Xiamen Bay, Fujian, China is located at the southeast coast of mainland China ( Figure 1 ). It is a semi-closed bay connecting the Jiulong River Estuary with the Taiwan Strait. There are more than 30 islands distributed in Xiamen Bay, including Xiamen Island (158 km²), which is a main part of Xiamen city, Fujian province, and Dajinmen Island (151.65 km²). Due to the islands, caps, and small bays, Xiamen Bay is characterized by strongly irregular coastlines and extremely complicated topography. In this study, Xiamen Bay is divided into six parts for convenience of analysis: Tong’an Bay, the west Xiamen water, the Jiulong River Estuary, the south Xiamen water, the east Xiamen water, and the Dadeng water. The former three parts are grouped into the Inner Bay Area, and the others are the Outer Bay Area.

The Xiamen Bay is a macrotidal bay. The averaged tidal range at the Xiamen gauge station ( Figure 1 ) is 4.0 m, and the recorded maximum tidal range is 6.92 m, which occurred on 22 October 1933 (Huang et al., 2018). The main four tidal constituents at the Xiamen gauge station are M₂ tide (1.91 m), S₂ tide (0.54 m), K₁ tide (0.34 m), and O₁ tide (0.27 m). Thus, Xiamen Bay also belongs to a semidiurnal tidal regime [F< 0.25, where F = (a _K1+a _O1)/(a _M2+a _S2) is the form number (National Ocean Service, 2000)].

Xiamen Bay is struck by TCs frequently from July to October. Statistics show that about 2.7 TCs can affect Xiamen Bay per year on average (Miao et al., 2022). TCs always raise coastal flooding in the low-lying land of Xiamen Bay. For example, the TCs Iris (5903) and Dan (9914) induced strong coastal flooding at Xiamen City, leading to 79 and 245 km² of flooded agricultural areas, respectively (Li et al., 2000; Yuan et al., 2022). During 2016, although Xiamen Island is well protected by seawalls, the TC Meranti (1614) flooded over them and inundated over 10 km² of low-lying land (Luo and Wu, 2022).

The Finite-Volume Community Ocean Model (FVCOM, version 4.1) (Chen et al., 2003; Chen et al., 2013) is employed to simulate the coastal inundation of Xiamen Bay. The unstructured triangular mesh that FVCOM employs can fit the irregular coastlines and islands, and be freely refined around large slope gradients. Moreover, FVCOM contains a widely used wet/dry scheme (Leendertse, 1970; Zheng et al., 2003). The computational domain covers the maximum flooding area, and the wet and dry points are distinguished by the local total water depth. Therefore, FVCOM is rather appropriate for this study.

The model domain covers all the coastlines and islands of Fujian province and the Taiwan Island of China, extending 1000-km seaward to the Pacific Ocean ( Figure 2A ). As TCs can move to a large extent during their lifetime, this large-scale model coverage can well capture the seawater movement in Xiamen Bay under the influence of TCs and also eliminate boundary effects associated with smaller regional meshes, thus improving the accuracy of flooding simulations (Thomas et al., 2022). In Xiamen Bay, the model domain extends landward and covers the coastal lowland to the 20 m topographic contour ( Figure 2B ). The mesh resolution is about 80 km at the open boundary and increases to a high resolution of 100 m in Xiamen Bay, which is maintained in all the coastal lowland domains. Thus, the model mesh has 241278 nodes and 473744 triangular elements.

The bathymetry data used in this model are primarily based on the ETOPO1 Global Relief Model (Amante and Eakins, 2009). Although the resolution of ETOPO1 is as high as 1 arc-minute, it cannot assess the extremely complex bathymetries in Xiamen Bay precisely. Therefore, we gather 64 naval electric nautical charts around Xiamen Bay to obtain the precise estimations of its underwater topography. The topography of coastal lowlands around Xiamen Bay are estimated by a 10-m-resolution DEM dataset offered by Fujian Provincial Department of Ocean and Fisheries ( Figure 2C ). The seawalls around Xiamen Bay offer a robust defense mechanism against coastal flooding; they are considered in the FVCOM model simply by elevating the depth of corresponding mesh nodes ( Figure 2C ). All the simulations run 15 days with a spin-up of 7 days.

The tidal levels of the open boundary condition are derived from the latest released satellite-assimilated tidal model TPXO9-atlas (Egbert and Erofeeva, 2002), with eight tidal constituents (M₂, S₂, N₂, K₂, K₁, O₁, P₁, and Q₁) specified. Four vertical sigma layers are used in the model. FVCOM calculates the bottom friction and wind stress using the quadratic drag law:

The drag coefficient for bottom friction is determined by matching a logarithmic bottom layer to the model at a height z _ab above the bottom, namely:

where κ = 0.4 is the von Karman constant, z is the distance from the seabed to the position of u and v, and z ₀ is the bottom roughness parameter. We set z ₀ to be a constant of 0.0002 in this study.

It is commonly recognized that the wind drag coefficient in high-wind conditions does not increase linearly with surface wind speed, but the specific relationship is still uncertain (Peng and Li, 2015; Donelan, 2018). In this study, we set the wind drag coefficient to be constant during low- and high-wind conditions following Yin et al. (2021a). The equation is expressed as follows:

where U ₁₀ is the wind speed (m·s^-1) at a height of 10 m above the sea surface.

A wind field combining a reanalysis background wind product and a parametric typhoon model is adopted to simulate the coastal flooding. The reanalysis background wind product is retrieved from the hourly ERA5 dataset (Hersbach et al., 2020) of the European Centre for Medium-Range Weather Forecasts. The ERA5 dataset has a high time resolution, but its spatial resolution is relatively coarse, which is not sufficient to precisely capture the structures of pressure and wind within a scope of hundreds of kilometers near the center of a TC. A parametric typhoon model is thus applied to reassess the wind structure surrounding a TC’s center. The best-track dataset from China Meteorological Administration (Ying et al., 2014) is employed to drive the parametric typhoon model.

The pressure profile is estimated by the equation proposed by Fujita (1952):

where P_c is the central pressure, P_e is the environmental pressure, R _max is the radius of maximum wind, and r is the distance to the TC’s center.

The wind speed field is determined by the gradient wind model, in which the force of the pressure gradient force is balanced by the Coriolis force and centrifugal force (Schwerdt et al., 1979). The solution is expressed as follows:

where W ₁ is the gradient wind speed at radius r; f = 2ωsinφ is the Coriolis parameter, ω = 7.272 × 10^-5 rad·s^-1 is the Earth’s rotational angular speed; φ is the latitude; and ρ_a = 1.15 kg·m^-3 is the air density.

Equation (5) produces a circular symmetric typhoon wind field. Actual observations indicate that the wind structures of TCs are not always symmetric, mainly relating to the translation of the TC, the blocking action from an adjacent anticyclone event, the meridional gradients of the Coriolis acceleration (β effect), the asymmetric convection of bottom boundary layer friction, and landfall (Olfateh et al., 2017). In this study, the asymmetry is considered only by superimposing a moving wind field on the circular symmetric wind field, which is a popular method in modeling surges (Li et al., 2022):

where W ₁ is the circular symmetric wind speed calculated by Equation (5); W ₂=e ^πr/500000(V _x ,V _y )  is the moving wind field (Miyazaki, 1962), ϕ is the angle between the vector from the TC’s center to the position and due east, β is the angle between the wind direction with the isobars, which can be specified as a constant (for example, 18° in this study); and c ₁ and c ₂ are correction coefficients, which are set as 0.8 and 1.0 here, respectively.

Thus, the wind field can be reconstructed by combining the asymmetric circular wind field assessed by Equation (6) and the ERA5 background wind field:

where e is a weight factor to ensure a smooth transition between the two wind fields (Carr and Elsberry, 1997):

In addition to TC Meranti (1614), TCs Trami (1312) and Dujuan (1521) also induced prominent storm surges in Xiamen Bay. Observed tidal level data during the three TC events at Xiamen and Shijing ( Figure 1 ) are obtained from Fujian Provincial Department of Ocean and Fisheries. The numerical model performance is evaluated by comparing total water levels with the observations. Root-mean-square error (RMSE), relative error (RE) to tidal range, and skill score (SS) are used to quantify the model’s performance:

where X m i   and X o i are the model simulation and observed data at time i, respectively; N is the number of observed records used for the comparison; TR_o is the observed tidal range, and X o ¯ is the average value of the observed data. When SS > 0.5, the model performance can be considered highly reliable.

The water level comparisons at stations Xiamen and Shijing are displayed in Figure 4 , and the error skills are also labeled. The averaged RMSE at Xiamen and Shijing are only 0.25 m and 0.22 m, respectively, which are small enough comparing to the macro tidal range (4.0 m) at Xiamen Bay. The REs are all less than 5%, and the SSs are all larger than 0.95. All the skills show that our model can accurately reproduce the storm tides in Xiamen Bay during TC events.

Figure 5A displays the water level timeseries at station Xiamen induced by the five TC tracks of 1000-year pressure at the current climate. The results show that the water levels are highly dependent on the TC tracks, implying the necessity of multi-tracks blend method used in this study. Peak water level is highest on track L1, which is 70 km (or about 2.5 times the maximum wind radius) south of Xiamen Bay, but lowest on track R2. The maximum difference of peak water level between each track can be over 3.0 m. After removing tidal signals from the water level timeseries by subtracting the water levels from a tide-only run, the storm surges are achieved. As Figure 5B shows, track L1 obtains a maximal storm surge of 4.37 m, but the time that it occurs is 1.5 h later than that of the peak water level, which may relate to the well documented tide-surge interaction.

The tracks-blended results show that stronger storms induce higher water levels and storm surges (see solid lines in Figures 5C, D ). The peak water levels of the 10-, 100-, and 1000-year pressures reach 4.99, 5.89, and 6.63 m, and the corresponding peak storm surges are 2.34, 3.27, and 4.38 m, respectively. The storm surge under the 1000-year pressure can reach 2.0 m higher than that of the 10-year pressure, but the corresponding peak water level is only 1.6 m higher. Nevertheless, the time differences between peak water levels and peak surges are 0.2, 0.5, and 1.5 hours under the 10-, 100-, and 1000-year pressures, respectively. These features imply a stronger tide-surge interaction in stronger storms.

Higher water levels lead to more extensive flood inundation around Xiamen Bay. Herein, we consider all wet nodes (or cells) with model depth< 0 to be flooded areas. Astronomical tides can periodically inundate the large intertidal zones in the Jiulong River Estuary and the widespread low-lying coastal farms and salt pans between Dadeng Island and the mainland ( Figure 6A ). Our statistics ( Figure 7 ) show that 126.5 km² of low-lying land can be flooded during high tides, with over 80% of this flooding occurring in the Jiulong River Estuary and the Dadeng water.

The flooded areas largely expand in Xiamen Bay under TC events—increasing 3.1-, 4.0-, and 4.7-fold during astronomical high tide under TCs of 10-, 100-, and 1000-year pressures, respectively ( Figure 7 ). TCs impact the Inner Bay Area at a higher rate than the Outer Bay Area. The proportion of flooded area in the Inner Bay Area increases from 42% during tide-only situation to about 70% under TC events.

Under the influence of TCs with a 10-year pressure, all the low-lying shoals in the Jiulong River Estuary are expected to be potentially inundated ( Figure 6B ), leading to an extremely significant expansion of flooded area of about 180 km² ( Figure 7 ). In the west Xiamen water, the elevated water levels force seawater flood into Xinlin Bay, resulting in an increase of over 10 km² of flooded area. The flooded area in Tong’an Bay can intrude inland for more than 5 km and lead to about 12 km² of additional flooded area. Mainly due to the significant increase in flooded area in the Jiulong River Estuary, the Inner Bay Area results in 67% flooded area in Xiamen Bay, which is much more than the 42% under tide-only conditions. Significant flooded area expansion (nearly 40 km²) can also be seen in the Dadeng water of the Outer Bay Area. These additional floods mainly happens on the low-lying coasts opposite Dadeng Island. Despite the high expansion of flooded areas, Xiamen Island, which is protected by high-standard seawalls, is little impacted by TCs with 10-year pressure.

Under the influence of TCs with 100-year pressure, the flooded areas continue to expand in all six parts of Xiamen Bay under the influence of higher peak water levels. In the Jiulong River Estuary, the flooded areas intrude more inland and expand by about 55 km² ( Figure 6C ). In the west Xiamen water, the Maluan Bay is also flooded, resulting in a significant expansion of flooded area of over 25 km². We should note that part of the low-lying southwest Xiamen Island (about 4 km²), where a wetland park is located, can be flooded under this condition.

TCs with 1000-year pressure can result in extremely severe flood disaster in Xiamen Bay when meeting high tides. Nearly 600 km² of low-lying land can be flooded, and over 50% of this occurs in the Jiulong River Estuary ( Figure 7 ). Coastal flooding can extensively intrude into Xinlin Bay, Maluan Bay, and low-lying southwest Xiamen Island, leading to 60 km² of flooded area in the west Xiamen water ( Figure 6D ). In Tong’an Bay, the extremely high water level generates expansive flooding along both coasts, leading to over 40 km² of flooded area.

Flood damage typically increases disproportionately with flood depth (Scawthorn et al., 2006). The average flood depth is 1.3 m during astronomical high tide and increases to 2.3, 2.8, and 3.4 m with superpositions of TCs with 10-, 100-, and 1000-year pressures, respectively ( Table 2 ). The flood depth is little area-varying under the sole influence of tide but strongly area-varying under the influence of TCs. The TCs with 10-year pressure can result in average flood depths of 2.4 m and 2.3 m in the Jiulong River Estuary and the Dadeng water, respectively, but that of the west Xiamen water and east Xiamen water reaches only 1.9 m. The average flood depth in the Jiulong River Estuary spread to 3.2 m under TCs with 100-year pressure, but it increased little in the west Xiamen water. Under the influence of TCs with 1000-year pressure, the averaged flood depth in the Jiulong River Estuary can dramatically spread to 4.0 m, whereas that in Tong’an Bay, west Xiamen water, and east Xiamen water are less than 3.0 m.

Flood duration is reported to be an additional critical parameter responsible for flood damage (Javelle et al., 2003; Wagnaar, 2012; Feng et al., 2017), since many infrastructures, such as the power system (Cruse and Kwasinski, 2021) and the transportation system (Pyatkova et al., 2019), are concerned with how long the facility will be offline. Figure 8 displays flood duration map under influence of tide-only and TCs of 10-, 100-, and 1000-year pressures. When TC is absent, long flood duration is found in the river channels of the Jiulong River Estuary and the low-lying coastal farms and salt pans in the Dadeng water ( Figure 8A ). TCs prolong flood durations in all flooded areas. The additional flooded area suffers relative short flood duration. The proportions of area suffering flood longer than 10 hours are 40%, 46%, 51%, and 53% under scenario of tide-only and TCs of 10-, 100-, and 1000-year pressures, respectively. Accordingly, the area-average flood durations of each scenario are longer than 10 h.

Integrated maps of flood areas, flood depth and flood duration, the most severe flooding caused by TCs always occurs in the Jiulong River Estuary due to the extent of low-lying shoals, followed by the Dadeng water, which is mainly due to the low-lying coastal farms and salt pans between the island and mainland. Relatively, the south and east Xiamen water is little flooded by TCs. We should note that although Xiamen Island is well protected by seawalls, its low-lying southwest portion can be flooded under TCs with 100-year pressure meeting high tide. Although the flooded area is limited, it should be strongly considered due to the extremely dense population and numerous buildings on the island.

Global SLR can exacerbate coastal flood. With an SLR of 0.77 m on the open boundary, the peak water level at the Xiamen station can be elevated substantially by about 0.89 m ( Figure 5 ). The bolstered rise in peak water level may be related to the amplified tides under SLR, which has been well documented by many studies (e.g., FitzGerald et al., 2008; Khojasteh et al., 2021). A simple example is that the succedent low tidal level is elevated by only 0.67 m. The 0.77 m SLR expands the flood areas during high tide by ~ 77 km² in Xiamen Bay, over half of which occurs in the Jiulong River Estuary ( Figures 6 , 7 ). Moreover, the average flood depth increases to 1.6 m, with the largest increase occurring in the Dadeng water ( Table 2 ).

Comparing the dashed and solid lines in Figure 5D , the surge levels change little under intensive 10-year pressure, whereas they increase by 0.20 m and 0.10 m under intensive 100-year and 1000-year pressures, respectively. This result illustrates that changes in TC climatology can increase the influence of SLR on peak water levels, although SLR may predominate the elevated peak water level in Xiamen Bay. This phenomenon is similar to results from New England, but it is very different compared to those from the Gulf of Mexico, where the effect of TC climatology change is likely to be larger than the effect of SLR (Marsooli et al., 2019). The combination of SLR and TC climatology change lead to TCs with 10-year pressure inducing a peak water level that is nearly equivalent to that induced by TCs with 100-year pressure under the current climate, and TCs with 100-year pressure will induce a peak water level nearly equivalent to that induced by TCs with 100-year pressure under the current climate ( Figure 5C ).

As a result of the substantial elevated peak water levels under the combination of SLR and TC climatology change, the flood area is expanded by about 90 (23%), 80 (16%), and 85 (14%) km² under TCs with 10-, 100-, and 1000-year pressures, respectively ( Figure 7 ). The main expansion happens in the Jiulong River Estuary, following by the west Xiamen water and the Dadeng water. The averaged flood depth and averaged flood duration will increase by ~ 0.5 m and ~ 1.0 h for each scenario by 2100 under SSP5-8.5 projection, respectively.

Comparing Figures 6B, F , under the condition of TCs with 10-year pressure, the SSP5-8.5 climate change projection can flood more area in the west Xiamen water, including Maluan Bay and some of the low-lying area in the southwest of Xiamen Island, although the flood depth and duration of these places is very small. This change results in a remarkable expansion (21 km²) in flood areas of the west Xiamen water and a slight decrease in the average flood depth ( Figure 7 ; Table 2 ). In other parts of Xiamen Bay, the SSP5-8.5 projection leads to flood area expansion and longer flood duration accompanied by flood depth increase, especially in the Jiulong River Estuary, where the flood area is expanded by 35 km², the average flood depth and duration increases from 2.3 m and 10.40 h in the current climate to 2.9 m and 11.91 h.

Under the condition of TCs with 100-year pressure, the SSP5-8.5 projection by 2100 leads to flood area expansion and increased average flood depth for all six parts of Xiamen Bay, except for some lagging in the average flood depth in Tong’an Bay, where significant flood area expansion occurs (comparing Figures 6C, G ). The flooded area (583 km²) and flood depth (3.3 m) in the whole Xiamen Bay are nearly equivalent to that induced by TCs with 1000-year pressure under the current climate (598 km² and 3.4 m). The low-lying land in the southwest of Xiamen Bay, which can only be entirely flooded under TCs with 1000-year pressure, will be flooded under TCs with 100-year pressure by 2100 (SSP5-8.5). The averaged flood duration increases significantly (from 10.72 h in the current climate to 11.78 h), mainly due to the notable longer flood duration in the Jiulong River Estuary.

TCs with 1000-year pressure can dramatically flood Xiamen Bay with an area of over 680 km² by the year 2100 under climate change projection SSP5-8.5, over half of which occurs in the Jiulong River Estuary. Compared with the flood area of the current climate, a significant increase in flood areas (over 20 km²) is found in the Jiulong River Estuary and the west Xiamen water. The flood depth also increases substantially in all six parts of Xiamen Bay. Especially in the west Xiamen water, the flood depth increases by as much as 1.3 m compared to the current climate. This may due to there being nearly no low-lying land that can be flooded further, as an increase of flood area of only 6 km² occurs compared with that induced by TCs with 100-year pressure. The limited low-lying areas lead to only 0.5 h increase in flood duration in the west Xiamen water, but in the Jiulong River Estuary with large-scale low-lying areas available for flooding, the flood duration experiences a significant increase of 1.2 h.

Figure 9 shows the land exposure of different land use categories to flooding and the number of populations directly affected by each TC intensities. On the condition of current climate, over 80% of the area flooded periodically by astronomical tides are wetlands (57.5%) and water bodies (23.6%). No people is influenced by sole astronomical tides. With influence of TCs with a 10-year pressure, exposed artificial surface increases significantly from ~ 12 km² to ~ 77 km², and the flooded agricultural area expands by about 100 km². There are over 300 K people will be affected by flooding under this condition. This result implies that the flood risk increases notably since the high value of artificial surface and agricultural areas. Under the influence of TCs with 100-year pressure, the flooded areas of each land use category increase progressively. Since most wetlands has been flooded by TCs with 10-year pressure, the area of flooded wetlands increases little. While, the highly urbanization around the Xiamen Bay leads to the flooded artificial surface continues to increase by 56.8 km², which is the most significant category. And the affected population increases to 487 K at this condition. There are nearly 600 km² of flooded low-lying land when a TC with 1000-year pressure meets astronomical high tide. Among them, 29% are artificial surface, 26% are agricultural areas, and 38% are water bodies and wetlands. The flooded artificial surface continues to increase by over 40 km². The affected number of people grows to about 697 K.

It is projected that climate change will lead to more area exposure of each land use category and also more population exposure to flooding ( Figure 9 ). Under the SSP5-8.5 scenario by 2100, 38 K people will live below the high-tide line, implying tide-only coastal flooding event will occur frequently with SLR. Similar result is found in Sydney, Australia (Hague et al., 2020). Relative to current climate, there are 77 km² more area will be inundated by high tide. 39% of them are water bodies, 19% are wetlands, and 16% are agricultural areas. Under influence of TCs of the three intensities, the most remarkable increase in exposed area are all found in the artificial surface category, following by the agricultural area category. About 70% of the increase in exposed area are in the two categories. The affected population increases to 436 K, 665 K, and 901 K under influence of TCs with a 10-, 100-, and 1000-years pressure, respectively. This result means that flood disaster in the Xiamen Bay will be much stronger in the warmer future.

In the context of dramatic increases in flooded area and exposed population in the warmer future, adaption and protection of coastal communities is a critical question. A traditional simple option is reinforcement and heightening of coastal seawalls and dikes. Meanwhile, it is expected that the required seawall heightening is much higher than the SLR values, mainly due to the expected increase of wave run-up (e.g., Zhang et al., 2021), the higher extreme tides relating to stronger storminess, the larger tidal ranges due to hard coastlines (e.g. Feng et al., 2019; Zhang and Li, 2019) and the accelerating land subsidence relating to rapid coastal urbanization (Syvitski et al., 2009). Seawalls and dikes are costly in construction and maintenance, but only provide protection from floods under a given hazard severity. If the flood exceeds such severity, serious disaster may occur. Typical examples are Hurricane Katrina flood in 2005 on the northern Gulf of Mexico (1392 casualties and 125 billion USD economic losses) and Storm Xynthia flood in 2010 over the Europe (64 reported casualties and 4.5 billion USD economic losses).

Alternative option is to build soft flood defenses based on natural ecosystems (Temmerman et al., 2013). Vegetated foreshores can substantially lower sea levels and reduce incoming wave energy (Mi et al., 2022), resulting in less constraining on seawall design, thus reduce initial construction costs and maintenance costs (Chaumillon et al., 2017). Such soft flood defenses for coastal protection is increasingly become a consensus for many countries and regions, such as the Netherlands and Mississippi River Delta in the United States.

Since the dense population and limited routes for crowed dispersal in Xiamen Island, it is necessary to reinforce and heighten the seawalls around Xiamen Island, especially that southwest of the island. In addition, current coastal wetlands and green space must be well protected, and restored as they act as key soft defenses, protecting population and property from exposure to flood hazards (Arkema et al., 2013), but are strongly reclaimed since 1950s (Wang et al., 2010). As a coastal city, there are also other solutions for adaptation to future stronger coastal flooding, including optimizing urban drainage systems to increase resilience to extreme flood events (Ferrans et al., 2022), and constructing flood-resistant buildings (Amini and Memari, 2020).

This study gives a reasonable estimation of flooded area and affected population in Xiamen Bay under three TC intensities based on a high-resolution numerical model. the flood situations in warmer future are also projected. Climate change induced SLR and TC climatology changes substantially exacerbate the flood risk in Xiamen Bay. About 70% of the increased exposed area are artificial surfaces and agricultural areas. The affected population also increase significantly. Xu et al. (2016) adopted an approach based on bathtub method to have an assessment of the flooded risk of Xiamen City, which is located in Xiamen Bay. As a result, 116-171 km² land and 440-720 K population will be flooded under 4.75-5.86 m extreme sea-level. The maximum water level at station Xiamen under influence of TCs with a 10- years pressure by 2100 is 5.84 m. In our estimation, there are only 101 km² flooded land (exclude the flooded area in Jiulong River Estuary and Dadeng water) and less than 430 K affected people. The flooding in our estimation is obviously weaker than the assessment by Xu et al. (2016). Since the dynamical tide approach is more suitable for assessing coastal inundation (Neumann and Ahrendt, 2013; Hsiao et al., 2022), we are confident in our estimation.

We acknowledge that there are several limitations in this study. The most important is that the exclusion of fluvial flood. TCs always bring simultaneous heavy rainfall, and it is expected that TC rainfall is in an increasing trend of about 1.3% per year (Guzman and Jiang, 2021). The compound flooding can result in higher flood depths and larger flood extents along the river and coastlines compared to their individual occurrences (Svensson and Jones, 2004; Hsiao et al., 2021). That means flood risk in Xiamen Bay may be underestimated in this study. While, the annual mean river discharge of the Jiulong River is ~ 380 m³/s, and the recorded highest discharge is 15540 m³/s. Even under such high discharge condition, the fluvial flooding only raises water levels in the upper Jiulong River Estuary, but throw little influence on those in the lower Jiulong River Estuary and around Xiamen Island (Yang et al., 2017). Since we focus on the flooding in the whole Xiamen Bay, the influence extent of pluvial flood is small compared to the study area.

Moreover, we neglect the remarkable spatial variation in SLR, which is mainly related to vertical land movement, steric effects and mass redistribution (Brown et al., 2016; Qu et al., 2019). While, Feng et al. (2019) found that it is highly similar between tidal response to spatially-varying SLR and that to uniform SLR. And sea level at Xiamen is projected to rise 0.51-1.94 m by 2100 under RCP 8.5, and the spatial difference of SLR along the Fujian coasts is small (Qu et al., 2019). Although the 0.77 m SLR used in this study from IPCC (2021) may not be the exact value in Xiamen Bay, it is in the projected SLR range and can be acceptable. In addition, we only consider TC climatology changes by simply enlarging the pressure decrease, maximum wind speed, and the maximum wind radius according to previous studies. But studies show that all the TC features will be changed by a warming climate, including the tracks, translation speed, and also structures (IPCC, 2021). However, because this study captured the essence of the impact of climate change on sea level and TC climatology, these limitations do not undermine our conclusions.