Observer‐driven pseudoturnover in vegetation monitoring is context‐dependent but does not affect ecological inference

Abstract

AbstractAimsResurveys of vegetation plots are prone to several errors that can result in misleading conclusions. Minimizing such errors and finding alternative approaches for analyzing resurvey data are therefore important. We focused on inter‐observer error and excluded other sources of variation. Our main questions were: How large is the inter‐observer error (i.e. pseudoturnover) in vegetation surveys, and can it be reduced by simple data aggregation approaches? Which factors are affecting pseudoturnover and does it vary between morphological species groups or change over time? Is ecological inference robust against inter‐observer differences?LocationSwitzerland.MethodsOver seven years, we double‐surveyed a total of 224 plots that were marked once in the field and then sampled by two observers independently on the same day. Both observers conducted full vegetation surveys, recording all vascular plant species, their cover, and additional plot information. We then calculated mean ecological indicator values and pseudoturnover.ResultsAverage pseudoturnover was 29% when raw species lists were compared. However, by applying simple aggregation steps to the species list, pseudoturnover was reduced to 17%. Pseudoturnover further varied among habitat types and declined over the years, indicating a training effect among observers. Most overlooked taxa, responsible for pseudoturnover, had low cover values. Mean ecological indicator values were robust against inter‐observer differences.ConclusionsTo minimize pseudoturnover, we suggest continuous training of observers and species‐list aggregation prior to analysis. As mean ecological indicator values were robust against inter‐observer differences, we conclude that they can provide a reliable estimate of temporal vegetation and ecological changes.