The analysis of long‐term changes in plant communities using large databases: The effect of stratified resampling

Abstract

AbstractQuestion:Releves in large phytosociological databases used for analysing long‐term changes in plant communities are biased towards easily accessible places and species‐rich stands. How does this bias influence trend analysis of floristic composition within a priori determined vegetation types and how can this bias be reduced by resampling methods?Location:The Netherlands.Methods:Basis for the analysis was the Dutch national phytosociological database, consisting of more than 460 000 relevés. For the Cirsio dissecti‐Molinietum, Genisto anglicae‐Callunetum and the Stellario‐Carpinetum betuli, we analysed differences in floristic composition (species frequencies) before and after 1970, using X2‐statistics. We resampled the original dataset, using different types of stratification. The results of the trend analysis with the resampled datasets were compared with the results using the original dataset.Results:In the resampled datasets a much smaller number of plant species showed a significant trend over both periods, compared to the results form the original dataset. Differences could be related to the smaller size of datasets after resampling, but also to avoiding over‐representation of certain geographical localities in the datasets. The influence of stratification was different for the three plant communities, due to the different distribution of releves in time and space.Conclusions:Analysis of long‐term changes in plant communities is strongly affected by the uneven spatio‐temporal distribution of releves in different research periods. These effects are reduced by geographical stratified resampling. The results of the trend analysis after stratified resampling are expected to give a more reliable representation of true changes, than the results after using the original data. The scale of stratification is dependent on the structure of the data.