Mapping tree biomass of northern boreal forest using shadow fraction from Quickbird imagery

摘要

Forest mapping from satellite remote sensing images is a convenient approach for regions with limited or absent forest inventories. We developed and tested a method to map above-ground biomass of black spruce (Picea mariana) stands in northeastern boreal regions of Canada using high resolution satellite images. Development of the method involved: 1) calculating shadow fraction (SF) using either classification (pixel-based) or segmentation and classification (object-based) algorithms, (ii) generating linear regression relationships between SF and biomass from ground sample plots using several combinations of method parameters towards defining the best options, (iii) calculating a global linear regression applicable for all sites using the best options, and (iv) mapping biomass as a grid layer for each site using the global regression. The linear relationships were calibrated using biomass estimates of 108 ground sample plots and the shadow fraction of tree crowns calculated from QuickBird images representing three test sites. The global regression relationship produced R{sup}2, RMSE and bias in the range of 85 to 88% (except one case at 44%), 14 to 18 t/ha and -3 to 8 t/ha, respectively. The results suggest that the method may be an efficient means of mapping biomass of black spruce stands in northern Canada.