Study area
The study area is found in southern Patagonia, near the town of Puerto Deseado, on the northern coast of Santa Cruz Province in Argentina (47°45’S, 65°53’W). Part of the fieldwork was carried out at the imperial cormorant breeding colony of Isla Chata, situated in the Parque Interjurisdiccional Marino Isla Pingüino (47°55’S, 65°44’W), one of the largest breeding sites for the species, hosting 6613 ± 1541 breeding pairs (Morgenthaler 2019). The other location is the non-breeding site of Peninsula Foca, situated at the entrance of the Ría Deseado; a long (> 40 km), narrow inlet formed by the partial submergence of a river valley by the sea, belonging to the protected area Reserva Provincial Ría Deseado (47°44’S, 65°50’W). Peninsula Foca is a small rocky island connected to the mainland during low tide, which is the roosting ground of around 200-300 non-breeding imperial cormorants (pers. obs.). The two sampling sites, Isla Chata and Peninsula Foca, are situated 20 km apart.
Conventional diet sampling
At Isla Chata, pellets were collected in December 2012 and 2013, during chick-rearing stage, around nests situated at the periphery of the colony. At Peninsula Foca, they were collected between late November and early-February of the same seasons, by walking throughout the roosting area. The pellets were analysed with a binocular microscope and hard prey remains were used to quantify and identify prey at the lowest taxonomic level possible. Identification was carried out by using our own collections and available literature and catalogues (Lombarte et al. 1991; Boschi et al. 1992; Gosztonyi and Kuba 1996; Pineda et al. 1996; Piacentino 1999; Volpedo and Echeverría 2000; Tombari et al. 2010). The frequency of occurrence (%FO) and the number of occurrences (%N) were calculated for all prey items, and expressed as percentages. Published or our own allometric regressions were used to estimate the average total length (TL) and wet weight (W) of different prey types (Pineda et al. 1996; Koen-Alonso et al. 2000; Torroglosa et al 2012). The length of not worn fish otoliths and cephalopod mandibles were used for TL and W calculations. Finally, the Shannon-Weaver diversity index was calculated for each species (Tramer 1969).
SIA sample collection and processing
Whole blood samples of adult and chick imperial cormorants were collected during two consecutive breeding seasons (2012 to 2013) at Isla Chata by late-December for SIA (overall N = 30). Adults and two-to-three week old chicks were captured directly from nests by hand or with the help of a pole. The manipulation of each animal lasted less than five minutes. Chicks were then put back promptly to their nest. At release, adults flew directly to the water, and returned to their nests shortly afterwards. Approximately 0.3-0.5 mL of blood was extracted from the brachial vein of adults and chicks and preserved in 70% ethanol before processing in the laboratory (Hobson et al. 1997).
The pellets used for conventional diet analysis at Isla Chata were collected on the same day and at the same place as the blood sample. The membranes of these pellets were cleaned, all prey items or parasites were removed, and rinsed with distilled water in order to be used for SIA.
The samples were dried at 60°C for >24 h for whole blood, and for >48 h for pellet membrane, and ground to a fine, homogenized powder. Carbon and nitrogen isotope ratios were measured in the Center for Stable Isotopes at the University of New Mexico, USA, by Elemental Analyser Continuous Flow Isotope Ratio Mass Spectrometry using a Costech ECS 4010 Elemental Analyser coupled to a Thermo Fisher Scientific Delta V Advantage mass spectrometer via a CONFLO IV interface. Isotope ratios were reported using the standard delta (δ) notation relative to AIR and Vienna Pee Dee Belemnite (V-PDB), respectively, and expressed in units per thousand (‰) as follows: δ = (Rsample/Rstandard - 1), where Rsample and Rstandard are the molar ratios of the heavy to light isotopes (13C/12C or 15N/14N) of the sample and standard, respectively. Average analytic precision based on routine analysis of a laboratory protein standard was < than 0.1‰ (1σ). The laboratory standard was calibrated against IAEA-N-1, IAEA-N-2, USGS 42 and USGS 43 for nitrogen and NBS 21, NBS 22 and USGS 24, USGS 42 and USGS 43 for carbon.
Five potential prey sources were chosen for SIA according to our preliminary dietary results and information of the same area (Gandini and Frere, unpublished data). These were the three demersal-benthic fish groups: the channel bull blenny (Cottoperca gobio), the Patagonian rock cod (Patagonotothen cornucola) and the eelpouts (zoarcidae fish), and two different cephalopods: the demersal-pelagic squid (Dorytheutis gahi, known as Patagonian squid) and the benthic octopus (Enteroctopus megalocyathus, known asred octopus). Specimens of four of these species were collected in the study area and period, and their SIA values have been previously published in Morgenthaler et al. (2016, 2020) (Table 1). Since no specimens of Cottoperca gobio were collected at the study site, SIAvalues from samples collected in 2014 at the central Patagonian Shelf break between 45-47°S and 60-61°W were used for the mixing models (Zhu et al. 2018) (Table 1).
Table 1 Stable isotope values of main prey of imperial cormorant used in the mixing models. Data are presented as means with standard deviation (in parentheses). All data are corresponding to the study site and period (originally published in Morgenthaler et al. 2016, 2020), but Cottoperca gobio which samples were collected at the central Patagonian Shelf break in 2014 (Zhu et al. 2018)
PREY
|
n
|
δ13C (s.d.)
|
δ15N (s.d.)
|
Cottoperca gobio (channel bull blenny)
|
5
|
-17.9 (0.3)
|
14.4 (0.3)
|
Patagonotothen cornucola (rock cod)
|
10
|
-14.9 (1.2)
|
17.4 (0.6)
|
Zoarcidae (eelpout)
|
8
|
-14.3 (0.6)
|
17.8 (0.9)
|
Dorytheutis gahi (Patagonian squid)
|
6
|
-18.1 (1.0)
|
14.2 (1.3)
|
Enteroctopus megalocyathus (red octopus)
|
4
|
-16.6 (1.0)
|
17.0 (0.3)
|
Trophic discrimination factors (TDF) selection and mixing polygons inspection
Since no diet to whole blood trophic discrimination factors (TDF) has been experimentally determined for the imperial cormorant, choice was made to use SIDER package in R (Healy et al. 2018), which has been successfully used for Neotropic cormorant at the same study site (Morgenthaler et al. 2021). SIDER package, run in R 3.6.3 (R Core Team 2020), uses a phylogenetic regression model using Bayesian inference-based on a compiled dataset to estimate the TDFs of a consumer considering its tissue type and feeding ecology. This package uses a standard database with numerous worldwide published TDF values, however since P. auritus and S. magellanicus TDF values (Craig et al. 2015, Ciancio et al. 2016) were not included in this database, they were added manually before running the models (see Morgenthaler et al. 2021). TDF values obtained with SIDER werechannel bull blenny: Δ13C: 0.70 (±1.40); Δ15N; 2.20 (±0.93), Patagonian rock cod: Δ13C: -0.24 (±1.45); Δ15N:1.59 (±1.01), eelpouts: Δ13C: -0.40 (±1.47); Δ15N: 1.51 (±1.03), Patagonian squid: Δ13C: 0.75 (±1.40); Δ15N: 2.24 (±0.91) and red octopus: Δ13C: 0.32 (±1.41); Δ15N: 1.67 (±1.01).
To the authors’ knowledge, no experimental TDF values for pellet membrane have been published for any bird species. By comparing the isotopic values of imperial cormorant pellet membranes with those of the different prey in our study, it appeared that their ranges were similar (Fig. 2, Table 3). Applying any positive Δ15N values to perform the mixing models would have increased the probabilities of consumers falling outside the mixing space. So the choice was made to maintain null TDF values for both isotopes for all prey to perform the mixing models of pellet membrane as an experimental approach, and to discuss the results obtained with these values and the implication of such choice (see discussion).
For each model (year-tissue) we applied the simulation method of Smith et al. (2013), using the aforementioned TDF values, to ensure that the consumer data were situated within 95% of the source isotopic mixing polygon. The assumption was made that no important source was missing. In the case of whole blood, the 2012 average of probabilities of consumer falling in the mixing polygon was 0.57 (min: 0.46; max: 0.73), and the one for 2013 was 0.58 (0.43- 0.77). In the case of pellet membrane, the 2012 average of probabilities of consumer falling in the mixing polygon was 0.36 (0.11-0.52), and in 2013, it was 0.46 (0.20-0.68).
Stable isotope mixing models
The relative contribution of the potential prey to the diet of the imperial cormorant from Isla Chata based on their isotopic values (whole blood and pellet membrane) was estimated with Bayesian mixing models from the 'simmr' package (Parnell and Inger 2016) in R 3.6.3 (R Core Team 2020). Models were run for each year and tissue separately due to inter-annual differences of cormorant isotopic data. Two different models were run for each year: 1) an initial model, with no prior information, and 2) an informed model with the following priors based on conventional diet estimates from Isla Chata (%W): channel bull blenny: 0.6 (±0.2); rock cod: 0.2 (±0.1); eelpout: 0.1(±0.05); Patagonian squid: 0.05 (±0.02) and red octopus: 0.05 (±0.02). A standard deviation was associated with each of these priors to account for uncertainty in the prior information (Parnell and Inger 2016).
Statistical analyses
For conventional diet, multivariate similarity analyses (ANOSIM) using the R 'vegan' package were used to test for differences in the biomass estimates of the main prey types among years and among breeding stages (Oksanen et al. 2016). The isotopic centroid positions were examined using nested linear models and residual permutation procedures (Turner et al., 2010). Centroid locations were compared between years for each tissue, and between tissues within each year, and were considered different if the Euclidean Distance (ED) between centroid locations was significantly greater than zero (Turner et al. 2010).