Search

Forest ecosystem models are an important tool for estimating carbon pools and fluxes in the terrestrial biosphere. Models can provide dynamic estimates through time and can predict the results of changes inforest management practices or changes in the forest ecosystem resulting from natural disturbances. Vegetation and landcover regions identified through unsupervised classification of Global Vegetation Index(GVI) data provide an appropriate ecosystem and species description for model input parameters. The timing and magnitude of photosynthesis as indicated by NDVI observed from four year average monthly GVI composites were used to identify 42 distinct regions of the former Soviet Union (FSU) . These regions represent areas of similar vegetation and land cover at a higher level of spatial detail and with more thorough species description than provided by available continental scalethematic maps. The image classes provide a consistent framework of vegetation and land-cover information across the FSU. Qualitative comparison on a pixel-by-pixel basis with detailed topographic maps and other data showed that, in general, despite the widely acknowledged problems with GVI, surface conditions were well identified by the GVI classification, The image class descriptions for the continental scaleanalysis required a supplemental description of the species specificto regional ecosystems before they could be used as a forest ecosystem model input parameter. Model predictions for carbon pools in test sites located in the Amur region of Russia compared well to carbon estimates made using other techniques. While GVI-based image classes appear to be appropriate for continental scale analysis, there is stilla need to validate the GVI classification approach using higher resolution remote sensing data and field investigations. [ABSTRACT FROM AUTHOR]