Dealing with biases introduced by lipids in stable carbon and nitrogen isotope analyses: a solution based on 28 marine invertebrate, fish, and mammal species

ABSTRACT: Stable carbon (δ¹³C) and nitrogen (δ¹⁵N) isotope ratios are widely used in marine food web and habitat use studies. However, lipids are naturally depleted in ¹³C relative to proteins and are variable in content, biasing δ¹³C of bulk samples, with consequences for the accuracy of conclusions. This issue can be resolved either by extracting lipids from samples prior to analysis, a resource-intensive process that can also alter δ¹⁵N, or by estimating lipid-free δ¹³C using one of several equations that differ in degree of sophistication and generalization across taxa. Here, δ¹³C and δ¹⁵N were measured in bulk and lipid-extracted muscle samples from over 2000 specimens of 28 species of marine invertebrates, fishes, and mammals. Our objectives were to compare the effect of lipid extraction on δ¹³C and δ¹⁵N across taxa and evaluate the performance of 5 normalization models, overall and using subsets of species, to propose a model to revert lipid-extracted δ¹⁵N back to their bulk values and to identify the best approach for dealing with lipid-related biases. Lipid extraction caused an uneven enrichment in δ¹³C and δ¹⁵N across species. Model taxonomic specificity increased estimation accuracy for both isotopes. While models from Logan et al. (2008; J Anim Ecol 77:838-846) and McConnaughey & McRoy (1979; Mar Biol 53:257-262) were the best at predicting lipid-free δ¹³C, a linear model reliably estimated δ¹⁵N values of lipid-free samples using δ¹⁵N values of bulk samples. This study presents a method for reliably estimating δ¹³C and δ¹⁵N values of muscle tissue without resorting to duplicate analyses. This represents a major step toward the harmonization of data sets generated using bulk and lipid-extracted samples.