Cost-sensitive rainfall thresholds for shallow landslides

Abstract The risk management of rainfall-induced landslides requires reliable rainfall thresholds to issue early warning alerts. practical application these often leads misclassifications, either false negative or positive, which induce costs for the society. Since missed-alarm (false negative) and false-alarm positive) cost may be significantly different, it is necessary find an optimal threshold that accounts minimises such costs, tuning rates. In this paper, we propose a new methodology develop cost-sensitive thresholds, also analyse several factors produce uncertainty, as accuracy intensity values at landslide location, time occurrence, minimum amount define non-triggering event, variability scenarios. Starting from detailed mapping occurred during five large-scale events in Italian Central Alps, first developed curves with ROC-based approach by using both rain gauge bias-adjusted weather radar data. Then, based on reference scenario quantified items missed alarms alarms, cost-curve (Drummond Holte 2000). Finally, studied sensitivity items. study confirms how important information regarding site development thresholds. Although use bias-corrected strongly improves values, large uncertainty related exact occurrence still remains, negatively affecting analysis. Accounting different misclassification because combinations make increase decrease convenient. our scenario, most convenient lower than seeks minimise number whereas are almost seven times greater costs. However, scenarios, vary significantly, much half order magnitude.