Why Accurate Stable Isotope Discrimination Factors are so Important: A cautionary tale (involving kea)

Post provided by AMANDA GREER

Stable isotopes as a tool for ecologists

Our research into the foraging ecology of this cheeky parrot (kea: Nestor notabilis) prompted us to develop a simple method to establish discrimination factors © Andruis Pašukonis

Isotopes are atoms that have the same number of protons and electrons but differ in their number of neutrons; they are lighter and heavier forms of the same element. Unlike radioactive isotopes, stable isotopes do not decay over time.

The ratio of heavy to light stable carbon (δ¹³C) and nitrogen (δ¹⁵N) isotopes in an animal’s tissues depend on its diet, although offset by a certain amount. This integration of δ¹³C and δ¹⁵N from an animal’s diet into its tissues allows ecologists to use stable isotope analysis to investigate a species’ present and historical diets, food-web structures, niche shifts,  migration patterns and more.  

Dietary reconstruction studies

One of the most popular uses that ecologists have for stable isotope technology is to help reconstruct an animal’s diet over a given time period. Dietary reconstructions have been significantly advanced by the development of Bayesian stable isotope mixing models which use likelihood estimations to investigate what proportion of the total diet is composed of each food source. This type of analysis can be particularly useful in studying difficult to follow species – for example pilot whales and polar bears – as it is often much easier to collect a tissue sample than it is to track an individual animal to see what it eats. The technique is becoming increasingly prevalent, with the number of citations in the ecological literature for stable isotope mixing models increasing dramatically over the past decade. However, the results of all these mixing models rest on the accuracy of the discrimination factors – the offset between the stable isotope ratios of an animal’s diet and it’s tissue. If the discrimination factors are out, the estimated dietary proportions will be incorrect. This lesson was driven home to us during our first stable isotope dietary reconstruction attempt, which failed rather spectacularly.

The number of citations of studies with the keywords stable isotope(s) and mixing model(s) over the last decade. Data from ISI Web of Science TM. © Amanda Greer

An unlikely result

Our research explores the foraging ecology of different sub-populations of kea Nestor notabilis, a parrot species which lives primarily in high-altitude habitat along New Zealand’s Southern Alps. Although not as difficult to follow as pilot whales, researching

One of our main study sites is the Hawdon Valley in Arthur’s Pass, New Zealand. While kea can easily move from one peak to another, it takes rather longer for researchers. © Andruis Pašukonis

kea presents its own set of challenges. In their mountain habitat it can take kea less than two minutes to fly to a point that is more than half a day’s trek away for researchers. In order to reconstruct the kea’s diet when we could no longer observe them, we turned to stable isotope analysis.

We collected a large number of kea feather samples and samples of the plants and invertebrates that they eat in their high-altitude habitat then analysed them for δ¹³C and δ¹⁵N. When the first batch of results came back we immediately sat down to plot the data. As no discrimination factors for parrots had yet been established, we substituted discrimination factors commonly adopted by ecologists: 0.4‰ for carbon and 3.4‰ for nitrogen. The resulting isospace plot looked something like this:

Isospace plot of kea data points (dots) when plant and animal ratios were adjusted by inaccurate discrimination factors ©Amanda Greer

Oh.

Given that the diet of kea in high-altitude locations has been estimated at being more than 70% vegetarian, and that our own observations had confirmed this trend, this is not how we had expected the plot to look. Even worse, when we plugged the data into our Bayesian mixing model (MixSIAR) it estimated that this population was eating about 85% animal matter! We realised that something had to be seriously amiss with the discrimination factors we were using.

Reading further into the issue we discovered that having errors as small as 1‰ can seriously affect the dietary estimations produced by mixing models. It became clear that, if we hoped to continue with dietary reconstructions, we would have to figure out a way to establish kea-specific discrimination factors.

Traditional methods to establish discrimination factors

Typically, diet-to-tissue discrimination factors are determined by holding a group of animals to a strict diet of no more than three food items until the tissues of interest have fully acclimatised to that diet. Tissue and dietary stable isotope ratios are then sampled and compared. This methodology offers unparalleled control over dietary traits, such as quantity and quality of protein and δ¹³C and δ¹⁵N values, which have all been used to great effect to study the possible effects of diet type on discrimination factors.

Most kea live in high-altitude habitat where research has shown they have a predominantly vegetarian diet ©Andruis Pašukonis

However, housing and feeding animals for extended periods of time is costly and time-consuming. In addition, welfare recommendations may not permit certain species to be held on such a restricted diet over a long time-period (eg. kea). What’s more, some species, particularly those classed as endangered or that require large/special housing, are not easily accessible to the average researcher and so must be studied with the co-operation of a zoo or research facility and they’re not always open to the idea of diet manipulation. The upshot is that for most animal species, no specific discrimination factors are available. For example, for terrestrial birds, discrimination factors have been established for just 15 species*.

This lack of data forces stable isotope ecologists to rely on values determined for the most closely related species, or results from review articles such as the often substituted 0.4‰ for the carbon discrimination factor and 3.4‰ for the nitrogen discrimination factor, or taxonomic-specific reviews like that of **Caut et al. As can be seen from the isospace plot above, these substitutes are not always appropriate. In fact, given that the discrimination factors of different tissues (eg. blood and hair) differ substantially within the same animal, it is impossible for a single carbon or nitrogen discrimination factor to always be appropriate.

A quick and easy alternative

To establish kea discrimination factors we needed the co-operation of a fantastic local zoo. Modern zoos typically keep excellent dietary records of what each group of animals is fed and try to approximate the animal’s natural diet. We obtained these dietary records for kea along with samples of the food items they were fed. Then we collected kea feathers from inside the enclosure. We calculated the δ¹³C and δ¹⁵N of the kea’s weekly diet based on the (dry-weight) contribution of each food item and compared these figures with the values from the kea feathers. This simple experimental protocol can be easily extended to allow researchers to control for metabolic routing of proteins (where proteins contained within food are preferentially routed to certain tissues); for the self-sourcing of foods within the enclosure; and for the animals’ food preferences. Our method is also straightforward to adapt to other taxa and tissue types and is very quick and cost-effective to run.

What a difference a discrimination factor makes

Armed with our new, kea-specific discrimination factors (4‰ for carbon, 3.1‰ for nitrogen) we revisited the original dataset and our results looked very different.  The proportion of animal matter in the kea’s diet had changed from 85% to 36%; a far more reasonable estimate for kea foraging in high-altitude habitat.

The mixing model results based on the incorrect discrimination factors (left) and the kea-specific discrimination factors (right). © Amanda Greer

We offer this cautionary tale to highlight how important it is to have the right discrimination factors before proceeding with dietary reconstructions using stable isotope mixing models. It is our hope that it will become standard practice to establish species-specific discrimination factors wherever possible before embarking on stable isotope mixing model studies. When this is not possible, we urge researchers to run a number of mixing models to investigate the effect that using different discrimination factors has on their results.

Hopefully, the availability of simple methodologies will lead to the publication of discrimination factors for many more species. The resulting dataset could be used to investigate what drives variation in discrimination factors across species – phylogenetic distance? divergent diets or physiologies? their environment? Any trends uncovered could result in better estimations of likely discrimination factors for species which cannot be kept in captivity. Perhaps they could also provide an answer to a question which has been niggling at us since we began this journey: why is the kea’s carbon discrimination factor so atypical in the first place?

To find out more about stable isotope discrimination factors, read ‘Simple ways to calculate stable isotope discrimination factors and convert between tissue types‘ by Amanda Greer et al.

* American crow, chicken, California condor, dunlin, garden warbler, house sparrow, Japanese quail, kea, peregrine falcon, snowy owl, song sparrow, upland buzzard, yellow-rumped warbler, yellow-vented bulbul, zebra finch

** For those of you wondering, when we used the feather-specific discrimination factors suggested by Caut et al., our kea were found to eat 63% animal matter – still a highly unlikely result.