Understanding the mechanisms of landslide path dependency (LPD): the role of topography, post-landslide soil properties, and earthquakes

Abstract

Landslides are one of the most important geohazards, causing widespread damage to human life, infrastructure, and ecosystems. They commonly occur in mountains and are controlled by a combination of conditioning factors, such as slope morphology, soil, and geology, as well as triggering factors, including rainfall, earthquakes, and volcanic eruptions. In addition to these controlling factors, a recent study demonstrates that landsliding is also dependent on its history. One such process in landslide-prone areas is landslide path dependency (LPD), which was detected in Collazzone (Italy) and later confirmed in central Nepal. LPD refers to the tendency of earlier landslides to cause subsequent landslides through one or more legacy effects. Although LPD was detected, the mechanism controlling it remains unclear. This dissertation investigates the mechanisms behind LPD by examining the impacts of topography, post-landslide alterations in soil physical properties, and earthquakes. The study areas for this dissertation are Collazzone in the Umbria region (Italy) and the Nepal Himalayas. In Chapter 2, results show that LPD is stronger in downslope directions than in lateral directions. This pattern is likely due to a landslide soil-feedback mechanism controlling subsequent landslide processes. In Collazzone, many landslides occurred on slopes with concave downslope and lateral profiles, so landslide deposits remain on the hillslope rather than being fully removed. During heavy rainfall, water accumulates and pore pressure increases in these deposits, enhancing downward susceptibility and therefore greater LPD in the downslope direction. In Chapter 3, the impacts of landslides on soil physical properties were investigated. Results show that soil organic carbon (SOC) stocks and surface soil resistance are significantly lower in landslide-affected soils than in adjacent stable soils immediately after failure. However, these differences gradually decrease with landslide age, and some soil properties within the landslide show a convergent trend with those of the adjacent soil within approximately 7-10 years. Chapter 3 results show that landslides transiently impact soil properties, creating conditions favorable for LPD. In Chapter 4, the impacts of extreme events, i.e., earthquakes, on LPD were investigated. To study the relationship between earthquakes and LPD, the Collazzone inventory (1937- 2014) was divided into pre- and post-earthquake periods associated with the 1997 and 2009 earthquakes. The results show that earthquakes in Collazzone did not affect rainfall-triggered landslide occurrences; however, they strengthened the overall LPD by increasing overlap between successive landslides, i.e., landslides tend to overlap more with earlier landslides after an earthquake. One possible explanation is that shaking intensity, as measured by the Modified Mercalli Intensity (MMI), was 4-5 (light to moderate shaking), indicating that these earthquakes, rather than producing widespread hillslope damage, preconditioned areas near earlier landslides, resulting in more overlaps and thus a greater LPD effect. To test whether this impact is consistent across different geomorphic settings, a similar analysis was conducted in Nepal following the 2015 earthquakes. In contrast to Collazone, the Nepal earthquakes were stronger and had a greater impact on the MMI scale (7-8.6, very strong to severe shaking). Correspondingly, rainfall-triggered landslide numbers were higher in Nepal in the few years after the 2015 earthquakes than in the pre-earthquake period, decreasing to pre-earthquake levels in 4 years, but the impact on LPD was low. One possible reason is that the 2015 earthquakes caused more direct hillslope damage wherever they were felt, triggering additional landslides, even in previously unaffected areas, resulting in a low apparent LPD. The conceptual model that emerges from this is that earthquakes under light to moderate shaking conditions can either precondition slopes near existing landslides and thereby strengthen LPD or, under strong shaking, may cause greater hillslope damage and trigger landslides even in previously landslide-unaffected areas, leaving LPD unaffected. Overall, this dissertation demonstrates that LPD is controlled by a combination of topographic, soil, and seismic processes. Topography and post-landslide-altered soil properties create a landslide-soil feedback loop that favors LPD, while earthquakes modify overall LPD.