Statistical methods for estimating species richness of woody regeneration in primary and secondary rain forests of Northeastern Costa Rica
MetadataShow full item record
Chazdon, R.L., Colwell, R.K., Denslow, J.S., Guariguata, M.R. 1998. Statistical methods for estimating species richness of woody regeneration in primary and secondary rain forests of Northeastern Costa Rica . Man and the Biosphere Series No.Vol. 20. In: Dallmeier, F., Comiskey, J.A. (eds.). Forest biodiversity research, monitoring and modeling: conceptual background and old world case studies. :285-309.
Permanent link to this item: http://hdl.handle.net/10568/17965
The study of plant communities requires a basic understanding of the abundance, distribution, and number of species present. Yet, in obtaining this information, scientists can rarely sample the entire community or area of interest. In practice, data from numerous small sub-samples provide a basis for extrapolating to a larger area, Such extrapolating must take into account the well-supported observation that estimates of local species richness depend strongly on the number of individuals and the area sampled (Gleason, 1922; Preston, 1948). Although researchers must rely heavily on extrapolations for many kinds of ecological studies, relatively little attention has been focused on improving the accuracy, applicability, and accessibility of species-richness estimators in vegetation studies, particularly in higly diverse tropical ecosystems. If robuts and accurate statistical estimators of species richness that are reasonably insensitive to sample size can be found, they can serve to provide a quantitative basis for identifying conservation priorities, for comparative biogeographic or regional studies, and for assessing long-term changes in species richness. Bunge and Fitzpatrick (1993) and Colwell and Coddington (1994) provided a broad overview of statistical approaches for estimating species richness form samples. Here, we evaluated the performance of several of these methods in estimating species richness of young woody regeneration in six tropical forest sites. We compared the performance of various estimation techniques within individual sites as well as across a range of sites differing in successional status and in woody species abundance and spatial distribution. We focused specifically on two size classes of wood regeneration: 1) established seedlings <1m in height, and 2) saplings and shrubs>1m in height, but <5cm in diameter at breast height (dbh), hereafter referred to as saplings. In most standard vegetation sampling approaches, the species composition and abundance of young regeneration (seedlings and saplings) are sub-sampled in numerous small quadrats within a larger plot. These sampling methods often do not yield similar densities of stems for different regeneration size classess. Moreover, fewer trees are sampled compared to seedlings or saplings. Therefore, comparisons of species richness among size classes within the same site many be biased by different sample sizes, different numbers and spacing of quadrats, and different total areas censused. Comparisons of species richness across a range of sites may also be confounded by sampling-induced biases such as differences in area sampled, overall density of individuals, and spatial distributions (patchiness) of individuals. Our study considered effects of sample size (number of quadrats or number of individuals), overall density, and non-random spatial distributions on the performance of different species-richness estimators within and among six tropical forest sites. Before evaluating the estimators, it is important to define the evaluation criteria. We defined three features of an ideal species-richness estimator. First, it would be independent of sample size (number of quadrats or number of individuals), beyond some minimum threshold, and would remain stable as sample size increased. When plotted on the same scale as the species-accumulation curve, it should be rapidly increase to Smax and remain constant. Second, the ideal estimator would be insensitive to patchiness of species distributions across the quadrats sampled. Third, it should be insensitive to sample order