Optimizing forecast-based actions for extreme rainfall events

Abstract

The last decade has seen a major innovation within disaster risk management through the emergence of standardized forecast-based action and financing protocols. Given sufficient lead time and forecast skill, a portion of relief funds may be shifted from disaster recovery to disaster preparedness, reducing losses in lives and property. While short-term early warnings systems are commonplace, forecasts at the monthly or seasonal scale are relatively underused, despite their potential value. Incorporating both, numerous relief organizations have developed operational early action protocols for natural hazards. These plans may have well-defined forecasts, trigger criteria, and identification of early actions ranging from weeks to months prior to a predicted disaster, but many have not been explicitly optimized to maximize financial or utilitarian returns. This study investigates the effect of different forecast methodologies, performance metrics, and levels of risk aversion on optimal decision-making through a sensitivity analysis of an early action protocol, using a case study in coastal Peru. Results suggest that the relative benefit of actions at different lead times plays the largest role in determining optimal decisions, with forecast methodology and risk aversion playing a lesser role. The optimization framework is designed to be applicable even in the absence of post-disaster monitoring and evaluation, supporting the proliferation of adaptive early action protocols more broadly. Plain language summaryForecast-based early actions for disasters are increasingly common, and some relief organizations have adopted standardized early action protocols to identify and respond to disasters. Because they are often new and untested, these protocols may not be optimized to provide the maximum return on investment. This paper presents a way to test different types of decisions in an early action protocol, including forecast type, willingness to take action, and ways in which to calculate benefits. We find that early preparation—that is, seasons or months in advance—is valuable, and that the value of preparation at different times before the disaster is more important than the accuracy of the forecasts or our willingness to take risks.