PARENT SESSION
Oral Session # 85: GIS and Remote Sensing II; Ecological Modeling III.
Presiding: Y Lin
Thursday, August 7. 1:30 PM to 5:00 PM, SITCC Meeting Room 203.
Estimating carbon storage for the boreal forest of St. Petersburg, Russia using satellite imagery.
Oetter, Doug*,1, Krankina, Olga2, Cohen, Warren3, Maiersperger, Thomas2, 1 Georgia College & State University, Milledgeville, GA2 Oregon State University, Corvallis, OR3 USDA Forest Service, Corvallis, OR
ABSTRACT- The boreal forests of northwest Russia play a critical role in sequestering global atmospheric carbon. Estimates of their current and potential carbon storage are important for modeling worldwide carbon sinks and sources. Using detailed forest inventory data in conjunction with multiple-date Landsat satellite imagery, we predicted regional forest age and above-ground biomass for Russia's 76,860 km2 St. Petersburg Region. The remote sensing analysis included multiple-image rectification and normalization, creation of a land cover map, and application of advanced regression techniques to estimate forest characteristics. Model development and map error assessment were based on field measurement of over 1500 forest ground plots, collected in 1992-93 by the Northwest State Forest Inventory Enterprise. Twelve Landsat Thematic Mapper (TM) images were used to map the full region. Following geographic registration and scene-to-scene radiometric normalization, unsupervised classification was used to create a 13-class land cover map. For the forest classes, forest inventory data were used to build regression models to predict mean forest age and above-ground biomass for each forested pixel. A canonical correlation analysis index of the TM bands was regressed against the vegetation variables using the reduced major axis technique. Regional estimates of forest cover and total biomass were in close agreement with ground-based estimates. Results were assessed against an independent subset of the forest inventory plots. Error assessment indicated that predictive accuracy varied by forest composition, and that the overall error was acceptable. This satellite-based forest mapping method proved to be an effective tool for estimating carbon storage for a large boreal forest region.
Key words: Landsat Thematic Mapper, remote sensing, boreal forest, carbon storage