ИСТИНА

The object of our research is the diversity and spatial structure of herbaceous vegetation (meadows and semi-ruderal herblands) on the surrounding area of Polistovsky Reserve. We’ve developed the vegetation classification and attempted to model the distribution of communities using GIS and remote sensing (RS) technologies. GIS and RS technologies open a vast perspective for plant community researches. Remote data allow to map the communities and reflect their features (terrain, photosynthetic activity, productivity and biomass, etc). So understanding the place of each certain community within the dimensions of remote sensed data help us to get ecological “portrait” of them. The technology was effective used for forests, but almost not tried for grassland communities, because 1) the spatial resolution of regularly available satellite images was not enough for grassland communities till 2016, and 2) along with the vision and infrared data, terrain is also critically important for community distinguishing - but the resolution of available digital elevation model is also not enough for grassland. Nowadays the improvement of data and techniques availability probably opens the new era for using RS for grassland researches. We use new available data (Sentinel-2 satellite images) and plan to include drone’s data for understanding spatial distribution of grassland communities. We used the data of 205 grassland releves (the point was marked by GPS) from the surroundings of Polistovsky state natural reserve, Pskov region, Russia. The syntaxonomy of grassland communities was developed with Juice 7.0 . As RS data we used Sentinel-2 images and drones data which have been got while the field research and processed with Pix4D.com. Using ArcGIS 10.5.1, we extracted the remote sensing data (the normalized Sentinel pixel values, Kawth-Thomas and NDVI indices values), then using R 3.4.4 (Mann-Whitney test) analysed the distinguishing between syntaxa of some syntaxonomic ranks (from classes to associations). Position of the communities within Sentinel and drone data dimensions was also analysed with discriminate function analysis and Random Forest predictive algorithm. The diversity of herbaceous vegetation identified as 14 associations, 7 variants and 1 facie from 4 classes: Molinio-Arrhenatheretea, Phragmito-Magnocaricetea, Epilobietea angustifolii, Artemisietea vulgaris. The discriminate function analysis shows some uniqueness of classes in the dimensions of Sentinel-2 bands and vegetation indices. But the optical data is not enough for accurate distinguishing and modelling the spatial distribution of communities. We’re waiting for crucial improvement of our result using drone terrain data and hope to share result and technology details by conference report. The reported study was funded by RFBR according to the research project № 18-34-00786.