A Vulnerability-Based, Bottom-up Assessment of Future Riverine Flood Risk Using a Modified Peaks-Over-Threshold Approach and a Physically Based Hydrologic Model

There is a chronic disconnection among purely probabilistic flood frequency analysis of flood hazards, flood risks, and hydrological flood mechanisms, which hamper our ability to assess future flood impacts. We present a vulnerability-based approach to estimating riverine flood risk that accommodates a more direct linkage between decision-relevant metrics of risk and the dominant mechanisms that cause riverine flooding. We adapt the conventional peaks-over-threshold (POT) framework to be used with extreme precipitation from different climate processes and rainfall-runoff-based model output. We quantify the probability that at least one adverse hydrologic threshold, potentially defined by stakeholders, will be exceeded within the next N years. This approach allows us to consider flood risk as the summation of risk from separate atmospheric mechanisms, and supports a more direct mapping between hazards and societal outcomes. We perform this analysis within a bottom-up framework to consider the relevance and consequences of information, with varying levels of credibility, on changes to atmospheric patterns driving extreme precipitation events. We demonstrate our proposed approach using a case study for Fall Creek in Ithaca, NY, USA, where we estimate the risk of stakeholder-defined flood metrics from three dominant mechanisms: summer convection, tropical cyclones, and spring rain and snowmelt. Using downscaled climate projections, we determine how flood risk associated with a subset of mechanisms may change in the future, and the resultant shift to annual flood risk. The flood risk approach we propose can provide powerful new insights into future flood threats. Plain Language Summary As the climate changes, we expect weather patterns to shift. We often discuss how strong and how frequent future storms might become. Scientists can estimate these changes with global models, called GCMs though we have some difficulty explaining that some of our climate predictions are reliable (like future air temperature) while others are less reliable (future rain intensity). Within a given region, we might know that floods are generally caused by certain storms (i.e., tropical cyclones, frontal systems). Based on climate projections, we often have reason to believe that these storms might be changing, but we do not have a simple way of letting these predictions tell us about future flooding risk. We propose a method of estimating the riverine flooding risk by looking at the dominant regional storm types that most commonly cause floods. We identify different types of storms and consider how often and intense these storms are at present. We use our approach to determine what exactly the GCMs can say about the future climate with confidence, and how that might alter flooding risk. Our results suggest that in the Northeast U.S. changes to future risk are linked to the most uncertain aspects of future climate predictions.