PRECISION IN MONITORING PLANT-SPECIES COMPOSITION IN MONTANE GRASSLANDS

Sampling distributions of species composition counts from four large data sets were collected by point techniques. Results show good agreement between observed and expected frequencies under binomial distributions for all species measured with the Tidmarsh wheel. However, with the Levy bridge where observations are not independent, greater variance occurred for species with low frequencies. The results were used to specify those sample sizes needed to achieve any pre-determined precision. The traditionally accepted sample size of 200 is adequate for detecting a change with 20% precision in only the dominant species (mean composition of over 25%). Detecting change in species with a frequency of less than 5% will require in excess of 1200 points to achieve a 20% precision. © 1990 Kluwer Academic Publishers.