Comparing Eco-Phytocoenotic and Eco-Floristic Methods of Classification to Estimate Coenotic Diversity and to Map Forest Vegetation

Abstract

The coenotic diversity of forests of the model region in southwestern Moscow oblast with an area of 51 500 ha has been assessed using data from field studies, remote sensing (Landsat-5 TM, Landsat-8 OLI, and TIRS), and digital terrain models of the landscape. Forest communities are classified using two different methods: eco-phytocoenotic and eco-floristic. We recognize 15 eco-phytocoenotic syntaxa at the level of group associations and 9 eco-floristic syntaxa. The high accuracy of grouping of releves is supported statistically for each classification approach. The quality of classification is evaluated by stepwise discriminant analysis based on the representation and abundance of species. It is higher for eco-floristic syntaxa (87.1%) than for eco-phytocoenotic ones (78.9%). The adjustment of composition and names of syntaxa of eco-phytocoenotic classification ensure the compliance of typological and mapping units. The prediction quality of syntaxa recognized from pixel brightness and topographic variables is 78.6%. The quality of discriminant analysis of recognized syntaxa of the eco-phytocoenotic model show a lower accuracy of mapping model (69.7%). Large-scale maps of forest vegetation for the model region based on both classifications have been developed. It is shown that representations of eco-phytocoenotic units have a higher accuracy, as these units correspond to recent state of plant communities at their actual succession stage. On the other hand, eco-floristic units provide insight into the potential vegetation composition of a habitat. The large number of syntaxa of eco-floristic classification (associations and subassociations) made it possible to trace general patterns of vegetation on largescale maps. This feature could be more informative in medium- and small-scale mapping.