Counts versus categories: choosing the more appropriate weed scoring method

Abstract

Weed sampling can be a time consuming and repetitive task if populations are to be adequately monitored. In situations where there are a range of species densities, high numbers of samples are required, and/or where the sample area is large, it may be appropriate to group weed densities into ordinal categories. An ordinal system is outlined in which weed densities were categorised into 8 groups. Weed species composition and density were assessed using the category and count methods, on 10 fields in northern New South Wales. A multinomial distribution was used within a generalised linear model framework, to assess the validity of the category method. A simulation study was also conducted to assess the performance of the 2 scoring methods. The main difference between the count and category predictions is that count data produce a population mean, whereas ordinal scores provide a mean within a categorical range. Collecting categorical data is less time consuming, particularly for higher weed densities. However, the simulations showed that to determine significant differences between population means required more samples using the category method (15–30) compared with less than 20 for the count method, assuming a satisfactory study power of 0.9. The category method is reliable, particularly where general trends or baseline information are required from the data. When the population density is low and more detailed information is required, counts are a more appropriate methodology.