| description abstract | Mitigating losses from a storm surge requires an accurate prediction of its peak, duration, and current speed along the coastline. Because this prediction depends on varying atmospheric and oceanic conditions, the uncertainty of predicted values must also be evaluated. Sensitivity analysis can serve to assess how variations in storm parameters can impact the surge characteristics. Such analysis is usually performed by varying one of the parameters, e.g., central pressure, by a certain percentage while assuming all other parameters, e.g., radius of maximum wind and forward speed, constant. A more reliable sensitivity analysis can be obtained when performed together with stochastic analysis. Uncertainty quantification approaches that are based on polynomial approximations of the output values with respect to input parameters can be effectively implemented to provide output sensitivities to variations in input parameters. Towards that objective, we implement a nonintrusive polynomial chaos expansion to a series of idealized hurricane storm surge simulations to quantify the sensitivity of storm surge height, duration, and current speed to variations in hurricane parameters, including size, speed, and central pressure. Particular attention is paid to the spatial variation in the sensitivity of the surge height along the shoreline, which is not well investigated. Physical reasoning behind quantified sensitivities are discussed. | |
