N-alkanes are long-chain saturated hydrocarbons occurring in plant cuticles that can be used as chemical markers for estimating the diet composition of herbivores. An important constraint of using n-alkanes to estimate diet composition with currently employed mathematical procedures is that the number of markers must be equal or larger than the number of diet components. This is a considerable limitation when dealing with free-ranging herbivores feeding on complex plant communities. We present a novel approach for the estimation of diet composition using n-alkanes which applies equally to cases where the number of markers is lower, equal or greater than the number of plant species in the diet. The model uses linear programming to estimate the minimum and maximum proportions of each plant in the diet, and avoids the need for grouping species in order to reduce the number of estimated dietary components. We illustrate the model with two data sets of n-alkane content of plants and faeces obtained from a sheep grazing experiment conducted in Australia and a red deer study in Portugal. Our results are consistent with previous studies on those data sets and provide additional information on the proportions of individual species in the diet. Results show that sheep induded in the diet high proportions of white clover (from 0.25 to 0.37), and relatively high proportions of grasses (e.g. brome from 0.14 to 0.26) but tended to avoid Lotus spp. (always less than 0.04 of the diet). For red deer we found high proportions of legumes (e.g. Trifolium angustifolium and Vicia sativa reaching maxi. mum proportions of 0.42 and 0.30 of the diet, respectively) with grasses being less important and Cistus ladanifer, a browse, also having relevance (from 0.21 to 0.42 of the diet).
- linear models
- linear programming