TRACKING BIG SUR LANDSLIDE DISPLACEMENTS IN 4D WITH SFM-BUILT TOPOGRAPHY

Active mountain-scale rockslides along the rugged Big Sur coastline in Central California move differentially in response to patterns of heavy rainfall, wave erosion, and tectonic uplift of the Santa Lucia Range. An accurate portrayal of deep-seated landslide deformation in heterogeneous Franciscan Complex rock dictates periodic imaging of the ground surface and reconstruction of displacement fields in three dimensions (3D). We test the utility of repeat fixed-wing photogrammetric surveys for quantifying 3D displacement fields from the changing position of rocks and other visible features on the landslide surface. We construct a sequence of high resolution topographic data sets using 4D Structure-from-Motion (SfM) techniques to co-align oblique image sets (8 total) collected between January and July 2017. The resulting point clouds (with a density up to 23 points/m²) capture spectacular details in the evolving landscape along Highway 1, including three landslides of interest: the Paul’s Slice / Hermitage Hill complex near Lopez Point and the Mud Creek landslide 19 km to the south.

Computed displacement fields for Hermitage Hill and the secondary, nested Paul’s Slice show ongoing deformation since January at an irregular pace. Averaged velocities for Paul’s Slice exceed those for the larger, underlying Hermitage Hill by a factor of ~2–4. Repeat differential GPS surveys corroborate spatial and temporal patterns of acceleration observed from SfM point clouds. Our results suggest two stages of acceleration for Paul’s Slice and Hermitage Hill, and the latter increase is roughly contemporaneous with epic failure of the Mud Creek landslide on May 20, 2017. This period of mutual instability is substantially out-of-phase with patterns of cumulative rainfall and suggests a broadly synchronous delay in elevated pore-fluid pressures. Notably, horizontal-to-vertical displacement ratios also vary by an order of magnitude in space and time, possibly reflecting differences in failure style. As such, we emphasize the advantage of a time-series of 3D displacement fields which, in contrast with ground surface changes from DEMs of difference, reveal velocity patterns and complex trends critical for the monitoring of landslide behavior.