Honest reliability: uncertainty and nonstationarity reduce flood reliability

Abstract:

Recent catastrophic flood events, increasing flood losses, and climate change challenge current estimates of reliability that design floods will not be exceeded over a planning horizon. These estimates depict an average value of reliability that mask uncertainty in streamflow, land use, and river channels. This presentation builds on existing research that describes reliability as a random variable whose distribution depends on the joint probability distribution of flood magnitude-frequency relationships and channel capacity. We quantify and map the distribution of reliability while accounting for uncertainty and nonstationarity in flood hazard estimates to garner a better understanding of potential risk profiles compared to average estimates. Monte-Carlo simulations of flood hydraulics are conducted at Little Sugar Creek, Charlotte, NC, USA representing uncertainty in land use, channel capacity, and streamflow. Floods of varying magnitude and frequency are simulated to characterize the spatial gradient of reliability across the landscape, while uncertainty and temporal changes in magnitude-frequency relationships are accounted for to quantify the impacts of nonstationarity and uncertainty on reliability through time. Our results indicate that accounting for uncertainty in channel and floodplain capacity reduces reliability estimates compared to estimates solely based on flood likelihood, and the divergence is greater in the presence of a nonstationarity flood series. Further, the distribution of reliability is spatially heterogenic due to topographic controls such as in-stream structures and valley form. Quantifying the distribution of reliability enables prescribing a desired level of confidence associated with an estimate. Similarly, we show that a desired level of confidence in reliability can be specified to estimate a design flood magnitude. These methods can provide a valuable flood risk communication and decision-making tool in a dynamic world. The distribution of reliability along floodplain corridors highlights an opportunity for nature-based infrastructure whose benefits might extend beyond flood mitigation.