Journal of Oceanology and Limnology   2019, Vol. 37 issue(6): 1929-1940     PDF       
http://dx.doi.org/10.1007/s00343-019-8240-8
Institute of Oceanology, Chinese Academy of Sciences
0

Article Information

LI Tao, WANG Fangdong, HOU Jingming, CHE Zhumei, DONG Jianxi
Validation of an operational forecasting system of sea dike risk in the southern Zhejiang Province, South China
Journal of Oceanology and Limnology, 37(6): 1929-1940
http://dx.doi.org/10.1007/s00343-019-8240-8

Article History

Received Jan. 16, 2019
accepted in principle Mar. 26, 2019
accepted for publication Apr. 30, 2019
Validation of an operational forecasting system of sea dike risk in the southern Zhejiang Province, South China
LI Tao1,2, WANG Fangdong3, HOU Jingming1,2, CHE Zhumei4, DONG Jianxi1,2     
1 National Marine Environmental Forecasting Center, Beijing 100081, China;
2 Key Laboratory of Research on Marine Hazards Forecasting of State Oceanic Administration, Beijing 100081, China;
3 Beijing Space Flight Control Center, Beijing 100094, China;
4 Marine Monitoring and Forecasting Center of Zhejiang Province, Hangzhou 310007, China
Abstract: In this study, an operational forecasting system of sea dike risk in the southern Zhejiang Province, South China was developed based on a coupled storm-surge and wave model. This forecasting system is important because of the high cost of storm-surge damage and the need for rapid emergency planning. A comparison with astronomical tides in 2016 and the validation of storm surges and high water marks of 20 typhoons verified that the forecast system has a good simulation ability. The system can forecast relatively realistic water levels and wave heights as shown under the parametric atmospheric forces simulated in a case study; the sea dikes in credible high risk were mainly located in the estuaries, rivers, and around the islands in the southern Zhejiang. Therefore, the forecast system is applicable in the southern Zhejiang with a support to the effective prevention from typhoon storm-surge damage.
Keywords: storm surge    sea dike    operational forecast    southern Zhejiang Province    risk calculation    
1 INTRODUCTION

The major marine threat to the low-lying coastal areas of China is storm surge, which accompanies typhoons or extratropical cyclones and can cause flooding, destruction of buildings and loss of life (Ministry of Natural Resources of the People's Republic of China, 2018). This risk has seriously affected the development of Zhejiang Province, which is considered one of the most economically vital provinces. Statistics have shown that the average direct economic loss over the last almost 10 years (2007–2016) was about 1.23 billion yuan, accounting for 10.58% of the total national marine disaster loss; nearly 96.37% of this could be attributed to storm surge damage (Ministry of Natural Resources of the People's Republic of China, 2018). To reduce losses in recent years, the local government has strengthened disaster prediction and emergency management before the typhoon season. The ongoing policy is to supply more accurate predictions and specific responses, such as constraining the period for the most serious period and broadening evacuation scope. As storm-surge damage is on the top list of marine disaster, accurate forecasting of storm surges is of great importance in protecting local constructions and lives from the damage or loss.

A traditional storm surge forecast is to collect information of the affected region, maximum storm surge, and warning level, and then will be sent to the authorities, local governments, major marine companies, and other agencies concerned. The receivers then will respond timely based on the information received and manage the emergency plans. A possible problem in this process is that the forecast is technical and the receivers might not make a full use of it because of a lack of technical knowledge. If more direct operating proposals can be synchronously supplied with the forecast, those who receive the information, especially the decision makers, can make more reasonable and proper responses to the storm surge risk. Thus, it is imperative to increase the content of storm surge forecasts.

To supply more direct operating proposals, such as inundation risks and evacuation plans, one shall compare and select a storm surge process where the forecasted typhoon has similar parameters to precomputed results based on climatologically generated synthetic typhoons. As mentioned by Zhang et al. (2013), the advantage of the method is that it is quick and convenient; however, computed typhoon parameters cannot precisely represent real storms, which might lead to incorrect inundation risks and inappropriate evacuation plans.

If a hydrodynamic model can satisfy, in real time, demands of computation time, the results will be more useful to receivers as it takes into account real typhoon parameters. Furthermore, circulation associated with storm surge should be coupled with waves in coastal applications as the short-wave effects produce an approximate 10%–15% increase in the peak water level (Funakoshi et al., 2008). In this study, we used the advanced circulation model (ADCIRC; Luettich et al., 1992) coupled with simulating waves nearshore model (SWAN; Booij et al., 1999) to produce a sea dike risk forecast system for southern Zhejiang Province to provide rapid, direct, and accurate operating proposals of sea dike risk to receivers along with the traditional storm surge forecast.

2 MODEL AND DATA ANALYSIS 2.1 Model description

Tide and storm surge computations were calculated using ADCIRC, which is a continuous Galerkin, finite element, a shallow-water model that accounts for water levels and currents at a range of scales (Dawson et al., 2006; Westerink et al., 2008). ADCIRC utilizes the generalized wave continuity equation to avoid spurious oscillations. Advancement of the solution in time can be computed using either a semi-implicit or explicit time-stepping algorithm (Tanaka et al., 2011; Dietrich et al., 2012). The twodimensional, depth-integrated implementation version (ADCIRC-2DDI) was selected to perform the computations in this study.

SWAN is a third generation and phase-averaged spectral wave model that has been extensively used to simulate waves in shallow water (Booij et al., 1999). The wave action balance equation was converted to solve a finite difference solution on an unstructured mesh (Zijlema, 2010). An implicit backward difference scheme was used for the solution advancement in time while a sweeping Gauss-Seidel method was employed to solve the subsequent system of equations, which permitted for highly efficient solutions at high spatial resolution in the nearshore area.

A tight coupling between ADCIRC and SWAN was implemented; both models share the identical unstructured mesh and parallel computing infrastructure and run sequentially in time (Dietrich et al., 2011). This integrated SWAN+ADCIRC system was adopted in our study as the computation efficiency is improved by having the same computational core running and local memory or cache communication, which is crucial for operational applications.

2.2 Data

To construct the model system, three kinds of fundamental data were used: water depths, coastlines, and sea dikes. The water depth near the coast of Zhejiang Province was measured by the Zhejiang Administration of Surveying Mapping and Geoinformation (ZASMG) from 2012 to 2016 (Fig. 1). The outer edge of the measured database was along the 26 m isobath, and the resolution was 1:10 000 for most of the Zhejiang coast and 1:2 000 for the regions around the Zhoushan Islands. The water depths for the northwest Pacific Ocean and offshore the remaining Chinese coast were obtained from the database of the National Marine Environmental Forecasting Center (NMEFC), which merged the Earth-topography-two-minute (ETOPO2) dataset from the National Geophysical Data Center (Smith and Sandwell, 1997) with the measured water depths obtained by other survey departments since 2010.

Fig.1 The simulation domain (in km) outlined in green and measured original offshore water depth (m) offshore Zhejiang Province

Similar to the water depth, the coastline also consists of two parts; information for Zhejiang Province came from ZASMG updated for 2016 at a resolution of 1:10 000 and the rest of East Asia were based on the 1997–2001 coastline data of the National Administration of Surveying, Mapping and Geoinformation. As there are more frequent changes to coastal regions due to land reclamation and coastal erosion, the relatively recent data we used ensured that the model grids captured the features of the current geography. Additionally, the high resolution near the coast of Zhejiang Province met the refinement requirements of the grids.

To prevent typhoon damage, plenty of sea dikes were constructed; as counted by ZASMG, the total number was about 1 352 at the end of 2013. The statistics included the name, location, type, top elevation, and parapet elevation of each sea dike. The measured factors fully covered the requirements to consider the influence of sea dikes in the forecasting system. In view of the need to calculate the influence of tides, tidal forcing along the open boundary also needed to be taken into account. This information came from the global ocean tide model NAO99 (Matsumoto et al., 2000) that contains about 16 major constituents with a spatial resolution of 0.5°.

2.3 Model setup

The coast along southern Zhejiang Province supports two prefecture-level cities, Taizhou and Wenzhou (TW); thus, the simulation domain takes this coastal area as the primary focus. Additionally, the typhoons that influence southern Zhejiang Province and the southeast coast of China are mainly generated in the northwestern Pacific Ocean, particularly around the eastern Philippines, and subsequently move northwest and then land or swerve along the coast of China (Tu et al., 2009). As shown in Fig. 1, the simulation domain was designed like a fan and mainly included the northwestern Pacific area, covering most of the East China Sea, Taiwan Strait, and Bashi Strait; the open boundary furthest to the east is along 129°E, which satisfies the above two criteria.

There are many islands and estuaries in the TW region (Fig. 2a) and the complex geography indicated that an orthogonal grid would not be a good choice. To accurately characterize the terrain and enhance the computation efficiency, the triangular mesh was used to refine the localized resolution of topographic gradients. The total computational grid contained 213 259 nodes and 395 306 elements, and the grid along the coast of TW was the most refined region, down to about 100 m (Fig. 2b). Taking the TW as the center, the grid resolution was gradually reduced southward and northward along the coastline. For example, the resolution near the middle of the southern Fujian Province coast was about 1 km, but the southernmost coastline of the simulation domain (east coast of Guangdong Province) was about 2 km. Similarly, as the distance to the coastline increased, the grid also became coarser. For the eastern boundary of the simulation domain, the resolution reached 25 km.

Fig.2 The shoreline, water depths and gauge stations (red dots) in Wenzhou and Taizhou with Yueqing Bay outlined in red (a), and unstructured grid in Yueqing Bay showing the sea dikes (b) (red lines with significant ones numbered)

Along the open boundary of the domain, the amplitude and the phase lag of eight dominant astronomical tidal constituents were interpolated from NAO99 for tidal forcing, including the diurnal K1, O1, P1 and Q1 constituents and the semidiurnal M2, S2, N2, and K2 constituents. In order for the initial transients to physically dissipate and dynamically correct tides to be generated (Bunya et al., 2010), each simulation required a spin-up period; 1 day before the starting time of the simulation was added to deal with this issue. The nodal factor and equilibrium argument for forcing tide constituents in ADCIRC were also changed based on the starting time of the spin-up.

The forecasting system was started from a standstill and the time step of ADCIRC was set to three seconds due to its semi-explicit formulation. SWAN had a much longer time step (1 h), which is a result of the sweeping method used by SWAN; the SWAN communicated every hour with ADCIRC.

2.4 Sea dike treatment

As the inundation process was not considered in the forecasting system, the onshore sea dikes were chosen based on some criteria. The candidate was located in the TW region, and the relatively important protective effect of one, such as sea dike 1 (Fig. 2b), was selected if two or more sea dikes were along the same coastline. Subsequently, the sea dikes were checked out if they were not close to, intersected or coincided with the shoreline, such as sea dikes 2 and 3 (Fig. 2b). The sea dikes were named separately if they were too long or had segments that were too far apart. In total, there were 337 sea dikes selected for the forecasting system; the location of each sea dike was also selected and its height (top elevation plus parapet elevation) above mean sea level was confirmed. This information was used to determine the sea dike risk.

Here, we used a simple difference between the dike height and the forecasted water level to determine the sea dike risk (State Oceanic Administration, 2016). As shown in Table 1, the sea dike overtopping was divided into four levels, each level corresponds to the risk of a sea dike overtopping during the storm surge. If a sea dike is categorized as high, this indicates that a severe storm surge will pose a serious threat to sea dike safety and the probability of sea dike overtopping is high. Local inhabitants protected by this sea dike should be evacuated and security inspections should be strengthened to prevent the occurrence of the sea dike breaking. In contrast, a sea dike in the low category would only be slightly affected by the storm surge.

Table 1 The standard divisions of storm surge overtopping
3 MODEL PERFORMANCE 3.1 Run times

The domain constraint of a real-time forecast system is the computation time, especially when a typhoon is approaching and the storm surge warning needs to be issued. Thus, it is important to produce the forecast in a short time to support subsequent evacuation planning. Here, we used the Sunway 3000A (Zheng and Yu, 2009) of NMEFC to test the effect of computation time. The computational time penalty decreased as the number of cores increased (Fig. 3); two cores utilized more than 5 000 min for a 4-day simulation while the penalty was reduced to less than 20 min with 256 cores. Although the forecast system could not be fully tested because of a limitation with the Sunway 3000A, the tests we did perform showed that the forecast system can fully meet the operational demand. The necessary forecast information can be produced in about 20 min with 128 cores, which is the same as other operational forecast systems used by NMEFC.

Fig.3 Timing results for simulation on Sunway 3000A The time shown is wall-clock minutes for a 4-day period.
3.2 Tidal validation

The TW is located on the coast of the East China Sea and is an area famous for a large tidal range. The average tidal range is about 1.65–5.54 m (Huang and Huang, 2005), so simulations from a hydrodynamic model that can be used to calculate water level is crucial in the TW. A year-long simulation, spanning 2016, was conducted by forcing the forecast system with eight dominant astronomical tidal constituents along the open boundary and the calculated water levels were the simulated astronomical tides. These calculated tides were compared with observed astronomical tides and determined by using harmonic analysis to analyze the observed water levels at gauge stations.

The comparison was conducted through the hourly values at nine gauge stations (Table 2) located in or near the TW (Fig. 2a). The mean absolute error at the gauge stations ranged from 14.30 cm to 36.69 cm with an average from all stations of 25.92 cm. The deviation was relatively low for gauge stations that face the open water (e.g., Dachen, Kanmen), and high deviations were found in the estuary and inland coastal gauge stations (e.g., Rui'an and Aojiang). The distribution of the relative error showed the same pattern because the shallow water constituent was not included in the simulations and tidal signals at the inland stations can be significantly affected by mesh resolution (Kerr et al., 2013). The range of relative error for the high and low astronomical tides between the gauge stations showed obvious differences although the mean values were almost the same. The error of low tides was 13.24% (15.66% from Sagangtou minus 2.42% from Kanmen), which was twice the error of high tides (6.68%, 11.90% from Aojiang minus 5.22% from Haimen). This indicated that the simulations of the high astronomical tides were more balanced than the low tides.

Table 2 Error analysis for the different gauge stations in 2016

In general, extensive typhoon damage is usually caused when the peak of the storm surge coincides with the high astronomical tide and this combination is particularly critical in strong tidal range areas. Overall, the astronomical tide was overpredicted (Fig. 4) at the inland gauge stations, such as Longwan, Rui'an, and Aojiang. However, for the other stations, the forecast system captured the tidal variety in TW and supplied a relatively realistic astronomical tide information.

Fig.4 Comparison of the maximum astronomical tide in each tidal periodicity as calculated by harmonic analysis and simulated by the model for 2016 The green dotted line is the best-fit line.
3.3 Storm surge validation

Twenty typhoons from 1970 to 2016 were used to validate the storm surge and high water mark simulation by the forecast system (Fig. 5). With the exception of one typhoon (9414 "Doug") that moved northward offshore of Zhejiang Province, the remaining 19 typhoons all made landfall on the southeast coast of China. However, all 20 typhoons caused serious damage to the coast, and produced more than 50 cm of storm surge at more than two gauge stations in TW. Typhoon parameters, including location, center pressure, maximum wind speed and radius to maximal winds, were derived every 6 h from the best-track dataset of the China Meteorological Administration (Ying et al., 2014). These data were used as the meteorological forcing and the input wind velocity and atmospheric pressure were internally calculated by the dynamic Holland model (Holland, 1980) of ADCIRC. Wind decreases too rapidly with increasing radius when it is greater than 2 or 3 times the radius to maximum wind (Willoughby and Rahn, 2004); thus, the gauge stations located within 2 times of the radius to maximum wind were chosen for the analysis.

Fig.5 The tracks of the 20 typhoons that used to verify the simulations

Two sets of simulations were conducted, one for astronomical tides and the other for total water level; the storm surge was calculated as the difference between the two. Peak simulated storm surge heights were compared against the peak recorded storm surge height from 64 total records at gauge stations (Fig. 6a). Compared with observations, about 62.5% of the simulated deviation was less than 10%, while only 9.38% was higher than 20%. The slope of the best-fit line (with an ideal value of one) was 0.99 and the correlation coefficient (R2), which describes how well a regression line fits a dataset, was 0.96, showing that the simulated storm surge fits well with the recorded surge. The modeled high water was compared to the recorded high water mark (Fig. 6b). Despite the fact that 73.44% of simulated deviation based on the mean sea level was less than 10%, the amplitude of deviation was much larger than the storm surge. The slope of the best-fit line was 0.91, indicating that the model tends to underpredict the high water and R2 was only 0.82. In general, storm surge and high water levels can be successfully captured by the forecast system, although the deviation of the latter is greater than the former.

Fig.6 Scatter plots of peak storm surge (a) and high water mark (b) (based on the mean sea level) observations versus the simulations Different colors indicate the relative error, and the red dotted line is the best-fit line.
4 FORECAST VALIDATION

Any operational system should be run quickly to provide available results to emergency managers for decision-making as soon as a typhoon advisory is updated, but, importantly, the in-site typhoon forecasts usually demonstrate obvious deviations from observations. Thus, forcing the forecast system with the in-site typhoon forecast can accurately reveal the forecasting ability of the system. Here, a compromised in-site typhoon forecast was applied using the typhoon Fitow that struck in 2013 as an example (Fig. 7). The middle 3 days were used as the forecast interval and the typhoon landed late on the third day. During this time frame, observations of typhoon parameters were taken after 24, 48 and 72 h and then linearly interpolated. Outside of the forecast interval, the typhoon parameters were observations. Based on the typhoon track and parameter observations every 24 h, the forecasts were obviously different from the observations.

Fig.7 The observed (blue line) and the in-site 72 h forecasted (red line) tracks for typhoon Fitow (2013) The green dots represent offshore buoys and the magenta dots are the six-hourly forecasted typhoon positions.

Fitow made landfall at the junction of Zhejiang and Fujian Provinces, costing about 3.5 billion yuan in direct economic losses. The observed maximum storm surge was 3.75 m at Aojiang gauge station and the high water marks at six gauge stations exceeded the local warning water levels (Ministry of Natural Resources of the People's Republic of China, 2018). As simulated by the forecast system (Fig. 8a), Fitow mainly triggered higher water levels in the south of Zhejiang and Fujian Provinces, particularly the estuaries near the landfall point where the maximum water levels exceeded 5 m. Figure 8 shows the comparison between observations and predictions of the peak storm surge at eight gauge stations. Overall, the forecast system accurately captured the peak storm surges with the magnitudes at four gauge stations (Wenzhou, Rui'an, Shipeng, Shacheng) well forecasted, but underpredicted at two stations (Shagangtou and Kanmen), which were distant from the typhoon track and lead to the relatively low modeled atmospheric forcing. An additional gauge station (Dongtou) also revealed an underprediction of peak storm surge possibly because this station is located on the leeward side of the Dongtou Island. In contrast, an overprediction was found at Aojiang station, but the relative error to the observation was only about 12%; thus, the deviation can be accepted with such a forecast outcome.

Fig.8 The map of maximum water levels (in m) (a) and scatter plots of storm surge simulations versus observations (b) The black line in (a) is the track of typhoon Fitow; the dashed line in (b) is the forecasted track.

The highest wave height reached 11 m in the open ocean with the high waves mainly located along the typhoon track (Fig. 9a). Due to the lack of a background wind field, the model cannot simulate well the effects of the swell. This issue was apparent in the interval before the effects of strong winds caused by the typhoon were visible, or far from the typhoon track, such as at buoy QF214 (Fig. 9b). In or near to the radius of the maximum wind, such as at buoy stations QF205 and QF210, the forecasted relatively high significant wave heights matched well with the observations, although the phase of the wave crests still showed some deviations that could be ascribed to the deficiency of the forecast typhoon track. Unlike QF205 and QF210, the magnitude of wave height at QF208 was well simulated, but the phase showed a lag; this buoy is located in the northern part of Taiwan Strait where special topographical features limited the performance of the Holland model (Yuan et al., 2014).

Fig.9 The map of maximum significant wave height (in m) (a) and hydrographs of the four buoy stations during typhoon Fitow (2013) (b) The black line in (a) is the same as Fig. 8a; the black lines in (b) are the computed significant wave heights and the red dots are the measured data.

The relatively high-risk sea dikes are located along the banks of three rivers, the Yueqing Bay coast and some islands in southern Zhejiang Province (Fig. 10). In particular, the risk is highest for the south bank of Aojiang River, indicating this region had the highest probability of overtopping during typhoon Fitow.

Fig.10 The simulated risk of sea dike overtopping during typhoon Fitow
5 CONCLUSION

Because of the rapid and extensive economic development in Chinese coastal zones, large amounts of damage caused by a severe storm surge are possible. To reduce the potential for losses from natural marine disasters, it is extremely urgent to increase the accuracy and usability of marine disaster forecasts, especially for the prevention of typhoon storm surge damage. These forecasts allow local governments to work out specific emergency procedures and make appropriate evacuation plans. To facilitate this, in recent years, NMEFC has used the coupled storm surge and wave model (ADCIRC + SWAN) to provide operational sea dike risk forecasts in some key coastal provinces. This study developed a forecast system in southern Zhejiang Province.

With the goal of increasing accuracy, the numerical forecast system needed to reduce the computational time penalty to shorten the time information is delivered to decision makers because it increases the amount of preparation time available for emergency responses. Therefore, a triangular grid and parallel computing were adopted in the developed sea dike forecasting system to find the coastal terrain and enhance the computational efficiency. The performance test showed that a 4-day simulation was completed in 20 min with 128 cores. The dynamic Holland model was used as the atmospheric force in operational forecasting. In contrast to actual overtopping, the movement of water over and around the sea dike was not considered in this study. A solid coastline was set in the forecast system and the distance to the coast or the priority of the area the sea dike protects determined whether the sea dike was retained in the forecast system. Four levels of sea dike risk were distinguished by comparing the difference between the dike height and forecasted water level.

By analyzing tidal simulations based on 2016 information, we found that the average error of nine gauge stations was 25.92 cm with errors in the estuaries or rivers larger than in the open ocean. Except for a very few high astronomical tides, the forecast system tended to slightly overpredict the astronomical tides. Comparing the simulations with observations for 20 typhoon storm surges, the high water marks and storm surge data showed slopes of the best-fit lines above 0.91, indicating that the forecast system accurately simulated the variations in water level and storm surge within the radius of maximum wind.

A case study was carried out by conducting a 3-day comprehensive in-site forecast for Typhoon Fitow, which struck in 2013. The results showed that the storm surges and water levels were well reproduced and the significant wave heights of buoys in or near the radius to the maximum wind were reasonable, confirming the forecasting ability of the system. Although there are no available overtopping data to verify the sea dike risk, it is likely similar to the simulated sea dike risk where overtopping mainly occurred in the estuaries, rivers and around the islands.

Compared with actual overtopping, this forecast system is relatively simple because it does not simulate physical processes after the overtopping of the sea dikes; subsequent work should include this, as the sea dike risk information will be more credible if these physical processes are included. Moreover, insitu typhoon forecasts had greater uncertainty whereas an ensemble forecast would provide more possibilities; thus, ensemble forecasting models will have more advantages in operational storm surge forecasting and will be implemented in the future. In addition, we found that atmospheric forces are also key factors that affect numerical forecasts of storm surges. If the regional weather forecast model is used to supply the operational atmospheric forces, the accuracy of the forecast may be better.

Finally, operational risks to sea dikes have been promoted in recent years and this forecast system was finished in early 2017. However, there were no obvious typhoon storm surges in 2017 to validate the forecasts despite the overtopping observation equipment being in place. Therefore, the forecast system needs to be further verified in the future.

6 DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from National Marine Environmental Forecasting Center (NMEFC) but restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available. Data are however available from the authors upon reasonable request and with permission of NMEFC.

References
Booij N, Ris R C, Holthuijsen L H. 1999. A third-generation wave model for coastal regions, Part Ⅰ, model description and validation. Journal of Geophysical Research, 104: 7649-7666. DOI:10.1029/98JC02622
Bunya S, Dietrich J C, Westerink J J, Ebersole B A, Smith J M, Atkinson J H, Jensen R, Resio D T, Luettich R A, Dawson C, Cardone V J, Cox A T, Powell M D, Westerink H J, Roberts H J. 2010. A high-resolution coupled riverine flow, tide, wind, wind wave, and storm surge model for Southern Louisiana and Mississippi. Part ñ:model development and validation. Monthly Weather Review, 138(2): 345-377. DOI:10.1175/2009MWR2906.1
Dawson C, Westerink J J, Feyen J C, Pothina D. 2006. Continuous, discontinuous and coupled discontinuouscontinuous galerkin finite element methods for the shallow water equations. International Journal for Numerical Methods in Fluids, 52(1): 63-88. DOI:10.1002/fld.1156
Dietrich J C, Tanaka S, Westerink J J, Dawson C N, Luettich R A Jr, Zijlema M, Holthuijsen L H, Smith J M, Westerink L G, Westerink H J. 2012. Performance of the unstructuredmesh, SWAN+ADCIRC model in computing hurricane waves and surge. Journal of Scientific Computing, 52(2): 468-497. DOI:10.1007/s10915-011-9555-6
Dietrich J C, Zijlema M, Westerink J J, Holthuijsen L H, Dawson C, Luettich R A Jr, Jensen R E, Smith J M, Stelling G S, Stone G W. 2011. Modeling hurricane waves and storm surge using integrally-coupled, scalable computations. Coastal Engineering, 58(1): 45-65.
Funakoshi Y, Hagen S C, Bacopoulos P. 2008. Coupling of hydrodynamic and wave models:case study for Hurricane Floyd (1999) hindcast. Journal of Waterway, Port, Coastal, and Ocean Engineering, 134(6): 321-335. DOI:10.1061/(ASCE)0733-950X(2008)134:6(321)
Holland G J. 1980. An analytic model of the wind and pressure profiles in hurricanes. Monthly Weather Review, 108(8): 1212-1218. DOI:10.1175/1520-0493(1980)108<1212:AAMOTW>2.0.CO;2
Huang Z K, Huang L. 2005. Tidal Theory and Calculation. China Ocean University Press, Qingdao. p.9. (in Chinese)
Kerr P C, Martyr R C, Donahue A S, Hope M E, Westerink J J, Luettich R A Jr, Kennedy A B, Dietrich J C, Dawson C, Westerink H J. 2013. U.S. IOOS coastal and ocean modeling testbed:evaluation of tide, wave, and hurricane surge response sensitivities to mesh resolution and friction in the Gulf of Mexico. Journal of Geophysical Research:Oceans, 118(9): 4633-4661. DOI:10.1002/jgrc.20305
Luettich R A Jr, Westerink J J, Scheffner N W. 1992. ADCIRC: an Advanced Three-Dimensional Circulation Model for Shelves, Coasts, and Estuaries: Report Ⅰ: Theory and Methodology of ADCIRC-2DDI and ADCIRC-3DL.Dredging Research Program Tech. Rep. DRP-92-6, US Army Corps of Engineers, Washington. 137p.
Matsumoto K, Takanezawa T, Ooe M. 2000. Ocean tide models developed by assimilating TOPEX/POSEIDON altimeter data into hydrodynamical model:a global model and a regional model around Japan. Journal of Oceanography, 56(5): 567-581. DOI:10.1023/A:1011157212596
Ministry of Natural Resources of the People's Republic of China. 2018. China marine disaster bulletin. http://www.mnr.gov.cn/sj/sjfw/hy/gbgg/zghyzhgb/. (in Chinese)
Smith W H F, Sandwell D T. 1997. Global sea floor topography from satellite altimetry and ship depth soundings. Science, 227(5334): 1956-1962. DOI:10.1126/science.277.5334.1956
State Oceanic Administration of China. 2008-2017. China marine disaster bulletin. http://www.mnr.gov.cn/sj/sjfw/hy/gbgg/zghyzhgb/. (in Chinese)
State Oceanic Administration of China. 2016. HY/T 195-2015 Technical Guide for Forecast of Storm Surge Overtopping of Dike. China Standards Press, Beijing. p.4. (in Chinese)
Tanaka S, Bunya S, Westerink J J, Dawson C, Luettich R A Jr. 2011. Scalability of an unstructured grid continuous galerkin based hurricane storm surge model. Journal of Scientific Computing, 46(3): 329-358. DOI:10.1007/s10915-010-9402-1
Tu J Y, Chou C, Chu P S. 2009. The abrupt shift of typhoon activity in the vicinity of Taiwan and its association with western North Pacific-East Asian climate change. Journal of Climate, 22(13): 3617-3628. DOI:10.1175/2009JCLI2411.1
Westerink J J, Luettich R A, Feyen J C, Atkinson J H, Dawson C, Roberts H J, Powell M D, Dunion J P, Kubatko E J, Pourtaheri H. 2008. A basin-to channel-scale unstructured grid hurricane storm surge model applied to southern Louisiana. Monthly Weather Review, 136(3): 833-864. DOI:10.1175/2007MWR1946.1
Willoughby H E, Rahn M E. 2004. Parametric representation of the primary hurricane vortex. Part Ⅰ:observations and evaluation of the Holland (1980) model. Monthly Weather Review, 132(12): 3033-3048. DOI:10.1175/MWR2831.1
Ying M, Zhang W, Yu H, Lu X Q, Feng J X, Fan Y X, Zhu Y T, Chen D Q. 2014. An overview of the China meteorological administration tropical cyclone database. Journal of Atmospheric and Oceanic Technology, 31(2): 287-301. DOI:10.1175/JTECH-D-12-00119.1
Yuan K R, Shang S P, Xie Y S, Zhang L, Zhang Y D, Zhang F. 2014. The simulation of typhoon waves in Taiwan Strait. Journal of Xiamen University (Natural Science), 53(3): 413-417. (in Chinese with English abstract)
Zhang K Q, Li Y P, Liu H Q, Rhome J, Forbes C. 2013. Transition of the coastal and estuarine storm tide model to an operational storm surge forecast model:a case study of the Florida coast. Weather and Forecasting, 28(4): 1019-1037. DOI:10.1175/WAF-D-12-00076.1
Zheng X, Yu T. 2009. High availability massive storage system of Sunway 3000A. Computer Engineering & Science, 31(S1): 40-41,45. (in Chinese with English abstract)
Zijlema M. 2010. Computation of wind-wave spectra in coastal waters with SWAN on unstructured grids. Coastal Engineering, 57(3): 267-277.