Linking Atmospheric Rivers with Landslide Frequency along the Northern Pacific Coast of North America

Keywords: atmospheric rivers, landslides, remote sensing, google earth engine, landsat, drone, UAS
Abstract Type: Virtual Paper Abstract

Abstract

Landslides are an important mechanism of sediment transfer in steep landscapes, acting as an ecosystem disturbance agent and posing hazards to human populations. Slope failures that lead to landslides are often triggered by heavy precipitation events, such as those during atmospheric rivers (ARs), which transport large amounts of moisture landward. However, our ability to predict landslide occurrence based on hydroclimatic, topographic, and other environmental parameters has been limited by the spatiotemporal availability of reliable landslide databases. Here, we focus on landslide occurrence modeling across the coastal British Columbia and Southeast Alaska, where ARs are known to bring extraordinarily heavy pulses of precipitation. To this end, we mapped landslide boundaries across the region using Landsat (TM,ETM+, and OLI) surface reflectance data collected between 1985 and 2022, within the Landtrendr algorithm in Google Earth Engine (GEE). Annual landslide boundaries were used to model the drivers of regional landslide frequency and determine the potential impact ARs had on historical landslide occurrence. The accuracy of remotely detected landslides was evaluated using delineated landslide boundaries within the Tongass National Forest, Alaska, and drone-based field surveys in British Columbia. Preliminary results indicate that our remotely sensed landslide mapping had moderate correspondence with the reference datasets. This ongoing research will contribute to our ability to anticipate the frequency and magnitude of natural disasters in response to changing climatic conditions.