Summary
Published in Canadian Journal of Remote Sensing., vol 39. p. 217-231. https://doi.org/10.5589/m13-030
The aim of this study was to develop an imputing method to map forest stand attributes (height and volume) using panchromatic aerial photographs with a spatial resolution of 30 cm and Landsat images. The method was tested on several black spruce-dominated sites in Québec, Canada. The method involved four procedures: (i) measuring the shadow fraction from panchromatic aerial photographs; (ii) generating regression models between the shadow fraction and forest attributes; (iii) locally mapping the forest attributes as a grid layer (30 m × 30 m); and (iv) expanding the forest attributes from the local maps to a large study area using an imputation approach. Regression models between the shadow fraction and stand attributes were calibrated with ground sample plots and a series of 73 aerial photographs.
The portion of each aerial photograph that was used was restricted to its centre (between 6.2° and 11.3°) to improve goodness-of-fit. Small local maps of stand attributes were thus produced from the shadow fraction and used as the training dataset for an imputing approach, a k nearest-neighbours algorithm, with the aim of mapping forest attributes over a large area.
Sector(s):
Forests
Categorie(s):
Scientific Article
Theme(s):
Cartography and Data, Imaging and LiDAR
Departmental author(s):
Author(s)
LEBOEUF, Antoine et Richard A. FOURNIER
Year of publication :
2012
How to get the publication :
PDF available upon request. Available at the Direction des inventaires forestiers
Keywords :
Boreal forest, Forest inventory, Mapping, Remote sensing, Stand attributes Volume
