ANALYSIS OF ELEVATION CHANGES DETECTED FROM MULTI-TEMPORAL LIDAR SURVEYS IN FORESTED LANDSLIDE TERRAIN IN WESTERN OREGON

We examined elevation changes detected from 2 successive sets of bare-earth, Light Detection And Ranging (LiDAR) imagery in a 23 km² forested area in the northern Coast Range of Oregon. The first set was acquired during vegetation leaf-on (September 2007) and the second set was acquired during leaf-off (December 2007). We successfully identified and mapped active landslides using a differential Digital Elevation Model (DEM) created from the two LiDAR data sets, but to do so required the use of 1) differential elevation change thresholds (0.50 m and 0.75 m) to remove “noise” from the differential data, 2) visual pattern recognition of contiguous elevation changes reflective of typical landslide morphology, and 3) supplemental satellite imagery (Quickbird). After mapping, we field verified 87% of the mapped landslides, but we could not detect active landslides with elevation changes less than 0.50 m. We calculated the change in volume at the verified landslide sites. About 1/4 of the total volume of landslide material was missing from the landslide sites and was likely a result of post-failure erosion or poor quality bare-earth elevations interpolated from a low number of ground points in the leaf-on data. We also examined the accuracies of 285 LiDAR elevations at 4 landslide sites using GPS and total station surveys. A comparison of LiDAR and survey data indicated an overall root mean square error of 0.50 m for the 285 locations, a maximum error of 2.21 m, and a systematic error of 0.09 m. Errors were largest in areas with abrupt changes in slope as well as in areas with reduced ground-point densities. LiDAR ground-point densities were lowest in areas with dense young forests and/or mixed conifer-deciduous and deciduous vegetation, which resulted in extensive interpolations of elevations in the leaf-on, bare-earth DEM. These interpolated elevations adversely impacted the accuracies of elevation changes in the differential DEM. For optimal use of multitemporal LiDAR data in forested areas, we recommend that all datasets be flown during leaf-off seasons or, if possible, the average number of points per square meter should be increased during leaf-on acquisition so that ground-point densities are similar to those collected during leaf-off conditions.