The Southern Ocean is biologically the world's most productive ocean. At the hub of the Antarctic marine food web, krill is the primary consumer of diatoms, the major prey for many species of fishes, penguins, seals and whales1,2,3 and a substantial commercial fishery resource1. The abundance of krill is very sensitive to climate change and has significant impacts on high trophic-level predators in the Southern Ocean ecosystems4,5. Recent Antarctic krill populations have been significantly influenced by rapid climate change, human removal of krill-eating predators and intense commercial fishing1,6,7. Therefore, the records of krill population change, especially those pre-dating the onset of human harvesting in Antarctic, are crucial for understanding and predicting responses of krill population to natural climate changes.

Modern krill abundance can be obtained directly from acoustic and net surveys. Krill population data from scientific trawls are available for the past ~30 years6. For historical krill abundance, only about 100 years of data have been inferred from Antarctic fur seals8. The long-term historical krill abundance prior to human intervenes remains unknown. Similar to Antarctic fur seal, Adélie penguin is an important land-based krill predator, which has been chosen as an indicator of changes in krill availability and abundance by the commission for the conservation of Antarctic marine living resources9. Furthermore, Adélie penguin is a circum-Antarctic distributed species; continuous and long time-series remains of Adélie penguin could be preserved in various lake sediments10. Thus, Adélie penguin is an ideal indicator for tracking krill availability and abundance over a long period of time.

Stable isotope analysis of animal tissues is a powerful tool for examining wildlife foraging habitats, diet and migration patterns11,12,13, especially in historical periods14,15. Keratinous tissues such as toe nails, feathers and hair and bone collagen can preserve dietary information for long periods of time16, particularly in the cold and dry Antarctic environment. They are ideal for investigating paleodiets of krill predators8. Hairs and faeces in lake sediments have been used successfully to infer past population dynamics of seals and penguins in Antarctica17,18,19. Similar stable isotopes and biomarkers have also been used extensively in northern high-latitude or Arctic regions to study the effects of climate change, seabird colonisation and past whaling activities on lake ecosystems20,21,22.

Here, we analyzed stable nitrogen isotope ratios (15N/14N, expressed as δ15N) of modern and ancient Adélie penguin bones and feathers from the Vestfold Hills, East Antarctica (Figure 123,24 and Figure 225), with the main aim of inferring the relative krill abundance over a long historical time. Several observations follow. First, δ15N values in penguin tissues show an enrichment of ~3–5‰ from prey to predator in marine ecosystems11. Second, as a diet of penguins, krill are much lower in δ15N values than fishes are15. Third, Adélie penguins feed preferentially on krill species26,27,28 of Euphausia superba in the Antarctic Peninsula and of Euphausia crystallorophias along the East Antarctic coasts. Lower δ15N values in Adélie penguins indicate diets based primarily on krill, while higher values indicate diets richer in fishes and other species of potentially higher trophic level (e.g., squid). The variations in δ15N values of penguin tissues indicate the changes in the proportion of krill in penguin diets and thus the krill availability and abundance in the foraging area29. The δ15N values in penguin tissues through time could serve as a proxy for krill availability and abundance in foraging areas.

Figure 1
figure 1

Map of the Vestfold Hills including the sampling site in this study and two sites in previous studies23,24 in East Antarctica.

(The map was drawn using Microsoft Excel 2010 and then converted to tiff format using Microsoft Office Visio 2007).

Figure 2
figure 2

Adélie penguin in the Vestfold Hills and the bones and feathers from the sediment core DG4 profile with conventional AMS 14C dates25 (yr BP: years before present).

Photo credit: T. Huang.


The δ15N values of penguin bones and feathers at different depths of a sediment core DG4, as well as modern samples from Magnetic, Zolotov and Gardner Island at Vestfold Hills, are given in Table 1. The δ15N values of main preys for Adélie penguin in East Antarctic30,31 are plotted in Figure 3. Sample sizes (n) reported below are the numbers of the distinct depths in the DG4 sediment at which samples were collected, as shown in Table 1. Modern Adélie penguins at Vestfold Hills have δ15N value (mean ± standard error of mean) of 10.1 ± 0.3‰ for bones (n = 6) and 10.0 ± 0.04‰ for feathers (n = 6), which are very close to those reported in Adélie Land (9.4 ± 0.09‰, n = 20) and MacRobertson Land (9.4 ± 0.2‰, n = 31), East Antarctica23,24 (Figure 3). The δ15N values of ancient Adélie penguins are much higher. They ranged from 12.5‰ to 18.1‰ in the bones with a mean of 15.3 ± 0.6‰ (n = 10) and ranged from 11.3‰ to 15.5‰ with a mean of 13.6 ± 0.4‰ (n = 11) in the feathers (Table 1 and Figure 4).

Table 1 Stable nitrogen isotope ratios (expressed as δ15N) of modern and ancient penguin tissues collected from Vestfold Hills, East Antarctica. The interpolated radiocarbon age of ancient penguin tissues were calculated based on AMS14C data from a previous study25. Also indicated are reconstructed warm/cold climate periods32,33,34 to which the sample at each depth corresponds (slash = neither warm nor cold)
Figure 3
figure 3

δ15N values (mean ± standard error of mean) of Adélie penguins in this study and previous studies23,24 and penguin prey δ15N values compiled by Cherel et al. (2008)30 for Euphausia crystallorophias, Euphausia superba and Pleuragramma antarcticum and Gillies et al. (2012)31 for Trematomus bemacchii.

(VH: Vestfold Hills, AL: Adélie Land, ML: MacRobertson Land).

Figure 4
figure 4

δ15N value fluctuations in Adélie penguin bones and feathers during the Holocene epoch, along with the episodes of climate change as inferred from ice cores and marine sediments in East Antarctic32,33,34.

The instrument error for the isotope analysis is ≤0.2‰. Red bands refer to warm climate periods and blue bands to cold periods and expanded sea ice condition.

The δ15N values in modern Adélie penguin bones and feathers at Vestfold Hills are similar (Wilcoxon rank sum test statistic T = 45, p = 0.35). For the same type of tissues, δ15N in modern and ancient samples are significantly different (for bones: T = 115, p = 0.0002 and for feathers: T = 132, p = 0.0002), with higher δ15N found in ancient penguins.

In addition, the δ15N values in ancient penguin feathers and bones between warm and cold climate conditions are also significantly different. Here, we determined the episodic warm periods (7500–6300 yr BP and 4800–2200 yr BP) and cold periods (8500–7600 yr BP and 6300–5700 yr BP) according to previously reconstructed climate change during the Holocene in East Antarctic based on ice cores32 and marine sediment cores33,34. Using information of warm/cold periods provided in Table 1, we were able to compare δ15N values in cold periods against those in warm periods. It was found that δ15N values of ancient Adélie penguin during cold periods are much lower than those during warm periods (Figure 4), in both bones (T = 3, p = 0.05, nwarm = 7, ncold = 2) and feathers (T = 10, p = 0.01, nwarm = 6, ncold = 4). Furthermore, when combining data from bones and feathers to form a single test, we observed even stronger evidence (T = 21, p = 0.0001, from a tissue-stratified Wilcoxon rank sum test). This implies the relative krill abundance in the cold periods is much higher as compared with that in the warm periods.


Different penguin tissues have different metabolic rates and their stable isotope signatures have different time scales, from days for blood plasma to weeks for feathers and to months of average turnover-time for collagen of bones16. Because our samples were collected at a breeding site during a summer season, the stable isotope values of archive feathers in the sediment core DG4 reflect those of the average diets at the time of feather growth. The nitrogen isotope values of penguin bones reflect those of the average diets of Adélie penguins during the last turnover of their bone collagen. Visual inspection showed that the ancient feathers in DG4 were from adult penguins whereas the source was uncertain for the bones (which can be from adults and/or juveniles). Since the trend and magnitude of δ15N in the ancient penguin feathers are similar to those in the bones (Figure 4), the possible age difference in the penguin bones seems unlikely to drive the temporal change of δ15N during the Holocene epoch.

Variation of the inferred relative krill abundance (represented by the inverse of δ15N values in penguin tissues) was found to be in accordance with previously reconstructed episodes of climate change during the Holocene in East Antarctic32,33,34. That is, periods with high krill abundance (lower δ15N values) correspond to cold climates32,33, whereas periods with low krill abundance (higher δ15N values) correspond to warm climates32,34. Krill is a stenothermal, cold-water planktonic species and is very sensitive to natural climate changes. Krill abundance has been linked with expanded sea ice cover (extent and duration) and increased latitudinal extent of cool water6,35,36, which are consistent with our findings. Moreover, our results from Adélie penguins are also consistent with those from Antarctic fur seal hairs from a sediment core spanning the past century in the West Antarctic Peninsula8. Taken together, these results suggest that the link between krill abundance and climate persists over the past 8,000 years in East Antarctica.

It has been reported37 that δ15N values of the primary producer (baseline) and thus penguins' preys may also be influenced by climates. This can in turn give rise to the change of δ15N in penguins in response to varied climates. However, we note that the δ15N values in the Southern Ocean phytoplankton (fragilariopsis kerguelensis, main food of Antarctic krill) are higher during cold climate periods and lower during warm periods38, opposite to the patterns seen in our penguin tissues. Based on these facts, we argue that the temporal variations of δ15N in penguin tissues cannot be attributed to the variation of δ15N of the krill themselves but primarily to the change of penguin's diet composition.

Modern penguin bones and feathers have much lower δ15N values than ancient ones. Similar low δ15N values were also observed in modern penguin eggshells by Emslie & Patterson (2007)15. These differences are most likely caused by substantial dietary changes between modern and ancient Adélie penguins (Figures 3, 4). There are two other possibilities which, however, do not seem to explain the differences observed in the current study. First, diagenetic processes may change dietary signals in tissues over time. This is unlikely for penguin feathers and bones because of the excellent preservation condition in the dry and cold Antarctic environment39 though the preservation state of ancient tissues has not been directly studied in the present study. Second, ammonia volatilization could cause older tissues to become more enriched in nitrogen-15 than younger or modern tissues40. This, however, only applies to sediment, soil and plants growing in the soil. In fact, the ancient penguin bones and feathers in DG4 have much lower δ15N values (12–18‰) than those in the sediments of DG4 core (25–39‰, Table S1) and ornithogenic soils (~30‰) (Mizutani & Wada, 1988)40. Furthermore, this would not explain the similar patterns of fluctuations in δ15N values observed in both feathers and bones if volatilization is a significant factor (see also Emslie et al. 201341).

Taken together, the lower δ15N values of modern tissues than in ancient tissues indicate higher proportion of krill in modern Adélie penguin diets and thus greater krill abundance in modern ages than in ancient ages, in the East Antarctic. This is consistent with the observations by Emslie & Patterson (2007)15, which supports the ‘krill surplus’ hypothesis, i.e., the higher krill abundance is likely due to the hunting of krill-eating seals and whales since as early as the 19th century.

One limitation of the present study is the small sample size for analysis. In the sediment core DG4, each 1-cm section contains only one or two pieces of feathers and bones. To address this issue, we have chosen a nonparametric statistical method to test the differences in δ15N values between different conditions. Since Adélie penguins and krill are widely distributed in the circum Antarctic, we expect to find more sampling sites and collect multiple sediments in future field work.

Our results corroborate that climate change has a pronounced effect on krill abundance and hence Adélie penguin diet composition26,27,28. The reported 8000-year record of the relative krill abundance, as inferred from Adélie penguin paleodiets, provides a unique insight into the change of regional marine food chains in natural status. This will be valuable for assessing impacts of future climate changes on this key species and thus for the conservation and management of Antarctic marine resources.


Sample collection and chronology

We collected a sediment core DG4 near a large Adélie penguin colony at Gardner Island, Vestfold Hills (Figure 1), East Antarctica, during the 2005–2006 austral summer. There are a large number of penguins in the colony in summer breeding seasons and penguin remains such as guanos, bones and feathers are deposited into the sediments each year (Figure 2). This sediment thus preserves a record of penguin activities at this colony since it was first formed. The core was sectioned at 1 cm intervals in the lab. Layers from 55 cm to the surface contained numerous penguin bones and feathers, but sample sizes per 1 cm section were quite small. Modern bone and feather samples were collected from dead adult penguins at Gardner, Magnetic and Zolotov Island, Vestfold Hills.

The chronology of DG4 has been established by AMS 14C dates on penguin tissues and bulk sediments in the core in a previous study25 and the penguin occupation appears to have been continuous since 8500 year BP to the present25.

Stable isotope analyses

Penguin bones and feathers were picked from each 1-cm section of the sediment core and used for stable isotope analysis. These samples as well as the modern adult penguin tissues were cleaned with Millipore water and 2:1 chloroform: methanol solution and then dried in an oven at 40°C. The cleaned samples were cut into small pieces and weighed in a tin capsule. Stable nitrogen isotope ratios in whole feathers and bone collagens were determined using isotope ratio mass spectrometer at the G.G. Hatch Isotope Laboratories, Earth Sciences, University of Ottawa, with a precision at ≤0.2‰. Isotope ratio (15N/14N) in samples are reported in δ notation and the units are per mil (‰) and defined as δ15N = [(Rsample − Rstandard)/Rstandard], where Rsample is isotope ratio of the sample and Rstandard of the atmosphere (See Supplementary Information for more detail).

δ15N as a proxy for krill abundance

We used the inverse of isotopic ratio of nitrogen, δ15N, in penguin tissues as a qualitative proxy for krill abundance. The δ15N in tissues has been used to reconstruct diet composition of marine species13,14,15. Generally, marine predators are 15N-enriched over their diet, because of the kinetic isotope effects in metabolic pathways involving the bond-breaking or synthesis of biochemical compounds, during which the molecules that contain the lighter nitrogen-14 are preferentially utilized over the same molecules containing the heavier nitrogen-1511. δ15N values can therefore be used as an indicator for the dietary change of marine predators.

Information of krill population can be estimated from penguin's diets based on traditional methods such as stomach lavage analyses42 which, however, often come with biases associated with sampling of gut contents. The stable nitrogen isotope method as used here can be regarded as a qualitative biochemical method and reflects the average amount of all the diets rather than specific ones. Since this method assesses paleodiets of wildlife through their ancient tissues8, there is no time limitation regarding availability of diets.

Adélie penguins feed primarily on krill and, when krill are scarce, on fishes26,27,28. Because krill are much more 15N-depleted than fishes15, δ15N in penguin tissues where diet information is well preserved can indicate the changes of the proportion of krill and thus their abundance. Specifically, high δ15N indicates low krill abundance and low δ15N indicates high krill abundance.

Statistical test for differences in δ15N between different conditions

To assess the significance of the difference in penguin δ15N isotope values between different time periods (modern vs. ancient), different tissues (feather vs. bone) and different reconstructed climate conditions (warm vs. cold), we adopted the nonparametric Wilcoxon rank sum test. Due to small samples sizes and presence of ties, asymptotic p value resulting from a Normal distribution approximation is invalid. Therefore, we used an exact two-sided p value obtained via a permutation procedure (i.e., randomly reassigning δ15N values across groups of samples) for each test as reported in “Results”. The function “wilcox_test” in the R package “coin” was used to perform these tests. The function also implements an extension of the Wilcoxon rank sum test to allow for stratifying on another factor. We've utilized this technique to compare δ15N in cold periods vs. δ15N in warm periods while accounting for the possible impact of tissues (bone or feather) on δ15N.