply that the advancement in arrival date was consistent for different populations of the same species, but also that different countries have experienced different levels of climate change and hence different degrees of advancement of arrival. The average trend was for earlier arrival by 0·28 days per year for first arrivals, and 0·18 days per year for mean arrivals.

Such results and conclusions rest on a number of assumptions that may cause smaller or larger degrees of bias (Lehikoinen et al. 2004, Lehikoinen & Sparks 2010).In order to obtain distributions of arrival and departure dates, information about the actual arrival (and departure) dates on the breeding grounds is ideally required. Thus daily (or less regular) data on arrival must be collected from individuals of known identity (i.e. ringed individuals). There are very few long-term datasets of such accuracy for birds of known breeding origin, and these are restricted to studies of Barn Swallows and a few other species. Most time series of arrival dates are rather based on observations or captures of birds on migratory passage at bird ringing stations, but these data pose many problems. They may be affected by variation in sampling effort among years, as well as by changes in population size. First arrival dates, in particular, will become artificially delayed in decreasing populations (Miller-Rushing et al. 2008).If migrating birds are not close to their actual breeding sites, samples of arriving individuals may be biased because more than one population may contribute to the sample recorded. Moreover, changing weather conditions can affect the frequency, timing and duration of stopover during migration, and such changes can affect not only the kinds of birds recorded or captured but also the frequency with which migrating birds land at ringing stations and hence have a probability of being captured. There are additional problems when comparing samples from different years, and when comparing spring and autumn migration, because any of the factors just mentioned (e.g. change in population size, uncertainty about breeding origin of migrants, effects of weather) may change in response to climate change. Using observations collected by amateur birdwatchers may introduce even more bias, for example owing to higher intensity of observations during weekends, or owing to observer activity increasing over time (Sparks et al. 2008).

Ideally, the entire distribution of arrival (or departure) should be recorded in order to provide the best information on timing of migration (Sparks et al. 2005). Not surprisingly, as noted above, there are only a handful of such studies involving individuals of known breeding origin for a sufficient number of years to allow meaningful analyses. Møller (1994b, 2008a) analyzed the distributions of arrival date for Barn Swallows over 32 years, using four key measures obtainable from statistical frequency distributions: mean, variance, scenes and kurtosis (“peakedness” of the distribution).These four measures, which all changed during the study, were related to environmental conditions in North Africa during spring migration as reflected by the Normalized Difference Vegetation Index (a measure of live green vegetation coverage obtainable from satellite data). During benign years when survival was high, mean arrival dates were delayed, showed increased variance, showed decreased skewness,and showed increased kurtosis. This shows that when conditions are better in Africa, not only do Barn Swallows arrive earlier, but weaker selection from mortality on passage lets through birds with a wider range of individual quality: for example, underfavourable conditions, poor quality birds survive and arrive later than the smaller number of survivors in normal years. Moreover, arrival time was independently affected by spring temperature on the breeding grounds, with mean arrival being earlier, the variance smaller and the kurtosis smaller in warm springs. These analyses suggest that different aspects of the arrival distributions change in response not only to climatic conditions en route, but also to weather conditions on arrival on the breeding grounds.

Consistency and geographical variation in arrival date

Given the many difficulties associated with time series data of arrival and departure dates, one might ask whether they nonetheless provide us with any worthwhile biological signals. Several studies have shown that they do. First, Rubolini et al. (2007)analysed more than 600 sets of time series data with a minimum span of 15 years,showing an overall advance in arrival date, especially so in first rather than in mean arrival dates. Importantly, different species varied significantly in advancement of arrival date, showing that while some species have responded to climate change, others have not. First arrival dates advanced more strongly at intermediate latitudes in Europe, and arrival dates advanced more for short- than for long-distance migrants. Second, as mentioned above, Lehikoinen & Sparks (2010) showed similar differences among species in advancement of first and median arrival dates for more than 3800sets of time series data. Thus, we can conclude with confidence (1) that different species vary in their response to climate change, (2) that the mean advancement in arrival date is clear and highly significant, and (3) that geographical patterns of advancement in arrival date follow geographical patterns of climate change, which has-been particularly pronounced in central Europe.

Evolution or phenotypic plasticity of arrival date in response to climate change

Despite numerous data being collected over more than 200 years, we still do not know to what extent changes in arrival dates are due to phenotypic plasticity (individuals changing during their lifetimes) or evolution (heritable genetic change) (Pulido &Berthold 2004, Pulido 2007, Sheldon 2010). Extent of migration, migratory direction and annual timing of migration are all heritable characters that can change rapidly in response to artificial selection, as demonstrated by elegant studies of the Blackcap (Sylvia atricapilla). Indeed, date of migratory restlessness by Blackcaps has advanced considerably among nestlings from the wild that were reared in the laboratory (Pulido& Berthold 2004, 2010), implying that even in a constant environment there is a significant advance in timing of migration. We also know that many species have altered their migration strategy during the last hundred years, and even during the last 20years. However, this does not resolve the question about the evolutionary basis for such changes. Because long-distance migrants live on a different continent during winter, and cannot readily be subject to environmental influences from the breeding range in winter, it has been suggested that their response must be due to micro-evolution(i.e. genetic change) (Jonzén et al. 2006). However, to demonstrate convincingly that a change in arrival date is caused by micro-evolution rather than phenotypic plasticity, we need to show that the individual genetic component of arrival date has changed in response to climate change. Such analyses in turn require separating the variation in arrival date into its environmental and genetic components, and evidence that the individual genetic component, the so-called breeding value of the individual, has changed. To achieve this, information on arrival date for a reasonably large pedigree of individuals is needed, with information on parents, offspring and grand-offspring. These requirements are difficult to fulfill because migratory birds generally have long natal dispersal distances, making it difficult to record such genealogical data. When estimates from a meta-analysis showing an average advance in arrival date were converted into rates of evolutionary change, the rates were within the limits of what cane considered possible from genetic change alone (Gienapp et al. 2007). However, only the study by Karell et al. (2011) on plumage coloration has produced conclusive evidence that climate change effects on birds are due to evolution rather than phenotypic plasticity (Gienapp et al. 2008, Sheldon 2010).

If genetic information on arrival is difficult to obtain, another approach is to test whether phenotypic plasticity alone can account for the observed changes in phenology.If phenotypic plasticity is the underlying mechanism, we might specifically predict that cross-sectional data (comparing individuals in a given year) and longitudinal data (following individuals throughout their lives) should show similar patterns(Przybylo et al. 2000). Saino et al. (2004) analysed cohorts of Barn Swallows and their temporal change in arrival date for an Italian population that winters in Central Africa, in relation to winter conditions estimated from the Normalized Difference Vegetation Index. They found that the arrival date of old (but not yearling) Barn Swallows was advanced during springs that followed winters with benign environmental conditions in Africa (Figure 19). These patterns of within-individual variation in arrival date subsequently affected the timing of breeding and offspring production, showing that environmental conditions in Africa have carry-over effects to the breeding season, with consequences for annual reproductive success. Interestingly, extensive analyses of another population of Barn Swallows breeding in Spain and wintering in West Africa provided slightly different results: ecological conditions during migration advanced arrival date, while conditions in the winter quarters delayed arrival date; again, however, phenotypic plasticity was sufficient to account for these effects (Balbontín et al. 2009b).

Is there any evidence for a role of genetic factors in changes to the timing of migration? To date, the only indication we have comes from an analysis of the genetic basis of arrival date in two populations of Barn Swallows. This showed that spring arrival date was both phenotypically and genetically negatively correlated with morphological traits involved in migration (wing and tail length) and life history traits(time until breeding) (Teplitsky et al. 2011). One implication of this finding is that negative genetic correlations may constrain adaptation to climate change, because selection for earlier arrival will necessarily be associated with a delay in breeding.

Breeding vs. winter and stopover effects on arrival date

 Changes in arrival dates for migratory birds to the breeding grounds may provide a misleading perspective on climate change if environmental conditions in the winterquarters are in fact the main determinants of timing of migration. However, this is notthe only possibility, because early arrival to the winter quarters may allow early moultand hence early departure, as seen in some species (Kok et al. 1991, Beaumont et al.2006, Chambers 2008, Møller, Flensted-Jensen & Mardal 2009), but not in others(Møller, Nuttall et al. 2011). Therefore, early departure from the winter quarters mayin turn be influenced by the timing of autumn migration. Since conditions on the breedinggrounds cannot be assessed from the winter quarters, it is difficult for migrants toadjust their arrival times to conditions at the breeding sites. There is some evidencesuggesting that temperature anomalies in Africa affect the arrival date of migrants tothe breeding grounds (Saino et al. 2007). Investigations of the effects of conditions inthe winter quarters on arrival to the breeding grounds have produced mixed results:some studies showed earlier arrival during benign conditions (Saino et al. 2004, Gordo& Sanz 2005, Gordo et al. 2005, Both, Sanz et al. 2006, Møller 2008a), another reporteddelayed arrival following a benign winter (Balbontín et al. 2009b), and anotherfound no effects (Tøttrup et al. 2008). Note, however, that only Both, Sanz et al.(2006), Møller (2008a), Saino et al. (2004) and Balbontín et al. (2009b) used breedingpopulations of known origin, so only these studies are likely to provide reliable tests ofhow winter conditions affect arrival time to the breeding grounds.

The question is even more complicated by the fact that migrants rest at stopover sites during migration, and that environmental conditions at these sites may affect migration and hence the date of arrival. The frequency of stopover and the time required for refueling appears to have changed in recent years (Bairlein & Hüppop2004), and it is known that environmental conditions for both European Pied Flycatchers and Barn Swallows in North Africa during spring stopover affect their arrival date to the breeding grounds (Both, Sanz et al. 2006, Møller 2008a, Both 2010a).Furthermore, weather conditions on migration also affect arrival date, especially as migrants approach their breeding destinations, perhaps because migrants at these locations are better able to adjust their migration activity to prevailing weather in large areas in northern Europe (Ahola et al. 2004, Tøttrup et al. 2008). These data suggest that the annual schedule of migration is determined by environmental conditions in all geographical areas encountered throughout the annual cycle, rather than pinpointing changes in migration schedules to events in a particular area during a specific period.

Timing and duration of autumn migration

There are many fewer data on departure dates from the breeding grounds than there are for arrival dates. Lehikoinen & Sparks (2010) analysed 683 time series datasets for 246species, including 374 last departure dates and 150 median departure dates. A total of60% of last departure dates became delayed over the study period, while only 36% of median departure dates became delayed. Studies of autumn departure at specific sites show similar trends, at least for long-distance migrants (e.g. Jenni & Kéry 2003). That changes in autumn departure dates should be less clear than those for spring arrival dates is as expected if there has been an advance in the start of reproduction, but little or no change in the end of reproduction. However, some studies suggest that autumn departure is advanced just as spring arrival, at least in single-brooded species with relatively short migrations (e.g. Jenni & Kéry 2003, Van Buskirk et al. 2009a).

Sex differences in arrival date and protandry

 Bird migration is under intense selection because early-arriving individuals have higher mating and breeding success than later-arriving individuals (e.g. Møller 1994b). Males and females do not benefit equally from early arrival, because competition for high quality mates is more intense among males, and because tertiary sex ratios (i.e. sex ratios among reproducing adults) are often male-biased, implying that a fraction of males do not mate. Therefore, males should experience stronger selection for early arrival than females, and the fitness benefits males might thus accrue can be quantified. Møller (2007b) analysed the fitness advantages of early arrival in male and female Barn Swallows during a period when spring arrival advanced in a Danish breeding population. Arrival date was quantified by weekly captures of adults at breeding sites, and the fitness advantage was quantified as survival to the next year, as well as by the number of offspring reared during the current year. Males arrived before females, and increasingly so in recent years (Figure 20A). Advancing early arrival was associated with increasingly better survival in males, while females showed the opposite pattern. While early-arriving males survived better and females worse at the start of the study, by the end of the study early-arriving individuals of both sexes survived worse than later-arriving individuals (Figure 20B). Further analyses revealed that this result arose because early-arriving males survived better when inter-clutch intervals were short, while early-arriving females survived better when inter-clutch intervals were long(Figure 20C). Therefore, patterns of selection in the two sexes change in response to advancing arrival dates, leading to a complex result.

The extent of sex difference in arrival date will depend on the cost and benefits of early arrival for males and females. Earlier arrival in males than in females (protandry)is caused by more intense competition among males (Kokko 1999, Morbey & Ydenberg2001). In Barn Swallows the degree of protandry (the number of days that the average male arrives before the average female) has increased in recent years because males, responding to warmer springs, arrive earlier, whereas females do not (Møller 2004a).Specifically, the degree of protandry grew as temperatures increased during April, which is when the first birds arrive. In addition, the sex difference in arrival date depended on condition during spring migration in North Africa. Recent years have been characterised by dry weather and resultant poor vegetation quality, as measured by the Normalized Difference Vegetation Index (Møller & Szép 2005). Barn Swallows are at their most vulnerable in North Africa, when they have just crossed the Sahara, and deterioration in environmental conditions here increased the mortality rate, allowing only males in prime condition to survive. Not surprisingly the degree of protandry was larger in years when mainly long-tailed males in prime condition survived ;by contrast, in years when even males in poor condition were able to survive,the sexes tended to arrive simultaneously (Møller 2004a).

While these analyses concern protandry at the population level, selection acts at the level of individuals, and any given male may breed with a female that arrives relatively early or relatively late. Does it matter for males if they mate with an early or late-arriving female? Males that arrived well before their mates gained a fitness advantage because they fertilised more eggs in their own nests than did later-arriving males. The fitness advantage of males mated to females that arrived earlier came from earlier breeding, which allowed their offspring to fledge early and hence have a high probability of survival (Møller, Balbontín et al. 2009). These results demonstrate the complexities of selection acting on arrival date when it advances in response to climate change.

Evolution of secondary sexual characters, colour and body size

Climate change is a forceful selective agent, and it may become even more so in the near future. Characters with high resolvability (i.e. having high genetic diversity on which natural selection can act) should therefore be expected to respond rapidly to climate change. Evidence for this has emerged in several recent studies of secondary sexual characters, coloration and body size, which are all traits that are indeed expected to have high resolvability.

Male Barn Swallows have outermost tail feathers that are on average 20% longer than those of females, and males with long tails enjoy mating advantages in terms of mating success, laying date, annual fecundity and paternity, among others (Møller1994a). Tail length is also strongly dependent on the individual’s condition, so climate change may affect the expression of this secondary sexual character through phenotypic plasticity depending on environmental conditions. Barn Swallows from the Danish population under study winter in southern Africa and migrate across the Sahara through North Africa. As noted above, rainfall has decreased in North Africa since the mid-1980s, resulting in reduced survival mainly of Barn Swallow individuals in poor condition. This change in climate has led to an increase in mean tail length of over 12 mm in males (but not females) at the population level (Figure 21; Møller &Szép 2005). This change in tail length is associated with changing temporal patterns of selection on male tail length due to survival.

A clear example of the effects of climate change on plumage coloration comes from Tawny Owls (Strix aluco) in Finland (Karell et al. 2011). Tawny Owls occur into colour morphs that are either brown or grey, as determined by simple Mendelian inheritance with two alleles. There is strong natural selection against the brown morphin winters with lots of snow, when they survive poorly, whereas survival of grey morphs is independent of snow depth. Such snowy winters have become rare, as increasing temperatures have caused snow depth to decline from 12 cm to 2 cm over the period1981–2008. This has produced a dramatic increase in the frequency of brown morphs, from around 12% in the early 1960s to 42% in 2005–2010 (Figure 22). Thus, climate change has clearly altered natural selection on colour morphs.

Body size of birds is the result of past selection for adaptation to local conditions. Bergmann’s Rule states that the body mass of populations of species increases with decreasing mean annual temperature, which is explicable in terms of larger body size reducing the surface-to-volume ratio, and hence minimising loss of energy. We should therefore predict that under climate change, body mass will decline as temperatures rise. The prediction was supported in a study of the size of five bird species in Israel, using body mass and tarsus length of museum specimens for 1950–1999. Body mass and tarsus length were found to have declined in four and two species respectively(Yom-Tov 2001), and various alternative explanations could not account for these trends. This pattern was further confirmed in a number of breeding bird species in the UK over a period of 40 years (Yom-Tov et al. 2006) and in migratory birds in Pennsylvaniaover a period of 50 years (Van Buskirk et al. 2009b). However, no such pattern was found in extensive data on locally hatched juveniles of twelve species of birds in Central Europe (Salewski et al. 2010). Thus, while there is evidence for reductions in body size of birds in response to climate change, the findings are not all consistent. It remains to be seen whether these inconsistencies result from differences in sampling, geographical differences in climate change, or other factors. Whether these changes in body size represent phenotypic or genetic changes also remains largely unknown. One study that was able to distinguish the two types of change involved Red-billed Gulls (Larus scopulinus) in New Zealand during 1958–2004, and showed 4% decrease in mean body mass as temperatures increased. In this case the change could be shown to have come about through phenotypic plasticity rather than genetic changes, underscoring the general pattern that commonly observed responses to climate change may arise from plastic adjustments rather than evolution (Teplisky et al.2008). Thus, this example provides evidence than changes in size can be entirely phenotypic rather than genetic.

Climate effects at the population level

Climate change and population dynamics

There is little doubt that climate can affect behaviour, reproduction, survival, migration and dispersal of individuals. However, that does not necessarily mean that these individual effects will trickle up to the population level and have an impact on population dynamics. How might climate impact population size? Several alternative mechanisms need careful consideration. For example, climate may have either density-independent or density-dependent effects, depending on whether the fraction of individuals affected by climate is independent of their density (e.g. through extreme weather events causing mortality), or has greatest effect at low and high densities(e.g. through an impact on food availability). We also need to consider whether climate mainly affects reproduction, or mainly survival. Last, we need to consider whether climate mainly affects the mean of demographic values such as population size or reproductive success, or rather their variance. For example, climate may, through hits effects on food availability, change mean reproductive success. But as already described, extreme climatic conditions can also dramatically increase the fraction of individuals without reproductive success, thereby increasing the variance in demography among individuals. These alternatives can have significant impacts on expected population responses, as we shall see below.

Population regulation may act during the non-breeding season, either through climatic conditions or through the effects of population density on non-breeding mortality(Lack 1954). Alternatively, populations may rather be regulated by climatic conditions affecting productivity, and hence the number of recruits entering the breeding population. There is an extensive literature analysing effects of weather on population sizes of birds, and a review (Sæther et al. 2004) has shown that weather has important consequences for changes in mean demographic variables (such as population size and composition), environmental stochasticity (environmental events affecting theentire population) and demographic stochasticity (variation in demographic parameters among individuals). It also showed that weather has clear effects on density dependence: northern temperate altricial species such as tits tend to have their populations regulated by winter climatic conditions, whereas it is weather during the breeding season that tends to regulate populations of precocial species such as duck sand partridges.

The diversity of population responses to climate change is large (Sæther & Engen2010). Three examples will suffice. The first concerns Emperor Penguins (Aptenodytesforsteri), which breed on the Antarctic ice shelf. There has been a dramatic decline in population size of Emperor Penguins associated with a decline in ice cover (Figure23), with adult survival strongly affected by sea ice conditions and temperature during winter (Barbraud & Weimerskirch 2001, Masson-Delmotte et al. 2003,Jenouvrier et al. 2005). Because population growth is mainly determined by adult survival, the negative impact of climate change on survival has reduced the population size by half. However, breeding success improved when population size declined, because less competition allowed Emperor Penguins to forage closer to their colonies. Nonetheless, because population size is determined by adult survival, this improved breeding success has not led at an increase in population size in recent years (Jenouvrier et al. 2005, 2009). In contrast to Emperor Penguins, other Antarctic penguin and seabird species have responded differently to these changes in local climatic conditions (Jenouvrier et al. 2003, 2005, Forcada et al. 2006). This shows that climate change can affect different demographic traits such as reproduction and survival in opposite directions, with complex effects on population size and population growth.

A second example concerns spatial patterns in population dynamics of tits and European Pied Flycatchers. We can understand the spatial dynamics of populations of these species because they have been studied in numerous places for decades. Surprisingly, even neighboring populations in the Netherlands can show a ten-fold difference in the impact of temperature on local recruitment (Grøtan et al. 2009). When analysing the amount of variance in local population dynamics of Great and Common Blue Tits explained by climate, after accounting for density-dependence and demographic stochasticity, there was large variation among populations, even to the extent that climate may affect different populations in different directions (Sæther et al. 2007).Some of the variation in the importance of climate is explained by latitude (Sæther etal. 2003): while the North Atlantic Oscillation (NAO) only weakly affected dynamics at low latitudes, this effect increased strongly at high latitudes. Among Great Titpopulations, NAO explained a larger proportion of the variance in annual size at higher latitudes, while in European Pied Flycatchers there was no latitudinal effect because the proportion of environmental variation in population dynamics also increased with latitude. What mechanism might explain such temporal patterns in effects of climate on population dynamics? One possibility is suggested by the finding that population declines in European Pied Flycatchers in the Netherlands occurred in sites with late food peaks and hence a mistiming between food peak and timing of breeding, while early-breeding populations that did not mistimed their reproduction remained stable(Both, Bouwhuis et al. 2006).

Population dynamics can be affected not only by processes within a given population, but also by immigration from elsewhere. However, dispersal and immigration rates are generally poorly known. A convincing example of the role of immigration comes from White-throated Dippers (Cinclus cinclus), which have experienced population size increases as a consequence of climate change. Cold winters are the main cause of mortality in the species (Sæther et al. 2000), and population increases during warm springs are attributable not only to improved success of locally hatched juveniles, but also to higher recruitment of immigrant yearlings from other populations(Figure 24). A surprisingly similar situation has been observed in Great Tit population sin the Netherlands, in which both local recruitment rate and immigration rate were positively affected by environmental variables (Grøtan et al. 2009).

Climate change and interspecific competition

Climate change will affect the relative abundance and the composition of different species, through its effects on timing of reproduction and changes in range size. This should in turn lead to changes in the intensity of interspecific competition. A clear example of interspecific competition comes from Dhondt’s (1977) classic experiment son tits. These showed that Common Blue Tits consume the early life stages of caterpillars, which therefore never have a chance to grow up and hence become suitable food for the larger Great Tit. More recently, this interspecific competition for food has-been shown to be affected by climate, since its intensity varies both among population sand with respect to local climate change (Dhondt 2011).

Climate change can also alter competitive relationships between resident and migratory birds, through its effects on timing of breeding or population density of the competing species. A potential example concerns competition for nest-holes in Europe between resident Great Tits and migratory European Pied Flycatchers that wintering tropical Africa (Ahola et al. 2007). Competition arises from individuals of the two species fighting over boxes in limited supply, often with a fatal outcome for one of the individuals. As the interval between laying date of European Pied Flycatchers and Great Tits has decreased due to climate warming advancing the arrival date of European Pied Flycatchers, the number of cases of European Pied Flycatchers dying in the attempt to take over a box of a Great Tit has increased. A higher density of European Pied Flycatchers exacerbated this effect, and the density of each species was respectively determined by climate in Africa for the migrant flycatcher, and climate in Europe for the resident tit.

Both of the foregoing examples provide clear evidence that shifts in the balance of competitive interactions between species are another consequence of climate change.

Climate change and predation

Predators survive by consuming prey, be they krill (Euphausiacea sp.) consumed by Emperor Penguins, tipulids by Eurasian Golden Plovers (Pluvialis apricaria), herring(Clupea harengus) by Atlantic Puffins (Fratercula arctica), caterpillars by Great Tits, or Great Tits by Eurasian Sparrow hawks. This diversity of predator-prey interactions can be affected by climate change in direct and indirect ways, and with significant consequences for populations and communities (Bretagnolle & Gillis 2010). The relationship between the abundance of prey and their predators depends both on the functional response (reflecting how the abundance of prey changes in response to consumption by predators), and the numerical response (changes in abundance of predators as a consequence of prey being consumed by and thus converted into predators).Both these responses can be affected directly and indirectly by climate change. The main factors involved are changes in (1) the distribution of prey and predator, (2)the phenology of prey and predator, (3) the population density of prey and predator, and (4) changes in behaviour of prey and predator.

The evidence to date for these four factors is as follows. First, changes in the distribution (in both space and time) of prey and predators may occur as a consequence of one or both parties responding differently to climate change, for example owing to different feeding ecology. Brommer & Møller (2010) showed that insectivorous birds, and in particular terrestrial herbivorous birds, have expanded their range margins the most under recent climate change; however, there was no evidence for differences in range expansion between birds of prey and insectivorous or granivorous birds. Second, an effect of phenology has been shown by Thackeray et al. (2010),who found a stronger phenological response of secondary consumers (predators) than primary consumers (herbivores), implying an increasing degree of phenological mismatch for the former (Visser et al. 2004). Even studies of bi-trophic food webs (birds and their predators) or tri-trophic food webs (insects, insect-eating birds and their predators) suggest that predators have responded less than their prey (Nielsen & Møller2006, Both et al. 2009). Third, evidence for an impact of climate change on predatorand prey density comes from Pearce-Higgins et al.’s (2010) work on Eurasian Golden Plovers, which have been affected by climate change-induced changes in the densityof their tipulid prey (Figure 25). Fourth, what is the evidence that climate change could affect predator and prey behaviour? To date no studies appear to exist showing that climate-induced changes in behaviour or morphology of prey or their predators have affected predator-prey interactions. One possible scenario we might predict is that the decrease in body size of many bird species linked to recent climate warming(Yom-Tov 2001, Yom-Tov et al. 2006, Van Buskirk et al. 2009b) may change the prey species preference of predators.

Irrespective of the mechanism, the effects of climate change on predator-prey interactions are likely to depend on whether predators are specialists on one or a few prey species, or generalists consuming a wide variety of prey. Generalists should be relatively unaffected by climate change, because they can switch from one prey species to another if any of their main prey species changes in abundance. However, such prey switches are then likely to affect not only competition among different prey species, but also competition among coexisting predator species. Increasing evidence supports such complex ecological interactions responding to climate change. For example, Millon et al. (2009) studied Eurasian Sparrow hawks in Denmark, where their main prey species are the Common Blackbird and the Eurasian Skylark (Alaudaarvensis). An increase in the abundance of blackbirds was followed by an increase in consumption by the sparrow hawk, resulting in a numerical response through increased sparrow hawk population size, which in turn reduced the abundance of the closely related Song Thrush (Turdus philomelos) to very low levels. A study of a specialist vole predator showed that climate change had a ten-fold stronger impact on the Eurasian Buzzard (Buteo buteo) than on vole abundance (Lehikoinen et al. 2009). Interestingly, the Eurasian Buzzard breeding at the northernmost limit of its range in Finland did not benefit from climate change despite increasing winter temperatures that should have advanced breeding, apparently because cold summer temperatures caused reductions in chick productivity and survival. An example of effects of climate on more than a single predator species comes from a study of two specialist predators, the Long-tailed Skua (Stercorarius longicaudus) and the Snowy Owl (Nyctea scandiaca),which feed almost entirely on nearctic collared lemmings (Dicrostonyx groenlandicus).While there were clear fluctuations in predators and prey until 2000, the density of lemmings subsequently remained very low, leading to a complete absence of Snowy Owls, and Long-tailed Skuas are predicted also to experience severe declines (Gilg etal. 2009). These effects were hypothesised to have arisen from climate change affectingthe abundance of prey and, therefore, the abundance of predators.

Climate change and host-parasite interactions

Parasites and infectious diseases are common causes of mortality in birds, albeit poorly studied ones. One way in which parasitism could be climate-dependent is that parasites that live away from their hosts for some of their life cycle necessarily depend on climatic factors such as temperature and rainfall. Therefore, climate change will affect the timing of emergence of many parasites, and hence also influence which part of the year parasites are active and can use for reproduction (Dawson et al. 2005, Hudson et al. 2006). Indeed, several bird ectoparasites, such as louse flies now emerge significantly earlier than just a few years ago (Møller 2010).

Such advances in date of emergence also apply to ticks, fleas and other ectoparasites, and to vectors of blood parasites such as black flies and mosquitoes (Merino & Møller 2010, Møller 2010).Parasites impose fitness costs on their hosts either by using resources that hosts could otherwise have used for themselves, or by inflicting pathological damage. Extensive studies of humans (Guégan et al. 2003, Guernier et al. 2004) and birds alike(Møller, Arriero et al. 2010) have shown that parasites inflict more severe costs on their hosts in the tropics. This latitudinal trend is apparently due to parasites maintaining high population densities throughout the year in warm climates, whereas at high latitudes, severe winters reduce parasite population size and rate of reproduction. In birds, nestling mortality caused by parasites increases from the northern temperate zone to the tropics (Møller, Arriero et al. 2010). Regardless of latitude, increasing temperatures could be expected to result in an increase in virulence (i.e. the impact of parasites on the fitness of their hosts), because more parasite generations can fit into longer growing season, resulting in an increased rate of parasite evolution. If so, then we should expect virulence to increase with global warming. Effects of increased virulence could be worsened if climate change negatively affects the condition or the nutritional status of the host, thus weakening their immune responses. Recent evidence for rapid changes in virulence came from a study of Barn Swallows, which experienced (by contrast) decreasing per capita costs of parasitism from tropical fowl mites (Ornithonyssus bursa) (Møller 2010) over a period of 30 years. Moreover, parasites can also have particularly negative effects on their hosts during adverse environmental conditions (de Lope et al. 1993, Merino & Potti 1996). This may help explain how parasites interact with climate to synchronise changes in population numbers of Rock Ptarmigan (Lagopus mutus) across different geographical locations in the UnitedKingdom (Cattadori et al. 2005).

Studies of bird parasitism under climate change are particularly important because avian parasites such as those causing malaria and avian influenza may also have implications for human public health. Two recent studies have shown a rapid increase in the proportion of bird hosts infected with malaria and other blood parasites (Møller2010, Garamszegi 2011). First, two species of blood parasites of the genera Haemoproteusand Leucocytozoon in the Barn Swallow more than doubled in abundance over a period of 20 years (Møller 2010). Second, an analysis of all studies of blood parasites in birds similarly revealed a strong temporal trend towards increasing prevalence of blood parasites, especially after 1990, when climate change started in earnest(Garamszegi 2011). These findings, albeit preliminary, suggest that blood parasites may increase in abundance with potentially negative impacts on their hosts. Blood parasites have already had significant negative impact on endemic bird populations in Hawaii and New Zealand, and climate change may allow mosquito vectors to move to higher altitudes, reducing the remaining populations of endemic birds that are currently surviving in forest remnants at high altitudes (Benning et al. 2002, Tompkins & Gleeson 2006).

The Common Cuckoo (Cuculus canorus) and its hosts provide a particularly illuminating example of the effects of climate change on parasites. Arrival dates of long distance and short-distance hosts have advanced at different rates, with short-distance migrants advancing on average by 0·38 days per year and long-distance migrants by0·16 days per year (Saino et al. 2009). Because the cuckoo is also a long-distance migrant, it has only advanced its arrival date by 0·13 days per year, thus becoming increasingly mistimed with its short-distance migratory hosts. In countries with no change in spring temperature, cuckoos used short-distance migratory hosts in about 80% of parasitized nests, but in countries with the strongest increase in spring temperature, this declined dramatically to only 5% of parasitized nests (Figure 26; Møller, Saino et al. 2011). If analysed in terms of change in use of short-distance migratory hosts in relation to their previous levels, countries with no change in spring temperature were predicted to experience an increase in frequency of short-distance migrant hosts by 20%, while the frequency of short-distance migrant hosts in countries with the greatest increase in spring temperature would decrease by an astonishing 60%(Møller, Saino et al. 2011). Cuckoos have host races (also often known as “gentes”)specialised to exploiting different host species by mimicking their eggs in terms of colour and volume. The above findings suggest that host races that rely on short distance migrants such as Meadow Pipits (Anthus pratensis), Northern Wrens (Troglodytestroglodytes) and European Robins may decrease radically in abundance and ultimately go extinct. This is serious because host races such as those exploiting Common Redstarts (Phoenicurus phoenicurus) differ by more than 2% between northern and southern populations, clearly comprising evolutionary important cryptic genetic diversity that exceeds the levels generally considered to reflect divergence at the species level (Johnsen et al. 2010).

Climate change may not only affect parasites, but also the ability of their hosts to defend themselves by means of immune responses. Components of the immune system depend strongly on the condition of the host, and any deterioration in climatic conditions can have severe negative effects on the ability of birds to produce an immune response. Indeed, size of the spleen and cell-mediated immunity are reduced in years with adverse weather conditions (Møller & Erritzøe 2003, Merino & Møller2010), although at present we have no long-term data on immunity in relation to global climate change.

Climate change and extinction risk

The vulnerability of bird populations and their ultimate risk of extinction may depend on climate change. Many ecologists have addressed questions about the extinction risk of populations, trying to provide reliable answers for decision-makers. Most recently Sæther et al. (2005) used data on the probability of individuals being recruitedinto the population to quantify extinction risk for different species. Environmental stochasticity (environmental events affecting the whole population, such as climate change) had the strongest immediate effect on the risk of extinction in birds, while long-term population persistence was most strongly affected by population growth rate. Species with smaller clutches were predicted to have the longest persistence, because stochastic effects on population dynamics were less severe in species with small clutches. Modelling exercises disregarding such stochastic effects have suggested that a large proportion of species may run the risk of extinction (Thomas et al.2004), although the reliability of these projections remains poorly known. Extinction caused by climate change or other factors has important consequences beyond the species involved, since extinction processes may affect wider ecosystem processes (Sekercioglu et al. 2004). Even species that are currently common may become extinction prone in the near future, with significant consequences for ecosystem processes such as decomposition, pollination and dispersal of seeds. Two recent studies highlighted the immediacy of potential climate-induced extinction. First, Sekercioglu et al. (2008) suggested that a large percentage of bird species inhabiting high altitudes in mountain ranges may risk going extinct as the weather changes and plant communities move upslope. Second, Huntley et al. (2010) judged that the risk of extinction among South African birds is greater than previously thought. However, such predictionsare fraught with uncertainty because many factors other than climate will affectfuture distributions of species, making it difficult to assign causality to different potential drivers of extinction (e.g. Vallecillo et al. 2009). Moreover, extinction risk sentirely based on climate variables and climate change may be exacerbated by other environmental drivers of extinction such as changes in land-use (Brook et al. 2008).When making predictions about extinction risks it is crucial to take account of these uncertainties, although this is frequently left aside (Thuiller et al. 2004).

With time, real data may come to exist to validate such theoretical predictions about extinction risk. It will certainly be very valuable to closely track changes in population sizes, but in particular to confront such changes with predictions, as environmental conditions enter a phase that has never been seen before in recent history.

Range expansion

Projected changes in distributions and range margins

Future projections of the distribution of species rely either on population viability analysis, which requires very detailed information on demography and population dynamics, or on habitat suitability modelling, which uses information on current distribution to predict the extent of the geographic area that in the future might be occupied by the species. Habitat suitability modelling relies on the observation that climatic conditions or other factors often predict current distribution ranges quite well, and therefore that we should be able to forecast potential future distributions based on our knowledge of current distributions and future climatic scenarios (Thuiller & Münkemüller 2010). There are many different types of models with their own advantages and limitations, and deviations from assumptions may have large effects on forecasts. One commonly used way to reduce the risk of drawing false conclusions is to use ensembles of climate scenarios—the list of climate change scenarios provided by the IPCC (Anon. 2007c)—rather than a single scenario, when predicting the future distributions of species (Araújo & New 2007).

It is possible to validate the forecasts of such models, and this is often done. One possibility is to test the performance of the model across space or time, for example by comparing predictions with observed population trends, or using retrospective data from the past to project them to current conditions (e.g. Peterson 2003, Araújo et al.2005, Hijmans & Graham 2006, Gregory et al. 2009). A slightly different approach into test a climate-driven model by removing the effects of potential confounding variables one at a time because the effect of each variable can then be partitioned. For example, Vallecillo et al. (2009) analysed the effectiveness of a model of bird distribution in relation to land-use change by testing it against another dataset with information on climate and topography as well as land-use. They subsequently analysed the efficiency of fires to predict local colonisation and extinction, showing that exclusion of such important variables as fire can dramatically change the predictions. Hockey et al. (2011) provided similar conclusions for South Africa.

Huntley et al. (2007) provided the most extensive effort so far to forecast the distribution of the entire European avifauna. This effort based on climate ensemble models provides many interesting insights, but also illustrates how such forecasting sometimes results in extreme situations such as habitats becoming suitable in faraway places like Iceland or Spitsbergen, when in reality dispersal would prevent colonization these areas. As well as dispersal, many other factors such as interspecific interactions may prevent such exercises from producing reliable scenarios that can be used in applied contexts such as conservation.

Observed changes in range margins

Ranges of many species of birds have changed during recent decades, as shown by studies in North America, Europe and Africa. In North America, the northernmost range margin of many species is limited by minimum winter temperature limits (Root(1988a,b), and similar patterns are found in Europe (Forsman & Mönkkönen 2003).In South Africa, by contrast, range margins have mainly changed in response to land use, with little or no direct evidence of effects of climate change (Hockey et al. (2011).The north-temperate studies do, however, provide a direct link between climate and distribution, and in these regions we should therefore expect consistent range expansions with increasing temperatures, especially in winter and spring.

The evidence suggests that this is a ubiquitous and common phenomenon (Parmesan& Yohe 2003, Root et al. 2003, Hickling et al. 2006, Parmesan 2006). Surprisingly, however, published changes in range margins are often modest in magnitude and not statistically significant; this arises from large differences in response among species, with some moving in the “right” and others in the “wrong” direction. One study that did find clear effects was Thomas & Lennon (1999), who used atlas mapping data to show that among breeding birds in the British Isles, those species with southerly distributions advanced their range margins northwards by 1·18 km per year, while species with northerly distributions did not contract their range margins in the south. These estimates were obtained by adjusting for the fact that range margins and population sizes may change simultaneously. This is necessary because it would be unsurprising if a species expanded its range when its population size doubled, but completely surprising if a species expanded its range while simultaneously declining. The latter would provide much more convincing evidence for a role of climate change, making it crucial to control estimates of effects of climate for effects caused by changes in population size and hence dispersal to other parts of the range. Subsequent analyses for Finland showed an even larger advance of 2·69 km per year for southernmost range margins, but again no significant change for northernmost range margins (Brommer 2004). Likewise, analyses of changes in range margins in the US have shown consistent changes in southernmost range margins alone: in the US east of the Rocky Mountains, range shave advanced by 2·35 km per year east of the Rocky Mountains (Hitch & Leberg2007), and by 0·64 km per year for New York State (Zuckerberg et al. 2009).

These changes all appear relatively modest given the extreme flying ability of most birds, especially their annual migrations of often thousands of kilometres. However, the foregoing studies all used distributions as mapped by bird atlases; bird censuses, by contrast, may provide superior data for quantifying changes in range margins. evictor et al. (2008) used breeding bird census data from France to estimate amen shift northwards for the entire breeding bird community of 5·35 km per year, which is considerably larger than the estimates listed above. A possible explanation is that many species in France are close to their southernmost distribution limits, and therefore most susceptible to climate-related effects.

How general are the changes in range margins observed to date? To answer this question, it is useful to know the degree of variation and consistency in the findings of different studies across different regions. Several species were shared between the foregoing four studies of change in range margins, allowing Brommer & Møller (2010)to calculate how consistent findings were among studies. Surprisingly, there was no statistically significant consistency among studies of the same species (Brommer &Møller 2010), implying that if a species shifted north by, say, 10 km per year in the UK, it was unlikely to do the same in Finland. Similar conclusions were reached when controlling for change in range size. These results imply that changes in range margin will depend on idiosyncratic features of local populations, and that it will be very difficult to predict how range margins will respond to climate change.

Even if species are not very consistent in how their range margins change, different categories of species might be broadly consistent, owing to their shared ecological characteristics. Indeed, an analysis of Finnish breeding birds found that groups defined by feeding ecology, migration and habitat were consistent in how their range margins expanded in response to climate change (Brommer 2008). This analysis was subsequently broadened by searching for similar patterns across the four geographical regions listed above (Brommer & Møller 2010). This showed that for Finnish species with a southern distribution, feeding ecology predicted a small fraction of range expansion, especially in seed-eaters and species using aquatic habitats. In contrast, there were no such effects in the UK or the US. Analyses of the Finnish data revealed more change in range margins in insectivores and terrestrial herbivores, as well as in partial and short-distance migrants. Thus there is a certain amount of evidence for the importance of feeding ecology in determining range shifts. However, other factors could also affect changes in range margins: these might include life history, dispersal distance, thermal range and brain mass (a surrogate for cognitive abilities). To date there is only scant evidence that such traits can predict species’ rate of change: analyses of data from Finland and the UK provided weak evidence that small brain mass was associated with larger range margin expansion, and yet weaker effects of adult survival rate and body mass (Brommer & Møller 2010).Finally, population trends may have impacts on changes in range margins because an increasing population will produce more emigrants that will disperse elsewhere and thus contribute to range expansion. Brommer & Møller (2010) analysed range margin expansions in relation to population trends in the four studies listed above. There was evidence of a clear positive relationship for species with southern distributions in Finland, the UK and the US, implying that southern species with an increasing population expanded their range margins the most.

Many of the findings reported here suggest that ecological characteristics and population trends are only poor predictors of change in range margins. The majority of studies of change in range margins are based on atlas projects that often have changing definitions of breeding records, and the research effort in each census plot is often inconsistent, albeit with a trend for more effort over time. Thus the data used for range shift analyses may be less than ideal, giving rise to noisy results and conclusions. Clearly the use of extensive breeding bird censuses using standardised methods, as done by Devictor et al. (2008), will be the best way forward, especially if estimates for different species can be derived.

Dynamics at range margins

Expansions in range margins are of course made up by the movements of individual birds, but as yet we know very little about whether these individuals differ from the population as a whole. Which are these mobile individuals and what are their characteristics? Are they individuals with extrovert and exploratory personalities? Are they individuals predisposed to be highly dispersive? These questions still need answering. We can however speculate about which geographical sections of populations are likely to be most affected by climate change. Clearly, some populations at the edges of distributions have changed, as shown by alterations in range margins, but we might further expect that greater fluctuations should occur in populations at the southern compared to the northern margins of a species’ range. This is because it is at the southern margins of populations that climate change should generally have been more severe due to increasing temperatures reaching threatening levels, pushing species closer to their extreme limits of heat tolerance. In support of this prediction, as noted above, Brommer & Møller (2010) found stronger evidence for changes in the range margins of species with a southern distribution, and this effect was consistent across studies in Finland, the UK and the US. Models based on future climate change scenarios (“consensus ensemble forecasts”) for this century similarly predict that climate change will have particularly strong consequences for the loss of biodiversity in southern Europe, with gains in biodiversity at high altitudes and latitudes (Thuiler et al. 2011).

What is the mechanism behind these fluctuations? All organisms have a fundamental niche that reflects the environmental conditions under which they are able to thrive. The actual environmental space a species occupies is partly determined by tolerance to heat and other biotic factors, and the results described above certainly dimply a link between temperature and population fluctuations, even though birds are homeothermic (body temperature relatively constant). The underlying mechanism relating temperature to population dynamics may be that fluctuations in food availability are greatest at the southern limit of breeding ranges (Pearce-Higgins et al. 2010).

Effects at the level of communities

Generalists and specialists

Birds vary enormously in their niche width and hence their degree of specialisation,and this should be expected to influence their susceptibility to climate change. A species’degree of specialisation can be quantified by its distribution across a range of predefined habitats recorded during standard bird censuses such a point counts (Julliardet al. 2003). For example, a specialisation index based on 18 habitats in France ranged from a value of 0·23 in the generalist Common Blackbird to 1·16 in the specialist Eurasian Skylark. In this study, specialists were more common in a given locality if the rest of the community was also specialised, and most species reached maximum abundance when the community in which they occurred had a specialisation index similar to their own index. Therefore, as habitats become degraded, habitat generalists come to predominate, and the overall level of specialisation of the community decreases. In France, specialists have been declining much more since 1990 than generalists, even when controlling statistically for other variables such as climatic requirements and life history, and this decline was probably caused by habitat degradation(Jiguet et al. 2007). This makes sense since most communities in environments influenced by humans are to a large extent composed of generalists, while specialists, owing to their peculiar habitat requirements, are poorly adapted to changing environments. Taken together, these results suggest that specialist species will be the most severely affected by climate change and consequent habitat degradation and domination of generalist species.

Impoverished communities

Climate change may affect different populations or species in different ways, with consequences for the composition of communities of birds (Stralberg et al. 2009).Extensive modelling has suggested that species richness may decline as a consequence of climate change (Huntley et al. 2007, Jetz et al. 2007). Changes in distribution of species derived from the European climatic bird atlas (Huntley et al. 2007) were used by Gregory et al. (2009) to predict large-scale patterns of population change for birds in Europe, and these predictions were subsequently tested against observed population trends of species with expanding and shrinking breeding ranges. The results showed that population trends of species benefiting from climate change and species suffering from climate change have grown increasingly more different over the last 20 years, again suggesting substantial climate-induced changes in community composition. Ina similar vein, Pounds et al. (1999) reported strong declines in species richness for Monteverde in Costa Rica, in this case owing to reduced dry-season mist causing changes in altitudinal limits of montane specialist birds.

The thermal range of individual species will clearly affect such changes in community structure, but those species worst affected may be most likely to come from groups with particular food sources or life histories (Devictor et al. 2008). Climate change may particularly impact long-distance migrants that are generally insectivorous, through its effects on the abundance and diversity of invertebrates. Such predicted changes in communities can be confronted with observed changes in an attempt to test how well actual observations fit predictions. Lemoine et al. (2007) and Lemoine & Böhning-Gaese (2003) showed a declining abundance of migrants in the Lake Constance area as predicted from spatial variation in abundance across Europe, providing one of the few examples where observed community composition has been found to reflect the prediction, even when considering several other possible explanations for a decline in abundance of migrants.

Conservation implications of climate change

Which are the species that are particularly susceptible to climate change?

Climate change clearly has important implications for conservation, given the expected(or indeed already observed) effects of climate change on population declines and risk of extinction. This is starkly illustrated by the finding that species that have failed to respond to climate change are precisely those that have shown the strongest recent declines in population size (Figure 27; Møller et al. 2008). These declines were specifically recent ones, subsequent to 1990, which is when climate change started to become most severe. This is strong evidence that it is specifically these species’ absence of response to climate change that is responsible for their decline, rather than some other factor. Other possible factors that are commonly related to population declines in birds include habitat use (such as whether species occur in agricultural landscapes), the northernmost limit of a species’ breeding ranges, and its body size, but none of these detracted from the climate change effect in Møller et al.’s (2008)study. We can take this analysis one step further by considering whether species that have shown such effects have any shared characteristics that put them at particular risk, such as use of agricultural habitats and long-distance migration.

Agricultural intensification is thought to be responsible for declines in abundance of temperate birds living in agricultural habitats (Fuller et al. 1995). However, very few studies have investigated the combined effects of climate change and agriculture on population trends of birds. Møller et al. (2008) showed a significant negative effect of agriculture for the period 1970–1990, but not after 1990 when response to climate change became the main statistical predictor of population trends. This is as expected because agricultural intensification was particularly strong during the first period (Fulleret al. 1995, Burfield & van Bommel 2004), although an alternative interpretation is that agricultural modification of bird populations during that period may have reduced populations so much that no further changes have happened.

Long-distance migratory birds have shown dramatic declines in recent years (Sandersonet al. 2006), but since long-distance migrants have also shown weaker responses to climate change (Lehikoinen et al. 2004, Lehikoinen & Sparks 2010, Végvári et al.2010), it is interesting to try to partition the role of climate change from the role of other factors affecting population declines of long-distance migrants. This can be done by considering the two simultaneously in the same statistical model, which revealed that declines in populations of long-distance migratory birds were actually relatively weak when the response to climate change is taken into account (Møller et al. 2008).This suggests that climate change is the key factor. A study from the Netherlands also emphasised the role of climate over migration per se: long-distance migrants in forests declined more than those in marshes, owing to a stronger effect of climate on the seasonal food peak in forests than in marshes (Both et al. 2010). Further correlational support for an effect of climate was the fact that population declines were the strongest in Western Europe, where spring temperatures have increased the most.

A final factor that seems likely to affect a species’ susceptibility to climate change is the latitudinal range of its distribution. Indeed, bird species with a northerly breeding distribution have shown greater population declines than species with a more southerly distribution (Julliard et al. 2003). More northerly species may be more susceptible to climate change owing to the greater negative effects of climate change on biological diversity (which may include, for example, their food supplies) at northern latitudes (Parmesan & Yohe 2003).

In order to improve our future understanding of the factors affecting the susceptibility of different species to climate change, it is important to consider how best to monitor the consequences of climate change for bird populations. The efficiency of bird monitoring programmes as means of studying bird populations has been analysed by relating interspecific variation in population trends (as assessed by monitoring programmes) to changes in potential breeding range between the late 20th and late21st centuries, as forecasted by climate envelope models (Gregory et al. 2009). Species that were predicted to increase the size of their breeding range increased in population size as shown by the European bird monitoring programme, whereas species predicted to decline in range size decreased in population size. Moreover, an index oft he impact of climate change on overall trends in bird populations also increased rapidly during the last 20 years, when climate changed particularly rapidly. These findings are encouraging in the sense that they suggest that existing monitoring programmes are capturing biologically meaningful data on population trends. However, the recent acceleration in population declines, coinciding with the worsening of recent climate change, needs careful examination to tease apart the effects of climate change from those of other potentially confounding or interacting factors. This is because numerous other factors have also increased in severity during the last two decades, including intensification of agriculture, forestry and fisheries, and increased urbanisation and pollution. Indeed, two studies of seabirds have shown that significant interaction sexist between the effects of climate change and other factors including fisheries (Barbraud et al. 2008) and agriculture (Møller et al. 2007). Given how many threats to the viability of bird populations have increased in the last 20 years just as climate has, caution needs to be taken in interpreting indices such as that described above. Ideally, an index of the impact of climate change should be developed that statistically controls for other potentially confounding drivers of population declines such as those listed above.

A partial attempt at disentangling these multiple confounding factors was made in a study of 61 species of rare breeding birds in the UK (Green et al. 2008). Population trends were correlated with trends in climate suitability (as estimated by climate envelope models), and this effect was shown to be statistically independent of several possibly confounding factors including proximity to the European breeding range, body mass, migration and habitat. However, several other variables that are well established to be important drivers of population trends in breeding birds (such as agriculture, forestry, fisheries, urbanisation and pollution) were unfortunately not investigated.

Overall, the relative role of climate change and other possible drivers of global change for bird populations is a major concern. Only a small minority of studies has considered both categories of drivers, so we know virtually nothing about their relative importance. One such study, by Jetz et al. (2007), analysed the relative impact of climate change and land-use change on global diversity of birds using the Millennium Ecosystem Assessment scenarios. Even under relatively benign environmental scenarios(taking into account both climate and land-use change), a staggering 400 species are projected to suffer a reduction in range by more than half by 2050, and no fewer than 900 by 2100. However, these dramatic predictions mainly arose from the effects of land-use change, with climate change having a substantial effect mainly at high latitudes. The species in danger of range loss are mainly endemics with narrow ranges in the tropics. However, these analyses may be considered overly optimistic because the geographic ranges of species were assumed to remain stationary overtime, whereas in reality range margins are likely to shift owing to climate change. Likewise, the assumption that all bird species will have sufficient time to adapt to climate change seems particularly optimistic given the strong negative relationship between threat status and response to climate change (Møller et al. 2007).

These modelling uncertainties imply that we need to look for empirical evidence to assess the issue. Clavero et al. (2011) analysed whether climate change indicators can be influenced by changes in land-use that are in fact not directly driven by climate change. They did so by simultaneously examining how strongly indicators of climate change were related to land-use (with respect to land abandonment, fire and urbanisation)as well as to temperature. Surprisingly, the three land-use indices had as strong an impact on indicators of climate change as temperature. This means that changes inland-use can reverse, hide or even exacerbate apparent impacts of climate change if not controlled statistically. Clearly, many studies have been overly naïve by considering that a correlation between a single climate variable and a biological phenomenon would reflect a causal effect of climate alone, when in fact many other temporally varying factors may play an equally important role. These results call for caution in interpreting models that project changes in the abundance or distribution of species based on climate change alone, since they may not prove to be informative in identifyingthe most crucial conservation issues facing bird populations.

What can we do to mitigate the effects of climate change?

Given the grim outlook imposed by climate change exacerbated by the effects of numerous other global changes, what can we do to reduce the negative impact of climate change on bird populations? The first objective must be to do rigorous and thorough research to identify the problems. Many amateurs have collected particularly valuable data in the form of ringing records of individual breeding birds over very long periods. Recently I made a compilation of such long-term studies of individually identifiable birds, which showed that there have been more than 200 such studies covering time period of more than ten years. This is more than the number of comparably longtermstudies of all other kinds of animals combined. Many of the studies reviewed in this essay have used such data, through collaborations between amateur and professional biologists. This emphasizes how such data are extremely valuable for analyzing the effects of climate change, as well as for addressing many other fundamental scientific questions. There is scope for many more such studies.

Numerous amateurs and professionals contribute to national bird census programmes, one of the objectives of which is to study the effects of climate change on bird populations. Such long-term schemes are extremely valuable for monitoring populations and for generating information on the factors that constitute threats to birds, and therefore to numerous other kinds of animals and plants. There are many fewer such detailed population studies or monitoring programmes in the tropics and the southern hemisphere, and this lack of knowledge is a major obstacle to our efforts to assess and tackle problems of climate change.

There is no time for complacency, because there is a sizeable community of climate change sceptics who only wait for errors to be made. Too many studies show only a relationship between a biological variable and a single climatic variable, while not considering alternative explanations. Numerous global changes occur simultaneously, and each of these needs careful consideration and analysis. This is crucial to avoid making an open invitation to criticism and reducing the credibility of the scientific community at large.


Birds can be viewed as harbingers of climate change, and they are in a way all canaries in the coal mine of our common environment, experiencing dramatic and sometimes irreversible changes. Numerous other global changes have negative effects on birds and other organisms, and they are likely to interact synergistically with the effects of climate change. These combined threats make it much more difficult for birds to adapt phenotypically and genetically to global change.

A prominent question often raised is as follows: why care about birds when hundreds of millions of humans are starving, and hundreds of additional millions are about to appear into the world? Scientific insights that we learn from birds and other organisms may inform general mitigation of climate change. Climate change will impact on humans, but disproportionately so on poor people in developing countries. There are no simple or easy solutions to the problems created by climate change, because they are a consequence of short-sighted human decisions. We must all act at the personal, community, national and international levels to mitigate and reverse these trends. Only through common and concerted action by many people showing an interest in our common environment will it be possible to reverse the continuing decay of our planet, to the benefit of all living beings alike.


Anders Pape Møller



J. del Hoyo and A. Elliott kindly invited me to write this foreword. N. Saino and many other colleagues influenced my thinking on effects of climate change on birds.


The full references to the citations included herein can be found towards the back of the volume, in the General List of References