Decision-related sensitivity analysis of a flood risk model assessing the benefit-cost ratio of a flood polder at the Bavarian Danube

Floods are one of the most hazardous natural phenomena in Germany, which calls for thorough and comprehensive flood risk mitigation strategies. Considering this, a series of controlled flood detention basins, so-called flood polders, are planned along the Danube as one part of the Bavarian flood protection program 2020plus. By means of a reduction of the peak discharge of large flood events, flood polders can reduce the load on downstream located flood protection structures and therewith lower the probability of dike breaches.

A cost-benefit analysis is used to evaluate the economic efficiency of a flood polder location. Therein, the costs of construction and maintenance of the polder are compared with the expected monetarized flood risk reduction over the lifetime of the polder. However, assessing the economic consequences of flood risk on a trans-regional level to quantify the benefit of the flood protection measure is challenging and subject to significant uncertainties. These include uncertainties on the characteristics of the hydrological inputs and their occurrence probability in the future, on the material characteristics and resistances of the dike structures, on population and asset developments, on the economic impacts of flooding, on the discount rate as well as on model choices and simplifications.

This highlights the necessity of conducting a thorough uncertainty and sensitivity analysis. Since our model serves to support decision-making, it seems natural that the sensitivity of the benefit-cost ratio to input and model uncertainties should be measured in the context of this decision. Decision sensitivity measures have been proposed [1], however, to our knowledge they have not been applied to the assessment and management of natural hazards. In this contribution, we utilize the expected value of information for sensitivity analysis on the cost-benefit analysis of a flood polder at theDanube River in Bavaria.

[1] Felli, J. C., & Hazen, G. B. (1998). Sensitivity Analysis and the Expected Value of PerfectInformation. Medical Decision Making, 18(1), 95–109.