This presentation was held at the DUE Permafrost 2014 workshop, 12th of Febuary 2014, Frascati, Italy
Workshop: climate-cryosphere.org/meetings/due-permafrost-2014
Siewert, Matthias and Gustaf Hugelius
Department of Physical Geography and Quaternary Geology, Stockholm University,, Stockholm Sweden
Email:
Soils in periglacial landscapes have been identified as a major pool of soil organic carbon (SOC). To map SOC, soil pedon data is commonly upscaled using landcover classifications (LCC). Higher spatial resolution of the satellite imagery input will increase the landcover classification accuracy, but as imagery with very fine resolution is used, traditional LCC approaches become problematic. When discrete landscape objects are represented by many pixels, no unique spectral signature can be assigned. Here a case study from the Russian Arctic tundra is presented. Spatial autocorrelation of soil pedon data and satellite images are analyzed using semivariograms to determine a minimum sampling distances for the soil pedon data. Furthermore, postprocessing is investigated to improve LCC accuracy and textural characteristics of the satellite images are analyzed using local variance plots to find an optimal spatial resolution for LCCs.
Keywords: Soil organic carbon, upscaling, resolution, landcover classification