Scrutinizing functional patterns and assembly rules estimated from transect data

Transect sampling is routinely used in vegetation science. Yet, the reliability of this type of sampling for assessing community assembly rules has not been scrutinized. Local interactions and limited dispersal have essential role in the organization of plant communities. Information about vegetation patterns in the neighbourhood is probably crucial for assessing and understanding community assembly rules. The "one-dimensional" transect sampling does not provide information about adjacent vegetation patterns in the neighbourhood, therefore vegetation characteristics and assembly rules might be incorrectly estimated from transect data. In this study, alkaline, loess and sand grasslands were sampled. Presence of plant species was recorded in large elongated grids of 10 x 1040 units of 5x5cm resolution, which corresponded to a spatial extent of 50 cm x 5200 cm. Vegetation characteristics (taxon-based alpha and beta diversity and trait-based functional diversity) derived from the central 1x1040 units long transect were compared to characteristics derived from the full grid data. In the second analysis the central transect was analyzed by variography and then the patterns in the neighbouring transects were approximated by kriging. These simulated patterns were compared to the original patterns of the neighbouring transects. Community assembly rules were assessed by comparing realized patterns of functional diversity to null models. The main vegetation characteristics were similar between sampling types. Abundances and patterns of abundant species were reliably estimated from transects while the patterns of rare species varied between sampling types. Functional diversity (expressed by Rao index) was also properly estimated from transects. Local vegetation characteristics were autocorrelated within 50cm in each grasslands. Consequently, the patterns of adjacent transects, running in the close neighbourhood, could be well approximated by geostatistics. The strongest spatial dependence appeared in alkaline grasslands. Due to differences in the abundance estimates of rare species, null models derived from grids and transect were slightly different with more power in the case of grid-based data. However, the spatial structure of the dominant species and the related assembly rules did not differ between sampling types. Due to strong autocorrelations at fine spatial scales, transect sampling proved to be effective for estimating functional patterns and assembly rules.