Assessing non-parametric estimators of species richness. A case study with birds in green areas of the city of Puebla, Mexico

Assessing non parametric estimators of species richness. A case study with birds in green areas of the city of Puebla, Mexico. Our objective was to evaluate the performance of non parametric estimators of species richness with real data. During the 2003 breeding season, bird communities were sampled in two green areas in the city of Puebla (Mexico), and the corresponding sample based rarefaction curves were obtained. Mean data were adjusted to two non asymptotic and seven asymptotic accumulation functions, and the best model was selected by means of reliability criteria in information theory. The cumulative Weibull and the Beta P functions were the best fit models. Bias, precision and accuracy of five non parametric estimators of species richness (ICE, Chao 2, Jackknife 1, Jackknife 2, and Bootstrap) were then assessed for increasing sampling efforts (1-53 sampling units) against the asymptote of the selected accumulation functions. All the non parametric estimators here evaluated underestimated true richness most of the time, specially in one of the sites. However, after combining data from the two assemblages, only ICE, and Jackknife 1 and 2 exhibited bias below 10% with different sampling efforts, and only Jackknife 1 was globally accurate (scaled mean squared error x 100 < 5%, even with low sampling efforts, ca. 20% of the total). Therefore, we propose using the Jackknife 1 non parametric estimator as a lower bound to measure bird species richness in urban sites similar to those in the present study.