1. Behavioral and feeding flexibility are crucial components of the adaptive capacity of species regarding environmental change. Understanding the individual variety of behaviour remains a key issue in conservation practice, with difficulties in measuring this variability continuously and linking it to biotic or environmental covariates. 2.
[...]
In this article, we propose a parametric statistical approach to construct and explain isotopic richness, as seen as a component of dietary behavioral diversity. We place ourselves in the framework of trait probability density and assume an explicit parametric model on our data. 3. Our goal is to maintain ecological relevance and interpretability throughout our modelling process. We draw on quantile regression to propose a model with minimal assumptions on the data. We are less interested in a point-by-point estimate of isotopic signature than in accurately predicting their variation. As a consequence, we develop a validation criterion allowing our parametric model to fit our stable isotopes data, in the sense of a similarity of distributions. A detailed algorithm is proposed to describe our procedure. 4. We apply our methodology to real stable isotopes data from a dataset of New Caledonian coral reef fish. We can exhibit the effect of the interaction of diet and reef state on the variability of both Carbon and Nitrogen isotopic signatures. This illustration also shows some adequacy between known ecological results and the output of our model, by, for example, demonstrating an increase in trophic diversity for herbivore-detritivores and a decrease in trophic diversity for corallivores in a degraded coral reef. Our model could be applied to other behaviors, such as habitat selection or home range estimation.
