The non-breeding period plays a major role in seabird survival and population dynamics. However, our understanding of the migratory behaviour, moulting and feeding strategies of non-breeding seabirds is still very limited, especially for small-sized species. The present study investigated the post-breeding behaviour of three distant populations (Kerguelen Archipelago, south-eastern Australia, New Zealand) of the common diving petrel (Pelecanoides urinatrix), an abundant, widely distributed zooplanktivorous seabird breeding throughout the southern Atlantic, Indian and Pacific oceans. The timing, geographical destination and activity pattern of birds were quantified through geolocator deployments during the post-breeding migration, while moult pattern of body feathers was investigated using stable isotope analysis. Despite the high energetic cost of flapping flight, all the individuals quickly travelled long distances (> ~2500 km) after the end of the breeding season, targeting oceanic frontal systems. The three populations, however, clearly diverged spatially (migration pathways and destinations), and temporally (timing and duration) in their post-breeding movements, as well as in their period of moult. Philopatry to distantly separated breeding grounds, different breeding phenologies, and distinct post-breeding destinations suggest that the common diving petrel populations have a high potential for isolation, and hence, speciation. These results contribute to improving knowledge of ecological divergence and evolution between populations and inform the challenges of conserving migratory species.

Methods

Study site, study species and animal instrumentation

The study was conducted at three field sites representing separate populations: Ile Mayes (Kerguelen Archipelago, southern Indian Ocean, 49°28’S 69°57’E, hereafter referred as Kerguelen); Kanowna Island (south-eastern Australia, 39°10’S 148°16’E); and Mana Island (New Zealand, 41°06’S 174°46’E). In order to avoid potential bias due to inter-annual variations, individuals from all three colonies were sampled during the same non-breeding period (November 2017 to September 2018). Additional data was collected for Kerguelen (in 2015-16 and 2016-17) and Kanowna Island (in 2016-17 and 2018-19). Kerguelen is located in subantarctic waters sensu lato (between the Polar Front and the Subtropical Front), while both Kanowna and Mana Islands are located farther north, in the temperate subtropical region.

In the present article, migration is defined as the period during which there is an annual to-and-from movement of populations between the breeding site and a migration area [1]. The post-breeding period is defined as the period between the last burrow attendance at the end of the breeding season, and the first return in a burrow the following season. The non-breeding period is the period between two successive breeding cycles. It includes the post-breeding migration away from breeding sites, followed by a more or less prolonged period in the vicinity of breeding sites before initiating the next breeding season.

To study the at-sea distribution of individuals from the three populations during the non-breeding period, adult birds were equipped with leg-mounted GLS (Migrate technology, model C65). The mass of the device attached to a plastic or metal ring with a cable tie was < 1.5 g, corresponding to on average 1.07 ± 0.1% of body mass (117 g – 175 g). Previous use of such devices on small seabird species have shown limited impact on the feeding ecology or future breeding [27, 41]. Breeding individuals were equipped at the end of the breeding season and were recaptured during the following breeding season for removal of the device. Sex was determined by DNA analysis of a small blood sample (0.1 mL) collected from the brachial vein for the individuals from Kerguelen (Laboratoire Analyses Biologiques, CEBC, France), and of a body feather for those from Kanowna Island (DNA solutions, Australia). The sex of individuals at Mana Island was determined from their sex-specific calls [5, 42].

Information on moult pattern in Pelecanoididae species is very limited, largely because individuals can be at sea during this stage [5]. Similarly to alcids, the majority of diving petrel species seem to undergo an annual almost synchronous wing molt of flight feathers during the non-breeding period [44, 45]. For CDP, in south-eastern Australia and New Zealand, adults are thought to start wing moult at sea after chicks have fledged [5]. In South Georgia, Kerguelen and Heard Island, moult of primaries seems to be initiated just before the end of the chick-rearing period [5, 43], although sources are discordant [46]. However, the accuracy of these records is complicated by the lack of information on the breeding status of the observed individuals. In the present study, the timing of wing moult was inferred from information provided by the GLS on the time spent on the sea surface. Since flight feather renewal directly affects flying ability [45, 47, 48], in particular for species with high wing loading (body mass / wing area) such as diving petrels [49], a peak in time spent on the water is likely to reflect the wing moult of flight feathers.

For body feathers, it is likely to be a protracted process initiated during the post-breeding period, similarly to other species of small petrels [50, 51]. The continuous moult of body feathers throughout the non-breeding period allows the birds to progressively renew their plumage while maintaining its vital role in thermoregulation [52]. Stable isotope ratios in body feathers can be a proxy of the location and trophic level of the individuals when they were synthetised [19, 20]. Therefore, to investigate the isotopic niche and the moult of body feathers of individuals throughout the non-breeding period, four contour feathers were collected from the middle and lower back of each tracked individual at recapture. Additional samples were collected on adult individuals that were not tracked in order to increase the sample size and to assess temporal variation (three years of data for both Kerguelen and Kanowna). Handling time at deployment (measurements, banding and GLS attachment) and recapture (device removal, weighing, blood and feather sampling) were minimized and took on average < 5 min.

Data processing and analyses

The GLS measured light intensity every minute and recorded the maximum value for each 5 min interval. The determination of twilights (dawn and dusk), linked to a time base, enabled longitude (timing of local midday and midnight) and latitude (duration of day and night) to be estimated, providing two positions per day with an average of 186 ± 114 km (mean error ± SD; [53]). Processing and calculations were conducted using the GeoLight package in the R statistical environment [54, 55]. As latitude estimations around the equinoxes are unreliable, data for two weeks before and after the equinoxes were excluded before spatial analysis was conducted [56]. Additionally, latitude or longitude estimates that were clearly inaccurate (unrealistic speed > 1500 km/day, trajectory or spikes) were removed [31]. When outward and inward movements were occurring during the equinox periods (March-April and September-October), the timing of arrival and departure from the breeding colonies was determined from longitudinal directional movements using the raw data [31, 48].

Filtered locations were used to generate kernel utilization distribution (UD) estimates with a smoothing parameter (h) of 1.8 (corresponding to a search radius of ~ 200 km) and 1º x 1º grid cell size (based on the mean accuracy of the device). For each population and year, the 50% (core foraging area) and 95% (home range) kernel UD contours were obtained [57]. The core area was used to estimate the centroid position (mean latitude and longitude) during the post-breeding period for each individual. Spatial analyses were performed using the adehabitatHR R package [58].

The period of maximum proportion of time spent on the water (wet-dry sensor being wet > 90% per day; [35]) was used to determine the moulting period of flight feathers for each individual [48]. This was recorded differently at the three study sites. For Kerguelen and Kanowna Island, wet-dry data were sampled every 30 s with the number of samples wet and maximum conductivity recorded every 4 h. At Mana Island, wet-dry data were sampled every 30 s with number of wet samples and maximum conductivity recorded every 10 min.

The dates of last and first burrow attendance were determined by combining information on activity (wet-dry: 100% dry for a period > 4 h), temperature (for Kerguelen and Kanowna Island only, sampled every 5 min with maximum recorded every 4 h) and movement data (presence of the bird in the vicinity of the breeding region). These data were then used to estimate the duration and the total distance travelled during the post-breeding migration.

For stable isotope analyses, feathers were cleaned of surface lipids and contaminants using a 2:1 chloroform:methanol solution in a ultrasonic bath, followed by two successive methanol rinses and air dried 24 h at 50°C. Each feather was then cut to produce a fine powder for homogenization before carbon and nitrogen isotope ratio determination using a continuous flow mass spectrometer (Delta V Plus or Delta V Advantage both with a Conflo IV interface, Thermo Scientific, Bremen, Germany) coupled to an elemental analyser (Flash 2000 or Flash EA 1112, Thermo Scientific, Milan, Italy) at the LIENSs laboratory (La Rochelle Université, France). Stable isotope values were expressed in conventional notation (δX = [(R_sample/R_standard) – 1]) where X is ¹³C or ¹⁵N and R represent the corresponding ratio ¹⁵N/¹⁴N or ¹³C/¹²C. R_standard values were based on Vienna Pee Dee Belemnite for ¹³C, and atmospheric nitrogen (N₂) for ¹⁵N. Replicates of internal laboratory standards (Caffeine USGS-61 and USGS-62) indicate measurement errors < 0.10 ‰ for δ¹³C and 0.15 ‰ for δ¹⁵N.

As the Southern Ocean is marked by a latitudinal gradient of δ¹³C and δ¹⁵N (Jaeger et al. 2010), low (δ¹³C < -19.5 ‰ and δ¹⁵N < 9.9 ‰) and high isotopic values (δ¹³C > -19.5 ‰ and δ¹⁵N > 9.9 ‰) in body feathers were interpreted as corresponding to feathers moulted in Antarctic/Subantarctic and Subtropical/neritic waters, respectively [36, 59].

All spatial and statistical analyses were conducted in the R statistical environment 3.6.1 [55]. To investigate among-population, inter-annual and sex-related variations on post-breeding migration parameters (departure and return dates, migration duration, total distance travelled and maximum range), general linear models (GLM) were fitted using the lme4 package [60]. For all models a Gaussian family was selected (error structure approached the normal distribution), all combinations of variables were then tested and ranked based on their Akaike’s Information Criterion (AIC), and the global models were checked to ensure normality and homoscedasticity of the residuals [61] before further statistical tests. Inter-population differences were quantified using analyses of variance (ANOVA or Welch’s ANOVA), and post-hoc tests were conducted using t-tests (parametric), or Kruskal-Wallis and Mann-Whitney U tests (non-parametric) depending on the data distributions. Before these analyses, data were checked for normality (Shapiro-Wilk test) and equality of variances (Levene test). Estimates are presented as means ± standard error (unless specified).