We conducted our study in Álftanes (southwest Iceland, 64° 06′N, 22° 00′W), a main spring staging site for light-bellied brent geese of the East Canadian High Arctic population (ECHA). We carried this study out during the brent goose spring migration staging (01/05–27/05/2017), a critical period of intense foraging, where birds accumulate energy stores for migratory flight and breeding (Inger et al., 2008, 2010). In late April, the population has a staggered migration from wintering grounds along the Irish coasts to staging sites in west Iceland (Inger et al., 2008). They typically resume migration en masse approximately one month later to their breeding grounds in the Canadian High Arctic (Gudmundsson et al., 1995; Clausen et al., 2003; Robinson et al., 2004). The ECHA population amounts to approximately 35000 individuals of which over 10% are uniquely marked with coloured Darvic leg rings displaying individual alphanumeric combinations (Cleasby et al., 2017). We conducted this study on marked individuals only and all observations in this study were made from inside of a vehicle, through an Opticron HR 80 GA ED telescope with an SDLv2 40936 zoom eyepiece (×54 zoom) and Meopta Meo Star B1 12 × 50 HD binoculars, causing minimal disturbance to geese. We observed birds at opportunistically selected field sites around Álftanes from 6.00 h to 22.00 h.

The age of geese (juvenile/adult) was visually determined in the field by observation of plumage colouring along the edges of wing coverts, clearly distinguishing juveniles from adults (Reed, 1993). We identified relationships between individuals [partners/family with parent(s) and offspring/single adults/single juveniles] by observing proximity and cohesion during birds’ movements, which was found to represent a reliable method of identifying first-order relatedness among individuals (Harrison et al., 2010). We recorded flock size and considered individuals at least 20 goose lengths apart as a different flock.

We opportunistically re-sighted marked birds and recorded fat deposition on foraging geese by visually determining abdominal profile indices (API; Owen, 1981; Boyd et al., 1998; Clausen et al., 2003; Madsen and Klaassen, 2006) on a seven-point scale (1–7; 1 being lowest and 7 the highest score; Inger et al., 2008) with accuracy of 0.5. The API score is linearly related to body mass and highly correlated with the size of abdominal fat reserves (Fox et al., 1999; Madsen and Klaassen, 2006). It is commonly used on geese and shorebirds as they are known to predominantly store energy in their abdominal cavity (Owen, 1981; Wiersma and Piersma, 1995; Boyd, 2000; Tinkler et al., 2007). We used APIs as an indication of body condition and weight gain. We have evidence from this population and other closely related taxa that it correlates with body mass and subsequent reproductive success (Ebbinge and Spaans, 1995; Tinkler et al., 2007; Inger et al., 2008; Harrison et al., 2011; Harrison et al., 2013).

We measured foraging behaviour of individual geese through focal sampling of peck and pace rates. Identifiable geese were initially randomly selected then resampled each week of the staging period. We classified pecks as movement of the beak toward the ground, thereby touching the substrate, and paces as every step taken by an individual (Inger et al., 2006a). We used tally counters to record the number of pecks and paces over 1–2 min and obtained a rate by dividing the number of pecks/paces by the total length of observation. We did not include observations in which birds simultaneously did not peck nor pace in this study. Additionally, we recorded length and occurrence of sitting behaviour when geese sat on the ground during a focal observation.

We used faecal sample collection of marked geese for stable isotope analyses. Given brent geese feed on saltmarsh (marine) and terrestrial grasses, which are isotopically distinct with δ¹⁵N values being lower in terrestrial grasses and δ¹³C values being considerably higher in marine grasses, we classified all field sites accordingly accounting for the resulting difference in carbon and nitrogen stable isotope ratios in faecal samples (Inger et al., 2006b). For collection of faecal samples, we split the field season into three time periods: early (May 1–10) reflecting arrival and the onset of hyperphagia, mid (May 11–19) reflecting a full hyperphagia and late (May 20–27) reflecting the cessation of hyperphagia and preparation for migration. We chose these time periods in response to the feeding rate of geese measured across the spring, i.e., initial increase of feeding rates and API (onset of hyperphagia), reaching a peak toward the end of the staging season (full hyperphagia) and a subsequent drop of feeding rates and API until departure (cessation of hyperphagia). Such a mass trajectory and feeding rate pattern has also been recorded in other Branta populations during spring staging, for example in dark-bellied brent geese (Branta bernicla bernicla) on their last spring staging site before embarking on a long-distance flight to their Arctic breeding grounds (Dokter et al., 2018). In the early period, we collected samples from a range of identifiable individuals and then resampled from these target individuals in each subsequent time period, to track individual variation throughout staging. We collected faecal samples from known individuals with one person observing the targetted goose until it defecated and a second collecting the sample by following directions of the observer through two-way radios. Faecal samples were stored at −20°C in conventional plastic-zip lock bags until analyses.

We subsampled faecal samples along a central horizontal plane of the cylindrical pellet. We freeze-dried samples, ground them into a homogenous powder, and weighed them to 0.7 mg (±0.1 mg) in tin capsules. Stable carbon and nitrogen isotope measurements were determined by a Thermoquest EA1110 elemental analyser interfaced with a Europa Scientific 2020 isotope ratio mass spectrometer (Elemtex Ltd., Cornwall, United Kingdom). Stable isotope ratios were expressed as δ notation, measured in parts per thousand (‰) relative to international standards Pee Dee Belemnite for carbon and air for nitrogen according to the following equation: δX = [(R_sample/R_standard) – 1] × 1000 where X is ¹⁵N or ¹³C and R is the corresponding ratio of ¹⁵N/¹⁴N or ¹³C/¹²C. We report stable isotope ratios as δ¹³C values for carbon and δ¹⁵N values for nitrogen and data are corrected for sample mass and instrument drift. Measurement precision (standard deviation), based on within-run replicate measures of the laboratory standard, was better than 0.25‰ for δ¹⁵N and 0.21‰ for δ¹³C isotopic values.

Past studies have shown that long-distance migrants will time their departure to maximise wind support (Grönroos et al., 2012). Using the package “RNCEP” (Kemp et al., 2012) in R, we interpolated the 10m above ground U wind (east/west) and V wind (north/south) components (m/s) for a central point on the Álftanes peninsula (64° 06′N, 22° 00′W) using the NCEP/NCAR dataset (Kalnay et al., 1996), provided by the NOAA/OAR/ESRL PSD, Boulder, Colorado, United States¹. Measurements were interpolated on the hour each day between 15 May (15 days prior to departure) and 5 June 2017 (5 days post departure) inclusive.

To model U and V wind as a function of wind support and crosswind we estimated the birds flight direction (heading) on migratory departure using the “geosphere” package (Hijmans et al., 2017). By studying the migratory tracks obtained from birds in this population, it is known that light bellied brent geese fly west north west toward Greenland, therefore a reference point on the southeast coast of Greenland was used to calculate heading (65° 42′N, 37° 42′W). Wind support and crosswind were calculated using methods described by Safi et al. (2013) (Figure 1). Positive values associated with wind support represented a tail wind and negative values a head wind. Crosswind was taken as an absolute value irrespective of direction (Safi et al., 2013).

We conducted all data analyses using R software version 3.3.1. (R Core Team, 2016). We determined the sex of marked individuals as described in Bodey et al. (2020) by using a discriminant function based on morphological differences between wing length and head-bill of male and female individuals, recorded during catches when darvic rings are attached. All birds that did not fit male or female morphological criteria were classified as “unknown” and were excluded from statistical analyses.

The linear plateau model indicated that individuals in this study population reached a plateau in weight gain measured as an abdominal profile index (1–7), of approximately 6 between May 24 and 25 (n = 187 observations, 187 individuals, critical day value = 24.5, 95% CIs: 20.8–28.2, plateau y = 5.7, p < 0.001; Figure 1). Furthermore, a significant positive linear correlation between API and the number of days into staging was observed until May 24 (n = 172 observations, 172 individuals, r² = 0.70; est. = 0.15, SE = 0.01, p < 0.001) but no such correlation was observed thereafter (n = 13 observations, 13 individuals, r² = 0.41; est. = 0.05, SE = 0.22, p = 0.81). However, after May 24 female geese had higher API compared to males (n = 13 observations, 13 individuals, r² = 0.41; est. = 22120.96, SE = 0.33, p = 0.02). A sigmoid function modelling API by day shows both males and females experience the inflection point in weight gain 15.3 days prior to departure (n = 187 observations, 187 individuals; r² = 0.80, est. = 15.31, SE = 0.71, p < 0.001; Figure 1).

The top ranking peck rate model, selected by lowest AICc value, included Julian date, the quadratic term of Julian date and resource type as independent variables (AICc = 1757.8, AICc null model = 1759.2; Table 1; model selection table in Supplementary Material 1). As might be expected given the trajectories, only the quadratic effect of Julian date had a statistically significant negative influence on peck rates during staging (n = 154 observations, 73 individuals, est. = −0.133, SE = 0.054, p = 0.013), meaning that peck rates of geese initially increased in the earlier part of the staging period and then decreased again at the end of staging (Figure 2).

The best performing pace rate MAM only included Julian date as an independent variable but only differed marginally from the null model (AICc = 1,357.2, AICc null model = 1,357.5; see Supplementary Material 2 for AICc model selection table). Julian date did not influence pace rates statistically significantly (n = 154 observations, 73 individuals, est. = −0.089, SE = 0.058, p = 0.121). Yet, we observed that sitting behaviour differed statistically significantly between the four weeks of staging (F₁,₃ = 6.727, p < 0.001) and a Tukey post hoc test revealed that sitting behaviour was higher in the last week of staging compared to the previous three weeks (week 1 vs. week 4: p < 0.001; week 2 vs. week 4: p = 0.044; week 3 vs. week 4: p = 0.004).

The best performing MAM for δ¹³C included habitat type as response variable whereas the best performing MAM for δ¹⁵N included habitat type as well as individual API. For C:N the best MAM included API and habitat type as well as their interaction as response variables. All three stable isotope measures showed a significant difference between samples collected on terrestrial or marine habitats, specifically δ¹³C and δ¹⁵N were both higher on marine sites (n = 130 observations, 59 individuals, δ¹³C: r² = 0.35; est. = −1.42, SE = 0.17, p < 0.001; δ¹⁵N: r² = 0.29; est. = −1.05, SE = 0.36, p = 0.004). We found δ¹⁵N was statistically significantly higher in birds with higher API (n = 130 observations, 59 individuals, δ¹⁵N: r² = 0.29; est. = 0.33, SE = 0.11, p = 0.004, Figure 3C) and lower C: N ratios were found in faecal samples of individuals with a higher API (n = 130 observations, 59 individuals, r² = 0.28, est. = −1.72, SE = 0.76, p = 0.02, Figure 3A). However, the results show no significant relationship between an individual’s API and habitat type they are feeding on (n = 130 observations, 59 = individuals; r² = 0.83, est. = 0.24, SE = 0.14, p = 0.08). Using a Tukey post hoc test we also found that individuals with higher API than the population average had significantly different δ¹⁵N ratios in their faeces compared to those with lower than average API (p = 0.03, Figure 4C). Both individuals with above and below average API were feeding on similar resources with no significant difference in the δ¹³C values (p = 0.99, Figure 4B), there was also no difference in the C: N ratio in faeces (p = 0.16, Figure 4A).

Geese experienced an average wind support of +1.23 m/s between 15 May and 5 June, indicating a tail wind (Figure 5A) and a crosswind of 5.21 m/s (Figure 5B). On the evening before departure the geese experienced a strong head wind (−6.78 m/s) but this dropped at midnight to −1.11 m/s and became a light tail wind (0.17 m/s) at 06:00 30/05/21. There was an above-average crosswind (6.89–7.51 m/s) present between midnight and 11:00 on the day of departure. Favourable wind support (i.e., tail wind) and crosswind conditions suitable for migratory departure were frequently present between 24 May when API attenuates and 30 May when the geese depart en masse and in the days following (Figure 5).

Throughout the month of spring staging in Iceland, light-bellied brent geese significantly changed foraging rates and locomotion behaviour in relation to the time of staging. Initially geese arrived with low API, however despite this, we observed low foraging rates. This is often explained by the “gut limitation hypothesis,” whereby an individual experiences slow regeneration of digestive efficiency after a period of fasting (McWilliams and Karasov, 2005). The approximately 1,500 km non-stop migratory flight from Northern Ireland induces a period of fasting likely to cause atrophy of disused organs such as the digestive system. Regeneration of organs requires time and energy, hence an initial slow foraging period upon arrival in Iceland (McWilliams and Karasov, 2005). This has been observed in lesser snow geese, Chen caerulescens caerulesces where lower foraging rates were associated with individuals carrying atrophied digestive organs (Ankney and MacInnes, 1978). Furthermore, this hypothesis was also supported in passerine experiments, showing low food intake rates after a period of fasting even when food was offered ad libitum (Klaassen and Biebach, 1994; Hume and Biebach, 1996; Klaassen et al., 1997; Karasov and Pinshow, 2000; Karasov and McWilliams, 2005).

We found the time from arrival was the strongest predictor of variation in foraging rates during spring staging. After the initial slow start geese entered a state of hyperphagia, excessively eating in an effort to store enough energy to fuel the continuation of migration and initiation of breeding upon arrival in the High Arctic. As abdominal fat deposits increased so did the foraging rate but both foraging rates and fat gain decreased in the last period of staging, likely due to individuals approaching the target mass required to fuel migration and breeding. Additionally, it becomes more energetically advantageous to maintain a threshold weight as opposed to continue increasing bodyweight (Lindström, 2003), so in this last period before departure fatter individuals were observed sitting more. On average males had a lower threshold weight than females, potentially the result of mate guarding behaviour and a difference in energetic demand experienced by the sexes during long-distance migration and a capital breeding strategy. The reproductive success of female brent geese can change from 8 to 65% with an increase in their body mass at the breeding grounds (Inger et al., 2008). Light-bellied brent geese from the adjacent Svalbard/Greenland population are known to lay eggs within a week of arriving (Clausen et al., 2003), highlighting the importance of accumulating large fat deposits prior to departing the spring staging site.

The delay in mass departure despite attenuation of weight gain by both sexes suggests that there are other physiological limitations prolonging the staging period in Iceland. In response to fasting and in-flight starvation waterfowl and wader migrants have been shown to experience protein turnover within their organs (Bauchinger and McWilliams, 2010). Lesser snow geese, a similar-sized waterfowl were found to atrophy organs during a 5–6 day period of decreased foraging rates (Ankney and MacInnes, 1978). This is also commonly seen in migratory wader species such as red knots, Calidris canutus found to have a pre-departure turnover of protein, with the digestive organs shrinking and flight muscles (i.e., pectoral) increasing in size (Piersma et al., 1999). This phenotypic flexibility is a potential explanation for the behavioural changes observed in the brent geese during their pre-departure week of spring staging.

The stable isotope component of this study in conjunction with the behavioural observations provides further evidence of phenotypic flexibility occurring in the light-bellied brent geese during spring staging in Iceland. We found significant changes in the C:N ratio and δ¹⁵N in the faeces of brent geese, in response to reaching peak abdominal fat. Both are indicators of a higher rate of protein turnover that would be expected in animals undergoing fasting and tissue-specific reorganisation, in this instance in preparation for the non-stop migratory flight to their breeding grounds in the Canadian High Arctic.

In comparison to other capital breeding migrants brent geese are physiologically constrained by a large energy requirement to body size ratio (Buchsbaum et al., 1986; Battley and Piersma, 2005). Their solely herbivorous diet has low nutrients and is high in structural components, with cellulose in plant cell walls making it hard to breakdown and digest (Buchsbaum et al., 1986). Therefore, they require a large gizzard and large small intestine to efficiently acquire nutrients and accumulate crucial energy deposits, constituting ∼95% lipids and ∼5% proteins, needed to fuel migration (Sedinger et al., 1989; Jenni and Jenni-Eiermann, 1998; Klasing, 1998; Battley and Piersma, 2005; Rozenfeld and Sheremetyev, 2016). Female geese in particular have a high energetic demand requiring significant fuel accumulation during the month of spring staging to rapidly initiate breeding on arrival and successfully breed. However, a large digestive system takes up crucial space within the abdomen and is metabolically costly during flight and capital breeding. Digestive organs constitute approximately 10% of a bird’s lean mass making it a significant burden on wing load (Pierce and McWilliams, 2004). The measured decrease in C:N ratio in brent goose faeces associated with heavier individuals as the staging month progressed is a likely result of a physiological trade-off occurring. As adipose fat is accumulated in the abdomen we can infer that the digestive organs are compromised to maximise available space (Battley and Piersma, 2005). Waterfowl are known to compensate for the low nutrient concentrations in their herbivorous diet by having a long small intestine improving nutrient absorption and also recycle excretory and endogenous nitrogen in the caecum, improving overall nitrogen balance (Sedinger, 1997; Barboza et al., 2008, p. 156). Given this, a compromised digestive system and resulting reduction in organ size will likely cause a reduction in assimilation efficiency causing the excretion of dietary nitrogen (Barton and Houston, 1993; Sedinger, 1997).

The results also show a significant increase in δ¹⁵N of faecal samples as the month progresses and individuals accumulate abdominal fat stores. A higher δ¹⁵N is usually indicative of an animal feeding at a higher trophic level (Bearhop et al., 2000; Inger and Bearhop, 2008). However, brent geese are strictly herbivorous and we did not observe a change in their resource use. A study by Fuller et al. (2005) confirmed that δ¹⁵N values were not only influenced by dietary intake but also fluctuations in the body’s nitrogen homeostasis, suggesting that the increase in δ¹⁵N in the goose faeces might have an internal physiological explanation. Migratory birds are known to build, dispense and restore components of their body throughout their annual life history stages, strategically enhancing fuel storage and reducing flight costs (Fox and Kahlert, 2005). Empirical evidence from terminal studies on long-distance migratory water birds demonstrates pre-departure atrophy with up to 50% reduction in digestive tissues occurring at the staging site (Piersma et al., 1993; Jehl, 1997; Piersma and Lindström, 1997; Piersma and Gill, 1998).

The reorganisation of internal organs prior to migratory departure enables a bird to reallocate proteins away from the expensive disused digestive system into the synthesis of flight muscles (McLandress and Raveling, 1981; Jehl et al., 2015). When an animal’s body is in a stable feeding state, assimilation of dietary protein tends to result in light ¹⁴N being preferentially excreted, resulting in an increase in body tissue δ¹⁵N, relative to the diet (Bearhop et al., 2000; Inger and Bearhop, 2008; Krausman and Cain, 2013). However, the enrichment of ¹⁵N seen in the faeces of heavier brent geese is commonly found during periods of nutritional stress when higher rates of protein catabolism can often occur (Hobson et al., 1993). In this case, we have inferred that the increase in δ¹⁵N represents a controlled reallocation of proteins in preparation for migratory departure.

An alternative explanation for the increase in δ¹⁵N in the faeces is that the geese are entering a state of starvation in response to the low food intake, initiating the catabolism of proteins as a source of energy. However, this is unlikely as birds and mammals are known to prioritise lipid utilisation until fat stores drop below a critical limit (Jenni et al., 2000). For migratory bird species, this was shown to be as low as 5–10% of fat remaining and only then do they shift to catabolising protein to meet energy requirements (Cherel et al., 1988; Schwilch et al., 2002). Given that abdominal fat stores remain constant once a threshold weight is reached, it is unlikely that the protein catabolism measured in this study using stable isotope techniques is the result of foraging rates decreasing (i.e., starvation), rather shows phenotypic flexibility in the migratory light-bellied brent geese.

Despite female brent geese carrying a higher energetic burden upon reaching the breeding grounds, the sexes showed no difference between the δ¹⁵N values and C:N ratio in their faeces. This highlights the significant cost of carrying potentially redundant organs and the importance of pre-departure protein catabolism and synthesis in an effort to reduce wing load and improve flight efficiency.