As the risks and impacts of inland flooding grow, it is increasingly important to evaluate the economic implications of alternative flood management approaches. Methods for estimating the economic benefits of structural mitigation (e. g. , levees) are well established but do not address the analytical needs of information-based approaches, such as flood forecast and early warning systems (FFEWS), which must account for how information affects protective action decisions. Meanwhile, although decision-analytic value-of-information (VOI) methods have been applied to evaluate weather, climate, and other hydrometeorological forecasts, they have not been widely used to quantify the benefits of FFEWS for flood-affected communities. The purpose of this study is to use an existing forecast and user group—the US National Weather Service’s hydrologic forecast for the Mississippi River at Saint Paul, Minnesota, and the city’s emergency managers who act on the forecast—to demonstrate the advantages and challenges of applying VOI analysis in a real-world FFEWS application. We focus on two main types of actions—temporary levee closures and relocation of impounded vehicles—and construct payoff matrixes summarizing protective action costs and flood losses from alternative combinations of action and flood levels. Using flood exceedance probabilities based on historic data, these matrixes allow us to compare the expected value of cost-plus-loss outcomes under alternative with- and without-forecast scenarios. For our preferred specification, we estimate that a perfect forecast would provide expected benefits of roughly 260, 000 per year to the city, mainly through avoided protective action costs. We then conduct a sensitivity analysis examining how benefits are affected by forecast errors, using the range of errors observed over the last 30 years, and find that the decline in value is relatively small (12% or less). We discuss how these estimates are constrained by data limitations and are sensitive to key assumptions, particularly the without-forecast counterfactual.
Houtven et al. (Tue,) studied this question.