Hiperspektralna metoda badania i kartowania roślinności wysokogórskiej

Basing on recent studies with the use of hyperspectral data and on the results of vegetation characteristics comparison, it was assumed, that hyperspectral airborne images allow for identification and monitoring of chosen plant species in high mountain environment. Algorithms based on spectral mixture models give the possibility of describing pixel value as an aggregation of individual spectral responses. This helps in assessing percentage of each species cover in the plant community. Main aim of this study was the continuation of previous studies on identification of spectral characteristics of high mountain plant species, with the view of creating local spectral library of those species. Furthermore, a method of automatic identification and mapping of high mountain vegetation should be developed. The study involved analysis of 5 species: Luzula spadicea, Juncus trifidus, Calamagrostis villiosa, Deschampsia fl exuosa and Nardus stricta, and was conducted in Gąsienicowa Valley in Tatra National Park in Poland. For the assessment of possibilities of high mountain plant species identification and mapping, airborne hyperspectral images from DAIS 7915 scanner were used. Field measurements, used as a reference for airborne data, were taken with the use of GER 3700 and ASD FieldSpec spectrometers in 13 measurement locations. They were also used for analysis of spectral reflectance differences between the species. Identification and mapping of plant species were realized using 4 different classification methods, of which Spectral Angle Mapper (SAM) method gave the best results. The results were compared to test data for accuracy assessment.