Canadian Forest Service Publications
Biomass mapping of Canadian northern boreal forests using a k-NN approach and sample plots from high resolution QuickBird images. 2005. Beaudoin, A.; Guindon, L.; Leboeuf, A.; Luther, J.E.; Ung, C.H.; Côté, S.; Lambert, M.C. Paper No. 74 in Proceedings of the 26th Canadian Symposium on Remote Sensing. 14-16 June 2005. Wolfville, Nova Scotia.
Issued by: Laurentian Forestry Centre
Catalog ID: 33479
Availability: PDF (download)
The limited availability of ground sample plots (GSP) in huge and remote regions requires mapping methods with greater dependence on modeling and remote sensing. We developed and tested a method to map the biomass of black spruce (Picea mariana) stands of Canadian subarctic forest using a k-NN approach applied to Landsat Thematic Mapper imagery with sample plots generated from high resolution QuickBird images. Biomass was mapped locally using three QuickBird panchromatic images and a global regression model of above ground biomass as a function of shadow fraction (SF) of the images (reported in ). Herein, the QuickBird-derived biomass maps provided surrogates to traditional GSP for mapping biomass at the regional scale. The k-NN method was applied to the full extents of three Landsat images, representing three test sites. RMSE and bias were calculated using (i) GSP to estimate the combined error of scaling from the plot to the regional level and (ii) SSP to estimate the scaling error from application of the local biomass maps to the regional level. The RMSE of the regional scale biomass maps ranged from 10.1 to 19.6 t/ha with an overall RMSE of 17.2 t/ha based on the GSP. Bias estimates were only slightly positive with an overall bias of 2.8 t/ha for the three test sites. Application of the k-NN method using SSP produced good estimates of biomass over the three test sites with very low biases and relative errors in the order of 20-30% depending on the test-site. Further developments will consider extension of the method across large areas of Canadian subarctic forest.