Quantifying learning in biotracer studies.

Area of Science:

• Ecology
• Quantitative Biology
• Bioinformatics

Background:

• Mixing models are essential for analyzing biotracer data, particularly stable isotope ratios, to determine dietary sources in consumers.
• Interpretation of Bayesian mixing models requires caution, as they can default to prior information when data poorly resolve source contributions.

Purpose of the Study:

• To apply information theory to quantify the learning gained from new biotracer data regarding a consumer's diet.
• To develop a power analysis approach using a priori simulations to determine optimal sample sizes for biotracer studies.

Main Methods:

• Application of information theory to assess the amount of information gained about consumer diets from biotracer data.
• Utilizing a priori simulations to perform a power analysis for determining optimal sample sizes.
• Examining two example datasets to test the proposed information theory approach.

Main Results:

• Variation in source isotope ratios inherently limits the precision of consumer diet estimates, even with large sample sizes.
• The information theory approach serves as a power analysis tool to identify optimal sampling strategies.
• Biotracer data alone have fundamental limitations in discriminating consumer diets.

Conclusions:

• Information theory provides a robust method for quantifying learning and optimizing sample sizes in biotracer studies.
• Integrating other data types, like gut content analysis, as prior information can enhance model learning.
• Information theory can guide the development of efficient sampling protocols, especially when resources or ethical considerations are limiting.