A Comparison of Point Clouds Derived from Stereo Imagery and Airborne Laser Scanning for the Area-Based Estimation of Forest Inventory Attributes in Boreal Ontario

Abstract

Point clouds derived from the photogrammetric pixel matching of 35-cm Leica ADS40 imagery (∼2.4 points/m2) were compared to those derived from airborne laser scanning (ALS; 1.1 returns/m2) in terms of their capacity to predict core forest inventory attributes at 400-m2 resolution on a boreal landscape in northeastern Ontario, Canada. These attributes described average stem size (top height, dominant–codominant height, quadratic mean stem diameter, mean stem volume) and growing stock (basal area, gross merchantable stem volume, sawlog volume, stem density), as calibrated from 426 400-m2 plots distributed across 8 forest types. Predictions derived from image-based point clouds for 10 independent validation plots in each forest type exhibited accuracies equivalent to ALS, however, some losses in precision were evident. Excluding mean stem volume and stand density, losses in precision corresponded to increases in coefficients of variation (CVs) of 4 percentage points or fewer for predicted versus observed plot values. CVs for mean stem volume and stand density increased by as many as 11 percentage points with image-based predictions. This result implies that forest inventories that are supported by an accurate, preexisting digital terrain model can be acceptably updated with optical imagery as the primary data source.