QUALITY OR QUANTITY? ASSESSING THE PREDICTIVE POWER OF STABLE-ISOTOPE MIXING MODELS IN AQUATIC ECOLOGY

R. R. Doucett. Stable Isotope Laboratory, Department of Biology, University of New Brunswick, Fredericton, NB, Canada E3B 6E1

Stable-isotope data are commonly incorporated into mixing-model equations to calculate relative contributions of primary food sources to top predators in complex food webs. Few studies, however, account for sampling error in all model parameters, calling into question the reliability of results. The objectives of this study were: (1) to determine the importance of terrestrial and aquatic inputs to juvenile Atlantic salmon in Catamaran Brook, New Brunswick, using stable isotopes of carbon and nitrogen and a two-source isotope mixing model, (2) to place 95% confidence limits on model output using a statistical resampling technique, and (3) to examine the predictive capacity of the model under different scenarios. When variability was incorporated into all model parameters, poor performance was observed when food sources differ by 5 per mil or less. Model performance was enhanced when the difference between the 2 sources was increased to 8 to 10 per mil, but the predictive power of the model was invariably dependent on sample size. Power curves will be presented to aid researchers in attempts to utilize stable isotopes for quantitative purposes in food-web studies.