On Meteo-Tsunami Forecasting with the Alaska Tsunami Forecast Model, Forced by the High-Resolution Rapid Refresh Product

The Alaska Tsunami Forecast Model (ATFM) is a numerical model used to forecast propagation and inundation of tsunamis, developed at the National Tsunami Warning Center (NTWC). ATFM uses the non-linear shallow-water equations of motion, averaged in depth, with multiple nested grids associated in two-way communications between domains of each parent-child pair as waves approach coastal waters. The model physics has been extended to incorporate atmospheric surface pressure and wind stress fields to forecast meteo-tsunami, seiche and storm surge using mesoscale weather forecasts.

On May 15, 2018, a persistent mesoscale convective system developed over parts of the Northeast, stretching 470 miles from western Pennsylvania to southern New England to Maryland. The associated frontal system enhanced by convective outflow moved off the coast along southern Connecticut and Long Island at around 22:00Z, advancing southeast into the Southern New England Bight, nearly normal to ocean isobaths. A meteo-tsunami was generated, however, the moving direction of the frontal system was not optimal for the tsunami to grow due to Proudman resonance.

High-Resolution Rapid Refresh (HRRR) is an hourly initialized forecast model developed by the NOAA Earth System Research Laboratory (ESRL). Its primary domain covers the contiguous United States and neighboring coastal waters, with a separate domain for Alaska, both at a horizontal grid spacing of 3 km. HRRR assimilates conventional observations, satellite and radar data, and generates output in hourly intervals (or 15 min at shorter forecast lengths) out to lead times of 18 to 36 hours. Wind and surface pressure forecasts used in this study were obtained from HRRR version 3, which ran experimentally (as HRRRx) at ESRL at the time of the event, and is now run operationally at the National Centers for Environmental Prediction.

In this study, HRRRx forecasts of the event, available from ESRL, are used to test the dynamic response of model ocean to the forecast pressure anomaly and surface wind fields. This is a new and pilot approach investigating the development of a warn-on-forecast system. The initial results appear to be promising. Once being fully established, the system has the potential to provide the capability for more timely and reliable warnings, out to at least 18 hours in advance.