STRUCTURE AND DYNAMICS OF BOREAL ECOSYSTEMS: ANOTHER APPROACH TO LANDSAT IMAGERY CLASSIFICATION

Abstract

An alternative approach to information extraction from Landsat TM/ETM+ imagery is proposed. It involves transformation the image space into visible 3D form and comparing location in this space the segments of the ecosystem types with expressed graphically typology of forest and mire cover (biogeocenotic scheme). The model is built in LC1-LC2-MSI axis (the two first principal components of the image matrix in logarithmic form and moisture stress index). Comparing to Tasseled Cap, this transformation is more suitable for study area (north taiga zone of Eastern Fennoscandia). The spectral segments of mature and old-growth forests line up from the ecological optimum (moraine hills) along two main environmental gradients: i) lack of water and nutrition (fluvioglacial sands bedrock) and ii) degree of paludication (lacustrine plains). Thus, the biogeocenotic complexes are identified. The succession trajectories of forest regeneration through spectral space are also associated with the type of Quaternary deposits. For mire ecosystems spectral classes accurately reflect the type of water and mineral nutrition (ombrotrophic or mesotrophic). Spectral space model created using measured by the scanner physical ecosystem characteristics can be the base for developing objective classification of boreal ecosystems, where one of the most significant clustering criterions is the position in the spectral space.

Keywords

• boreal ecosystems
• geoinformation modeling
• multispectral imagery classification
• quaternary deposits