INTEGRATING EMPIRICAL RAINFALL THRESHOLDS AND HYDROLOGIC MONITORING FOR DEVELOPING PRELIMINARY LANDSLIDE EARLY WARNING CRITERIA

Landslide early warning systems can be extremely valuable tools for managers of transportation organizations and emergency planners working in landslide-prone terrain. Early warning prior to widespread landsliding events can help them to prepare for disasters, reduce losses, and optimize the safety of operations. Development of effective early warning systems requires the establishment and testing of robust criteria that are directly and/or indirectly associated with increased probability of landslide initiation. Here we evaluate the utility of integrating empirical rainfall thresholds with near-real-time rainfall and subsurface hydrologic-response monitoring to inform possible criteria for an early warning system. We test this approach for the railway corridor between Seattle and Everett, Washington, where frequent shallow landslides along steep coastal bluffs often lead to costly service interruptions. The empirical rainfall thresholds are based on (1) cumulative totals of recent versus antecedent rainfall, (2) intensity and duration of the current storm, and (3) an antecedent wetness index accounting for evapotranspiration and drainage. The near-real-time, telemetered monitoring system was installed in 2015 to record rainfall, volumetric soil water content, soil water pore pressures, and shallow water table fluctuations at multiple locations along the coastal bluffs. Preliminary analysis of these data using a limited inventory of landslides that disrupted commuter railway service between August 2015 and June 2016 suggests that the monitoring system could be used to develop more informed operational decisions related to shallow landslides. Current practice is to close the railway for 48 hours following a landslide that blocks the rails, but information from the rainfall thresholds and monitoring systems could allow managers to promptly reopen service after a landslide if the risk of additional slides is deemed low, or preemptively cancel passenger rail service when the landslide risk is heightened. While thresholds can all be easily interpreted, they may be less accurate than the hydrologic monitoring data for characterizing landslide probability. However, the monitoring data may require more advanced understanding and analysis to be an effective tool for railway managers.