Determining sample size for assessing species composition in grassland¹

Abstract

Species composition is measured in grasslands for a variety of reasons. Commonly, observations are made using the wheel‐point apparatus. However, the problem of determining optimum sample size has not yet been satisfactorily resolved. In this study the wheel‐point apparatus was used to record 2 000 observations in each of a range of sample sites. A random number generator was used to select 30 sets of samples of size N (number of point‐observations) from each sample site. A measure of similarity for each sample with every other in a set was calculated to establish the variation that may be expected for that sample size. The Richards function was used to determine the sample size (number of point‐observations) required to retrieve a minimum level of precision. It is proposed that 100 point‐observations are sufficient for coarse management scale investigations, 200 for monitoring programmes and 300 for detailed scientific studies.