Analyzing mixing systems using a new generation of Bayesian tracer mixing models

Area of Science:

• Ecology
• Quantitative Ecology
• Biogeochemistry

Background:

• Existing tracer mixing models present a confusing array of software with varied inputs and assumptions.
• There is a need for a unified and flexible framework for ecological mixing model analyses.

Purpose of the Study:

• Introduce MixSIAR, an open-source R package for Bayesian tracer mixing models.
• Provide guidance for implementing mixing model analyses, including data inputs, error structures, and prior specifications.
• Enhance mixing model capabilities with covariate analysis and model selection tools.

Main Methods:

• Developed MixSIAR as an inclusive, flexible Bayesian framework in R.
• Outlined practical differences in mixture data error structures and their relation to study designs.
• Described methods for establishing informative priors and discussed their influence on model outputs.
• Detailed options for source data inputs and guidance for combining sources.
• Incorporated fixed and random effects as covariates and model selection via information criteria.

Main Results:

• MixSIAR consolidates disparate mixing model tools into a single platform.
• The package offers diversified parameterizations and improved analytical capabilities.
• Demonstrated the utility of MixSIAR with a case study on alligator diet partitioning.

Conclusions:

• MixSIAR provides a unified and advanced platform for Bayesian tracer mixing model analyses.
• The framework supports researchers in ecological diet partitioning and related studies.
• Offers a foundation for future development and improvement of mixing model applications.