observations (e.g., Band and Moore, 1995; Sorensen and
Seibert, 2007; Vaze
et al
., 2010). Slope values can be expected
to become more intermediate when finer-resolution grid cell
neighborhoods are averaged into larger cells through
DEM
coarsening. Our observation that the filtered version of a
given
DEM
resolution exhibited a smaller variance than the
unfiltered version for planform curvature suggests that low-
pass filtering has a greater impact on local surface roughness
than
DEM
coarsening by mean cell aggregation.
Classification and Interpolation of Lidar Returns
Total station surveys resulted in greater
RMSE
for locations
with terrain features, such as boulders or fallen trees, than for
non-feature survey points. This suggests that either the lidar
pulses tended to miss terrain features or, more likely, that
DEM
interpolation algorithms smoothed such features. This may
present challenges for researchers interested in using
DEMs
for
studies at the hillslope scale, or to evaluate processes depen-
dent on micro-topography. For example, hummocky terrain
and large boulders may affect surface and subsurface water
flowpaths, but our observations suggest that lidar interpola-
tion algorithms mitigate roughness from these terrain features
during
DEM
generation. Researchers interested in micro-topog-
raphy may be able to utilize a
DEM
generated from unclassi-
fied lidar returns although our research suggested it could be
difficult or impossible to distinguish boulders, hummocks, or
fallen tree boles from low-lying vegetation.
Plate 1. UAA (m
2
) computed for: (a) 1 m, (b) 3 m, (c) 5 m, and (d) 10 m
wmnf dem
s
with 3 m contour interval. Thick black line denotes the
calculated catchment boundary for each DEM.
(a)
(b)
(c)
(d)
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
May 2015
393
