The Holland (2010) parametric wind model has been extensively utilized in tropical cyclone and storm surge-related coastal hazard mitigation and management studies. The only remaining input parameter, the radius of maximum wind speed ($R_m$), is usually generated by previously proposed empirical relations which are, however, sensitivity to study areas in producing better performed numerical results. In order to acquire optimal $R_m$ formulations over the region of Zhoushan Archipelago, East China Sea, 16 empirical relations were compiled into the Holland (2010) model to produce time series of the pressure, wind speed, and wind direction in comparison to observational records taken at three stations during the tropical cyclone events of Ampil and Rumbai. Their respective agreements were evaluated by error metrices including the root mean square error, correlation coefficient, mean bias error, and scatter index, whilst the overall performances of the 16 formulations were ranked according to a proposed comprehensive error. In the following order, the $R_m$ formulations of Lu (2012), Zhou (2005), Kato (2018), and Jiang (2008) ranked the best for both events in terms of their minimum comprehensive errors; however, recommendations on the application of specific empirical formulations for the region of Zhoushan Archipelago are also provided herein from the perspective of conservation and accuracy.

Storm surge current-time estimations are strongly influenced by recent water-level conditions, while the temporal dependence among samples from the same typhoon event can complicate the assessment of model generalization to unseen events. This study evaluated storm surge estimation at the Lianyungang tide gauge using 2168 samples from 28 typhoon events during 2002–2024. A 41-feature extreme gradient boosting (XGBoost) model integrating historical surge and physics-motivated information was evaluated using fully nested event-grouped cross-validation, with complete outer-test typhoon events excluded from hyperparameter optimization, early stopping, and model fitting. The 41-feature XGBoost model achieved a mean absolute error (MAE) of 0.083 m, a root mean square error (RMSE) of 0.128 m, and a coefficient of determination (R²) of 0.794. On the common valid sample subset, its RMSE was 9.15% lower than that of Persistence, and it achieved a lower event-level RMSE in 18 of the 28 independently held-out typhoon events. Controlled information-source experiments showed that historical surge information accounted for most of the aggregate predictive skill, whereas adding core storm-state variables and the complete set of physics-motivated descriptors produced only limited changes in overall performance. The relative improvement over Persistence was larger for samples in which the shortest available historical surge lag was 3–6 h than for those with a 1 h lag, but this advantage did not extend consistently to the highest-surge conditions, where systematic underestimation remained evident. These results demonstrate the importance of event-independent validation for assessing machine-learning storm surge estimation and show that the model skill depends strongly on both information availability and the surge magnitude. The framework should be interpreted as a retrospective, single-station current-time estimation approach rather than an operational forecasting or extreme-surge warning system.

The nonlinear interaction between storm surge and astronomical tide controls coastal flood levels in macrotidal regions, but the effect of tidal phase on the surge of a given typhoon is rarely quantified. Using a high-resolution MIKE 21 FM model of the East China Sea forced by ERA5, we hindcast the storm surge of super typhoon Bebinca (2024), which made landfall at Shanghai. Tidal RMSE at four gauges is 0.11–0.37 m and surge RMSE is 0.11–0.24 m. The surge evolved from pressure-dominated offshore setup to wind-dominated nearshore setup and then to slow shallow-water decay, reaching 2.0–2.5 m in the Yangtze River Estuary and northern Hangzhou Bay. Keeping the typhoon forcing identical, we shifted the open-boundary tide by ±6 h. The two shifted cases differed by 0.5–0.9 m in total water level at landfall, but by less than 0.1 m in surge, so that difference was the tide itself. Relative to the baseline, in which landfall nearly coincided with high water, both shifted cases lowered the landfall surge by about 0.9 m on the northern coast of Hangzhou Bay and by about 0.7 m at the Yangtze River Estuary peak. Between the shifted cases, surge differed appreciably only at the bay head, by up to about 1 m before landfall.