The main anthropogenic source of CO2 emissions is the combustion of fossil fuels, while the clearing and burning of forests contribute significant amounts as well. Vegetation represents a major reservoir for terrestrial carbon stocks, and improving our ability to inventory vegetation will enhance our understanding of the impacts of land cover and climate change on carbon stocks and fluxes. These relationships may be an indication of a series of troubling biosphere-atmospheric feedback mechanisms that need to be better understood and modeled. Valuable land cover information can be provided to the global climate change modeling community using advanced remote sensing capabilities such as Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) and Airborne Synthetic Aperture Radar (AIRSAR). Individually and synergistically, data were successfully used to characterize the complex nature of the Canadian boreal forest land cover types.

The multiple endmember spectral mixture analysis process was applied against seasonal AVIRIS data to produce species-level vegetated land cover maps of two study sites in the Canadian boreal forest: Old Black Spruce (OBS) and Old Jack Pine (OJP). The highest overall accuracy was assessed to be at least 66% accurate to the available reference map, providing evidence that high-quality, species-level land cover mapping of the Canadian boreal forest is achievable at accuracy levels greater than other previous research efforts in the region. Backscatter information from multichannel, polarimetric SAR utilizing a binary decision tree-based classification technique methodology was moderately successfully applied to AIRSAR to produce maps of the boreal land cover types at both sites, with overall accuracies at least 59%.

A process, centered around noise whitening and principal component analysis features of the minimum noise fraction transform, was implemented to leverage synergies contained within spatially coregistered multitemporal and multisensor AVIRIS and AIRSAR data sets to successfully produce high-accuracy boreal forest land cover maps. Overall land cover map accuracies of 78% and 72% were assessed for OJP and OBS sites, respectively, for either seasonal or multitemporal data sets. High individual land cover accuracies appeared to be independent of site, season, or multisensor combination in the minimum-noise fraction-based approach.