A hindcast of coastal flooding from hurricane Irma

• A hindcast of winds and water levels in Hurricane Irma (2017) is performed • Model performance is good and varies spatially and across two sources of meteorology • Water level response is complex, and varying along 1500 km of coast • Winds create a cross-shelf wave with quasi-resonant interactions near landfall • Error analysis of USGS storm tide sensors shows notable uncertainties in some cases Hurricane Irma produced coastal flooding across roughly 1500 km of U.S. coastline that caught many by surprise. The atypical surge patterns were driven by nuanced characteristics of the storm’s evolution that, when paired with the coastal geometry, led to a widely varied storm surge response. Here, we perform a hindcast of the coastal flooding response to Irma with the coupled ADCIRC+SWAN circulation and wave model, and two sets of meteorological forcing, one a reanalysis and the other a parametric model. This allows us to both assess model performance and to illustrate the mechanisms behind the complex surge response. Irma’s track along the length of the Florida peninsula affected water levels along much of the west Florida coast as well as the southeast Atlantic coast from Florida to South Carolina. We find the fidelity of the water level response along much of the west Florida and the southeast Atlantic coasts closely tracks the fidelity of the wind speed and direction. However, the rapid change from strong offshore to strong onshore winds and resulting across-shelf flow near the location of landfall in southern Florida makes the water level response there highly sensitive to the details of the storm. In particular, the strong initial setdown created a cross shelf pressure gradient that reinforced the surge driven by onshore winds on the back side of the storm in a resonant-like cross shelf response that may have further enhanced the surge. We also assess the quality of U.S. Geologic Survey storm tide sensor data products, which have become a widely used source of storm-related water level data. Several sources of error are identified that can combine to yield a sizeable overall error in the processed water level data that may be unapparent to a user who accesses these data.