Changes in the spatial patterns of avian migrations: Evidence, mechanisms and causes

Understanding the suite of environmental changes impacting migration and disentangling these from the dependencies between migratory stages is fundamental to understanding the drivers and mechanisms causing spatial shifts to migratory bird distributions. In this review, our objectives are to: (i) provide evidence of spatial change at all migratory stages, (ii) identify the key mechanisms driving change, (iii) discuss how anthropogenic change influences migratory patterns, (iv) highlight interdependencies between migratory stages and (v) offer a framework for understanding migratory pattern changes.


| INTRODUC TI ON: MI G R ATI ON IN A CHANG ING WORLD
Avian migrations are some of the most remarkable journeys on the planet and are highly intertwined with the life histories of the species that undertake them (Soriano-Redondo et al., 2020).While our understanding of basic ecology is limited in many migratory taxa, among better studied systems and regions there is clear evidence for systemic shifts in key aspects of bird migration (La Sorte & Thompson, 2007;Socolar et al., 2017).Migratory timings of numerous species are changing in response to global warming, leading to phenological shifts in life history events including breeding ground arrival (Horton et al., 2019) and egg laying (Dickey et al., 2008;Thorup et al., 2007).The breeding and non-breeding distributions of many migratory birds are also changing and this has been demonstrated on continental scales (Lehikoinen et al., 2021).However, the rate of change in migratory traits does not always match rates of environmental change (Devictor et al., 2008), and such mismatches are implicated in population declines (Jones & Cresswell, 2010).
Quantifying patterns of change in migratory traits (including phenology and space use) is vital to understand how anthropogenic change will impact migratory species in the future.
Here, we focus on spatial changes in avian migrations as they have received relatively less attention compared to changes in phenology, but are still incredibly important when understanding species responses to environmental change (Chen et al., 2011).We define spatial changes as shifts in the population-level distributional centroid of a migratory stage where the centroid is the average location of all individuals (e.g.Curley et al., 2020;La Sorte & Thompson, 2007).In this case, a migratory stage can be a destination (breeding or non-breeding site), a stopover (stationary periods interrupting active migration) or a flight route (active flight between any of the other two types of stages).While no wild study can know where all individuals are at all times, there are a variety of data that can be used to estimate spatial change under our definition and the mechanisms driving change.These include standardised survey data (Curley et al., 2020;Lehikoinen et al., 2021), citizen science records (Fink et al., 2020;Wilson et al., 2013), colour-mark resightings (Gill et al., 2019;Oudman et al., 2020) and tracking data (e.g.GPS or geolocator; Dufour et al., 2021;Teitelbaum et al., 2016).These methods importantly differ in the number of different individuals accounted for and whether they record data at the population level or individual level.Standardised survey data and citizen science records take into account a large number of individuals at the population level, so are suited to estimating spatial change at the population level under our definition.In comparison, colour-mark resightings and tracking data typically take into account fewer individuals an follow them longitudinally at the individual level, so they are more useful for understanding the mechanisms driving change.
Spatial changes in avian migrations are highly variable and multi-faceted (Curley et al., 2020).The use of multiple, often widely separated, sites within an annual cycle makes understanding patterns and drivers of change complex.Climatic and landscape-level changes vary across spatial scales and in turn drive variation in the magnitude and direction of spatial change across species, populations and different migratory stages (Lehikoinen et al., 2021).Changes at one migratory stage are interdependent on others throughout the migratory cycle (Potvin et al., 2016;Visser et al., 2009).For instance, spatial shifts in breeding grounds could alter migration distances, locations of stopovers, the durations spent refuelling and/or precipitate carry-over effects in subsequent seasons (Bearhop et al., 2004;Inger et al., 2010;McKinnon et al., 2015).As such, spatial changes in migration need to be considered within the perspective of the full annual cycle to understand their root causes and consequences on distribution, phenology and fitness.When spatial and environmental changes throughout the annual cycle are considered together, it is possible to disentangle the direct effect of environmental change from the indirect influence of change at other migratory stages.
Here, we review the literature on spatial changes in avian migration to fulfil the following aims: 1. provide evidence of spatial change at all migratory stages, 2. identify the key mechanisms driving change in spatial migratory patterns, 3. identify and discuss how anthropogenic change influences migratory patterns, 4. highlight the interdependencies between migratory stages, and 5. offer a framework for understanding migratory pattern changes in the context of full, multi-stage migratory journeys.

| ME THODS
We conducted a literature search using Web of Science in

| E VIDEN CE OF S PATIAL CHANG E
We first define the components of migratory journeys more comprehensively and then provide evidence and patterns of directional change at the population level for each component (Curley et al., 2020;Maclean et al., 2008;Välimäki et al., 2016).The three key spatial components are defined as follows: 1. Destinations: breeding and non-breeding areas, typically the two points at extreme ends of the migratory journey.Breeding areas are generally discrete due to spatial limitations imposed by nests/chick rearing, but non-breeding areas can be diffuse with individuals using multiple sites within a non-breeding period (Burgess et al., 2020), or using itinerant or eruptive movement strategies (Trierweiler et al., 2013).These diffuse ranges can be defined either by averaging locations during a fixed time period, or taking the main period of stationarity outside of the breeding season (Burgess et al., 2020).For species with multiple distinct non-breeding sites, spatial change could be measured at each distinct site.
2. Stopovers: location of stationary periods used for refuelling, recovery, information gathering or moult (Linscott & Senner, 2021) that are between migratory destinations.This includes 'classic' stopover sites where visits are brief (hours to days), and staging sites where stationary periods are longer (days to weeks; e.g.Fox et al., 2002;Klaassen et al., 2017).

Flight routes: space use during periods of active migratory
flight that can connect a stopover to a destination, link stopovers to other stopovers or connect one destination to another destination.

| Evidence for spatial shifts in destinations
Poleward distributional shifts are predicted under climate change, especially for resident species (Parmesan & Yohe, 2003).This pattern is evident in some migrant studies, but these tend to focus on a single group of closely related species or are from a limited geographic area meaning the results may not be generalisable (Lehikoinen et al., 2013;Maclean et al., 2008;Visser et al., 2009).Maclean et al. (2008) found consistent north-eastern shifts in the non-breeding grounds but only for seven species of Charadrii wading birds, and Lehikoinen et al. (2013) found the same pattern for only three species of European duck.Studies examining a broader range of taxa across larger spatial scales often find highly variable patterns, with species shifting poleward, equatorially, longitudinally or even showing stationarity (Curley et al., 2020;Hovick et al., 2016;La Sorte & Jetz, 2012;Pavón-Jordán et al., 2019;Potvin et al., 2016).
For instance, among 77 North American migrants, 28.3% of breeding distributions were found to move poleward, 30.9% moved towards the equator, 18.0% moved longitudinally, and 22% did not move (Curley et al., 2020).The most extreme destination shifts occur in partial migrant populations as the ratio of migrants to residents changes within the population (Buchan et al., 2020;Meller et al., 2016).This can stem from a relative change in the proportion of residents at the migrant breeding grounds, such as those seen in European Shags Phalacrocorax aristotelis (Grist et al., 2017), and Blackbirds Turdus merula (Møller et al., 2014), or a change in the proportion of residents at migrant non-breeding grounds, as observed in Barnacle Geese Branta leucopsis (Feige et al., 2008).
A few studies have assessed how breeding and non-breeding distributions shift in tandem for multiple species, instead of analysing one destination in isolation.They found that non-breeding distributions are shifting faster than breeding distributions for migrants breeding in Europe and North America (Lehikoinen et al., 2021), and that the directionality of change in non-breeding distributions is more consistent than for breeding distributions (Figure 1; Curley et al., 2020).This inconsistency in directionality results in both increased and decreased migration distances, measured as the distance between the population-level breeding and non-breeding distributional centroids (Curley et al., 2020;Potvin et al., 2016;Visser et al., 2009).For instance, Potvin et al. (2016) found that 11 of 29 migrant species breeding in Finland showed decreasing migration distances and the remainder were either increasing or exhibited no detectable change.

| Evidence for spatial shifts in stopovers
Shifts in the location of stopovers are less well studied compared to destinations but have been reported for larger, more conspicuous species, for example Eurasian Ruff Calidris pugnax (Verkuil et al., 2012), Pink-footed Geese Anser brachyrhynchus (Clausen & Madsen, 2016) and Barnacle Geese (Tombre et al., 2019).These examples often relate to highly philopatric species that congregate at discrete sites and/or rare habitats, allowing monitoring of long-term changes in numbers and distribution.For example, Semipalmated Sandpipers Calidris pusilla congregating at discrete intertidal stopovers have shifted to sites where predation risk is lower, in response to population recovery of a key predator (Hope et al., 2020).Some species use larger areas for stopovers and do not congregate at discrete sites, for example Pied Flycatchers Ficedula hypoleuca stage over much of the Iberian Peninsula (Bell et al., 2022), and thus inherently require large-scale data sets to detect species-level change.
In many documented instances, newly colonised stopover sites are located along the ancestral flyway, with the relative contribution of each migratory flight to the overall journey changing (Clausen & Madsen, 2016;Tombre et al., 2019;Verkuil et al., 2012).Staging areas of Barnacle Geese in Norway have shifted northward, lengthening the initial flight from Scotland but decreasing the distance to the breeding grounds in Svalbard (Figure 2; Tombre et al., 2019).
The inverse is found for Siberian breeding White-fronted Geese Anser albifrons, which have shifted their staging areas southward in Russia (Grishchenko et al., 2019).For many species, particularly passerines, there is limited information regarding the extent to which populations/species show stopover site philopatry, the spatial scale at which philopatry can operate and the extent to which discrete sites are used at the population level (Cresswell, 2014).Without this prerequisite knowledge, it is hard to monitor species for directional change in stopovers.

| Evidence for spatial shifts in flight route
Migratory flights are perhaps the most flexible part of migratory journeys, but the high variability in flight route exhibited by tracked individuals makes determining directional population-level change difficult.For example, individual Wood Thrushes Hylocichla mustelina take different routes across the Caribbean each year (Stanley et al., 2012), as do Egyptian Vultures Neophron percnopterus that cross the Sahara Desert (López-López et al., 2014).Therefore, separating directional shifts in flight routes at the population level from individual variation is difficult.Many migrations also include nocturnal, or high-altitude flight, that is hard to observe; and migratory fronts are often too broad to determine shifts in the weighted distributional centroid.However, direct observation suggests that Common Cranes Grus grus have increased the relative use of an established migratory route across Northern Italy towards Spain (Mingozzi et al., 2013), demonstrating spatial shifts can be detected for large soaring species.There is also evidence of a continental westward shift in the routes taken by some individuals from East Asian passerine populations breeding in Siberia, for example Richard's Pipit Anthus richardi (Figure 3) (Dufour et al., 2021).These populations normally winter in South East Asia, but a small proportion now migrate through Europe during autumn en route to non-breeding areas in Western Europe (Dufour et al., 2021;Jones et al., 2019).

| MECHANIS MS DRIVING S PATIAL CHANG E
The following mechanisms, none of which are mutually exclusive, can give rise to our definition of spatial change: 1. Relative changes in abundance among the migratory strategies (the combination of sites used across the annual cycle) of a population due to differences in demography associated with different strategies.
2. Individual-level flexibility allowing switching between existing migratory strategies, either within an individual's lifetime, or through offspring switching away from their parental strategies.

Individuals colonising new sites within and outside of existing flyways.
There are also factors that can constrain these mechanisms either physically, for example ecological barriers (Marjakangas et al., 2023), or genetically, for example inheritance of migratory orientation (Sokolovskis et al., 2023).It is important to identify the mechanisms that give rise to spatial change and the factors that limit the capacity for change in order to understand, predict and potentially mitigate the effects of anthropogenic change.Certain species traits may be linked with particular mechanisms that enable or constrain spatial change therefore allowing the species most susceptible to anthropogenic change to be identified.

| Population level mechanisms of spatial change
The weighted centroid of a migratory stage may shift if the relative abundance of individuals using different sites changes without individuals themselves switching sites or migratory strategy.This may occur if environmental change alters the survival or breeding success of different strategies (Buchan et al., 2020).For instance, four Afro-Palearctic migrants (Garden Warbler Sylvia borin, Tree Pipit

Anthus trivialis, Willow Warbler Phylloscopus trochilus and Common
Cuckoo Cuculus canorus) have experienced breeding population declines in England but have shown increases in Scotland, resulting in a northward shift in their weighted breeding distribution (Morrison et al., 2013).Also, the survival of Common Cuckoos breeding in England and Scotland is higher for birds that take a more easterly autumn migration route to Africa (Hewson et al., 2016).Over time, this could cause an eastward shift in stopovers and flight routes, as the relative abundance between easterly and westerly migratory routes alters.Abundance changes at one migratory stage could alter abundance at subsequent migratory stages over variable spatial scales due to differences in migratory connectivity and the diversity of migratory strategies between species (note that much evidence for this is theoretical; Finch et al., 2017;Gilroy et al., 2016;Taylor & Norris, 2010).Localised habitat degradation of the non-breeding grounds generally causes local declines in non-breeding ground abundance.However, on the breeding grounds, this would translate to diffuse changes in breeding abundance for species with weak migratory connectivity and a high diversity of migratory strategies, and localised changes in breeding abundance for species with strong connectivity and low diversity (Finch et al., 2017;Taylor & Norris, 2010).This is because individuals disperse across a broader range of breeding sites from any localised area on the non-breeding grounds when connectivity is weak and diversity is high.

| Individual -level mechanisms of spatial change
Spatial shifts can also occur by individuals altering migratory strategies within their own lifetimes (flexibility) or by their offspring adopting different strategies (developmental plasticity; Åkesson & Helm, 2020).High individual flexibility is currently driving a northward shift in the spring staging areas of Barnacle Geese in Norway (Oudman et al., 2020), while high developmental plasticity is driving a northward non-breeding ground expansion in Black-tailed Godwits Limosa limosa with juveniles leading the colonisation of northerly sites (Gill et al., 2019).Individual flexibility can vary between species from near complete itinerancy between years, for example Pink-footed Goose (Clausen et al., 2018), to high philopatry, following initial breeding and non-breeding site selection, for example Pied Avocet, Recurvirostra avosetta (Chambon et al., 2019).
Within species, flexibility can also vary across stages of the annual cycle, and higher flexibility of non-breeding site choice over breeding site choice (Guillemain et al., 2013) could drive the larger spatial shifts in non-breeding grounds recorded by Lehikoinen et al. (2021).
Developmental plasticity is often lower for breeding ground site choice as many individuals return near to natal sites to breed (Förschler et al., 2010).On the other hand, first non-breeding site choice is often more flexible, especially for young birds that migrate independently of adults (Gill et al., 2019).(Trierweiler et al., 2013).Site choice flexibility occuring within seasons would perhaps allow faster spatial changes compared to inflexible site choice in species such as Black-tailed Godwits where non-breeding site philopatry is high following natal site choice (Gill et al., 2019).
In addition, the variability of environmental conditions as opposed to the mean trend may alter flexibility and developmental plasticity.
We give examples here of two competing mechanisms that lead to environmental variability having opposing effects on flexibility and developmental plasticity.On one side, theoretical models show that high variability in the environment can drive high natal site philopatry, as individuals return to natal breeding sites as a place of known survivable quality, rather than risk the costs of site switching to a new and uncertain environment (McNamara & Dall, 2011).
Conversely, the climatic extremes produced by high environmental variability, especially tropical storms and strong winds, could facilitate vagrancy -individuals occurring outside of known population distributions.This, in turn, might enable the colonisation of new sites outside existing flyways that would promote spatial change (Lees & Gilroy, 2021).
Innate migratory direction (Helbig, 1991) and activity (Berthold & Pulido, 1994; which is correlated with migratory distance) are genetically heritable traits that can be altered by environmental conditions, for example agricultural chemical exposure (Eng et al., 2017).The mechanisms of genetic inheritance differ between the few species in which it has been studied.For Blackcaps Sylvia atricapilla (Helbig, 1991) and Swainson's Thrushes Catharus ustulatus (Delmore & Irwin, 2014), there appears to be co-dominant inheritance with young exhibiting an intermediate migratory direction compared to both parents, while for Willow Warblers, a dominant inheritance mechanism means most individuals exhibit the migratory direction of a single parent (Sokolovskis et al., 2023).These differences potentially influence the likelihood that juveniles adopt different migratory strategies to their parents, with dominant inheritance being more constraining.However, the variation in these genetically controlled traits within a species is arguably just as important for spatial changes to migration as the inheritance mechanism itself.In some instances, the range of phenotypic traits can be limited in certain directions and constrain the capacity for change.For instance, individual Swainson's Thrushes originating from inland populations are constrained to take an indirect autumn migration that retraces post-glacial range expansion (Ruegg et al., 2006).Individuals with extreme phenotypes for migratory direction (Thorup et al., 2012) and distance (Veit, 2000) are often those that occur as vagrants outside of existing ranges.
Individuals with the largest genetically encoded migratory distances may be important in northward range expansions linked to climate change (Veit, 2000).While these endogenous genetic programmes broadly shape the orientation and rough duration of the initial migration, stopovers and final non-breeding site choice are also influenced by environmental conditions encountered (see above; Vansteelant et al., 2017), social interactions (see below; Oudman et al., 2020) and epigenetic changes in response to environmental factors, although evidence of this in the context of bird migration is currently weak (Merlin & Liedvogel, 2019).
Social interactions influence flexibility and development plasticity as they can override genetically controlled endogenous programmes and shape choice of migratory strategy (Byholm et al., 2022;Oudman et al., 2020;Teitelbaum et al., 2016).For example, juvenile Whooping Cranes Grus americana follow experienced adults to recently colonised non-breeding sites (Teitelbaum et al., 2016).These socially inherited migratory strategies have a high likelihood of suitability, as often many other individuals have successfully adopted the strategy and survived, as opposed to genetically inherited strategies where migratory strategy is inherited from only the parent(s).The use of social cues has been demonstrated in many larger taxa (e.g.geese, swans and cranes; Byholm et al., 2022;Harrison et al., 2010;Teitelbaum et al., 2016) as the techniques, such as colour-mark resightings or tracking data, required to demonstrate the use of social cues are generally restricted to larger taxa.It is unclear if the use of social cues extends more broadly across avian taxa, but circumstantial evidence is mounting that social cues are more widely used, even in species traditionally considered as solitary migrants, for example Common Cuckoos (Piersma, 2022).While widespread social transmission has the potential to cause rapid spatial change, it should theoretically only enable shifts to sites/routes that are already colonised.However, it could be important in the growth of recently colonised sites.As demonstrated by Pink-footed Geese colonising new breeding sites in Novaya Zemlya, Russia, where social transmission of this new migratory route has ignited exponential popualiton growth in the region (Madsen et al., 2023).

| Constraints on mechanisms of spatial change
The geography and topography of land masses can guide spatial shifts of stopovers and destinations in certain directions.Arctic species at the northern edge of their breeding range, or high-altitude breeders, have little capacity for northward or elevational changes (Hayhow et al., 2015).For instance, isotherm convergence in highaltitude areas of the European Alps constrains montane species, for example Alpine Accentor Prunella collaris, due to a lack of higher altitude habitat within dispersal distance (Brambilla et al., 2022).More broadly, across 2092 breeding bird communities in Europe, it has been found spatial shifts are larger if the communities are further from coastline and they tend to move in the direction of least elevation change (Marjakangas et al., 2023).Ecological barriers, for example oceans, deserts and mountain ranges, are often found along the shortest migratory path between destinations.There are costs associated with crossing barriers that heighten the risk of mortality.These include acquiring and carrying large fuel loads (Alerstam, 2001), the lack of uplift over open water crossings (Becciu et al., 2020) and few/no refuelling opportunities.However, if individuals cross ecological barriers directly, then they benefit by taking the shortest migratory path, decreasing migration duration, overall energetic expenditure and predation risk (Gill et al., 2009).These costs and benefits vary with species-level and individual level traits.For example, if the costs exceed the benefits, and alternative migratory routes are available, then individuals may circumvent the barrier crossing.This can create detours or funnels in migratory flights, especially at narrow land bridges, meaning that flight routes and stopovers are geographically constrained during the circumvention.This is illustrated by many raptor species, for example Black Kite Milvus migrans and Booted Eagle Hieraaetus pennatus, where migratory flight routes are constrained to go through the straits of Gibraltar to minimise the Mediterranean sea crossing (Hahn et al., 2014).Conversely, if the benefits exceed the costs, then individuals cross the barrier directly.Prior to barrier crossings, individuals require fuel loads that enable a successful crossing (Strandberg et al., 2010), as failure to reach suitable fuel loads increases the risk of mortality.After barrier crossings, individuals can accumulate in high abundance (Lafleur et al., 2016) as they need to refuel, and often exhibit reduced selection for high-quality sites as they do not have adequate fuel stores for costly searching behaviour (Buler & Moore, 2011).These fuelling requirements constrain stopover site choice prior to and post barrier crossing.As such, the availability of suitable stopovers near the interfaces with ecological barriers will act as constraints on the potential for change.

| HOW ANTHROP OG ENIC DRIVER S LE AD TO CHANG E S IN MI G R ATORY PAT TE R N S
Two important anthropogenic drivers leading to changes in migratory patterns are land use change and climate change (summarised in Figure 4; Zurell et al., 2018).They are widespread and spatially variable drivers, causing both slow and rapid change to landscapes and flyways, with direct and indirect effects on migratory patterns (Zurell et al., 2018).Land use and climate change may interact and new sites that fall within shifting climatic optima may already be severely degraded due to anthropogenic land use change.There are broader reviews on the impacts of land use change (Stanton et al., 2018) and climate change (Thomas, 2010) on range shifts for resident species, and these provide a more detailed summary that has many parallels to migrants.Here, we focus on elements that are important to migrants while still covering the key effects.

| Land use change
The anthropogenic conversion or modification of natural habitats due to deforestation, agricultural expansion and urbanisation accounts for 60% of habitat change globally (Song et al., 2018).
Reductions in habitat extent and/or increases in habitat degradation reduce site carrying capacity (Haran et al., 2020) leading to redistributions in abundance as populations track areas of suitable habitat.
However, land use change can increase and decrease habitat suitability for different species with differing ecology.For example, deforestation for agriculture negatively impacts woodland specialists; deforestation in the Sahel reduces Subalpine Warbler Sylvia cantillans numbers (Cresswell et al., 2007), but has positively impacted waterfowl populations sizes and promoted redistribution of waterfowl into areas of agricultural expansion (Feige et al., 2008;Gauthier et al., 2005).Similarly, urbanisation results in habitat loss for many species, but urban-forest fragments can be high-quality stopovers for Neotropical migrants (Oliver et al., 2011) with populations aggregating at higher densities than expected (Greco & Airola, 2018).
Other factors associated with land use change such as noise pollution, light pollution and anthropogenic disturbance can further degrade habitats and/or alter the energetics of migration and lead to spatial redistributions (Béchet et al., 2003).Areas of increased noise pollution are generally associated with lower densities of birds (Barbosa et al., 2020;Klingbeil et al., 2020) and is analogous to habitat degradation from land use change (Ware et al., 2015).The predominant effect of light pollution is as an attractant for birds actively migrating (Spoelstra & Visser, 2013).A single intense light source in New York impacted c1.1 million individuals during nocturnal migration over just 7 days (Van Doren et al., 2017).This can increase energetic costs of migration as individuals travel further due to the initial attraction and then subsequent retention at the light source.
Other manmade structures such as windfarms can similarly increase energetic costs due to avoidance behaviour increasing migratory F I G U R E 4 Flowchart detailing how current anthropogenic drivers directly impact migrant avian populations and how the possible responses of these migrants may lead to a multitude of spatial shifts in their migratory patterns.
distances (Masden et al., 2009;Plonczkier & Simms, 2012).Lastly, disturbance associated with anthropogenic activities can alter energetic balances and fitness, as individuals increase energetically costly flight behaviour and subsequently have reduced foraging time (Béchet et al., 2004).For example, the introduction of Snow Goose Anser caerulescens spring hunting resulted in a westward stopover shift as individuals oriented towards sites with lower disturbance (Béchet et al., 2003).Migratory routes associated with higher energetic costs will likely reduce fitness and will over time stimulate the redistribution of abundance and/or individuals using these routes.
Other changes in land use not directly associated with habitat loss or degradation can also increase mortality.Many migrants are killed by hunters (Gallo-cajiao et al., 2020) or by collisions with human infrastructure (Cusa et al., 2015), such as wind farms, power lines and buildings.This can alter the abundance of individuals using different migratory strategies and the abundance of migrants in partial migrant populations.The risk of collision is spatially heterogeneous, and for 27 migrants in Europe and North Africa, risk was highest around the strait of Gibraltar and the Bosporus in Turkey (Gauld et al., 2022).In a partial migrant population of Great Bustards Otis tarda in the Iberian Peninsula, increased collision rates with power lines could be driving the decline of migrant individuals (Palacín et al., 2017).

| Climate change
Avian migrants have a 'thermal niche' (Socolar et al., 2017), and distributions are driven by resource availability that is correlated with temperature and to some extent by physiological limitations at climatic extremes (Quintana et al., 2022).Therefore, as climate change shifts isotherms northward and to higher elevations, resource landscapes also shift along the same axis.Consequently, populations either redistribute along latitudinal and elevational clines or persist and face the repercussions of climate change (Chen et al., 2011;La Sorte & Thompson, 2007).One of the key implications of this pattern, specific to migrants, is the differential impacts on within hemisphere migrants vs. trans-equatorial migrants.For species migrating within the same hemisphere, breeding and non-breeding grounds would shift in the same direction on average if shifting isotherms were tracked.For trans-equatorial migrants, breeding and nonbreeding would shift in opposing directions on average, lengthening migration distances (Howard et al., 2018).However, there is much variation in the rate of temperature change within each hemisphere (Cohen et al., 2014;Loarie et al., 2009), and therefore, this is not a uniform trend and even species migrating within a hemisphere could experience very different rates of warming at breeding and non-breeding grounds.This may be particularly true for species with varying habitat requirements across the migratory cycle as the rate of temperature change varies with habitat.Climate change is occuring rapidly in flooded grasslands, mangroves and deserts, while it is comparatively slower in coniferous forests and montane grasslands (Loarie et al., 2009).
Increases in the frequency of extreme climatic events (Cohen et al., 2014;Coumou & Rahmstorf, 2012) and changes to the strength and direction of prevailing winds (La Sorte & Fink, 2017) may have specific effects on migrants.Extreme climatic events can cause high levels of mortality over short time periods during active migration (Newton, 2007) and therefore reduce abundance of individuals using migratory strategies that spatially or temporally overlap with climatic extremes.In a population of partial migratory European Shags in Scotland, two extreme winter storms over 9 years caused low survival of resident individuals and selection against this strategy (Acker et al., 2021).Shifts in prevailing winds will alter energetic costs associated with different migratory flight routes both positively and negatively, and in turn the viability or desirability of certain routes (Nourani et al., 2017) or breeding sites (Krietsch et al., 2020).For Neotropical passerines migrating across the Caribbean, the strength of energetically costly crosswinds in spring are predicted to decrease (La Sorte & Fink, 2017), thereby decreasing the risk of mortality.This could increase the proportion of individuals migrating over the ecological barrier as opposed to circumventing it via the Central American land bridge in species where this dual strategy exists, for example Red-eyed Vireos Vireo olivaceus (Sandberg & Moore, 1996;Smolinsky et al., 2013).
Increasing desertification (Burrell et al., 2020) will shift resource abundance due to habitat degradation.Therefore, desertification will have similar population-level impacts to habitat degradation associated with land use change, but spatial shifts will occur along the same axis as desert expansion.In the Sahel region south of the Sahara Desert, vegetation zones have shifted 600 m southward per year (Gonzalez, 2001), meaning Afro-Palearctic migrants spending the non-breeding season in this region should shift southward to track suitable habitats.Desertification also lengthens ecological barrier crossings and subsequently increases migratory costs by raising energetic expenditure (Schmaljohann et al., 2009) and physiological stress (Eikenaar et al., 2020).There is currently no direct evidence for this altering migratory patterns.Nevertheless, this could cause shifts in flight routes to circumvent potential desert crossings (Smolinsky et al., 2013), or intensify stopover site use at the very edges of ecological barriers to minimise barrier crossing distance, and/or allow recovery immediately after barrier crossings (Blackburn et al., 2019).

| INTER DEP END EN C IE S B E T WEEN MI G R ATORY S TAG E S A S PART OF AN OVER ALL FR AME WORK
Any spatial change in a migratory stage will have cascading consequences across other aspects of the migratory cycle.Here, we focus on the fact that spatial change can bring about further spatial changes at other stages of migration.This is due to dependencies between destinations, stopovers and flight routes that are implicit when migration is considered as a chain of interlinked steps rather than a set of discrete stages.In this section, we highlight some specific cases of interdependencies that are likely to act (although direct evidence is limited), and then present a framework that explores how spatial change may arise.

| Inter-stage interactions
Large changes in migratory orientation will cause many stages of migration to change in tandem; however, examples of very large changes in orientation are not that common.One recent example is from a small subset of Richard's Pipits that have shifted their migratory orientation almost 180° to now migrate westward from East Asia to Western Europe, instead of southward into South East Asia (Dufour et al., 2021).This drastic change will result in changes to Changes in destinations often cause changes in overall migratory distance and duration at the population level and individual level, which can have implications for the number and spatial configuration of stopovers.Individuals can buffer against longer migrations, and the associated increased energetic costs, by increasing the number of stopovers (Lindström et al., 2019).Stopover frequency is flexible at the individual level (Carneiro et al., 2020) and can allow for additional refuelling.Increases in the number of stopovers used at the individual level has been predicted for Afro-Palearctic migrant passerines in response to northward breeding ground shifts caused by global warming (Howard et al., 2018).Where these additional stopover(s) are accommodated in the annual cycle will ultimately depend on the availability of suitable habitat at the required points along migratory journeys.It should be noted that the increased energetic requirements could also be met by changes to temporal aspects of stopovers, that is stopping for longer or increasing refuelling rates (Lindström et al., 2019).
Stopover sites often shift within the bounds of original flyways, as for Pink-footed Geese and Barnacle Geese where stopovers have moved northward along existing flyways (Bauer et al., 2008;Tombre et al., 2019).Therefore, the relative contribution of each migratory step to the overall journey and the relative departure fuel loads required could change.In the case of these two geese species, the distance from the final stopover to the breeding grounds is shortened and could allow individuals to reach sites further north without stopping, thereby altering breeding distributions.The reverse would be true if the final migratory flight increased in distance.This is perhaps the case for White-fronted Geese now staging further south in Russia, due to agricultural abandonment further north (Grishchenko et al., 2019).Therefore, the change in distribution of suitable staging habitat could prevent birds from reaching breeding areas at their northerly range limit early enough to allow for breeding.
Many migrants benefit from supportive wind directions to survive migrations (Erni et al., 2005), with flight routes being flexible to local wind conditions (Senner et al., 2019;Vardanis et al., 2011Vardanis et al., , 2016)).If individuals track changes in prevailing wind, then this could affect the length and direction of migratory flights, and in turn where birds decide to settle at stopovers and/or destinations (Krietsch et al., 2020;Vansteelant et al., 2017).In Eurasian Honey Buzzards Pernis apivorus, individual tracking has shown that the distance and direction of non-breeding sites from the breeding grounds are dictated by different wind conditions experienced during flight (Vansteelant et al., 2017).Therefore, in this instance, population level shifts in non-breeding distribution could be driven by changes in flight routes caused by alterations to prevailing winds.

| Framework
Given the diversity of spatial changes we have covered, we present many of the possible changes to migratory patterns that can occur (Figure 5).This groups spatial changes into whether they are

| FUTURE DIREC TIONS
Monitoring multiple stages of migratory journeys in tandem for a single population would allow us to assess how spatial changes at one migratory stage may alter distributions at subsequent stages, therefore following on from the joint tracking of breeding and nonbreeding distributions by Lehikoinen et al. (2021).Global citizen science initiatives such as eBird offer considerable hope for quantifying shifts in destinations and stopovers, and methodological advancements to account for spatiotemporal biases will improve their utility (Johnston et al., 2021).The potential to identify spatial shifts in destinations and stopovers from such rich data are illustrated by assessments of phenological changes in migratory birds (Horton et al., 2018;La Sorte & Horton, 2021).Detecting changes in flight routes is challenging, particularly for smaller taxa.Radar technology offers some potential, especially if radar networks could measure change over broader spatial scales that relate to our framework.These datasets used in combination could examine the extent to which migratory stages are interdependent facilitating the exploration of novel inquiries.For instance, we could address questions such as the consequences of breeding ground shifts on stopover distribution and layout.They may also allow us to account for the interdependencies between migratory stages, and model how full migratory distributions will change under future environmental change.
There is a need to deepen understanding of the extent and drivers of flexibility in migratory strategies to understand the capacity for spatial change.Biologging will continue to be one of the main tools, allowing insights into how flexible adults are in site and route choice between years (Åkesson & Helm, 2020;Dias et al., 2011;Grecian et al., 2019;López-López et al., 2014).Further work should focus on the environmental and intrinsic drivers of flexibility, and which cues individuals use in site choice decision-making.In comparison, we know even less about the extent and drivers of developmental plasticity and the factors that govern natal site and route choices, but see (Studds et al., 2008;Vansteelant et al., 2017).This is important as natal site selection often governs population-level distributions for species with high adult philopatry (Gill et al., 2019).
Extensive tracking of juvenile birds of known breeding origins could help to determine what factors govern natal site and route choices, but this is rarely done due to low return rates for archival tracking devices deployed on juveniles, but see (Borrmann et al., 2021;Byholm et al., 2022).
Biologging tags can be currently used on larger birds (e.g.waterfowl and raptors) to measure the energetic and physiological costs of migration (Weegman et al., 2017), mortality risk (Buechley et al., 2021) and breeding performance (Ozsanlav-Harris et al., 2022) associated with different migratory strategies.Migratory strategies linked to higher survival or reproductive output could therefore be identified to better predict the directionality of future spatial shifts.et al., 2018) and how future wind regimes may affect migratory flight routes (La Sorte & Fink, 2017;Nourani et al., 2017).While there are benefits to tracking potential environmental change, the actual capacity for change may be limited by energetic and physiological costs of migration (Howard et al., 2018), time constraints on breeding (Duijns et al., 2019) and interdependencies between migratory stages.Therefore, the extension of climate change projection models to encompass these costs and constraints could offer insights into the limitations of tracking climate envelopes and even determining when migration might cease entirely (Buchan et al., 2020).
Improving predictive models will increase the accuracy of future projected distributions, thus allowing us to identify key areas to conserve for future population viability, and identify species most at risk due to high constraints on the capacity for spatial shifts.

ACK N O WLE D G E M ENTS
We thank Liam Langley and Rhianna Hughes for providing constructive feedback on later versions of this manuscript.The lead author was funded jointly by the University of Exeter studentship and the Wildfowl and Wetlands Trust.

CO N FLI C T O F I NTE R E S T S TATE M E NT
None to Declare.

PE E R R E V I E W
The peer review history for this article is available at https:// www.webof scien ce.com/ api/ gatew ay/ wos/ peer-review/ 10. 1111/ ddi.13785 .
November 2020 using the terms in Appendix S1.This was not a systematic literature search, and instead sought to find examples of spatial changes in migratory patterns, and the mechanisms driving change.The search returned 7178 hits, which were ordered anthropogenic change and assessing future population viability within this vulnerable and extensive taxonomic group.K E Y W O R D S breeding, climate change, distribution, flexibility, migratory behaviour, non-breeding, range shift, stopover by relevance, and the top 1000 hits were screened for relevant papers.The titles and abstracts were scanned for papers mentioning birds or a bird species, and any aspect of migration to determine whether they were relevant.Papers were then retained if they measured change, or investigated the mechanisms of change, in any spatial aspect of avian migration at the population level or individual level.This resulted in 81 retained papers published 1991-2020.We supplemented these with material referenced within retained papers and those known by the authors prior to commencing the review.

F
Spatial changes in the non-breeding (panel a) and breeding (panel b) distributions of 77 North American Migratory birds from 1990 to 2015 (Curley et al., 2020).Each arrow represents a single species.The direction and length of each arrow represent the direction and magnitude (in kilometres) of the spatial change in the migratory destination from 1990 (0, 0 on the graph).The scale of distance moved longitudinally and latitudinally is shown on the x and y axis, respectively.Note that the arrow direction is significantly more consistent in panel (a).Arrows above/below the horizontal dashed line indicate north/south movements; arrows to the right/left of the vertical dashed line represent east/west movements, respectively.Diversity and Distributions, Volume: 26, Issue: 4, Pages: 415-425, first published: 03 February 2020, DOI: (10.1111/ddi.13036)© 2020 The Authors.Diversity and Distributions published by John Wiley & Sons Ltd.

F
I G U R E 2 Northward shift in the spring stopover distribution of Svalbard-breeding Barnacle Geese (Tombre et al., 2019).The main map depicts full spring migration route from the UK to Svalbard via Norway (arrowhead lines in green).The main wintering areas on the Solway Firth (most southerly green dot) and the two main Norwegian stopover areas (more northerly red and blue dots) are also shown.The inset map shows the main staging areas of geese in Helgeland (southerly red shaded region) and the newly colonised area in Vesterålen (northerly blue shaded region).The colonisation and increasing use of Vesterålen mean that the centroid of spring stopover distribution is moving further North.Global Change Biology, Volume: 25, Issue: 11, Pages: 3680-3693, first published: 02 September 2019, DOI: (10.1111/ gcb.14793) © 2019 The Authors.Global Change Biology published by John Wiley & Sons Ltd.

F
I G U R E 3 New western migratory route in Richard's Pipit, resulting in a change in flight route, stopovers and non-breeding site(Dufour et al., 2021).The breeding, ancestral migration range and ancestral non-breeding range are used by the majority of the population and taken from BirdLife International.The western non-breeding areas and autumnal migration route have been recently colonised and in this instance were shown using geolocator tracking.The inset panel shows the locations of where birds were caught to be fitted with the geolocator devices.Current Biology, 32, Dufour, Paul; de Franceschi, Christophe; Doniol-Valcroze, Paul; Jiguet, Frédéric; Guéguen, Maya; Renaud, Julien; Lavergne, Sébastien and Crochet, Pierre-André, A new westward migration route in an Asian passerine bird, 1-7, © 2021 Elsevier Inc (2021), with permission from Elsevier.
Spatial shifts due to flexibility or developmental plasticity can involve individuals changing to sites/routes that are already used, colonising new sites/routes within the existing flyway, or new sites/ routes outside of existing flyways.These distinctions are important as they can occur through a different set of mechanisms and can be subject to different constraints.Here, we discuss three factors that are important in driving and/or constraining flexibility and developmental plasticity: (a) environmental change and variation, (b) genetic inheritance of migratory behaviour and (c) social transmission of migratory behaviour.Individuals can detect environmental and habitat change across seasons and adjust migratory behaviour accordingly (Thorup et al., 2017).Therefore, flexibility and developmental plasticity can manifest as individuals encounter optimal ecological conditions and are able to occupy these sites.This spatial response to the environment could be a genetically encoded reaction norm, similar to other non-migratory traits (Chik et al., 2022), and distributions change over time as individuals choose to settle in different sites to track trends in mean climatic conditions.In some species, individuals also behave flexibly within non-breeding periods and adaptively alter site use within a season in response to environmental change.For example, Montagu's Harriers Circus pygargus track shifting prey availability across the Sahel during the non-breeding season the non-breeding destination, stopovers and flight routes; all interconnected and only discernible in this manner because this study encompassed the full migratory cycle.Not all examples are this drastic and below we give examples of how: (a) changes to destinations could alter stopover distributions, (b) changes to destinations could alter breeding distributions, and (c) how changes to flight routes could alter non-breeding distributions.
caused by shifts in the destination, or the journey (stopovers and flight routes), and what direction migratory distance changes in.Migration distance is chosen to delineate change as it has important implications for phenology and/or stopover site choice (Packmor et al., 2020; Schmaljohann, 2019), and it is only possible to calculate when all spatial changes across the migratory cycle are considered.Spatial change in any of the three migratory components (Figure 5) may occur directly in response to environmental change at the stage in question, or indirectly in response to environmental and/or spatial change at previous stages of the migratory cycle that carry-over to the stage in question (Figure 6).Disentangling the direct effects of the environment from the interdependencies between migratory stages is key to understanding the drivers and mechanisms underpinning spatial changes.It requires all stages of the migratory cycle to be considered, and not just a single stage in isolation.Figure 6 is not comprehensive but illustrates how spatial migratory change may come about.Environmental change can influence migration (Figure 6, top row) and lead directly to changes in destinations, stopovers and flight routes (Figure 6, middle row).The direct changes may cause further indirect change to other stages of migration (Figure 6, bottom row).Following the far left path of the diagram, we see how environmental change can lead to southward shifts in wintering grounds that then require an additional stopover to reach.This framework encourages researchers to consider how changes to the destinations may impact the migratory journeys between these destinations, inducing changes in flight route tortuosity and stopover configuration, and vice versa.
For instance, measuring the energetic and physiological costs of migrating in different wind conditions could enable determination of flight routes that maximise fitness and how future shifts in prevailing F I G U R E 5 A framework of spatial changes in avian migratory patterns.Changes are placed in the context of the full migratory journey and grouped according to whether shifts in destinations (row a) or stopovers/flight routes change (row b) cause spatial change and what the ultimate effect of these changes is on migration distance.In (row a), we show how changes in one or both destinations can impact migratory patterns.While we visualise changes here as seemingly north-south shifts in destinations, this framework could be extended to destinations shifts in any direction.In (row b), we show how changes in stopovers and flight routes lead to changes in migratory patterns.We group these together as changes in one are often implicit of changes in the other.Working from left to right across (row b) we show the following changes in migratory patterns: (1) original baseline migration, (2) decrease in route tortuosity and middle stopover shifts, (3) final stopover becomes winter destination, (4) increase in route tortuosity and first two stopovers shift and (6) first stopover shifts but no change in flight route.winds could redistribute abundance throughout the migratory cycle.Multi-sensor biologging data also allow measurement of net energetic gain (e.g. via intake rates and energy expenditure) from across a population's range.This would allow researchers to predict abundance redistributions with greater accuracy, but perhaps more importantly identify migratory sites where populations are more energetically constrained, and would therefore be more susceptible to anthropogenic change.Researchers have begun to model how migratory destinations could shift under current climate change projections (Howard F I G U R E 6 Direct and indirect spatial changes in avian migrations.Environmental change can affect migration (top row) and bring about spatial change in a migratory component(s) (middle row).This direct spatial change can then cause further indirect spatial change in other migratory stages (bottom row).A single migratory stage is shifted in this diagram but multiple stages could be simultaneously impacted.If the original migration (top row) is oriented along a north-south axis, then this corresponds to southward shifts in non-breeding distribution or eastward shifts in breeding distribution (left side, middle row).Then, this may indirectly cause an additional stopover to be used as a southward shift in non-breeding distribution increases migratory distance or an eastward shift in the flight route to reach the more easterly breeding distribution (left side, bottom row).