Knowledge-based classification of remote sensing data for the estimation of below- and above-ground organic carbon stocks in riparian forests

摘要: Floodplain forests play a crucial role in the storage of organic carbon (Corg). However, modeling stocks these dynamic ecosystems remains inherently difficult. Here, we present spatial estimation Corg riparian woody vegetation and soils (to depth 1 m) Central European floodplain using very high resolution remote sensing data auxiliary geodata. The research area is Danube National Park Austria, one last remaining wetlands with near-natural Europe. Different types within show distinct capacities to store Corg. We used distinguish following types: meadow, reed bed hardwood, softwood, cottonwood forests. Spectral knowledge-based classification was performed object-based image analysis. Additional knowledge rules included distances river, object area, slope information. Five different schemes based on spectral values additional were compared validated. Validation for accuracy derived from forest inventories topographical maps. Overall higher combination spectral- than alone. While water, beds meadows clearly detectable, it remained challenging types. total quantified Monte Carlo simulation all classified types, distribution mapped. average study site 428.9 Mg C ha−1. Despite certain difficulties this method allows an indirect floodplains.