RECENT INSIGHTS ON THE RELATIONSHIP BETWEEN LANDSCAPE FORM AND LANDSLIDING DURING EARTHQUAKES (Invited Presentation)

Strong earthquakes in steep and unstable terrain provide opportunity to study large populations of landslides generated by a common forcing. Prediction of the number and size of moderate to large landslides is one of the most elusive parameters that affects seismic risk assessments. Progress is impeded by a number of factors that are poorly described at regional scales and include the effects of amalgamation of adjacent landslides into larger features, the paucity of information on regional material properties, and the lack of 3D data with which one can more accurately characterize landslide kinematics and landslide volume. Recent developments and application of photogrammetry through camera mounted UAVs, as well as stereo-pair, high-resolution satellite imagery provides continually improving topography models, allowing the geometry of coseismic landslide populations to be described at ever-finer resolution over regional scales. These advancements contribute insight into landslide mechanics at a regional scale, which can be used to estimate coseismic landslide risk under a variety of conditions. We present examples from recent earthquakes for which we have combined satellite image mapping with UAV- and satellite- based digital terrain models to describe both 2D and 3D landscape change. It can be shown that hillslope length distributions in a given landscape closely follow simple finite slope stability models for variable Coulomb strength parameters, and agree well with field based measurements and inversion of coseismic landslide inventories for strength parameters. Based on satellite imagery, we find that the “roll over” described in normalized frequency area landslide distributions is not an artifact of image resolution and can be better understood as the median of a log-normal distribution of landslide areas. Further, the effects of amalgamation during mapping can strongly influence frequency-area and volume-area scaling parameters, but ongoing work indicates that they do not strongly influence total volume. We speculate that a common denominator among these observations is the distribution hillslope lengths/relief in a given landscape Thus landscape form (rather than local slope alone) lends some predictability as to the size and distribution of landsliding during earthquakes.