Risk assessment for typhoon-induced storm surges in Wenchang, Hainan Island of China

Abstract A risk assessment for storm surge disasters can provide scientific support for coastal management as well as marine disaster prevention and mitigation. By taking the Wenchang City of Hainan Province as a pilot area, a risk assessment approach for typhoon-induced storm surge disasters is introduced in detail in this article. First, a numerical simulation system for storm surge inundation is developed, which is applied for the simulation and calculation of the probable maximum storm surge in Wenchang, and the obtained inundation areas and depths are used to assess the storm surge hazard. Then, the data of land use and disaster-bearing bodies are used to classify and assess the vulnerability of Wenchang. Finally, by taking the community as unit, the risk assessment for typhoon-induced storm surge affecting Wenchang is performed by combining the assessment results of both hazard and vulnerability, thereby obtaining a level map of risk distribution. The results show that there are risks of typhoon-induced storm surge in most of the coastal areas of Wenchang, especially for its northeast and east coastal areas, where the risks reach Levels I–II. This approach can be conveniently applied to risk assessments of storm surge in other coastal areas. Note that this approach focuses on the current risk of typhoon-induced storm surge in coastal areas, and further studies will be conducted on the assessment methods for the potential risk.


Introduction
China's east coast, bordering the Northwest Pacific, is the area with the most frequent and widespread typhoon-related disasters over the world. Typhoon-induced storm surges have the potential to threaten public safety and cause economic losses in many to storms in Cadiz Bay Natural Park in Southwest Spain and drew risk zoning maps. Based on the Ise-wan Typhoon, the largest typhoon event in recorded history, Japan has created the storm surge hazard map from the results of possible maximum storm surge (PMSS) inundation for Japan's coastal areas, providing a scientific support for the disaster prevention and mitigation of the government (Cabinet Office 2004). In addition, the risk assessments of storm surge have also been piloted at the national, provincial, city and county scales in China. Moreover, key technologies such as the calculation methods of the PMSS have been developed (Wang et al. 2018), and the guideline for the risk assessment and zoning of storm surge disasters has been further compiled. By taking Wenchang of Hainan Province as the pilot area, this article introduces in detail the approach for risk assessment of typhoon-induced storm surge,  Figure 1(a) denote the locations of Haikou and Qinglan stations whose measurement data are used in Figure  3(a-f). Circles in Figure 1(b) denote the locations of stations in Table 2. which provides significant scientific support for the siting and fortification of major coastal engineering projects, the decision making of government for disaster prevention and mitigation as well as the evacuation of people during disasters. Besides, the approach can be conveniently applied to the risk assessment of storm surge in other coastal areas.
Wenchang City is located in the northeastern Hainan, which experiences the most frequent and severe typhoon-induced storm surge disasters in the South China Sea region. Hence, we choose Wenchang as the study area.
The remainder of this article is organized as follows. Section 2 describes the study area, assessment method, datasets, model configuration and verification. The results and analysis are presented in section 3, including the hazard assessment, vulnerability assessment and risk assessment. The conclusions are summarized in section 4, together with a discussion of broader issues.

Study area
Wenchang (19 21'N-20 10'N, 110 28'E-111 03'E) is located in the northeastern Hainan, which is bordered on the east and southeast by the South China Sea and on the north by the Qiongzhou Strait (Figure 1(a)). The elevation of Wenchang decreases from its southwest (inland area) to northeast (coastal area), and Dongzhai Harbor in the north and Bamen Harbor in the south are vulnerable to storm surge inundations due to the lower elevation. Wenchang is susceptible to the typhoons from the Western Pacific and South China Sea, and the typhoon-induced storm surges often bring great damages to infrastructures and coastal marine industries. Typhoon storm surges affecting Wenchang generally begin in May and end in November, with an average of 2-3 times per year and a maximum of 8 times. Super Typhoon Rammasun (No. 1409) made landfall in Wengtian Town, Wenchang City, Hainan Province, on 18 July 2014, with a central pressure of 910 hPa and a maximum wind speed of 60 m/s. During the landfall, the highest total water elevation in Dongzhai Harbor reached 4.5 m, and the storm surge disaster caused six deaths (including missing) and direct economic losses of 2.732 billion yuan (MNR 2015). On 16 September 2014, Typhoon Kalmaegi (No. 1415 landed in Wengtian Town, with a central pressure of 960 hPa and a maximum wind speed of 40 m/s. The maximum total water elevation at Xiuying tide gauge station reached 4.52 m, setting a historical record. The direct economic losses were up to 926 million yuan (MNR 2015).

Method
The storm surge risk assessment in Wenchang is performed by four steps: data collection and processing, hazard assessment, vulnerability assessment and risk assessment. The data to be collected and processed include historical hazards, basic geographic information, land use data and the data of disaster-bearing bodies in Wenchang. Moreover, based on the conditions of data collection, additional field surveys may be required if necessary. For the hazard assessment, we first determine the typhoon- induced PMSS based on the characteristics of storm surge in Wenchang, and then, calculate the inundation area and submerged depth through numerical simulation to classify the risk levels. The vulnerability assessment adopts land use types as an indicator to perform the qualitative vulnerability classification. The vulnerability level can be adjusted according to the actual situation when there is a critical disaster-bearing body in the assessment unit. The risk assessment of storm surge disaster is carried out based on the hazard assessment and the vulnerability assessment in the study area. Finally, zones with different risk level are divided by considering the spatial homogeneity of risk level distribution and administrative division comprehensively.

Datasets
The datasets adopted in this article are shown in Table 1. Here, the ocean bathymetry, Digital Elevation Model (DEM) terrain and seawall data are used to build the numerical model for the simulation of typhoon-induced storm surges. The wind field data is used to establish the external forcing field for the numerical model. The typhoon observation data is used to validate the performance of the model and provides data support for the risk assessment of storm surge in Wenchang. The data of land use, seawall and disaster-bearing bodies are used for the vulnerability assessment. The vertical datum used in this study is the local mean sea level, and datasets with other vertical datum are converted to the local mean sea level (Table 1).   (Figure 2(a)). The land area contains the area below the 10-m contour in Wenchang (Figure 2(c)). The grid resolution reaches 22-50 m in the coastal area of Wenchang. Besides, the model range contains 446,715 grid points and 874,441 triangular grids. ETOPO1 global data (Amante and Eakins 2009) with a resolution of approximately 1.8 km is used for oceanic bathymetry with nearshore modifications, based on the sea chart bathymetry from the Hydrographic Office of the Chinese Navy (Figure 2(a,b)). The topography over land is described by the DEM terrain data with a resolution of 0.025 km (Figure 2(c)). The time step of the model is set to 1 s, and the initial water elevation and flow rate are 0. A wet and dry scheme is used to simulate inundation and flooding. Tidal elevations at the open boundary during the entire simulation period are computed by using the Oregon State University Tidal Prediction Software (Egbert et al. 1994;Egbert and Erofeeva 2002) with eight tidal constituents (M2, S2, N2, K2, K1, O1, P1 and Q1) being included. The total water elevation or storm surge simulation results are outputted per hour. The storm surge model is driven by wind stress and atmospheric pressure gradient acting on the surface using the combination of Jelesnianski parametric wind model (Jelesnianski 1966) and National Centers for Environmental Prediction (NCEP) Final Operational Global Analysis (FNL) reanalysis datasets ( Table 1). The Jelesnianski parametric wind model is used when radial distance from the typhoon center < 500 km: where, r is the radial distance from the typhoon center; V G and P r are wind speed and atmospheric pressure, respectively, as functions of r; p a and p c are the ambient atmospheric pressure and typhoon central pressure, respectively; R max is the radius of maximum wind speed (RMW); and V max is the maximum wind speed.  (Pawlowicz et al. 2002), and surges (including effect of waves) are measured by subtracting astronomic tidal elevations from the total elevations. From Figure 3, it can be found that the simulated astronomic tide and storm surge changes are generally consistent with observations. During Typhoon Nesat, the simulated tidal phases have a correlation coefficient of 0.76 at Haikou and 0.84 at Qinglan stations, and the simulated surge phases are strongly correlated with observed surges,  with correlation coefficients > 0.86 (all correlation coefficients exceed the 95% confidence level based on the Student's t-test). The tidal amplitudes at Qinglan station are nearly identical as observed, and the peak surges are overestimated <10% compared to observations at both stations (Figure 3(a,b)). During Typhoon Rammasun, all of the correlation coefficients between simulated and observed results are very high (>0.83), especially with regard to the tide at Qinglan and surge at Haikou stations (>0.9). The tidal amplitude at Qinglan station is simulated accurately, and the peak surges have approximately 10% biases at both stations (Figure 3(c,d)). During Typhoon Kalmaegi, the simulated and observed phases have correlation coefficient > 0.84 at Haikou and Qinglan stations, and the peak tides and surges are estimated nearly precisely (Figure 3(e,f)). In general, the performances of the simulations are excellent, demonstrating that the numerical model with the external forcing could be considered reliable for the hazard assessment.

PMSS evaluation
The Joint Probability Method (JPM) and the Joint Probability Method with Optimal Sampling (JPM-OS) are recommended in PMSS evaluation by FEMA and researchers Resio et al. 2009;Niedoroda et al. 2010;FEMA 2014), which could provide probabilistic surge and flood elevation maps with various annual chance of occurrence. The JPM adopts a parametric storm description involving five or six hurricane descriptors and it develops probability distribution for each parameter. These distributions are each discretized into a small number of representative values, and all possible parameter combinations are simulated using a hydrodynamic model (FEMA 2014). To reduce the computation burden, the JPM-OS is developed, which selects storms for simulation through optimal parameter selection with associated weighting and interpolation methods Resio et al. 2009). Yang et al. (2019) compares the results of JPM and JPM-OS, and finds that results of JPM-OS and JPM are very similar. In addition, they compared the performance of various JPM-OS methods. However, in this article, we do not adopt the JPM and JPM-OS approaches due to the lack of adequate storm surge observations and the shortage of storm climatology knowledge in Wenchang. We employ a deterministic-probabilistic approach (Wang et al. 2018) to explore the storm surge climatology through sensitivity experiments and evaluate the PMSS. The hazard assessment in Wenchang is calculated based on the inundation area and submerged depth of the PMSS. In terms of the PMSS in Wenchang, based on historical observations and the results of sensitivity experiments, a set of possible maximum typhoons within a certain area is constructed to simulate the maximum storm surge and maximum total water elevation. The key parameters selected include typhoon central pressure, peripheral pressure, RMW, the typhoon approach speed and approach direction. The deterministic-probabilistic approach we use is described in a published article (Wang et al. 2018), and in this article we just cite the conclusions.
3.1.1.1. Typhoon intensity. The mean value of sea level pressure during the typhoon season (from May to October) in Wenchang, which is 1008 hPa, is taken as the peripheral pressure.
The area with a radius of 400 km around Wenchang is set as the assessment area. Based on the typhoon best-track datasets of 1949-2015, the lowest central pressures of the typhoons that have passed through the assessment area over the years are counted. A Gumbel distribution (Gumbel 1958) of annual typhoon pressure extremes revealed the lowest typhoon central pressure for a 1000-year return period was 866 hPa. Therefore, the pressure deficit of 142 hPa (the pressure difference between the peripheral and the central regions) is set as the typhoon intensity of PMSS in Wenchang (Figure 4).
3.1.1.2. Radius of typhoon maximum wind speed (RMW). The lack of RMW observations brought us to empirical methods. Based on the historical observations from Tropical cyclone yearbook  and previous empirical statistical formula, Jiang et al. (2008) obtained a RMW empirical equation that could apply to the Wenchang area: where R max denotes the RMW, DP is the typhoon pressure deficit. Based on the typhoon pressure deficit of 1000-year return period (142 hPa) calculated in the previous chapter, the RMW of the PMSS in Wenchang is determined as 20.6 km.   (Wang 1989). Considering the complex and nonlinear storm surge responses to typhoon approach directions along the irregular coastline of Wenchang, the storm surge cases during the typhoon's landfall are constructed with h of 67.5 , 75 , 82.5 and 90 , and the PMSS parameters obtained above are adopted as the values of typhoon intensity, RMW and approach speed. Table 2 shows that the peak storm surges at 16 representative stations (as seen in Figure 1(b)) along the Wenchang coast all reach the extreme values when h of 67.5 and 90 are chosen. Therefore, 67.5 and 90 are chosen as the typhoon approach direction for PMSS.
3.1.1.5. Typhoon tracks. According to the above calculation, a central pressure of 866 hPa, an peripheral pressure of 1008 hPa, a RMW of 20.6 km, an approach speed of 20 km/h as well as two approach directions of 67.5 and 90 are determined for the PMSS cases. On this basis, 30 tracks are constructed for the approach direction of 67.5 at an interval of 0.25 times of the RMW (Figure 5(a)), and each track corresponds to a PMSS case. The similar procedure is performed for the approach direction of 90 ( Figure 5(b)), so a total of 60 PMSS cases are simulated, covering all of the typhoon tracks that may cause strong storm surges in Wenchang. Note that the continuous computation lasts 90 h for each track, and Figure 5 reveals the typhoon start, landfall, and termination locations of all the PMSS cases.

Result analysis
The envelope diagram of inundation ranges and submerged depths for 60 PMSS cases (as shown in Figure 6(a)) is used to perform the hazard assessment of typhooninduced storm surges in Wenchang. Obviously, there are different extents of inundation along the coast of Wenchang, most notably on the eastern side of Dongzhai Harbor and near Bamen Harbor. The reason is that the terrain on the eastern side of Dongzhai Harbor in the northern Wenchang is relatively flat and the seawall is not closed, which makes it easy for seawater to overflow from the north and south sides of the seawall to the land during storm surges and cause inundation, with the submerged depths up to 2-5 m. Besides, in Bamen Harbor in the south of Wenchang, during the storm surges, the upstream movements of seawater pass through several rivers, resulting in the inundation in low-lying coastal areas with the submerged depths of up to 2-6 m.
For the hazard assessment of storm surges affecting Wenchang, the submerged depth is used to classify the hazard levels (Table 4), and the community is set as the evaluation unit. The highest hazard level of each grid in the community is determined as the hazard level of the evaluation unit, thereby obtaining the hazard level distribution of storm surges affecting Wenchang (Figure 6(b)). The hazard level on the east side of Dongzhai Harbor is Level I, for instance, parts of Puqian Town are inundated to a depth of more than 3 m. Most of the areas along the coast of Bamen Harbor are submerged to a depth of more than 3 m, especially in areas where the rivers pass through, and the hazard level reaches Level I, such as some communities in Towns Dongjiao and Wencheng. The hazard Level II appears in the communities in the northern Wenchang and most of the communities along the coast in the east, such as Towns of Jinshan, Wengtian, Changsa and Longlou, with submerged depths between 1.2 m and 3 m. The hazard Levels III-IV (submerged depth of 0.15-1.2 m) are dispersedly distributed in the inland areas of Wenchang. The other inland towns, such as Baoluo and Donglu, are classified as no inundation risk zones due to a submerged depth of less than 0.15 m.

Vulnerability assessment
Based on the land use status data and taking the community in Wenchang as the evaluation unit, the vulnerability assessment is performed by the vulnerability coefficients corresponding to each land use type and the weighted comprehensive evaluation method. The vulnerability coefficients (shown in Table 3) adopt values from the operation guidance for storm surge risk assessment in China, which are developed through an expert scoring method and extensive discussions with central/local governments and groups (MNR 2019b). The evaluation equation is as follows.
where, A represents the vulnerability coefficient of a certain community, a i is the weight for the land use type of No. i, V i is the vulnerability coefficient for the land use type of No. i, and n is the number of the land use types. In terms of the assessment units with some critical disaster-bearing bodies, the vulnerability coefficients of the communities need to be revised by considering the vulnerability coefficients of critical disaster-bearing bodies and their corresponding weights. On this basis, the distribution of vulnerability levels in Wenchang city with the community as the assessment unit during storm surges (Figure 7) is obtained based on the relationship between vulnerability levels and vulnerability coefficients for storm surges (Table 4).
Analysis shows that the land use types with the widest distribution in Wenchang are cultivated land, garden plot and forestland, accounting for about 54.9% of the total area of Wenchang city. Areas with higher vulnerability levels include lands used for urban construction, mining, port terminals, railway and highway. The important disaster-bearing bodies in Wenchang are scattered and account for a relatively small proportion of the land area, among which the port terminals are mainly located along the southeast and northwest coasts of Wenchang. Figure 7 shows that the vulnerability Level I mainly distributes in the northeastern Wenchang where 10 communities in total are included, which belong to urban land, railway land, etc. Level II is mainly in the northeastern and eastern coasts of Wenchang, with 20 communities being included. Level III mainly distributes in the eastern coasts of Wenchang, affecting 37 communities. Most areas of Wenchang belong to Level IV, including 223 communities, belonging to cultivated land, garden plot, forestland, etc.

Risk assessment
The risk assessment model for storm surge disaster is as follows.
where, R denotes the risk, H is the hazard, and V is the vulnerability. The risk level is evaluated by combining the hazard level and the vulnerability level based on the unit of community. The risk area for storm surge hazard is divided into four levels: extremely high risk zone (Level I), high risk zone (Level II), medium risk zone (Level III) and low risk zone (Level IV). Where an area has no inundation, it is considered that there is no risk in this area (Table 4). Figure 8 shows that there is a risk of typhoon-induced storm surge disasters in most of the coastal areas of Wenchang. Risk zones of Level I are mainly located in the northeastern Wenchang, including some communities in Towns of Puqian, Jinshan and Wentian. All of those zones have a hazard level of above Level II, while the vulnerability level ranges from Level I to II due to a dense population distribution. Risk zones of Level II are mainly distributed along the eastern coast of Wenchang, including Towns of Wentian, Changsa and Longlou, where the hazard level is Level II and the vulnerability level ranges from Level II to III. Note that there is a hazard Level I along the coast of Bamen Harbor, but with a vulnerability of Level IV. Besides, risk zones of Level III are mainly located in the eastern coastal area of Dongzhai Harbor, with a high hazard level and a high risk of inundation. Nevertheless, the vulnerability level is only Level IV due to the predominance of cultivated land and forestland, so the risk level in these zones is low. The other inland regions are classified as risk zones of Level IV, or no-risk zones.

Conclusions and discussion
Taking the typhoon-induced storm surge disasters in Wenchang as examples, this article introduces an approach of risk assessment for storm surges, which can be widely applied to other coastal regions. The approach consists of four parts-data collection and processing, hazard assessment, vulnerability assessment and risk assessment. The data of ocean bathymetry, DEM terrain, wind field, typhoon observations, land use and disaster-bearing bodies in Wenchang City are adopted in this article, with the spatio-temporal scales, resolutions, sources and processing methods for each data described in detail.
During the hazard assessment, based on the models of ADCIRC and SWAN, we build a numerical simulation system for typhoon-induced storm surges in Wenchang, and ensure that the system could catch the main characteristics of storm surge disasters by comparing the simulations of historical cases with observations. Based on the typhoon central pressure, the typhoon RMW, typhoon approach speed and direction, peripheral pressure and astronomical tide determined in this article, the PMSS cases in Wenchang were constructed, and the simulated inundation ranges and submerged depths were used to determine the hazard levels of storm surge disasters in this region. The results show that the hazard level along the coastal area of Wenchang is relatively high (Levels I-II), especially in the east coast of Dongzhai Harbor and the coastal area of Bamen Harbor, where the maximum submerged depth can reach 2-5 m along Dongzhai Harbor and 2-6 m along Bamen Harbor.
By using the data of land use status and disaster-bearing bodies, the vulnerability assessment was conducted based on the corresponding relationship between the classification of land use and the vulnerability levels, which are further classified by taking the community as a unit. Areas with vulnerability Levels I-II in Wenchang account for about 10.6%, mainly in the communities of the northeastern and eastern Wenchang. The areas with vulnerability Level III are mainly located in the eastern coastal area of Wenchang. Most areas of Wenchang are in vulnerability Level IV, which belong to land use types of cultivated land, garden plot and forestland.
The classification of risk assessment integrates both the assessments on hazard and vulnerability. The results show that there are risks of typhoon-induced storm surge disasters in most of the coastal areas of Wenchang. Risk zones of Levels I and II mainly locate in the northeastern Wenchang, the coastal area of eastern Wenchang and Bamen Harbor. Risk Level III mainly appears in the eastern Dongzhai Harbor, and the other inland areas belong to zones of risk Level IV or no risk.
The results of risk assessment for typhoon-induced storm surge disasters play an important role in the disaster prevention and mitigation as well as coastal management for local government, and some targeted and hierarchical risk management measures can be carried out accordingly. For example, in risk zones of Levels I-II, it is possible to raise the flood control standards of seawalls and coastal engineering, and adjust the plan of land use to demonstrate the feasibility of the newly built major coastal projects for storm surge disaster prevention. In addition, it is also essential to establish a comprehensive monitoring and early warning system for storm surge disaster, setup marine disaster shelters, prepare and publish emergency evacuation maps. Furthermore, enhancing the public awareness of marine disaster prevention and mitigation, conducting publicity and education as well as emergency drills of marine disasters are also necessary. In risk zones of Levels III-IV, a strategy of accepting the risk can be adopted, but the monitoring and early-warning for storm surge disasters need to be strengthened, and necessary measures for disaster prevention and mitigation should also be taken.
There are several discussions for this study, as follows.
1. The purpose of this study is to provide a risk assessment for storm surge disasters which is a scientific support for coastal management as well as marine disaster prevention and mitigation. An intuitive and concise risk map is preferred by the local decision makers, so the PMSS scenario is selected to perform the hazard assessment, and then the classification of risk assessment is obtained by integrating the assessments on hazard and vulnerability. The PMSS scenario, which is composed of flooding maps that could capture the range of storm surge possibilities in Wenchang, has taken the most severe disasters into account and is suitable for decision making.
2. The disadvantage of our method is that the flood map is not probabilistic, and the uncertainty is difficult to analyze. However, the shortcomings of our method could be made up by JPM/JPM-OS methods. The JPM/JPM-OS methods could provide probabilistic flood maps corresponding to any return period, and the results are scientifically more robust. However, the methods require abundant observations and a knowledge of the storm climatology of the study region, and JPM-OS method will be used in the future PMSS evaluation when conditions permit. 3. The risk assessment is obtained based on the PMSS, so the results of risk assessment fall within the ranges of possible maximum levels, thus, there is limited guidance for real-time warning of storm surge disasters. For the real-time evaluation on risk levels of typhoon-induced storm surges, it is necessary to assess the risks of typhoon-induced storm surges with different levels and tracks, thereby establishing a multi-scenario risk database of typhoon-induced storm surges. 4. The risk assessment in this study focuses on the current risks in a certain area, but the potential risk of storm surge disasters is poorly estimated. For example, the eastern coastal area of Dongzhai Harbor has a very high hazard level (Level I) but a low vulnerability level (Level IV), so it is classified as the risk zone of Level III. Therefore, it is not easy to attract the attention of administrators in disaster prevention and mitigation. However, if sensitive disaster-bearing bodies appear in the future, these regions will face a very large risk of storm surge disasters. Therefore, weight coefficients will be introduced in future work, thus both the current risk and potential risk could be comprehensively considered in risk assessments of marine disasters. 5. It is important to keep the assessment results up-to-date with different environment situations. In the background of changing risk due to sea-level rise, protective measures such as newly built seawalls and significant changes of disasterbearing bodies or land use, the risk of typhoon-induced storm surges should be re-evaluated so as to provide more accurate supports to the disaster prevention and mitigation.

Disclosure statement
No potential conflict of interest was reported by the authors.

Funding
This work was supported by the Integration and Application Demonstration of Marine Dynamic Disaster Observation and Early Warning System (2018YFC1407000).