Comparison of Bayesian and numerical optimization-based diet estimation on herbivorous zooplankton

Area of Science:

• Ecology
• Biomarker research
• Aquatic ecosystems

Background:

• Consumer diet estimation is vital for understanding trophic interactions and ecosystem dynamics.
• Fatty acid (FA) profiles serve as biotracer tools for diet analysis.
• Zooplankton play a critical role in aquatic food webs.

Purpose of the Study:

• To evaluate the performance of various diet estimation methods for herbivorous zooplankton.
• To compare Bayesian and numerical optimization approaches using FA data.
• To identify the most accurate and efficient method for estimating zooplankton diet composition.

Main Methods:

• Utilized consumer FA profiles and a resource library from feeding experiments.
• Assessed four Bayesian and three numerical optimization diet estimation methods.
• Evaluated method performance based on absolute errors, probability intervals, primary resource identification, and absence detection.

Main Results:

• All tested methods could identify the primary food resource most of the time.
• Numerical optimization methods (QFASA-CS and QFASA-KL) showed the smallest errors and best performance.
• QFASA methods accurately detected the absence of resources and were computationally faster than Bayesian methods.
• Some Bayesian methods yielded ambiguous results and had longer processing times.

Conclusions:

• Fatty acid (FA) models are effective for accurately estimating complex dietary mixtures in herbivorous zooplankton.
• QFASA-CS and QFASA-KL are recommended for their accuracy, efficiency, and reliability in diet estimation.
• This research advances the use of lipids as trophic biomarkers in ecological studies.