Testing the spatial significance of weed patterns in arable land using Mead's test

There is a need in weed science for statistical tests for patchiness and spatial pattern. The objective of this study was to investigate the performance of Mead's test for detecting patterns in synthetic data and in real weed counts made in maize, and making a first assessment of its applicability in ecological studies on weeds. In an extension to Mead's test, made here for the first time, we merge original quadrat count data into rectangular cells of m by n quadrats. Care was taken to rule out the effect of starting point on the test result. Using the synthetic data, we demonstrate the ability of the test to detect both patchiness and homogeneity as deviations from randomness. The first deviation results in right-sided significance, and the second in left-sided significance of the test. Analysis of the real weed patterns demonstrated patchiness at many scales for five of the six investigated species and lack of any deviation from randomness in the sixth: Taraxacum officinale. The latter was the only wind dispersing species in the dataset. No deviation towards homogeneity was found in any of the real weed species at any scale. All patchy patterns showed anisotropy, being elongated in the direction of field traffic. As it turns out, Mead's test is well suited to detect departures from randomness in observed weed patterns and enhances the suite of diagnostic tools that can be employed by weed ecologists.