<b>Data from: </b><b>Experimentally increased food availability allows for earlier departure in a long-distance migratory shorebird</b>
收藏资源简介:
Study population Red knots of the subspecies islandica spend the non-breeding period in north-west Europe, mostly in the Wadden Sea (Buehler & Piersma 2008; Piersma et al. 2005). In late March red knots undergo a prenuptial moult where they replace their body contour feathers, going from winter into summer plumage (Buehler & Piersma 2008) (Figure 1). From mid-April onwards birds start to store energy, gaining ~80g (2/3 of their lean body mass) in just 2 – 3 weeks’ time (Piersma et al. 1996, 2005) (Figure 1). Between early and late May birds depart from the Wadden Sea towards Iceland or northern Norway (Wilson et al. 2011) (Figure 1), from where they fly to their breeding grounds in Greenland and north-east Canada (Davidson & Wilson 1992). We conducted experiments in the spring seasons of 2021 – 2023, using red knots captured in the Wadden Sea. The experimental aim was to manipulate time access to food during the period of fuelling body stores and spring prenuptial moult, after which the birds were tagged and released in late April to study the timing of their migration departure from the Wadden Sea. In the first year (2021), the experiment was carried out as a pilot study where we kept individuals in captivity up to late May and did not measure timing of migration departure. Bird capture and housing The 97 red knots used in this study were captured in mist nets at the island of Griend (53.15°N, 5.16°E) in the Wadden Sea in autumn and winter (September - January) 2019, 2021 and 2023 (Table S1). Upon capture, we took biometric measurements of birds, extracted a small blood sample (< 75 μL) for molecular sexing (Van Der Velde et al. 2017) and aged birds as either “1st calendar-year”, “2nd calendar-year” or “adult” (older than 2nd calendar-year). We only selected adult birds for the experiment as younger birds over-summer in their non-breeding range rather than migrate (Martínez‐Curci et al. 2020). Birds were housed in outdoor aviaries (7 – 8 birds per aviary) at the NIOZ Royal Netherlands Institute for Sea Research on the island of Texel, following the methods described in (Buehler & Piersma 2008; Vézina et al. 2009), see supplemental material for details. Treatments The experiments started in winter / early spring after capture, except for 2021, where birds had already been in captivity since October 2019 (Table S1). Birds were randomly divided in three treatment groups, which differed in the time that food was accessible: 6, 12 or 24 hours per day. Per treatment, we used either 1 aviary (2021) or 2 aviaries (2022 and 2023). Birds and treatments were assigned randomly to each aviary. The 12-hour treatment represents the foraging time that red knots experience outside in their natural habitat (Bulla et al. 2017; Piersma et al. 1994). The 6- and 24-hour treatments can therefore be considered as constrained and increased accessibility to food respectively. Food accessibility was controlled by using automated cat feeders (Cat Mate C300, closerpets.eu) that open and close at set times. To avoid overlap between feeding times and biweekly bird measurements, food in the 6- and 12-hours treatments was accessible between 00:00 and 06:00 and 18:00 – 06:00 (times in UTC), respectively. Like most shorebirds, red knots forage during daylight as well as dark conditions (Thomas et al. 2006; Van Gils & Piersma 1999). To reduce competition between birds we placed three feeding units in every aviary. Treatments lasted until release (late May in 2021, late April in 2022 and 2023, Table S1). Measurements in captivity During the experiment birds were captured from aviaries twice a week (Tuesdays and Fridays) for measurements. In addition, we measured the birds on 29 April, the last day of the experiment before birds were released in 2022 and 2023. We measured body mass by weighing birds on an electronic scale with 0.1g precision. Plumage status was scored as the fraction of summer plumage (red-brown feathers) relative to winter plumage (grey feathers) on both belly and back, and scored in percentage categories of 0%, ~5%, ~25% ~ 50% ~ 75%, ~95% and 100% summer plumage. Body mass and plumage change To determine the onset and rate of body mass deposition and plumage moult, we compiled trajectories of body mass (Figure S1, S2) and plumage status for every individual bird over the course of the experiment. From these trajectories we determined the start and end of body mass deposition and plumage status increase, as well as the rates of change in this period. To determine the start of body mass deposition a breakpoint analysis was conducted on body mass trajectories from 1 March up to the end of the experiment using the function ‘selgmented’ in the r-package ‘segmented’ (Muggeo 2024), which compares models with different numbers of breakpoints (with a maximum of 3). We chose the number of breakpoints in the best-performing model and then determined the start of body mass deposition as the first breakpoint after which the slope coefficient was greater than 0.5 g/day. When none of the slope coefficients were greater than 0.5 g/day, we did not determine a start of body mass increase (27 out of 94 birds). These were typically birds did not initiate body mass increase and kept a low body mass throughout the experiment. The maximum rate of body mass deposition was determined as the strongest slope coefficient between the breakpoints. The start of plumage change was determined as the first measurement date at which plumage status reached 25% summer plumage or higher. For one bird which started the experiment with 25% summer plumage we did not determine this. We determined the end of prenuptial moult as as the first date on which the maximum plumage status was reached. The rate of plumage status increase was determined as the difference between the maximum and minimum fraction of summer plumage divided by the days between the start and end date. The rate of change of birds for which no start or end date could be determined, as they did not increase their plumage status, was set to 0 (47 out of 93 birds). For every bird, we also extracted the body mass (divided by length of the tarsus to correct for variation in body size) and plumage status measured on the day closest to the release day, i.e. 29 or 30 April. Individual plots of mass trajectories can be found in the supplements (Figure S1, S2). Departure timing On 29 April 2022 and 2023 we released respectively 27 and 32 birds in the Wadden Sea (from the north-east side of the island Texel, 52.12N°, 04.90E°). Each bird was equipped with two radio transmitters: (1) a WATLAS transmitter (4.4 g), which emits signals to a network of receiver stations to determine tracks by reverse-GPS, currently operating in the western Wadden Sea, (Bijleveld et al. 2022); (2) a Lotek NTQB2-3-2 VHF nanotag (0.7 g) which can be picked up by the MOTUS network (Taylor et al. 2017), which covers most of the Wadden Sea except the central and eastern Dutch Wadden Sea (Figure 2a) where birds can be detected in flight up to 10 - 15 km (Taylor et al. 2017). WATLAS transmitters were attached to the skin of the birds’ rump with cyanoacrylate glue after cutting away the vanes of feathers in a small circle, the same size as the transmitter. Thereafter MOTUS transmitters were glued on top of the WATLAS transmitter. A MOTUS receiver was present at the release location (see above) to ensure tags were picked up in this network. Birds were released in flocks of 9 – 11 birds, each flock consisting of a mix of birds from every treatment. We determined date of departure from the Wadden Sea using both tracks received within the WATLAS network (Figure 2b) and detections received from the MOTUS network in the Wadden Sea region (Figure 2a). To improve data quality, both WATLAS and MOTUS data were filtered and smoothed (see supplemental materials). For the analyses, we only included birds that had left the Wadden Sea before 15 June, as we considered later departures might no longer represent spring migration flights but rather movement between summer staging areas. Of the 59 released birds, 44 birds (19 in 2022 and 25 in 2023) yielded recorded signals that show departure movements out of the Wadden Sea. When birds were last detected by the WATLAS network on a northward movement out of the Wadden Sea (often moving northward over the Wadden Sea, Figure 2b) we set the date of this movement as the departure date. When birds were last detected in the WATLAS network moving in another direction (mostly towards the eastern Wadden Sea), we checked the MOTUS network for detections outside of the WATLAS network. For tags that were last detected by a MOTUS receiver, we visually inspected the series of bursts received on the last day of detection. We only included detections that showed a typical pattern of a ‘departure event’, with a peak in signal strength followed by a gradual decrease in signal strength (Müller et al. 2018). For one bird (Z101007) we excluded the last series of 4 consecutive bursts because they did not correspond to the typical departure pattern. Instead, we used the previous series of detections (that did show this pattern) as the departure event. Further details on birds for which we could not determine departure events, as well as individual plots and maps of MOTUS detections and WATLAS tracks can be found in the supplementary materials (Figure S3 and S4). Statistical analyses We tested how treatment affected (i) body mass deposition and (ii) plumage moult in captivity, and how (iii) treatment affected date of departure from the Wadden Sea, as well as how (iv) departure was affected by body mass deposition and plumage moult. (i, ii) We used generalized linear models (GLMs) to test the effects of treatment on (1) the start and (2) rate of body mass deposition, (3) final body mass (reached on 29/30 April), as well as start (4) and rate (5) of plumage status change of each individual, and (6) final plumage status (on 29/30 April). As explanatory variables we included treatment as ordered factor (6 < 12 < 24 hours of food accessible), sex as factor (female / male) and year as factor (2021, 2022 and 2023). For models 1, 3 and 4 we fitted GLMs with Gaussian errors. For models 2 and 5 we fitted Tweedie compound Poisson GLMs with a log link, and for models 6 we fitted Beta regression GLMs. We explored the data following the protocol described in Zuur et al. (Zuur et al. 2010). We checked for outliers using Cook’s distance test in the ‘check_outliers’ function in the package ‘performance’ (Lüdecke et al. 2019). We explored collinearity by testing whether variance inflation factors (estimated from function ‘vif’ in the ‘car’ R-package (Fox et al. 2024)) were below 3.0. Models were validated by visual inspection of the square root of the absolute standardized residuals plotted over the fitted values, and fitted over co-variates within the model (treatment, sex and year) and co-variates not in the model (cage number). Within each model we tested for significance of linear and quadratic trends of treatment, as well as significance of sex and year, using the t-statistic, or Wald’s z-statistic for the Beta regression GLM. (iii) We used Cox regression models to test how date of departure from the Wadden Sea (expressed as days between release and departure) was affected by either treatment or individual body mass deposition and the extent of plumage development at release. We included data for 44 individuals with departure dates as well as 3 individuals that did not depart from the Wadden Sea, censoring the individuals from the latter group. We explored the data as described above. First, we tested the effects of treatment, including treatment as ordered factor, sex as factor and year as factor as explanatory variables. Within the model we tested for significance of linear and quadratic trends of treatment, as well as significance of sex and year using the Wald’s z-statistic. Due to high variability of individual onset of mass increase, not all individuals started mass increase during the experiment. To explore how this affected our results, we also repeated the above analysis using a dataset of only the individuals that did start mass increase while in captivity (n = 30). (iv) Thereafter we tested the effects of body mass deposition and extent of plumage development on departure date using mixed effect Cox regression models. We included data from the 47 individuals as described above, after which we omitted 3 individuals for which the start of body mass increase could not be determined. We included (centred and scaled) start and rate of body mass deposition, final body mass, rate of plumage status change, final plumage status, sex and year as explanatory variables. We did not include start of plumage status change as more than halve of the individuals did not show plumage changes. As rates and final scores for body mass and plumage status were correlated (variance inflation factors > 3.0) we did not include rates and final scores in the same model. As these models included parameters affected by the experimental treatment rather than the treatment itself, we used AICc model selection to determine the best model, which was the model with the lowest AICc, or the model with the least amount of parameters within 2 ΔAICc of this model (Arnold, 2010). A list of models is included in Table S2.



