Canadian Forest Service Publications
Forest attribute estimation of northern boreal forests using Quickbird imagery and a shadow fraction method. 2012. Leboeuf, A.; Fournier, R.A.; Luther, J.E.; Beaudoin, A.; Guindon, L. Forest Ecology and Management 266: 66–74.
Available from: Atlantic Forestry Centre
Catalog ID: 33416
This study demonstrates a method to map forest stand polygons based on four forest attributes (volume, basal area, height, and crown closure) using shadow fraction values estimated from high spatial resolution QuickBird panchromatic images. The method was tested over three test sites in northeastern Canada that were largely dominated by black spruce (Picea mariana (Mill.) BSP.). The method involved four sets of procedures: (i) estimating shadow fraction from the panchromatic band of QuickBird images, (ii) generating site-specific and global regression models linking shadow fraction with each of the four forest attributes, (iii) mapping the forest attributes as a grid layer (30 X 30 m) for each test site using the global regression models, and (iv) generating stand polygons from the raster layers. Each test site was covered by a QuickBird satellite image with 0.6 X 0.6 m of spatial resolution. Between 2002 and 2004, 108 ground sample plots were acquired to develop local regression models. Goodness of fit (R2) of the linear regression models between shadow fraction and stand attributes ranged from 0.55 to 0.79. Relative root mean square error and bias estimates were, respectively, 0.32 and 0.15 for volume; 0.29 and 0.14 for basal area; 0.32 and ~0.04 for crown closure; and 0.17 and 0.05 for height. Stand volume maps were produced from both the shadow fraction method and conventional forest stand maps (derived from aerial photo-interpretation) for a test site. Volume patterns were similar, and total volume for the test site differed by only 5.6% between the two maps. Lastly, the raster images derived from the shadow fraction method were used to produce a stand map following guidelines similar to those used by provincial inventory. In all cases, our results suggest that the shadow fraction method is a reliable and convenient way to map forest stand polygons and related attributes of black spruce stands of northeastern forests of Canada.