Résumé
RATIONALE: The bias associated with lipid contents in fish tissues is a recalcitrant topic for trophic studies using stableisotopes. Lipids are depleted in the heavy carbon isotope (13C) and the lipid content varies considerably among species,tissues and in both time and space. We have applied and assessed different correction methods for tropical tuna tissues.METHODS:We tested two types of normalization methods to deal with variable lipid content in liver, gonads, and whiteand red muscles of yellowfin, bigeye and skipjack tuna: a chemical extraction using dichloromethane and a mathematicalcorrection based on three modeling approaches (linear, non-linear and mass balance models). We measured isotopicratios of bulk and lipid-free tissues and assessed the predictive ability of the correction models with the lipid-freemeasurements. The parameters of the models were estimated from our dataset and from results from published studieson other species.RESULTS: Comparison between bulk, lipid-free and lipid-corrected isotopic ratios demonstrated that (1) chemicalextraction using dichloromethane did not affect δ15N values; (2) the change in δ13C values after extraction was tissuespecific;(3) lipid-normalization models using published parameter estimates failed to predict lipid-corrected δ13C values;and (4) linear and non-linear models using parameters estimated for each tissue from our dataset provided accurate δ13Cpredictions for all tissues, and mass balance model for white muscle only.CONCLUSIONS: Models using published estimates for parameters from other species cannot be used. Based on a rangeof lipid content that do not exceed 45%, we recommend the linear model to correct the bulk δ13C values in theinvestigated tissues but the parameters have to be estimated from a proportion of the original data for which chemicalextraction is required and the isotopic values of bulk and lipid-free tissues are measured.