Deriving tropical cyclone associated flood hazard information using clustered GPM-IMERG rainfall signatures: a case study in Dominica

Abstract

Abstract Different stakeholders are now looking for methods suitable for communicating the potential impacts of tropical cyclone (TC) associated rainfall and the subsequent flood hazard. We developed a new solution that utilizes GPM-IMERG satellite precipitation estimates to characterize TC rainfall spatial-temporal patterns and derives precipitation curves appropriate for TC-related flood hazard assessment. The method was demonstrated when modeling floods on Dominica due to the rainfall of Tropical Storm Erika in 2015. We performed three experiments on the procedure, starting with a time-series clustering analysis using the K-means algorithm for optimal clusters K = 5, 4, and 3. For each experiment, we excluded pixels associated with very low precipitation intensities and amounts likely disparate from the TC. We then introduced an intensity threshold of 10mm/hr to define the onset of the storm’s rain and align the pixel time series before deriving cluster representative precipitation signals (RPS) based on time step quantiles. The RPSs were used as precipitation inputs for the openLISEM, an event-based hydrological model, to simulate the resultant flood characteristics. RPSs from K = 4 were selected as Tropical Storm Erika’s final rainfall scenarios, which resulted in different flood scenarios. These results demonstrate the underlying variability in the rainfall of a single TC which should be accounted for to achieve a reliable flood hazard assessment.