Storm surges along the Tottori coasts following a typhoon

Sooyoul Kim a,n, Yoshiharu Matsumi a, Tomohiro Yasuda b, Hajime Mase b

a Graduate School of Engineering, Tottori University, Koyama-cho Minami, Tottori 680-850, Japanb Disaster Prevention Research Institute, Kyoto University, Gokasho, Kyoto 611-0011, Japan

a r t i c l e i n f o

Article history:Received 25 November 2013Accepted 7 September 2014

Keywords:Storm surgeEkman setupAsymmetric and symmetric typhoon fieldParametric wind and pressure modelWeather Research and Forecasting (Wrf)model

a b s t r a c t

In the present study, the after-runner surge that maximum surge height appears 15–18 h later along theTottori coasts facing the Sea of Japan/East Sea (SJES) after typhoons undergo a change in shape andintensity as extratropical cyclones is investigated using asymmetric and symmetric wind and pressurefields of Typhoon Songda (2004). For the asymmetric wind and pressure field, the Weather Research andForecasting (WRF) model is used, while for the symmetric wind and pressure field, a parametric windand pressure model is used. The results indicate that both models simulate fairly well the 10 m levelwind and the sea level pressure along the Pacific Ocean, while the WRF model shows better agreementwith the observations over the SJES. Subsequently, from storm surge simulations for Typhoon Songda, itis found that using the deformed and asymmetric meteorological field of typhoon structures agrees wellwith observations. The study shows that the after-runner surge's characteristic comes from the Ekmansetup in the presence of the Coriolis force over the Tottori coasts. It is critical that its behavior should betaken into account for the safety design of coastal defense structures around the Tottori coastal region.

& 2014 Elsevier Ltd. All rights reserved.

1. Introduction

A storm surge produces a rise in sea level. In Japan, it is inducedby tropical cyclones (hereafter, the word ‘typhoon’ is used). It isimportant to understand storm surge behavior in order to reduceand prevent coastal disasters by employing countermeasuresagainst inundation of coastal areas. In general, sea level rise alongthe coasts of Japan appears several hours before a typhoon arrives,and the maximum surge coincides with the time of typhoonlandfall. However, the time history of the surge elevation isdifferent depending on the various regions in Japan; while themaximum surge appears at the time of landfall on coasts along thePacific Ocean, it appears 15–18 h later along the Tottori coastsfacing the Sea of Japan/East Sea (SJES) after veering northwards.Here we define the storm surge, occurring after a typhoon haspassed, as the after-runner surge. For example, Typhoon Songda'smaximum surge along the Tottori coasts was recorded 15 h afterthe typhoon passed in September 2004.

Kennedy et al. (2011) introduced that a large water level increase(called the forerunner surge) appeared along a substantial section ofthe western Louisiana and northern Texas coasts 12–24 h in advanceof the landfall of Hurricane Ike (2008). They diagnosed the fore-runner surge as being generated by Ekman setup on the wide andshallow shelf. In contrast to the forerunner surge, in this study, we

will show a sea surface level rise that appeared along the Tottoricoast approximately 15 h in retreat of the landfall of Typhoon Songda(2004). We will discuss that the after-runner surge comes from theEkman setup and the predominant factor is the Coriolis force overthe Tottori coast. Understanding of the after-runner surge is impor-tant for the safety design of coastal defense structures and the riskmanagement, however no relevant study about after-runner surge inthe SJES region has been previously reported.

The SJES is a sea bounded by the Korean Peninsula and theJapanese islands of Hokkaido, Honshu and Kyushu. It is linked toother seas through the Tartary Strait, La Perouse Strait and theTsushima/Korea Strait. The surrounding mountainous terrainsdeform the core structure and intensity of typhoons (e.g., Benderet al., 1987). According to Kitabatake (2008), typhoons oftenundergo extratropical transition when they move into the mid-latitudes because of energy budgets (Palmen, 1958; DiMego andBosart, 1982) and structural changes induced by a baroclinicenvironment (e.g., Klein et al., 2000). In particular, Kitabatake(2008) showed that the typhoon intensity is directly downgradedto an extratropical cyclone (maximum wind speedo8.7 m/s) orsometimes reduces to a tropical cyclone (8.7omaximum windspeedo17.5 m/s) above 401N in the SJES and North Pacific Ocean.

Since parametric wind and pressure models for typhoon areintroduced by Fujita (1952) in Japan, many advances have occurred(e.g., Fujii and Mitsuta (1986); Veltcheva and Kawai, 2002).Recently, improved and new formulations were provided forpredicting extreme waves generated by hurricanes along Gulf ofMexico (e.g., Jeong et al., 2012; Panchang et al., 2013). However,

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/oceaneng

Ocean Engineering

http://dx.doi.org/10.1016/j.oceaneng.2014.09.0050029-8018/& 2014 Elsevier Ltd. All rights reserved.

n Corresponding author.E-mail address: sooyoul.kim@sse.tottori-u.ac.jp (S. Kim).

Ocean Engineering 91 (2014) 133–145

parametric models are difficult to reproduce downgraded extra-tropical cyclones above 401N in the SJES despite of their improve-ment and advance. Recently, meteorological circulation models fortyphoon wind and pressure fields have been used to reproduce thedeformation of typhoon structure and to predict extreme wavesand surges (e.g., Lee et al., 2010; Mase et al., 2008).

We investigate the effects of the deformation of the typhoonstructure on after-runner surges in the present study in order tounderstand the after-runner surge. The event studied is TyphoonSongda in 2004. The Weather Research and Forecasting (WRF)model developed by the National Center for Atmospheric Research(NCAR) and Schloemer's (1954) parametric pressure formula,combined with a wind model of Fujii and Mitsuta (1986), are usedin order to reproduce the pressure and wind fields. The WRFmodel is chosen to reflect the physical change in the typhoonstructure and intensity. On the other hand, the parametric modelis selected with no shape factors employed to determine thedeformation of the typhoon structure. Subsequently, the after-runner surge due to Typhoon Songda is simulated by inputting themeteorological data sets into a coupled model of Surge, Wave andTide (SuWAT) developed by Kim et al. (2008, 2010) to address theorigin of after-runner surges.

The models, together with the experimental conditions used inthe study, are explained briefly in Section 2. The experimentalresults and a discussion are provided in Section 3, with the mainfindings summarized and conclusions drawn in Section 4.

2. Model descriptions and simulation conditions

2.1. Model descriptions

2.1.1. Weather Research and Forecasting modelWe used the Weather Research and Forecasting model (WRF)

to reproduce the Typhoon songda's meteolorogical field in thepresence of the deformation in shape and the change in intensity.The WRF model, version 3.0, developed by NCAR (Skamarock et al.,2008) is a three dimensional, fully compressible, non-hydrostaticmodel formulated within a terrain-following mass coordinate inthe vertical direction. Based on the previous studies over the SJES(Kim et al., 2012, 2013), the model physics in the WRF are chosen:the microphysics of Thompson et al. (2004), the land surface ofChen and Dudhia (2001), the planetary boundary layer of Honget al. (2004), the shortwave radiation of Chou and Suarez (1994),and the cumulus parameterization of Kain (2004).

The three domains from the outermost, 2nd and 3rd, which arethe two-way nesting domains shown in Fig. 1(a), are used for WRFsimulations. The outermost domain (D01) with a grid size of 12 kmcovers the overall typhoon tracks. The intermediate domains (D02and D03) are downscaled to grid sizes of 4 km and 1.3 km,focusing on the Tottori coasts to be suitable for storm surgesimulations. In all the domains, the 40 vertical layers are used;the lowest vertical height near the surface is approximately within10 m, the top pressure is 1000 hPa. The initial and boundaryconditions for the WRF model are obtained from the NationalCenters for Environmental Prediction (NCEP): we use the NCEPFNL (Final) Operational Global Analysis data of 11�11 in space and6 h interval in time (referred to hereafter as FNL data). A bogusdata assimilation scheme to generate the initial typhoon structureis not employed in this study.

2.1.2. Parametric wind and pressure modelA parametric wind and pressure model embedded in the

SuWAT model is based on the formulas of Schloemer (1954) forthe pressure and of Fujii and Mitsuta (1986) for the wind. The Fujiiand Mitsuta (FM) model has been used widely when simulating

storm surges in coastal engineering (e.g., Kim et al., 2010).In recent years, parametric wind and pressure models have beenimproved to reproduce the asymmetric structure of typhoons (e.g.,Veltcheva and Kawai, 2002). However, those models, studied overcoasts of the Pacific Ocean, are considered only for the shape oftyphoons. In addition, the change of intensity of extra-tropicalcyclones has not been studied in the SJES. Therefore, the presentstudy employed the original FM model to reproduce the TyphoonSongda's meteorological field without shape deformation.

The FM model is briefly described as follows: the ratio G(x) of asurface wind to a pressure gradient wind is a function of x¼ r=r0,where r is the radius from the typhoon center to a particularlocation and r0 denotes the radius to the location of maximumwind speed:

GðxÞ ¼ Gð1Þþ½GðxpÞ�Gð1Þ�ðx=xpÞk�1exp ð1�1=kÞ½1�ðx=xpÞk�1�D E

ð1Þin which k¼2.5, xp ¼ 0:5, GðxpÞ ¼ 1:2 and Gð1Þ ¼ 0:6667 are usedoriginally in Fujii and Mitsuta (1986). The radius r0 is determinedby the least square method for the best fitting radius based on theobserved sea level pressure at 148 stations, which are operated byJapan meteorological Agency. The pressure gradient wind isreduced by a factor Gð1Þ of 0.6667 to take into account thetopographical friction in this study. The surface wind is obtainedby multiplying the pressure gradient wind speed by the ratio G(x)and setting the inflow angle between the wind vector and isobarto 301. Finally, the wind at 10 m height is acquired as the vectorsum of the surface wind and the typhoon's speed of movement.The deformation of the typhoon structure is not taken intoaccount in this model.

2.1.3. Coupled model of surge, wave and tideThe SuWAT model, developed by Kim et al. (2008, 2010), is

used to simulate the after-runner surge caused by Typhoon

Fig. 1. Typhoon Songda's best track and the computational domains: (a) theoutermost domain D01 and intermediate domains D02 and D03; (b) positions oftide and wave gauges in the intermediate domains D04 and D05, and innermostdomain D06.

S. Kim et al. / Ocean Engineering 91 (2014) 133–145134

Songda. The SuWAT model consists of the depth-integrated non-linear shallow water equations for the surge module and a third-generation wave model (SWAN version 40.41) for the wavemodule. The staggered Arakawa C grid in space and the leap-frog method in time are used in the surge module. The wavemodule and surge module exchange information at a regularinterval about the sea surface level, current, radiation stress andsea surface drag coefficient. The SuWAT model is able to contain anarbitrary number of domains by way of an appropriate messagepassing interface with the nesting scheme through which the seasurface level and current on the boundary are interpolated fromthe coarse grid to the fine grid with a 1/3 ratio. The detailedparallel strategy was described in Kim et al. (2010).

2.2. Simulation conditions

The SuWAT model is composed of six domains coveringthe whole Yellow Sea to the west, the Philippine Sea to the south,the SJES to the north and the north Pacific Ocean to the east in theoutermost domain in the study, downscaling the grid sizes from12 km to approximately 50 m, as shown in Fig. 1. A tide gauge isinstalled in the middle of the waterway linking to the NakaumiSea, and a wave gauge is located to the south-east of the tidegauge. The water depths in the outermost (D01) and intermediate(D02 and D03) domains were derived from the General Bathy-metric Chart of the Oceans (GEBCO, 2003). For the rest of the

domains, the water depths were derived from the Digital Bathy-metric Chart and the Grid Bathymetric Data published by theJapan Hydrographic Association.

Three numerical experiments for the wind and pressure fieldsinduced by Typhoon Songda were carried out using the followingmodel configurations in the WRF and FM models:

1) a WRF run with a nudging technique for all computationaldomains (referred to as WRF-NA),

2) a WRF run with a nudging technique for only the outermostdomain (denoted WRF-NF),

3) an FM run (denoted FM).

By using the estimated wind and pressure data sets from theabove three runs, six simulations were performed of TyphoonSongda's after-runner surge by adopting two types of sea surfacedrag coefficients:

1) a wave-dependent drag coefficient (Janssen, 1989, 1991),2) the sea surface drag coefficient of Honda and Mitsuyasu (1980).

We ignore tidal variation in the simulations because the tidalrange is quite small so that annual average is approximately 0.3 mat the Sakai Minato port in the SJES (Kim et al., 2008); then, theeffects of tidal variations on storm surges were neglected.

125°E 130°E 135°E 140°E 145°E20°N

25°N

30°N

35°N

40°N

09/04/0300 JST

09/05/0300 JST

09/05/1200 JST

09/06/0000 JST

09/06/1200 JST

09/07/0000 JST

09/07/0600 JST

09/07/0900 JST

09/07/1500 JST

09/07/1800 JST

09/07/2100 JST

09/08/0000 JST

09/08/0300 JST

09/08/0600 JST

1.LTF2038

2.LTF2039

3.LTF20424.LTF2049

5.LTF2099

6.LTF2098

7.LTF20978.LTF2096

9.LTF2030

10.LTF209311.LTF2022

12.LTF2019

13.LTF2087 14.LTF2082

15.LTF2054

16.LTF205317.LTF2052

18.LTF2020 19.LTF2091

20.LTF2051 21.LTF2041

22.LTF203223.LTF205524.LTF2056

25.LTF204626.LTF2092

27.LTF203128.LTF2028

29.LTF207230.LTF2071

31.LTF202932.LTF2068

33.LTF2063 34.LTF1021

35.LTF1008

Fig. 2. Typhoon Songda's best track and meteorological stations of which data are used in the study to assess the computed wind and pressures.

S. Kim et al. / Ocean Engineering 91 (2014) 133–145 135

3. Simulation results and discussion

3.1. Comparison of the simulated tracks and best track

Typhoon Songda initially developed as a tropical depressionnear the Marshall Islands at 15:00 JST on August 26th according toJapan Meteorological Agency. Fig. 2 shows the best track of TyphoonSongda in JST. The typhoon reached its maximum intensity with acentral pressure of 925 hPa and a wind speed of 46.3 m/s at15:00 JST on August 31st and it traveled west-northwest to Okinawauntil 15:00 JST on September 6th. The typhoon changed trajectoryto north-northeast towards western Kyushu with a central pressureof 945 hPa and a wind speed of 38.6 m/s at 09:00 JST on September7th. After the typhoon entered the SJES, it began to weaken rapidly.The typhoon had become embedded in the mid-latitudes as anextratropical cyclone by 09:00 JST on September 8th.

The modeled typhoon tracks are plotted in Fig. 3, with colors toshow the central pressure difference between the observations andthe calculations. As expected, the simulated track obtained from theFMmodel matches well with Typhoon Songda's best track and centralpressures, because the typhoon's observed positions are used directlyin the FM model. As a result, the differences in the central pressuresbetween the FM model and the observations are within 10 hPa. Anydisparities are due to the differences in the typhoon center positionand the grid on the outermost domain D01 (see Fig. 3(b)). There arealso discrepancies between the best track and the estimated tracks intheWRF-NF andWRF-NA runs after the typhoon entered the SJES; thetwo tracks show a tendency to move to the left of the best track. Thetrack estimated in the WRF-NA model deviates from the best trackafter it has entered domain D03. TheWRF-NF estimated track deviatesfrom the best track after it has entered the intermediate domain D02.The differences between the central pressures in the WRF-NF and

Fig. 3. (a) Difference between Typhoon Songda's best track and estimated tracks. (b) Difference between observed and modeled central pressures.

S. Kim et al. / Ocean Engineering 91 (2014) 133–145136

WRF-NA models are up to 45 hPa in the early stages of the runs.However, as the typhoon approached Kyushu, the differences in theestimated central pressures are smaller, reducing to within 30 hPa

after 00:00 JST on September 7th (Fig. 3(b)). The differences betweenobservations and the central pressures estimated in the WRF-NAmodel are generally larger than those estimated in theWRF-NFmodel.

Fig. 4. Comparisons of observed data with estimations for wind speeds and directions along the best track.

S. Kim et al. / Ocean Engineering 91 (2014) 133–145 137

There would be other changes in surface pressure field if thenudging is applied to the intermediate domain. Further the physicsparameterizations and the number of vertical layers are alsocritical factors affecting the pressure fields and the tracks of atyphoon. In this study, these results are attributed to the differentnudging method used for the four-dimensional data assimilation,applying to all domains in the case of WRF-NA but only to theoutermost domain in the case of WRF-NF. In addition, the FNL dataused for the initial and boundary conditions in the WRF modelaffects the accuracy of the hindcast meteorological field. Theaccuracy of the central pressures and positions of tropical cyclonesmay be improved by implementing a bogus scheme in the WRF(e.g., Low-Nam and Davis, 2001). However, no such scheme wasused in the present study.

3.2. Comparisons of the wind and pressure fields

According to the previous studies (e.g., Yeh and Elsberry, 1993a,1993b; Jankov et al., 2005), it was found that appropriate physicsparameterizations in WRF improve typhoon's wind and pressurefields. It was addressed that the proper planetary boundary layerscheme increases wind fields up to 4% over the SJES (e.g., Kimet al., 2012, 2013). In the present study, the nudging method wasonly assessed in WRF to reproduce the typhoon‘s asymmetry field.

Figs. 4 and 5 compare the observed wind speeds and sea levelpressures at several stations with the corresponding estimatedvalues. Since the anemometer heights are different at each station,we adjusted the observed wind speeds to the 10 m level using thefollowing simple method by Hsu et al. (1994):

u2 ¼ u1ðz2=z1Þp ð2Þwhere u2 is the wind speed at the reference height of z2, u1 is thewind speed measured at height z1, and the value of p is taken to be0.11 for most ocean conditions.

Before the typhoon enters the SJES, it is seen that the FM modelprovides poor estimates of the wind speed and direction, forexample at station 4.LTF2049 (see Fig. 4). Sea level pressures areoverestimated at stations 1.LTF2038 and 7.LTF2097 when comparedwith the adjusted observations (see Fig. 5). These results would beimproved if asymmetric wind and pressure models are used. On theother hand, it is seen that the estimated wind speeds and directionsfrom the WRF-NF and WRF-NA runs match reasonably well withthe observations at station 4.LTF2049 (Fig. 4). In the same fashion,the results for the wind predicted by the WRF runs are alsomatched by the adjusted observations for station 12.LTF2019, whichis more than 300 km from the best track. For station 14.LTF2082 atthe border of the Seto Inland Sea, any models overestimate thewind speed before the wind direction changes from 451 to 2701,since this station is bounded by mountainous regions to the west,the north and the east in these simulations which used the fixednesting grid system, even though it would be improved by increas-ing the spatial resolution in the WRF simulations. The sea levelpressures estimated by the WRF-NF and WRF-NA runs are close tothe observations at station 1.LTF2038 (Fig. 5). Although onlyrepresentative results are shown in the present paper, it was foundthat similar tendencies, as described above, could be seen for mostof the stations bordering either the Pacific Ocean or the SetoInland Sea.

After the typhoon has made landfall along the Tottori coast, theFM results begin to change. At stations 20.LTF2051 and 29.LTF2072, the FM run overestimated the maximum wind speedby a factor of two, even though it successfully simulates both theincrease and the decrease in wind speed. In contrast, the WRF runsprovide good estimates of the maximum wind speed, but theincrease in wind speed before and the decrease after the max-imum are slightly shifted in time. The wind direction showed

more complicated changes at stations bordering the SJES than thatat stations in the Pacific Ocean. Such complicated changes in thewind direction are well simulated in the WRF runs. Nevertheless,

Fig. 5. Comparisons of observed pressures with estimated values.

S. Kim et al. / Ocean Engineering 91 (2014) 133–145138

for the sea level pressure, the FM run provided the best perfor-mance. In particular, the minimum sea level pressures are satis-factorily simulated in the FM run, as shown at stations of 13.LTF2087, 18.LTF2020 and 21.LTF2041.

Fig. 6 shows the wind speed and sea level pressure field inspace obtained from the FM and WRF-NF runs. After landfall of thetyphoon at Kyushu, the sea level pressure field according to the FMmodel shows no deformation in structure, but that from the WRF-NF run shows a noticeable deformation in shape, especially alongthe mountain areas on the Korean Peninsula and in Japan.Although the wind vectors from both runs reveal the generalfeature that the wind speed is higher on the right side of the trackand the wind direction is counterclockwise, the differences in theFM and WRF-NF wind vectors are significant. The wind fieldasymmetry from the WRF-NF model are controlled by the moun-tains facing the SJES: it can be seen from Fig. 4 that the simulatedwind direction and speed at each station of the Tottori coasts arecloser to the observations.

We have summarized the accuracy of the wind speeds obtainedfrom the WRF-NF, WRF-NA and FM runs against the observationsby using three indices: the correlation coefficient (CC), the rootmean square error (RMSE) and the maximum speed difference(MSD). The results are shown in Fig. 7. The variations in CCbetween stations are similar for the WRF-NF and WRF-NA runs.CC values for the FM run are generally smaller than those for theWRF runs, especially at station No. 15 located on the border of theTsushima/Korea Strait linking to the SJES. The values of CC rangefrom 0.24 to 0.95 for the FM model, 0.45 to 0.95 for the WRF-NAmodel and 0.55 to 1 for the WRF-NF model. The averaged values ofCC are 0.73, 0.79 and 0.80 for the FM, WRF-NA and WRF-NFmodels, respectively.

The RMSEs are smaller than 1 m/s at stations Nos. 2–6 locatedon the Pacific Ocean. The average RMSEs are 3.74 and 4.12 m/s forthe WRF-NA and WRF-NF runs, respectively. On the other hand,the RMSEs range from 3 to 9 m/s and the averaged error is 5.42 m/sfor the FM run. Furthermore, the FM MSD is much larger at somestations than the MSDs for the WRF runs. Thus, compared withusing the FMmodel, the WRF model improved the simulation of thewind field of Typhoon Songda around the SJES.

As with the wind speed, the accuracy of the estimated sea levelpressures has been evaluated and the results are shown in Fig. 8.The sea level pressures from the WRF-NF and WRF-NA runs havelarge CCs over 0.9 at most stations, while the CCs of the FM modelshow relatively low values after the typhoon entered the SJES. TheFM RMSEs range from 3.5 to 11 hPa, while those from the WRFruns range from 1.0 to 4.5 hPa. The maximum pressure differences(MPDs) are similar at stations Nos. 1–11. At the other stations, theMPDs of the WRF runs show small differences compared withthose from the FM run. Overall the sea level pressures estimatedby the WRF-NF and WRF-NA models show good agreement withthe observations, especially when compared to those from theFM model.

3.3. Comparisons of waves and surges

Waves and surges induced by Typhoon Songda were simulatedby three different wind and pressure fields using the FM, WRF-NAand WRF-NF models. The numerical experiments were conductedunder the conditions as described above. In the first run (referredto as WD), the coupling of wave and surge was taken into accountin calculating storm surge. A wave-dependent drag coefficientproposed by Janssen (1989, 1991) was used. In the second run

Fig. 6. D01 domain views of calculated wind speed vectors and sea level pressure contours according to the FM and WRF-NF models in the D01 domain.

S. Kim et al. / Ocean Engineering 91 (2014) 133–145 139

(referred to as HM), coupling was not used and surge wassimulated with the sea surface drag coefficient proposed byHonda and Mitsuyasu (1980), given by

Cd ¼ð1:290�0:024WÞ � 10�3ðWr8m=sÞð0:58þ0:063WÞ � 10�3ðW48 m=sÞ

(ð3Þ

As approaching the typhoon, the east short-fetch wind along theTottori coasts generated the waves showing their peaks coincidingwith the highest wind around 17:00 September 7th. After passing thetyphoon, the north-west long-fetch wind on the typhoon tail regionthe long wave period appeared from 01:00 September 8th. Fig. 9compares the predictions of the significant wave height and periodwith the observations. As noted above, the wave characteristics werecalculated for the WD runs using the three different meteorologicalfields estimated by the WRF-NF, WRF-NA and FMmodels. The FM runoverestimated the significant wave heights ðHsÞ, particularly when thetyphoon approached closest to Sakai Minato. This is attributed to theoverestimated wind speeds produced by the FM model. However, the

time series of Hs obtained from the WRF-NF and WRF-NA modelsare in good agreement with the observations, especially with regard tothe peak of Hs. However, after the maximum Hs has been reached, theWRF-NA model reproduces the decrease in Hs rather better than therun using theWRF-NFmodel. The three sets of predicted wave periodsðTsÞ are similar during the period September 7th to 8th as the typhoonapproaches, but they are notably different during September 5th to7th and after September 8th. In general, Ts values are underestimatedby all models. The WRF-NF run is best in predicting the observedperiods until September 8th, while the FM run shows the best resultsafter that date.

Fig. 10 compares the calculated after-runner surges driven by thedifferent meteorological fields with the observations at Sakai Minato.A large oscillation in the time series was seen in the FM model whenthe typhoon passed near Sakai Minato and these results are notshown. The fluctuations in the sea level may be attributed to strongsynoptic background winds predicted by the FMmodel in the SJES. In

Fig. 7. Statistical indices: (a) correlation coefficient (CC), (b) root mean square error(RMSE) and (c) maximum speed difference (MSD) for wind speeds at every station. Fig. 8. Statistical indices: (a) correlation coefficient (CC), (b) root mean square error

(RMSE) and (c) maximum pressure difference (MPD) for sea level pressures atevery station.

S. Kim et al. / Ocean Engineering 91 (2014) 133–145140

addition, the SuWAT model is run with a spin-up, involving pre-calculations to obtain stability in the sea surface level and the currentin an entire computational domain, for some periods prior to thetyphoon event. During the spin-up, the current and the sea levelshould propagate into the entire domain to be stable. However, asshown in Fig. 1, the SJES has the characteristics of an enclosed sea,and this geophysical feature might prevent the current and the sealevel from propagating sufficiently into the SJES during the spin-upcalculation under the meteorological field obtained from the FMmodel. In addition, it is believed that the weak synoptic backgroundwind over the SJES during the spin-up might produce unstablecirculations in the SJES. However, it is clear that much additional

work is required before a complete understanding can be reached incase of using a parametric wind and pressure model in the SJES.

The results of the WRF-NF runs with the different sea surfacedrag coefficients are shown in Fig. 10(a) which shows that thedifference is ignorable. The results imply that the origin of the after-runner surge does not come from the wind stress predominantly.We will show its origin in detail in Section 3.4. The estimated after-runner surges agree reasonably well with the observations near thepeak, regardless of the sea surface drag coefficient. However, whenthe typhoon approached Sakai Minato, the two runs underesti-mated the sea level from 0900 on September 8th. These under-estimations may be caused by the fact that the estimated typhoon

Fig. 9. Comparisons of observed and calculated significant wave height Hs and period Ts simulated from SuWAT runs with the wind and pressure fields obtained from theWRF-NF (nudging in only the outermost domain), WRF-NA (nudging in the entire domain) models and FM models.

Fig. 10. Comparisons of observed and calculated sea surface levels simulated from SuWAT runs with the wind and pressure fields obtained from the WRF-NF (nudging inonly the outermost domain) and WRF-NA (nudging in the entire domain) models, and different drag coefficients of the wave dependent drag coefficient (WD) and the Hondaand Mitsuyasu drag coefficient (HM).

Fig. 11. Comparisons of calculated sea surface levels at Sakai Minato with and without Coriolis force in the case of SuWAT runs using the meteorological field obtained fromthe WRF-NF model and the wave dependent drag coefficients.

S. Kim et al. / Ocean Engineering 91 (2014) 133–145 141

track was inclined to the left side of the best track (see Fig. 3(a)).Fig. 10(b) shows the sea level changes using the two sea surface dragcoefficients for the WRF-NA run. The results indicate that the sealevel rises are well simulated until 1500 on September 7th. Theestimated typhoon track was close to the best track at that time (seethe diamond symbols in Fig. 3(a)). Later, the after-runner heightsusing the WD and HM coefficients are lower than the observations.The difference between the WRF-NF and WRF-NA runs was thenudging scheme for the four-dimensional data assimilation executed

in all domains or in the outermost domain alone. In summary,accurate wind and pressure fields in a typhoon affected by topologyare crucially important in simulating after-runner surges.

3.4. Effect of Ekman setup

We conducted additional calculations as turning Coriolis forcingoff on the momentum equations for the condition of WRF-NF runand compared the results with those obtained from the WRF-NF

128°E 130°E 132°E 134°E 136°E

000.1

0.2

0.2

0.2

0.20.2

0.3

0.30.3

0.3

0.335°N

36°N

37°N

30m/s

with Coriolis Force

128°E 130°E 132°E 134°E 136°E

0

0.1

0.1

0.1

0.1

0.2 0.20.2

0.2

0.20.20.2 0.2

0.30.40.5

35°N

36°N

37°N

30m/s

with Coriolis Force

128°E 130°E 132°E 134°E 136°E

0.1

0.1

0.10.1

0.2

0.2

0.2

0.3

0.30.3

0.40.5

35°N

36°N

37°N

30m/s

with Coriolis Force

128°E 130°E 132°E 134°E 136°E

0.1

0.10.1

0.1

0.20.2

0.20.2 0.2

0.20.2

0.3

0.3

0.4

35°N

36°N

37°N

30m/s

with Coriolis Force

128°E 130°E 132°E 134°E 136°E

00.1

0.1

0.2

0.2

0.2

0.3

0.3

0.335°N

36°N

37°N

30m/s

without Coriolis Force

128°E 130°E 132°E 134°E 136°E

0.1 0.1

0.20.2

0.3

35°N

36°N

37°N

30m/s

without Coriolis Force

128°E 130°E 132°E 134°E 136°E

0.10.1

0.1

0.2 0.2

0.3

35°N

36°N

37°N

30m/s

without Coriolis Force

128°E 130°E 132°E 134°E 136°E

0.1

0.1

0.1

0.10.1

35°N

36°N

37°N

30m/s

without Coriolis Force

Sakai Minato Sakai Minato

Fig. 12. D03 domain views of calculated sea surface level contours and wind vectors with (left panels) and without Coriolis force (right panels) in the case of SuWAT runsusing the meteorological field obtained from the WRF-NF model and the wave dependent drag coefficients. (a) 17:00 JST September 7th, (b) 22:00 JST September 7th,(c) 02:00 JST September 8th, (d) 08:00 JST September 8th, (e) 17:00 JST September 7th, (f) 22:00 JST September 7th, (g) 02:00 JST September 8th and (h) 08:00 JST September 8th.

S. Kim et al. / Ocean Engineering 91 (2014) 133–145142

run in order to investigate Ekman setup at the coast. Ekman setup isdue to a geostrophic balance between the Coriolis force acting onthe along-shelf current and the across shelf pressure gradient (e.g.,Freeman et al., 1957). As shown in Fig. 11, the differences betweentwo time series of computed sea surface levels are apparent afterthe typhoon landfall: the observed second sea level rises are wellsimulated when considering the Coriolis force, while those no occurin case of the absence of the Coriolis force. The detailed processesare shown in Fig. 12. When the typhoon is located near 1301E and371N in the SJES, it is shown that the sea level rises are 0.3 mgenerated along the left coast's edge of Sakai Minato (Fig. 12(a)).At that time, the sea surface level at Sakai Minato is going down upto �0.1 m after the first sea level rises occur (see, Fig. 11). As the

typhoon moves to the northeast further, the surge wave generatedis propagating to Sakai Minato, toward the east (Fig. 12(b) and (c)).The figures indicate that the predominantly shore-parallel windspeeds less than 30 m/s are calculated over most of the Tottori coastduring the period of the second surge waves observed at SakaiMinato. A strong, mainly, shore-parallel depth-averaged currents upto 100 cm/s are widely computed over the Tottori coast (Fig. 13(a)).These strong currents could force a setup of 0.5 m on the shelf ofTottori coast. However, in case of in the absence of the Coriolis force,the figure shows that the weaken current mainly indicates towardoffshore against the Tottori coast (Fig. 13(b)). The processes forthe surge wave in the absence of the Coriolis force are shown inFig. 12(e)–(h): the figures show that the second sea level rise no

Fig. 14. The contribution of the Coriolis force to the calculated maximum surge heights when using the wind and pressure field of WRF-NF with the wave dependent dragcoefficients.

Fig. 13. D03 domain views of calculated depth-averaged current fields and vectors with (upper panels) and without Coriolis force (lower panels) at 02:00 JST September 8th.

S. Kim et al. / Ocean Engineering 91 (2014) 133–145 143

occurs. Then, the surge waves lastly disappear as passing throughHokaido in case of in the presence of the Coriolis force (Fig. 12(d)).From these results, it was found that the after-runner surge appears tobe forced by the Ekman drift. Its effect during Typhoon Songdamay beseen over the shelves of Korea and Japan as well in Fig. 14, whichshows the contribution of the Coriolis force to the calculated max-imum surge heights: it is up to 70% along the Tottori coast. Our findingwas in line with the conclusions by Kennedy et al. (2011) that aforerunner surge generates in advance of hurricanes in Gulf of Mexicoas the Ekman setup on the wide and shallow shelf.

4. Conclusions

The highest surges appear along the western coasts of the TottoriPrefecture 15–18 h after typhoons have passed or have moved nearto 401N. Typhoons often undergo extratropical transition in themid-latitudes of 351N–551N. In the present study, we simulated thewind and sea level pressure fields due to Typhoon Songda in 2004using a parametric wind and pressure model, and a regionalatmospheric circulation model. For the parametric typhoon model,Schloemer's formula (1954) and Fujii and Mitsuta's wind model(1986) were used. The weather research and forecasting (WRF)model was employed as the regional atmospheric circulationmodel. The NCEP FNL (Final) Operational Global Analysis data wereused for the grid nudging method in the WRF runs, applied to theoutermost domain alone and also to all domains.

Comparing each model with Typhoon Songda's best track data,the WRF models showed differences in the central pressures andthe estimated tracks over the SJES. Nevertheless, the wind speedsand the sea level pressures for all domains showed that the WRFmodels were in better agreement with observations over the SJESthan the FM model. Unlike coasts along the Pacific Ocean wherethe parametric typhoon model successfully reproduces themeteorological field, the deformation of the typhoon structure,resulting from geophysical features or extratropical transitions/intensity changes, should be taken into account for the SJES.

It was found that applying asymmetric wind and pressure fieldis preferable to the after-runner surge simulations for TyphoonSongda in the SJES: using the grid nudging method only to theoutermost domain of the WRF model provides fairly good agree-ment with observation at Sakai Minato, while the WRF modelemploying nudging to all domains underestimates the after-runner surge. Clearly, the prediction of accurate wind and pressurefields in a typhoon is important to simulate the after-runner surgealong the coasts of the Tottori Prefecture.

The study indicated that the after-runner surge's geostrophiccharacteristic comes from the Ekman setup and the predominantfactor is the Coriolis force over the Tottori coast. These after-runner surges have been observed from many typhoons, forexample, Vera (1986), Dinah (1987), Caitlin (1991), Mireille(1991), Rusa (2002), Maemi (2003) and Megi (2004) along thiscoastal region. Therefore, it is important that the after-runnersurge's behavior should be considered for the safety design ofcoastal defense structures around the Tottori coastal area.

Acknowledgments

This work was partially conducted under the framework of the“Projection of the change in future weather extremes using super-high-resolution atmospheric models” supported by the KAKUSHINProgram, and the Grant-in-Aid for Scientific Research (B) of theMinistry of Education, Culture, Sports, Science, and Technology(Grant no. 23760464). The authors would like to thank Eur Ing T.S.

Hedges, University of Liverpool, UK, for his many helpful sugges-tions and a detailed reading of the manuscript.

References

Bender, M.A., Tuleya, R.E., Kurihara, Y., 1987. Numerical study of the effect of islandterrain on tropical cyclones. Mon. Weather Rev. 115, 130–155.

Chen, F., Dudhia, J., 2001. Coupling an advanced land-surface/hydrology model withthe Penn State/NCAR MM5 modeling system. Part I: Model description andimplementation. Mon. Weather Rev. 129, 569–585.

Chou M.-D., M.J. Suarez, 1994. An Efficient Thermal Infrared Radiation Parameter-ization for Use in General Circulation Models. NASA Tech. Memo 104606, vol. 3.pp. 85.

DiMego, G.J., Bosart, L.F., 1982. The transformation of tropical storm agnes into anextratropical cyclone. Part I: The observed fields and vertical motion computa-tions. Mon. Weather Rev. 110, 385–411.

Freeman, J.C., Baer, L., Hung, C.H., 1957. The bathystropic storm tide. J. Mar. Res. 16,12–23.

Fujii, T., Mitsuta, Y., 1986. Synthesis of a stochastic typhoon model and simulation oftyphoon winds. Ann. Disaster Prev. Res. Inst., Kyoto Univ. B-1 (29), 229–239.

Fujita, T., 1952. Pressure distribution within Typhoon. Geophys. Mag. 23, 437–451.GEBCO: General Bathymetric Chart of the Oceans, 2003. British Oceanographic Data

Centre Centenary Edition.Honda, T., Mitsuyasu, H., 1980. Experimental study of wind stress in the sea surface.

Annu. J. Coast. Eng., JSCE 27, 90–93.Hong, S.-Y., Dudhia, J., Chen, S.-H., 2004. A revised approach to ice microphysical

processes for the bulk parameterization of clouds and precipitation. Mon.Weather Rev. 132, 103–120.

Hsu, S.A., Meindl, E.A., Gilhousen, D.B., 1994. Determining the power-law wind-profile exponent under near-neutral stability conditions at sea. Appl. Meteorol.33 (6), 757–765.

Jankov, I., Gallus, W.A., Segal, M., Shaw, B., Koch, S.E., 2005. The impact of differentWRF model physical parameterizations and their interactions on warm seasonMCS rainfall. Weather Forecast. 20 (6), 1048–1060.

Janssen, P.A.E.M., 1989. Wave induced stress and the drag of air flow over seawaves. J. Phys. Oceanogr. 19, 745–754.

Janssen, P.A.E.M., 1991. Quasi-linear theory of wind-wave generation applied towave forecasting. J. Phys. Oceanogr. 21, 1631–1642.

Jeong, C.K., Panchang, V., Demirbilek, Z., 2012. Parametric adjustments to therankine vortex wind model for Gulf of Mexico hurricanes. J. Offshore Mech.Arct. Eng. 134 (4) 041102-041102-17, http://dx.doi.org/10.1115/1.4006148.

Kain, J.S., 2004. The Kain-Fritsch convective parameterization: an update. J. Appl.Meteorol. 43, 170–181.

Kennedy, A.B., Gravois, U., Zachry, B.C., Westerink, J.J., Hope, M.E., Dietrich, J.C.,Powell, M.D., Cox, A.T., Luettich Jr., R.A., Dean, R.G., 2011. Origin of the HurricaneIke forerunner surge. Geophy. Res. Lett. 38, L08608. http://dx.doi.org/10.1029/2011GL047090.

Kim, S.Y., Yasuda, T., Mase, H., 2008. Numerical analysis of effects of tidal variationson storm surges and waves. Appl. Ocean Res. 30 (4), 311–322. http://dx.doi.org/10.1016/j.apor.2009.02.003.

Kim, S.Y., Yasuda, T., Mase, H., 2010. Wave set-up in the storm surge along opencoasts during Typhoon Anita. Coast. Eng. 57, 631–642. http://dx.doi.org/10.1016/j.coastaleng.

Kim, S.Y., Matsuura, T., Matsumi, T., Tom, T.H., Yasuda, T., Mase, H., Nishino, H., 2012.A study of sensitivity analysis of WRF parameters for meteorological predic-tions at the Sanin Coast. J. Jpn. Soc. Civil Eng., Ser. B2 (Coast. Eng.) 68,I_1236–I_1240 (In Japanese).

Kim, S.Y., Matsuura, T., Matsumi, Y., Kazuhisa, T., Yasuda, T., Tom, T.H., Mase, H.,2013. Effects of WRF parameterization on meteorological predictions in themid-latitude region -planetary boundary layer scheme and cloud microphysics.J. Jpn. Soc. Civil Eng., Ser. B2 (Coast. Eng.) 69, I_516–I_520 (In Japanese).

Kitabatake, N., 2008. Extratropical transition of tropical cyclones in the westernnorth pacific: their frontal evolution. Mon. Weather Rev. 128, 947–972.

Klein, P.M., Harr, P.A., Elsberry, R.L., 2000. Extratropical transition of western NorthPacific tropical cyclones: an overview and conceptual model of the transforma-tion stage. Wea. Forecast. 15, 373–396.

Lee, H.S., Kim, K.O., Yamashita, T., Komaguchi, T., Mishima, T., 2010. Abnormal stormwaves in the winter East/Japan Sea: generation process and hindcasting usingan atmosphere-wind wave modeling system. Nat. Hazards Earth Syst. Sci. 10,773–792.

Low-Nam, S., C. Davis, 2001: Development of a tropical cyclone bogussing schemefor the MM5 system. The Eleventh PSU/NCAR Mesoscale Model User's Work-shop, Boulder, Colorado. pp. 130–134.

Mase, H., Yasuda., T., Tracey, T.H., Tsujio, D., 2008. Forecast and hind castsimulations of wind waves which caused disasters along Toyama Coasts onFebruary 2008. Annu. J. Coast. Eng. (JSCE) 55, 156–160 (In Japanese).

Palmen, E., 1958. Vertical circulation and release of kinetic energy during thedevelopment of Hurricane Hazel into an extratropical cyclone. Tellus 10, 1–23.

Panchang, V., Jeong, C.K., Demirbilek, Z., 2013. Analyses of extreme wave heights inthe Gulf of Mexico for offshore engineering applications. J. Offshore Mech. Arct.Eng. 135 (3) 031104-031104-15. http://dx.doi.org/10.1115/1.4023205.

Schloemer, R.W., 1954. Analysis and synthesis of hurricane wind patterns over LakeOkechobee, FL. Hydromet. Rep. 31, 49 (Govt. Printing Office, No. C30.70:32).

S. Kim et al. / Ocean Engineering 91 (2014) 133–145144

Skamarock, W.C., et al., 2008. A Description of the Advanced Research WRF version3. NCAR Tech. Note NCAR/TN-4751STR, 113pp.

Thompson, G., Rasmussen, R.M., Manning, K., 2004. Explicit forecasts of winterprecipitation using an improved bulk microphysics scheme. Part I: Descriptionand sensitivity analysis. Mon. Weather Rev. 132, 519–542.

Veltcheva, A., Kawai, H., 2002. Investigation of the typhoon pressure and wind fieldwith application for storm surge estimation. Rep. Port Airpt. Res. Inst. 41 (2),23–45.

Yeh, T.-C., Elsberry, R.L., 1993a. Interaction of typhoons with the taiwan orography.Part I: Upstream track deflections. Mon. Weather Rev. 121 (12), 3193–3212.

Yeh, T.-C., Elsberry, R.L., 1993b. Interaction of typhoons with the Taiwan orography.Part II: Continuous and discontinuous tracks across the island. Mon. WeatherRev. 121 (12), 3213–3233.

S. Kim et al. / Ocean Engineering 91 (2014) 133–145 145