2.1 Study area and species

Iceland is located in the Atlantic Ocean, close to the Arctic Circle. Its climate is influenced by the Gulf Stream and the temperature is considerably higher than expected at this latitude. Monthly mean temperatures vary from -3 to 3°C in January and reach 8 to 15°C in July. Precipitation is high, ranging from 400 to 4000 mm annually (climateknowledgeportal.worldbank.org). The temperatures usually prevent the shorelines from freezing throughout the year. On average, the mean annual temperature has been above 2°C during the last decades, whereas it was below 2°C during the last century. Iceland is also free from pack-ice and thereby remains isolated [31].

Icelandic Arctic foxes can adopt two different resource use strategies, referred to as coastal and inland. Interior habitats are more subject to seasonal variations in temperature influencing the availability of resources. The resident rock ptarmigan (Lagopus muta) and migrating birds like waders, geese and passerines make up for most of the inland foxes’ diet [16, 19, 25, 32]. The Western part of Iceland bears the most productive seashores, with a greater productivity than northern, southern and eastern Iceland combined [33]. It also supports most of the large seabird colonies that nest on the cliffs during summer [25]. Ice-free shores contribute to the stability of food supply [34], enabling coastal foxes to benefit from carrion from marine mammals, fish and marine invertebrates in addition to seabirds and some terrestrial preys [19]. All foxes can occasionally consume sheep (Ovis aries) and reindeer (Rangifer tarandus) carcasses, as well as cattle and horse carcasses that are used as baits by hunters [16]. Because of their more stable food resources, coastal foxes are thought to be more territorial whereas inland foxes are more mobile [19]. This subdivision is reflected in genetic differentiation between coastal foxes from the north-western part of Iceland and the foxes from the rest of the country [35]. Because of this clear distinction, we carried out all analyses addressing coastal and inland foxes separately.

2.2 Arctic fox samples

Fox mandibles were obtained from the collection of the Icelandic Institute of Natural History in Reykjavik, which consists of 12,200 mandibles from Arctic foxes culled from 1979 to present. Legally killed fox carcasses are donated voluntarily by hunters from all over Iceland. All foxes have been aged by counting annual cementum lines of canine tooth roots [36] at Matsons laboratory (United States). To include both coastal and inland foxes and assure a sufficient sample size, we chose jaws of foxes culled in Nordur Isafjardarsysla (henceforth NIS) and Nordur Múlasysla (NMU), two counties respectively representing the Western productive seashores and the Eastern inland areas (Fig 2). Individuals from NIS were culled within 5km from the coast and can thereby be qualified as coastal whereas foxes culled in NMU were located further than 50km from the coast and are thus defined as inland. We included 106 samples from Carbonell Ellgutter & al. [28] culled in the same counties, along with samples coming from 5 other counties whose ecology was either similar to NIS or NMU (Ester Unnsteinsdóttir, personal communication), and culled within 5km from the coast or further than 50km from respectively. Altogether, this added up to a total of 256 samples, 127 of them being from inland areas and 129 being from coastal areas, covering a period of 40 years, from 1979 to 2018. The last 14 years, characterized by large fluctuations in abundance, were sampled more intensively. All the foxes analyzed were between one and two years old. As previous research showed that resource use as inferred from δ¹³C and δ¹⁵N does in general not differ between sexes [32, 37] male and female individuals have been chosen at random (S1 and S2 Tables).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Map of Iceland displaying the different culling locations.

https://doi.org/10.1371/journal.pone.0282128.g002

The isotopic composition of bone collagen has a slow turnover rate [38, 39]. It can reflect a lifetime average dietary intake although it is biased towards the period of greatest growth [3], which is until 8–9 months old for Arctic foxes [25]. Here, we will assume that the isotope signatures are representative of the diet of Arctic foxes during their first year of life, thereby reflecting their average resource use during this period. Collagen has been extracted from lower jaws following the same protocol as Carbonell Ellgutter & al. [28], based on a standardized method from Brown & al. [40] and modified according to Richard & Hedges [41]. Bone powder has been extracted with an electric drill and demineralized in a 0.25 M HCl solution for 6 days or until decalcification. After being filtered, rinsed and dried, the bone powder was grinded using a mixer mill (Tissuelyzer II, QIAGEN) before we proceeded to the lipid extraction. Lipid extraction was performed twice with 2:1 chloroform/methanol. The collagen samples were finally freeze dried before they were sent to analysis for stable isotopes of carbon and nitrogen at the Stable Isotopes in Nature Laboratory (SINLAB) at the Canadian Rivers Institute, University of New Brunswick. A Finnegan Delta V Plus isotope-ratio mass spectrometer was used coupled with a Finnegan Conflo IV and CE NC2500 Carlo Erba element analyzer. Stable isotope ratios were expressed using the standard δ notation in parts per thousand (‰) [42]. The international standards, i.e the Vienna PeeDee Belemnite for carbon and atmospheric air for nitrogen, were used as reference [42]. Data were calibrated using a three-point calibration curve and international reference material as standards (Nicotinamide, BLS, MLS, USGS61) were used for calibration while N2 and CH7 were used as check standards. Additional technical information can be found in S3 Table.

The quality of the inferred isotopic composition of the collagen samples was checked following Guiry and Szpak [43] using the atomic C:N ratio as quality criteria. Based on collagen composition reported by the authors and on relationships between δ¹³C and C:N in our data we excluded data with an atomic C:N ratio exceeding 3.33, the highest ratio reported for mammalian collagen (see S4 Table for a justification of this threshold). We thus included 25 samples with an atomic C:N ratio between 3.28 and 3.33, which need to be considered with caution as they may be contaminated to a certain degree with non-collagen material. We excluded 23 samples with an atomic C:N ratios above 3.33.

Coastal areas are shown in blue, and inland areas in green. Culling locations are shown as a circle when hunting does not occur in the whole region. The number of foxes included in our dataset is specified under the abbreviation for each region. Reprinted from the Icelandic Institute for National History under a CC BY license, with permission from Hans H. Hansson, original copyright 2023.

2.3 Prey samples

Greylag goose (Anser anser) muscle and egg samples along with northern fulmar (Fulmarus glacialis) muscle samples have been provided by the Icelandic Institute of National History. Both muscle and egg samples have been prepared for stable isotope analysis following the method from Ehrich & al. [44]. Egg samples were dried and subdivided into a sample to analyze for δ¹⁵N, and a sample subjected to lipid extraction with 2:1 chloroform-methanol, before measuring δ¹³C. Muscle samples were cleaned with ethanol, dried, grinded in a mixer mill and subjected to chemical lipid extraction. Samples were analyzed for stable isotopes in carbon and nitrogen at SINLAB, along with the collagen samples. Additional prey signatures of ptarmigan, common eider (Somateria mollissima), wood mouse (Apodemus sylvaticus), golden plover (Pluvialis apricaria), whimbrel (Numenius phaeopus), sheep, horse, kittiwake (Rissa tridactyla), starfish (Asteria rubens), redshank (Tringa totanus), common snipe (Gallinago gallinago), and black guillemot (Cepphus grille) were obtained from Carbonell Ellgutter & al. [28].

2.4 Data analysis

The statistical analysis was performed using R version 4.0.4 [45]. We corrected both foxes and preys’ raw δ¹³C values for the Suess effect, which consists in a depletion in δ¹³C in the biosphere driven by the input of CO₂ from fossil fuel since the Industrial Revolution [46]. To be consistent with Carbonell Ellgutter & al. [28], we followed the same method, i.e using a mean δ¹³C rate of change of -0.026% per year [47] and correcting all the δ¹³C isotopic ratios to levels, which correspond to the first year of the study (1979). The resulting correction was close to measurements obtained from terrestrial ecosystems [48].

The distribution of Arctic foxes’ isotopic signatures with respect to their potential preys was assessed graphically by plotting individual values from both habitats. Prey values were corrected for isotopic discrimination, which corresponds to the amount of change in isotope ratios occurring as a resource is incorporated into the consumer’s tissue [42, 49]. Since the discrimination factor for Arctic fox bone collagen has not been determined experimentally, we used a combination of the experimentally determined trophic discrimination values from Arctic fox blood cells (+0.49‰ and +2.56‰ for δ¹³C and δ¹⁵N respectively) [37], and a correction factor between red blood cells and bone collagen estimated for wolves by Adams & al. [50] (+2.6‰ and +0.8‰ for δ¹³C and δ¹⁵N respectively). The resulting estimate for the discrimination factor from the diet to bone collagen of Arctic foxes was used to adjust our prey signatures with: 3.09 ± 0.25‰ and 3.36 ± 0.37‰ for δ¹³C and δ¹⁵N respectively. This discrimination factor is an approximation and cannot fully account for the complexity of all the molecular processes involved in isotope discrimination during resource assimilation, thus some uncertainty about true discrimination remained.

Temporal changes in Arctic fox isotopic values were analyzed using generalized additive models (GAM) with the mgcv package to allow for non-linearity [51]. Changes in δ¹³C and δ¹⁵N have been modelled as a smooth function of the birth year of each individual. An interaction allowed to fit different changes over time for inland and coastal foxes, and their different means with respect to the two isotopes ratios were modeled as a fixed effect. We chose the restricted maximum likelihood method and used the default parameters of the package for both smoothing parameter and the number of basic functions. Both gam.check and concurvity functions were used to test for the fit of the model as recommended by Ross & al. [52] (S5 Table).

The proportions of different prey in Arctic foxes’ diet over time were estimated using Bayesian stable isotope mixing models as implemented in the MixSIAR package [53] and were illustrated with boxplots. As mixing models perform best with few potential sources, the different prey items were grouped considering the similarity of their isotopic signatures and their ecological relevance. We created four distinct groups: cliff nesting seabirds (black guillemot, northern fulmar and kittiwake) called seabirds, eider ducks that had distinct isotopic signatures from the other seabirds were grouped with whimbrels to marine resources, the rock ptarmigan was kept as a focal species, and all other terrestrial prey (common snipe, greylag goose, goose egg, redshank, golden plover, wood mouse, horse, sheep) were grouped to terrestrial resources. Because of their more marine signature, whimbrels were placed with eider ducks, but the other wader species were placed in the terrestrial resources. To address potential changes in dietary composition, we defined four periods characterized either by growth or decline of the Arctic fox population (Fig 1). The mixing models were run separately for coastal and inland foxes. To account for the uncertainty considering the discrimination factor and because the results of mixing models can depend on how correct the discrimination value is [49], the analysis was repeated with the discrimination factor used by Carbonell Ellgutter & al. [28] (S6 Table, S1 Fig), in addition to the factor described above. We ran the analysis following the MixSIAR manual recommendations, and did 1 000 000 or 300 000 000 MCMC replicates, (preceded by 500 000 or 1 500 000 burn-in), depending on when convergence was reached, and used a residual*process error [54]. The performance of mixing models being sensitive to the selection of priors [55], and as all priors are informative in a mixing context, we used priors based on the known dietary preferences of the foxes from both habitats, i.e. predominantly marine diet for coastal foxes and terrestrial diet for inland foxes [16, 32]. Thus, the prior attributed to coastal foxes was 2/3 of marine diet components (1/3 seabirds and 1/3 marine resources) and 1/3 terrestrial diet components shared between the two groups (1/6 rock ptarmigan and 1/6 terrestrial resources). For inland foxes the prior was opposite with 2/3 terrestrial diet components shared between the two groups and 1/3 marine diet components. The analyses were also run with uniform priors for comparison. The convergence of the MCMC estimations was assessed based on Gelman-Rubin and Geweke diagnostics, and we also inspected the correlation between different sources. In addition, we checked that the posterior distributions were unimodal. The isospace plots for the converging models are shown in S2 Fig.

The use of stable isotopes to infer a population’s trophic niche width is increasing as isotopic niches are considered a good proxy for ecological niches [56–58]. For each period and habitat, we determined the isotopic niche breadth using standard ellipses containing 80% of the data as calculated by the SIBER package. The areas of those ellipses were computed using the Bayesian approach (SEAb, parametrized as detailed in Jackson et al. [59], S3 Fig). The ellipses being unbiased for sample size, it was possible to compare them even though the periods did not contain an equal number of individuals.

4.1 Variations in resource use

Our results confirmed the previously described differences in diet between coastal and inland foxes, with coastal foxes having overall more marine isotopic signatures than inland foxes [28, 34]. The stable carbon isotope composition among inland foxes over the study period was combined with a slight increase in nitrogen istotopic composition in recent years. Coastal foxes, on the opposite, showed no statistical evidence for changes in δ¹³C and δ¹⁵N values, and the changes in δ¹³C they underwent during the beginning of the study had a low statistical support, and were likely due to the small number of individuals sampled during the first years of the study. These overall rather stable diets both at the coast and in inland habitats were in agreement with the estimations of dietary composition from isotopic mixing models. Both analyses also suggested some slight fluctuations, although not significant, in the inland. This was in agreement with the temporal changes in isotopic niche breadths, which showed important changes in the resource use of inland foxes while coastal foxes seemed to have a rather stable diet and constant niche width throughout the years.

In agreement with Carbonell Ellgutter & al. [28], who highlighted the importance of marine resources in the diet of Icelandic Arctic foxes, the mixing model analysis showed that marine preys were the main resource consumed by coastal foxes, and that they also were important to a certain degree for inland individuals. However, contrary to Carbonell Ellgutter & al. [28], we separated the marine resources in two groups and our results suggested that preys from lower trophic levels such as common eiders and whimbrels could be more important than cliff-nesting seabirds. Although the separation between the two groups of marine resources is not strongly supported by the mixing model (high correlation between the posterior distributions and overlap of credibility intervals), it appears likely that marine resources like common eiders, which are the most common waterfowl available throughout the year and are widely distributed in coastal areas, could represent an important part of the diet of Icelandic Arctic foxes. However, predation on eiders in Iceland is thought to be lower than for other populations in other areas since their protection is one of the reasons for the Icelandic fox culling program [26, 60]. A high importance of common eider in Arctic fox diet is also contrary to findings from prey remains at dens [20], where fulmars were the most common species. This unexpected result could possibly be due to the fact that many of the coastal foxes were culled at eider colonies (Unnsteinsdóttir, personal communication).

Alternatively, whimbrels, which were included in our marine preys, could have been an important resource for foxes from both habitats as they are abundant and accessible during the breeding season [61]. Unfortunately, the grouping used in the mixing models analysis makes it hard to determine whether or not this particular wader species was more important than the others, and no information was found in literature to either support or contest this assumption.

The rock ptarmigan, also a year-round resident in Iceland, appeared to be the preferred terrestrial resource. They were especially important for inland foxes but were also consumed in significant amounts by coastal foxes (Fig 5b).

During the breeding season, waders and geese are increasingly available to foxes, and goose populations are especially increasing [20]. In contrast with Pálsson & al. [20] who suggested that these resources were the main preys available to inland foxes, our results suggested that these prey items were of minor importance. This confirmed the results from Carbonell Ellgutter & al. [28] who did not find support for an increased use of geese, despite isolating the greylag goose as a focal source in their mixing model analysis. The lack of species such as the pink-footed goose in our prey signatures could explain the apparent minor importance of this group since this species has been shown to be especially important for Iceland Arctic foxes [16, 20]. The inclusion of terrestrial species like reindeer, which can occasionally be consumed by the foxes, would have given a more detailed picture of the diet of inland individuals, while the lack of marine species such as puffins (Fratercula arctica), crustaceans and other invertebrates might result in a underestimation of the proportion of marine resources, especially in the diet of coastal foxes. The use of freshwater prey by Arctic foxes in Iceland has not been reported in the literature [16, 19]. Although the isotopic composition of aquatic resources can be influenced by complex and diverse factors [39], the turnover rate of collagen reflects the lifetime dietary intake of the foxes. Thus, in the event of an occasional consumption of freshwater resources, it is unlikely that it would have affected our results. It is also important to note that the discrimination factor we used is an estimation as the fractionation has not been calculated for collagen in Arctic foxes yet. However, we assessed the robustness of our analysis using also the values from Carbonell Ellgutter & al. [28].

4.2 Driver of population change

As suggested by Unnsteinsdóttir & al. [22], the fluctuations in the population size of Arctic foxes likely result from changes in carrying capacity due to changes in abundance of main resources. Accordingly, the results of Carbonell Ellgutter & al. [28] showing a constant and important use of marine resources indicated that increasing populations of seabirds could have been a major driver of the long population increase of the foxes from the 1980’s. For coastal foxes, our results showed a stable diet composition over the whole study period, despite considerable fluctuations in population size (Fig 1). This is also consistent with a bottom-up regulation of the fox population. Indeed, after a period of increase for many seabirds in Iceland up to the turn of the millennium, populations stabilized and some dramatically declined during the last decades [62, 63] when the numbers of foxes hunted in coastal areas started to fluctuate considerably. Breeding failure of several species was observed in 2005, while in 2010–2011 Puffin reproduction failed totally [64]. This was attributed to a lack of sandeel (Ammodytes marinus), a major resource for many seabird species [64, 65] notably for Fulmars [66], a key prey of Arctic foxes [20].

The strong decrease in fox numbers in 2011–2012 that was more pronounced for the coastal population might be a direct consequence of this event. Moreover, our results suggested that Common Eider is likely to be an important prey for Arctic foxes. Their populations stabilized and partly declined in Iceland in the 2000s after an overall increase in the end of the last century [67] and notably declined in Westfjords in western Iceland after 2000 [68]. This resource may thus also have contributed to flattening out and periodic decline of the coastal fox population.

For inland foxes, Ptarmigan were the main resources during the whole study period, but their population trends cannot really explain the increase in the Arctic fox population as many Icelandic populations showed declining trends in the last decades [69]. As suggested by Carbonell Ellgutter & al. [28] it is likely that the increase in marine resources also resulted to an improved situation for inland foxes during the growth period. This is supported by the fact that the crash of seabird populations together with declining Eider populations in 2010–2011 led to an abrupt decline in foxes hunted in the inland as well, although the decline was not as strong as for the coastal population. Notably fulmars were indeed breeding on cliffs far inland and preyed upon by Arctic foxes, but at present, these inland colonies are less active (ER Unnsteinsdóttir, personal observations). During these years, the marine input in the diet of the foxes declined and the proportion of ptarmigan increased. However, according to Fuglei & al. [69] ptarmigan populations were in the low phase in these years and could thus not compensate the lack of marine resource for the foxes. After this decline phase, the diet of inland foxes returned to its previous state. The non-significant trend for an increase in nitrogen isotopic composition in the end of the study period might indicate increased use of geese and waders that experience positive population trends in inland areas [20], at least by some foxes–although such a diet shift was not detected by the MixSIAR analysis. In the future, inland foxes might adapt their diet to the changing resource situation shifting from declining Ptarmigan and seabirds to other terrestrial preys.

4.3 Population isotopic niche breadth

The high hunting pressure in Iceland leads to a high turnover in territorial foxes, and Unnsteinsdóttir & al. [22] suggested that Icelandic Arctic foxes engage in contest competition as they adapt their territory size in response to variations in carrying capacity.

Consequently, we predicted that Arctic foxes’ niche breadth would vary over time, but found no support for this hypothesis in the coastal habitat. Although the diet of coastal foxes seemed not to vary over the study period, one could have expected that the decline in the availability of seabirds would have led to a narrower niche breadth or to a shift in the diet. The apparent consistency in their isotopic niche at the population level could hide some variations at a finer scale—the individual scale.

The results from isotopic niche analysis suggested that coastal foxes globally have broader niches than inland foxes. In previous research, Dalerum & al. [34] pointed out the same phenomenon and suggested that these wider niches were due to a diversification of individual strategies likely dictated by the local abundance of resources. The heterogeneity of coastal areas could lead to increased individual specialization of some foxes, especially since coastal foxes are more territorial [19]. This assumption would support that the fluctuations in carbon isotope composition observed among coastal foxes are likely to be influenced by individual variations in the diet rather than a global shift in the resource used at the population scale. This specialization would be a way of reducing the potential dietary overlap among foxes, in response to an increasing intra-specific competition pressure [70].

Inland foxes on the contrary showed more variations in isotopic niche space over time, as well as a marked reduction in their niche breadth during the period of decline. This fits with the increased use of rock ptarmigan suggested by the mixing model results during the decline period. In previous research, Hersteinsson [19] showed that the number of occupied fox dens was positively correlated to ptarmigan abundance, highlighting the importance of this prey item.

The reduction in the niche space matches the years of low productivity of seabirds, and illustrates the low availability of these resources, thus narrowing the niche. As the fox population recovered, the niche size widened, which could indicate that some marine species were available again.