On the power of plotless density estimators for statistical comparisons of plant populations

Abstract

Reliable density estimations of plant populations and their statistical comparison are fundamental in many ecological studies. Typical sampling methods are plotless density estimators (PDE) and plot counts. A simulation study was performed to examine the power of these sampling methods for the statistical comparison of the densitiesof two plant populations. The variable area transect estimator, the ordered distance estimator, and Byth’s T-square estimator were considered and compared with the quadrat count estimator. The relative root-mean-squared error and the relative bias were used to explore the quality of the estimators, and robust confidence intervals and standard deviation estimators were given. Tests were developed to compare the population densities of two independent populations, and the performance of the tests was examined. All simulations were run for spatially random and aggregate data patterns. For completely random data, all estimators and all tests behaved well if they were based on the same sampling intensity. For the aggregate pattern, all PDE’s were negatively biased, but the quadrat count estimator was unbiased. Nonparametric confidence intervals had the most robust performance for aggregate data. The results of the presented simulation study demonstrate that plant ecologists should avoid PDE for both density estimation and statistical testing except when the pattern of a population is known to follow a completely random spatial pattern. Instead, the use of the plot count estimator is recommended.