New Zealand, a hotspot for winter-breeding seabirds
One of the 25 world-wide biodiversity hotspots is located in New Zealand (Myers et al. 2000). Originally part of the Gondwana supercontinent, Australia and New Zealand have been separated by the opening of the Tasman Sea (~ 80 Ma years ago, Molnar et al. 1975). Since then, collisional tectonism, crustal thickening and volcanic activity have progressively lifted the land mass and shaped this archipelago (Hicks & Campbell 2012). The surrounding marine environment evolved as well when Tasmania separated from Antarctica (40 ± 5 Ma ago), creating a cold water flow coming from the South Indian Ocean when seabirds already existed (Fordyce & Jones 1990, Nelson & Cooke 2001). Subsequently, the Drake Passage opening between Antarctica and South America (30 ± 5 Ma ago) strengthened the presence of oceanic fronts with the Antarctic Circumpolar Current (ACC, Deacon 1982, Orsi et al. 1995, Rintoul et al. 2001). Lately, the Quaternary glacial – interglacial cycles have affected the ice cover on land, the sea level, the current flow and the front’s positions (Darvill et al. 2016).
Currently, the New Zealand archipelago ranges between 29 – 52° S and 166 – 174° E, at the interface of the ACC and the South Pacific gyre. Its surrounding marine realm is characterised by zonal belts of warm, cool and cold surface waters, separated by oceanic fronts marking sharp changes in vertical water structure, temperature, salinity and nutrients: the subtropical front at 43 – 45° S, the subantarctic front at 50° S and the Polar front at 60° S (Heath 1985). These fronts and the ACC interact with the submarine relief (Butler et al. 1992) along their way on the great ocean conveyor belt (Broecker 1991). The archipelago, settled by humans with the consecutive arrivals of Polynesian (1280) and European migrants (1769), underwent drastic landscape modification and invasive mammal introductions that caused species extinctions (Wilmshurst et al. 2008).
Despite these losses, New Zealand remains a biodiversity hotspot with high endemism for both terrestrial and marine biodiversity. Indeed, its waters host 56 marine mammal taxa out of the 120 species listed worldwide (Schipper et al. 2008, Baker et al. 2010), and the largest seabird assemblage worldwide with 78 species, including 33 endemic (Croxall et al. 2012). Whereas winter-breeding occurs in the entire southern ocean, the winter-breeding species richness differs geographically: 12 species occur in the southern Atlantic sector, against 8 species in the southern Indian sector, and 21 in the southern Pacific sector. New Zealand archipelago stands out for hosting the greatest concentration (17 species), where the highest density occurs on the New Zealand South Island with 10 species.
During winter, most species migrated out the New Zealand waters by trans-equatorial migrations towards the North Pacific (Shaffer et al. 2006), latitudinal migrations towards South America (Deppe 2012, Rayner et al. 2012) or in Southern Ocean frontal zones (NIWA 2016). These migrations result in a drop of breeding seabirds numbers from > 43 million individuals in summer to < 3 million in winter (NZ Bird online 2019). How these winter-breeding species are able to sustain the cost of reproduction during the apparent challenging winter, while others species migrate to survive, is poorly understood. Learning about such fundamental biological processes is crucial to understanding the evolutionary costs and benefits of winter-breeding. Although the nesting sites of winter-breeding species are scattered on New Zealand in coastal areas with phylogeographic partitioning (Rawlence et al. 2014), at sea, several reports indicate some overlap of their foraging range. Particularly, the waters off the west coast of New Zealand’s South Island are reported to be used by the highest concentration of winter-breeding species. In this area, five species are reported to breed and forage: the Westland petrel Procellaria westlandica (Freeman et al. 1997, Waugh et al. 2018), the Little penguin Eudyptula minor (Heber et al. 2008), the Fiordland penguin Eudyptes pachyrhynchus (Warham 1974), the southern Buller’s albatross Thalassarche bulleri bulleri (Sagar & Weimerskirch 1996, Stahl & Sagar 2000a, Stahl & Sagar 2000b) and the Spotted shag Phalacrocorax punctatus (Heather & Robertson 2000). Diet studies indicate that these species all feed on fish, cephalopod and crustacean with common prey taxa (Van Heezik 1989, Freeman 1998, James & Stahl 2000, Flemming et al. 2013).
Study site and field procedures
The study was conducted on Taumaka / Open Bay Island (43.859 °S, 168.885 °E), 4.5 km off the west coast of New Zealand’s South Island ( Figure 2-1). This 20 ha island hosts one of the largest populations of Fiordland penguins with at least 150 breeding pairs (McLean & Russ 1991). Individuals were sampled (once throughout the study) over different breeding stages (incubation, guard and crèche) and over two breeding seasons (2016 and 2017). Individuals were observed from hides as they returned to the island and captured by hand as they walked to, or at, the nest. Once captured, they were fitted with a cloth hood to minimise stress and weighed using a spring scale (± 20 g, Pesola Ltd, Switzerland). To determine at-sea movements and marine habitat use, the birds were equipped with a GPS 21 data logger (i-gotU Mobile Action Technology, Inc., Taiwan) accurate to ± 10 m (Morris & Conner 2017). The device was removed from its original casing to reduce drag effect and sealed in a waterproof heat-shrink tubing (Tyco Electronics, Switzerland). During incubation, model GT 600 (45x39x13mm, 26.5 g) units were deployed, programmed to sample at 2 min intervals. During guard and crèche stages, model GT-120 (45x25x12mm, 15 g) units were deployed, programmed to sample at 1 and 2 min intervals, respectively. The data loggers were attached to the mid-line dorsal feathers on the lower back with black waterproof tape (TESA 4651, Beiersdorf AG, Germany) following Wilson et al. (1997).
To obtain information on diving behaviour, same individuals were simultaneously instrumented during the guard and crèche stages with a dive behaviour/accelerometer data logger (40x15x11 mm, 6.5 g Axy-Depth, Technosmart, Rome, Italy, or 21x13x4 mm, 1.7 g WACU, MIBE/IPHC, Strasbourg, France) recording depth (± 5 cm) and temperature (± 0.1 °C) every second and tri-axial body acceleration at 25 Hz. In total, attached devices represented ca 1 % of the penguins body mass in air and ca 2.6 % of their cross-sectional surface area and, therefore, are likely to have had negligible impact on the foraging behaviour (Wilson et al. 1986, Agnew et al. 2013). Handling procedures lasted 10-15 min before animals were released near the nest to resume normal behaviours.
After one or two foraging trips to sea, individuals were recaptured and the data loggers were removed. Culmen length and bill depth were measured using Vernier callipers (± 0.1 mm) to sex the birds using a discriminant function previously established at this site (Murie et al. 1991). A 0.1 mL blood sample was collected by venipuncture of a tarsal vein and stored in 70 % ethanol for later analysis of stable isotopes for trophic niche estimation.
Data handling and processing
The GPS data were processed within the R statistical environment (R Core Team, 2017). Foraging trips were defined as the time spent at sea between the departure and the return (land-based points removed). A speed filter with a threshold at 8.2 km·h-1, corresponding to the upper 22 range speed of the similar-sized Macaroni penguin Eudyptes chrysolophus (Brown 1987) was applied to remove erroneous locations. The following trip metrics were calculated using the package adehabitatHR (Calenge 2006): the trip duration, total horizontal distance travelled, the mean horizontal speed, the maximum distance from the colony; and compass bearing to the most distal point.
Raw depth data were downloaded from the data loggers and converted using the software Axy Manager (Technosmart Europe S.r.l., Rome, Italy). Submergences of greater than 1 m depth were considered as dives and were processed within the R software to calculate their parameters (maximum depth, duration, descent and ascent rates, bottom time, post-dive interval) using the package diveMove (Luque, 2007). Maximum depth of successive dives were compared to assess the Intra Depth Zone (IDZ; Tremblay & Cherel 2000). For each individual trip, the proportion of time spent diving and the dive frequency were determined. For individuals providing sufficient data, the behavioural aerobic dive limit (bADL) was estimated from the dive duration and the post-dive interval. As 98 % of post-dive intervals were ≤ 60 s (and those lasting longer were not related to dive sequences), this threshold was used to investigate the bADL. The bADL was determined from the intersection of quantile regressions fitted on a moving average of 5 successive dives (supplementary Fig S 2-1), to account for oxygen payoff delay (Horning 2012). Lastly, dive data were visually inspected to ascertain dive shape profiles (Wilson et al. 1996). Dive occurrence was investigated in relation to local sunrise, sunset and nautical twilight times (www.timeanddate.com, retrieved for Haast, 10 km east of the colony) in order to describe their temporal distribution (dawn, day, dusk and night). The foraging effort was assessed from the vertical dive rate (total vertical distance/trip duration, m·h-1).
The acceleration data along the three axis (surge, heave and sway) were analysed within IGOR Pro software (Wavemetrics Inc., USA, 2000, Version 7) to infer behaviour at each second of the trip. The gravity-related static acceleration, determined by a one second running 23 mean, was subtracted from the raw acceleration (Wilson et al. 2006, Shepard et al. 2008) to calculate Vectorial Dynamic Body Acceleration (VeDBA = √( 2 + 2 + 2)), a proxy of whole body activity (Qasem et al. 2012). Assuming that sudden sharp and rapid movements were associated with prey pursuits (Ropert-Coudert et al. 2006, Chimienti et al. 2016), peaks above 0.15 g during dives on a 1s moving average VeDBA were considered to represent Prey Encounter Events (PEE, supplementary Fig S 2-2) following the method described in Sánchez et al. (2018). Intervals of ≤ 2 s between VeDBA peaks were considered to be related to the same prey item as inspection of the frequency distribution of interval durations revealed a sharp decrease in the number of PEE above this threshold. The PEE were then aligned with the diving behaviour and the interpolated GPS coordinates in order to describe their vertical and spatio-temporal distribution along the foraging trip.
All the statistical analyses were conducted within the R software (R Core Team 2017). Different tests were used to assess the variable normality (Shapiro-Wilk test), variance homogeneity of unbalanced samples (Bartlett test), to look at the distribution between departures and returns times (Wilcoxon test) and their variation between breeding stages (Kolmogorov-Smirnov tests). The uniformity of range bearings was tested with a Kuiper’s test. To investigate the variables influencing the foraging behaviour of the instrumented individuals, different mixed models were fitted to account for the hierarchical structure of the tracking data (Bolker et al. 2009) using individuals as a random factor. Generalized Linear Mixed Models (GLMMs) were used to investigate the influence of the year, breeding stage and sex (explanatory variables) on the maximum distance from the colony, mean horizontal speed, trip duration and total horizontal distance travelled (response variables) using the package nlme (Pinheiro et al. 2014). Allowing non-linear relation between the response variable and its covariables (Wakefield et al. 2009), Generalized Additive Mixed Models (GAMMs) were used to investigate the influence of the year, breeding stage and sex (explanatory variables) on the dive rate, depth, duration, and percentage of anaerobic dives (response variables). The dives temporal autocorrelation was taken in account by an autoregressive component (rho) using the package mgcv (Wood 2004). A GAMM was also used to investigate the influence of the oceanographic variables on the foraging effort index. Prior to modelling, oceanographic variables were scaled and checked for collinearity, with a cut-off criterion of rs = 0.5 for inclusion in the model. GLMs were used to investigate the influence of the year, breeding stage and sex on the isotopic values (δ13C and δ15N).
Model selection was based on the Akaike’s Information Criterion corrected for small sample sizes (AICc), ranking all the candidate models using the package MuMIn (Barton 2016). The best-supported model was chosen, or in the case of several candidate models with substantial support (ΔAIC < 4), a model averaging procedure was used to identify the important predictor variables (Burnham et al. 2011) and to select the individual model including them (Supplementary Table S 2-1). Selected models were validated by examination of the residuals (Zuur 2009). Variable influence was stated with significance at P < 0.05. Unless otherwise stated, data are presented as Means ± SE.
At-sea movements and habitat use
Due to logistical constraints, not all stages could be sampled equally in both years of the study. Device malfunctions and loss at sea of some devices meant that complete data sets were not available from all individuals (Table 2-1). Data on at-sea movements were obtained for 35 individuals (37 trips), totalling 1427 h at sea and 22470 filtered locations (27 removed). There was a significant difference between the sexes in the body mass of departing birds, with males weighing 3.4 ± 0.1 kg and females 2.8 ± 0.0 kg (ANOVA, P < 0.001), but not between year or breeding stage. When recaptured, the same birds were weighed with an average mass gain of 44 ± 33 g. Movements from the colony occurred throughout most of the day (01:00 – 22:00 h) with departures occurring mainly before sunrise (67 %), while returns to the colony were spread throughout the afternoon until sunset (70 %). This pattern was consistent across the breeding stages. During incubation, individuals conducted multi-day trips (3 – 5 d), while during the guard stage trips were mostly for a single day (8 – 15 h, 64 %) or shorter multi-day trips (2 d, 36 %). During the crèche stage, all trips consisted of longer multi-day trips (2 – 7 d).
Individuals travelled in a non-random direction across breeding stages (Kuiper test, D = 4.5, P < 0.01), in a northerly direction over a narrow peri-insular shelf (Figure. 1). They reached a maximum distance from the colony of 115 km, making a total home range area of 3877 km2. Their total horizontal distance travelled ranged from 11-379 km, covered at a mean horizontal speed of 3.1 ± 0.1 km·h-1. The GLMMs indicated an influence of the breeding stage and year on the trips parameters (Table 2-2). The guard stage trips were closer to the colony by 62 ± 11 km, they had trip duration that were shorter by 70 ± 9 h, and the birds travelled shorter horizontal distances by 187 ± 25 km (all P < 0.001). The year 2017 had shorter maximum distance from the colony (17 ± 8 km) compared to 2016. Sex had no influence on trip parameters.
Table of contents :
Chapter 1: General introduction
1.1 Foraging and breeding: the essentials of life
1.2 Seabird foraging
1.3 Timing of breeding in seabirds
1.4 New Zealand, a hotspot for winter-breeding seabirds
1.5 Research objectives and thesis structure
Chapter 2: Foraging ecology of a winter breeder, the Fiordland penguin
2.3 Materials and methods
2.3.1 Study site and field procedures
2.3.2 Data handling and processing
2.3.3 Statistical analyses
2.4.1 At-sea movements and habitat use
2.4.2 Diving behaviour and foraging effort
2.4.3 Trophic level and isotopic niche width
2.5.1 At-sea movements and habitat use
2.5.2 Diving behaviour and foraging effort
2.5.3 Trophic level and isotopic niche width
2.6 Supplementary material
Chapter 3: Fine scale foraging behaviour of southern Buller’s albatross, the only Thalassarche provisioning chicks through winter
3.3 Materials and methods
3.3.1 Study site and field procedures
3.3.2 Data handling and processing
3.3.3 Statistical analyses
3.4.1 At-sea movements and activity budget
3.4.2 Foraging behaviour and trophic niche
3.5.1 At-sea movements and activity budget
3.5.2 Foraging behaviour and trophic niche
3.6 Supplementary material
Chapter 4: Foraging niche overlap during chick-rearing in the sexually dimorphic Westland petrel
4.3 Materials and methods
4.3.1 Study site and field procedures
4.3.2 Data processing and statistical analyses
4.4.1 At-sea movements and activity budgets
4.4.2 Fine-scale foraging behaviour
4.5.1 At-sea movements and activity budget
4.5.2 Foraging behaviour and trophic niche
4.6 Supplementary material
Chapter 5: Niche segregation in New Zealand winter breeding seabirds
5.3 Materials and methods
5.3.1 Study sites and field procedures
5.3.2 Data processing and statistical analyses
5.4.1 At-sea movements
5.4.2 Temporal pattern of foraging and diving behaviour
5.4.3 Diet and isotopic niche
Chapter 6: General Discussion
6.1 Foraging patterns of winter- and summer- breeding species
6.2 Marine trophic resource in winter around New Zealand
6.3 Relation between winter-foraging and winter-breeding
6.4 Perspectives and future research