Assessment and Detection of Landslide-Generated Tsunamis through Numerical Modelling and Instrumental Data
Date
2024
Authors
Nemati, Fatemeh
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
This PhD dissertation focuses on improving the understanding and mitigation of landslide-generated tsunamis in coastal British Columbia. With a history of landslide tsunamis, much of the coast is at risk of future damage. Here, I demonstrate the hazard of tsunami waves triggered by a large potential landslide in the Strait of Georgia, and develop a proof-of-concept methodology for the detection of landslides and triggered tsunami waves in the Douglas Channel region, using data from a network of seismic, hydroacoustic, and bottom pressure instruments.
The first study details numerical simulations of a potential large subaerial landslide on the coast of Orcas Island and resultant tsunami waves in the Strait of Georgia. Landslide motion and tsunami generation are modelled using the non-hydrostatic physics-based NHWAVE model. The simulated failure moves downslope at up to 13.6 m/s, traveling 732 m before coming to rest after 85 s. Tsunami propagation is continued using the fully nonlinear and dispersive Boussinesq wave model FUNWAVE-TVD in a succession of layered and nested grids. In the near-source region, modelled waves have peak amplitudes of 15-20 m, current speeds up to 10 m/s, and up to 30 m runup. Significant waves occur throughout the region surrounding Orcas Island. In the tsunami propagation direction, runup reaches 7.5 m at Neptune Beach near Lummi Bay. Both initial and reflected waves cause significant runup (> 1.5 m) along much of the shoreline between Point Roberts and Lummi Bay. The findings show that significant coastal impacts may result from landslide-generated waves in the region. Such waves would arrive with little or no warning, highlighting the need to improve tsunami hazard assessment and mitigation strategies.
The second study investigates the use of seismic, hydroacoustic, and hydrostatic pressure instrumental data, to determine an optimal method for landslide detection in the Douglas Channel region, with potential future application in a system to provide early warning of landslide-generated tsunami waves. A new landslide detection method was developed, integrating a Deep-Learning AI model known as EQT with the pre-existing SSNAP earthquake detection model to form the SSNAP-EQT model. Using waveform data from a network of 8 broadband seismic stations, the model was tested to determine if it could detect and locate a number of landslides documented in the region in 2017, 2022 and 2023. The results demonstrate the effectiveness of SSNAP-EQT in detecting landslides and even microseismic earthquakes. Limitations of the existing system include gaps in the station distribution and in data availability that affect the accuracy of event detection and location. Event detections were validated through the analysis of hydroacoustic data from a hydrophone near Kitimat. Analysis of hydroacoustic spectrograms shows great promise in enabling further characterization of landslide events. Data from a bottom pressure sensor near Kitamaat Village were used to demonstrate an effective method to detect and measure potential tsunami waves. An expanded network of hydrophones, pressure sensors, and seismic stations, strategically distributed across the study area, would significantly enhance the precision of landslide detection and enable effective early warning of triggered tsunami waves.
Description
Keywords
landslide-generated tsunamis, landslide detection, Numerical modeling, Instrumental data, Orcas Island, Douglas Channel, Seismometer, Hydrophone, Pressure sensor