ИСТИНА

Hyperspectral reflectance spectra, covering the visible and near infrared spectrum with a spectral resolution of around 10 nm or finer, are finding an increasingly important role in environmental remote sensing. The much higher dimensionality of the feature space of hyperspectral compared with multispectral imagery potentially allows the discrimination of many more ground cover types and end-members for spectral unmixing approaches. Detailed spectral features can be identified, and attributed to specific biophysical or biochemical features and processes. Most hyperspectral imagery is currently collected from airborne sensors although some spaceborne imagery is already available. Interpretation and analysis of hyperspectral imagery depends on the availability of high-quality systematic in situ spectra measured for homogeneous materials. Here we describe the collection of nearly 300 spectra from subarctic vegetation in northern Europe (central Kola Peninsula, Russia, and Abisko, Sweden), spanning most vegetation groups (trees, shrubs, dwarf shrubs, grasses, mosses and lichens). The vegetation was measured in boreal forest, treeline, and mountain tundra ecosystems, and some species in the collection were sampled in a variety of growing conditions (such as altitude, moisture, degree of technogenic pollution). The data will be made freely accessible through the online library SLAP (spectral library of arctic plants. In this paper we identify common features in the spectra and propose a number of general spectral types on the basis of statistical grouping of spectra. Such grouping may allow the identification of specific vegetation types which are distinctly different from all other groups, and, conversely, find geobotanically different, but spectrally similar vegetation, as well as define some of the driving factors for the spectral similarities. We hope that results of our research could help to advance automated knowledge-based interpretation of hyperspectral imagery and vegetation mapping in polar regions.