Observer and relocation errors matter in resurveys of historical vegetation plots

Aim Revisits of non‐permanent, relocatable plots first surveyed several decades ago offer a direct way to observe vegetation change and form a unique and increasingly used source of information for global change research. Despite the important insights that can be obtained from resurveying these quasi‐permanent vegetation plots, their use is prone to both observer and relocation errors. Studying the combined effects of both error types is important since they will play out together in practice and it is yet unknown to what extent observed vegetation changes are influenced by these errors. Methods We designed a study that mimicked all steps in a resurvey study and that allowed determination of the magnitude of observer errors only vs the joint observer and relocation errors. Communities of vascular plants growing in the understorey of temperate forests were selected as study system. Ten regions in Europe were covered to explore generality across contexts and 50 observers were involved, which deliberately differed in their experience in making vegetation records. Results The mean geographic distance between plots in the observer+relocation error data set was 24 m. The mean relative difference in species richness in the observer error and the observer+relocation data set was 15% and 21%, respectively. The mean “pseudo‐turnover” between the five records at a quasi‐permanent plot location was on average 0.21 and 0.35 for the observer error and observer+relocation error data sets, respectively. More detailed analyses of the compositional variation showed that the nestedness and turnover components were of equal importance in the observer data set, whereas turnover was much more important than nestedness in the observer+relocation data set. Interestingly, the differences between the observer and the observer+relocation data sets largely disappeared when looking at temporal change: both the changes in species richness and species composition over time were very similar in these data sets. Conclusions Our results demonstrate that observer and relocation errors are non‐negligible when resurveying quasi‐permanent plots. A careful interpretation of the results of resurvey studies is warranted, especially when changes are assessed based on a low number of plots. We conclude by listing measures that should be taken to maximally increase the precision and the strength of the inferences drawn from vegetation resurveys.