Flyway connectivity and exchange primarily driven by moult migration in geese

GPS tracking

Extraction of migration stopovers, nest and moult sites

For detection of overlap of the two (in winter disparate) populations, we selected tracks between 1 March – 15 November, thus including spring migration, breeding, moult and autumn migration. We extracted migration stopovers as sites where geese spent at least 2 days in an area of 30 km radius [32] during spring (1 March – 1 June) and autumn (15 August – 15 November). Nesting sites were defined where geese spent at least 10 days in an area of 2 km [33] during 15 May – 15 July, thus including successful and failed breeding events. Failed breeders were determined as birds moving > 100 km from their nesting sites in the time of 10–26 days after nest initiation. Moulting sites were determined where geese spent at least 21 days in an area of 30 km radius between 1 July and 1 September.

Detection of population overlaps

For each of those periods and for each individual goose, we calculated the utility distribution area with dynamic Brownian bridges (function “brownian.bridge.dyn” in R-package “move” [34], window size = 31, margin size = 11) as raster maps with grid cell size 5 × 5 km for migration stopovers and 750 × 750 m for nesting and moulting sites. In order to combine those to population level utility distributions, we defined “used grid cells” as those where at least one individual bird was expected to be for > 15 min (threshold based on data resolution). For each population and the four periods (see Extraction of migration stopovers, nest and moult sites) we determined overlapping used grid cells and calculated the proportion of cells that overlapped. For further analyses, the overlapping sites were grouped by period and distance into regions. In order to detect if the region use by members of the two populations also overlapped in time, for each such region we extracted the number of individuals per population that used it (i.e. had GPS positions less than 30 km from the cells’ borders; threshold based on stopover site definition) during or outside the stopover/nesting/moulting time and quantified the time and duration of use and overlap.

Example tracks for population switching

The data set includes tracks of two adult males that were caught in either The Netherlands (2014/15; goose NK3) or northern Germany (2015/16, goose 424), and were observed to switch to wintering in Hungary in one or more subsequent years. We present their tracks, migration stopovers, nesting and moulting sites (as above) during the two available years of data each and indicate likely regions of population switching.

Migratory stopovers, nesting and moulting sites

Geese of the North Sea population used on average 6.0 (± 0.2, standard error, range: 1–11) stopover sites during spring migration and 4.9 (± 0.2, range: 1–8) stopover sites during autumn migration. The stopovers were rather evenly distributed along the flyway, especially in spring (Fig. 1a, d). Geese of the Pannonic population used 7.0 (± 0.5, range: 5–9) stopover sites during spring and 5.7 (± 0.7) stopover sites during autumn migration. Stopovers of Pannonic geese were more concentrated in Kazakhstan and not as evenly distributed over the flyway (Fig. 1a, d).

From the North Sea population we detected a nesting site for 54 birds (67%) and a moulting site for 57 birds (70%). Of the 54 breeders, 20 birds (37%) failed and moved up to 2000 km to moult on Taimyr peninsula (Fig. 1b, c). Additionally, of the 27 non-breeders, 9 birds (33%) moulted on Taimyr, so in total 36% of the birds tracked in the North Sea population moulted there. From the Pannonic population we extracted a nesting site for 5 birds (50%) and a moulting site for 9 birds (90%). Two birds nested on Gydan peninsula and 3 birds nested on Taimyr. All 9 (i.e. 100%) extracted moulting sites were situated on Taimyr (from the two birds nesting on Gydan, one likely failed breeding and moved to Taimyr, whereas from the other bird a moulting site was missing).

Regions and timing of population overlap

The two populations of geese had overlapping sites during spring migration stopover, moult and autumn migration stopover, but not for nesting sites. The two overlap regions during spring migration were very small and both located close to the border between Russia and Kazakhstan (Fig. 2b), in the following named West Kazakhstan and North Kazakhstan. Those overlap regions summed to less than 1% of the total spring migration stopover area of both populations (Table 2). From each population, only one bird stopped in either of the two regions, but timing overlapped well (Fig. 3). Geese stayed for 7–20 days (Table 3), so there might have been possibility for exchange.

	North Sea population	Pannonic population
West Kazakhstan	19.8	6.8
North Kazakhstan	16.6	13.7
West Pyasina	21.8 (±12.7, N = 3)	41.3
Pyasina River	35.8 (±2.0, N = 8)	34.9 (±1.6, N = 2)
East Pyasina	33.6 (±9.1, N = 4)	40.8
Pyasina autumn	15.7 (±4.3, N = 8)	13.8 (±7.1, N = 3)
Gydan	17.3 (±4.7, N = 5)	6.5 (±6.5, N = 2)
Bely Island/ North Yamal	50.8 (±1.4, N = 2)	21.0
Central Yamal	5.3 (±3.7, N = 3)	–

Standard error and sample sizes (number of individuals in the region) indicated in brackets, if none are indicated sample size is N = 1 and therefore no standard error

All overlapping sites during moult were situated on Taimyr at the Pyasina delta (Fig. 2c), for comparison we have split them into three regions: West Pyasina, Pyasina River and East Pyasina. The proportion of those overlap cells from the Pannonic population was comparatively high (16.2%; Table 2) and also from the North Sea population it was higher than for the migration stopovers (1.5%). This indicates a small area for a high density of North Sea and Pannonic birds that moult in the Pyasina delta and much possibility for exchange. Timing overlapped well for the many tracked birds that moulted there (Fig. 3) and staying duration was 22–41 days (Table 3). Several birds, especially of the North Sea population, shortly passed by before or after moulting there or elsewhere.

There were large areas of overlapping autumn migration stopovers close to the breeding and moulting sites, some still in the Pyasina delta, others on the adjacent peninsulas and islands. We define four regions for further use: the Pyasina delta (hereafter called Pyasina autumn), Gydan, Bely Island/North Yamal and Central Yamal (Fig. 2d). The proportion of overlap cells were again small, but larger than for spring migration stopovers (Table 2), which might be due to the closeness to the moulting sites. The timing of region use overlapped well for the two populations (Fig. 3), maybe less so for Central Yamal due to low sample size. Staying durations were more variable than during spring or moult, namely 5–51 days (Table 3). Many geese seemed to pass by the Pyasina autumn region long before autumn migration, but these were birds that had moulted in the same area before, now likely fuelling for autumn migration flight (Fig. 3).

Observed population switches

The first goose (NK3) for which we observed a population switch, followed a southern route after departing from The Netherlands during its first tracked spring migration (2015). We could not detect a nesting attempt, but instead it moved far to Eastern Taimyr for moult (Fig. 4a). The subsequent autumn migration started off far south, close to the tracks of the Pannonic population. However, that year it changed direction, crossed the Ural mountains westward and joined the North Sea population birds towards its previous wintering site in The Netherlands. The second tracked spring migration (2016) was even further south, but still in the flyway of the North Sea population, again skipping to nest and towards Eastern Taimyr for moult. Also that year’s autumn migration started off towards the south, however this time the bird stayed in the flyway of the Pannonic population east of the Ural mountains, doing a stopover in Northern Kazakhstan (see other Pannonic birds in Fig. 1c), ending up in Hungary to winter. It is likely that this bird switched from the North Sea population to the Pannonic population during moult in 2016. Unfortunately, we could not follow its spring migration in 2017 after it was shot in Hungary in January 2017.

The second goose (424) showed an even clearer population switch in its first tracked year (2016). It migrated from its wintering site in The Netherlands on a southern spring route via the Ukraine and Kazakhstan (Fig. 4b), already close to the Pannonic population’s flyway, which overlaps with the North Sea population’s flyway there. However, when approaching the Arctic coast it crossed the Ural mountains westward like some other North Sea birds (Fig. 1a) approaching its moulting site on Central Taymir via the North Sea flyway, along the Arctic coast, Yamal and Gydan peninsula. We could not detect a nesting attempt that year nor the following. Goose 424 started its first tracked autumn migration along the Arctic coast, stopping long on Bely Island and Northern Yamal. Then it turned south and joined the flyway of the Pannonic population, staying east of the Ural mountains and moving via Kazakhstan to a wintering site in Hungary. So it has switched from the North Sea population to the Pannonic population in 2016 during moult or autumn migration. During its next year’s spring and autumn migration it strictly followed the migration routes of the Pannonic population, indicating that the switch was long-term.

The importance of moult migration

Previous research documented population exchanges in swans and geese during migration [15] or in winter [14], but in our case moult migration of failed or non-breeders was the key phenomenon creating overlap between the two flyways. Such extensive moult migrations are a well-known phenomenon in waterfowl [35–40], but have not been recognized before as a key factor in flyway population exchange. In support of our findings, recently a mixed moult site of two Northern American populations of greater white-fronted geese has been detected by ringing data [41].

During wing moult, waterfowl shed all their flight feathers simultaneously, and cannot fly for some weeks until the flight feathers have regrown. Waterfowl are precocial and nidifugous species, meaning that their young are mobile and leave the nest shortly after hatch, weeks before they can fly. During this period, the young are accompanied by their parents, which also cannot fly at that time as they undergo wing moult soon after breeding. So, only birds that refrain from breeding or that fail their breeding attempt are free to move to moulting areas at flying distance. In our study, moult migrations were up to 2000 km long. It is generally thought that such moult migrations take the birds to areas that are relatively safe [37] and/or offer good feeding opportunities that become available later in the season than the breeding areas [36].

Overlap in space and time

In the case of greater white-fronted geese, overlap outside the moulting phase was rather limited. In fact, the flyways ran much less close to each other than previously thought [29]. The birds from the North Sea and Pannonic population basically flew west and east from the Ural mountains, and this mountain range therefore forms a natural barrier between the two flyways. Both the wintering and breeding ranges are also separated in space. Wintering areas were located in two lowlands separated by mountainous regions including the Alps, and nesting areas were west and east of the foothills of the Ural mountains. However, non-breeders, or birds with failed breeding events, in particular from the North Sea population, fly further to the east, north of the Ural foothills, to moult at Taimyr peninsula, where geese from the Pannonic population breed and moult. As a result, the largest overlap between the two flyways occurs at this peninsula.

Timing is crucial when considering migratory connectivity [42]. In North-America, within a single flyway, three populations of greater white-fronted geese largely overlap in space, but apart from using different breeding areas, differ largely in timing of stopover as well as habitat use within stopovers [43, 44]. On the contrary, greater white-fronted geese from the two populations in Europe followed different routes, but did not differ in timing where they met. However, the birds from the North Sea population gradually follow the advancement of spring on their way to the breeding grounds [32], while those from the Pannonic population stage a long time in Kazakhstan before jumping towards their breeding grounds in the tundra. This jumping over the taiga zone has also been described for swans and geese further to the east in Asia [45, 46], and is probably due to the lack of suitable stopover sites in this vast region.

Flyway switching in social migrants

To some extent the exchange between flyway populations is surprising, as geese are social migrants like swans and cranes, species in which the migration route is learned from more experienced individuals, often the parents [47, 48]. Especially greater white-fronted geese are known for their very long family bonds that can last well beyond the return migration [49–51]. Some juveniles are however splitting from the family already in the first winter (after their first autumn migration), and perhaps earlier (AK, unpublished data). In barnacle geese (Branta leucopsis), it has been suggested that novel migration strategies arose when families disrupted before the onset of migration [52]. In this species, parental care used to last well into spring migration, but this seems no longer the case since migration was delayed because of changing environmental conditions [53].

Male-biased switching?

Waterfowl migration systems are also unique in that they show female-biased nest site fidelity [54]. This might be especially pronounced in (partly) capital breeders, like white-fronted geese [55], where it is most important for the female to accumulate reserves by following the green wave in spring [32]. In another study of barnacle geese, the large majority of individuals switching between breeding populations (within the same flyway) were males [17]. In waterfowl, pair formation is generally believed to take place in the wintering grounds or during spring migration, so males tend to follow females to the breeding grounds [54]. However, ring observations in Greenland white-fronted geese (Anser A. flavirostris) suggests that pair formation occurs mainly during spring or perhaps even summer [56]. The adult males in which we recorded switching of flyways may have lost their mate and were able to move more freely. The fact that they moved to the population of presently higher reproductive success might be adaptive, but could also have been caused by uneven sample sizes (i.e. more available GPS tracks from the North Sea population).

Conservation and management implications

Our findings support the notion that the North Sea and Pannonic populations of greater white-fronted geese should be considered as a meta-population complex. Thus, management actions like increased derogation shooting or hunting in the wintering sites [30] of one population can affect the other. Due to its low breeding success, the North Sea population, even if presently stable at high numbers, should be of concern. As has been shown for e.g. Greenland white-fronted geese [13], it might be a cryptic sink that is kept stable by immigration from other wintering sites, like the Pannonic population or others further east.

Furthermore, the high breeding failure in the North Sea population can lead to higher levels of moult migration and, in light of our results, higher flyway exchange rates. Thus, careful monitoring and a quantification of the exchange with the Pannonic population are needed to fully understand the dynamics of this meta-population. This might become more important in the light of presently high levels of climate and habitat change, as population exchange events can increase or decrease in frequency and magnitude [57] due to weather related events [14, 15], Consequently, population dynamics can change, possibly leading to new migration flyways if switching is not complete [14] or complete population shifts if switches are permanent.