Estimation of crown and stem water content and biomass of boreal forest using polarimetric SAR imagery

Characterization of boreal forests in ecosystem models requires temporal an d spatial distributions of water content and biomass over local and regiona l scales, In this paper, we report on the use of a semi-empirical algorithm for deriving these parameters from polarimetric synthetic aperture radar ( SAR) measurements. The algorithm is based on a two layer radar backscatter model that stratifies the forest canopy into crown and stem layers and sepa rates the structural and biometric attributes of forest stands. The structu ral parameters are estimated by training the model with SAR image data over dominant coniferous and deciduous stands in the boreal forest such as jack pine, black spruce, and aspen. The algorithm is then applied on AIRSAR ima ges collected during the Boreal Ecosystem Atmospheric Study (BOREAS) over t he boreal forest of Canada. The results are verified using biometry measure ments during BOREAS-intensive held campaigns. Field data relating the water content of tree components to dry biomass are used to modify the coefficie nts of the algorithm for crown and stem biomass, The algorithm was then app lied over the entire image generating biomass maps. A set of 18 test sites within the imaged area was used to assess the accuracy of the biomass maps. The accuracy of biomass estimation is also investigated by choosing differ ent combinations of polarization and frequency channels of the AIRSAR syste m. It is shown that polarimetric data from P-band and L-band channels provi de similar accuracy for estimating the above-ground biomass for boreal fore st types, In general, the use of P-band channels can provide better estimat es of stem biomass, while L-band channels can estimate the crown biomass mo re accurately, When AIRSAR images are also used to simulate the data from e xisting spaceborne radar systems, it was found that the combination of L-ba nd HH polarization (JERS-1), C-band HH polarization (RADARSAT), and C-band VV polarization (ERS-1) had limited capacity for mapping boreal biomass (63 % accuracy).