Landslide Warning Systems with Short-Term Observation: Using Available Data for Adaptive Thresholds

Abstract:

Recent advances in hillslope monitoring in landslide-prone areas has improved the ability to predict potentially hazardous conditions and warn communities. In particular, hydrologic monitoring on susceptible hillslopes has become more accessible due to low-cost, low-power sensor technology and wireless data transmission. In 2015, more than 40 debris flows initiated during a period of intense rainfall in Sitka, Alaska, causing extensive property damage and killing three residents. In response, we worked with the community to implement a network of wireless hydrologic monitoring stations in 2019 and 2020. Long term (>5 years) observation will provide data for improved landslide warning thresholds and forecasting, but the community is already facing difficult decisions regarding hazard mitigation, property value and insurance issues, and emergency response plans. As Sitka and other small communities gain access to real-time data that indicate landslide potential, scientists and emergency planners must consider how to implement warning thresholds that are adaptable as we gain new information and understanding. In Sitka, we integrate the short-term hydrologic monitoring data with shallow landslide modeling (SHALSTAB), machine-learning predictions of soil saturation, precipitation records, and a limited inventory of recent landslides in Southeast Alaska. We conclude that statistical investigation of multiple incomplete data streams maximizes understanding of the physical processes that control landslide initiation. We also evaluate the predictive value of each dataset, and the shortcomings, and recommend preliminary landslide warning thresholds. This synthesis of available data provides tools that will allow Sitka to take initial actions to minimize landslide risk before long-term data are available.