Sampling Effects on Food Availability Estimates: Phenological Method, Sample Size, and Species Composition 1

It is currently recognized that the method used to collect phenology data can affect the resulting pattern. However, to date, the underlying influences have not been examined. To examine potential methodological biases, we investigated the effects of phenological method, sample size, and species composition on phenological patterns using data collected to estimate food availability for three primate species in Ranomafana National Park, Madagascar. Two phenological methods were compared: selected tree observations and systematic transect monitoring. By generating bootstrapped subsamples derived from the transect tree data set, we simulated two selected tree data sets and subsequently compared the observed and bootstrapped values. Although the observed values fell within the bootstrapped confidence inrervals, suggesting no significant effects of sampling protocol or sample size, additional lines of evidence suggest otherwise. Observed samples composed of different plant species, whether based on species attributes such as life‐form or categories such as food versus nonfood plants, consistently produced different phenological patterns. Wide confidence intervals of the bootstrapped samples indicate high individual variation in reproductive activity within the species sampled. Finally, we compared how well the selected tree and the transect methods represented food items used by all three primate species studied and found that transect methods sampled a wider diversity of food items, including rare foods, and a higher percentage of main primate food items than selected tree methods.