Critical Zone Controls on Shallow Landslides Near Coos Bay, Oregon

Understanding what controls landslide occurrence and characteristics is important for disaster preparedness. Although previous mechanistic models have been moderately successful in predicting potential landslide locations, we still do not fully know what controls their exact size, depth, and timing. Importantly, it remains unclear where groundwater flow in weathered bedrock will generate shallow landslides through groundwater exfiltration or lateral flow convergence in the critical zone (CZ). Here, we investigate how the CZ and rainfall influence landslide occurrences by modeling soil saturation and landslide locations and sizes using different CZ structures. We look at an instrumented site (CB1) and surrounding catchments (~0.6 km2) near Coos Bay, Oregon, where past studies documented the occurrence of shallow landslides in successive years. We first model subsurface CZ structures based on 1) distance to channel and 2) least compressive stress (LCS) thresholds from the boundary element stress model Poly3D. Next, we assume that distance to channel and LCS influence the extent of subsurface weathering and hydraulic conductivity (Ksat), which were checked against measurements from a deep borehole at a narrow ridge. We explore different CZ structures using constant Ksat set by either distance to channel or an LCS threshold or variable Ksat as a function of LCS. With these structures, we predict saturation at the soil-weathered rock boundary for a given rainfall using the hydrologic model GEOtop. Then, we model landslides induced by the predicted saturations using a multidimensional slope stability model coupled with a spectral search algorithm. Preliminary analysis shows that various CZ structures induce spatially varying soil saturation patterns, which can impact shallow landslide occurrences. We also reproduce the approximate size and location of a well-monitored CB1 landslide triggered by a record storm in 1996. Using the parameters checked by our reproduction, we examine how these scenarios affect landslide size and location distributions as well as relative abundance in the surrounding catchments. Our work shows that weathered bedrock structures may be sufficiently predicted such that hydrologic models can then successfully predict spatial patterns of soil saturation and landslide hazards.