Real-time prediction of rain-triggered lahars: incorporatingseasonality and catchment recovery

Abstract. Rain-triggered lahars are a significant secondary hydrological and geomorphic hazard at volcanoes where unconsolidated pyroclastic material produced by explosive eruptions is exposed to intense rainfall, often occurring for years to decades after the initial eruptive activity. Previous studies have shown that secondary lahar initiation is a function of rainfall parameters, source material characteristics and time since eruptive activity. In this study, probabilistic rain-triggered lahar forecasting models are developed using the lahar occurrence and rainfall record of the Belham River Valley at Soufrière Hills Volcano, Montserrat collected between April 2010 and April 2012. In addition to the use of peak rainfall intensity as a base forecasting parameter, considerations for the effects of rainfall seasonality and catchment evolution upon the initiation of rain-triggered lahars and the predictability of lahar generation are also incorporated into these models. Lahar probability increases with peak one-hour rainfall intensity throughout the two-year dataset, and is higher under given rainfall conditions in year one than year two. The probability of lahars is also enhanced during the wet season, when large-scale synoptic weather systems (including tropical cyclones) are more common and antecedent rainfall and thus levels of deposit saturation are typically increased. The incorporation of antecedent conditions and catchment evolution into logistic regression-based rain-triggered lahar probability estimation models is shown to enhance model performance and displays the potential for successful real-time prediction of lahars, even in areas featuring strongly seasonal climates and temporal catchment recovery.