Population limitation in migrants

Abstract

Unlike resident bird species, the population sizes of migratory species can be influenced by conditions in more than one part of the world. Changes in the numbers of migrant birds, either long-term or year-to-year, may be caused by changes in conditions in the breeding or wintering areas or both. The strongest driver of numerical change is provided in whichever area the per capita effects of adverse factors on survival or fecundity are greatest. Examples are given of some species whose numbers have changed in association with conditions in breeding areas, and of others whose numbers have changed in association with conditions in wintering areas. In a few such species, the effects of potential limiting factors have been confirmed locally by experiment. In theory, population sizes might also be limited by severe competition at restricted stopover sites, where bird densities are often high and food supplies heavily depleted, but (with one striking exception) the evidence is as yet no more than suggestive. In some species, habitats occupied in wintering and migration areas, and their associated food supplies, can influence the body condition, migration dates and subsequent breeding success of migrants. Body reserves accumulated in spring by large waterfowl serve for migration and for subsequent breeding, and females with the largest reserves are most likely to produce young. Hence, the conditions experienced by individuals in winter in one region can affect their subsequent breeding success in another region. Such effects are apparent at the level of the individual and at the level of the population. Similarly, the numbers of young produced in one region could, through density-dependent processes, affect subsequent overall mortality in another region. Events in breeding, migration and wintering areas are thus interlinked in their effects on bird numbers. Although in the last 30–40 years the numbers of some tropical wintering birds have declined in western Europe and others in eastern North America, the causes seem to differ. In Europe, declines have mainly involved species that winter in the arid savannas of tropical Africa, which have suffered from the effects of drought and increasing desertification. In several species, annual fluctuations in numbers and adult survival rates were correlated with annual fluctuations in rainfall, and by implication in winter food supplies. In North America, by contrast, numerical declines have affected many species that breed and winter in forest, especially those eastern species favouring the forest interior. Declines have been attributed ultimately to human-induced changes in the breeding range, particularly forest fragmentation, which have led to increases in the densities of nest predators and parasitic cowbirds. These in turn are thought to have caused declines in the breeding success of some neotropical migrants, which is now too low to offset the usual adult mortality, but as yet convincing evidence is available for only a minority of species. The breeding rates and population changes of some migratory species have been influenced by natural changes in the availability of defoliating caterpillars. In other species, tropical deforestation is likely to have played the major role in population decline, and if recent rates of tropical deforestation continue, it is likely to affect an increasing range of migratory species in the future. Not all such species are likely to be affected adversely by deforestation, however, and some may benefit from the resulting habitat changes.

In the limitation of their populations, migratory birds differ from residents in at least one important respect. Their population sizes may be influenced by conditions in more than one part of the world: in areas that are used for breeding, as well as in areas that are visited at other times of year. For many migrant species, the breeding and wintering ranges are widely separated geographically, and might differ greatly in the numbers of birds they can support. Hence, factors operating in the migration or wintering range might limit the numbers that can occur in the breeding range, or vice versa. This is most clearly apparent where changes in the numbers of a species over a period of years are associated with changes in conditions in one area, but not in the other. In this review, I address some of the issues involved in the population limitation of migrants, using as examples species whose long-term or year-to-year numerical changes can be clearly linked to events in breeding or non-breeding areas. Special attention is given to landbird migrants that breed in Eurasia and winter in Africa (Moreau 1972, Newton 1994), or that breed in North America and winter in Central and South America (Terborgh 1989, Rappole 1995, Sherry & Holmes 1995). Most such species spend less of each year in their breeding areas than in their migration and wintering areas.

In understanding population changes, it is helpful to separate long-term trends from year-to-year fluctuations about the trend, because different factors might be involved (Newton 1998). For example, some species may undergo long-term change in numbers, resulting from progressive habitat change, but may also continue to fluctuate from year to year in response to different factors, such as annual variations in rainfall and associated food supplies. Moreover, within species, population trends and limiting factors may vary across the range, and also at different time periods, so that local research findings cannot necessarily be extrapolated over wider areas or longer periods (see Peterjohn et al. 1995 for analysis of recent regional trends in North American species). Because of the problems of studying birds that routinely occupy two or more different areas each year, it has seldom been possible to examine the same individuals throughout the year, and most of the evidence on population limitation in migrants is indirect.

Spacing and movement patterns

The spacing behaviour of birds in the non-breeding season can vary from solitary and territorial to gregarious and flocking, depending on the species, the habitat, the distribution of food supplies, predation pressures and other factors, as well as on the social status of the individuals concerned (Newton 1998). In addition, whereas some migrant birds remain in the same place for the whole non-breeding period from arrival to departure, others occupy two or more sites in succession, at different points on the migration route, or they may remain continually on the move, again depending largely on the spatial and temporal occurrence of food supplies. In the northern continents, such itinerancy is shown by boreal finches and others, whereas in the drier parts of both Africa and South America, some migrant species follow rain-belts for the food sources they promote. Hence, if migrants are limited by conditions in wintering areas, for some species there may be more than one such area involved, and these areas may differ from year to year. Some birds may spend different amounts of time in the same localities each year, and reach localities in some years that remain vacant in others. The bulk of the population may thus be concentrated in different areas in different years, depending on the distribution of food. African examples include the Lesser Spotted Eagle Aquila pomarina and White Stork Ciconia ciconia (both studied by radiotracking, Meyburg et al. 1995, Berthold et al. 2002), and South American examples on a smaller spatial scale include the Eastern Kingbird Tyrannus tyrannus and Swainson's Hawk Buteo swainsoni (Brown & Amadon 1968, Morton 1971). They illustrate a major difference between the breeding and non-breeding seasons: in the former, breeding birds are tied for up to several months to the localities in which they nest, but in the non-breeding season, individuals of at least some species are free to move around, concentrating wherever food is plentiful. Those migrants that ‘winter’ in the southern hemisphere, while the local birds are breeding, thus have an advantage that the local birds then lack, and in theory this could reduce competition between the two groups.

SOME GENERAL PRINCIPLES

If we accept that, for various reasons, the total carrying capacities of the breeding and non-breeding habitats of particular populations need not necessarily correspond, then two scenarios are possible when birds return to their nesting areas each year and compete for territories or nest-sites (Fig. 1):

Figure 1. Model showing seasonal changes in total bird numbers in relation to the carrying capacity of the breeding area (thick line). In the lower curve (A) numbers left at the end of the non-breeding season are fewer than the nesting habitat could support, and in the upper curve (B) numbers are greater than the nesting habitat could support, leading to a surplus of non-territorial, non-breeders. In (A), breeding numbers are limited by conditions in the area occupied in the non-breeding season, and in (B) by conditions in the area occupied in the breeding season. From Newton (1998).

• 1
  Too few birds are left at the end of the non-breeding season to occupy all nesting habitat fully, so that practically all individuals of appropriate age and condition can breed. In this case, breeding numbers would be limited by whatever factors operate in the non-breeding season, and alleviation of the factors that influence the extent or carrying capacity of non-breeding habitats would be needed before breeding numbers could rise. For convenience, the breeding numbers in such populations could be described as ‘winter-limited’.
• 2
  More birds are left at the end of the non-breeding season than the available nesting habitat can support, producing a surplus of non-territorial, non-breeders. In this case, alleviation of the factors that influence the extent or carrying capacity of the nesting habitat would be needed before breeding numbers could rise. For convenience, the breeding numbers of such populations could be described as ‘summer-limited’.

The same two types of scenario could hold at the end of the breeding season, as birds return to their non-breeding quarters. At that time, the total numbers of adults and young might be insufficient to use the resources of non-breeding areas to the full, leading to good survival through the non-breeding season; alternatively, the total numbers may exceed the carrying capacity of non-breeding areas, leading to intense competition and poor survival.

Because bird reproduction is by definition confined to the breeding range, numbers there might be ‘summer limited’ in a different way, namely if breeding success were so poor that subsequent breeding numbers could not reach the level necessary to fill either the available breeding habitat or the available non-breeding habitat. This situation could arise even with relatively good year-round survival. In other words, whereas failure to occupy all wintering habitat must be due to events in breeding or migration areas, failure to occupy all breeding habitat could be due to events in either breeding, migration or wintering areas. Understanding the limiting mechanisms in any one population may not be easy, especially if the situation changes from year to year. There is no reason why a species might not be summer-limited in one year or area and winter-limited in another year or area (for examples see Newton 1998).

Effects of habitat loss on migrants

It is a matter of observation that, when areas of habitat are lost or added through human action, bird numbers often change accordingly. To take some recent examples from Britain, Siskins Carduelis spinus have increased in numbers and expanded in range as new conifer plantations have matured and provided additional breeding habitat (Gibbons et al. 1993). By contrast, Redshanks Tringa totanus have declined and contracted as summer nesting habitat has been destroyed by land drainage (Norris et al. 2004), and Twite Carduelis flavirostris have declined as areas of winter salt marsh have been lost to reclamation projects (Atkinson et al. 2004). Many other examples have been described in the literature from Europe and North America, including the massive declines in numbers of waterbirds that followed the drainage of marshes. Although many such examples may represent causal relationships, some may be due to coincidence between population decline (caused by some other factor) and habitat loss. Clearly, not all bird population changes occur in response to habitat changes, and some species seem to have far more potential habitat than they currently occupy. In the case of migrants, this again raises the possibility that numbers may be limited at one end of the migratory terminal at a level lower than habitat at the other end would support.

In recent years, much thought has been given to predicting the effect of habitat loss (equivalent to food loss) in both resident and migratory bird species. For resident birds, in which breeding and wintering areas are the same, population declines should be roughly proportional to habitat loss, if habitat were of uniform quality and fully occupied throughout. In other words, if half the habitat (or food supplies) were lost, we would in general expect the population to be roughly halved (but see later). The situation is more complicated for migrant birds because they occupy separate areas in winter and summer, and habitat loss may occur in one area or both (Fig. 2). The actual population change following loss of breeding or wintering habitat would be expected to depend on where the tightest bottleneck occurred; that is, the relative strengths of density-dependent constraints in the two areas (Sherry & Holmes 1995, Sutherland 1996). Such constraints include those various pressures, such as competition for space and food, that can affect an increasing proportion of individuals as their density rises, resulting in an increased per capita mortality or decreased per capita reproduction. In the wintering area, the strength of density-dependence is measured by the per capita rate of increase in mortality that occurs as a result of rising population size (or decreasing area) (slope d). In the breeding area, the strength of density-dependence is measured by the per capita rate of decrease in reproduction with rising population size (or decreasing area) (slope b). If slope b > slope d, loss of breeding habitat would have most impact on overall population size, and if d > b, loss of wintering habitat would have most impact. If the two density-dependent relationships were known, the effect of loss of habitat (or food supply) on equilibrium population size could in theory be calculated as b/(b + d) for the breeding area, or as d/(b + d) for the wintering area.

Figure 2. Model of the relationship between per capita mortality and reproduction (continuous lines), and equilibrium population size (E). Per capita winter mortality increases, and per capita reproduction decreases, with increasing population size. The equilibrium population size is where the two lines intersect. The model shows how loss of habitat (or food supply) results in population decline. If wintering habitat (or food supply) is reduced by 50%, this results in a displacement of the relationship between mortality and total population size in direct proportion to the degree of habitat loss (dashed line). The equilibrium population size is reduced accordingly (E₁). A 50% loss of breeding habitat (or food supply) similarly results in a displacement of the relationship between net breeding output and total population size (dashed line) and a reduced equilibrium population (E₂). In this example, 50% loss of breeding habitat results in a smaller reduction in equilibrium population size than does 50% loss of wintering habitat, because density-dependence is stronger in winter than in summer. Modified from Sutherland (1996).

If, in an extreme case, all the density-dependence occurred in winter habitat (say), with no density-dependence in breeding habitat (which at prevailing population levels was present in excess, as in Fig. 1a), then loss of winter habitat would cause a matching reduction in population size. In this situation, loss of breeding habitat would have no effect up to the point at which density-dependent decline in breeding success set in. Although as yet there can be few species for which enough information is available to test the model in Figure 2, or to judge the form of density-dependent relationships over a wide range of densities at both seasons, attempts have been made for the Oystercatcher Haematopus ostralegus in Britain (Goss-Custard et al. 1995, Sutherland 1996).

The above considerations lead to a number of conclusions regarding the effects of habitat (or food) loss on the equilibrium population sizes of migrants: (1) knowledge of the density-dependent response within just the wintering or breeding area cannot be used to predict precisely the effects of habitat or food loss in either, because it is the ratio of density-dependence in the two areas that is important; (2) unless there is no density-dependence acting during one of the seasons, a loss of habitat or food supply in either summer or winter areas could result in population decline; and (3) the consequence of habitat or food loss is greatest for the season in which density-dependence is strongest (winter in Fig. 2). In practice, all migrants are likely to be affected more by changes in one area than the other, although whether breeding or wintering areas are most important in this respect may change through time. They could also change from year to year in species subject to large annual fluctuations in habitat, food supplies or other conditions.

The above generalizations on the role of summer and winter conditions hold most clearly for populations limited by resources – by the available habitats and food supplies. They could also hold for populations limited below the levels that resources would permit by factors such as parasitism, predation and human persecution. However, in some circumstances, the latter factors can also kill an unsustainably large number of individuals each year, sending populations into decline, and leaving a surplus of unused habitat and food. For example, if for some reason the predation pressure on eggs and chicks in the breeding areas increased so much that loss of annual production could not be offset by improved annual survival, the population would decline below the levels that both breeding and wintering habitats would support. Similarly, if shooting pressure on full-grown birds increased in winter quarters, so that the loss could not be offset by improved reproduction or natural survival, the population could again decline below the carrying capacities of both breeding and wintering habitats. In both these examples, decline would continue while that situation held (eventually to extinction), the trend being driven primarily in whichever area the per capita effects of adverse factors on reproduction or survival were greatest.

The buffer effect and density dependence

All habitats seem to vary from place to place in quality and attractiveness to the birds they support. One known mechanism through which density dependence in mortality or reproduction could occur during a period of population growth involves the ‘buffer effect’. This occurs when birds occupy the best habitat areas first, and as they fill these areas to capacity, they spread increasingly to poorer areas as their numbers continue to rise. As survival or reproduction is lower in the poorer areas, the mean per capita performance in the population as a whole declines as overall numbers grow, in a density-dependent manner. Sequential habitat fill of this type is seen: (1) as birds arrive in their breeding areas in spring, or their wintering areas in autumn, when they occupy the best places first, so that later arrivals are relegated to poorer places (e.g. Brooke 1979, Lundberg et al. 1981, Goss-Custard et al. 1984); (2) in the annual fluctuations of populations, where numbers remain more stable from year to year in the preferred habitats (or territories) than in the secondary habitats (Kluijver & Tinbergen 1953, Zimmerman 1982, Rodenhouse et al. 2003); and (3) in the progressive occupation of habitat areas (or territories) of different quality, as a population grows over a period of several years (e.g. Mearns & Newton 1988, Ferrer & Donázar 1996, Löhmus 2001). All these processes, which result from habitat variation, can help to regulate bird populations (for further examples of each type see Newton 1998). Food and other resources are involved in the regulation because they influence the quality and carrying capacity of habitats.

The Icelandic population of the Black-tailed Godwit Limosa l. islandica wintering in Britain has risen four-fold since the 1970s, but rates of increase within individual estuaries have varied from zero to six-fold (Gill et al. 2001). In accordance with the buffer effect, rates of increase were greatest on estuaries with low initial numbers, and Godwits on these sites were found to have lower prey intake rates and lower survival rates than Godwits on longer-occupied sites with stable populations. Godwits from the poorer, more recently occupied, wintering sites also arrived later in spring on their Icelandic breeding areas. Their breeding has not been studied, but by analogy with other species, later arrival usually means relegation to poorer habitat and poorer breeding success. In this species, therefore, population growth could have resulted in a progressively larger proportion of the population wintering in poorer habitat, with measured consequences on feeding rates and migration dates, and possible consequences on breeding success. The buffer effect, acting on a large spatial scale, could therefore have been a major density-dependent process acting to constrain population growth in this migratory species.

A similar spread to poor sites during a period of population growth was earlier noted in Grey Plovers Pluvialis squatarola wintering in different parts of Britain (Moser 1988), in Brent Geese Branta bernicla wintering in the Netherlands (Ebbinge 1992) and in Great Cormorants Phalacrocorax carbo wintering in Switzerland (Suter 1995). In the last of these, the process was stepwise, and each category of habitat experienced a rapid build-up in numbers, followed by stabilization, before the next type of habitat was occupied. In each of these studies, the first-filled habitat was assumed to be better, in which case survival would have been poorer in the secondary habitats, leading to progressive decline in mean per capita survival as the population grew, although this was not confirmed (but for effects on reproduction of Brent Geese see Ebbinge 1992). Hence, although the presence of secondary habitat would permit a population to attain a higher level than it could in the primary habitat alone, the poorer performance of individuals in secondary habitat would put ever increasing constraints on further population growth.

EXAMPLES OF SPECIES AFFECTED BY EVENTS IN BREEDING OR WINTERING AREAS

In some species, year-to-year changes in overall population levels have been clearly driven largely by conditions in breeding areas. On the North American prairies, rainfall varies greatly from year to year, and this influences the amount of wetland habitat available to nesting waterfowl. In wet periods, populations increase, and in dry periods they decline. So important are these prairie wetlands as nesting habitat that they influence the entire continental wintering populations of several species, including the American Coot Fulica americana (Fig. 3). They show how year-to-year conditions in the breeding areas can largely determine year-to-year fluctuations in total populations.

Figure 3. Relationship between the numbers of American Coots Fulica americana shot each year in the United States (reflecting total population size) and the number of ponds on the prairies in the preceding summer (reflecting the habitat and feeding conditions). Significance of relationship: b = 0.21, r² = 0.54, P < 0.001. Redrawn from Alisauskas and Arnold (1994).

By contrast, the numbers of several migrant songbird species counted each spring on their European breeding areas have fluctuated according to rainfall (and hence food supplies) in their African wintering areas. Examples include the Sedge Warbler Acrocephalus schoenobaenus, Sand Martin Riparia riparia and Purple Heron Ardea purpurea (den Held 1981, Peach et al. 1991, Bryant & Jones 1995, Szép 1995). They show how conditions in wintering areas can largely determine year-to-year fluctuations in total populations.

In some migratory species, breeding density declined, and it was not immediately obvious whether the causal factors lay in breeding or wintering areas. However, where fecundity and survival rates were monitored in the same population during periods of both increase (or stability) and decrease, the comparison provided useful pointers to where the cause of the decline might lie. Thus, where annual reproduction declined while annual survival stayed the same, the problem lay in the breeding areas, but where survival had decreased, the problem lay in either breeding or wintering areas, depending on the time of year the extra deaths occurred. Decline in a European Golden Plover Pluvialis apricaria population was associated with a decline in survival rate but no change in reproduction, whereas a decline in a Northern Lapwing Vanellus vanellus population was associated with a decline in reproduction but no change in survival (Parr 1992, Peach et al. 1994, Yalden & Pierce-Higgins 1997).

In yet other species, long-term population declines were associated with reductions in both breeding and survival. The White Stork, for example, has suffered from reduced reproduction on its European breeding grounds, caused by drainage and pesticide-induced food shortages, and also from reduced survival on its West African wintering grounds, caused mainly by drought-induced and pesticide-induced food shortages (mainly locust control) (Dallinga & Schoenmakers 1989, Kanyamibwa et al. 1993, Bairlein 1996). Thus, in any avifauna we can expect to find species whose numbers are changing because of events in breeding or non-breeding areas or both, and routine monitoring of fecundity or survival rates can often be enlightening.

Other examples of summer-influenced and winter-influenced population changes in migrants, involving more than 53 different populations of 44 species, are given in Table 1. In most populations, the evidence is entirely circumstantial, and based on long-term or year-to-year correlations between changes in breeding numbers and changes either in (1) conditions in breeding or wintering areas, or in (2) associated breeding or mortality rates. In some populations, however, potential causal relationships were subsequently confirmed by experiments involving manipulation of likely limiting factors (see later).

Table 1.  Migratory bird species in which temporal changes in breeding density have been linked with changes in previous winter conditions (A), with changes in summer conditions (B), or with changes in both winter and summer conditions (C). Updated from Newton (1998).

Species	Location	Long-term upward or downward trend	Annual fluctuations	Source
A. Associated with change in winter conditions/survival
Twite Carduelis flavirostris	Germany–Netherlands	+		Dierschke (2002)
Snow Bunting Plectrophenax nivalis	Germany–Netherlands	+		Dierschke (2002)
Shorelark Eremophila alpestris	Germany–Netherlands	+		Dierschke (2002)
Sedge Warbler Acrocephalus schoenobaenus	England		+	Peach et al. (1991)
	Netherlands		+	Foppen et al. (1999)
Blackcap Sylvia atricapilla	Britain		+	Baillie and Peach (1992)
Whitethroat Sylvia communis	Britain	+	+	Winstanley et al. (1974), Baillie and Peach (1992)
	Sweden		+	Hjort and Lindholm (1978)
Willow Warbler Phylloscopus trochilus	Britain	+	+	Baillie and Peach (1992), Peach et al. (1995)
Loggerhead Shrike Lanius ludovicianus	Minnesota	+		Brooks and Temple (1990)
Barn Swallow Hirundo rustica	Denmark		+	Møller (1989)
	Britain		+	Robinson et al. (2003)
Sand Martin Riparia riparia	Britain		+	Cowley (1979), Bryant and Jones (1995)
	Hungary		+	Szép (1995)
Avocet Recurvirostra avosetta	England		+	Hill (1988)
Golden Plover Pluvialis apricaria	Scotland	+		Parr (1992), Yalden and Pearce-Higgins (1997)
Puffin Fratercula arctica	Scotland	+	+	Harris and Wanless (1991)
Night Heron Nycticorax nycticorax	France		+	den Held (1981)
Purple Heron Ardea purpurea	Netherlands		+	den Held (1981), Cavé (1983)
Barnacle Goose Branta leucopsis	Svalbard	+		Owen (1984)
Dark-bellied Brent Goose Branta bernicla	Netherlands	+		Ebbinge (1991)
Lesser Snow Goose Chen caerulescens	Canada	+		Francis et al. (1992)
B. Associated with change in summer conditions/survival/breeding rate
Chaffinch Fringilla coelebs	Russia		+	Sokolov (1999)
Willow Warbler Phylloscopus trochilus	Russia		+	Sokolov (1999)
Icterine Warbler Hippolais icterina	Russia		+	Sokolov (1999)
Kirtland's Warbler Dendroica kirtlandii	Michigan	+		DeCapita (2000)
Black-throated Blue Warbler	New Hampshire	+	+	Holmes et al. (1991, 1996), Rodenhouse et al. (2003)
Dendroica caerulescens
Prairie Warbler Dendroica discolor	Indiana		+	Nolan (1978)
Wilson's Warbler Wilsonia pusilla	California		+	Chase et al. (1997)
Bell's Vireo Vireo bellii	California	+		Griffith and Griffith (2000)
	Missouri	+		Budnick et al. (2000), Rothstein and Robinson (1994)
Black-capped Vireo Vireo atricapillus	Texas	+		Hayden et al. (2000)
Pied Flycatcher Ficedula hypoleuca	Finland		+	Virolainen (1984)
	England		+	Stenning et al. (1988)
	Russia		+	Sokolov (1999)
American Redstart Setophaga ruticilla	New Hampshire		+	Sherry & Holmes (1992)
Swainson's Thrush Catharus ustulatus	California		+	Johnson & Geupel (1996)
Wood Thrush Hylocichla mustelinus	Illinois		+	Robinson (1992)
	Delaware		+	Roth and Johnson (1993)
Corncrake Crex crex	Britain	+		Green et al. (1997), Green (1999)
Common Sandpiper Actitis hypoleucos	Britain		+	Hollands and Yalden (1991)
Dunlin Calidris alpina	Fennoscandia	+	+	Soikkeli (1970), Jönsson (1991)
Curlew Sandpiper Calidris ferruginea	South Africa		+	Summers and Underhill (1987)
Little Stint Calidris minuta	South Africa		+	Summers and Underhill (1987)
Lapwing Vanellus vanellus	Britain	+		Peach et al. (1994)
Stone Curlew Burhinus oedicnemus	England	+		Aebischer et al. (2000)
Mallard Anas platyrhynchos	North America		+	Reynolds (1987)
Various duck species	Iceland		+	Gardarsson and Einarsson (1994)
Brent Goose Branta bernicla	Britain		+	Summers and Underhill (1987)
Pink-footed Goose Anser brachyrhynchus	Svalbard	+		Madsen et al. (2002)
Arctic Tern Sterna paradisaea	Britain	+		Suddaby and Ratcliffe (1997)
C. Associated with change in both winter conditions (overwinter survival) and summer conditions (previous breeding rate)
White Stork Ciconia ciconia	France–Germany	+	+	Dallinga and Schoenmakers (1989), Kanyamibwa et al. (1993), Bairlein (1996)
Great Skua Stercorarius skua	Scotland	+	+	Klomp and Furness (1992)

In some of the bird populations mentioned in Table 1, breeding numbers increased in years that followed a good breeding season and decreased in years that followed a poor breeding season (Fig. 4). In these populations, spring–summer conditions in breeding areas evidently had most influence on subsequent year-to-year changes in breeding numbers, and such populations were therefore below the limit imposed by winter habitat (at least in most years). The same was true for other populations whose survival rates changed over the years, according to conditions on breeding areas. In yet other populations, breeding numbers varied from year to year according to previous winter conditions, implying that such populations were close to the limit imposed by winter habitat (or food supplies).

Figure 4. Demography and population change. (a) Relationship between annual survival of adult Barn Swallows and annual change in breeding density. From Møller (1989). (b) Relationship between annual breeding output of Pied Flycatchers and annual change in breeding density. From Virolainen (1984). Significance of relationships: Barn Swallow, b = 3.44, r² = 0.93, P < 0.001; Pied Flycatcher, b = 0.15, r² = 0.74, P < 0.001.

If we can assume that breeding rates were influenced primarily by conditions in breeding areas and mortality rates primarily by conditions in wintering areas (unless otherwise specified), then 22 of the populations in Table 1 were winter-limited, another 29 or more were summer-limited, and two were influenced by both summer and winter conditions. However, because both the extent and carrying capacities of habitats vary from year to year, and from area to area, we can expect that the same species might be winter-limited in some years or areas and summer-limited in other years or areas, as in the different Willow Warbler Phylloscopus trochilus populations given in Table 1. Each case must be judged on its particular circumstances.

Over several years, the same population might change from one state to another, as its status with respect to available habitat changed. Several species of geese increased during the latter half of the 20th century in response to reduced shooting pressure in their wintering areas, but then came up against food shortage in the breeding areas, as growing numbers competed for favoured food plants. This increased competition resulted in reduced chick survival in Lesser Snow Geese Chen caerulescens in the central Canadian arctic (Francis et al. 1992), and in reduced summer survival among adult Pink-footed Geese Anser brachyrhynchos on Svalbard (Madsen et al. 2002). The major constraint to further population growth thus shifted from the wintering to the breeding areas as the populations grew. In some other goose populations, studied in less detail, increasing competition was manifest chiefly in declining proportions of young in wintering flocks (Fig. 5). In Brent Geese wintering in western Europe, total numbers fluctuated from year to year around the long-term upward trend, according to annual variations in predation rates on eggs and chicks in Siberian breeding areas (Summers & Underhill 1987).

Figure 5. Reduced production of young by Greylag Geese Anser anser and Barnacle Geese Branta leucopsis, as their numbers have grown. The populations concerned breed in Iceland (Greylag) and Svalbard (Barnacle) and winter in Britain, where counts were made. The graphs show 5-year moving mean values. Adapted from Owen et al. (1986).

The pattern in which breeding density fluctuated from year to year in parallel with the previous year's breeding success was recorded only in short-lived species, in which individuals breed in their first year of life (passerines and dabbling ducks). It would not be expected in longer-lived species, in which individuals do not breed until they are two or more years old, and in which annual recruitment rates are naturally low. In such species, breeding success would need to be poor over several years before any effect on breeding numbers became obvious, unless accompanied by a simultaneous increase in mortality or emigration (for Arctic Tern Sterna paradisaea see Suddaby & Ratcliffe 1997).

Another indication of the importance of winter conditions in influencing population changes comes when closely related species that breed in the same area show different trends according to where they winter. Of the various waterfowl species that breed in Siberia, those that migrate to western Europe have all increased in numbers in recent decades, following their greater protection from winter hunting. Examples include the western population of the Greater White-fronted Goose Anser albifrons and the Dark-bellied Brent Goose Branta b. bernicla. By contrast, all those populations that winter in south-east Asia have continued to decline, in association with rising persecution in that region (Syroechkovski & Rogacheva 1994). Examples include the eastern subspecies of the Bean Goose Anser fabalis serrirostris and A. f. middendorffii and the Baikal Teal Anas formosa. These various waterfowl species share similar nesting habitats, which are largely undisturbed by people, and their divergent population trends have been attributed to conditions in their different wintering areas (Syroechkovski & Rogacheva 1994).

Another divergent pattern, evident in temperate regions after a hard winter, is that resident species are often found to have declined greatly, whereas summer visitors from tropical wintering areas have not (Dobinson & Richards 1964, Graber & Graber 1979, Cawthorne & Marchant 1980, Holmes & Sherry 2001). This difference provides another indication of the importance of winter conditions, at least for the resident species (Newton 1998). Some summer visitors may in fact benefit from the scarcity of resident competitors. The unusually large numbers of European Pied Flycatchers Ficedula hypoleuca that bred in Britain in 1917 and 1947 were attributed to the greater availability of nest cavities in those years, occasioned by the scarcity of resident tits caused by preceding hard winters (Elkins 1983). This early observation also suggested that Flycatcher numbers were limited in some areas by shortages of nest-sites, an inference supported by subsequent experiments involving either the removal of competing tit species (Gustafsson 1988), or the provision of additional nestboxes (e.g. von Haartman 1971, Currie & Bamford 1982, Newton 1998). Many other species of cavity-nesting migrants (excluded from Table 1) also increased in breeding density and distribution following the provision of artificial nest-sites, as did various hirundines and swifts that use buildings (Erskine 1979, Newton 1994, 1998, Evans et al. 2003). The implication is that nest-site availability (and hence, a feature of the breeding area) is important in limiting the overall population levels of such species.

Aseasonal weather and other events

Where population declines affect both residents and migrants simultaneously, this has often been attributed to late cold springs that affect the migrants on return passage as well as the residents. Such declines affecting both groups were recorded on the Courish Spit on the southern Baltic coast during a succession of cold springs in 1958–77 and 1985–98 (Sokolov et al. 2000). Storm-induced mortality among migrating birds is outside the scope of this paper, but spring weather events, killing large numbers of migrants soon after their return to breeding areas, have been recorded infrequently in many species, especially insectivores (e.g. Buss 1942, Ligon 1968, Vespäläinen 1968, Whitmore et al. 1977, Brown & Brown 1998). Among Common Sandpipers Actitis hypoleucos, which migrate from Africa to breed in Europe, annual survival fluctuated in one area according to the weather in April when they arrived (Hollands & Yalden 1991). The mean annual survival over 13 years was 79%, but following late snowstorms in 1981 and 1989, survival fell to 39% and 50%, respectively, and breeding pairs from 21 to 14 and from 20 to 12. In general, losses that occur in spring, when numbers are near their seasonal low, are much more likely to affect subsequent breeding density than are losses that occur in late summer or autumn, when numbers are near their seasonal high and subject to density-dependent effects over winter. Recoveries in numbers from serious spring events would normally be expected to take at most a few years.

Carry-over effects

Most discussion in the literature of population limitation in migrants carries the implicit assumption that conditions in wintering areas have no effects on subsequent breeding performance, and that conditions in breeding areas have no effect on subsequent winter survival. However, these assumptions are not always justified. Among geese and other waterfowl, foraging conditions in wintering and migration sites have long been known to affect reproductive success in the subsequent breeding season, through the effect of food supplies on body condition (for Snow Goose see Ankney & MacInnes 1978, Bêty et al. 2003, for Canada Goose Branta canadensis see Hanson 1962, Raveling 1979, for Mallard Anas platyrhynchos see Krapu 1981, Pattenden & Boag 1989), and in a few such species the mechanisms have been studied. Thus in Brent Geese in the Netherlands, the favoured spring staging habitat is saltmarsh whose plants allow the geese to fatten rapidly. However, the number of geese that can feed in saltmarsh is limited, so as the population grew over a period of years, increasing proportions of birds were relegated to less nutritious agricultural grassland. The geese used body reserves accumulated in spring for migration and subsequent nest defence, egg production and incubation, and individuals that had fed on saltmarsh showed better breeding success than those that had fed on grassland (Ebbinge 1992). Females that had accumulated the greatest body reserves at a spring stopover site were more likely to return with young in the following autumn than were females that accumulated smaller reserves, whereas males, which accumulated smaller reserves than females, showed no such relationship (Ebbing & Spaans 1995).

In some other species of geese smaller proportions of females laid, and clutch-sizes were smaller, in years when feeding conditions in staging areas were poor than in years when they were good (for Barnacle Geese Branta leucopsis see Cabot & West 1973, for Lesser Snow Geese see Davies & Cooke 1983). In Whooper Swans Cygnus cygnus, the proportions of young each year in wintering flocks in Sweden were correlated with the mean temperature in the preceding winter, implying that winter temperatures influenced subsequent breeding success (presumably through effects on feeding conditions and body reserves) (Nilsson 1979). From these and other studies, it is clear that the breeding success of some migratory waterfowl depends partly on the body reserves accumulated in wintering and staging areas, and that the effects of such reserves are evident at the level of the individual bird and at the level of the population. By contrast, in some shorebird species, formerly thought to depend for egg production partly on similar reserves, isotope analysis revealed that eggs were formed from terrestrial rather than coastal foods, and hence were produced from food eaten after arrival in breeding areas (Klaassen et al. 2001).

One important factor contributing to reproductive success among migrant birds is date of spring arrival and commencement of nesting. Within populations, individuals that arrive and start nesting early in the season do better in terms of habitat quality, territory acquisition and number of young raised, than those that arrive and nest late (for Willow Ptarmigan Lagopus lagopus see Moss 1972; Common Wheatear Oenanthe oenanthe see Brooke 1979, Currie et al. 2000; Pied Flycatcher see Lundberg et al. 1981; Painted Bunting Passerina ciris see Lanyon & Thompson 1986; Great Reed Warbler Acrocephalus arundinaceus see Bensch & Hasselquist 1991; Barn Swallow Hirundo rustica see Møller 1994; Savi's Warbler Locustella luscinioides see Aebischer et al. 1996). In some such species, the same sequence of territory settlement held from year to year, even though the occupants changed, and even though some early settlers were displaced by former owners that arrived later. Variations in arrival dates sometimes exceeded a month, and late arriving individuals were often in poorer condition.

In a few studies, poor body condition and late arrival have been associated with winter habitat. Marra et al. (1998) used analyses of carbon isotopes in muscle tissue of American Redstarts Setophaga ruticilla to make this connection. Individuals that wintered in moist forest in Jamaica had lower tissue ¹³C values than those found in poorer, secondary scrub habitat, a pattern mirrored in the available insect prey. Also, those individuals occupying the poorer scrub habitats were in poorer body condition than those in better habitats. Muscle tissue ¹³C values provided a tracer to link newly arrived birds in North America with the quality of the habitat they had occupied in winter. Those birds that arrived earliest in the North American breeding areas had lower ¹³C values than birds arriving later, indicating that the earlier birds had come from the best wintering habitat. This work thus provided another link between the conditions experienced by individuals in winter, their subsequent migration dates and breeding success.

Whatever the date of departure from wintering areas, adverse weather encountered en route may delay arrival in breeding areas and thus affect reproduction (Johnson & Herter 1990, Richardson 1990). In 1997, many White Storks were late in leaving their African wintering areas; this was attributed to poor food supply (Berthold et al. 2002). In addition, some individuals (including a radiotagged bird) were delayed for another week en route, as they hit a severe cold spell. These circumstances led to late arrival in European breeding areas, and depressed breeding success over wide areas, providing another link between conditions in wintering and migration areas and subsequent breeding success.

Although I know of no specific examples of events in breeding areas affecting overwinter (post-migration) survival in migrants, at the level of the population, good breeding success is likely to result in large populations. In species limited in wintering areas, mortality is likely to be density-dependent, resulting in high mortality following good breeding years, and lower mortality following poor breeding years. Among migrants, overwinter loss has been shown to be density-dependent in populations of Sedge Warbler, Blackcap Sylvia atricapilla, Common Whitethroat S. communis, Willow Warbler, European Pied Flycatcher, Common Redstart Phoenicurus phoenicurus, Barn Swallow, Redshank, Mallard, Northern Shoveler Anas clypeata and Barnacle Goose (Mihelsons et al. 1985, Järvinen 1987, Kaminski & Gluesing 1987, Stenning et al. 1988, Owen & Black 1991, Baillie & Peach 1992, Whitfield 2003). In most of these species, overwinter loss was also the key factor governing year-to-year change in breeding numbers (reviewed by Newton 1998).

CAUSAL FACTORS

Over the past several decades, various species of intercontinental landbird migrants have declined in Europe and North America, but the main causal factors seem to have differed between the two regions. The European species winter in the Afrotropics, and most of those that have declined occupy the arid scrub habitats prone to drought. Their declines have been widely attributed to events on wintering areas, notably reductions in rainfall and associated food supplies. The North American species winter in the tropics of Central and South America, and most of those that have declined occupy forest, with smaller numbers in scrub and grassland habitats. Their declines have been widely attributed to events in breeding areas, notably forest fragmentation and the associated increases in densities of predators and parasitic Brown-headed Cowbirds Molothrus ater. The lines of evidence proposed in favour of these various hypotheses are summarized below, along with some alternative explanations applicable to at least a minority of species.

Eurasian–Afrotropical migrants

In much of Africa north of the equator, rainfall during the northern summer largely determines the state of the vegetation for the overwintering migrants, and hence may be assumed to influence the food supplies of many birds, whether they consume plant or animal matter. Rainfall varies greatly from year to year, but over much of the region since the late 1960s has in most years been well below former levels, owing to a failure of the rain-belts to extend so far to the north. Hence, drought conditions have been most severe along the northern edge of the Sahel zone, lying immediately south of the Sahara desert, and have diminished southwards across the savannah zones, towards the equator. In the Sahel zone, rainfall deficits were particularly marked in 1968, 1973, 1983 and 1984, and again in 1990. It was in 1969 (after the 1968 drought) that the importance of conditions in African wintering areas was first impressed upon European bird-watchers, when massive declines were apparent in the returning numbers of Common Whitethroats and Common Redstarts, two species that are strongly represented in the northern Sahel (Winstanley et al. 1974). In Britain, where the Whitethroat was among the species monitored by the Common Bird Census of the British Trust for Ornithology, numbers dropped by about 70% between 1968 and 1969 (Fig. 6). The Sand Martin and Sedge Warbler were also badly hit, even though most individuals spend only 4–6 weeks in the Sahel in October–November, before moving south, and an even shorter period on return passage in March (Cowley 1979, Jones 1985).

Figure 6. Above: rainfall trends in the Sahel region of Africa expressed as departures from the long-term mean (- - -) for 1940–88. Data from Grainger (1990). Below: population trend (log scale) of Common Whitethroats Sylvia communis breeding in Britain, as revealed by the Common Bird Census of the British Trust for Ornithology, 1962–88. Data from Marchant et al. (1990).

Other species that winter in the Sahel zone, or pass through in autumn and spring, showed lower population levels in Britain in 1984, 1985 or 1991 than in any previous year since 1962 (when the Common Bird Census started). They included, besides the two species just mentioned, the Barn Swallow, Grasshopper Warbler Locustella naevia, Eurasian Chiffchaff Phylloscopus collybita and Spotted Flycatcher Muscicapa striata. The Chiffchaff subsequently recovered but the Willow Warbler continued to decline, as did the Spotted Flycatcher (Peach et al. 1995, 1998). Moreover, annual fluctuations in the numbers of some species followed annual fluctuations in the preceding year's rainfall in the Sahel zone, as shown for the Sedge Warbler in Britain and the Netherlands (Table 1; Peach et al. 1991, Foppen et al. 1999).

With the lapse of further years, it has become apparent that rainfall in the western Sahel has a major impact on the breeding populations of several species that nest in Europe, including some that overwinter there and others that pass through. Of the 25 species of Palaearctic–African migrants that were adequately monitored in Britain, 11 showed substantial net declines over the period 1962–88 (Marchant et al. 1990). The pattern of decline differed between species, the Common Whitethroat suffering a catastrophic crash mainly in one year, and others showing more gradual declines over ten or more years, with some subsequently recovering and others not. These differences may reflect differences in wintering areas between species, or in the mechanisms involved, and in some species may be confounded by simultaneous changes in breeding areas. Most of the remaining European–Afrotropical migrant species showed no marked net change over the 27-year period, but five showed a substantial rise. The latter winter well south of the Sahel zone, however, and some extend south of the equator to the austral summer.

Further long-term data on population trends of migrants were obtained at the Mettnau trapping station in southern Germany, where birds were caught in mist-nets operated annually between late June and early November (Berthold et al. 1993). In the 20 years from 1972, 12 out of 21 trans-Saharan migrant species showed significant overall declines, and only one (Common Nightingale Luscinia megarhynchos) showed an overall increase. By contrast, only two of 14 other species (residents and short-distance migrants) showed significant declines and none showed a significant increase. Declines were evidently more widespread among the long-distance migrants that winter in Africa than among the residents and short-distance migrants that winter in Europe. However, as the trapping occurred after the breeding season, the totals included adults and young, and so reflected both breeding numbers and breeding success. Comparisons of count data from different countries have shown that year-to-year changes in numbers are often correlated over large parts of the European breeding range but, as expected, the counts from some regions show different trends, and some show no evidence for links with Sahel droughts (Svensson 1985, Marchant 1992, Sokolov et al. 2001). Range-wide trends would perhaps not be expected in species whose west European populations winter in West Africa and east European populations in East or southern Africa, and are thereby exposed to different climatic regimes.

It is not just counts that point to the importance of wintering areas in influencing numbers. In key-factor analyses, variation in overwinter loss was found to account for most of the variation in total annual loss in seven migrant passerine species monitored in Britain or Denmark (Baillie & Peach 1992, Møller 1989). In addition, annual survival rates of Sedge Warblers and Common Whitethroats trapped each summer at specific sites in Britain were also correlated with rainfall in the Sahel zone (Fig. 4; Peach et al. 1991, Baillie & Peach 1992), as were survival rates of Sand Martins trapped at sites in Scotland and Hungary (Bryant & Jones 1995, Szép 1995). Major crashes in numbers occurred in 1968/69, 1983/84 and 1990/91, all following years of poor rainfall. Similarly, annual survival rates of Barn Swallows breeding in an area of Denmark were correlated with March rainfall in their southern African wintering areas (Møller 1989). In British Swallows, breeding numbers were not correlated with rainfall in their South African wintering areas, but with rainfall in the western Sahel, at its driest during spring migration (Robinson et al. 2003). Although factors causing long-term trends in bird numbers are not necessarily the same as those causing annual fluctuations, they appear to be the same in these cases. These studies provided further circumstantial support for the proposed causal chain: low rainfall → low winter food supplies → lower overwinter survival → low breeding population.

It is not only insectivorous passerines that have shown such links. The numbers of Purple Herons nesting in the Netherlands over a 19-year period, and their annual survival rates, were correlated with wetland conditions in their West African wintering areas, as measured by water discharge through the Niger and Senegal Rivers (den Held 1981, Cavé 1983). The wetter the conditions in West Africa, the better was the overwinter survival, and the greater the number of herons that returned to Holland to breed each year. A similar relationship was apparent among Night Herons Nycticorax nycticorax counted at colonies in southern France, and to a lesser extent among Squacco Herons Ardeola ralloides in the same localities (den Held 1981). All these species, and the Little Bittern Ixobrychus minutus, which also winters in drought-prone areas of West Africa, have decreased in western Europe since the 1960s (Marion et al. 2000).

Similarly, the numbers of White Storks that breed in western Europe have declined greatly since 1960, but those in eastern Europe much less so. Over this period annual adult survival was lower in the western population, and annual variations (measured from ringed birds) were closely linked to rainfall in the western Sahel zone where the birds winter (Kanyamibwa et al. 1993, Barbraud et al. 1999). The eastern population, which winters mainly in East Africa, showed no such relationship with rainfall, perhaps because rainfall has declined less there, and because East Africa offers better conditions and greater variety of habitats than West Africa (Dallinga & Schoenmakers 1989). Moreover, the mean numbers of young raised per pair of White Storks in western Europe was correlated with rainfall in West Africa in the preceding winter, presumably because the winter food supplies influenced the condition of adult Storks in spring, and not only their survival (Bairlein & Henneberg 2000). Differences in rainfall patterns may also explain the differences in population dynamics of some other conspecific populations of birds wintering in different parts of Africa, such as western and eastern populations of the White Wagtail Motacilla alba (Svensson 1985). Overall, however, drought in African wintering areas has been suggested as the major causal factor in the regional population declines of at least 17 Eurasian breeding species.

Other possible factors in population declines

Decreasing rainfall is not the only factor that may have caused declines in the numbers of some European birds that winter in West Africa. The burgeoning human population, and associated overgrazing, burning and woodcutting, have all accentuated the process of desertification, so that even in relatively wet years many areas now hold little habitat for birds (Jones 1985). So even if rainfall returns to its earlier levels, some migrant species may not regain their former numbers because of the habitat degradation that has occurred in the interim. With the increasing use of pesticides, locusts and other insects eaten by birds are likely to have become less available, and in various wintering areas many birds have themselves been killed directly by pesticide use (e.g. Mendelssohn & Paz 1977, Mullié & Keith 1993, Goldstein et al. 1996). At least four Palaearctic species, which feed extensively on locusts and grasshoppers on their African wintering grounds, have declined in recent decades, including the White Stork discussed above, the Lesser Kestrel Falco naumanni, the Montagu's Harrier Circus pygargus and the Pallid Harrier C. macrourus. In all these species, however, factors in the breeding range may also have been involved (for Lesser Kestrel see Donázar et al. 1993, for Montagu's Harrier see Arroyo et al. 2002).

Unusually, the felling of rainforest in West Africa may have affected very few migrant species adversely, because the forest is replaced by more open secondary habitats (‘derived savanna’), which many species favour (Morel & Morel 1992). The same is true for miombo woodland in southern Africa, and is simply a consequence of the small number of Eurasian migrant species that winter in closed-canopy woodland compared with those that favour more open areas.

Not all Eurasian–African migrant species are necessarily limited primarily by factors acting on the wintering range, and for any one species the situation may differ from one part of the breeding-wintering range to another. Examples of species whose changes in breeding numbers have been linked with events in breeding areas include the Pied Flycatcher for year-to-year changes (Fig. 4; Virolainen 1984), and the Turtle Dove Streptopelia turtur for long-term decline (Browne & Aebischer 2001). In some species, as mentioned above, the relative importance of breeding and wintering conditions in influencing breeding numbers may well change from one period of years to another or from one region to another.

Nearctic–Neotropical migrants

Notwithstanding the ongoing debate about the number of species involved, and the regions affected, most studies in North America have attributed declines in tropical migrants to events in breeding areas (Askins et al. 1990, Hagen & Johnston 1992, Sherry & Holmes 1992, Martin & Finch 1995, Latta & Baltz 1997). For some eastern forest species, problems are thought to stem from forest fragmentation – the division of the once extensive deciduous forest into small urban and rural woodlots. Such fragmentation could have caused population declines much greater than expected from the areas of habitat lost. The evidence consists partly of the apparent contrast in population trends between birds in small forest fragments and those in extensive forest areas, and partly from current species distributions, which show that many species are now almost absent from small forest fragments. Relevant studies span a wide area from the Atlantic to the Mississippi (Bond 1957, Robbins 1979, 1980, Whitcomb et al. 1981, Ambuel & Temple 1983, Howe 1984, Lynch & Whigham 1984, Hayden et al. 1985, Freemark & Merriam 1986, Blake & Karr 1987, Robbins et al. 1989, Askins et al. 1990).

Many studies have entailed assessment of the numbers and kinds of species breeding in forest fragments of different sizes in the same region. The main finding was as expected, namely that total species numbers increased with increasing size of forest. Whereas certain species were found in both small and large forest fragments, others were found entirely or almost entirely in large ones, mainly in the forest interior. More specifically, larger forests tended to have more species per unit area of neotropical migrants, some species of which tended to be scarce or absent in small forests.

In most studies, forest area accounted for most of the variation in species richness and total density of birds, and especially of neotropical migrants (Askins et al. 1990, Blake & Karr 1987). Other variables, such as the degree of isolation from other forests, or vegetation structure and heterogeneity, explained relatively little additional variance in species numbers, as judged from the findings of multiple regression analyses. However, vegetation variables influenced the kinds of species that were present, and their local densities (Ambuel & Temple 1983, Lynch & Whigham 1984). As expected, edge species occurred at greatest density in small forests.

The relative importance of forest area and isolation to species occurrence seemed to vary between studies (Lynch & Whigham 1984, Robbins et al. 1989, Askins et al. 1990), probably because both features depended on the distribution of woodland in the landscapes concerned, on the particular bird species represented, as well as on the way that isolation was measured: whether by the distance to the nearest wood, by the proportion of woodland in the surrounding area or by some combination of these measures. Similar relationships between species numbers and area have been found in forest and other habitat areas in other parts of the world, but European woods seem to hold few species that could be considered as forest-interior specialists (Newton 1998).

Predation

Forest fragmentation is held to have promoted increases in the numbers of predators and brood-parasites. In eastern North America, as in Europe, many small generalist predators, which eat eggs and chicks, reach much higher densities in suburban and farming areas than in more natural areas (Wilcove et al. 1986, Small & Hunter 1988, Hoover et al. 1995), benefiting from the additional food provided by human activities. Such predators include mammals, notably Grey Squirrels Sciurus carolinensis, Raccoons Procyon lotor and feral cats, and birds such as American Crows Corvus brachyrhynchos and Blue Jays Cyanocitta cristata. All these species tend to concentrate their hunting along woodland edges (Gates & Gysel 1978, Whitcomb et al. 1981, Paton 1994).

Several studies have shown that the nesting success of songbirds was higher in large forests than in small woods, and that in large forests nest success increased from edge to interior. In total, overall predation rates decreased with increasing area of forest patch in all of eight studies, and decreased from centre to edge in ten of 14 studies involving artificial nests, and in four of seven involving natural nests (Paton 1994). In each case, predators that entered the woods from nearby farmed or suburban areas accounted for most of the losses. In one study in southern Wisconsin, for example, the overall nest success of 13 species was only 18% within 100 m of forest edge (n = 96), compared with 58% at 100–200 m within forest (n = 98) and 70% at more than 200 m into forest (n = 82) (Temple & Cary 1988). The entire area of most small woods lay within the most vulnerable zone. It is not hard to imagine therefore that as forest becomes more fragmented, overall nest success could decline, eventually to the point at which insufficient young were produced to offset the annual adult mortality, leading to regional population declines (Brawn & Robinson 1996). Other (but not all) studies using artificial or natural nests have given similar results elsewhere in North America (Gates & Gysel 1978, Wilcove 1985, Small & Hunter 1988, Yahner & Scott 1988, Donovan et al. 1995, Hoover et al. 1995, Robinson et al. 1995, Weinburg & Roth 1998, Manolis et al. 2002), as have studies with artificial nests in Europe (Andrén & Angelstam 1988, Sandström 1991, Andrén 1992).

If nest predation has increased with forest fragmentation, why should it have affected some neotropical migrants more than other birds? Three reasons have been commonly suggested. First, with few exceptions, the neotropical migrants that breed in the eastern United States construct open nests, either on the ground or in low bushes, in sites that are most vulnerable to predators (Wilcove & Whitcomb 1983). By contrast, none of the residents or short-distance migrants that breed commonly in suburban settings places its nest on the ground, and many use tree-cavities, which provide some of the safest sites. Secondly, long-distance migrants start breeding later in the year (mostly June) and usually raise only one brood per year, occasionally two (Greenberg 1980, Whitcomb et al. 1981). By contrast, many resident species and short-distance migrants start in April or May and routinely attempt second and even third broods. Thirdly, many neotropical migrants are small and unable to drive off nest predators in the way that some larger resident species can. For these various reasons, then, neotropical migrants are supposedly disadvantaged with respect to nest predators, which could together cause a disproportionate reduction in their breeding rates, and contribute to the selective disappearance of tropical migrants from fragmented eastern woodlands. Through nest predation alone, Wood Thrushes Catharus mustelinus in Pennsylvania produced insufficient young to offset the usual annual mortality (Hoover et al. 1995), and through a combination of predation and cowbird parasitism the same was true for Wood Thrushes in Illinois (Robinson 1992).

In an analysis of the population trends of different species in relation to various features of their biology, Böhning-Gaese et al. (1993) found that nest location and nest-type were the best predictors of population trends. Species with high nest locations did better than those with low, and species with closed nests did better than those with open nests. These are both aspects of nesting biology that influence rates of predation and probably also of parasitism, as discussed below.

Parasitism

The Brown-headed Cowbird lays its eggs in the nests of a wide range of passerine host species, thereby reducing their own production of young (Mayfield 1977). Originally confined to the grasslands of mid continent, following deforestation the species gradually spread to occupy most of the continent north to the boreal forest (Mayfield 1977, Brittingham & Temple 1983). Within this newly occupied terrain, its numbers continued to rise at least to the 1960s, as it benefited from the increased food supplies provided by waste grain in cereal fields, cattle feedlots and garden feeders. Its overall numbers probably increased more than ten-fold during the 20th century (Brittingham & Temple 1983).

Forest fragmentation increased the amount of edge, along which Cowbirds have access to the nests of forest species previously immune to attack. Several studies have shown greater parasitism of nests in small than large woods, and near the edges of woods than in the interior (Gates & Gysel 1978, Brittingham & Temple 1983, Temple & Cary 1988, Donovan et al. 1995, Robinson et al. 1995). Many of these new hosts, including neotropical migrants, have no innate defences, and will not desert or eject strange eggs, but raise the resulting young. Because Cowbirds use many host species, moreover, they are not vulnerable to decline in any one, and could in theory parasitize favoured species to extinction, while being maintained at high density by other common hosts (May & Robinson 1985). The same features that make neotropical migrants nesting near the edges of woods more vulnerable to nest predators than residents also make them more vulnerable to parasitism, namely their use of accessible nest-sites and their predominant single-broodedness.

During the 1960s, Cowbirds parasitized up to 70% of nests of the rare Kirtland's Warbler Dendroica kirtlandii, reducing the production of young Warblers to fewer than one per nest (Mayfield 1983, Walkinshaw 1983). This was fewer than needed to maintain the population level, and during 1961–71 the Warbler population declined by 60%. Then, with the start in 1972 of a programme of Cowbird removal, parasitism dropped to 6%, Warblers fledged about three young per nest and their breeding numbers stabilized (Kelly & DeCapita 1982, DeCapita 2000). Later, they increased more than threefold when the habitat expanded as a result of a forest fire (Rothstein & Robinson 1994), but to facilitate this increase, and to maintain Kirtland's Warbler in the long term, Cowbird control was continued. The historical suppression of wild fires had clearly not helped this species, which prefers young stands.

Cowbird removal also resulted in large increases in both breeding success and subsequent breeding numbers of the Least Bell's Vireo Vireo bellii pusillus in southern California and the Black-capped Vireo Vireo atricapillus in central Texas (Griffith & Griffith 2000, Hayden et al. 2000). On the other hand, removal of Cowbirds from Mandarte Island off British Columbia led to an improvement in the nest success of Song Sparrows Melospiza melodia, but to no increase in subsequent breeding density, which was determined by other factors (Smith 1981). The same held for South-western Willow Flycatchers Empidonax traillii in California (Whitfield 2000). Thus, of five experimental removals of Cowbirds, all five led to improved nest success in the host species, but only three led to an unequivocal response in host breeding numbers. Nonetheless, some other species seem no longer to be producing enough young to offset known adult mortality. In a mainly agricultural landscape in central Illinois during 1985–89, nests of forest-nesting neotropical migrants contained an average of 3.3 Cowbird eggs per nest. End-of-season juvenile to adult ratios averaged 0.1 in neotropical migrants compared with > 1.0 for year-round residents and short-distance migrants (Robinson 1992). The neotropical migrants could not have maintained their numbers under this level of parasitism without large-scale immigration.

Other populations, which may have declined following greater contact with this generalized brood-parasite, include the Golden-cheeked Warbler Dendroica chrysoparia in parts of Texas, the Black-capped Vireo in parts of Oklahoma (Grzybowski et al. 1986), and the Eastern Warbling Vireo Vireo gilvus in parts of California and British Columbia (Verner & Ritter 1983, Ward & Smith 2000), White-crowned Sparrow Zonotrichia leucophrys in parts of California (Trail & Baptista 1993), and Bell's Vireo Vireo bellii in Missouri (Budnik et al. 2000). At present, Cowbird removal forms part of the management action for several endangered North American migratory birds, including the Kirtland's Warbler, Black-capped Vireo, Least Bell's Vireo and South-western Willow Flycatcher.

It thus seems, then, that many songbirds in eastern North America which nest in forest fragments in suburban and agricultural areas are now experiencing the dual effects of increased nest predation and parasitism, particularly near forest edges, and that for various reasons these effects fall most strongly on many neotropical migrants. So far, attempts to assess these impacts on population levels have been made only in a minority of species, such as the Kirtland's Warbler and Wood Thrush in which survival, as well as breeding success, have been measured. Studies on a greater range of species, and in a greater range of areas, are needed before the generality of these findings can be assessed. However, many other species are suffering such high rates of nest failure in some areas that population decline seems inevitable in these areas without continuing immigration. Although the predisposing fragmentation of the eastern deciduous forest occurred mainly in the 19th century, the associated increase in generalist predators and cowbirds has occurred mainly in the latter half of the 20th century. Decline through excess predation and parasitism would be expected to leave increasing amounts of breeding habitat unoccupied. In this scenario, it is not shortage of nesting habitat as such that limits numbers, but shortage of ‘source’ habitat in which annual per capita reproduction matches or exceeds annual per capita mortality. Habitat where the reverse conditions hold acts as a sink.

Forest fragmentation may also reduce the food available per unit area. This aspect has received much less attention than predation or parasitism, but could be important in some species. For example, among Ovenbirds Seiurus aurocapillus nesting in different sized forest fragments in Ontario, the density and pairing success of territorial males increased with area of the woodlot (Burke & Nol 1998). Within Ovenbird territories, the biomass of invertebrate prey was 10–26 times greater in large woods than in small ones, associated with the deeper leaf litter in larger woods.

The relationship between forest area and species numbers was found to vary regionally: deficiencies in species numbers in small forests tended to be more marked in agricultural than in more forested landscapes (Freemark & Collins 1992). These regional differences were in turn associated with similar regional variations in songbird nest success. In nine migratory species, predation and parasitism rates, measured in standardized ways, were found to increase with reduction in the proportion of forest present. Landscapes with little forest acted as sinks for some species, in that reproduction was insufficient to offset annual mortality, so that breeding numbers could not be maintained in the absence of continual net immigration from more forested areas (notably Wood Thrush, Hooded Warbler Wilsonia citrina, Ovenbird and Kentucky Warbler Oporornis formosus). There were some exceptions to the trends, as local factors – including those influencing Cowbird abundance – modified the overall patterns (compare the studies on Wood Thrushes in Illinois, Indiana, Pennsylvania and Delaware: Robinson 1992, Roth & Johnson 1993, Hoover et al. 1995, Weinberg & Roth 1998, Fauth 2000).

Other factors operating in breeding areas

Other evidence points to events in breeding areas (as opposed to wintering areas) as being a major cause of population fluctuations in some neotropical migrants. For many years, some migrant warblers have undergone long-term changes in breeding densities in association with changes in caterpillar abundance. Most outbreaks of defoliating caterpillars in eastern North American hardwood forests are sporadic, occurring in one location in one year and somewhere else the next. The probability that any particular stand will experience an outbreak in any given year is low. Every now and then, however, one caterpillar species of boreal regions, the Spruce Budworm Choristoneura fumiferana, increases over a period of years to reach plague proportions over wide areas, causing extensive defoliation. Several migrant bird species show strong numerical responses to this insect, including the Bay-breasted Warbler Dendroica castanea, Blackpoll Warbler D. striata, Tennessee Warbler Vermivora peregrina, Cape May Warbler D. tigrina, Black-billed Cuckoo Coccyzus erythropthalmus and Yellow-billed Cuckoo C. americanus (Kendeigh 1947, Morris et al. 1958, Crawford & Jennings 1989). In one study, Bay-breasted Warblers increased in abundance from 2.5 pairs per 10 ha in uninfested stands to 300 pairs per 10 ha during an outbreak, Blackburnian Warblers D. fusca increased from 25–30 pairs to 100–125 pairs per 10 ha, and Tennessee Warblers from zero to 125 pairs per 10 ha (Morris et al. 1958). Because in each outbreak the numbers of the birds rose over several years in parallel with the caterpillars, it was hard to tell how much the increase was due to immigration and how much to the high local breeding success, though both were involved.

The role of previous breeding success in the annual population changes of some neotropical migrants was confirmed in studies in the Hubbard Brook Experimental Forest in New Hampshire (Holmes et al. 1991, 1996, Rodenhouse & Holmes 1992, Sherry & Holmes 1992). Over a period of years, and in different forest plots, warbler breeding densities fluctuated in parallel with caterpillar biomass the previous summer. There was thus a correlation between the production of young in one year and the numbers of breeders returning the next. In other words, breeding success rather than overwinter survival had most influence on the annual changes in breeding numbers, and increases in breeding density were marked by greater recruitment of yearlings (Sherry & Holmes 1988, 1992). These findings were apparent in the American Redstart setophaga ruticilla, Black-throated Blue Warbler Dendroica caerulescens, Black-throated Green Warbler D. virens, Red-eyed Vireo Vireo olivaceus and others, so in these species the year-to-year changes in breeding success and in subsequent breeding numbers were attributable to fluctuations in summer food supplies. Elsewhere, a relationship between the annual recruitment of yearlings to a breeding population and fledgling production the previous year was noted in another neotropical migrant, the Prairie Warbler Dendroica discolor (Nolan 1978), paralleling the Pied Flycatcher and others in Europe (Fig. 4; Virolainen 1984, Stenning et al. 1988).

The abundance of foliage-gleaning birds in the Hubbard Brook Forest, especially of warblers and vireos, increased in the early 1970s, coincident with a major outbreak of defoliating caterpillars (Holmes 1990). Furthermore, natural and experimentally induced declines in food (mainly caterpillar) abundance were shown to reduce the frequency of re-nesting and second-brood attempts, as well as growth-rates, hatching and fledging success, and to increase the frequency of nestling starvation. The general pattern seemed to be that food was abundant for birds in this forest only during lepidopteran irruptions, but such occasions occurred infrequently, perhaps once every 10–20 years in any one forest stand. Between these outbreaks birds probably experienced prolonged periods of food limitation, as caterpillars were scarce for several years in succession. The neotropical migrants thus showed population trends in this forest that differed from those of residents and short-distance migrants, whose breeding numbers fluctuated from year to year mainly in line with winter severity.

The importance of these findings in our present context is twofold. First, they show that events in breeding areas that influence breeding success can influence subsequent breeding densities, apparently over-riding any effects of events in wintering areas. Secondly, they show that food supply can affect breeding success and breeding numbers in addition to any effects of predation and parasitism. The question then arises as to what extent recent declines in the numbers of certain migrant species in North America were due to recent declines in caterpillar densities. Budworm caterpillars were abundant over wide areas in the late 1940s and again in the 1970s. One would therefore expect a decline in the numbers of some species in the 1950s−60s and again in the 1980s−90s. In addition, since the 1950s, insecticides have been used widely in eastern Canada to suppress caterpillar outbreaks. This practice is likely not only to have obliterated the ‘good food years’, but also to have suppressed all insect populations in the affected areas in the years of spraying. The influence of food supply on migrant population trends could thus be substantial and may have contributed to the declines in some species witnessed in the 1980s−90s.

Another factor thought to operate on breeding areas to reduce migrant numbers is summer drought, again through its influence on insects and other food supplies. Severe declines in the numbers of several neotropical migrant species in Wisconsin and Michigan during 1986–88 coincided with drought (Blake et al. 1992), as did similar declines in Illinois (Robinson 1992). Because the migrants do not start nesting until June, they are more vulnerable to the effects of drought than are residents and short-distance migrants that start in April or May. Such effects would be expected to be short-lived, however, and followed by population recoveries as the weather returned to normal. In addition, long-term changes in the species composition of particular forest areas would be expected from the successional changes that occur as the forest matures, making the habitat less suitable for some species and more suitable for others (for examples see Litwin & Smith 1992, Holmes & Sherry 2001).

Further evidence for the importance of events in breeding areas for declines in Nearctic–Neotropical migrants has come from studies in wintering areas. For example, between 1972 and 1990, birds were systematically netted in the Guanica forest in Puerto Rico (Faaborg & Arendt 1992). Over the period concerned, the numbers of wintering migrants steadily declined. Some species, such as Northern Parula Warbler Parula americana and Prairie Warbler Dendroica discolor, were common at the start, but virtually absent in later years, whereas other migrants, such as Black-and-white Warbler Mniotilta varia, and various resident species maintained their numbers. As no obvious change occurred in the forest itself during the study, the declines were attributed to events in breeding areas.

Most of the evidence thus points to changes in the eastern North American breeding areas as being the main cause of recent declines in the numbers of some forest-dwelling neotropical migrants. Such changes may not have occurred over the entire range or in every wood, but they have clearly occurred in a sufficiently large proportion of small woods to put the regional populations of some species into decline. Only time will tell how long the declines will continue.

Factors operating in wintering areas

Over the same period, much tropical forest has been destroyed. The highest rates of deforestation have occurred in Central America and the Caribbean, the very regions where forest migrants are most concentrated (Myers 1980, Terborgh 1989). On present evidence, it is impossible to say how much tropical deforestation has contributed to the declines of forest migrants, but if it continues at the recent rate, it may soon overtake events in the breeding areas as the major cause of declines. However, not all species are likely to have been affected adversely because, as mentioned above, some thrive in the secondary habitats that replace the forest.

One species already affected by tropical deforestation, according to Terborgh (1989), is Bachman's Warbler Vermivora bachmanii, now probably extinct. At the turn of the century, this species bred widely in damp broadleaved woods across the southern United States, but by the 1950s it could be seen reliably only at a few places in coastal South Carolina. All wintering records came from Cuba, from the dense evergreen thicket that once covered large parts of the island, and which has since been largely replaced by sugar cane. Almost certainly, this warbler has succumbed primarily through destruction of its wintering habitat, for substantial areas of breeding habitat still remain.

All migrants are affected by conditions in both breeding and wintering areas, and in recent years species have presumably been adjusting to changes in one or other area. However, the habitats occupied by some species, in both breeding and wintering areas, are so vulnerable to human activity that such species seem destined to decline markedly in the coming years. The Cerulean Warbler Dendroica cerulea has suffered extensive loss of breeding habitat in the past 200 years (Robbins et al. 1992). Its favoured nesting habitat, mature floodplain forest with tall trees, has become scarce over most of its original nesting range, and its apparent sensitivity to fragmentation of remaining tracts gives it an additional disadvantage. In winter, the species is restricted to primary humid evergreen forest along a narrow elevation zone at the base of the Andes. This zone is among the most intensively logged and cultivated regions of the Neotropics. From 1966 to 1989, the Cerulean Warbler showed the most precipitous decline of any North American Warbler (3.4% per year nationwide).

Non-forest species

Migrants dependent on natural grassland, such as the Sprague's Pipit Anthus spragueii and Upland Sandpiper Bartramia longicauda, declined greatly after the ploughing of the prairies, but species that can live in agricultural habitats, such as the Indigo Bunting Passerina cyanea, greatly expanded after the felling of the eastern deciduous forest. Of more recent changes, some have been attributed to changes in breeding areas and others to changes in wintering areas. Thus the Bobolink Dolichonyx oryzivorus has been declining in the eastern United States since the early 1900s. It nests mainly in hayfields, and decline has been attributed to procedural changes that make hayfields less suitable as habitat (Bollinger & Gavin 1992). By contrast, an earlier decline of the Dickcissel Spiza americana, which inhabits brushy pastures, was put down to destruction of wintering habitat, resulting from overgrazing by cattle (Fretwell 1980).

Most arctic nesting shorebirds that winter on neotropical coastlines have so far been spared the effects of large-scale habitat destruction. Their breeding habitats are still largely intact, whereas their wintering habitats may have expanded as a result of the soil erosion following deforestation, and the resulting expansion of coastal mud. During part of the 20th century, some shorebird species increased in numbers in response to lessened shooting pressure, which had reduced their numbers in the past. Some raptors, notably Swainson's Hawks, suffered from pesticide-induced mortality in wintering areas in Argentina. Around 5000 individuals were picked up under their communal roosts, after the spraying of their food organisms (grasshoppers) with the organophosphate compound, monocrotophos, now banned in this region (Goldstein et al. 1996). Earlier declines in Peregrine Falcon Falco peregrinus numbers were attributed to the use of organochlorine pesticides in both breeding and wintering areas, and recent recoveries in Peregrine numbers have followed reductions in organochlorine use (Cade et al. 1988). It is more difficult to explain recent declines in wetland and coastal species, such as Least Tern Sterna antillarum, which in eastern North America declined by 14% per year during 1978–88 (Sauer & Droege 1992).

Comparison between Old and New World systems

Although in the last 30–40 years some tropical migrant birds have declined in western Europe and others in eastern North America, the causes seem to have differed. In Europe, declines have mainly involved species that winter in the arid savannas of tropical Africa, which have suffered from the effects of drought and increasing desertification. In several species, annual fluctuations in numbers and adult survival rates were correlated with annual fluctuations in winter rainfall, and by implication in winter food supplies. There were no obvious changes in breeding success that could be linked with population changes.

In North America, by contrast, declines have affected many species that breed and winter in forest. In eastern forest regions, declines have been attributed to human activities on the breeding range, particularly forest fragmentation and associated agricultural and suburban developments, which have led not only to loss of forest, but to increases in the densities of nest predators and parasitic cowbirds. Declines in the numbers of some migrants are thought to result from reduced breeding success, which is now too low to offset the usual adult mortality, though as yet convincing evidence is available for only a few species, such as Kirtland's Warbler and Wood Thrush in some areas. In other species, such as Bachman's Warbler, tropical deforestation seems to have played the major role in population decline, and is likely to affect an increasing range of species in the future. Whereas for the Palaearctic–Afrotropical system, a considerable consensus exists on the causes of declines, for the North American system, much of the evidence is as yet little more than suggestive, and no one explanation can account for all the facts (for discussion see James & McCulloch 1995).

Experimental evidence on population limitation in migrants

Most of the evidence discussed above on the causes of population changes, whether for Old World or for New World species, is based on correlation and inference. Experiments are almost impossible to perform on regional scales, but widespread changes in land-use or legislation have sometimes been followed by widespread changes in bird populations. Examples include the increases in many waterfowl and shorebird populations that followed the introduction of protective legislation or other conservation measures, or the increases of bird-of-prey populations that followed reductions in the use of organochlorine pesticides (Newton 1979, 1998, Cade et al. 1988). In addition, local population increases have followed experimental manipulations of potential limiting factors or the introduction of specific conservation measures. Examples include increases in breeding success and breeding density that were associated with: (1) removal of predators (for various ducks see Duebbert & Kantrud 1974, Duebbert & Lokemoen 1980, for Sandhill Crane Grus canadensis see Littlefield 1995); (2) removal of Brown-headed Cowbirds (for Bell's Vireo see Griffith & Griffith 2000, for Black-capped Vireo see Hayden et al. 2000); (3) provision of extra nest-sites for species with special needs (for many species of cavity-nesting and other birds see Newton 1994, 1998); (4) removal or exclusion of competing species (for effects of tit Parus removal on Pied Flycatcher densities (via access to nest-sites) see Gustafsson 1988); and (5) change or intervention in destructive agricultural procedures (for Corncrake Crex crex see Green et al. 1997, Green 1999). Similar measures have led to population increases in some resident bird species too, as has the provision of additional winter food (Newton 1998). Restrictions in hunting pressure and pesticide use have affected wintering populations, as well as local breeding populations, but the remaining types of experiments were restricted to breeding populations. They confirm that many migrant species were at that time and location limited by conditions in breeding areas.

IMPORTANCE OF STOPOVER SITES

Although most studied declines in the numbers of migratory bird species have been attributed to events on the breeding or wintering areas (or both), some might have been caused by events at localities that lie on the migration routes, in crucial staging areas that individual birds may visit only for days or weeks each year, but nevertheless provide essential refuelling points. In order to accumulate the body fat necessary for migration, birds need to obtain more food per day than is usual. Moreover, because the same stopping sites can be used by large numbers of birds at a time, competition is often intense, and as wave after wave of migrants passes through, food supplies can be severely depleted. The potential for limitation on staging areas is perhaps especially acute in shorebirds and waterfowl, which in many regions have only a limited number of possible refuelling sites. The quality of any stopover site depends, of course, not just on the available food supplies, and levels of competition, but also on the security that the site offers against predation, disturbance and other threats. For example, in springs when Greater Snow Geese Chen caerulescens atlantica were exposed to hunting on their main staging area in eastern Canada, they migrated in poorer condition, and laid later and smaller clutches than in years when hunting at this crucial site was banned (Bêty et al. 2003).

Marked declines in food supplies, mainly through depletion, have been measured at stopover sites during the migration season. This was done by excluding birds from some places, and comparing the trends in prey populations inside and outside the exclosures. For example, a 60% decline in the total standing crop of macro-invertebrates and 7–90% declines in different prey species were recorded during the shorebird passage (July–September) in Massachusetts (Schneider & Harrington 1981). Similarly, some passerine migrants were found to depress food supplies rapidly by up to 67% at coastal stopover sites in spring following their migration over the Gulf of Mexico (Moore & Yong 1991).

The rate of fattening in particular species has varied between sites or years according to local food supplies, and was assumed to influence the overall speed of migration. At particular sites, individual birds remained longer and accumulated fat more slowly at times when food was scarce, or at times when the density of competing birds was high (Bibby & Green 1981, Cherry 1982, Moore & Yong 1991, Kelly et al. 2002, Whalen & Watts 2002). For example, at two different stopover sites in France, Sedge Warblers gained weight at 0.42 g (se = ±0.07) per day and at 0.05 g (se = ±0.10) per day (Bibby & Green 1981). Similarly, the rate at which migrants gained weight varied from year to year at the same site: Sedge Warblers gained 0.40 g, 0.05 g and 0.55 g per day in 1973, 1974 and 1975, respectively (Bibby et al. 1976). In 1974, when the rate of weight gain was low, 84% of Sedge Warblers stayed more than 2 days. But in 1973 and 1975, when the rate of gain was greatest, only 46% stayed two or more days. Similarly, Dunlin Calidris alpina stayed on a Moroccan estuary for an average of 16 days when food was scarce, but only 11 days the following spring when food was abundant (Piersma 1987). Slow rates of fattening, which delay migration at early staging sites, could affect the rates of fattening at later sites if birds arrive after food supplies have been depleted (the domino effect of Piersma 1987), and migrants that reach breeding areas late or in poor condition may fail to breed that year (Newton 1977, Johnson & Herter 1990, Berthold et al. 2002).

Density-dependent patterns of habitat settlement during stopovers (e.g. Veiga 1986) also implied competition, as did observations revealing that dominants gained weight faster than subordinates (Lindström et al. 1990). Some species behaved territorially at stopover sites, and individuals were unable to gain weight until they had acquired a territory, with its associated food supplies (Rappole & Warner 1976, Mehlum 1983). Competition may thus result in a longer stopover, or cause a migrant to leave with low fat reserves, reducing the distance it could travel before the next stop. In many species, adults trapped at stopover sites were heavier, with greater body reserves, and stayed for shorter periods than first-year birds; and within age-groups, males were heavier, with greater body reserves, and stayed for shorter periods than females (Veiga 1986, Ellegren 1990, Morris et al. 1996, Woodrey & Moore 1997, Yong et al. 1998). Such differences could influence the travel speeds, departure and arrival dates, and survival chances of individuals during migration.

A link between the body condition achieved at migration sites and the subsequent performance of individual birds has emerged in several studies. The body weight achieved at stopover sites has been found in some species to correlate with subsequent survival or, more strictly, return rates (for Great Reed Warbler see Nisbet & Medway 1972; for Semipalmated Sandpipier Calidris pusilla see Pfister et al. 1998; for Red Knot Calidris canutus see Baker et al. 2004). The assumption is that return rates do indeed reflect survival, and not simply the movement of poor-performing birds to other staging or wintering sites, but this remains uncertain. No such doubt hangs over the relationship between spring body condition and subsequent reproductive success, as shown in Brent Geese and others (see above). In general, geese in good condition in spring were more likely to return in autumn with young than were geese in poor condition in spring (Ebbinge & Spaans 1995). These various findings indicate that poor conditions on wintering and stopover sites can reduce the survival and reproduction of individuals; but whether these effects influence subsequent population size depends on the extent to which they are offset by improved survival or breeding in the remaining birds.

Where a food supply is depleted by passage migrants, latecomers arriving after most of the food has been eaten could be penalized by reduced rates of fat accumulation, and eventually by the inability to form reserves. This situation is exemplified by Bewick's Swans Cygnus columbianus bewickii studied at a staging site in the White Sea, the last stop on spring migration before arrival on the Siberian breeding grounds (Nolet & Drent 1998). In this locality, the Swans could feed on their main food, tubers of Fennel Pondweed Potamogeton pectinatus, only during low tide. In the course of the staging period, the Swans tended to forage at progressively lower water levels, indicating that they gradually depleted this food supply, and exploited increasingly deeper parts of the tuber bank as the days went by. This depletion reduced the Swans’ main foraging period from 6.0 h per tide on 20 May to 3.3 h per tide on 28 May. The authors calculated that this must have greatly reduced the rate of refuelling during the staging period. In accordance with this, Swans arriving early stayed for shorter periods than those arriving late. It seemed of paramount importance for the Swans to arrive at the stopover site as soon after ice break-up as possible, because a month later the tubers were greatly depleted and those remaining began to sprout. The first Swans to arrive could also leave the site first, and (in theory) reach the breeding grounds earliest, get the best territories and achieve the highest reproductive success. The latest Swans to leave would have arrived on the breeding grounds too late to breed that year. In such a situation, the more Swans that fed there, the greater the proportion likely to have been excluded from breeding.

Hence, this swan study provided an example of how competition for limited food supplies at a stopover site, used for no more than a few weeks each year, could have helped to regulate the population. The White Sea provides the only sizeable stopover site for swans on this part of the spring migration route, so with such severe competition at this crucial site, the birds would be limited in how much they could respond to any improvements in conditions that might occur in their breeding or wintering areas.

The above seven paragraphs illustrate the main types of evidence proposed to support the view that events at stopover sites could act to limit migrant populations. Food supplies at staging sites can be heavily depleted, slowing rates of fattening, which in turn delays migration, in some individuals in spring to the extent that it reduces breeding success, or prevents breeding altogether. So stopover events clearly affect individuals, but whether these effects are sufficient to reduce population sizes below what they would otherwise achieve remains, for most species, an open question. In only one study known to me has a marked decline in a bird population been firmly tied to a change in conditions at a major stopover site. This is the recent catastrophic decline over three years (from 51 000 to 27 000 individuals between the years 2000 and 2002) of the Red Knot C. c. rufa population that breeds in arctic Canada and winters in Tierra del Fuego. Decline coincided with collapse (through human overfishing) of the Horseshoe Crab Limulus polyphemus population, the eggs of which form the food of Knots at their stopover site in Delaware Bay (Baker et al. 2004). This locality is the last refuelling site of these birds en route to their arctic breeding areas. From 1997 to 2002, increasing proportions of Knots studied in the Bay failed to reach the threshold departure mass of 180–200 g. Survival of adults fell by 37%, and the proportions of second-year birds in wintering flocks by 47%. Of birds caught in the Bay, known survivors were heavier at initial capture than were birds not seen again. In view of this example, the situation for spring migrants passing through the Sahel zone of Africa in drought years might well repay more study.

CONCLUDING REMARKS

In this brief review, I have been concerned mainly with changes in the numbers of migratory birds, and with the relative importance of events in the breeding and wintering areas in influencing the changes. In the longer term, however, population limitation in migrants is probably a dynamic process, involving both areas. If conditions in the wintering range permit increased survival, a species could become more numerous, and may expand its breeding distribution, perhaps into places where the production of young is lower. Similarly, if conditions in the breeding range permit increased production of young, the species could become numerous enough to expand its wintering range, perhaps into areas where survival is lower. In this way, the summer and winter ranges of migrants will tend to expand until reproduction and mortality balance (see also Cox 1985). It is only during a period of change, as in recent decades, that the breeding or wintering range may emerge as providing the stronger limitation. These speculations assume that there are indeed vacant areas into which migrants could expand if their numbers rose. For most species this is probably true; either in other habitats within the existing range or in other suitable areas outside it.

Various populations of geese, which have increased greatly in the past 50 years, provide some instructive examples. As their numbers have grown, they have expanded both breeding and wintering ranges, and in some populations, constraints to further population growth have shifted increasingly from wintering to breeding areas. This is shown in various ways, notably by the gradually decreasing proportions of young in some wintering populations (Fig. 5).

Migratory birds depend on encountering suitable conditions at all staging points on their routes. If conditions deteriorate at any one point, a bottleneck might develop that could begin to limit the population. When conditions are deteriorating everywhere at once, it becomes hard to pinpoint that bottleneck except in the most obvious cases. But the fact that migrants use two or more essential areas each year means that they are inevitably more susceptible to the effects of habitat destruction than are resident birds. Residents suffer only if their particular area is destroyed, but migrants could suffer if any one of several areas important to them is lost. In this sense they have, on average, more chances than residents of being affected – adversely or otherwise – by human action. They live in multiple jeopardy.

Another much neglected aspect concerns the interactions that occur between different populations or different sectors of the same population. Thus, if individuals from two or more breeding populations occur together in the same staging or wintering areas (like many shorebirds) where their numbers are limited, and feed on the same prey, the dynamics of the separate breeding populations can be interlinked, as the survival rate of individuals from one population is likely to depend on the overall size of both populations (Dolman & Sutherland 1995). Further complexities arise where individuals from a single population winter in different habitats or in different geographical regions, as in partial migration or in differential migration where birds of different sexes or ages migrate to different regions. In such cases, adversity in one region could have complex effects on populations, leading to population decline through poor recruitment, unequal sex ratios or other types of demographic disruption, as yet largely unexplored.

With so many travel lanes in bird migration, populations from different breeding or wintering areas may often use the same stopover sites. This gives great potential for competition: if one population passes first, it may deplete the food stocks for later ones, and if different populations are present at the same time, the individuals in both may suffer from depletion and interference. In other words, although living apart for most of their lives, the annual few weeks of contact on stopover sites could ultimately influence the size of one or both of any two competing populations. This is an aspect of stopover biology that has so far received little attention, but it could have enormous repercussions for competing populations.

Acknowledgments

I am grateful to Andy Gosler, in his capacity as editor, for encouraging me to write this article, and to Will Cresswell and Bill Sutherland, in their capacity as referees, for helpful comments on the manuscript.