Premium
Assessing vegetation change using vegetation‐plot databases: a risky business
Author(s) -
Chytrý Milan,
Tichý Lubomír,
Hennekens Stephan M.,
Schaminée Joop H.J.
Publication year - 2014
Publication title -
applied vegetation science
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.096
H-Index - 64
eISSN - 1654-109X
pISSN - 1402-2001
DOI - 10.1111/avsc.12050
Subject(s) - vegetation (pathology) , plot (graphics) , database , multivariate statistics , vegetation classification , geography , physical geography , statistics , mathematics , computer science , medicine , pathology
Aim Data from vegetation plots can be used for the assessment of past vegetation change in three ways: (1) comparison of old and new records from permanent plots established for vegetation monitoring; (2) revisiting historical phytosociological plots and subsequent comparison of old and new records; (3) comparison of large sets of old and new phytosociological records from the same area but different plots. Option (3) would be the cheapest in regions where large vegetation‐plot databases are available, but there is a risk of incorrect results due to a spatial mismatch of old and new plots. Here we assess the accuracy of such analyses. Methods We used three data sets of permanent plots from C zech mountain bogs and D utch oak forests and heathlands to quantify vegetation change. We selected subsets to simulate analyses based on (1) data from permanent plots or revisited phytosociological plots, i.e. containing old and new records from the same plots, (2) vegetation‐plot databases with old and new records from different, randomly selected sites, and (3) vegetation‐plot databases with old and new records from different but close sites. We repeated each subset selection 1000 times and analysed vegetation change in each of the three data sets and each variant of subset selection using permutational multivariate analysis of variance. Results For data sets with no actual vegetation change, analyses of some subsets simulating vegetation‐plot databases incorrectly suggested significant changes. For a data set with real change, a change was detected in analyses of simulated vegetation‐plot databases, but in several cases it had a different direction or magnitude to the real change. Conclusions The assessment of vegetation change using vegetation‐plot databases should be either avoided or interpreted with extreme caution because of the risk of incorrect results. Analyses such as these may be used to propose hypotheses about past vegetation change, but their results should not be considered valid unless confirmed using more reliable data. In many contexts, re‐visitation studies of historical phytosociological plots may be the best strategy to assess past vegetation change, while new networks of carefully stratified permanent plots are preferable for monitoring future change.