TITLE:
Quantitative Methods for Comparing Different Polyline Stream Network Models
AUTHORS:
Danny L. Anderson, Daniel P. Ames, Ping Yang
KEYWORDS:
LiDAR, Stream Channels, Accuracy, RMSE, Sinuosity, Terrain Analysis
JOURNAL NAME:
Journal of Geographic Information System,
Vol.6 No.2,
April
4,
2014
ABSTRACT:
Two techniques for
exploring relative horizontal accuracy of complex linear spatial features are
described and sample source code (pseudo code) is presented for this purpose.
The first technique, relative sinuosity, is presented as a measure of the
complexity or detail of a polyline network in comparison to a reference
network. We term the second technique longitudinal root mean squared error
(LRMSE) and present it as a means for quantitatively assessing the horizontal variance
between two polyline data sets representing digitized (reference) and derived
stream and river networks. Both relative sinuosity and LRMSE are shown to be
suitable measures of horizontal stream network accuracy for assessing quality
and variation in linear features. Both techniques have been used in two recent
investigations involving extraction of hydrographic features from LiDAR
elevation data. One confirmed that, with the greatly increased resolution of
LiDAR data, smaller cell sizes yielded better stream network delineations,
based on sinuosity and LRMSE, when using LiDAR-derived DEMs. The other
demonstrated a new method of delineating stream channels directly from LiDAR
point clouds, without the intermediate step of deriving a DEM, showing that the
direct delineation from LiDAR point clouds yielded an excellent and much better
match, as indicated by the LRMSE.