Page 414 - Contributed Paper Session (CPS) - Volume 6
P. 414
CPS2007 Jai-Hua Yen et al.
sampling limitation of time or other re-sources, completely species inventories
in the wild field are almost unattainable goals. Therefore, the observed
richness in the sample always underestimates the true species richness in the
assemblage. In the literatures, among the discussed estimation approaches of
species richness, the nonparametric methods are widely used in practical
application, which include first order Jackknife approach, second order
Jackknife approach by Burnham and Overton (1978) and Chao1 (or Chao2)
lower bound estimator by Chao (1984). They all use the observed rare species
in the sample (i.e. singletons and dou-bletons) to estimate the unseen richness
in the sample. However, species identity error almost occurred in each survey
especially in vegetation sampling was ignored before and recently discussed
in the literatures by Vittoz and Guisan (2007), Burg et al. (2015), and Morrison
(2015). This identity error may seriously make observed richness biased and in
turn the estimation of true richness will be seriously biased. Therefore, without
error adjustment, the species richness estimation will be inaccurate based on
original sampling data. In this manuscript, we have proposed a modify
approach to revise the biased sampling data caused by species identity error.
From the results of simulation study in secession 3 show that our adjusting
approach can be nearly unbiased to revise the biased observed richness,
singleton and doubleton richness. Also, the richness estimators based on the
revised data effectively correct the bias caused by the species identity error.
2. Methodology
In this article, we choose Chao2 lower bound estimator for incidence data
as our species richness estimator. Since we assume that species identity error
exists in the process of sampling, adjustment of richness estimator should be
considered.
First, we need to estimate the mean species identity error rate of observer
or investigator. Plant inventories from subplot of the area which the survey is
conducted. We assume that the number of species ( ) and the categories
of species in the subplot are known by the experiment designer but unknown
by the observer who goes conducting inventories. After conducting
inventories, we have the information that the number of observed species
belongs to the subplot ( , ) and the number of observed species does not
exist in the subplot ( ,0 ). represents the record status of the survey of
species . When = 1, species i has been recorded. When = 0, species
has not been recorded. We assume the species identity error () is a random
variable follows the distribution of () with mean ̅. denotes the mean
probability that a species is misidentified into another species which belongs
to the sampling plot. ,0 equals to the number of species which is
misidentified and recorded as species do not exist in the subplot. Also, if plant
inventories of the subplot are correct, then , should be equal to
403 | I S I W S C 2 0 1 9