Using Landslides Induced by Earthquakes for Paleoseismology, Hazard Prediction, and Spatial Analysis

Abstract

Landslides are a dangerous natural hazard that can kill people and damage property. Mass movement on hillslopes also erodes mountains and shapes landscapes over the longer term. By studying both landslides and their causes, we can better understand how they influence the environment and how to minimize the risk they pose to people and property. Landslides can be caused by shaking from an earthquake. Unlike landslides induced by other forces such as rainfall, earthquake induced landslides all happen at roughly the same time, which can amplify their hazards and impact on the landscape. I use this pulsed nature of earthquake induced landslides to investigate past earthquakes on the Seattle Fault. The Seattle Fault represents a major hazard for the city of Seattle, but has not experienced a major earthquake since approximately 1000 years ago. Because of this, earthquakes on the fault are poorly understood. We date past landslides in the area using a novel roughness based landslide dating technique. By looking for times when more landslides occurred than normal in places that would likely have strong shaking from an earthquake, I find landslide clusters that likely represent past earthquakes on the Seattle Fault. This method can be generalized to other areas prone to landsliding and infrequent but strong earthquakes. Next, I examine different scenarios to predict where a modern Seattle Fault earthquake would induce landslides. By combining the ground motions from 18 different scenario Seattle Fault kinematic earthquake models and a multi-modal landslide hazard model, I find where landslides are likely to occur and what controls these patterns. Over all scenarios, I find that there are always high numbers of modelled landslides (10% of slopes failing with translational landslides, and 4% of slopes failing in rotational landslides) and that these landslides mostly occur in the same places as modern rainfall induced landslides. Between scenarios, water saturation and local geologic strength are more important factors than earthquake properties such as hypocenter and magnitude. Finally, I investigate the assumptions behind the roughness based landslide dating technique used previously. I measure the roughness of earthquake induced landslide inventories from both the 1994 Northridge CA, USA and 2016 Kaikoura, NZ earthquakes. The roughness based dating method assumes that these landslides are all about the same roughness when they occur, however I find that these landslides actually have a wide range of initial roughness values. The range of initial roughness values can be reduced by filtering landslides by geology and landslide size. I also test how landslide deposits smooth over time by comparing repeat lidar surveys of the 2014 Oso, WA, USA landslide over a ten year period. These data show that geology also plays an important role in determining the rate at which the landslide smooths. Overall I found that when different geologic units are isolated and analysed separately, the variation in roughness in large landslides (Area greater than of equal to 10,000 sqm) would lead to a variation in estimated ages of 500 and 200 years. This is similar to the error (200 years) used in Chapter 2 and validates the use of this technique for landslide age estimation.