A Statistical Forecast Model for Tropical Cyclone Rainfall and Flood Events for the Hudson River

Tropical Cyclones (TCs) lead to potentially severe coastal flooding through wind surge and also through rainfall-runoff processes. There is growing interest in modeling these processes simultaneously. Here, a statistical approach that can facilitate this process is presented with an application to the Hudson River Basin that is associated with the New York City metropolitan area. Two submodels are used in sequence. The first submodel is a stochastic model of the complete life cycle of North Atlantic (NA) tropical cyclones developed by Hall and Yonekura (2011). It uses archived data of TCs throughout the North Atlantic to estimate landfall rates at high geographic resolution as a function of the ENSO state and of sea surface temperature (SST). The second submodel translates the attributes, of a tropical cyclone simulated by the first model, to the streamflows at specific points of the tributaries of the Hudson River. That points are the closure sections of five different watersheds. The second submodel is splitted in two processes: 1) determine the peak of the discharge; 2) determine the shape of the discharge hydrograph. To determine the peak of the discharge a Bayesian Simultaneus Quantile Regression approach translates the TC attributes ( track, SST, Velocities,..) into peaks; then a multivariate normal distribution associates a hydrograph shape to position and barometrical data. Finally the streamflow tributaries of the Hudson River are to be used as inputs in a hydrodynamic model that includes storm surge dynamics for the simulation of coastal flooding along the Hudson River.