Fatty acid signatures and classification trees: New tools for investigating the foraging ecology of seals

Analysis of the fatty acid composition of milk lipids in marine mammals offers a potential means of determining changes in diet and lactation stage. However, the large number of fatty acids routinely identified (over 60) relative to the number of animals usually sampled can limit the usefulness of standard multivariate statistical models for characterizing these patterns. Classification trees or tree-based models, which are not limited by the number of variables, were used here to study the fatty acid patterns in the milk of female harbour seals (Phoca vitulina) at parturition and during lactation. Tree analyses correctly classified 44 of 51 seals based on milk fatty acid composition to four stages of lactation, which corresponded to states of fasting versus increasingly intensive feeding. The fatty acid 16:2n-6 was quite effective in differentiating between seals at parturition and those 4 days or more later. Seals were grouped into early and late lactation by fatty acid 24: 1n-9. A comparison between classification rules derived from classification trees and discriminant analysis showed that each gave similar rates of misclassification but that the latter required a method for the a priori choice of which fatty acids to analyze.