The use of numerical classification methods and vegetation plot databases opened new avenues towards the understanding of vegetation patterns on broad spatial scales. Until now such works relied predominantly on species occurrences or cover-abundances as descriptors of individual stands (recorded as sample plots). Recent centuries in the field of ecology reinforced the view that, instead of species identity, the traits of plants constituting a community are more important for understanding how ecosystems function. Traits are measurable attributes of plant individuals which influence their survival, growth or reproduction success. In our study, we investigate the possibility of using plot-level averages of trait values (so-called community-weighted means, CWMs) to classify vegetation. We compare a trait-based classification with the species-based syntaxonomical system which is a widely used reference for vegetation typology in Central Europe.
Our study system includes wet and mesic grasslands of Poland. Such grasslands are widely distributed across Poland. In most cases, they had been created by man on forest clearings, and maintained by regular mowing and/or by pasturing for several decades or even centuries. Besides their agricultural importance, they have a high priority for conservation since they used to be species rich and host a number of rare plants. For the analysis, we used 6985 vegetation plots stored in the Polish Vegetation Database. This data provide lists of species and their cover-abundance information for each plot. We compiled species-level averages of trait values from public databases LEDA and Clo-Pla. The choice of traits for any trait-based analysis is crucial in understanding ecosystem properties, since different traits reflect different ecological functions. We included five traits representing basic plant strategies: specific leaf area (related to resource acquisition), canopy height (related to competitive ability), seed mass (linked with reproduction), clonal spread (reflecting ability to avoid disturbance), and bud bank (a proxy for the ability to regenerate after minor damage). After calculating trait CWMs for each plot, we used numerical methods to classify them in a way that clusters contain plots with similar CWMs. We compared these clusters in terms of their ecological background and geographical distribution, and investigated their correspondence with existing syntaxonomical units.
The overall correspondence of the trait-based classification with syntaxonomy was moderate. The classification mirrored the main gradients structuring grasslands in Poland, although some types with the strong dominance of functionally unique species appeared more distinct than how they are treated in syntaxonomy. On the other hand, plots which were dominated by different species possessing similar trait values were not separated in the trait-based classification. We developed an index relying on functional uniqueness and variation in the dominance of species across plots to predict their effect on the classification. We conclude that numerical classification based on CWMs of traits is a promising approach for delineating vegetation types with stronger correspondence with ecosystem properties. Further investigation of different vegetation types and different ecological and spatial coverage should gain further insight into the potentials and limitations of trait-based vegetation classification.
This is a plain language summary for the paper of Lengyel et al. published in the Journal of Vegetation Science (https://doi.org/10.1111/jvs.12850).