Abstract

There is a well-reported tendency for canopy height to be underestimated in small-footprint airborne laser scanning (ALS) data of coniferous woodland. This is commonly explained by a failure to record treetops because of insufficient ALS sampling density. This study examines the accuracy of canopy height estimates retrieved from smallfootprint ALS data of broadleaf woodland. A novel field sampling method was adopted to collect reference canopy upper surface measurements of known horizontal (x, y) and vertical (z) position that had sub-metre accuracy. By investigating the z differences between ALS and reference canopy measurements with matching x and y locations, the effects of ALS sampling density were removed from the analysis. For raw point-sample ALS data, a negative bias of 0.91 m for sample shrub canopies and 1.27 m for sample tree canopies was observed. These results suggest that for broadleaf woodland, a smallfootprint laser pulse hitting the upper surface of a canopy often advances into the canopy before reflecting a signal strong enough to be detected by the scanner as a first return. The depth of laser pulse penetration will vary with canopy structural characteristics and ALS device configuration. Interpolation of the point-sample ALS canopy measurements into a grid-based digital canopy height model (DCHM) propagated the observed errors, resulting in a negative bias of 1.02 m for shrub canopies and 2.12 m for tree canopies. Here the sampling density in relation to canopy surface roughness was important.