The task of identifying coincident locations in the lidar data
was facilitated by a suite of image-processing and terrain
visualization tools. We first interpolated a patch of lidar
data around the terrain feature (typically ∼50×50 m) to a
high-resolution DEMusing a Delaunay triangulation gridding
algorithm.We then attempted to identify the precise location
of the terrain point in 3-D space using terrain visualization
software (Quick Terrain Modeler, Applied Imagery). In the
cases where it was possible to confidently identify each
distinctive terrain feature (e.g. a mountain peak or ridge
point) in a shaded relief visualization of the lidar DEM, point
markers were placed onto the DEM surface at the appropriate
location (Fig. 3b). Each point marker was then imported into
a model of the raw lidar point-cloud data (Fig. 3d), and
the 3-D coordinates of the closest raw lidar point to each
marker location were extracted. In those instances where
shaded relief visualization of the terrain surface alone was
insufficient to confidently locate the GCP, we overlaid the
laser signal intensity return over the DEM surface (Fig. 3c).
These data provided additional information with which to
identify coincident points between the aerial images and the
lidar elevation data, in particular when proximal to areas of
very high (e.g. light snow surfaces) or low (e.g. ponded liquid
water) laser reflectivity. Using these methods we were able to
locate a total of 50 GCPs throughout the images comprising
the photo block (Fig. 4). Three-dimensional coordinates
extracted from the nearest raw lidar point to the GCP marker
location were assigned to each relevant control point when
measured in the aerial photographs. As some authors have
reported lidar errors to increase with off-nadir scan angle
(e.g. Baltsavias, 1999a), we selected marker locations as
close as possible to the centre of each of the nine swaths
comprising the full dataset. GCP extraction was therefore