Options for using remote sensing and its reliability in structural and spatial determination of forest ecosystems

Researches on vegetation cover are multidisciplinary and rely on scientific achievements in biogeography, landscape and conservation ecology, physical geography, phytocenology and other scientific disciplines, and the constant development of information technologies has led to a growing importance of remote sensing technique and methodology, combined with GIS technology, in regards to identification, valuing and monitoring of forest ecosystems. Structural and phenological variations between forest communities, and tree species they are composed of, create characteristic spectral images which can be recorded by sensors. Processing and analysis of these images by specialized software offer various useful information about development and distribution of forest communities. This paper presents options for using remote sensing techniques (supervised and unsupervised classification) for preparing vegetation maps, where Landsat images were used to determine vegetation cover. The level of reliability and applicability of this methodology was tested by comparison of areas under forests and clearings, areas under individual stands, and by analysis of their spatial coincidence with standard terrestrial methods of determination thereof. A reference area used in the process was the Management Unit “Topolik” managed by the Forest Estate “Novi Sad”. As for unsupervised classification, the area under forests that resulted from images is smaller by 31.52 ha, areas under clearings is larger by 38.60 ha, with lower spatial coincidence of these categories compared to terrestrial methods. Supervised classification offered better results in regards to identification of these categories – area under forests is smaller by 12.28 ha аnd the area under clearings is larger by 14.36 ha. When it comes to stands, identification by using the image diverts substantially compared to the situation in the field. Differences are two-way and range from 861.98 ha for poplar plantations to +3.96 ha for devastated ash forests, with significantly different spatial distribution. The use of higher quality images, hardware and software would generate more reliable and more applicable results, which is the intention when it comes to implementation of modern technologies for data collection in the forestry of Serbia.