A Baseline Storm Surge Ensemble Methodology for Pacific Islands – the PACSURGE System

Storm surge is inherently sensitive to tropical cyclone track, speed, intensity, and wind structure, rendering 2- to 5-day deterministic forecasts impracticable. Ensemble surge forecasts therefore are recommended for emergency preparedness during a pending tropical cyclone impact. High-resolution simulations are typically finite-element models such as ADCIRC, FVCOM, or Delft3D which require supercomputers, labor-intensive grid configurations, and numerous track scenario tests to identify and fix instability issues. Such models, while useful for case studies or hindcast of surge events, generally are unfeasible for operational ensemble runs due to the computer requirements and resource development. Other optimized models, such as SLOSH, still require resources for grid configurations and lack community model support. However, since storm surge is governed primarily by four components which bound the solutions – wind setup, pressure setup, current-induced geostrophic adjustment processes, and wave setup – and because of forcing errors, such advanced systems may not be necessary for ensemble forecast spread.

Based on these constraints, this poster discusses a storm surge ensemble guidance two-part package developed for rapid deployment and fast solutions. The initial focus target are Pacific islands, but other ocean basins and coastal continents are applicable. The first product is a 1D model called PACSURGE-M, and the second product assumes steady-state 1D peak surge conditions at landfall called PACSURGE. Each wind forcing run is from official tropical cyclone wind speed probabilities (DeMaria et al. 2009, 2013) disseminated through operational tropical cyclone forecast centers. However, any gridded wind ensemble product, such as the COAMPS-TC ensemble system, can be used. Wave input is generated by a wave forecast ensemble that is consistent with this wind speed probability product (Sampson et al. 2016). If pressure is not available, it is computed from the Courtney and Knaff (2009) relationship. Options for the geostrophic adjustment are for an empirical term by Fitzpatrick and Lau (2016) or specifically computed based on data from Chiang et al. (2016). Options for wave setup are based on a percentage of inshore breaking wave height or offshore significant wave height (FEMA 2005). In cases where a coral reef attenuates wave setup impact, or the geostrophic adjustment process is negligible (as in small islands), either term can be turned off. The tide range is superimposed on the surge, providing additional ensemble members.

Surge validation results of 8 tropical cyclones for Okinawa’s Buckner Bay (Nakagusuku Bay), using water level data from the Japan Oceanographic Data Center On-line Service System (J-DOSS), are presented. An example ensemble application for Typhoon Jelawat (2012) will be shown.

Both codes are written in FORTRAN and internally documented. An earlier version of PACSURGE-M is available in python and available on drfitz.net . A spreadsheet of PACSURGE is also available for checking the FORTRAN version, and is ideal for classroom instruction on storm surge in conjunction with notes on Fitzpatrick’s teaching website weatherclasses.com . Beta versions of PACSURGE and PACSURGE-M are ready for deployment, just requiring bathymetry cross-sections and tidal data for new locations.

References are available by request.