Predictive Ability of Delft3D-Based Storm Surge Forecasts Using Historical and Forecasted Typhoon Tracks
Christine B. Mata, Clein Winslee Duquez, Olivia C. Cabrera, tristan Janryll A. Mata, Nathaniel R. Alibuyog

Mariano Marcos State University

Abstract

The DELFT3D Flexible Mesh (D3D FM) model with an unstructured grid extending up to 10 m elevation above mean sea level inland was generated to simulate storm surge in coastal areas of Northwestern Luzon (NWL), Philippines. The objective of this research is to examine the forecasting ability of the model at different lead times during extreme events by using historical versus forecasted data. Three events were selected: Super Typhoon Mangkhut (2018), Tropical Storm Pankhar (2017), and Typhoon Hato (2017). Two sea-level stations were used to validate the model: Currimao (CM) and San Fernando (SF) stations. The model^s performance revealed an average NSE = 0.834, RMSE = 7.8 cm, and MAE = 6.2 cm, indicating very good agreement between observed and simulated water levels at both CM and SF stations. Using forecasted tracks of the same events by the Japan Meteorological Agency (JMA), the average RMSE at 72-, 48-, and 24-hr lead times are 5.6 cm, 7.5 cm, and 10.6 cm, respectively, while the average MAE is 4.3 cm, 6.9 cm, and 8.9 cm, respectively. The forecasting ability of the model performs very high and indicates suitability for operational surge forecasting, although it becomes less reliable closer to landfall, especially for intense typhoons with rapid structural changes. Model performance is generally better for moderate typhoons with simpler wind and pressure fields. Overall, the model can be readily used to support disaster preparedness and early warning systems in coastal communities.

Keywords: storm surge, typhoon track, lead time, sea level, forecast

Topic: Topic B: Applications of Remote Sensing