MIRE SITE TYPE MAPPING OF BOREAL PEATLANDS WITH HYPERSPECTRAL AIRBORNE HYMAP IN NORTHERN FINLAND

• Main Authors: Hilkka Arkimaa, Maarit Middleton, Eija Hyvönen, Viljo Kuosmanen, Jukka Laitinen, Raimo Sutinen
• Other Authors: The Pennsylvania State University CiteSeerX Archives
• Format: Text
• Language: English
• Subjects: Northern Finland
• Online Access: http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.381.7697; http://www.earsel.org/workshops/IS_Warsaw_2005/papers/Geology/35_Arkimaa_329_336.pdf

Peatlands are important ecosystems in the cool and humid boreal climatic zone covering over one third of the landscape in northern Finland. Classification of mire site types, based on soil nutrient and water regimes driven plant communities, provides primary information for peatland inventories and ecological studies. In this study, the treeless pristine mire site types were classified from high resolution (5 m) airborne hyperspectral (126 bands, 450–2450 nm) data. The geo- and atmospherically corrected HyMap TM imagery was processed with hyperspectral mapping methods and Geological Survey of Finland (GTK) peatland inventory data was used for ground referencing. The preliminary results of signature separability and mixture tuned matched filtering analyses resulted in a 61.3 % overall classification accuracy (kappa 0.526) of the four most common treeless pristine mire site types. Carex (sedge) fens (mire site types: ‘tall-sedge fen’, ‘flark fen’) were easily distinguished from Sphagnum bogs (‘Shagnum fuscum bog’, ‘low-sedge bog’). The abundance of sedge, herbaceous and broad-leaved species along with the wet surface conditions in the Carex fens produced both a prominent green peak and red well, and lower NIR and SWIR reflectance values as compared to those of Sphagnum bogs. High heterogeneity of peatland surface conditions, even within a single mire site type, is one of the contributors to the classification uncertainty. Future research will involve a detailed vegetation inventory and surface moisture measurements directed at revealing the true effectiveness of hyperspectral remote sensing in peatland surface mapping.