Low impact of weather effects
Surprisingly, our findings provided no evidence for weather as a migration trigger in curlews, in contrast to various other bird species [12, 24, 31–32]. However, our data confirmed that flight speed increased with increasing TWC, in accordance with previous studies [6, 29]. Based on the clear benefit of faster flight speeds during tailwind conditions, this suggested that curlews would mainly select days with suitable tailwind conditions for their departure from (and arrival at) their wintering grounds in the Wadden Sea. However, in contrast to our assumption, the current results revealed virtually no effect of wind or other weather parameters on the departure/arrival day, even when the actual weather during departure was compared with the mean weather conditions 4 days and/or 4 years prior to departure. Although curlews rarely departed during hours with precipitation (in accordance with other bird species), this predictor had no significant influence in our final model. The current findings were therefore not in agreement with previous studies, which found that the probability of departure of curlews from a pre-breeding staging site was reduced during precipitation . However, our findings were in accord with observations of departing curlews in China , and of the closely related whimbrel (Numenius phaeopus islandicus), which also demonstrated no significant influence of wind force or wind direction [53–54].
Although weather parameters had no impact on departure/arrival decisions, we found a significant negative correlation between TWC and flight altitude. Unfortunately, there were no available data for wind speed (and direction) for different air layers for this study; nevertheless, the pattern clearly suggested that curlews tried to find more favourable wind conditions at higher altitudes if they encountered headwinds at lower altitudes. In temperate latitudes, the prevailing westerly wind conditions in the higher air layers suggest wind assistance when ascending . This behaviour has previously been recorded for nocturnal songbird migrants [22, 33], as well as for diurnal long-distance migrants using radar techniques .  found intensive songbird migration in air layers up to 3 km altitude in temperate regions, when the birds encountered headwind conditions close to the surface. The authors demonstrated that migrating birds benefited from the wind conditions in higher air layers by ascending, exclusively during their spring migration. This might explain why there was no significant relationship between flight altitude and TWC in arriving curlews during their autumn migration. The current results clearly suggest that curlews depart (and stay) at lower altitudes when wind conditions close to the surface are beneficial, and use higher air layers during spring migration when they encounter headwinds.
Interestingly, curlews arrived at significantly slower flight speeds and lower altitudes compared with departing birds. Meteorological reasons for this can be excluded, given that the wind conditions and TWC were similar for departing and arriving individuals. One likely reason is that the linear distance to the nearest stop-over in arriving curlews was far smaller than for departing individuals, which might explain why departing curlews ascended to higher altitudes and had faster flight speeds compared with arriving birds.
Timing of migration according to resource availability
Our data clearly supported the concept that the onset of migration respected resource availability in the breeding grounds, given that distance to the breeding site was the single (highly significant) predictor affecting the day of departure in our LASSO analyses. Although no field data on food resources within the breeding sites were available for this study, previous studies demonstrated that arthropod densities in Arctic and sub-Arctic breeding grounds peaked shortly after snowmelt, resulting in higher chick growth rates if birds started nesting early [14–15]. This suggests that curlews should aim to arrive at their breeding sites as soon as the areas are free from snow and ice. This could in turn explain why curlews that breed further from their wintering grounds (e.g. in the eastern parts of Russia in this study) might delay their migration to allow the snow and ice in their Arctic and sub-Arctic breeding grounds to melt, and/or to ensure that they encounter optimal arthropod densities. Similarly, a previous study  showed that colour-ringed curlews breeding in Fennoscandia departed later than birds breeding further west. Although this study dealt with a different sub-population, the results were in agreement with the patterns found in the current study.
We found that curlews wintering in the Wadden Sea departed within a very short time window (i.e. mostly between mid-April and mid-May). This contrasted with birds wintering in south-west Britain, which had already started to depart during February and March . However, in contrast to British curlews that breed in north-western Europe , all but one of the curlews in the present study bred in Russia, i.e. much further east. The more condensed departure window in our study might thus reflect the relatively late availability of breeding sites due to snow and ice melt. This emphasises the involvement of an internal clock to ensure their timely departure (see discussion on repeatability and genetic trigger below).
Temporal patterns of migration start
The main window of arrival of birds in the Wadden Sea was late June to mid-July, which is about 2 weeks later than reported for birds breeding in central or northern Europe . Although desertion of offspring by females is common in shorebirds and has been shown for curlews , we found no sex differences in departure patterns, and only a non-significant tendency for females to arrive earlier (in contrast to our hypothesis). The reason for this is unclear. It is possible that some birds failed to rear chicks successfully, leading to the earlier arrival of at least some males.
Departures of songbirds usually occur at night and around sunset [17, 25] and the same is true for many long-distance migratory wader species, given that birds are thought to take advantage of favourable atmospheric factors during the night and to calibrate their orientation systems before they start in the evening hours [57–58]. Our results support these patterns, with more curlews departing during the early evening or early nighttime. However, an earlier study of curlews departing from a final pre-breeding stop-over site in China showed high variability in terms of the time of day for departures . The reason for these different findings remains unclear. Given the higher number of individuals arriving during morning and afternoon, our results suggest the existence of a strong temporal trigger regulating departure decisions, but a weaker such trigger for arrivals.
Correlations with first and last stop-overs
We expected that the departure day and TWC would be significantly related to the distance to the nearest stop-over site and the stop-over duration; however, no such correlation was found. Curlews did not stage for shorter periods if they departed later, nor did they stage for longer if they encountered headwind conditions during the first part of their migration to allow more time for re-fuelling. This finding is in accordance with studies of songbirds, which likewise showed no or only weak relationships [59–60]. Similar results were also found for whimbrels, with no impact of wind conditions on stop-over patterns; however, in contrast to our study, they tended to skip a potential stop-over when they departed later .
High repeatability in departure patterns suggest genetic triggers
Studies on repeated individual-based migration patterns across consecutive years are scarce. To the best of our knowledge, the current study is among the first to record high-resolution GPS-movement data in birds across consecutive years (see  for a study on black-tailed godwits (Limosa limosa limosa) equipped with geolocators). We found a high degree of repeatability in curlew departure and arrival dates for the same individuals across subsequent years. Given the absence of clear relationships between departure decision and wind/weather parameters, this supports the concept of an internal genetic trigger [see also reviews in 16–18], which seemed to play the most important role in departure and arrival decisions in our model species
A recent review  presented evidence for genetic control of the timing of bird migration. However, the authors also found considerable individual variation in this genetic programme, as a result of interactions with environmental and social factors, and individual learning. Given the highly repeatable, conservative time pattern and lack of any relationships between departure date and weather parameters for our model species, the current results suggested that such intraspecific variation in the genetic programme may be very low. Our results thus provide robust support for the concept of an internal clock, responsible for timing bird migration [16–17]. In contrast to these results however, black-tailed godwits tracked across subsequent years exhibited a much broader window of inter- and intra-individual timing of migration . The authors suggest that this broad window of repeatability indicated weak selective forces with respect to migration timing. This assumption would suggest that such forces are strong for curlews, although for the specific mechanisms affecting the high inter- and intra-individual repeatability of migration timing evidence is currently lacking. In contrast,  hypothesized a conservative annual-cycle strategy in long-distance migrants that was thought to minimise risks and reduce carry-over effects. Our study may support this hypothesis, given that curlews clearly showed constant, conservative patterns in terms of their departure decisions, independent of the weather conditions at their wintering sites. These findings are in line with other Arctic breeding birds, such as bar-tailed godwits (Limosa limosa baueri) from New Zealand, which also showed high repeatability .