Does wintering north or south of the Sahara correlate with timing and breeding performance in black‐tailed godwits?

Abstract Migrating long distances requires time and energy, and may interact with an individual's performance during breeding. These seasonal interactions in migratory animals are best described in populations with disjunct nonbreeding distributions. The black‐tailed godwit (Limosa limosa limosa), which breeds in agricultural grasslands in Western Europe, has such a disjunct nonbreeding distribution: The majority spend the nonbreeding season in West Africa, while a growing number winters north of the Sahara on the Iberian Peninsula. To test whether crossing the Sahara has an effect on breeding season phenology and reproductive parameters, we examined differences in the timing of arrival, breeding habitat quality, lay date, egg volume, and daily nest survival among godwits (154 females and 157 males), individually marked in a breeding area in the Netherlands for which wintering destination was known on the basis of resightings. We also examined whether individual repeatability in arrival date differed between birds wintering north or south of the Sahara. Contrary to expectation, godwits wintering south of the Sahara arrived two days earlier and initiated their clutch six days earlier than godwits wintering north of the Sahara. Arrival date was equally repeatable for both groups, and egg volume larger in birds wintering north of the Sahara. Despite these differences, we found no association between wintering location and the quality of breeding habitat or nest survival. This suggests that the crossing of an important ecological barrier and doubling of the migration distance, twice a year, do not have clear negative reproductive consequences for some long‐distance migrants.


| INTRODUCTION
Seasonal migration allows animals to exploit areas that are inhospitable at other times of the year (Alerstam, Hedenström, & Åkesson, 2003).
Although seasonal migration is widespread throughout the animal kingdom, it is especially well developed in birds (Newton, 2010). In shorebirds, seasonal migration is the rule rather than the exception (Piersma, 2003(Piersma, , 2007, but the distances travelled vary greatly between species. Some species migrate as far as 12,000 km one way (e.g., bar-tailed godwit Limosa laponica baueri: Battley et al., 2012) and cross ecological barriers such as oceans, mountain ranges, and deserts, while others remain close to the breeding grounds, even if that means remaining at sub-Arctic latitudes (e.g., rock sandpiper Calidris ptilocnemis ptilocnemis: Ruthrauff, Dekinga, Gill, & Piersma, 2013). Not only is there considerable among-species variation in migratory distances, there is also variation among populations within shorebird species (Reneerkens et al., 2009), and even among individuals within the same population Hötker, 2002).
Migrating long distances requires time and energy that may entail trade-offs with reproductive events, that is, costs are carried over from one season into the other (Alves et al., 2012;Harrison, Blount, Inger, Norris, & Bearhop, 2011;Klaassen, Lindström, Meltofte, & Piersma, 2001;Norris & Marra, 2007;Senner, Conklin, & Piersma, 2015). This effect may be enhanced when individuals need to cross ecological barriers, because they lack the opportunity to replenish energy during migration (Newton, 2010). Besides avoiding energy costs associated with large displacements, remaining close to the breeding grounds during the nonbreeding season may be advantageous as proximity should enable individuals to better time their arrival and breeding in accordance with local weather conditions (Both, Bouwhuis, Lessells, & Visser, 2006). For instance, in warm springs pied avocets (Recurvirostra avosetta) which wintered further away from the breeding grounds arrived later than individuals which wintered nearby (Hötker, 2002). Mistimed arrivals, which is nowadays mainly referring to arriving too late due to global warming (Knudsen et al., 2011), in turn, may compromise the timing of clutch initiation and, consequently, reproductive success .
Because breeding performance can be compared between individuals breeding in the same area but with different migration distances (Ketterson & Nolan, 1983), populations with disjunct nonbreeding ranges allow descriptive tests of possible trade-offs between migration distance and reproduction. In Eurasian spoonbills (Platalea leucorodia leucorodia) breeding in the Netherlands, individuals wintering south of the Sahara have a lower survival probability, breed later, and recruit fewer offspring than birds which remain in Europe (Lok, Overdijk, Tinbergen, & Piersma, 2011;Lok, 2013), suggesting that wintering further away from the breeding grounds and crossing an ecological barrier may be associated with increased fitness costs. Conversely, Icelandic black-tailed godwits (Limosa limosa islandica) wintering in Portugal not only arrive at their breeding sites in Iceland earlier than individuals wintering almost 2,000 km further north in Ireland and England, but also have a higher reproductive success and survival (Alves et al., 2012(Alves et al., , 2013. This emphasizes that the fitness costs of migration may be context dependent, for example, whether migration involves crossing an important ecological barrier such as the Sahara Desert (Lok, Overdijk, & Piersma, 2015), alternatively, the quality of the winter site may influence breeding performance.
The great majority of Continental black-tailed godwits (Limosa limosa limosa; hereafter, "godwits") breed in the Netherlands (Kentie et al., 2016) and spend the nonbreeding season either south of the Sahara-predominantly in the West African countries of Senegal and Guinea-Bissau -or north of the Sahara in the Iberian Peninsula (Márquez-Ferrando, Hooijmeijer, Groen, Piersma, & Figuerola, 2011; see also Figure 1). Although the godwit population has declined by over 75% in the past half-century (Gill et al., 2007;Kentie et al., 2016), the number of godwits wintering in southern Spain has grown especially in Doñana Wetlands, Spain; the percentage of the flyway population wintering here increased from 4% in the late 1980s to 23% in 2011 (Márquez-Ferrando, Figuerola, Hooijmeijer, & Piersma, 2014). This increase has been suggested to reflect an increase in the availability of suitable foraging habitat in the form of man-made fish ponds and irrigated rice fields in the region, at the same time that degradation of wetlands in West Africa has increased (Zwarts, Bijlsma, van Kamp, & Wymenga, 2009). As southern Spain is ~3,000 km closer to Dutch breeding areas and does not necessitate a potentially arduous migration across the Sahara Desert (Klaassen et al., 2014;Lok et al., 2011), it is possible that godwits remaining north of the Sahara may experience higher fitness due to an earlier arrival on the breeding grounds and lower energetic expenditure during migration.
F I G U R E 1 Locations of resightings of color-marked black-tailed godwits south of the Sahara in West Africa, September-March (dark gray), and north of the Sahara on the Iberian Peninsula in October (light gray). The Dutch breeding study site is marked with a large gray circle

0°N
Here we examine whether wintering in distinct regions is related to variation in reproductive parameters in a Dutch breeding godwit population. We examine whether wintering north of the Sahara is advantageous in terms of differences in arrival and laying date, habitat choice, and reproductive parameters. Moreover, we analyzed whether godwits wintering north or south of the Sahara differ in the repeatability of their phenology, exploring whether godwits wintering closer to their breeding area may be more flexible in their timing of arrival.
Observational studies linking events during the breeding season with outspoken differences in migration strategies are still thin on the ground (but see Alves et al., 2013;Lok et al., 2011Lok et al., , 2015 and will help determine whether the costs and a different length of migration influences population dynamics in a declining long-distance migratory bird.

| Study sites
Since 2004, we monitored godwits breeding in a region of dairy farming in southwest Friesland, the Netherlands (52°55′N, 5°5′E; Kentie, Both, Hooijmeijer, & Piersma, 2015), an area that traditionally held high densities of breeding godwits (Mulder, 1972 and was expanded again in 2012, with the addition of another adjacent 1,500 ha. Using measures of the occurrence of foot drains, high field water levels, and herb richness , we categorized all 2,811 fields in the study area as one of two categories reflecting the intensity of agricultural use. "Herb-rich meadows" represent traditionally managed meadows, often managed especially for birds by postponing mowing until after the incubation period, with the help of agricultural subsidies or as meadow bird reserves, whereas "agricultural monocultures" are industrialized grasslands and arable fields harvested several times a year. We have shown previously that godwits prefer herb-rich meadows over monocultures (Kentie, Both, Hooijmeijer, & Piersma, 2014) and that godwits breeding in herb-rich meadows have higher nesting success and chick survival than godwits breeding in monocultures (Kentie, Hooijmeijer, Trimbos, Groen, & Piersma, 2013;Kentie et al., 2015). On this basis, we consider herb-rich meadows to be the higher-quality breeding habitat.
We have systematically monitored godwit flocks in Extremadura and the Tejo and Sado estuaries for color-marked godwits in late winter since 2006 (Lourenço, Kentie, et al., 2010;Masero et al., 2010). In November 2010, we also began surveying Doñana and the surrounding region for color-marked godwits; since 2011, the resighting intensity has increased to a minimum of biweekly surveys each winter from October-March

| Defining an individual's wintering region
Data from 82 satellite transmitters and three geolocation tracking devices mounted on adult godwits both in a Spanish spring staging area (see Senner, Verhoeven, Abad-Gómez, et al., 2015) and on the breeding grounds (unpubl. data T. Piersma, stored at www.movebank. org and viewable at volg.keningfanegreide.nl) revealed that adult godwits were faithful to wintering location at the regional level. We thus adult birds. We therefore only used sightings of godwits caught in the Netherlands as adults, or, for birds marked as chicks, only after they were older than 3 years (see Table 1 for sample sizes).
T A B L E 1 Sample sizes of total observations (Total N) and of uniquely color-marked individual black-tailed godwits (Ind) wintering north and south of the Sahara used in the analyses. The sample of arrival dates is reported for females and males; nest success is reported for nests located in low-and high-quality habitats; observations of breeding habitat choice excluding duplicate observations of individuals that did not change their habitat  (2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015), we systematically searched the breeding study area and surroundings daily for color-marked godwits, from the arrival of the first adults in early March until the beginning of the laying phase in April. Resighting probability was high, with on average 3 ± 1 days between resightings (Senner, Verhoeven, Abad-Gómez, et al., 2015). We used the date of the first observation of an individual at the breeding area in a given year as its arrival date. We confined our analyses to individuals first seen before 30 April, as our resighting effort decreased after this point in order to focus on monitoring nests. During the laying and incubation phases between April and June, we surveyed the study area for finding nests by walking side-by-side through meadows to locate nests during the laying phase or, later, by flushing incubating adults from nests. When a nest was found, we measured each egg's width and length to the nearest mm. Godwits, like other shorebirds, usually have a clutch size of four eggs (Arnold, 1999). We used the egg flotation method (van Paassen, Veldman, & Beintema, 1984) to age the nest and predict its laying and hatching date (25 days after laying the first egg).
Egg volume was estimated with the formula: length × width 2 × 0.524 (from Romanoff & Romanoff, 1949). We identified the parents of the nest by reading color rings with a spotting scope, or with camera traps.
We considered a nest successful if at least one chick was found in the nest, or, as chicks leave the nest within 24 hr (Lind, 1961), if we found an indication of successful hatching such as broken eggs without blood or yolk and membranes clearly visible, and tiny egg fragments in the bottom of the nest. A nest was considered unsuccessful if we found all eggs abandoned, egg remains without membranes, egg remains with yolk or blood, or an empty nest without any egg remains.
To prevent nest abandonment, and because the birds are easier to capture close to hatch, parents were caught at the end of the incubation stage. Adults were either caught in a walk-in-trap, with a mist net placed over the nest, or occasionally by hand from the nest. Each individual was uniquely marked with four plastic color rings, a colored flag, and numbered metal ring. We measured bills and tarsus length to the nearest 0.1 mm, and head and wing length, and the length of the tarsus plus mid-toe without nail to the nearest mm. Prefledging chicks of >10 days were considered large enough to wear a color ring combination, although few were captured because of the difficulty locating them. We also collected a blood sample (30 μl from the brachial or tarsal vein) from each individual for molecular sexing (see Trimbos et al., 2011); when a blood sample was lacking, we used morphological measurements (see Schroeder et al., 2008) to sex the remaining 11% of the birds. This work was done under license numbers 4339E and 6350A following the Dutch Animal Welfare Act Articles 9 and 11.

| Data analyses
To examine whether arrival date at the breeding area differed between individuals wintering north or south of the Sahara, we used a linear mixed-effect model (LMM) with a normal error structure.
We included sex and its interaction with wintering region as factors, because males normally arrive earlier than females (Lourenço et al., 2011). Individual and year were included as random terms. Although resighting probability of arriving godwits is lower than one, we do not expect differences in resighting probability between the two groups.
We then examined whether wintering north or south of the Sahara correlated with the probability to breed in high-quality breeding habitat. Although we have observations for 13% of the individuals from multiple years, a generalized linear mixed-effect model (GLMM) with year and individual as a random effect did not converge because too few individuals changed breeding habitats between years. To overcome this problem, we used a generalized linear model (GLM) with the proportion of breeding attempts in good quality areas per individual over the years (with the function cbind()). We used a binomial error structure and a logit link function.
Next, we determined whether lay dates of females differed between individuals from the two wintering regions with a LMM with a normal error structure and individual and year as random terms.
Godwits can lay a second clutch when the first failed , and we used the lay date of first nest if we knew of other attempts. However, for the majority of nests we were unable to correct for repeat nesting attempts because, in most cases, a pair's full nesting history was unknown. If nest success were to differ between groups, the group with the lowest nest success would likely have more second nesting attempts, potentially biasing our estimates of lay dates if we missed a large number of first attempts. However, there was no effect of wintering region on nest success (see Section 3).
We then tested whether the average egg volume of a clutch differed between females from different wintering regions with a LMM with a normal error structure and individual and year as random terms.
We included lay date in the model because a previous study found an effect of lay date (in some years) on egg volume (Lourenço et al., 2011), and included an interaction between lay date and wintering region. Because female size may influence egg size, we checked with an ANOVA whether females wintering north or south of the Sahara differed in structural body size. Structural female size was described by the first component of a principal component analysis in which we added the biometric variables (bill length, total head length, wing length, tarsus length, tarsus-to-toe length). This principal component explained 51% of the variation.
We calculated the individual repeatability for arrival date by fitting a LMM with individual as random term. Repeatability is the betweenindividual variation divided by the sum of between-individual variance and residual variance (see Nakagawa & Schielzeth, 2010). We did this separately for each wintering region. We then compared the 95% confidence intervals of the repeatability of both wintering regions.
We also tested for an effect of wintering region on daily nest survival probability in a mark-recapture framework, because it accounts for nests not found before they were lost (Dinsmore, White, & Knopf, 2002). In the model in which we estimated the effect of wintering region on nest success, we always included year due to known strong year effects , and tested for an effect of sex, and the interaction between sex and wintering region.
We knew the wintering location of both adults of 29 of 532 nests with identified adults (belonging to 16 pairs; godwits are monogamous, see Kentie et al. (2014)). If the wintering locations of both adults from a nest were known, we randomly chose one of the adults for inclusion in our analyses. Of these 16 pairs, in six cases one partner wintered south while the other wintered north of the Sahara.
We used Program R (v. 3.1.1; R Core Team 2014) with the packages "lme4" (Bates, Maechler, Bolker, & Walker, 2014) for LMM, and "RMark" (Laake, 2013) for daily nest survival analysis, and "rptR" using 10,000 iterations to estimate repeatability (Schielzeth, Stoffel, & Nakagawa, 2016). Arrival date and lay date were log-transformed after visually checking residual plots. For the LMMs, we calculated the significance of the fixed and random effects, and the 95% confidence intervals for parameters with a parametric bootstrap (1,000 iterations).
Fixed covariates were excluded from the fullest model in a stepwise backward elimination if they failed to explain variation (p > .05).
Models of daily nest survival were compared with the second-order AIC (Akaike information criterion) for small samples (AICc) (Burnham & Anderson, 2002).

| RESULTS
Male and female godwits were found in equal proportions to winter north or south of the Sahara (χ 2 = 0.277, df = 1, p = .60, N = 495, CI), respectively. These differences in arrival date were significant, and males arrived significantly earlier than females (Figure 2a, Table 2a).
The interaction between wintering region and sex was not significant.
Females that wintered south of the Sahara on average initiated clutches on 20 April, which was 5 days (3-8 days, 95% CI) earlier than females wintering north of the Sahara (Figure 2b, Table 2b). Although early laying females generally tend to lay larger eggs (Schroeder et al., 2012), in our study females wintering south of the Sahara laid smaller eggs: Egg volume differed by 3% between wintering regions, with females wintering south of the Sahara laying eggs of 40.1 cm 3 (39.4-40.9 95% CI) and females wintering north of the Sahara laying eggs of 41.3 cm 3 (40.7-41.9 95% CI; Figure 2c, Table 2c). The interaction between lay date and wintering region had no significant effect on egg volume. Females wintering north or south of the Sahara did not differ in structural body size (F 1,187 = 0.038, p > .5).
On average, an individual had a 71% (67-75 95% CI) chance to be found in herb-rich meadows, and wintering region and sex had no effect on this probability (Table 2d). Daily nest survival rates varied among years, ranging between 0.981 in 2015 and 0.998 in 2013.
However, daily nest survival did not differ for birds from the two wintering regions (Table 3; ∆AICc = 1.9, but with one additional parameter not competitive (Arnold, 2010)), or among sexes from the two regions either (∆AICc = 0.6, but again not competitive with one additional parameter). This analysis only involved nests for which an adult with known wintering region was identified, biasing the daily nest survival toward higher probabilities as we could not always immediately identify the attending adults when the nest was found, thus potentially excluding predated/abandoned nests before adult identification was possible.

| DISCUSSION
Like several other European migrant birds (e.g., Flack et al., 2016;Lok et al., 2011;Reneerkens et al., 2009), Continental black-tailed godwits have a disjunct wintering area and can either be found north or south of the Sahara Desert. Godwits migrating to the southern wintering region cross the Sahara Desert twice and a return trip covers a total (beeline) distance of ~10,000 km, while godwits remaining at the northern location travel less than half the distance: ~4,000 km. We predicted that, because of their longer flight and the lack of stopover sites in the Sahara, godwits wintering furthest south would arrive later because they would wintering north of the Sahara. The eggs of godwits wintering north of the Sahara, however, were 3% larger than eggs of godwits wintering south of the Sahara, which could not be explained by body size differences; females wintering north and south of the Sahara were of similar structural size. Although southern wintering godwits arrived earlier, we found no relationship between wintering location and habitat quality of the acquired territory, or with daily nest survival probability.
Previous studies of other long-distance migratory birds with similarly disjunct populations have found a positive correlation between migration distance and the timing of spring arrival at the breeding grounds, and timing of clutch initiation (i.e., Bregnballe, Frederiksen, & Gregersen, 2006;Lok, 2013;Woodworth et al., 2016). We were surprised to find that godwits wintering south of the Sahara arrived and laid their clutches earlier than godwits wintering north of the Sahara.
However, a recent study involving the closely related Icelandic blacktailed godwit found that individual arrival dates were primarily related to breeding and wintering habitat quality, which resulted in individuals wintering furthest away arriving earliest in Iceland (Alves et al., 2012;Gunnarsson et al., 2006). In combination, these studies suggest that the effect of migration on phenology and reproduction is species and context dependent and should not a priori be assumed to be greater for long-distance migrants. Reference level for wintering region is "north of Sahara" and "female" for sex. Variables removed from the final model are in italics. Arrival and lay dates were log-transformed, and breeding habitat and nest success are on a logit scale. Random effect estimates refer to standard deviations.  Lourenço et al., 2011;Schroeder, Hooijmeijer, Both, & Piersma, 2006), we found no indication that birds wintering in West Africa had higher nest success. On the contrary, females wintering north of the Sahara laid larger eggs than females wintering south of the Sahara. In general, larger eggs hatch chicks of better quality (Krist, 2011) and, in godwits, egg volume is strongly correlated with hatchling mass, which is in turn correlated with chick survival (Hegyi & Sasvari, 1998;Schroeder et al., 2012). This might suggest a trade-off between early arrival, or a longer migration, and maternal investment in eggs. Unfortunately, our study was unable to examine the potential longer-term consequences of these differences because the sample size of recruits was too low for sufficient statistical power. Thus, it remains to be tested whether laying earlier and a difference of 3% in egg size represents a significant biological trade-off, and whether it is enough to affect chick condition or survival.
Repeatability in timing of migratory movements is generally high among long-distance migrants (Conklin, Battley, & Potter, 2013;Gill et al., 2014;Senner, Hochachka, Fox, & Afanasyev, 2014). We predicted that, as in pied avocets (Hötker, 2002), godwits wintering closer to the breeding grounds would vary more strongly in their arrival and therefore lay dates, as they would be better able to respond to changing local conditions at the breeding grounds. However, we found that arrival dates were similarly repeatable among individuals for individuals wintering north and south of the Sahara, and comparable with a previous study on repeatability in arrival date of this study population (Lourenço et al., 2011;between 0.18 and 0.29). The repeatability of arrival dates is likely to be underestimated, as not all individuals would have been seen immediately upon arrival. This repeatability in arrival dates in both groups does not suggest differences in the ability to predict conditions on the breeding grounds (see Winkler et al., 2014 for a discussion of this cognitive process). This is corroborated by the fact that during a spring with extremely cold temperatures, godwits departed from their Iberian staging areas at the same dates as in other years (Senner, Verhoeven, Abad-Gómez, et al., 2015). Nonetheless, all individuals were flexible enough to respond in such a way that reproductive success was not affected (Senner, Verhoeven, Abad-Gómez, et al., 2015).
There is thus no clear evidence to suggest that the recent and rapid growth of the Iberian nonbreeding population (Márquez-Ferrando et al., 2014) is associated with differential reproductive success related to choice of wintering sites or migration distance.
Perhaps the increasing proportion of birds remaining north of the Sahara is better explained by higher survival, as the mortality rate of godwits equipped with satellite tags was highest during the crossing of the Sahara Desert (unplubl. data). In any case, the costs of migration and consequences of choosing among different wintering sites for godwits, and other species, require further study to fully understand and predict the ecological responses of long-distance migrants to environmental changes (Webster, Marra, Haig, Bensch, & Holmes, 2002). This is especially the case in a system in which some individuals cross a barrier such as the Sahara, which may increase their mortality rate in years with more hazardous wind and weather conditions (Klaassen et al., 2014;Lok et al., 2015), but provide them with potential benefits (e.g., early arrival) in other years. Ideally, future studies will incorporate site-specific seasonal survival estimates (see Piersma et al., 2016;Rakhimberdiev, van den Hout, Brugge, Spaans, & Piersma, 2015) and consider habitat quality at nonbreeding sites.
Such work will help to elucidate how migration influences patterns of demography, biogeography, and the diversity of migratory strategies used by birds.