Everyone makes mistakes: Sampling errors in vegetation analysis - The effect of different sampling methods, abundance estimates, experimental manipulations, and data transformation

Abstract

Understanding of causes for recent changes in vegetation structure and species richness of natural habitats is crucial for their maintaining for future generations. However, to avoid misinterpretation of vegetation changes in time, we should be aware of limits and errors of methods used for vegetation sampling. In a specific vegetation type, i.e. species rich wet meadow, we quantified sampling error in vegetation sampling at four different sampling levels (visual cover estimation, detailed recording in a grid of small cells, detailed assessment during clipping for biomass, biomass sorting), compared differences among three abundance estimates (frequency, cover and biomass), and assessed the effect of data transformation, and rapid change in vegetation structure caused by experimental species removal. At the 1 m2 scale the captured proportion of species missed by classical relevé sampling was on average 16%. Subsequent detailed subquadrat sampling captured the majority of previously overlooked species. The chance of a species being overlooked increased both with rarity, and the species richness of the area sampled. Where abundance was measured using metrics of cover and biomass, common species were overvalued, but when abundance was measured using frequency, common species were undervalued. In this study, logarithmic transformation of values provided a more reliable characterization of vegetation, than binarized or untransformed values. With the exception of species abundance, the number of species overlooked, quadrat species richness, and vegetation characterization were all affected by the experimental treatment. Our findings highlight the potential effects of error when conducting vegetation sampling and analyses of community dynamics. Due to these effects, we need to consider the reliability of conclusions drawn when assessing temporal changes in plant dynamics. Data transformation modifies the effect of sampling error in analyses of vegetation data.