Q̓welq̓welústen/Mount Meager Volcanic Complex, British Columbia: Inter-eruptive landslide susceptibility assessment using statistical machine learning techniques

Abstract

Over the last 60 years, six landslides with volumes from 10⁵ m³ to 5.3 × 10⁷ m³ have occurred around Q̓welq̓welústen (Mount Meager Volcanic Complex; MMVC), southwestern British Columbia, Canada. The dormant volcanic massif has experienced significant uplift leading to exposed and incised Miocene and older basement rocks and volcanic rocks associated with the MMVC. The topographic relief and variation in rock types creates a complex environment where the interplay between lithology and landslide initiation can be captured. We developed models using logistic regression and random forest methods to assess the most critical parameters affecting inter-eruptive landslide susceptibility in the area. Both approaches indicate that Devastator Peak and Plinth Peak volcanic assemblages are of notably higher landslide susceptibility. Large landslides originated in higher areas, but long runout poses the main hazard. Results support the need for detailed geological mapping for landslide susceptibility assessments in elevated volcanic massifs worldwide.