Abundance‐ and incidence‐based estimation of total number of rare species in under‐sampled sites. Issue 1 (16th March 2022)

• Record Type: Journal Article
• Authors: Wu, Yongbin; Chen, Youhua; Li, Feng‐Quan; Shen, Tsung‐Jen; Nielsen, Scott E.
• Editors: Botta‐Dukát, Zoltán
• Is Part Of: Applied vegetation science. Volume 25:Issue 1(2022)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2022-03-16
• Subjects: Bayesian theory -- biodiversity data formats -- biodiversity metrics -- field surveys -- sampling theory -- species rarity; Plant ecology -- Periodicals; Plant communities -- Periodicals; Plant populations -- Periodicals; Nature -- Effect of human beings on -- Periodicals; 581.705
• DOI: 10.1111/avsc.12649
• Languages: English
• ISSNs: 1402-2001

Abstract

Abstract: Questions: Ecologists collecting field samples of biological data have a keen interest in addressing the following question: how many rare species are there in as‐yet unsurveyed additional samples? Depending on the size of a targeted additional sample, statistical models for estimating the number of rare species have not been systematically established and compared. Location: Global. Methods: For fairly comparing and predicting rare‐species richness at the same sample‐size baseline, we systematically developed and compared four estimators for rarefaction and extrapolation of rare‐species richness with a given specific abundance. These four estimators included a uniformly minimum variance unbiased (UMVUE), Bayesian‐weighted, Chao‐derived unweighted and naïve estimator. Results: After extensive numerical tests, for conducting rarefaction of rare‐species richness (i.e., when additional sample size was not larger than the original one) it is recommended to implement UMVUE, as it has zero bias and coverage percentage closest to 0.95. However, the performance of Bayesian‐weighted and Chao‐derived estimators is also satisfactory. By contrast, for conducting extrapolation of rare‐species richness (i.e., when the additional sample size is larger than the original one), the Bayesian‐weighted estimator is recommended, as it has the best performance among the four estimators (here UMVUE is inapplicable). Conclusions: There was no absolute winner, as the different estimators have their own merits and are recommended under different settings. When conducting rarefaction of rare‐species richness, UMVUE, which has the highest accuracy, is recommended. By contrast, when conducting extrapolation of rare‐species richness, the Bayesian‐weighted estimator is recommended, as it has the overall best performance. To facilitate the potential application in the comparison and prediction of rare‐species diversity using rarefaction and extrapolation techniques, an R package (fRSE) has been developed; it is freely distributed at the following URL: https://github.com/ecomol/fRSE . Abstract : We developed novel statistical estimators to predict the number of rare species that ecologists will likely encounter in further sampling. Among these methods, a Bayesian‐weighted estimator is recommended for its overall satisfactory performance and ecological relevance, seamlessly incorporating the information of compositional difference between original and additional samples in the inference. We believe the methods can help ecologists develop cost‐effective field‐sampling strategies and inform biodiversity conservation, particularly for rare species.