A storm surge operational forecast framework based on publicly-released meteorological data

In this study, we aim to develop a publicly-deployable regional storm surge operational forecast framework based on limited openly-released tropical cyclone (TC) forecasts, which can be deployed in less-developed regions with restricted access to real-time high-resolution wind-field ensemble forecast products. The framework consists of data imputation, wind-field reconstruction, and hydrodynamic modules. Specifically, in the wind-field reconstruction module, a size-informed correction of the Holland wind model is proposed to improve reconstructed pressure and wind fields. The size-informed wind-model correction notably improved the accuracy of reconstructed wind fields and predicted storm surges, reducing peak wind speed errors by up to 97% and peak water level errors by up to 94% in hindcasts. The framework has been applied to practical operational forecasts of storm surge in the Pearl River Estuary during 13 TCs since 2023. In this paper, its performance in forecasting storm surge in Macau during Typhoon Talim, Super Typhoon Saola, and Super Typhoon Koinu is evaluated. In 48-, 24-, and 12-h-lead forecasts of storm surge during Talim and Koinu, the framework yielded peak water-level errors mostly below 0.35 m, with Skill values above 0.92. However, its performance degraded notably in the case of Saola as there were considerable uncertainties in TC forecasts. We further investigate the uncertainties in storm surge forecasts due to the errors in publicly-released predictions of TC track and intensity. It is shown that the timing of peak storm tide is primarily sensitive to track errors, while peak water level is sensitive to both track and intensity errors.