Spatial modelling-based approach to phytogeographical regionalization using grassland vegetation data

Abstract

We presented a methodology for drawing continuous boundaries in the landscape differentiating between regions with different floristic composition. A region in Central Slovakia covering 2,445 km2 was investigated. Ecological indicator values for temperature (EIT) in 1,978 grassland polygons were analysed. Ordinary kriging was used to interpolate EIT across the study region. Lattice wombling was used to identify the most intensive gradients in EIT and to draw boundaries, while ANOVA was used for post-classification analysis. A strong pattern of spatial continuity was present in EIT assigned to species in grassland polygons allowing for drawing continuous boundaries in the landscape. The study region was divided into 15 districts using the proposed method. Post-classification analysis indicated that 17 out of 23 adjacent districts were found to differ significantly in term of mean value of source samples. The results implied the need for incorporating spatial autocorrelation in sample data into post-classification analysis; such factor is often neglected in ecological research. The presented findings suggested broader applicability of the proposed method for spatial modelling, as vegetation data is widely accessible in databases for many regions of Europe.