気象研究所研究報告 第46巻第1号 9-17頁 平成7年8月 Papers in Meteorology and Geophysics VoL46,No.1,pp,9-17,August1995

An experimental storm s皿ge prediction for the western part of

the lnlana Sea with application to Typhoon9119

by

Tatsuo Konishi

.ルZ6彪ozologJ6‘zl R召s6‘zz6h乃zsガオz話召, 乃z6陶z66‘z,Z6‘z名とzゐ4 305ノψ‘zn

(Received March27,1995:Revised July12,1995)

Abstract

It has been said that we are entering a stage in which super come to with greater frequency than in the past.That makes people uneasy about the occurrence ofbigstorm surges.Actually,the19thtyphoon of1991(Mireille)c&usedremarkablestrom surges exceeding3m in height accompanied by great imndation along the coast in the westem part of the Inland .In order to provide much more information on possible storm surges in advance than is done today,a dynamica1-numerical model mustbe developed for the forecasting of them.This paper proposes a numerical forecast model and a procedure on a real time basis,and shows the result of an experimen- tal forecast applied to Typhoon9119.It includes a new metho(10f determining the parameters of the modeL

there is no storm surge record in the past. 1.Introduction Moreover,it sometimes mderestimates a peak The19th typhoon of1991(Mireille,here- surge (iue to a typhoon taking a particular after referred to as Typhoon9119)caused course(Konishi,1990).Therefore,a dynamical remarkable storm surges exceeding 3m in -mmerical mo(iel must be developed for the height with great inun(iation resulting along the forecast of future storm surges. coast in the westem part of the Inland Sea of Many simulation studies of storm surges Japan. As shown in Konis阜i(1995),the fre- have been conducted for past events(Miyazaki quency of high storm surges in Japan has in- 6渉α1.(1961),Unokiασ1.(1964),Konishi and creased recently,being associated with the Tsuji(1995)).But investigations aiming at real same tendency of the frequency of typhoons time forecast of them are very few,especially hitting Japan.Quantitative forecast of storm in Japan. Konishi(1989)is one of these,but is surges before event is very important for disas- based on the supposition that observed meteor- ter prevention.At present,the Japan Meteoro- 010gical data during the passage of typhoons logical Agency(JMA)apPlies an empirical for- are all given。 Flather(1994)apPlied the North mula to the real time forecast of storm surges Sea model of England to the storm surge in the (JMA,1994).The method has rather good Bay of Bengal and performed similar experi- accuracy,but also has a limitation such that we ments to the present study.They utilized only can not forecast storm surges at a point where the meteorological data from real time forecast condition,and forecast the disaster in April ◎1995by the Meteorological Research Institute 1991,but the observed data on tide were so poor 10 Konishi,T. Vo1.46,No.1

that comparison between observation and cal- g is the gravity acceleration,and f is the Cor- culation was very difficult. Hubbert6!磁 iolis parameter.Parameterizations of聴x and (1991)described the forecast model of storm zbx are given as follows: surges on a real time basis in Australia. る、=ρ、CdlWIWx,為x=ρwCdwlVIVx, Jelesnianskiαα1.(1992)described the USA’s mode1.From these studies and also thehin(icas- whereρa is the density of the atmosphere,Cd is ting studies of past events,the accuracy of thedragcoefficientattheseasurface,Wisthe storm surge forecast has been found to depend wind velocity vector,Wx is the x component of largely on those of wind and surface pressure W,ρw is the density of sea water,Cdw is the forecast. drag coefficient at the sea bottom,V is the The purpose of this paper is to propose a mean water velocity vector,and Vx is the x procedure of storm surge forecast on a real component of V.τεy andτわy are parameterized time basis and to show the result of an experi- in a similar way.Water head蜀is originated mental forecast applied to Typhoon9119.It from the effect of atmospheric pressure and the includes a new method of determining the relation isρ、gζ6=一(Ps-Pn).Ps and Pn are the parameters of the typhoonmode1.These points surface pressure of the atmosphere and the are presented in the following order.Section2 peripheral pressure of the typhoon,respective- introduces the stormsurge modeL Thetyphoon 1y.Drag coefficients,Cd.and C(iw,are assumed model and method of determining the storm to be3.5×10-3and2.6×10-3,respectively.The parameters are explained in Section3.Actual values are selected from discussion of past procedure for Typhoon9119is presented in stu(1ies,especially,Konishi and Tsuji(1995). Section4.Results of estimated meteorological Boundary conditions are based on Konishi components and experimental forecast of (1989). On the coastal boundary,the condition storm surge,and discussion ofthem aregiven in of zero normal velocity is applied.As open Sections5and6,respectively. boundary conditions,modified radiation condi- tions, 盟=〉切(ζ一ζb),for the the eastern bound- 2.Storm surge moael ary, 2.1 Basic equations and related parameters 〃’=一禧万(ζ一ζb),for the westem boundary, The method of numerical simulation uti1- and (4) ized in the present study is the same as that in ハπニーゆ(ζ一ζ6),for the southen boundary Konishi(1989).The two-dimensional equations are used. of motion and continuity are expressed as 躍魂釧万慾房 ヲ ( 蜘蜘 非躍}’砒 鴇一み 十恥ρ耐画xω 一砺一ρ舞ヱ ル フ 一 一 一ψψ募十 (1) 3. Typhoon model In the state of the art,the accuracy of ζ ) (2) prediction of a typhoon track is not sufficient for storm surge prediction,especially for disas- (3) ter prevention.Therefore,several calculations where the origin,0,is within the equilibrium must be done to predict storm surges for the level of the sea surface,Ox points east,Oy possible courses of the track.For this,we are north,Oz is directed vertically upwards.We obliged to use a parametric typhoon model in define u and v as the x and y components of place of the results of numerical weather pre- velocityandthenMandNarethexandy diction. Fujita’s formula(Fujita,1952)was components of volume flux given by supposed as the distribution of atmospheric 躍〒∫o繍,N一∫o畝 pressure,which is given by ∠P り り Ps-jPn一 (5) respectively.ζis the sea surface displacement 厩アフ from the mean sea surface leve1,D is the depth, where∠P is the central pressure depression of 1995 Anexperimentalst・rmsurgepredicti・6withapPlicati。nt。Typh。。n9119 11 a typhoon,r is the distance from the center,名o numerical model.Storm surges are sensitive on is the parameter which is related to the radial the wind stmcture near the typhoon center.In distribution of the atmospheric pressure. In order to inclu(ie information on the stucture in this case,Pn was supposed to be lO10hPa皐s estimating名o,we intend to evaluate the radius conventionally used in simulation studies(Kohi- of the maximum wind,Rm,which has not been shi and Tsuji,1995).Wind in the typhoon has determined in our rea1-time forecast system. been calculated for hindcast of storm surges by Chubachi(1988)performed a regressioh analy- Miyazaki’s formula(Miyazakiαα1.,1961) sis between the positions of maximum surface written in the form, winds from aircraft recomaissance due to the W=C1・G’十C2・C・exp(一π7/7.), (6) Joint Typhoon Waming Center(JTWC)at where GI is a wind vector,C is the velocity and the observed clou(1pattems of vector of the center of the typhoon and7.is an typhoons from the geostationary meteorologi- attenuation factor. The magnitude of7θis cal satellite“Himawari”.He made three types normally assigned to500km in the formula and of the regression equations for three categories the same value was set for7.in the present of cloud pattems,EYE,CDO,and OTHERS. study.G〆has the same magnitude as the gradi- EYE pattem means that the eye of a typhoon is ent wind vector G,and the direction ofαiS clearly found in CDO(Central Dense Overcast). tumed inward of the typhoon at a certain angle CDO pattem means that no eye but CDO is from G to take the effect of friction between found in the satellite image.OTHERS indi- the atmosphere and the earth7s surface into cates that there is no eye and no organized account.The magnitude of G is calculated by cloud image.Good correlations,about O.6to the equation, O.7,were obtained for EYE and CDO pattems, ノ IGl一髪・[一・+1・+(ρα舞・)静}],(7) and the regression equations of the position of maximum winds are where the notations are the same as those in Rm=3.857*E-0.271*Mw7 Section2and Ps is given by equation(5).There 十〇.531*Cus十〇.227*PcT179.613 afe insufficient rea1-time data to allow a direct for EYE pattem (9) analysis of the central region of mosttyphoons. Rm=1.452*Cuw-0.470*Mw7 Therefore,equation(6)is simplified and sup- 十〇.061*Tyn十27.552 posed to be written as for CDO pattem (10), w=σ(G’十c・exp(一π7/7、)), (8). where E is the length of the malor axis of the The inflow angle between GI and G is supposed eye inunit of O.1degrees,Mw7is themaximum to be30degrees as in many past studies(e.g. wind speed of700hPa in unit of kt,Cus and Unoki6オα1.,1964). Base(10n these supposi- Cuw are the extensions of the convection cloud tions,the parameters we must determine in real towards south and west respectively in unit of time are∠P,名o,andσof equations(5)and(8) O.1degrees,Pc is the central pressure in unit of an(1positions of the typhoon.The latter gives hPa,Tyn is the latitude of the typhoon center in the typhoon velocity C in equation(8)乙 unit of O.l degrees.The Meteorological Satel- At fixed times,the forecaster in charge Iite Center of the JMA detemines these cloud determines the position of the typhoon center parameters except Mw7routinely at6-h inter- and the intensity of the typhoon,an(i predicts vals.So we can use them for forecast of storm the course and intensity change of the typhoon surges.However,no relevant data for Mw7are 120r24hours later.Information on the present fomd.The maximum wind on700hPa is a intensity includes central pressure,Pc,maxi- little stronger than that at the surface l on the mum surface wind,Wm,radius of over50-kt other hand the observed wind by an aircraft is winds,R50,and mean radius of lOOOhPa,R1000. weaker than the maximum wind.These have We determine∠P of(5)from Pc and assignthe oPPosite tendencies to each other,an(i it is interpolated positions between the present posi- expectedthatMw7maybenearlyequaltoWm. tion and that120r24hours later to those in the The maximum surface wind determined by the 12 Konishi,T. Vol.46,No,1 forecaster,Wm,is applied to Mw7.Now we κ’fos配π7 can evaluate all terms inthe right-hand sides of 132E 1 equations(9)and(10);and for the patterns of o 一. 80πoda 7 EYE and CDO,Rm can be estimated reason- 0 60 0 ably. Using the estimation of Rm and the ・20 o 6 22 50 ts∬γam information ofthe forecasterwith equations(5), 部 (7),an4(8),the magnitudes of名o and C’were ⑩ 50 3 21 determined in the following way.Firstly,the 14 15 16 7akada value of名o was selected,which minimizes the 例 100o一 square error between Rm and the position of 17 18 19 maximum wind calculated by equation(8). もb Next,usingthe名o,C’wasdeterminedbyminim- 33N izing the square error between wind speeds of 100 Wm and50kt andthecorrespondingcalculated values. ● 00 0 200 0 For the category of OTHERS,we can not 1000m o obtain a proper value of Rm,and without infor- mation on Rm,the parameters are detemined only from the forecasters〆information,such as Wm,R50,and RlOOO.This can be validated from the following viewpoint.The state of no Fig.1 Domain of computation. The eye and no organized cloud,as t耳e OTHERS numeral with solid circle corre- pattem,means that the spatial scales of varia- sponds to the station mmber in tion of wind and pressure fields become large Figs.4 and 6. Contours of the and the position of maximum wind also depth are drawn. becomes obscure.Therefore,it is not effective for calculations of storm surges in OTHERS 0.85. Fig.2 indicates the positions of the pattems.In this case,we applied a recursive typhoon centers with time in hours. Solid procedure to estimating the apPropriate名o, circles are from the JMA7s best track data and which is similar to Ueno(1994). open circles are from interpolated positions whichwere determinedusing the JMA’s predic- tion at9h on27th Sept.The direction of the 4.Estimated parameters of Typhoon9119 forecast track is similar to the best track, and comPutational conditions southwest to northeast,but the forecast devi- Fig.1shows the computational domain ates east of the best track and the forecast which covers the westem part of the Inland speed is rather low. Sea.In this region,peak surges were observed The computations started at Oh on27th from18h on27th Sept.1991through22h on the Sept. and en(ie(1at6h on28th Sept. Initial same day(Konishi,1995). Conside血g the conditionsweresetasζ,MandN=Oandonly actual forecast condition,sufficient lead time the tidaLanomaly or storm surge was comput- must be provided for the protection of life and ed. Before9h on27th,the positions of the property.We selected about12hours as the typhoon were given as the best track,名o and C〆 leadtime andtyphoonparametersmentionedin are taken as the value estimated above and∠P the previous section were determined at9h on was taken as the observe(i pressure depression. 27th Sept.The magnitude of∠P was taken as 75hPa from the intensity identification of the 5。 Results of experimental forecast forecaster. The procedure gave that ro was 49(km)and C/was O.9.We also calculated the 5。1 Results of press皿e and wind profiles case of‘best track》and then C‘is taken to be Fig.3shows the radial distribution of 1995 An experimental storm surge pre{iiction with application to Typhoon9119 13

Pressure an(1 Win(1 3 0 巽壕遮… 3N6 945 1000 “ … ●胃 …}… 『●; 隈 .喚..駈 謡.. ● 門 r 『…点… O… ●ゑ . 0 0(の\自)眺ρ一〇〇一①> も自帽き 1 、; ●℃ し (娼餌ρ)①』βののΦ』山 ” ● ・f 45 9 5 0… 9 『『 r Or 『h …汐蕊….● 、 『墨筏癖、 ・ ・ 『O ’ 馳● ” 『継無囎-烹… 9 5 O

= 』 『『= 』=二み== ・ = ;. 35 α. .… 3 -を試…Y曙 『O』『… 2『一〇 ㎜ …掌-三e弘 ぎ L 乱S…劉鐸日羅U『U読…、 、 皿r;二噌、..乱……いe 榊 9 ◎ … … 、、晶r..命..

Q 「S ・w 94 .弓.● ..一 . 34

馬 o 細bs em pa t f 2 to nlan ea 900 0 100 200 300 4 0 Sa 1 33 Distance from center(k皿) 1 35 o Fig.3 Calculated radial profiles of sur- κ hμ face pressure and wind for the 9坐5 1 4 case of the forecast track. Solid squares are the positions of the γ maximum wind and the wind of50 knots,which were determined by the forecaster.

129 130 -131 132 133E1R formula overestimates a little around the posi- Fig.2 Track of Typhoon9119. Solid tion of the maximum wind and underestimates circles indicate positions of around the marginal area of the typhoon. typhoon centers on the best track Actual wind observation in the case of at the time ofthe attachednumber on Sept.27th,1991,with central Typhoon9119also supports our method of pressures of the typhoon in hPa. estimation of Rm.Maximum wind of49.7m/s Open circles indicate interpolated wa30bserved at Nomozaki at15:20,when the positions of forecast at9h on27th. center of the typhoon was located about40km west of Nomozaki.That wind was the highest atmospheric surface pressure and surface wind of wind observations in the Kyushu area(Fu- for the case of the forecast track.These distri- kuoka District Meteorological Observatory, butions were fixed at a relative position of the 1993).Onthe otherhand,the samemethod as in typhoon center throughout the calculation. Section3was applied to the satellite data at15 Surface pressure is calculated from(5)and 0’clock an(i it gave48km of Rm,corresponding wind is from(8).The abscissa of Fig3is the to39km of70.Observation at Nomozaki is distance from the typhoon center whose direc- almost consistent with our estimation of the tion is120degrees clockwise from the course of ra(1ius of maximum wind.Thus,our method the progress of the typhoon,the maximumwind was verified by the observed data at15:00。 existing on the line.Solid squares show the maximum wind and50kt wind which were 5.2 Results of experimental forecasts of determined by the forecaster.The radius of storm surges maximum velocity is taken as Rm=58(km),the A comparison of calculated peak surges maximum velocity as Wm=45(m/s)and the with those observed is shown in Fig.4.For radius of50kt as R50=160(nautical mile).The station number in the abscissa,see Fig.1。 14 Konishi,T. VoL46,No.1

Several cases are calculated,whose conditions are large(1ifferences at the stations Nos・1-8,21 are tabulated in Table1.Case l is the result and22in either case.We can find that the using the typhoon model with the parameters difference of the track between Case l and Case determined in Section 4,whose track is 21eads to a difference of lm in storm surge。 forecast(}d at9h.Case2is the result due to the The axis of the maximum wind deviates east in best track but the central pressure is the same Case l and it raises the sea level in the eastern as at9h. In Case3,0bserved pressure in as- part. There remains considerable difference signed to the central pressure for Case2.There between the observed storm surges and the results of calculated cases,even if we select the best track as the typhoon track.The reasons Peak surge (forecast) are(1iscussed in the next section. 5 …k} The temporal variations of the calculated F曜(”r臆胃 … 5e…4… ガa…漁τt…0町 五r㎜ ”r…”0” ”9… ”一一 ” …9…す storm surges are shown for Onoda,Takada, B 冒 一 胃 and in Fig.5.In the (自)4 ■ 〔皿工P・ ”乾…0 ・ ・ Ω…4 }恩 figure,the observed and the calculated storm ,a surges for the best track are also shown.The Φbo』⇒の図邸①山 3 n∠ 、… …、 ナ best track data corresponds to Case3in Table 諾牟酒キ嫉〜4μメ、 ’ { 、N… 〆 1.In these figures,a delay of peak surge in the .覧、 …、… forecast case is commonly observed.A lag of { 、堂 ,1 ” , two or three hours can be seen in the calculated ズ storm surges.This time lag comes from the …,γ彫7 difference of the typhoon speed between tthe 玉 、・ ’ f7 ” ノ一’ …、 b e r e 、 forecast case and the actual case,that is,the 1 ~’テ… speed of the forecast typhoon is slow,as seen in Fig.2. Actually,the peak surge due to the typhoon passing along the best track occurs at 0 2 4 6 8 10121416182022 about the same time as the observed one.The Station number time lags of calculated peak surges in Takada and Matsuyama which are found in the results Fig.4 Comparison of forecast peak for the best truck are discussed in Konishi and surges with those observed. Explanation of the cases are given Tsuji(1995).It is due to wind structure in the in Table1.Forthe stationnumber typhoon,a temporary wind lull which caused a on the abscissa,see Fig.1. seiche an(1delayed peak surge.

Table l Computing cases

Case Track Pc ro Cy a (hPa) (km) (C1,C2) degree

1 Forecast 935 49 0.9 30 2 Best 935 49 0.85 30 3 Best Obs. 49 0.85 30 4 Best Obs. 20(*) 0.9,6/7 50 5 Best Obs. (Estimated wind)(#) (*)For calculation of wind.Different value was selected for calculation of pressure.For details,see Konishi and Tsuji(1995).The column C7consists of CI and C20f eguation(6)on this line. (#)Wind field was inversely estimated by tidal data(Konishi and Tsuji,1995). 1995 An experimental storm surge prediction with application to Typhoon9119 15

4 ONODA 1991 9/27-28 5 HIR〔〕SHINA 1991 9/27-28

Σ)の﹈Oα⊃uり ΣαO一の 十 十 B R T N 十 4 o C T ■ 一 E A 、Σ)の﹈O匡⊃の Σ偲O』『の 十 B R T N 、 み 、 し覧十 o C T ’’ 響 一 E A t 十 、 監 十 、 十 ノ 、 / ’ 息 ’ 0 箪 十 、 宇 准 十 、 十 、 / ’ 十篭1 !’ ’ 、 ’ \ r1 0 6 12 18 0 亀 TIME(H〔〕URS) 覧 ÷ 見 十 ㌧ ’ Fig.5.1 一! Comparison of the computed ’ ’ storm surge with that observed 、』ノ

(marked by十)at Onoda.Solid 一2 0 6 12 18 0 and dashed curves correspond to TIME(HOURS) the calculations with the typhoon tracks of forecast and the best Fig.5.3 Same as in Fig.5.1,but for Hiro- track(Case3in Table l),respec- shima. tively.

1991 9/27-28 4 TAKADA 4 MATSUYA酬A 1991 9/27-28

.Σ)の国O配⊃の Σ配Oトの Σ)の﹈O叱⊃の Σ叱O」「のB O 『」hK C T T A B O ER C T T An 十 - N 十 一 N

丸 \

,’ t

, ’ 、 十 、 0 、㌧ 0 騰、・ 、 ノ

,’り 、 )

『10 一10 6 12 18 0 5 6 12 18 D 6 TIME(H〔〕URS) TI卜1E(H〔3URS)

Fig.5.2 Same as in Fig.5.1,but for Ta- Fig.5.4 Same as in Fig.5.1,but for Ma kada. tsuyama.

result were due to the detailed wind stmcture of 6. Discussion and conclusion the typhoon,such as wind convergence,a lull in According to Konishi and Tsuji(1995)’s the storm and an outburst of westerly.On the result,the conventional method,which com- other hand,the forecast of storm surge based prises the same storm surge model as in Section on meteorological data about12hours before 2and the typhoon model of equations(5),(6), overestimated by O.4-1.5m in the eastem part, and(7)in Section3,simulated the storm surge even if we selected the best track as the for the case of Typhoon9119(see case4in Fig. typhoon track (see Fig。4). It is because the 6)within and accuracy of about50cm。In the forecast win(1model can not express the nar- method,the storm model with parameters to fit rowness of the strong wind or a convergent calculated meteorological elements to observed wind in a Iimited strip.This appears as the elements of a typical station was used。Differ- difference between simulated名o,20km(corre- ences between observation and the calculated spondingly,Rm二25km),which was estimated 16 Konishi,T. Vo1.46,No.1

Peak surge (si皿ulate(1) 5 2ゆ

1 .● 罰 H 〜● φ剛猛 躍 ,1 ●5躍”匿罵 殊 ● ・ぐ。: 彰 昌■ 一呪’” 臼σ’””』 ”oy『 o” ! 薦.潟峯郵風 6養 r = ~ 曳 20 9◆) }」 Dl■ 。.3 7 一 窩 7 (目)4 35 .象-禽◆,ン濤 鑑グ迄 黙 0 属職 . t欄 晶 貫36周喉.躍“、ノ為 蛭.. E. .t a,. ,e ..[r碗臼 .n 09 ● 鴎 ”晶甲 ’’’”一 H.ノ 0イ《 罵 Φヒρ』5の屑dΦ山 3 0∠ 謬 属

ヴ翼躍 Z罵 ノ A ■・.乞 9凸 』貫異 ρ窃 懲疑 ノ 零 竃 、

,.覧....臼甲1 ■、冒9 ・翼 、 U , 、、 O 真._.P 、9 ’ ”』■『”’辱 33 ー - 1凸 ノ16 ■ 》 貿 閥 ㌔鷺 99 / 2 7 嘱冨● ■ 翼㍑へ.民 ・ 覧 ’o ”’ン 、..、 . 丁〔 ’ 一 ー ー ー ー O 剛 6 ◆臥α匹o〇〇 〇〇〇尊 - 『 一 幻襯η翔 ’” 32:0αα 軸 〇〇 000000 ..∠ 臼聾’画’…’ 〇〇 〇〇 ”篭ゴ…” …2 32 2轟馬㌦ウz62 ’餌 2“o翼 藍OO .乞工ααα e r e 、 爲 【 曙 星 ー ‘ 塞8ワ 露 〇〇αO-OO5!0 η瑚即襯刀 05LOO00010。 『 一 曽 一oo 1 oo 戸 / 罵 露 ▼ oo oo m

31 i29 030 13耳 重32 133 聖34 135

02468101214’16182022 Fig.7 Distribution of damage rates of Station number houses in Kyushu,Cyugoku,and Districts,after Mitsuta Fig.6 Same as in Fig。4,but for simulat- and Fujii(1993).The largeststorm ed cases. surge was observed at Onoda, marked as A. Another belt of fmm the analysis of observed dataαfter the strong wind is found in line B。 event,and forecast名o,49km(Rm=58km).As a Dashedlines,AandBareadded. matter of fact,the estimation of70with our by the present author. method also gives a smaller value when the typhoon apProaches Japan.As shown in Sec・ errorS in StOrm Surge fOreCaSt. tion5.1,名o is estimated as39km(Rm=48km)at As shown in Fig.6,the present storm surge l5:00.Mitsuta andFujii(1994)investigatedthe model with the typhoon model in Section3with (1amage rate ofhouses in Kyushu,Chugoku,and parameters to fit calculate(1meteorological Shikoku districts.The rate was defined as the elements to the observed during the high加ind ratio of destroyed houses(tota11y or partially) period of the storm gives good simulated to all households in a given city.The figure is results.On the other hand,from the results of cited in Fig.70f the present study.They show the experimental forecast of Typhoon9119,it that high damage rates of houses are limited was fecognized that,besides improving the within a region of20to40kmfromthetrack of accuracy of typhoon track forecast,we must the typhoon center near the Suoh Sea.Maxi- ta.ke into account the variation of the wind field mum of peak surges,3.1m,in the westem part due to topographic effect or the internal struc- of the Inland Sea was observed at Onoda,vウhich ture of the typhoon in order to raise the accu- is situated in this strip,near the point marked racy of storm surge forecast.In conclusion,the asA(seealsoFig.1).lnFig.7,anotherbeltof most important task,especially for the predic- high damage rates can be fomd in the Iyo Sea, tion of storm surge in a region of complicated which is marked as B in the figure.Thus,there topography like the Inlan(1Sea,is to forecast seem to be two high wind belts in the region. the detailed wind stmcture in the stormyperiod The stations where there is a Iarge difference at the forecast time with a sufficient lead time. between calculation and observation in Fig.4 are situatedbetween these high wind belts.It is Acknowledgment:I would like to thank Mr. regarding such local wind structure leads to K Chubachi,:Kumagaya Local Meteorologica1 1995 An experimental storm surge prediction with application to Typhoon9119 17

Observatory,for his kind suggestion to use his hurricanes,NOAA Tech.Rep.NWS48,Silver regression equations.I also wish to express my Spring,MD,71pp. thanks to Mr.S.Yamada,Forecast Division of Konishi,T.,1989:“Numerical forecast of storm JMA,and ML Y.Fujita,Meteorological Sate1- surges on a real time basis”,066伽og勿hづ6 孤㎎1召伽6,39,21-42. lite Center,for providing the data of past Konishi,T.,1990:Characteristics and forecast of typhoons. storm surges,U初云o So窺(SEA AND SKY), 69,187-202(in Japanese). ノ~碑泥%66s Konishi,T.,1995:“Storm surges in the westem Chubachi,K.,1988:Theposition ofmaximum wind part of the Inland Sea,caused by Typhoon9119 within the distance of200km from typhoon -analysesoftidalrecords,documentsonflood- center,Document of task force in JMA for ing and field inspection”,U吻孟o So昭(SEA analysis of typhoon(mpublished document). AND SKY),70,171-188(in Japanese). Flather,R.A.,1994:A storm surge prediction Konishi,T.and Tsuji,Y.,1995:“Analyses of the model for the northem Bay of Bengal with storm surges in the westem part of the Inland application to the cyclone disaster in April Sea of Japan caused by Typhoon9119”Conti- 1991,ノ1.恥s. Oceanogr.,24,172-190. nental Shelf Res.,15,1795-1823。 Fujita,T.,1952:“Pressure Distribution Within Mitsuta,Y.and T。Fujii,1993:Relation between Typhoon”,060φ勿sJo81〃噂召勿z6,23,437-451. wind speed in typhoon9119(Mireille)and dam- Fukuoka District Meteorologica10bservatory age rate of houses,Annuals Disas.Prev.Res. 1993:Report on Typhoon Kinna(9117)and Inst.,Kyoto Univ.,36,B-1,1-16(in Japanese). Typhoon Mireille(9119),Memoirs of Fukuoka Miyazaki,M.,and T.Ueno,and S.Unoki,1961: Meteorologica10bservatory,48,388pp(in “The theoretical investigations of typhoon Japanese). surges along the Japanese coast (1)”,0α3‘z多zo- Hubbert,G.D.,G.」.Holland,L.M.Leslie and M.」. g名砂hJ6‘zlノ肱‘z8召zJn6,13(1),51-75. Manton,1991:A rea1-time system for forecas- Ueno,M.1994:A monologue consisting of four ting tropical cyclone storm surges,勉砿h67 papers on tropical cyclone motion,JMA/NPD ‘z%4Eo名60偽あ%9,6,86-97. Technical Report,32,28-29. Japan Meteorological Agency,1994:Tide Tables Unoki,S.,1.Isozaki and S.Otsuka1964:Simula- for the Year1995,265. tions of storm surges in Tokyo Bay,Report of Jelesnianski,C.P.,J.Chen,andW.A.Shaffer,1992: the Second District Port Construction Bureau, SLOSH l sea,lake,and overland surges from 125pp.(in Japanese).

台風9119号による瀬戸内海西部における 高潮の予報実験

小西達男 近年わが国へ来襲する台風の個数が増加しており,台風災害,特に高潮災害の発生が懸念されている。実際, 1991年の台風第19号(台風ミレーユ,以下台風9119号とよぶ)は瀬戸内海西部に3mを超す高潮を起こし, その沿岸に大きな浸水被害をもたらした。将来の高潮に備えて,より精度の良い情報を提供するため,高潮予 測のための数値・力学モデルが求められている。本論文では,リアルタイムでの使用を目的とした数値高潮予 測モデルとその利用手法を提案し,台風9119号に適用した場合の予報実験結果と問題点を示す。この過程で開 発された台風モデルのパラメータを定める新しい手法も提案する。