How does vegetation sampling in different parts of the growing season influence classification results and analyses of beta diversity?
Article first published online: 3 JAN 2014
© 2014 International Association for Vegetation Science
Applied Vegetation Science
Volume 17, Issue 3, pages 556–566, July 2014
How to Cite
Vymazalová, M., Tichý, L., Axmanová, I. (2014), How does vegetation sampling in different parts of the growing season influence classification results and analyses of beta diversity?. Applied Vegetation Science, 17: 556–566. doi: 10.1111/avsc.12087
- Issue published online: 17 JUN 2014
- Article first published online: 3 JAN 2014
- Manuscript Accepted: 27 NOV 2013
- Manuscript Received: 1 OCT 2012
- Czech Science Foundation. Grant Number: GAP505/11/0732
- Beta-flexible method;
- Central Europe;
- Cover transformation;
- Deciduous forest;
- Dry grassland;
- Partition similarity;
- Permanent plot;
- Sampling date;
- Seasonal variability
Several studies have demonstrated that data sets that combine vegetation plots recorded in different parts of the growing season might contain significant seasonal variability. The effects of this variability might confound some data analyses, such as vegetation classifications or beta diversity estimates, but the magnitude of these effects is unknown. Here, we try to quantify how strong these effects are, depending on the range of seasonal variation within the data set.
Southern Moravia, Czech Republic.
We used two data sets of permanent plots (Forests and Dry Grasslands from the Czech Republic, each recorded in spring, summer and autumn) to analyse the similarity of partitions in hierarchical classifications with (1) different parameter settings (transformations of cover data and the beta parameter of the Beta flexible clustering method), and (2) different proportions of plots recorded in different parts of the growing season (added non-hierarchical k-means classification).
Single-season classifications based on the summer records were mostly robust to various cover data transformations and Beta settings, whereas spring and autumn records showed high variability in the resulting partitions. The comparisons of partitions based on the same parameter settings, but using two- or three-season data sets, revealed considerable discrepancies. In the analyses comparing summer records with seasonally heterogeneous data sets, the similarity of partitions rapidly declined after the substitution of a few plots recorded in different parts of the growing season, and non-hierarchical clustering showed higher partition similarity than hierarchical clustering alone in the Dry Grasslands. Compared to single-season data sets, we found higher beta diversity when combining spring and summer plots in both Forest and Dry Grassland data sets.
The sampling date might strongly affect the results of classifications of temperate deciduous forests and dry grasslands. Therefore, for classification, we highly recommend using only summer-recorded plots. These plots are most robust with respect to various classification settings, correspond approximately to the phenological optimum of these vegetation types and are the most represented in vegetation databases from temperate regions. When the summer-recorded plots are less represented, we suggest careful seasonal stratification and the inclusion of information concerning the seasonal ratio of analysed data sets into each study.