A Bayesian model for assessing organic matter supply in complex marine food webs using amino acid stable isotope analysis

While several software packages have been developed to solve stable isotope mixing models, none are currently equipped to trace the flow of organic matter through the lower trophic levels of planktonic food webs. To address this gap, we have developed a new Bayesian mixing model tailored for use with δ¹⁵N values of individual amino acids. This model simultaneously estimates trophic relationships between consumers and organic matter sources at the base of the food web, determines the relative contributions of these basal organic matter sources to consumers, and accounts for trophic discrimination affecting amino acid δ¹⁵N values during protozoan and metazoan trophic steps. This "Organic Matter Supply Model" is uniquely suited for applications where food web structure is unknown and trophic intermediaries, such as protozoan and metazoan grazers with distinct amino acid trophic discrimination factors, play a critical role in nutrient transfer. In this paper, we describe the model's basic structure, outline key considerations for adapting it to specific applications, evaluate its performance using simulated zooplankton data, discuss its strengths and limitations, and offer recommendations for its further development. By testing the model on simulated zooplankton amino acid δ¹⁵N data, we demonstrate that the Organic Matter Supply Model can enhance our understanding of the roles of small particles and diel vertical migration in deep-sea organic matter supply pathways. Furthermore, it provides a new framework for exploring the foundational role of heterotrophic protists in marine ecosystems. We find specific subsets of amino acids to be most useful as markers of trophic ecology (in this case including glutamic acid and proline) and to identify supply from basal organic matter sources (phenylalanine, lysine, and threonine). Other amino acids may be more ideal source tracers in other settings, although amino acids with inconsistent or poorly constrained isotope fractionation behavior (<i>e.g.</i>, isoleucine, valine) should be excluded to optimize model reliability.