The Effect of Seawater Spatial density heterogeneity on Storm Surges — A Case Study of Typhoon "In-Fa" and Hangzhou Bay and Its Adjacent Waters

Abstract. Previous studies have shown that considering density spatial heterogeneity can improve the simulation capability and accuracy of water level, including storm surges, but the underlying dynamical mechanisms remain unclear. In this study, we conducted two contrasting models by FVCOM— a heterogeneous ocean model and a homogeneous ocean model to simulate water levels in Hangzhou Bay during the passage of Typhoon In-Fa in July 2021. We analyzed the spatiotemporal distribution characteristics of storm surges, and investigated carefully the mechanisms of storm surges by which density spatial heterogeneity. The simulation results show that both models perform comparably and with high accuracy in simulating astronomical tides and storm surges outside Hangzhou Bay. However, there are significant differences emerge in simulating storm surges inside Hangzhou Bay, where the heterogeneous ocean model substantially outperforms the homogeneous ocean model, with a maximum error of less than 0.2 m compared to 0.9 m for the latter, approximately 25 % of the maximum storm surge. Based on the magnitude and duration of storm surge rise and fall, the interior and exterior of Hangzhou Bay exhibit two distinctly different situations: the interior is characterized by large surge amplitudes with durations exceeding 2 days, while the exterior shows smaller amplitudes with durations of approximately half a day. Dynamical analysis reveals that in regions where the storm surge duration exceeds the inertial period (e.g., the interior of Hangzhou Bay), the part of the storm surge affected by density can be expressed as a coupled form of the product of the density Laplacian and a quadratic term of the storm surge. In regions where the density Laplacian is negative (i.e., the density surface is convex upward), the part of the storm surge affected by density is negative during the surge rise phase, thereby reducing the overestimation bias of the homogeneous model; during the surge fall phase, it becomes positive, reducing the underestimation bias. In regions where the density Laplacian is positive, the above effects are reversed.