Abstract

Landslides are a recurring problem on hillslopes throughout the Puget Lowland, Washington, but can be difficult to identify in the densely forested terrain. However, digital terrain models of the bare-earth surface derived from LIght Detection And Ranging (LIDAR) data express topographic details sufficiently well to identify landslides. Landslides and escarpments were mapped using LIDAR imagery and field checked (when permissible and accessible) throughout Kitsap County. We relied almost entirely on derivatives of LIDAR data for our mapping, including topographic-contour, slope, and hill-shaded relief maps. Each mapped landslide was assigned a level of “high” or “moderate” confidence based on the LIDAR characteristics and on field observations.

A total of 231 landslides were identified representing 0.8 percent of the land area of Kitsap County. Shallow debris topples along the coastal bluffs and large (>10,000 m²) landslide complexes are the most common types of landslides. The smallest deposit mapped covers an area of 252 m², while the largest covers 0.5 km². Previous mapping efforts that relied solely on field and photogrammetric methods identified only 57 percent of the landslides mapped by LIDAR (61 percent high confidence and 39 percent moderate confidence), although nine landslides previously identified were not mapped during this study. The remaining 43 percent identified using LIDAR have 13 percent high confidence and 87 percent moderate confidence. Coastal areas are especially susceptible to landsliding; 67 percent of the landslide area that we mapped lies within 500 meters of the present coastline. The remaining 33 percent are located along drainages farther inland.

The LIDAR data we used for mapping have some limitations including (1) rounding of the interface area between low slope surfaces and vertical faces (that is, along the edges of steep escarpments) which results in scarps being mapped too far headward (one or two meters), (2) incorrect laser-distance measurements resulting in inaccurate elevation values, (3) removal of valid ground elevations, (4) false ground roughness, and (5) faceted surface texture. Several of these limitations are introduced by algorithms in the processing software that are designed to remove non-ground elevations from LIDAR data. Despite these limitations, the algorithm-enhanced LIDAR imagery does effectively “remove” vegetation that obscures many landslides, and is therefore a valuable tool for landslide inventories and investigations in heavily vegetated regions such as the Puget Lowland.

Version 1.0

Posted November 2008