Assessing and zoning of typhoon storm surge risk with GIS technique: A case study of the coastal area of Huizhou

Storm surge is one of the most destructive marine disasters to life and property for Chinese coastal regions, especially for Guangdong province. In Huizhou city, Guangdong province, due to the high concentration of chemical and petroleum industries and the high population density, the low-lying coastal area is susceptible to the storm surge. Therefore, 10 a comprehensive risk assessment of storm surge over the coastal area of Huizhou can delimit zones that could be affected to reduce disaster losses. In this paper, typhoon intensity for the minimum central pressure of 880 hPa, 910 hPa, 920 hPa, 930 hPa, and 940 hPa (corresponding to 1000-year, 100-year, 50-year, 20-year, and 10-year return period) scenarios were designed to cover possible situations. The Jelesnianski method and the Advanced Circulation (ADCIRC) model coupled with the Simulating Waves Nearshore (SWAN) model were utilized to simulate inundation extents and depths of storm surge over 15 the computational domain under these representative scenarios. Subsequently, the output data from the coupled simulation model (ADCIRC–SWAN) were imported to Geographical Information System (GIS) software to conduct the hazard assessment for each of the designed scenarios. Then, the vulnerability assessment was made based on the dataset of land cover types in the coastal region. Consequently, the potential storm surge risk maps for the designed scenarios were produced by combining hazard assessment and vulnerability assessment with the risk matrix approach. The risk maps 20 indicate that due to the protection given by storm surge barriers, only a small proportion of the petrochemical industrial zone and the densely populated communities in the coastal areas were at risk of storm surge for the scenarios of 10-year and 20year return period typhoon intensity. Moreover, some parts of the exposed zone and densely populated communities were subject to high and very high risk when typhoon intensities were set to a 50-year or a 100-year return period. Besides, the scenario with the most intense typhoon (1000-year return period) induced the very high risk to the coastal area of Huizhou. 25 Accordingly, the risk maps can help decision-makers to develop risk response plans and evacuation strategies in coastal communities with the high population density to minimize civilian casualties. The risk analysis can also be utilized to identify the risk zones with the high concentration of chemical and petroleum industries to reduce economic losses and prevent environmental damage caused by the chemical pollutants and oil spills from petroleum facilities and infrastructures that could be affected by storm surge. 30 https://doi.org/10.5194/nhess-2020-130 Preprint. Discussion started: 12 June 2020 c © Author(s) 2020. CC BY 4.0 License.


Introduction
Storm surge is the abnormal rise in sea level during tropical cyclones, and the surge is primarily produced by strong storm winds pushing water into shore. When a tropical cyclone makes landfall, the accompanying storm surge will lead to significant flooding in the surrounding coastal area. Therefore, storm surge associated with tropical cyclones is a devastating hazard and 35 frequently causes considerable deaths and property damage in many coastal regions of the world. For example, in August 2005, Hurricane Katrina struck the United States, and the storm surge from Katrina along the Gulf coastal area was ranging from 10 to 28 feet high. It caused 108 billion dollars in property damages and resulted in a death toll of 1833 persons (National Oceanic and Atmospheric Administration, 2016). In 2019, typhoon Hagibis caused severe storm surge flooding that inundated southern areas of Honshu and killed at least 99 people in Japan (Asian Disaster Reduction Center, 2019). 40 In China, storm surge is regarded as one of the most seriously marine disasters, which inflicts tremendous losses to  China, 2018). These economic damages and casualties will be further 45 increased because of the explosive population growth and the rapid development of society in the coastal area of China (Seto et al., 2013;Lichter et al., 2011;McGranahan et al., 2007). Moreover, with considerably rising sea levels caused by continued global warming, tropical cyclone-induced storm surge will be more destructive in the future (Intergovernmental Panel on Climate Change, 2018). Therefore, it is important to establish storm surge preparedness plans in advance to reduce economic losses and casualties for coastal cities. 50 Huizhou is the eastern city in the Pearl River Delta region in Guangdong province, China. In Huizhou, the petrochemicals and electronic information have developed into the dominant industries, with an annual production value of more than one trillion RMB. The Daya Bay Petrochemical Zone, located in the coastal area of Huizhou, is currently ranked the first in China in terms of the scale of integrating refinery and petrochemical production. The high concentrations of petroleum refining facilities and energy infrastructures in the Petrochemical Zone and the high density of population in low-lying coastal 55 communities make the coastal area of Huizhou especially vulnerable to storm surges. The storm surge risk that the coastal area of Huizhou faces could be increased with sea-level rise, population growth, and further petrochemical industry development.
Thus, it is necessary to assess the potential risk of typhoon storm surge in advance in Huizhou to help decision-makers understand the affected regions and allow them to develop mitigation strategies and future land use planning.
The terminologies and methods for coastal risk assessment vary in different scholars and organizations. In most case, the 60 risk assessment of storm surge flooding is determined by the combination of hazard, exposure, and vulnerability (Chrichton, 1999;Kaźmierczak and Cavan, 2011;Koks et al., 2015): (1). The hazard is defined as a natural event that causes impacts on people and infrastructures. The hydrodynamic models, wave models, and the statistical methods are applied to the qualitative evaluation of storm surge hazard. With the statistical

Study area
Guangdong is a coastal province, located in the southernmost part of China, and has a long coastline along the South China Sea, as shown in Figure 1 (a). Guangdong is one of the most prosperous provinces in China with the highest GDP of 9.73 trillion RMB and a population of 113.46 million in 2018. However, due to the geographical position, Guangdong is the most frequently affected province by tropical cyclones in China. The storm surge is regarded as the most seriously marine disaster 110 for the Guangdong province. During the period from 1949 to 2017, 263 tropical cyclones landed in Guangdong province (Ying et al., 2014). In 2018, 3 different typhoons including Ewiniar, Bebinca, and Mangkhut made landfall on Guangdong province, which left 4 people dead and caused a direct economic loss of 2.37 billion RMB.
The Huizhou city is located in the southeastern area of Guangdong province, and it occupies part of the Pearl River Delta megalopolis to the northeast of Hong Kong and Shenzhen. It, spanning from 22˚4'N to 23˚57'N latitude and 113˚51'W to 115 115˚28'W longitude, covers a land area of about 11347 km 2 and sea area of approximately 4520 km 2 . There are two districts (Huicheng and Huiyang) and three counties (Boluo, Huidong, and Longmen) in Huizhou, as shown in Figure 1   with academic publishing permission from Global Administrative Areas (https://gadm.org/license.html); (b) The map of Huizhou city and the study area, which was made using the ArcGIS 10.5; (c) The map of towns and their boundaries in the study area and the petrochemical 125 buildings are distributed in the Data Bay Petrochemical Zone, which was made using the ArcGIS 10.5. The maps and satellite images obtained from Google Earth or Google Maps can be used and printed in the research papers with permission from Google's website (https://www.google.com/permissions/geoguidelines/). In this paper, the region within a distance of 10 kilometers from the coastline in Huiyang district and Huidong county is chosen as a study area to understand the potential risk of storm surge in this region, as shown in Figure 1 (b) and Figure 1 (c). 130 In addition to high population density in coastal communities, the main reason for choosing this region is that high concentrations of petroleum facilities and infrastructures in the Daya Bay Petrochemical Zone make the study area vulnerable to storm surges. The Daya Bay Petrochemical Zone has an area of about 27.8 km 2 , as shown in Figure 1 (c), and is currently taking the first spot at the scale of petrochemical-refining integration in China. By the end of 2018, the Petrochemical Zone has been commenced business by many world's top 500 companies and industry-leading enterprises. These world chemical 135 and industrial giants including Exxon Mobil, Shell, and Clariant have invested 131.6 billion RMB to shape up the industrial chains of the oil refinery, ethylene, propylene, and butylene in Daya Bay Petrochemical Zone. In 2018, the oil refining capacity and the ethylene production capacity have been enhanced to 22 million tons/year and 2.2 million tons/year, respectively, and the petrochemical industrial output value reached 270 billion RMB (Huizhou Guangdong province, 2019;Huizhou, 2018). Now, the Daya Bay Petrochemical Zone is striving to develop into the world-class petrochemical base and planning to be one 140 of the world's top ten petrochemical industrial zones in the subsequent few years. Therefore, with the growing population density and particularly the rapid development of petroleum and chemical industries, the storm surge risk over the study area will be increased. The risk assessment and risk analysis are considered to be important strategies to identify the risk regions in the Daya Bay Petrochemical Zone, which can minimize the loss of life and property and prevent environmental damage caused by affected coastal petroleum facilities and infrastructures. 145

Datasets requirement
The datasets used in the study contain observed data and survey data obtained from various sources. The datasets can be employed to conduct the hazard assessment, vulnerability assessment, and risk assessment of storm surge in the study area for each of the different typhoon intensity scenarios. The datasets, which can be downloaded from the dataset (Si, 2020), are listed in Table 1 and described below. representative typhoon scenarios over the study area. The input parameters of each of the typhoon scenarios are used to generate wind field with the Jelesnianski method, which is the requirement for modelling storm surge.
2). The dataset of Digital Elevation Models (DEM). The dataset of DEM with a scale of 1:2000 was constructed in 2015 and is available from Huizhou Land and Resources Bureau. It is a raster dataset that depicts land heights in Huizhou's solid surface. The point in the dataset contains the elevation value for the region that point covers. The DEM dataset is used for 160 modelling storm surge. The DEM map of the study area is shown in Figure 2.   hourly water level records. The dataset can be used for validating the coupled model (ADCIRC-SWAN) over the study area by comparing simulated water levels and measured water levels.

Model description and validation 195
In this study, the Jelesnianski numerical method (Jelesnianski and Taylor, 1973), the well-established Advanced Circulation (ADCIRC) model (Luettich et al., 1992;Westerink et al., 1994) and Simulating Waves Nearshore (SWAN) model (Booij et al., 1999) are employed to simulate typhoon storm surge. The ADCIRC model is a two-and three-dimensional hydrodynamic circulation model, which can be utilized to model tides, wind-driven circulation, and storm surge. The ADCIRC model has been applied to simulate the hydrology in regions including the Gulf of Mexico, the Mediterranean Sea, and the South China 200 Sea (Kerr et al., 2013;Orlić et al., 2010;Li et al., 2020). The ADCIRC-2D in the study was run using a spherical coordinate system. It can provide both water surface elevation and the depth-averaged velocity of the current in coastal seas by solving continuity equation and momentum equations. As for the friction coefficient, Manning's n values derived from the dataset of land cover types are utilized. The dataset contains information about the types of land-cover related to Manning's values over the study area, which are listed in Table 2. These land types were first associated with Manning's values, and then 205 average the Manning's n values for the ADCIRC mesh. Open spaces 0.035 The SWAN model is a third-generation numerical wave model, which is used to simulate wind-generated wave propagation in coastal regions. The model computes the wave action density spectrum by solving the wave action balance 210 equation. It can be coupled to the ADCIRC model to simulate the storm surge on the same unstructured grid (Dietrich et al., 2011;Dietrich et al., 2012).
The computational domain in this study covered the coastal region of Huizhou, as shown in Figure   The procedure of modelling storm surge is as follows: the wind filed, which is generated by the Jelesnianski method, is 225 provided to the coupled model (ADCIRC-SWAN). Then, the ADCIRC model is operated to calculate the water level and current under the wind filed. Subsequently, based on the water level, the current, and the wind velocity, the SWAN model computes the wave spectrum, which is then passed back to the ADCIRC model to calculate the water levels in the next simulated round. Thus, the modelling typhoon event can be converted by the wave-current coupled model (ADCIRC-SWAN) into a storm surge event, outputting the data of water height corresponding to every grid node over the computational domain. 230 As the Jelesnianski method and the coupled model (ADCIRC-SWAN) have never been used in simulating typhoon storm surge over the coastal areas of Huizhou, the simulated performance of the coupled model is needed to be evaluated. The 10 representative typhoons (0812、0814、0906、1208、1319、1604、1622、1713、1720、1822), which caused high water levels in Huizhou gauging station and Gangkou gauging station (Figure 4), are selected to validate the coupled model (ADCIRC-SWAN) for the study area. Figure 6 shows all maximum simulated water levels, the highest observed water levels 235 from the 10 representative typhoons, and the timing of these peaks for these 10 representative typhoons. 245 Figure 6. The predicted water levels (in black line) and highest measured water levels (in red dots) recorded by the Huizhou station and Gangkou station during the typhoon events.

250
Two error statistic approaches are utilized to evaluate the performance of the coupled model with a comparison between maximum predicted water levels and the highest measured water levels. The Relative Error (RE) is calculated as the measured observed water level is above 100 cm. The Absolute Error (AE) is computed when the highest measured water level is below 100 cm. The statistical results from the two stations have been summarized in Table 3. It displays that the data with RE≤20% or AE≤20 cm account for 90% of all simulated data, which satisfies the criterion in the guideline. Therefore, the performance 255 of the coupled model is considered to be reliable regarding its ability to simulate storm surges in the study area.

Storm parameters
After validating the performance of the coupled ADCIRC-SWAN model, the input storm parameters must be set for designing the typhoon scenarios, which can be used to create the wind field with the Jelesnianski method to drive the storm surge in the coupled model. The storm parameters include the minimum central pressure, the radius of maximum winds, the maximum wind velocity, and the storm track. 265

Return period and minimum central pressure
The typhoon return period is the average time between typhoons with a certain intensity at a specific location. The typhoon return periods for Huizhou can be determined by making statistics of historical typhoon records in Huizhou. Based on the historical typhoon records, the minimum central pressures-return period relation for the study area was constructed with extreme value distribution Type-I, as shown in Figure 6. Thus, the minimum central pressure associated with a given return 270 period can be calculated. For example, it can be observed from Figure 6, that the minimum central pressure of typhoons corresponding to the return period of 1000-year was estimated at 880 hPa.

Radius of maximum wind
The radius of maximum wind (Rmax) has a strong connection to the maximum wind velocity and the minimum central pressure of the typhoon (P0). The researchers (Vickery et al., 2000;Cheung et al., 2007) developed empirical formulas to calculate the radius of maximum wind based on the Rmax-P0 relationship, as shown in Eq. from (1) to (3), where ∅ represents the latitude of the typhoon's center, ∆ indicates the pressure difference between central pressure and ambient pressure, Rk is empirical 280 constant with the value of 50.
Analyzing the historical radius of maximum wind data measured in the northwest Pacific hurricane records ( Therefore, according to the above empirical equations and historical observations, the radius of maximum wind corresponding to the minimum central pressure can be calculated. As shown in Table 3, averaging the empirical values from Eq. (1) to (4) and observed values, the radius of maximum wind with respect to the minimum central pressure at 880 hPa was 295 estimated at 30 km.

Maximum wind velocity
Based on the records from the dataset of Historical Tropical Cyclones, the regression statistic is made on the observed data containing maximum wind velocity (Vmax) and the minimum central pressure ( 0 ) with the wind-pressure empirical equation 300 (Atkinson and Holliday, 1977), as shown in Figure 8.

Tropical storm track
The tracks of tropical cyclones, which affected the coastal area of Huizhou during the period from 1949 to 2017, can be divided into four categories: moving northward (10.2% of total), moving northwestward (47.5% of total), moving west-southwestward 310 (15.5% of total), and moving northeastward (24.4% of total), as shown in Figure 9.   In order to provide complete geographical coverage for the study area, a set of deviated typhoon tracks were produced.
The 33 typhoon tracks deviating from the original Typhoon Mangkhut track were generated, as shown in Figure 11. These 34 335 tracks are spaced 5 km apart and represent the typhoon activity near Huizhou. In this paper, the tracks were used to simulate the storm surge in the coastal area of Huizhou with the Jelesnianski method and the coupled model (ADCIRC-SWAN).

Procedure for assessing risk
The procedure for conducting the risk assessment of storm surge in China is derived from the standard Technical guideline for risk assessment and zoning of marine disaster (Part 1: storm surge). The risk assessment is regarded as a combination of hazard assessment and vulnerability assessment. The procedure involves four steps: typhoon scenarios design, hazard 345 assessment, vulnerability assessment, and risk assessment as shown in Figure 12. As seen in Figure 12, the datasets used in the study are Land Cover, DEM, Historical Tropical Cyclones, Historical Sea 350 Level, Storm Surge Barriers, and District Boundaries. The wind field created with the Jelesnianski method is provided to the coupled (SWAN+ADCIRC) model, which simulates the storm surge for each of the design typhoon intensity scenarios.
Subsequently, the 12-hour time series data of simulated surge documented the temporal variation of inundation depth over the computational domain are generated. Then, these output data from the coupled (SWAN+ADCIRC) model for each scenario are converted to the ArcGIS 10.5 software. Eventually, the hazard maps, vulnerability maps, and risk maps for these 355 representative scenarios are made using the ArcGIS 10.5 software.

Typhoon scenario design
The typhoon scenario is parametrized by intensity, maximum wind velocity, radius of maximum winds, and track.
The lower the central pressure or the longer the year return period, the more intense the storm. Thus, the minimum central pressure or the year return period can be regarded as an indicator for the typhoon intensity. The comprehensive and 360 representative typhoon intensity for the minimum central pressure of 880 hPa, 910 hPa, 920 hPa, 930 hPa, and 940 hPa (corresponding to 1000-year, 100-year, 50-year, 20-year, and 10-year return period) scenarios were designed. The corresponding maximum wind velocity and radius of maximum winds were calculated for each of the designed scenarios according to the analysis in section 3.2.1-section 3.2.3, as shown in Table 4. The 34 constructed tracks discussed in section 3.2.4 were used as input data to create the wind field. 365

Hazard assessment
The hazard assessment is to identify the potential inundation extent and depth of storm surge caused by typhoon for the study area. The ADCIRC+SWAN model integrated with the Jelesnianski method was run on the datasets for scenarios with an increasing minimum central pressure from 880 hPa up to 940 hPa. Then, the outputs of the coupled model (ADCIRC-SWAN) 370 were imported to the GIS software.
The spatial extents of surge area and heights of surge water in given scenarios were displayed in the ArcGIS 10.5 software.
The inundation depth was calculated by subtracting DEM from the height of simulated surge water at each grid. The storm surge hazard for the study area was assessed based on the classifications of inundation depths as summarized in Table 5.
Accordingly, the different hazard levels were assigned to the inundation zones. 375 [300, +∞) Very High (I)

Vulnerability assessment
The exposure assessment aims at identifying elements affected by storm surge. The land cover can be considered as the 385 representation of affected elements. The land cover type is regarded as an indicator to assess the vulnerability in the study area to storm surge. The vulnerability values ranging from 0 to 1 are assigned to different land cover types, which were defined in the guideline according to their properties of susceptibility and resilience to storm surge. The value of 0 indicates no vulnerability and the value of 1 represents the highest vulnerability. The four levels of vulnerability were defined in the guideline (I, II, III, and IV) from very high vulnerability (I) to low vulnerability (IV). The Land Cover dataset was categorized 390 into 12 first classifications according to the guideline, as summarized in Table 6. Based on the vulnerability value corresponding to land cover type in Table 6, the vulnerability level over the study area was evaluated.

Risk assessment
As for the risk assessment, the inundated region is divided into several storm surge risk districts by integrating the inundation 395 hazard assessment and vulnerability of affected elements over the study area. The quantitative risk assessment and the risk matrix are the primary methods for evaluating risk. However, the quantitative risk assessment method is data demanding and it is difficult to quantify all populations and properties at risk. The risk matrix, a typical semi-quantitative approach, is utilized to solve these problems. The risk matrix is made of classes of hazard level on one axis and the vulnerability level on the other axis, as shown in two-dimensional Table 7. With the risk matrix approach, the degree of risk can be determined based on limited quantitative data. The degree of risk is evaluated by four levels (I, II, III, and IV) from very high risk (I) to low risk (IV). For example, low vulnerability combined 405 with low hazard can lead to a low risk, or the combination of very high vulnerability and low hazard can result in moderate risk in the area.

Hazard assessment
The coupled model (ADCIRC-SWAN) model and the Jelesnianski method were utilized to simulate the inundation extents 410 and depths for each designed typhoon scenario (Table 4). The 12-hours simulation of storm surge flooding over the study area for each of typhoon scenarios can be displayed in the ArcGIS 10.5 software. For example, for the typhoon intensity with the 880 hPa (1000-year return period), the simulated inundation depths and extents of storm surge at time intervals of 2 hours during the 12-hour period are shown in Figure 13. The inundation depth over the study area was divided into four categories according to the criterion in the guideline. 415 As shown in Figure 13, the inundation area progressively expanded from the coastline to the mainland and reached a maximum at approximately the 11th hour. Moreover, the maximum distance that storm surge flooding penetrates inland from the coastline is approximately 6 km and the inundation distances in other regions are less than 4 km. Furthermore, the simulated inundation depths over many coastal areas are more than 300 cm at the 11th hour. Figure 13. The maps display that the simulated inundation extents of storm surge over the study area during 12-hour simulation for the 880 hPa (1000-year return period) scenario. These maps were made using the ArcGIS 10.5 software based on the terrain base map layer, which was obtained from Google Maps (Map data ©2019 Google).

425
With the output data of simulated storm surge elevations at the 12th hour, the hazard map where regions are in different colors based on the inundation depths (Table 5) was made for each of the designed scenarios using the ArcGIS 10.5 software.
The higher the inundation depth reach, the higher the risk is. These maps of hazard assessment over the study area for the representative scenarios are shown in Figure 14 and the inundated areas for each scenario are shown in Figure 15.

Vulnerability assessment
The vulnerability assessment can be used to identify different vulnerable regions in the study area. Making an exposure map 465 is an important step before the vulnerability assessment. The Land Cover dataset obtained from Huizhou Land and Resources Bureau in 2016 can be categorized into 10 different land cover types and the exposure map over the study area is illustrated in Figure 16. In addition, the land areas and percentages of different land cover types are presented in Figure 17.  Figure 16 shows the distribution of different land cover types within the study area, which provides an overall view of the region where the forest is concentrated, the zone of settlements, or the location of urban infrastructure.
475 Figure 17. The land areas and proportions of different land cover types in the study area.
As seen in Figure 17, the forest land occupies most of the land surface of the study area (55.9%, approximately 633.17 km 2 ). The second-largest land cover type is settlements land, which occupies approximately 10.51% (119.1 km 2 ) of the study area's surface. The agricultural land, the garden plot land, and the water land have a large surface area, while mining storage 480 land, the urban infrastructure land, the pasture land, the transportation land, and other lands have a low surface area.
According to the relation between the exposure of land cover types and their corresponding vulnerability values described in Table 6, the four vulnerability levels with each covered zone to storm surge in the study area can be determined, as displayed in Figure 18. Moreover, the land area and percentage of each vulnerability zone to the study area are summarized in Table 8.   Table 8 shows that the total area is 1132.8 km 2 . The zone marked by low vulnerability level (IV) covers an area of 978.21 km 2 . The zone is mainly present in the forest, agriculture, garden plot, and water, and accounts for the greatest proportion of the study area. The zone assigned as the highest vulnerability level (I) is covered with settlements and its geographical area is 119.1 km 2 . The vulnerability level in the land area, which is covered by mining storage and transportation, is high level (II) 495 and it makes up 2.38% of the study area. The moderate vulnerability (III) zone is mainly under the urban infrastructure class with a total area of 8.42 km 2 . Therefore, these zones would most likely suffer significant losses from storm surge.
Furthermore, the Huizhou port (C) is located on the coast, playing a critical role in global trading, which could leave residents and assets with greater exposures to the storm surge than these located on the further inland. Accordingly, the vulnerability level in the Huizhou port (C) is high (level II).
Moreover, some urban infrastructures, transportations, and mining storages that are situated along the coastline of 510 Huizhou are sensitive to the storm surge, and the vulnerability level for these coastal zones is moderate (III) or high (II).
In addition, the most common land cover types over the study area are forest land, agriculture land, the garden plot land, and the water land. These land cover types are hardly affected by natural disasters. Thus, most regions in the study area colored by blue are under the low vulnerability level (IV).

Risk assessment 515
With the risk matrix approach (Table 7), the risk map in the study area can be made by the combination of the hazard map and the vulnerability map. The risk region was categorized into four dangerous zones represented by different colors, as shown in Figure 19. The statistics of the areas of different risk level zones for each of the design scenarios are summarized in Figure 20.

530
The statistics data in Figure 20 indicates that the total area at risk level and the land area labeled with a very high-level decrease with the increasing minimum central pressure of typhoon. The regions under the moderate risk level take the largest portion of the total area at risk level for each of the typhoon scenarios. For example, approximately 66.39 km 2 area is exposed to the moderate risk as the minimum central pressure is the 880 hPa, and an area of about 22.57 km 2 would be at moderate risk when the minimum central pressure is the 940 hPa. 535 Figure 19 shows that the high concentrations of petroleum industries and the high density of the population in the Daya Bay Petrochemical Zone (B) make its vulnerability very high. When the minimum central pressure is 880 hPa, the Daya Bay Petrochemical Zone (B) is in deep inundation status (level I). Thus, the risk level in this zone is very high (level I). As the minimum central pressure is 910 hPa or 920 hPa, the risk levels over a wide area of the Daya Bay Petrochemical Zone (B) are high and moderate. However, the Daya Bay Petrochemical Zone (B) is largely free from risk due to the protection provided 540 by the storm surge barriers along the coastline when the minimum central pressure is 930 hPa or 940 hPa. Moreover, the Huizhou Port (C) and the southeast of the Daya Bay Development Zone (D) are classified as very high or high-risk regions because the regions occupied by transportation land or human settlements (Figure 18) are combined with high or very high hazard levels to storm surge ( Figure 14). In addition, the regions featuring very high or high-risk levels are the area to the north of the White Sand Shore (A), the Renshan Town (G), Pinghai Town (H), and the Huangbu Town (K). 545 Without the protection from the barrier system, these regions mainly occupied by humans are under the moderate risk level even as the minimum central pressure is 940 hPa. Although the vulnerability level is high in the Gilong Town(J) and the Tieyong Town (N), there is no sign of the risk of storm surge due to their locations far from the coastline when the minimum central pressure is 910 hPa, 920 hPa, 930 hPa, or 940 hPa.
The total area of hazard and risk of storm surge under different typhoon intensities are shown in Figure 21. It can be 550 observed from Figure 21, that although the 83.4% (227 km 2 ) of the total inundated area is a very high hazard zone to storm surge as the minimum central pressure is the 880 hPa, only 8% (21.97 km 2 ) of the total inundated area is belonged to the very high risk zone to storm surge and 25% (66.39 km 2 ) of the entire inundated area falls on moderate risk category due to their moderate or high vulnerability level. For other typhoon scenarios, many inundated areas are at a high hazard level for storm surge but most of them change to moderate risk zones, which indicates that the area at a high hazard level cannot represent 555 that area at a high-risk level.

Conclusions 560
In this paper, the application of the coupled model (ADCIRC-SWAN) and the Jelesnianski method for semi-quantifying potential risk assessment of storm surge under different typhoon intensity scenarios in the coastal area of Huizhou. The typhoon intensity scenarios were designed to the minimum central pressure of 880 hPa, 910 hPa, 920 hPa, 930 hPa, and 940 hPa (corresponding to 1000-year, 100-year, 50-year, 20-year, and 10-year return period). The coastal dikes and levees, which are supposed not damaged during the modelling period, were included in the ADCIRC+SWAN model and Jelesnianski method to 565 simulate the storm surge. The possible inundation extents and depths of storm surge under five different typhoon intensities were computed and the risk assessments were performed based on coastal storm surge maps using the ArcGIS 10.5 software.
The results indicate that the whole Daya Bay Petrochemical Zone and most of the coastal area of Huizhou are not at risk to the storm surge generated by low recurrence interval typhoon (20-year, and 10-year return period) due to the protection provided by coastal dikes and levees. The maximum inundation extents and depths increase with increasing return periods. 570 Significant losses and damages might occur in some parts of the Daya Bay Petrochemical Zone and many coastal communities for the return periods of 50-year and 100-year scenarios. Moreover, the regions extending from 4 km to 6 km offshore, particularly in the Daya Bay Petrochemical Zone, are under high or very high-risk level to a 1000-year return period typhooninduced storm surge.
The study provides a comprehensive assessment and zonation of hazard, vulnerability, and risk of storm surge to reduce 575 disaster losses, which caused by designed typhoon scenarios (1000-year, 100-year, 50-year, 20-year, and 10-year return period) in the coastal area of Huizhou. The risk maps and escape route maps, which can be downloaded from the dataset (si, 2020), have been used in practice in Huizhou city, China. These maps can help decision-makers in Huizhou recognize the densely populated communities under risk levels and allow them to develop evacuation strategies to minimize civilian casualties.
Moreover, the study analyses the storm surge risk especially for the Daya Bay Petrochemical Zone, which is occupied by the 580 high concentrations of petroleum industries. This risk analysis provides a better understanding of the risk regions in the Daya Bay Petrochemical Zone, which can both reduce economic losses and prevent environmental damage caused by the massive chemical pollutants and oil spills from coastal petroleum industries that are affected by storm surge. Finally, the proposed methodology and procedure can be applied to any coastal cities in China for conducting risk assessments of storm surge.
In further research, the risk assessment should be undertaken in the following aspects: 585 (1). The evaluation method based on the different land cover classes is simple. The stage-damage function is regarded as one of the most effective solutions to storm surge damage assessment. Therefore, the vulnerability curve rather than the land cover types should be utilized to conduct quantitatively vulnerability assessment in the study area.
(2). Because of increasing typhoon intensity and rising sea levels caused by climate change in the future, the increased storm surge will be taken into consideration when assessing future risk and making hazard mitigation plans in the study area. 590 (3). When the maximum inundated depths and extents are calculated under different intensity typhoons, the levee breach along the coastline will be included in the modelling process to improve simulation precision.