Rainfall-Induced Landslides in a Changing Climate

Abstract

Rainfall-induced landslides pose a great risk to society and cause catastrophic consequences including environmental damages, economic losses, deaths, and injuries all around the world. Such landslides are typically relatively shallow and occur frequently on hillsides with the capacity to evolve into destructive debris flows. The frequency of rainfall-induced landslides is expected to increase due to the ongoing climate change and the corresponding changes in the rainfall patterns. Moreover, the expansion of human settlement towards landslide-prone areas will increase the risks due to negative effects of human activities on slope stability and more severe consequences. The expected higher risks in the future necessitate better landslide risk assessment and management strategies. Therefore, this PhD thesis examines several topics that contribute to mitigating societal risks resulting from rainfall-induced landslides.

Development and implementation of efficient landslide risk mitigation measures rely on accurate and reliable spatiotemporal prediction of rainfall-induced landslides. Providing such predictions is often a challenging task due to the uncertainties in landslide prediction model parameters. Spatial variability of the model parameters contributes significantly to the uncertainties and limits the capacity of models to provide accurate spatial and temporal predictions. The model parameters might greatly vary over space and affect the landslide predictions. This PhD thesis examines the effects of spatial variability on the prediction of rainfall-induced landslides. This study proposes a new probabilistic three-dimensional landslide susceptibility model that accounts for spatial variability. Spatial variability of landslide model parameters is modeled by random field approach. In addition, Monte Carlo method is utilized in the developed model to quantify the effects of the uncertainties in model parameters on landslide predictions. The developed landslide model is validated using benchmark problems from literature and extensive simulations using a finite element-based program. Results of this study reveal the importance of spatial variability on the predictions of spatially distributed rainfall-induced landslides.

In addition to developing more accurate landslide prediction models, there is a need to quantify the effects of climate change on rainfall-induced landslides. Climate change is becoming more visible as climate abnormalities and corresponding catastrophic events are happening more frequently. There exist climate projections to understand how climate will change based on different socioeconomic narratives. For many countries worldwide, these climate projections display more intense and frequent rainfall events. The changes in rainfall patterns will have effects on the occurrence of rainfall-induced landslides but are not explicitly quantified. Through the quantification of the climate change impact on rainfall-induced landslides, mitigation strategies can be applied to strengthen the resilience and adaptive capacity of society to climate change and corresponding changes in landslide risk. This PhD study addresses this issue with a framework coupling climate and landslide modeling chains. A novel probabilistic framework is proposed for the integration of the modeling chains. In this framework, impacts of extreme intense rainfall events are scaled with their occurrence probability to obtain an ‘overall’ climate change impact. This approach provides a more realistic basis for the quantification of the climate change impact on rainfall-induced landslides without bias due to extreme rainfall events. Using the proposed approach, a comprehensive study is conducted on a landslide-prone study area in Trøndelag, central Norway. The study reveals the overall climate change impact on rainfall-induced landslides with increased probabilities of landslide initiations in the future.

Improved assessment of landslide risk contributes to better landslide risk management strategies. Among the different strategies relying on structural and nonstructural solutions, a landslide early warning system (LEWS) is recognized as an efficient strategy due to its lower costs and higher flexibility in comparison to alternative solutions. Landslide risk can be mitigated by these systems, which issue early warning to take necessary actions, such as evacuating people, moving mobile infrastructure, or closing road or railway sections. Landslide monitoring can support LEWSs and improve their reliability by providing consistent and reliable hazard assessments based on collected data. This PhD study examines a landslide-prone study area in Trøndelag, central Norway for the deployment of a hydrological monitoring system. The hydrological monitoring system is supported by state-of-the-art IoT-based technologies that provide efficient data acquisition and transmission. The system was deployed at two locations in the study area. The response of the slopes to seasonally cold climate conditions is monitored by volumetric water content sensors, suction sensors, and piezometers. The deployed system collected valuable information on the effects of ground freezing and thawing, rainfall, and snowmelt on the monitored parameters. A pilot study was implemented to develop an automated landslide prediction model, which integrates collected data with a physical-based landslide prediction model. This pilot study showed the potential of the collected data to be used in combination with a landslide prediction model, which can be a basis for a landslide warning model