2.3. Storm Tide Simulation
We use the ADvanced CIRCulation model (ADCIRC) to simulate storm tides (i.e., the total water level including storm surge and astronomical tide). ADCIRC, originally developed by Luettich et al. (1992) and Westerink et al. (1994), is a well-established open-source hydrodynamic model applicable to free surface circulation and transport problems. The model adopts the hydrostatic pressure assumption to numerically solve the governing equations in space and time using, respectively, the finite-element and finite-difference numerical methods. ADCIRC solves the 2-D or 3-D momentum equations to calculate velocity fields. For storm surge modeling in the coastal ocean, the 2-D depth-integrated model is usually preferred to the 3-D model because it is computationally cheaper and largely captures the dominant processes that control the storm-induced water level (Resio and Westerink, 2008). In the present study, we use the 2-D depth-integrated version of ADCIRC. Due to the extremely large computational resource that would be required to simulate wind waves associated with the large number of synthetic storms, we do not apply wave simulations here. Thus we neglect the effect of wave setup on the total water level, which has been estimated to contribute up to 10-20% to total extreme water levels in the Shanghai region (Melet et al., 2018).
Using the mesh generation module of Surface-water Modeling System (SMS), we develop a computational mesh that extends between latitudes 21°N and 41°N and longitudes 116.5°W and 129.5°W. The computational domain consists of an unstructured triangular mesh with 108,440 nodes and 209,844 elements. The mesh resolution is <100 m nearshore and gradually increases to > 50 km in deeper waters. The bathymetry of the model is based on the bathymetry survey of Yangtze River Estuary in 2012 (Zhu and Bao, 2016) and GEBCO (GEneral Bathymetric Chart of the Oceans)-2014 global 30 arc sec gridded data (https://www.gebco.net/data_and_products/historical_data_sets/). The vertical datum is in mean sea level (referenced to year 2000). The manning’s n bottom roughness coefficient varies spatially over the grid based on a global database for sediments on the seabed (http://portal.gplates.org/static/html/seafloor.html). The water level at the open boundaries is specified by eight major tidal constituents (K1, K2, M2, N2, O1, P1, Q1, and S2). The tidal information, including the amplitudes and phases, are obtained from the global model of ocean tides TPXO8-ATLAS with a 1/30resolution (Egbert and Erofeeva, 2002). At closed boundaries, i.e., island and mainland shorelines, the normal flow is set to zero whereas the tangential velocity is determined based on the free-slip boundary condition. While it is difficult to validate the ADCIRC modeling for storm surges in Shanghai, as a critical TC parameter –radius of maximum wind–is not available in the historical record for the Northwest Pacific Basin, the ADCIRC model has been found to produce storm surge estimations with relatively high accuracy in a number of studies (e.g., Westerink et al., 2008; Marsooli and Lin, 2018). Our evaluation on the model performance in simulating the astronomical tides, which can contribute to over 75% of the storm tides in Shanghai, shows that the ADCIRC model satisfactorily captures the dominant dynamics of major tide constituents along the Shanghai coast (Fig. S2).
We use analytical/parametric models to calculate the surface wind and pressure fields associated with the synthetic TCs to drive the hydrodynamic modeling. We use the parametric model of Holland (1980) to calculate the radial profile of pressure, which is related to the maximum wind speed, radius of maximum wind, and pressure deficit. We use the analytical model of Emanuel and Rotunno (2011) to calculate the wind field. The model calculates the 1-min axisymmetric wind field (associated with the storm) at the gradient level based on the storm characteristics including maximum wind speed and radius of maximum wind. We convert the gradient wind to the surface level (10 m above the sea surface) with a velocity reduction factor of 0.85 (Georgiou et al., 1983) and an empirical expression of inflow angles (Bretschneider, 1972). We add to the storm wind the surface environmental wind estimated as a fraction (0.55, rotated counter-clockwise by 20o) of the storm translation velocity, to account for the asymmetry of the wind field (Lin and Chavas, 2012). Finally, for use in the ADCIRC modeling we adjust the 1-minute wind to a 10-minute average with a reduction factor of 0.893 (Powell et al., 1996). This framework of TC modeling for storm surge analysis has been implemented by various previous studies (e.g., Lin et al., 2012; Marsooli et al. 2019). Also, Xu et al. (2020) used a similar method to estimate the wind field, and their estimated wind hazard based on the NCEP synthetic dataset compared well with historical wind observations in the Shanghai region.