気象砺究所研究報告 第36巷 第2号 61－118頁 昭和60年6月 Papers in Meteorology and Geophysics、ア01． 36，No． 2，PP，61－118．June 1985

A Climatological Study of Typhoon Formation and Typhoon Visit to Japan

by

Takashi Aoki＊

ハ4α¢070Jog1αzlノ～2s2α76h lnsオπ㍑彪，ITsz漉％わα，1わα7α勉，305∫‘功αη

（Received Feb．28，1985；Revised March28，1985）

Abstract

The formation of typhoons in the westem North Pacific and typhoon visits to Japan are investigated from a climatological standpoint．The relationship between the frequency of typhoon formation or typhoon visits and the sea surface temperature in the North Pacific is also studied． First，the formation of typhoons for the30．year period from l953to1982is examined． The average number of typhoons formed is27a year． For the amual var量ation of monthly『 frequency of typhoon formation，the maximum frequency occurs in August and the minimum in February。 Typhoons are formed most frequently in the ocean east of the Philippine Islands．The latitude of typhoon formation moves northward in summer and then retreats equatorward． Secular variation in the frequency of typhoon formation is studied by trend analysis and power spectrum analysis。The frequency showed a peak in the middle of the1960ンs．The frequency of typhoon formation increased until the middle of the1960’s an（i then decreased・ There exist two periodicities of3to4years and6to7years in the secular variation of typhoon formation． In the frequent／infrequeht months of typhoon formation the following characteristics are found． The frequency shows a marked increase north of15。N in the case of frequent typhoon formation．The polar vortex is weak／strong．The zonal index is high／low，because the Aleutian low is active／inactive and the subtropical anticyclone and the intertropical con－ vergence zone exist north／80uth of their normal positions． The negative／positlve anomalies of sea surface temperature are extensive in the subtropical ocean south of Japan。 Next，typhoon visit to Japan is investigated for the perio（i of 70 years from l913to 1982．When a typhoon apProaches Japan and comes within a distance of apProximately300 km from the coast，it is（iesignated as a typhoon that visits Japan。The average amual number of typhoon visits is9． The maximum monthly frequency occurs in August and there are no typhoons from January tσMarch in Japan。 The area of the most frequent visits is the sea south of the Okinawa Islands。Typhoon visits are the least fre〔1uent in the sea northeast of the Hokkaido District。 The∫requencyr of typhoon visits is high in the open sea and straits and is relatively low near the islands an（1momtains．The typhoon season opens負rst near the Nansei Islands，and then the fre－ quency increases also in northern areas。 In the later typhoon season，typhoon visits occur frequently in the sea southeast of the Pacific coast． Secular variation of typhoon visit is examined by trend analysis and power spectrum analysis． The frequency is lower from the latter half of the1920ンs to the1930’s．There exist h三gh frequencies of visit from the latter half of the1940’s to the early1950’s and

＊Present af五liation Japan Meteorological Agency，Tokyo100，Japan。 62 T．Aoki Vo1．36， No．2

around1960。 The results also show the existence of variations of approximately2to2，5 years and 5to6 years periodicities，in addition to a perio（iicity of40 years． In years of frequent／infrequent typhoon formation，typhoons frequently／infrequently visit the coast from the Tokai District to the Kanto District。 The annual variations of regional typhoon visit to Japan are studied by principal com－ ponent analysis． The first four eigenvectors account for97．5％ of the total variance and the pro丘1e of the annual variation in frequency of regional typhoon visits is adequately describe（i by the four eigenvectors． AregionaldivisionofJapanisproposedbyusingtheamplitudecoe伍cientscorrespon（1－ ing to these four eigenvectors． Ten regions are obtained． Broadly speaking，these regions can be divided into four groups． The Pacific Ocean side and the Japan Sea side form one boundary while the Kinki and Shikoku Districts form another． Finally，the correlations between the frequency of typhoon formation or typhoon visits and sea surface temperature are analyse（L The frequency of typhoon formation shows a minimum during El Ni五〇events and maximum frequency is observed two years Iater．There is simultaneous correlation in the sense that as the sea surface temperature in the eastem Equatorial Pacific rises，the fre－ quency of typhoon formation（1ecreases，and vice versa。 There are significant correlations between the number of typhoons formed and the sea surface temperature in the preceding year and two years before． The same relationship hOlds true for typho6n visit． High positive correlations are found for the sea surface tem－ perature in the eastem Equatorial Pacific，while there are negative correlations for the northwestem part of the North Paci丘c． For typhoon visit，there are significant positive correlations with the sea surface temperature of the summer of the preceding year in the ocean south an（i southeast of Japan． To predict the frequency of typhoon formation and typhoon visit，．multiple regression equations are develope（i by hsing sea surface temperatures as independent variables．The multiple correlation coefncients for typhoon formation and typhoon visit are O．854and O．825， respectively． These equations may be useful in long－range forecasting．

Contents

1 Introduction．。． 63 L L Review of Climatological Studies of Typhoon Formation and Typhoon Visit to Japan．．．．．． 63 L2．Purpose of This Study 65 1．3． Data 65 2 Climatology of Typhoon Formation 66 21。Frequency of Typhoon Formation．。．。 66 2．1。1．Distribution of Typhoon Formation 66 2．1．2。Annual Variation of Typhoon Formation．．．．． 69 2．2．Secular Variation of Typhoon Formation 75 2。2．1．Trends in Frequency of Typhoon Formation． 75 2．2．2．Power Spectrum Analysis of Typhoon Formation 76 2．3．Contrastive Study for Frequent and Infrequent Months of Typhoon Formation 77 2．3。L Typhoon Formation in Frequent and Infrequent Months 77 2．3・2・Large－Scale Circulation of the Atmosphere in Frequent and Infrequent Months 79 2．3。3．Cloudiness in Frequent and Infrequent Months．．．． 80 2．3．4。Sea Surface Temperature in Frequent and Infrequent Months 81 2．4．Summary。 83 3 Climatology of Typhoon Visit to Japan 84 3．1，Frequency of Typhoon Visit to Japan．．．．． 84 3．1。1．Distribution of Typhoon Visit to Japan 84 3．L2。Annual Variation of Typhoon Visit to Japan． 86 3．2．Secular Variation of Typhoon Visit to Japan．．．． 90 1985 AClimatologica1StudyofTyphoon 63

3．2。1。Characteristics of the Secular Variation of Typhoon Visit to Japan．．．．．．．．．． 90 3．2。2，Power Spectrum Analysis of Typhoon Visit to Japan ．．，．．．．，．．．．．．．＿．．， 91 3・2・3・Relationship between Frequency of TyphoQn Formation and Typhoon Visit toJapan＿．。99．9．．甲．．．．．．臼．．．9．．．．．噸．り9．9．－9．．．．響．－－．－．．．曾．．．．．．．．．臼．．．9991 3。3。Regional Division of Japan Based on Typhoon Visit．．。．．．．．．．．．．．．．．．．。．．．．．．．． 94 3、3。L Principal Component Analysis of Annual Variations of Regional Typhoon Visit 94 3．3。2，Method of Classi丘cation．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．。．．．．．．．．．．．．．．．．．．．．95

3．3。3．Results of Analysis．．．．．．．．．．．．。．．．．．．．．．．．。．．．．．。．．．．．．．．．．．．．．．．．．．．．．．．． 97

3・4。Summary．．9．．．．．乳．曾．．．9．9．．噸．．．．9．．．．．9．．．．．．曾．．．．．．．9．．．．．．り．．．．．．．．．．．．9．．．曾 99 4。Relationship between Typho6n and Sea Surface Temperature．．．．．．．。．．．．．．．．．．．．．．．．100 4・1・Correlation between Frequency of Typoon Formation and Sea Surface Temperature lOO 4．L l．EI Ni五〇Events and Frequency of Typhoon Formation．．．．．．。．．．．．．．．．．．．．．．100 4・L2・Seasonal Variation of Correlation Pattems for Frequency of Typhoon Formation IOl 4・1・3・Distribution of Correlation Coef五cients for Fre（1uency of Typhoon Formation．．102 4，2。Correlation between Frequency of Typhoon Visits to Japan an（l Sea Surface Temperature．ロ．．．9．．9．9．．99．．．．．D．D．．．．．．9一．曾．9．．．．．．．．．．．．．．6．｝．．．．．9．．．9．9．．！05 4・2・L Seasonal Variation of Correlation Pattems for Frequency of Typhoon Visits 105 4．2，2．Distribution of Correla亡ion Coe伍cients for Frequency『of Typhoon Visits．．．．106 4．3．Multiple Regression Analysis。．．．．．．．．．．．．．．．．．．．．．。．．．．．．．．．．．．．．．．．．．．．．．。．．．108

4．3。1。Possible Independent Variables．．．．．．．．。．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．。．109 4・3・2・Development of Multiple Regression Equations．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．U1 4。4．Summary．．．．．．．り．．．．り9．り．．9．．曾．．．．．．99．．．．9．．9．．．．9．．．．．．．り．．．．．．．．．．．．．．．．．9．ll2

5． Conclusions．．．．．．し．．．．．．．．．．．．．．．9．．．．9．．．．．．．9．．．．9．．．．．．．．．．9．．曾．．．．．．．．．．．．．．．．． 113

Acknowledgements．．．．．．．．．．。．。．．．．．．。．．．．．．．．．．．．．．．。．．．．．．．．．．．．．．．．．．．．．．．．。．．．．．．．115

References．9．．9．．．．．．◎．．．．．．．．．．．．．．．雪．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．99 115

和文概要3．D．．．．9一．．．．．．9．D．．．．．6臼．い．．9C．．9．．9．99－9．．．9．9．．．．．．．．．．．．．9．．．．．．．．9．．9一．．．118

1．Introduction Meteoro1ogical activities in Japan have developed greatly since the beginning of this 1．1Rc漉ωofα伽α孟oZo9εcαZS飢αεε80f century，and meteorological stations have been T〃P尭ooπForηεα孟εoπαπδT〃phooπ 1Vεs甜 established on various islands as far away as 加」αPαn！ Hachijojima，Chichijima，the Nansei，Islands Atyphoonisoneofthemostviolent and Formosa，as shown in Fig．1．Meteoro－ storms in the world．A number of typhoons 10gical data obtained from ships have been are formed in the westem North Paci行c every collected via marine ra（1io since l910．Con－ year，some of which strike Japan．Strong sequently，numerous observational data on winds and heavy rainfalls associated with typhoons have become available． typhoons are a major cause of natural disasters Otani （1940） investigate（l the average in Japan．The damage suffered from typhoons monthly tracks of typhoons。He found that in terms of human an（i property losses is the track of a typhGon is in its westem『most enormous and it is always necessary to take positlon in July，followed by a曲ift eastward． countermeasures to avoid the destructive force He also pointed out that the latitude of recur一・ of a typhoon． vature moves northerly U血til August，retreat－ Since typhoons form over the tropical ing southerly thereafter． oceans，where’meteorological stations to report Arakawa and Tutumi（1944）studied sta一・ on them were often sparse，the detection of tistically typhoon formation and visit to Japan typhoons was a primary problem in the early based on data obtaiPed from l891 to l940． stage of meteorological activities in Japan。 The frequency of typhoon landfalls on the It was difncult to study typhoons in detail in mainlan（l of Japan was examined by Arakawa those days owing to a lack of su缶cient ob－ （1947）。Typhoonlandfalls occurmost frequently servational data． along the east coast of the Shikoku District， 64 T．Aoki Vo1．36ン No．2

statistics concerning typhoons，using Precise datacoIlectedforaperiodofabout30years 〆 ！ from various observational networks． Chen and Ding（1979），Nyoumura（1979） 4Q’ and Nyoumura and Miyazawa（1980）studied typhoon formation，visits and landfalls。Nyou－ 6 mura（1981），Suzuki a血d Miyahira（1982）an（i

‘．騨 0、 ●o Aoki（1983）investigate（1regional secular varia－ ・、 tions of the fre（：luency of typhoon visits。

30’ There were a number of investigations o’ on global observational studies of typhoons and tropical storms by Dunn（1951），Jordan ノ an（i Ho （1962） and Gray （1968，1975，1977， 1979）． Gray studied the characteristics of

120。 130’ 4 typhoons in the westem North Pacific，com－ Fig。1．Meteorological stations in1913。 paring the results with those of other areas。 He showed that about one－third of all tropical next most frequently along the west coast of storms form in the western North Paci倉c， the Kyushu District，fo110we（i by a large making it the area of the greatest number of number of landfalls along the south coast of tropical storms in the world． the Kanto District． Forecasting typhoon formation and track Takahashi（1948），Daidoji（1948），Sakai is a very important aspect of tropical meteo－ （1952）［and Murakami（1953）investigated the rology and an important tool in preventing annual an（l secular variations of the frequency disasters．In order to forecast the movement oftyphoonformation，thepointoftyphoon of a typhoon， the direction and spee（l of formation and the frequency of typhoon land－ movement were studie（1statistically by Bel1 falls on the mainland of Japan． （1963），Shimizu and Shirai（1974），Aoki an（i As mentioned above，the early stage of Nomoto（1977），Aoki（1979）and Nyoumura an（i statistical studies of typhoon formation and Miyazawa（1980）．Nomoto and Okamura（1976） visit to Japan covered the period up to the also presented a review paper on the move一・ 且rst half of the l950’s． Observational data ment of typhoons． supPlie（1by reconnaissance aircraft hαve been The importance of climatological factors、 available since 1949， and，consequently the such as seasonal variations of typhoon tracks amount and quality of typhoon data have in－ was pointed out by Aoki（1979）．In addition， creased greatly． Statistical studies inclu（ling a statistical mode1（the PC Method）for the these data were carried out in the l960フs． prediction of typhoon move卑ents in the Tachi・（1961）studied points of typhoon westem North Pacific was derived by multiple formation an（i the frequency distributions of regression analysis．The PC Method is now typhoons in the westem North Paci丘c using widely used operationally，not only in Japanア dat参from l940to l959。 Tachi（1968）also but also in various other countries in South－ statistically analyze（i typhoon landfalls on east Asia． mainland Japan． Aoki（1981）studie（1the characteristics of¶ Shimizu（1971）investigate（1the frequency typhoon tracks using principal component． distribution of typhoon formation and the analysis，and showed that the丘rst two eigen－ number of typhoons crossing each2。1atitu（1e。 vectors are suf巨cient for the representation of 10ngitude square in the western North Pacmc the typhoon track． He also developed fore＿ for the perio（i from l940to l969． casting equations for the typhoon track． In the latter half of the1970’sシa number Long－range forecasting of typhoon forma－ of studies were done to re－examine previous tion and visits were investigate（l by Hudimoto 1985 A Climatological Study of Typhoon 65

and Neyama（1962），Tokita（1972）and Tashiro are extracted from the“Geophysical Review”， 〈1981）． issued by the Japan Meteorological Agency Lagged correlations between the fre－ （1953－1982）．The frequency of typhoon forma． quency of typhoons and sea surface tempera－ tion during the period from1953to l982is ture in the North Pacific were examined by examined． Yoshinoαα1。（1983），Xieαα1．（1983）an（i Cyclones that originate over the tropical Aoki and Yoshino（1984）．According to their ocean are called tropical cyclones． They are researches，the correlation coe伍cients between classified accor（iing to their intensity as the frequency of typhoons formed and the sea follows： surface temperature of two years prior to （a）tropical depression，with winds up to typhoon formation are signi且cant． 33knots． Variations of a longer－time scale in the （b）’tropical storm，with winds from34to frequency of typhoon formation were（lis－ 63knots． cussed by Takahashi（1962），Hayami and （c）typhoon or hurricane，with winds of Ouchi（1968a，b），Ouchi（1970）and Asakura 64knots or higher． 召印1．（1975）． In spite of these definitions，unless otherwise specified，a ‘‘typhoon”is defined as a tropical 1。2 P郡r皿08εof Thε8S6彫d〃 cyclone with wind speeds exceeding34knots． In the Hrst part of the present study，the （2） Typhoon Visit to Japan following two characteristics of typhoon for－ The frequency of typhoon visits to Japan mation and visit to Japan are studied clima－ during the70years from l913to1982is tologically： studied． Typhoon visits to Japan are exam－ （a） frequency distributions， ined carefully using the follow呈ng materials： （b） annual an（i secular variations of the （a） Atlas of Typhoon Tracks（Central frequencies． Meteorologica10bservatory，1944） The distributions of the pronle of annual （b）Typhoon Summary’of Japan（Japan・ variation of typhoon visits to Japan are also Weather Association，1973） investigate（玉by apPlying Principal component （c） Annual Atlas of Typhoon Tracks（Japan analysis． Regional division of Japan，based Meteorological Agency，1971－1982） on typhoon visits is also studied． （d）Daily Weather Maps（C．M．0．and Secondly，lagged correlations between sea J．M．A．，1913－1982） surface temperature in the North Pacific and （e）Geophysical Review（C．M．0．an（i frequency of typhoon formations and typhoon J．M．A．，1913－1982） visits are examine（1．Lagged correlations up In or（1er to ensure a homogeneous data to two y6ars prior to typhoon formation are set for the investigation of the frequency of analyzed and regression equations are de－ typhoon visits to Japan，the selection of ty－ veloped． Nextンthe usefulness of sea surface phoons is base（1 0n the criteria that the sea temperature data for long－range forecasting level pressure shonl（1be less than l，000mb is shown． or that the surface wind speed should be more On the basis of these investigations，the than10m／s，at least one meteorological station author wil1，in this paper，attempt to clarify in the Japanese Islands． Typhoons are ex－ the climatological characteristics of typhoon cluded from the data set if they do not meet formation and visit to Japan． theSe COnditiOnS． （3） 500mb Geopgtential Height 1．3 0α6α Geopotential heights at500mb are ob－ In this section，the author intends to（1e－ tained from compilations of the Long－range scribe the data used in the study． Forecast Division of the Japan Meteorological （1） Typhoon Formation Agency． Data used in this stud夕consist of The data conceming typhoon formation monthly means for grid points at10。latitude． 66 T．Aoki VoL36，No．2

10ngitude intervals between 20。一80。N and tained winds of a tropical cyclone have at－ 30。E－120。W． tained a speed up to34knots．This definition The primary sources of data for500mb of typhoon formation will apply throughout geopotential heights are the historical weather the study． maps published by the U．S．Weather Bureau for l946to l956，the synoptic weather maps 2．1 17rεg砿επ6“of T“phooπFor配α涜o犯 issued by the Japan Meteorological Agency 2．1．1 Pガsオ7の厩づoπo∫Tッヵhoon Fo7ηzαガoη for l957 to 1963， and data analyze（i by the （1）Typhoons in the Westem North Electronic Computation Center of the Japan Pacific Meteorological Agency，using an operational Atotalnumberof812typhoonswere objective analysis program for the years fol－ observe（i in the westem North Pacific and lowing l964． the average annual fre（luency is27ユ in the （4）Cloudiness 30years from l953to1982． The annual The mean monthly cloudiness in the fre（1uency of typhoon formation varies over Pacific Ocean from satellite observations is a fairly wide range from a maximum of39， obtained from Sadler2オα1．（1976）and Shideler which occurred－in1967，to a minimum of lg an（i Sadler（1979）．The data consist of areal in1969． The standard deviation is5．1， average（1cloudiness on2。50 1atitude－longitude The distribution of typhoon formation is intervalS． investigate（i in each 5◎ 1atitude－longitude Nephanalysis was use（i in（leriving cloud－ square in the westem North Pacific for the iness and is assumed to be roughly equivalent period from l953to l982．When a tropical to the standard World Meteorologica10rgani－ storm or hurricane that originates east of l80。 zation octas of clou（1iness． Square daily moves westward，as some typhoons do on rare values are totaled for each grid．The monthly occasions，it is treated in the statistics as one average was then obtained by dividing the formed in the square from l75。E to180。． total by the number of observations． Figure2 presents the total number of （5） Sea Surface Temperature typhoons formed in each of the5σ1atitude－ Monthly sea surface temperatures are 10ngitude squares in the30－year period．The prepare（l for gri（1 points at 50 1atitude an（i formation of typhoons is frequent in the region longitude intervals in the North Pacific be－ from5。N to250N and from1！0。E to l60。E． tween500N and lO。S，exclu（1ing the area The number of occurrences in the area ac－ south of lO。N and west of180。．・Sea surface counts for84％of all typhoons formed in the temperatures from1949to l976are available westem North Pacific．There are three areas in the published data book of“Monthly・Sea of maximum typhoon formation：（a）the ocean Surface Temperatures in the North Pacific” to the east of the Philippine Islands，（b）the by the Institute of Geography，Academia ocean near the Mariana Islands and（c）the Sinica and the Shanghai Central Meteoro－ ocean in the central and『northern parts of 10gical Observatory （1979）， the South China Sea． Typhoons are formed For additional data，6‘Fishing Information” most frequently in the area between IO。一15。N and“Oceanographic Monthly Summary”com． and l25。一130。E to the east of the Philippine piled by NOAA（1977－1982）were used for the Islands． period l977－1980and l981－1982，respectively． Only six typhoons forme（i during the same period in the area between the equator and 5。N，where the Coriolis parameter is 2． Climatology’of Typkoon Formation nearly zero．This number is Iess than l％ The author intends to clarify the clima－ of the total number of typhoons． tological characteristics of typhoon formation Formation of typhoons is also rare in the in this chapter． The formatioh of a typhoon area north of35。N－only three typhoons in is de負ne（i as the time when maximum sus． the30－year perio（1． In higher latitu（ies，there 1985 AClima．tologicalStudyofTyphoon 67

グ 400N

O

σ 甲 300N

＜7 ＠謬 20。N ○・ ，♂ l OON ・ ，釜 ・ 0 EQ IOO 200 300

FREQUENCY 100

50

lOO。E l IO。E I20。E l30。E I40。E l500E I600E l70。∈： 180。

Fig． 2。 Total number of typhoons formed in each5。1atitude and Iongitude square for the period from1953to1982。ThemarkMindicatestheareaofmaximumfrequencyoftyphoon formation，The thln line in the extreme left of the figure shows the frequency of typhoon formation in the South China Sea． are few warm ocean areas which are neces－ The number of typhoons formed in each sary for typhoon formation and a strong 5。1atitude belt is shown by a heavy solid vertical wind shear often suppresses typhoon line on the left side of Fig．2．Most of the formation． typhoons originate between 10。一20。N．The The number of typhoons formed in each latitude belt of maximum typhoon formation 5010ngitude belt is shown in the low6r part is in the lOo－15。N range，where the mean of Fig．2． As previously mentioned，three occurrence is8．4a year．Nyoumura and Miya－ areas of maximum frequency are identified zawa（1980）presented evidence showing the in the figure： （a）125。一135。E，（b）1400－150。E peak of typhoon formation to be in the15。一 and（c）llO。一120。E． Typhoons rarely form 20。N latitu（ie belt，based on statistics collecte（l east of l70。E． from l951 to 1977，a shorter period than the The small number of typhoons formed time span discussed in this paper． in the Central Pacinc can be attribute（i partly The thin solid line on the left side of to the lower sea surface temperature and Fig．2 refers to the number of typhoons larger vertical wind shear as compared with formed in the South China Sea．Maximum those of the westem North Pacific．Krishna－ typhoon formation arises between l5。一20。N． murti（1971）carried out computations deter－ The anmal mean occurrence is2．l in this mining velocity pqtentials for200mb and latitude belt，which accounts for52％of the gradient levels．These computations represent total number of typhGons formed in the South vertical circulation with rising motion over a China Sea． region of southeast Asia，and a descent over Gray（1968）indicated that most of the the mid－Pacific during the northem summer． initial disturbances from which typhcons later This fact provides another reason for the develop，originate on the poleward side of the lower typhoon formation in the Central Pacific． Equatorial Trough or in the intertropical con一 68 T．Aoki Vol．36，No．2

vergence zone（ITCZ）． Gray（1975），Tanaka form in the westem North PaciHc．Tropical （1980）and others have shown that the Equa－ storms form in the Northem Hemisphere torial Trough extends east－southeastward from about two times as frequently as in the South－ the South China Sea through the Mariana ern Hemisphere． Islands in summer．Therefore，it can be seen No tropical storms form in the South that the area of maximum typhoon formation Atlantic an（1in the eastem South Paci盒c． extends east－southeastward from the South The reason for this seems to be that these China Sea through the ocean east of the areas（10not contain the following necessary PhilipPine Islands． con（1itions of tropical storm formation． （2）Tropical Storms of the World One requirement is a wide ocean with The frequency of typhoon formation is surface temperatures higher than approxi－ ・compared with that of troplcal storms in other mately270C．The sea surface temperature areas．Statistics for tropical storms of the seldom exceeds270C in the eastem South world have been examined by Ramage（1974）， Pacific in summer（January to March）．This Gray（1979）and Mcbride andKeenan（1982）． is one of the most convincing reasons why Re6ults are summarized in Table1． tropical storms do not form in this area． Since tropical storms exceeding a max－ It is said that tropical storms frequently imum sustained wind speed of20－25m／s are form at the polar side of the Equatorial included in Table l，the amual number of Trough，but the Equatorial Trough remains tropical storms in the westem North Pacific north of the equator even in summer in the may appear somewhat lower than the values e皐stem Pacific． The Equatorial Trough in given above． However，this does not affect the Atlantic also rarely moves south of the the general picture of the fre（luencies of e（luator． Thus，a large－scale convergence tropical storms in various regions． area does not exist in the South Atlantic， Table l shows that tropical storms form where the sea surface temperature exceeds most frequently in the westem North PaciHc． 27。C．Therefore tropical storms seldom form Each year，approximately80tropical storms there． occur in the world，and one－third of them

Table1． Annual frequencies of tropical cyclones（maximum sustained winds exceeding 20－25m／s）from l958to1977and their active seasons．Area totals are rounded． （after Ramage，1974；Gray，19791and Mcbride and Keenan，1982）

Location Annual frequency Frequent Season Maximum Month

Westem North PaciHc 26．3 July－Oct． Aug。 Eastern North Pacific 13．4 Aug．一Sept． Sept．

Westem North Atlantic 8．8 Aug．一〇ct． Sept． Northern Indian Ocean 6．4 May－June，Oct．一Nov． Nov

Northem Hemisphere total 54．7 Southern In（lian Ocean 8．4 Jan．一Mar． Feb． Australian Oceans 10．3 Jan．一Mar． Jan． Westem South Paci丘c 5．9 Jan．一Mar． Feb．

Southem Hemisphere totaI 24．5

Global total 79．2 ．1985 AClimatologicaIStudyofTyphoon 69

2・1・2 ・4ηnzεα♂9Vα7iαオion o∫乃ρ九〇〇n 200 Fo7窺磁oη

（1） Frequency of Typhoon Formation by 150

Month ＞QZ回⊃OU鑑」 O O Figure3gives the monthly average fre－ quency of typhoon formation for l953－1982in

・the westem North Paci丘c．、It indicates an 50 August peak fgr typhoon formation，with a

fre（luency of5．6a month．The secon（1high－ O est number of 5．O occurs in September． JF阿AMJJASO卜‘D Fig． 3．Monthly frequency of typhoon formation February is the month that has the minimum in the westem North Paci丘c for the ：number of typhoons formed，with only eight period from1953to1982． Shaded areas in the whole30－year period． show the frequency of typhoon formation The shaded area in Fig．3shows the in the South China Sea． －monthly frequency of typhoon formation in 300 the South China Sea，which is part of the Q＞ N．HEM． 国 200Z westem North Pacific． The amual mean ⊃ o 肛 ’frequency of typhoon formation here is only S．HEM． four，which is approximately l5％of that in IOO the westem North Pacific． N．HE団． J F M A M J J A S O N D O The maximum frequency of typhoon for－ S．HEM， J A S O N D J F M A M J mation in the South China Sea occurs in Fig．4． Monthly frequency of tropical storms in September．Chen and Ding（1979）considered the Northern and Southern Hemispheres reiative to solar year for the period from that the September maximum is（1ue to the 1958to1977．（data a（1apte（i、from Gray， fact that the southwesterly monsoon predom－ 1979） inates in this area until August，starting to weaken in September． August（Gentry，1963）． Annual variation in the frequencies of Table l shows that the northem Indian typhoon formation in various regions of the Ocean（i。e．，the Arabian Sea and the Bay of world were examined and the following Benga1）has two seasons of frequent tropica1， lresults were obtained． storms－from May to June and from October Figure4in（11cates that September and to November． Here，tropical storms rarely 、January show peaks for tropical storm forma． form，from July to September．According to tion in both Hemispheres． January in the Ramage（1974），the sea surface temperature Southem Hemisphere corresponds to July in in the northem Indian Ocean reaches a mini－ the Northem Hemisphere．The tropical storm mum in August when the southwesterly season comes early in the Southem Hemi一 monsoon prevails．This is the reason why ・sphere as compare（1with that of the Northem this region has two seasons of frequent troP－ ．Hemisphere． ical storms，one pre－monsoon and one post－ September has the highest frequency of monsoon． tropical storm formation in the Northem （2）Monthly Distribution of ’Hemisphere，whereas August is the month Frequency of Typhoon Formation that has the greatest number of typhoons in Figure5shows the distribution of the the westem North Paci且c．This is because frequency of typhoon formation in each50 the maximum frequency occurs in September， 1atitude－longitude square according to the ’both in the eastern North Pacific an（i the month．From January to March，three－month westem North Atlantic． Especia11y，in the total frequencies are used due to scarcity of westem North Atlant1c，the fre（1uency in typhoon formation． The characteristics of September is about l。5times as high as in the distrlbution are as follows： 70 T．Aoki Vol．36，No．2

〃 400N ．〆 JAN．～MAR．

300N o 〃 200N σ “ IOON ノ慧＠ ④ O ④ EQ lOOOE 目O。E l20。E l30。E i400E 1500E i600E i70◎E l800

o

4QON APRlL

300N o 〃 20。N o ④ ◎ lOON M

○

E Q 100。E 日OoE ！200E I30。E l400E 1500E l600E l70◎E l800

Fig． 5（a）．Frequency of typhoon formation in each5。1atitude and longitude square for the periodfrom1953to1982。ThemarkMindicatestheareaofmaximumfrequency of typhoon formation．

（a）January to March： Few typhoons Philippine Islands and south of20QN．From form during these three months． Note that April to June，typhoons rarely form in the the frequency in the upper part of Fig．5（a） ocean east of l60。E． is the sum for three months． Typhoons （c）May： The frequency of typhoon formed in the ocean near the Palau Islands， formation increases in the South China Sea．． the Truk Islands and southwest of the Mar－ The maximum frequency occurs in the ocean shall Islan（is． east of the Philippine Islands and another （b）ApriL Most of the typhoon forma－ area of high．frequency exists near the Truk tion is confined to the ocean east of the Islands． 1985 A Climatological Study of Typhoon 71

4 400N MAY

奮 30。N o ∂ 200N

l OON 0 EQ IOOOE H OOE 1200E 1300E 140。E 1500E l600E l700E I800

9’ 400N JUNE

300N ⑤ ¢ ＠ 200N 0 むハ ＠ 気墨 lOON ＠ O

E Q lOOOE l lOOE 1200E 1300E 1400E 1500E l600E I700E l800

Fig．5（b）． Frequency of typhoon formation in each5。1atitu（1e and longitude square for the periodfrom1953tol982。ThemarkMindicatestheareaofmaximumfrequency of typhoon formation．

（d） June：The area of typhoon forma－ widens as compared with June．Many ty－ tion moves northward compared to May．Most phoons form in the two latitude belts of lO。一 typhoon formation is Iocated in the lO。一20。N 15。Nand20。一25。N，excludingtheSouthChina latitude belt． A few typhoons originate near Sea，and there is minimum typhoon formation the Nansei Islands north of25。N． Typhoons in the 15。一20。N latitu（ie belt． rarely form in the ocean east of the Mariana （f）August：This is the month in which Islandls。 typhoons form most frequently．It is charac－ （e） July： The frequency of typhoons teristic of this month that the fre（luency of increases an（1the area of typoon formation typhoon formation is high in the ocean north 72 T．Aoki Vol．36， No．2

4 400N 」ULY

300N

ρ o M M 200N ・㌍ o M ＠ 100N ノ慾 5 O

E Q lOOOE l io。E l200E l300E 1400E 1500E l600E l700E l800， ～

4’

400N AUGU＄T

留 ：300N

200N M ¢ M 5 lOON O

EQ 1000E I iOOE I200E l300E 1400E’ 1500E l600E l700E i800

Fig．5（c）． Frequency of typhoon formation in each5。Iatitude and longitude square for the periodfroml953to1982。ThemarkMindicatedstheareaofmaximumfrequency of typhoon formation． of20。N． Maximum numbers are fomd in frequently in the South China Sea． The area the ocean south and southeast of the Okinawa of the maximum number retreats southeast－ Islands and near the Mariana Islands．The ward except for the South China Sea．An peak frequency occurs in the latitu（ie belt of area of large typhoon formation is located in l50－20。N，where the frequency of typhoon the ocean near the Marlana Islands． formation is lower in July than in the neigh－ （h） October：The frequency of typhoon boring latitude belts．The frequency is small formation decrease in the ocean north of20。N． in the ocean south of l5。N． Approximately80％of the total amount of （g） September： Typhoons form most typhoon formation is concentrate（l in the lati一 1985 AClimatologica1StudyofTyphoon 73

o’

4・OON SEPTEMBER

拶 300N

4 （7 M 200N O M

lO。N 5 O

EQ IOOOE I IOOE I200E l3QOE 1400E 1500E l600E l700E l800

4 400N OCTOBER

300N ① 0 〃 200N 0 仏“ ど M M ρ l OON O 5 ◎

EQ 1000E l l OoE l200E l300E 1400E l500E l600E l700E l800’

Fig． 5（d）． Frequency of typhoon formation in each 5。 latitude and longitu（1e square 、for the periodfrom1953to1982．ThemarkMindicatestheareaofmaximumfrequency of typhoon formation．

tude belt of10。一20。N． east of the PhilipPine Islands． （i） November： The area of typhoon （3）Annual Variation for5。Latitude and formation continues to retreat southwar（1with Longitude Belts the main area of typhoon formation in the Figure6shows the annual variation of 5。一15。N latitude belt． The frequency is also monthly total frequency of typhoon formation large south of15。．N in the South China Sea． in each5。10ngitude belt from l953to1982． （j）December：Typhoon formation is less The following features emerge from Fig．6． frequent and formation areas are scattered． The maximum number occurs in August There are some areas of frequent formation in the125。一135。E longitude belt．September 74 T．Aoki VoL36，No．2

400N NOVEMBER o 300N 0 〃 200N o 《憾 、 書、

lOON 励 O

E Q lOO。E l I OOE l200E 130◎E 1400E 1500E 1600E i700E I800

ρ 400N ．〆 DECEMBER

300N 0 4 200N 0 も幌α lOON M O O EQ 畳OOOE 1卜OOE l200E i300E 1400E 1500E l600E I700E l800

Fig．5（e）． Frequency of typhoon formation in each 5。latitude an（110ngitude square for the periodfroml953tol982。ThemarkMindicatestheareaofmaximumfrequency of typhoon formation． is the month that has the maximum frequency Miyazawa（1980）pointed out that typhoons of occurrence in the 140。一145。E Iongitude may form even in northem latitudes in sum－ belt，west of the Mariana Islands．The peak mer．Inaddition，according to Gray（1975），the of typhoon formation occurs in October in the strong correlation between typhoon formation l55。一160◎E longitude belt． Thus in the area and easterly winds in the upper troposphere east of125。E，the maximum frequency of is well established． typhoon formation occurs later as one goes Figure7shows the amual variation in eastward． the monthly total frequency of typhoon forma－ Chen and Ding．（1979）and Nyoumura and tionat501atitude intervals of the1250－1350E 1985 AClimatologica1StudyofTyphoon 75

JFMAMJJASOND のk．k・ン north－south distribution of typhoon formation ⑤ in July shows a rather broad plateau for the た Iatitudes between10。N an（125。N． In August typhoons can be formed as far north as250－ 30。］N，which is the poleward limit of typhoon formation．After September the Iatitude of typhoon formation retreats equatorward． o の総⑤・ Notice that August is the month that has

O 10 the maximum frequency of typhoon formation oイ↑＼ at any latitude north of l50N． Howeverフfor 1000E I lOoE l200E I300E I400E l500E l600E I700E ！800 lO。一15。N，there is a minimum of typhoon ：Fig．6． Annual variation in the number of ty－ formation in August． phoons formed in each5。Iongitude belt The annual variation in latitude of typhoon for the period from1953to 1982． The formation，oscillating from south in winter to markMindicatesthemaximumfrequency of typhoon formation，while m corre－ north in summer，is consistent with that of sponds to the area of minimum formation． the subtropical ridge at200mb（Krishnamurti αα」．，1983）as shown in Fig．7． Although a 400N few typhoons form north of the subtropical ridge line，it is interesting to note that the

300N annual variation in latitude of typhoon forma－ tion is closely related to the changing of the subtropical ridge． 200N ㌧。 2．2Sε6乱Zα7 7αrεα撹oπof T〃P為oon lO F「oηπα孟ε0π 晃OON ④ 2．2．1 T72η4sプn F72g％2n61y o∫T』ゆhooη

｝ Fo7解αオづ0π EQ ハo Figure8shows secular variation of ty－ Fig。7。 Annual variation in the number of ty－ JF団AMJJASOND phoon formation in the westem North Paci丘c phoons formed in each5。Iatitu（ie belt at during the30years from l953to1982． The 125。一135。E for the period from l953to maximum frequency of typhoon formation for 1982． Dash－dotted line indicates the lati－ one year during this perio（l is39，which tude of the subtropical ridge at200mb． ThemarkMindicatesthemaximumfre． occurs in 1967． The next highest is36in quency of typhoon formation． 1971．The four consective years that saw the most frequent formation were1964to l967． 10ngitude belt during the30－year perio（1from

1953 to 1982． This貴gure also shows the 50 annual Iatitudinal variation of the subtropical ridge as average（1 0ver a ten－year perio（1at 40

200mb along l30。E by Krishunamurtiαα1、 ＞QZ国コG国匡」5 2O O （1983）． In general，this subtropical ridge runs from west to east．Easterly winds prevail to the south of the ridge Iine． I o In the l25。一135。E longitude belt，typhoons

0 form mainly between the5。一15。N during the 1955 1960 1965 i970 i975 1980 o丘season for typhoons． In July，the maxi－ Fig． 8． Time series of annual frequency of ty－ mum number of typhoons formed occurs in phoon formation for the period from1953 the area between200－250N．However，many to1982． The heavy line represents a5－ typhoons still form between100－150N and the year ruming mean． 76 T．Aoki VoL36，N6．2

The frequencies for the respective years are decreasing tendency in long－term variation． 34，32，35an（i39，which greatly exceed the However，the trend correlation coemcient， mean frequency of27．l for the whole30－year based on the（lata for each year，is only－O．07 period．The minimum frequency is l9，which and the coefncientわ is －0．08． Both coef丑一 occurs in l969，and the next lowest is21in cients are nearly zero． This is because the・ 1954，1973，1975and l977． year－to－year variation of typhoon formation The annual frequency of typhoon forma－ in the latter half of the period is much larger tion increased until the middle of the1960’s than that in the且rst half of the period，an（i and decrease（i afterwards as shown in Fig．8． because years of a sma11number of typhoon Linear equations are applie（1to the secular formation appeared in the early years of the variations for the entire 30－year period and latter half of the period，i．e．，1969an（i 1973． for some sub－periods： In ad（1ition， the trend to decrease stoPPe（i toward the end of the1970’s． γ＝α十わX 2．2．2 Poω¢7Sヵ召oオ7麗ηZ z4πα1』yszs o∫ whereγis the annual frequency of typhoon TツρhooηFoηn罐on formation，X is the time in years（subtracting The annual frequency of typhoon forma－ l900）and α and わare constants． Table2 tion shows periodic changes as well as long－ presents the calculate（1values ofαand～》an（l term tren（1s as shown in Fig．8．In order to trendcorrelationcoe伍cientsbetweenXan（iy． clarify the periodicity of the annual frequency The trend correlation coe伍cient and the of typhoon formation，power spectrum analysis coefacient 6 are both nearly zero for the has been done． The computational procedure entire30－year period．Namely，secular varia－ of spectral estimates used is the maximum tion of typhoon formation shows no systematic entropy method develope（1by Hayashi（1977）． trend and their amual frequency remains Results of this analysis are presente（i in steady between about20and40． Fig．9． The spectral peaks are found around In the且rst half of the period（1953－1967）， the3－4years and6－7years periodicities，but the positive value of the coefncientわshows are not very pronounce（1． The large spectral ・an increasing trend in frequency． The trend density of the periodicity of more than30 correlation coemcient is O．79． This value of years may be due to long－term variation which the correlation coemcient is significant at a cannot be qualified with only30years of data． O．1％ 1eve1． Figure10shows the results of spectrum In the latter half of the I period （1968－ analyses for the monthly frequencies of ty－ 1982），the trend of secular variation is some－ phoon formation from July to October for the what complicated．The trend correlation co－ 30－year period． emcient，using the5－year running mean data，is The primary peak of the6years perio－ 、一〇．75and the coefncientわis negative．Thus dicity is predominant in August，which is the frequency of typhoon formation shows a similar to the7years periodicity studied by Xu and Lu（1982）． There is also a sharp Table2． Trend correlation coefncients and spectral peak in the3years perio（1icity in July． linear trend COnStantS Of annual 150 frequency of typhoon formation． 》ヒωZ国O」（E■OU窃 O 5 O O correlation Y＝a十bX period coef匠cient b a

1953－1982 0．02 O．01 26．38 1953－1967 0．79 0．96 －29．79 1955－1967＊ 0．94 0．96 －18．28 0 1968－1982 －0．07 －0．08 31．96 Fig．9． ①50106 4 3YEAR2Power spectrum of annual frequency of 1968－1980＊ －0．75 －0．36 53．54 typhoon formation for the period from ＊， 5－year ruming mean of the frequency 1953tQ1982． 1985 AClimatologicaIStudyofTyphoon 77

50 Table3． List of frequent an（i infrequent months of typhoon formation．

Frequent Infrequent

O Month Year Frequency Month Year Frequency YEAR 2 Aug 1960 1 08809998888July 1953 l12122222 50 Aug 1962 July 1954 AUGUST Sept 1965 July 1956 Aug 1966 1 July 1957 Sept 1966 July 1959 O GO30106 4 5Yε剤R2 Aug 1967 Sept 1973 50 Sept 1967 July 1975 1968 1979 SEPTE岡BER Aug Aug July 1971 Aug 198Q Aug 1978 0 Aug 1981

co3010645YEAR2 8．6 1．7 50 Mean Mean

typhoon formation in this season is4．9per month．A month in which the frequency of 0 30 10 6 4 5 YE《R 2 typhoon formation is eight or more is（1efined Fig．10．Power＄pectra of monthly frequencies as a frequent month．On the other hand，a of typhoon formation for the period month in which two or less typhoons are from1953to1983． formed is deHned as infrequent． As liste（l in Table3，there are eleven In September，there exist two peaks near the frequent rponths and nine infrequent months． 4－5 years an（1 the 10－15y ears perio（iicities． The mean fre（luency of typhoons formed in There are no remarkable features in October． fre（1uent months is8．6and that of infrequent Spectral peaks in July and August coin－ months is1．7．Thus，the ratio of the number cide with those of annual frequency of typhoon of typhoons formed in frequent months to formation shown in Fig．9． This suggests that of infrequent months is about5：1． that the frequencies in July and August make The small dots in Fig．11show where large contributions to the secular variation of typhoons formed in frequent months．The annual frequency． The spectral peaks in spatial distribution of frequent months （heavy』 September are different from those of amual solid lines in Fig．11）is somewhat different frequency，which indicate that monthly fre－ from that of normal months（thin solid lines quency of typhoon formation in September in Fig．11）． has little effect on the secular variation of Above－normal numbers of typhcon forma－ annualfrequency． tion are observe（l in the north of 15。N in frequent months as shown on the left side of 2．3 Con診rαs涜oεS孟麗d妙for Frε9配επ孟 Fig．ll，while small below－normal numbers απd lnfrcg・ε肥撹盟ひη活hs of T』，phoo泥 are found south of 15。N． For規α撹oπ The latitude belt of maximum typhoon 2．3．1 T』yヵんoon Fo7ηzα蕗oη伽F70g麗nオOln4 formation is between l5。一20。N．In this point 1砂69㍑傭ハ40窺hs the profUe of latitudinal distribution is quali－ There were441typhoons formed in July， tatiVely Similar tO that fOr nOrmal mOnt長S． August and September during the30years However，it should be emphasized that the from1953to1982．The normal frequency of number of typhoons formed in this latitude 78 T．Aoki VoL36，No．2

●● ● ● ■ノ ． ．●・ ●． ●● ● ● ● ● 鴨 o ● ● 40●N

● ● ● ● ● ● ● ● ● ● ●● ● ・ ●．● ● ●● C ● ● ． 、 ぜ、 。 ノ ． ● ● ● ●● ● ● ●● φ ● ● ● ．の蕪 ● ● ● 30●N ● ● ●● ● ● ●● ●

●

● ● 200N ●

●

●

I O◎N

E Q IO F R2 EO Q U E5 NO C Y40

『 E5 0

5

lOOoE l『OoE l200E l300E l400E I500E I600E l700E 旧Oo Fig． l1。Location points of typhoon formation for frequent months。Heavy and thin Iines show the frequencies of frequent months and normals．

belt during frequent months is by twice as also noteworthy that the points of typhoon many as that of normal months． formation to the east of l55。E tend to con－ On the other hand the distribution of centrate at a relatively higher latitu（1e， i．e． longitude belts have rather less distinctive between20。N and30。N． features，as shown in the lower part of Fig． Figure12shows points at which typhoons ll．It should be noted that the number of formed in infrequent months．Typhoon forma－ typhoons forme（1in the longitude belts be－ tion in these months is limited to the area tween l350E and1500E near the Mariana between lO。一25。N and ll50E－160。E．By com－ Islands shows a considerable increase．It is paring Fig。12with Fig．11，it may be seen

〃

400N

30。N o焦曝 ●

● ● 20。N 8●

● ● ● ● ●

10。N

EQ lOOOE l IOoE I20。E I300E l400E l500E l600E l700E 1800

Fig． 12．Location points of typhoon formation for infrequent months． 1985 AClimatologicalStudyofTyphoon 79

that typhoons of infrequent months form in east of Siberia，with its center located in lower latitudes than those of frequent months． the eastem part of the Aleutian Islands．The There is no typhoon formation to the east of polar regions to the north of Siberia are Luzon Islan（l an（i to the south of the Nansei covered by positive anomalies．These anomaly Islands，where at other times we frequently pattems imply that cyclones such as the observe the formation of typhoons． Aleutian low are active and the polar vortex 2．3。2 五α7g¢一56α」2α76％」‘痂on o∫ごh6 is weaker than normal． ／1オηzos！）h272 ／n F72（7㍑2nオ αn4 Positive anomalies to the east of Japan 1π．ブ7（9（7㍑θnオ ノレ名onオhs show that the subtropical anticyclone intensi一 The difference in Iarge－scaie circulation 丘ed and moved farther north than normaL pattems between fre（1uent and infrequent As the result of frequent typhoon activities， months was studied with the ai（l of composite negative anomalies prevail in the westem maps． Figure l3shows a composite anomaly North Paci丘c，centering to the south of the map of the 500mb geopotential height in Kyushu District．This is coincident with a eleven months with active typhoon formation． large drop in the850mb geopotential height An elongate（1 negative anomaly area as analyzed by Tanaka（1980，1982）． stretches westward from south of Alaska In months of infreqUent typhoon forma－ across the Kamchatka Peninsula to the south一 tion（Fig．14），the elongated area of negative

十． 20 IO ず巽爺 800N 年 4’ lO 20 O ’， ｝4藤翻

600N． 十 ノ

O r 400，N 昏 十

O 十 20。N O 300E 60。E 900E 1200E 150。E 1800 1500W 1200W

Fig．13。 Composite map of anomalies from normal of500mb geopotential height（meter）for frequent months．

800N 等 弗o ぐ｝タ》o 0．

ヒっ ‘ 〆ノー2十、O 寡 く》o 600N 十房20 搾影じ岡書 ノ ’◎’ ρ I O 400N O ”二1’o’ O 十

0 7D O 2QON o O IO 十 10 O 300E 60。E 900E 1200E 1500E 1800 1500W 120。W

Fig．14． Composite map of anomalies from normal of500mb geopotential height （meter）for infrequent months。 80 T．Aoki VoL36，No．2

anomaly for frequent months tumed into a westerlies are stronger than normal and the positive one，with its center in the westem zonal index is high in frequent months． part of the Bering Sea an（1 Central Siberia． The zonal index define（l as the difference Signincant negative anomalies are found in between zonally （90。E－1700E） averaged polar regions near the Beaufort Sea，in（iicat－ anomalies of the500mb geopotential height ing an intense polar vortex and weak cold air at40。N and 60。N is 9．4for frequent months， outburst． whereas it is－20．6for infrequent months。 It is also noteworthy that negative anom－ The pattem of positive in middle latitudes alies extend eastward from the Yellow Sea and negative in Iower latitudes is associate（1 to the east of Japan，while positive anomalies with upPer easterlies to the south of Japan． are situated in subtropical portions near the This also provides favorable circulation for Philippine Islands．These features show that typhoon formation，and coincides with the the subtropical anticyclone exists south of its fact that a stronger－than－normall upPer normal position in months of infrequent ty－ easterly flow prevails over the subtropical phoon formation． westem North Pacific in months with many In order to show clearly the difference typhoons（Ding and Reiter，1983）． in circulation associated with the change in These characteristics of large－scale at－ frequency of typhoon formation，a difference mospheric circulation are consistent with map was prepared． It was made by sub－ synoptical analyses performed by Tokita tracting the composite anomaly map of the （1972），Tanabe（1979），Tao and Dong（1963）． 500mb geopotential height for infre（luent Matsuoka（1971），He（1978），Tashiro（1981） months from that for frequent months． and Ding and Reiter（1983）． Figure15shows the difference in anomaly 2．3．3 αo㍑01伽2ss伽F72gz69漉αnol obtained in this manner． Positive values ∫n加9麗傭ハ40n疏S indicate an increase in the anomalies of the Cloudiness as （lepicte（i by satellite obser－ 500mb geo口otential height when typhoon vations was analyzed using the same procedure formation is frequent． as for500mb geopotential heights．As stated Well－marked differences are evident in aboveンmost typhoons form on the poleward that negative and positive anomalies are in a side of the Equatorial Trough or the ITCZ line altematively extending from the westem （Gray，1968）。 Although it is not always North Pacific to the Polar region． The necessaryF that the ITCZ and maximum pattern of negative in higher latitudes and cloudiness zone shoul（i coincide， the use of positive in middle latitu（les shows the satellite cloud data is a goo（1 metho（i for

O、 40 60 60 40 800N 岡 ■O 十／ 20Oご 一21三………． 婦 一鴛q O 盆 60。N 十 ワ ノ ．，，r4ρ 拗 凡

’一20 ㌔ ヂ 0 0 400N ’ 十 O O

十 9D 200N 0 燃－20 0 300E 60。E 90。E 1200巨 1500E 1800 1500W 1200W

Fig．15、 Difference in5GO mb geopotential height（meter）． Positive values indiCate inCreaSeS inge・P・tentialheight狐・he孕thetyph・・nf・rmati・nisfrequent・ 1985 AClimatologica1StudyofTyphoon 81 determining the distribution of the ITCZ and 40。N 5 rainfall over the ocean． M 5 Making use of the availability of data for 300N 4 the period from l966through l977，composite 羅 maps of cloudiness were constmcted for six 20。N o of the frequent months and for㌻wo of the め 4 infrequent months． 10。N・ノ暗 For frequent months（Fig．16），one of 5 ① 4 一55 EQ M －the notable features is that the maximum 05 4 cloudiness area， having greater than five lOOS octas，ex亡ends eastward from the Philippine

Islands to the Mariana Islands，and there is 200S 5 2 2 3 4 、 、a broad zone between5。N and259、N exceeding IIOoE l20。E I300E I400E I50。E I600E I700E I8Qo four octas in the Central North Pacific． Fig、17。 Composite map of cloudiness（in octas） Another maximum cloudiness areas develop forinfrequentmonths．ThemarkM over New Guinia and Kalimantan．Minimum indicates the area of maximum cloud． iness，whilemcorrespondstothearea cloudiness（1ess than four octas）extends to the of minimum cloudiness． subtroplcs between25。N and30。N． Figure l7shows a composite map of in the Central Pacific is restricted to a nar＿ ・cloudiness during infrequent months． In the row zone between5。N and IO。N，but intensi一 subtropics，minimum cloudiness in infrequent 丘es to a maximum of greater than丘ve octas． months spreads into a broad zone between The area of maximum cloudiness near New 15。N and30。N replacing the narrow zone Guinea sprea（1s wi（iely and shifts southward． which occurred in the case of frequent An extensive area of maximum cloudiness ls months．A minimum of less than three octas also observed near Karimantan． occurs in the center of this area to the south＿ A difference map of cloudiness between east of Japa．n near 25。N． The equatorial frequent months and infrequent months is nlinimum centering at a．bout 2。S is （iistinct shown in Fig．18．Positive values show a in the Central Pacific， Iarger amount of cloudiness for frequent The nearequatorial maximum cloudiness months than for that of infrequent months． Aregionoflargeincreaseincloudiness extends east－northeastwar（l from the Philippine 5 40。N M 5 Islands． This region corresponds to that of 4 heavy precipitation pointed out by Tanaka 30。N 4 （1980，1982）． Another region of increase in 4 cloudiness is located over the ocean east of 20。N 0 l60。E between l5。N and25。N．A decrease r） M 景 I OON in cloudiness is observed to the south of New 5 4 M■ Guinea and Kalimantan． EQ These cloudiness features imply that the 4 4 ITCZ moves northward and that the west I OOS wind component in the lower Ievel，for ex－

200S ample850mb，tends to be strong in southem 3 2 5 4 IIOoE l20。E I5QoE l400E l500E l600E l700E l800 Asia when typhoons form frequently．These Flg．16． Composite map of clou（1iness（in octas） results agree with those presented by Tanaka forfrequentmonths．ThemarkMin－ （1980， 1982）an（i Pan （1982）． dicates the area of maximum cloudiness， 2．3．4 S2αS㍑7∫‘z62T6nψ召7αオz炉2伽F76g昭nオ while m corresponds to the area of ‘zη4 ∫η∫729π2nオ ハ40nオhs minimum cloudiness． One of the important factors in the forma一 82 T．Aoki Vo1．36，No．2

ness（Fig．16）and a combination of upwelling ＠ 0 400N 十 and vertical mixing of ocean water induced． by typhoons． 500N O Although SST decreases by1。C or so，it

200N sti11satis且es requirement that SST should be o 十 2「 o above27。C，because the normal of SST in IOON 領 1 this region is about29。C during summer． o 9。．∂ There are positive anomalies southeast of EQ

o『 隔 the Kuril Islands． Other positive anomalies も 0 ＿ are extensive across the central and the east－ IO。S O 十 旧。 em North Pacific near20。N． On the other

20。S 、 hand，negative anomalies are significant in the l IO。E 1200E I300E l400E I50。E I600E I700E l800 lower latitudes of the eastern Pacific，espe－ Fig．18． Difference in cloudiness （in octas）． ciallyoffshorePeru． Positive values in（1icate increases in In the case of infrequent months （Fig． cloudlness when typhoon formation is frequent， 20），positive anomalies are extensive from south of Japan to east of the PhilipPine Is－ tion of typhoons is sea surface temperature 1ands． This is possibly due to the result of （hereafter SST）．It is generally considered that abundant insolation an（1a calm sea． To the an increase in SST results in a supply of northeast of these positive anomalies， nega。 warm moist air in the boundary layer of the tive anomalies exten（l eastward from o任shore atmosphere su伍cient to provide favorable Japan to the south of Alaska．Large nega－ conditions for the formation of typhoons．For tive anomalies also exist over th．e Central Pa－ this reason it 玉s interesting to analyze the cific． difference in SST pattems between fre（luent Several features of the difference map of and infrequent months． SST（Fig。21）are noteworthy，as would be Figure19shows a composite anomaly expected from the previously shown figures． map of the SST of frequent months．It should Positive values in Fig．21show aロincrease be emphasize（l that negative SST anomalies in SST when typhoon formation is frequent． are observe（1unexpectedly in the ocean south There are several regions of large positive or of Japan where typhoons are often formed． negative values where i，the anomaly pattems These negative anomalies are the results of in frequent and infrequent months are exactly’ a decrease in insolat圭on due to heavy cloudi一 the oPPosite．

N 倉 ． ． の o ● o o O．00 O．OO へ 500 プ 囁 400 4鯵 300 49 200 蓄 亀 l oo 霧1…箋嚢i毒…，…

EQ りのロ O。25 ．※7925 灘iii欝撃 l oo O．OQ －Q．25＝’＝＿0．25． S 十 1200E l4・Oo I600 1800 1600 1400 1200 1000 800W Fig．19．Composite map of anomalies from norma1（in。C）of sea surface temperature for frequent months． 1985 AClimatologica1StudyofTyphoon 83

N ． 9 9 σρ O．00 O．OO、 500 十 囁 疹 じ 〇．OO 一〇．25 沁 400 ㊧’

300 O．25 9亀 4 0．75 200 ⑫ ．鷺・・251．rr 一〇・25．

l o o α・・運甥25⊂蛤 EQ ⑳，（霧震誌 騨凝・、 禰“懲5・ l oo O．OO O．00 一〇．25’ 一〇．25 S 120。E 140● 160。 i800 1600 140。 120。 ゆoo 80●W Fig．20。 CompQsite map of anomαlies from norma1（in。C）of sea surface temperaturefor infrequent months．

N 50。 箪 ぬ霧。 400 300 、． て 200 “ “ O．0／ l oo O．0ノ ．彊》．⑱ EQ ⑭ 『ご1 ＋ 爾．＝．o・5 OoS 0．0 ＿O，5’

1200E l400 1600 1800 1600 1400 1200 1000 800W Fig．2L Difference in sea surface temperature（in。C）。 Positive values indicate increases in sea surface temperature when the typhoon formation is frequent．

Negative values in the westem North Pa－ South China Sea the maximum frequency cific are directly relate（i to typhoon activities occurs in September． as mentioned above．Positive values in the Frequency is the highest in the ocean east Central Pacific and negative ones in the of the Ph皿ppine Islands．The peak in typhoon Equatorial Pacific suggeミt that the formation formation exists in the Iatitude belt from lO。N of typhoons also is associated with SST in to150N．The belt of maximum frequency is the Pacific at a great distance． The relation－ from200N to25◎N in summer and from5。N ship will be invest1gated ih Chapter4． to15。N in winter．Typhoons tend to form to the north of l5。N in months of frequent 2．4 S混醒配αr〃 typhoon formation，whereas those of infre－ A summary of the results obtained in quqnt months are limited to lower latitudes． this chapter is shown below。 The maximum frequency of typhoon for－ The average annual number of typho6ns mation in the30－year period from 1953to forme（1is27．l in the westem North Pacific． l9820ccurre（i in the mid（ile of the l960’s． The maximum frequency occurs in August In the且rst half of the period（1953－1967），the and the minimum in February，while in the number increased，but then decreased up to 84 T．Aoki VoL36，No．2 the middle of the 1970’s．However，there has been no decreasing tendency in recent ノ years．Periodicitiesof3to4yearsor6to 7years exist，in addition to a longer term variation of more than30years There are important differences between the pattems of the500mb geopotential height， cloudiness and sea surface temperature（SST） in frequent／infrequent months of typhoon formation． 〃 The polar vortex is weak／strong．The 0 subtropical anticyclone exists north／south of its normal position and the Aleutian low is 0 active／inactive． The zonal index in the Far グ East is high／low。The large／small amount of cloudiness exten（is from the PhilipPine Islands to the Mariana Islands．The decrease／ increase of cloudiness existst to the south of New Guinia and Karifnantan．SST in the ocean where typhoon formation is frequent Fig．22． Study area for typhoon visit to Japan． is low／high．Above／below nor血al SST is observe（1in the Central Pacific，whereas the 10ngitude square for the70years from l913 SST along the coast of Peru is low／high． to l982．The following characteristics emerge from Fig．23． Generally speaking，the number of typhoon 3．Climatology of Typhoon Visit to Japan visits is high in southem parts of Japan，Iow The direction and speed of movement of in northern． Typhoons are frequent along typhoons after formation vary widely from the Pacinc coast and infre（luent in the Japan one to another． This produces，as a conse－ Sea and the East China Sea coastal regions． quence，certain regions in Japan which be－ The area with most fre（luent typhoon come highly vulnerable to typhoons during visits is the ocean south of the Okinawa Is－ particular seasons，while other regions are lands．The maximum number of typhoon relatively free from typhoon danger． Thus， visits in the 70－year perio（1is83，0r 1．2per it is important to study the typhoon behavior year in1。latitude－10ngitude square．The area in the Japan area． The objective of this of minimum typhoon visits is the ocean east chapter is to investigate the climatology of of Wakkanai．Only twelve typhoons have typhoon visit to Japan． reache（i that area in70years，or approxima－ The term‘typhoon visit to Japan’is de－ tely One in eVery Six yearS。 fined as the visit of a typhoon to the study Next， detailed characteristics of the dis－ area as shown in Fig．22．The outer bound－ tribution of typhoon visits will be described． aries of the study area lie about 300km off （a） Among the Nansei Islands，typhoon the coast of Japan． visits are most frequent in the seas between the Okinawa Islands and the Sakishima Is－ 3．1 Frε9麗επ6〃of TgphooπV「εs麗謬o 1ands，and also between the Osumi Islands JαPαπ and the Amami Islands．By contrast，the 3．1．1 Z万sオ7貿）z410η o∫ T』y／）hoon i／！sπ ご0 number of typhoon visits is fewer around the 1⑫αn areas of the Sakishima Islands and from the Figure23shows the distribution of total Okinawa Islands to the Amami Islands．It frequency of typhoon visits at each l。latitute一・ should be noted that typhoons stear clear of 1985 A Climatological Study of TyphoDn 85

20 ノ ノ 20 o m ！

40 ・⑫ O m 娩玖

o o 30

o ゲ ～ ㌶ 〃 鴨 M 40 〃 60 ＠ 4、

o 細 60

Fig． 23． Fre（1uency of typhoon visits per1。1atitu（ie and longitude square for the perio（i from1913 to1982．ThemarkMindicatestheareaofmaximumfrequencyoftyphoons，whilem corresponds to the area of minimum frequency。 the islands and pass though open sea． （c） There is an elongated region from （b） Isopleths of the frequency of typhoon the sea south of the Shikoku District through visits in the Kyushu， the Shikoku and the the eastern part of the Shikok：u District and Chugoku Districts run from northeast to the westem part of the Kinki District to southwest． This pattem shows that visits Wakasa Bay，where typhoons pass frequently． are frequent southeast of these areas and The number of typhoon visits is also high infrequent northwest of them． in the area from the Izu Islands to the sea 86 T．Aoki Vo1．36，No．2

east of the Kanto District，going through the 250 southem part of the Kanto District．Typhoons pass here as frequently as in the southern 200 ＞OZU⊃OU匡」 5 0 0 0 part of the Kyushu District． It should be emphasized that the frequency of typhoon visits near Tokyo is much higher than near

Fukuoka． 50 （d）Frequency is relatively low from the Kii Peninsula to the Chubu－District and O JF図AMJJASOND the southem Tohoku District．According to Fig．24．Monthly frequency of typhoon visits to Arakawa（1961），some typhoons（ietour around Japan for the period from l913to1982。 mountains in the Chubu District as they ap－ proach the main island of Japan．There are a number of typhoons whose paths are altered typhoons are September，July and October in in this way．This may be the reason why that order．The four months from July to mountainous regions and their Ieeward sides October are thus the typhoon season in Japan， have few typhoon visits． and apProximately80％ of the total visits （e）There is an area of maximum fre－ occur （1uring this period． By contrast no quency of typhoon visits north of Sado Island typhoons visite（i Japan in January， February in the Sea of Japan．This may be due to and March in these70years．These char－ the fact that regular tracks of typhoons from acteristics are consistent with the frequency Wakasa Bay and from the center of the Sea of pressure patterns of typhoons as classifie（i of Japan meet one another at this area． The by Yoshino and Kai（1974）． regions traversing the Tohoku District，in－ Figure 25 shows the distributions of cluding this maximum area and the sea typhoon visits in 2。 Iatitude－longitude square southeast of the Hokkaido District， have for each month． Figure25presents the fol－ frequent typhoon visits． lowing characteristics of typhoon visits． （f） In a孕d around the Hokkaido District， （a）April：Typhoon visits are Iess fre－ the（1istribution of typhoon visits is affected quent an（i there were only three typhoons by topography similar to that found in the that reached Japan in April during the70－ Chubu Districtl the frequency is low over year period，or one in23years． the island of Hokkaido and high in the seas （b）May：The frequency is somewhat both east an（i west of it． higher in May than in April．Most typhoons 3．ヱ．2 且πη㍑αご『Vαγズαioπo∫Tlyヵんooη『Vズsπ in May visit the region of the Nansei Islands オo∫αρ碗 and the Pacific coast from the Kanto District In or（ier to investigate the annual varia－ to the Kyushu District．There was only one tion of typhoon visits to Japan， the number typhoon that entere（i the Sea of Japan after of typhoons entering the study area shown passing through the East China Sea． It then in Fig．22are totalled on a monthly basis． traversed through the Hokkaido District． When a typhoon visit covered part of two （c）June：Many areas have been hit by consecutive monthsンit was counted as one typhoons more than once in the70－year per－ for each month． iod． The frequency of typhoon visits is The following characteristics of annual high in the region from the Nansei Islands variation of typhoon visits to Japan become to the Kyushu District and also in the region clear from Fig．24．The frequency of typhoon from the Izu Islands to the sea east of the visits is highest in August．A total number Kanto District． of217typhoons were observed during the70－ （d） July：The number of typhoon visits year perio（1，0r3．1typhoons per year． The in JaPan inc「eases raPidly in July．The area months with the next Iargest number of of maximum typhoon visits exists south of 1985 AClimatologica1StudyofTyphoon 87 ，嘉／

〃

ノ 0 APRIL ゲ 瀞召 o 勿 初 勿 。！

ク 0 ○

○ ・σ1〆o o 財

O 肋

〃 0

MAY 5 ”｝69● 5 肋 0 0／ ノ5

0 oノ

ア D 9」 －

δ ＾

肋 づ 用 5 5 10 ノ －、㌧ 細 魂 げ 」UNE 野 ぴ〉” 9ρ

o‘ノ

勿 ’o o グ り －

10 Fig．25（a）． Monthly frequency of typhoon visits per2。latitude and longitude square for theperiodfrom1913to1982．ThemarkMindicatestheareaofmaximum frequencyoftyphoons，whilemcorrespondstotheareaofminimumfrequency． the Okinawa Islands The frequency is high southwest of the Kyushu District．It should from this maximum area to the sea between also be note（i that these of frequent typhoon the Okinawa Islan（1s and the Sakishima Is－ visits extend from southeast to northwest． 1ands． Typhoons also appear frequently in （e） August：As mentioned above，August the seas southeast of the Osumi Islands and is the month in which the frequency of 88 T．Aoki Vol．36，No．2 ね ノ5

ノ0 南 〃

〜．、詑 5

ノ0 JUじビ o〃

～ 。訊 ￥〜 0 ロ ～0

◎ ノ5

v』 20

～δ

勿 づ

ノ・ 加

AUGUST ～δ クo 聯0 彫． の

・． 20

0 40 3 0＼3 翁 齢3・ 0 20 20φ ！

SEPTEMBER ♂ 〃 々｝ ・ρ 肋 20

oノ 加 ・30 ao 《ク ～0 ρ～δ

～δ沢甜 Fig。25（b）． M・nthlyfrequency・ftyph・・nvisitsper201atitudeandl・ngitudesquaref。r theperiodfrom1913tol982．ThemarkMindicatestheareaofmaximum frequency・ftyph・・ns・whilemc・rresp・ndst・thearea・fminimumfrequency． typhoon visits is maximum for the year．The District to the Kanto District． There is a distributionofthefrequencyneartheNansei maximum frequency area in the sea south of Islands is similar to that in July．However， the Shikoku District．The northwar（l elon． the distribution pattern is very different from gated region from this maximum area to the that in July in the reglion from the Shikoku Sea・fJapanis・ne・ftheareas・fhighest 1985 AClimatologica1StudyofTyphoon 89

5 ，0 ノ 加 o

づ 20

ノ OCTOBER ’ 29 5 4 、観

50 ’ 30 。♂ン／ ロ 10 20

加 O づ

ノ

♂ 4 NOVEMBER ρ、

o〃 0 10

め ，嘉／

り

0 ’η 6

・く DECEMBER 噺棚 oノ 初

0 フ 0 －

Fig．25（c）． Monthly frequency of typhoon visits per2。1atitude and longitude square for theperiodfrom1913tol982．ThemarkMindicatestheareaofmaximum frequency of typhoons，while m corresponds to the area of minimum frequency． frequency． There is another frequent area altemating with each other along the Pacific extending from the Izu Islands to the sea coast from the Nansei Islan（is to the Kanto east of the Kanto District． District．The westerlies are weak in August It is interesting to note a high fre（luency and there is a fair chance for typhoons to area is situated next to a low frequency area， move north or northwest aroun（i Japan，which 90 T．Aoki VoL36，No．2

causes the distribution of typhoon visits as Few typhoons apProach the Sea of Japan。 described above． （h）November：The zero isopleth of （f）September：Typhoons are frequent typhoon visits runs from central Japan to in the eas亡em Kyushu，the Shikoku，the Kinki the Kyushu District． No typhoons have ap－ and the Tokai Districts． It should be noted peared north of this arga in the entire 70－ that the sea to the north of Sa（io Island is year period． The area of frequent typhoon also an area of high frequency． Typhoons visits is confined to the sea far south of frequently visit the region from this area to Japan． the sea southeast of Hokkai（10 by passing （i） December：Typhoons rarely strike through the Tohoku District． Japan and thefe have been only three typhoons The appearance of maxirbum frequency in December during the70years analyzed． in this area has a close relationship to the Typhoons approach the region from the jet stream situated over the Tsugaru Straits， Nansei Islands to the sea south of the Kanto according to averaged data of upper winds District． for the25－year perio（1from l956to1980at Sapporo and Akita． The jet stream prevents 3．2 Sεc麗Zαr IVαrεα撹oπof T〃pooπ1Vε8ε孟 typhoons from moving northwar（i by turning 置o JαPαn them eastward． 3．2．1 Chα7ααo万s〃6s o∫S26z4α7田Vα7宛オ20冗 The frequency of typhoon visits to the o∫乃ρhoon sea west of the Kyushu District and to the The total number of typhoon visits for Nansei Islands is relatively low in September． the70years from1913to l982is650．The There is a minimum frequency area in the mean amual frequency of typhoon visits and sea between the Okinawa Islands and the standard deviation are 9．3 an（i 3．2， respec－ Sakishima Islands where typhoons are fre－ tively． The average frequency of typhoon quently observed in July and August． formation is27．l a year． Therefore，one－ （g） October：Frequency is high in the third of the total typhoons formed reach sea to the southeast of the Japanese Islands． Japan． Isopleths of the frequency of typhoon visits Figure26shows the secular variation of show a high concentration along the PaciHc the annual number of typhoon visits to Japan coast．These isopleths show a great differ－ The year of most frequent visits was l950， ence in the frequency between visits to the when the frequency was21．The minimum Paci負c Ocean and to the Japan Sea coast． number is four，which occurred in1926and

30

20

＞OZ一⊃O国窪 O

O 1920 19：50 1940 1950 1960 1970 1980

Fig．26． Time series of annual frequency of typhoon visits to Japan for the period from 1913to l982。 The heavy line represents a5－year running mean． 1985 AClimatologicalStudyofTyphoon 91

1973，The frequency is somewhat low from Table4． Trend correlation coe伍cients and the latter half of the l920’s throughout the linear trend constants of annual entire period of the l930’s． frequency of typhoon visits to Japan。 Thereafter，from the latter half of the pericd COrrelation Y＝a十bX 1940，s to the early 1950，s and also around coefHcient b a

l960there were a number of typhoons which 1959－1973 一〇．61 一〇．48 42．25 reached Japan．Many of them caused great 1973－1982 0．40 0．33 －17．36 damage，such as Typhoon 4709 Kathleen，

Typhoon482110ne and Typhoon5028Jane， IOO 》↑一の面O」匿↑O出の some being very violent and disastrous such as Typhoon 5822 1da （the Kanogawa Ty－ 80 phoon），Typhoon5915Vera（the Isewan Ty－ 60 phoon）and Typhoon6118Nancy（the Second Muroto Typhoon）． 40 Frequency of typhoon visits decreased from the l960’s to the middle of the197αs．How－ 20 ever，in the past ten years， it has increased IO7 5 4 5 YEAR 2 somewhat． Fig． Trend correlation coefncientsare calculate（1 27． Power spectrum of annuaHrequency of typhoon visits to Japan for the period and linear equations are used to describe the from1913to1982． trends of the annual number of typhoon vi－ sits for two periods，1959－1973and l973－1982． in Japan（Hiyama，1972）． The methods used are the same as applied in Aoki（1983）pointed out the existence of a section2．2．The results are shown in Table longer periodic variation of20to30years in 4． the geographical variation of frequency of Coe伍cient b has a large negative value in typhoons in the westem North Pacific． Taka－ the former period（1959－1973），which indicates hashi（1982）found a periodicity of 35 years a decrease in frequency with time．The in secular variation of the frequency of tren（1correlation coefncient is－0．61． This typhoon visits （though his definition differs value exceeds a signi丘cance Ievel of5％． from ours）．The40years periodicity shown Coef丘cient b is positive in the latter period in Fig．27may have some relationship with （1973－1982），which may suggest an increase these long periodic phenomena．From these in the frequency of typhoon visits．However， considerations，it can be said that there may the trend correlation coefncient is O．40，which be secular variations over a long Period of is not significant at a5％1eveL More data time，such as several tens of years，in the must be collected in order to ascertain wheth－ frequency of typhoon visits to Japan． er or not the frequency is increasing with Figure27 also shows spectral peaks， time during the past ten years． though not as sharp，with periodicities be－ 3。2．2 Poz“27Sp26か％η¢ノ1η‘z／ッsJs o∫ tween 2 and 2．5years． This seems to 乃伽ooガV翻オ01伽n coincide with the periodicity of the quasi－ The time series of the annual frequency biennial oscillation in the val・iations of the of typhoon visits are investigated by calculat－ equatorial stratospheric winds． ing spectra based on the70－year data from 3．2．3 R61αオづoηshゆ加オzoo2πF7θg％onの20∫ 1913to1982using the same method applied 丁妙hooηFo襯疏onα磁7「妙hooη in section2．2．Results shown in Fig．27in－ 7翻sオo∫⑫侃 dicate sharp spectral peaks at periodicities of In this part，emphasis will be placed on 5to6years and approximately40years． the frequency of typhoon vis量ts to Japan in The5to6years periodicityP closely coin－ relation to the frequency’of typhoon formation． cides with those of droughts or mild winters As mentioned in section2．2，the frequency 92 T．Aoki VoL36，No．2

of typhoon formation ls highly variable．The the sea to the south of the Shikoku District mean of the top six years（1964，1965，1966， with non－recurving typhoons coming up north． 1967，1971，1974）during the perio（1from l953 from the sea south of the Izu Islands． to l982 is 34。7 typhoons per year． This Looking at the composite map of infre－ group is referre（1to as frequent years． The quent years（Fig．29），the most striking fea－ other six years （1954，1957，1969，1973，1975， ture is the minimum frequency of typhoon 1977）of the lowest frequency of typhoon visits to the southern Kanto District an（i formation is termed infrequent years and the south of the Tokai District．The westerlies mean frequency is20．8． and subtropical anticyclone may have moved Figure28shows a composite map of the farther south in the case of infrequent years． frequency distribution of typhoon visits at This tends to force typhoons to move west－ each2◎ 1atitude－longitude square for fre（1uent ward aroung the Nansei Islands and works years．Two features of Fig．28are note－ against their apProaching the eastern part worthy in comparison with the normal distri－ of Japan． bution shown in Fig．23．First，a maximum Figure30presents a scatter diagram used area in the frequency of typhoon visits to to show the relationship between the fre－ the southeast of the Nansei Islands is located quency of typhoon formation an（i typhoon in a rather northerly position． Secondly夕 visits to Japan． This figure clearly shows a another maximum area is located on the close correlation between frequency of typhoon coast from the Tokai District to the Kanto formation and the number of typhoons visit－ District． ing Japan． The correlation coe伍cient is O．52． These features re且ect the fact that many In Fig．30，there are four points rather typhoons form north of normal and dose to apart from the dense central group of points． Japan in frequent years，as stated in section The fre（luency of typhoon visits is high in 2．3．The maximum area from the Tokai 1959and l960in comparison with that of District to the Kanto District may be due to formation。On the contrary，visit fre（luencies convergence of the tracks of typhoons from are relatively40w in spite of high frequencies

ノ ノ ■ 5

m Io づ 10

〃 o 5 10 10 o〃

15 〆2。 ゆ ・・くソ グ 20

15

Fig．28． Frequency of typhoon visits per2。1atitude and longitude square fol・frequent years of typhoon formation。The mark M indicates the area of maximum frequency of typhoons，while m corresponds to the area of minimum frequency． 1985 AClimatologica1StudyofTyphoon 93

〆 。爵／

飢

づ

”・慣 5 ・、 m

ρo伊

lo恥 旦0 10

Fig．29． Frequency of typhoon visits per2。 1atitude and longitude square for infrequent yearsoftyphoonformation・ThemarkMindicatestheareaofmaximumfrequency of typhoons・while m corresponds to the area of minimum frequency。

20 September an（i October，as is usual for the ヒω一＞ZOO工隻ト ●1960 typhoon season in Japan．Few typhoon tracks・

●1959 stretched linearly westward．Many of them

量● recurve（i northward． Thus a number of ●● l O ● ● ●1967 typhoons apProached Japan． ●●1● 0 ： ●産964 （b） In l960，although the amual number of typhoons formed was near average，many ● typhoons formed in August．The frequency o o 20 30 40 was10in August of that』year，while the TYPHOON FOR図ATION average number in August is5．6in a normaL Fig．30． Comparison between the annual frequen－ year。 The subtropical high in the P3cific cies of typhoon formation and typhoon was situated farther north than in a normal visits to Japan， Dots accompanying the year。Many typhoons formed at higher lati－ number indicate exceptional years． tudes、that is，nearer Japan．The total number・ of formation in1964and1967．The correla－ of typhoons formed north of200N was l3． tion cqefncient will be the Iarge value of O。83 （c） In l964，the subtropical anticyclone in if these four years are ignored． The regres－ the Paci丘c developed well during the summer． sion Iine correspon（1ing to this coe伍cient is A consequence was that there were many also shown in Fig．30． typhoons moving westward， south of the The characteristics of typhoon formation subtropical anticyclone， and few typhoons and typhoon visits in these four exceptional recurvednorthwardtoJapan． Although the years w皿be describe（1as follows： annualfrequencyoftyphoon formation was （a） In l959，Typhoon 5915 Vera （the high，typhoons occurred mainly in late autumn Isewan Typhoon） hit Japan．The annual when typhoons have little effect on Japan．The frequency of typhoon formation was lower frequency of formation in November was the by血ve than in normal years，but14typhoons highest（six）in the30years from！953to l982． formed during the three months of August， （d） In l967，the annual number of ty一 94 T．Aoki VoL36，No．2

phoons formed was the largest in this30－year 30 period．However，there was a strong subtrop－ MEAN ical anticyclone prevailing over Japan and 20 the large majority of typhoons move（1west－ ward south of the subtropical anticyclone and 10 did not reach Japan．In addition，there were many typhoons which formed in the ocean OO far east of Japan．Most of them recurved A M J J A S O N D and moved to the northeast instead of ap－ proaching Japan． STANDARDST F－9O DEVIATIONDE 3．3 Reg’oπαZ jo加εsεoηof JαPαn Bαse4 31．Mean and standard deviation of 0n T〃phoon『Vεsε彦 the monthly frequency of re－ gional typhoon visits per2。 lati－ 3．3．1P吻6ゆαJCo勘on6nオ・4πα1卿so∫ tu（ie and longitu（1e square for ／1π照α♂『Vα7磁煎so∫1～ε9ガonα1 the period from1913to1982． 乃加ooη『V磁 In order to determine the objective and differ greatly with squares in these months。 quantitative characteristics of annual varia－ The standard deviation in September is com－ tions in regional typhoon visits， principal paratively small and frequency of typhoon component analysis is ma（le in this part of visits varies little according to area． thestudy． Using the above－mentioned data，eigen－ Monthly frequencies of typhoon visits for values and their associate（l eigenvectors （ε） each2。latitude and longitude square were are obtained from a symmetric covariance determined by counting the number of ty－ matrix（FF7191），where the prime denotes phoons that passed through each square for the the matrix transpose． Here，F is9by 191 70years from1913to1982．Figure31shows data matrix whoseηth column vector（fπ，π the mean and the standar（i dev圭ation for the ＝1，2，…，191） contains the 9 components monthly frequency of the typhoon visits in representing anomalies of frequencies frqm the square。Because there were no typhoons April to December for theηth square．An from January to March in Japan，annual anomaly is defined as the deviation from the variations are obtained from April to Decem－ mean monthly frequency of the l91squares． ber for each year． Principal component analysis gives the fo1－ Figure31in（1icates that the highest fre－ 10wing relations：fπcan be expressed as the quency of typhoon visits occures in August Iinear combination of the eigenvectors， and the second highest in September． By’ κΣ⇒ 加 負 19 2 1 9 1 M冗＝＝ comparing Figs．31and24，it becomes clear ち ＝葱 み （3．1） that the maximum frequency in August is ， not particularly outstanding in this case． where M is the number of components in The fre（luency in July is also relatively low the column vector（！π）and。4れwill be refer－ as compared with that of Fig．24． red to as the amplitude coe伍cient with the As shown in Fig．25（b），areas where l th eigenvector for the n th square。 ノ舞π typhoons visit frequently in July and August represents the projection of the ηth vector are somewhat limited． This is one of the rea－ on the！th eigenvector and plays a similar sons why mean frequency in July and August role as a coefncient in，for example，Fourier is not large in spite of the high frequency Series repl’esentation． of typhoon visits to Japan as a whole． The variances and cumulative variances Standard deviations as shown in Fig．31 explained by the且rst five eigenvectors， ex－ are large in July，August an（i October． This presse（1in percentages of the total variance， suggests that the frequency of typhoon visits together with the eigenvalues are presented 1985 AClimatologicalStudyofTyphoon 95 in Table5． These figures show that the shows that the pronles oftheannual variationof first four eigenvectors account for about95。7 regional typhoon visits vary mainly in phase． ％ of the total variance an（i this degree of This is because the component of the eigen－ accuracy has been deemed su伍cient for the vector has the same sign for each month． purpose of representing annual variation． ハ4， The component values of the eigenvector are the number of components in Eq．（3．1）has large in July，August and October，implying been reduced to four and we need not go that the frequencies of typhoon visits in these beyond且ve． months are relatively high or low simulta－ Figure32shows the first four eigenvectors neously． obtained by principal component analysis． （b） The second eigenvector describes23．6 The features of the first lfour eigenvectors 96 0f the total variance，having oPPosite signs will be described next．Here，the expression between October and July or August．This of high or low frequency of typhoon visits eigenvector demonstrates that frequencies are is determined in comparsion with the mean high／10w in July and August and are low／ frequency for l91 s（luares for each month high in October．Since the component value shown in Fig．31． of the eigenvector in October is large， the （a） The first eigenvector， explaining eigenvector is important in describing the 52．1％of total variance，identifies the principal regional difference of the frequency of ty－ mode of annual variation．This eigenvector phoon visits for this month。 （c） Thethirdeigenvector mainlydescribes Table5． Statistics of the first five eigenvectors． the regional difぞerence of the frequency of typhoon visits in September． In general，the Cumulative Ordinal Percentage number Eigenvalue variance percentage frequency of typhoon visits in a month has variance a correlation with the frequency of other －⊥∩∠つ」4rD130，86 52．1 52．1 months．All the correlation coefacients are 59．29 23．6 75．7 greater than O。5 except for frequency in 37．04 14．7 90．4 September，which has no significant correla－ 13．17 5．2 95．7 tion with any other month．The maximum 6．74 2．7％ 98．4％ correlation for September is only－0．3．This eigenvector for 14．7％ of the total variance． （d）The fourth eigenvector shows that EIGENVECTOR the number of regional typhoon visits is re1－ I O ー FlRST atively Iarge／small in July and September and small／large in August． This eigenvector accounts for only 5．2％ of total variance， ー 0 － SECOND which may mean the eigenvector has little significance． However， there are some squares where this eigenvector is the most IOITHIRD important factor in illustrating the profile of the annual variation of frequency，as will． be stated later．

O FOURTH 3．3．2 ハ42オhool o∫Clαss麺6α歪oη Much work has been done concerning the・ classi丘cation of meterological£eld and in

ー the determination of climatic regionality by A『v1JJASOND applying principal component analysis（e．g．， Sikka an（i Prasad，19811Gadgil and Sohi，1983； Fig． 32． First four eigenvectors for monthly frequency of regional typhoon vis量ts Kato，1983；and Nomoto （鉱α1．，1983）． How－ toJapan． ever，there have been no studies to date that ap一 96 T．Aoki Vol。36，No．2

Plie（i principal component analysis to meteoro－ the measure of the contribution of the i th 10gical phenomena such as typhoon visit． eigenvector on theブth square in the follow－ First，all of the l91 s（1uares used in the ing manner： previous section are divided into two groups

according to amplitude coe伍cients of the 1～（ど，ブ）＝。42（1，ブ）／Σ！蛋2（」，ブ）×100％． 乞＝1 first and secon（i eigenvectors． The first and secon（1eigenvectors together describe 75．7％ The pro丘1e of annual variation is prima－ of the total variance of annual variation． rily explained by the third or fourth eigen－ Thus，the first group consists of the majority vector in theブth square where1～（オ，ブ）con－ of squares in which the first and／or secon（i cerning the thir（10r fourth eigenvector （i．e．， eigenvector are the main determinants of 1＝30r4）exceeds50％． These squares are the profiles of annual variation． There is a regarded as the minority group。 minority group of s（luares in which．the third As for the majority group of squares and／or the fourth eigenvector play a major chosen in the manner described above， the role． co－ordinate of each square in the／1（1，ゴ）／／1（2， Let the amplitude coe伍cients correspond－ ブ）space is plotted in Fig．33．A small dis－ ing to the first an（i the secon（1 eigenvector tance between two points in this ngure im－ at the ノ●th sqtlare be ／1（1，ブ） and ／1（2，ブ〉， plies similarity of annual varlation in these respectively． Then the variable1）（ブ）is ob－ squares・ tained according to the following formula： To identify clusters， the procedure adopted here involves the following steps． Z）（ブ）＝（z42（1，グ）＋z42（2，ブ））1／2 First，the nuclei of prospective clusters are Since the contribution（1ue to the first and identified by visual inspection． Then，their the second eigenvectors is sm311at the l th nearest neighbour is ad（ie（i and an initial square where P（ゴ）is less than 5．0，such cluster is formed．This procedure was con． squares are exclu（ied from the majority group． tinued until a cluster was formed satisfying Next，we de£ne the variableノ～（ズ，ブ）as the conditions that the nearest neighbour of

20 JAo Jo ● ●● ● JAs ● ●● ●● ＼ ●● ●●●●●● ● ● lO ●● ●●●

● ●●● ● 8

● ． Sloα ●． ． ●●● ●●

●・ 3 ● ● ●● ●8 ● ●● ● ● ●．3．：・蕊・ ● ↑・ ・． ・・ ．38 ● ● ● ● 一 A2 ● ●● ● ●● ●● ● ● JOAs ● ●●●● ● ● ● 一IO ● ● ● ● ● ● ● 8 ● Oj も●

・● ’ 一20 一30 －20 －IO O I O 20 30 AI－ Fig． 33．Plot for the square between the amplitude coe伍cieユts corresponding to the 負rst （／11） and SeCOn（i （ノ12） eigenVeCtOrS． 1985 AClimatologica1StudyofTyphoon 97 every point within a cluster belongs to the positiv60r negative，but smaII，are added same cluster． The six clusters obtained are together into the seventh c玉uster． The other shown in Fig』33． squares belong to the eighth cluster． Squares which belong to the remaining The ninth cluster consists of squares minority group are classi且ed in the following where the sign of the amplitude coe伍cient manner． of the fourth eigenvector is positive． The First，the squares are divided into two remaining squares form the tenth cluster． groups according to the・values of R（3，ゴ） 3。3。3彪sπ1オso∫孟nα1ツsガs ・and」配（4，ブ），corresponding to the third an（i Figure 34 shows the IO regions into fourth eigenvectors，respectively． which Japan is divided according to typhoon In the squares where the third eigenvector visi亡s and the mean fre（luencies of typhoon d・minat旦s，thec・ntributiqn・fthesec・nd visits to the respective regions． Tables6 eigenvector is also relatively large． Next，the and7present the amplitude coefncients an（1 squares where the signs of the amplltude anomalies of the monthly frequency of coe伍cient of the second eigenvector are typhoon visits for each region，respectively． The regions thus obtained are represented in principle by the following combinations Table6． First four amplitude coe伍cients for of letters： for the typhoon season in Japan each region． from July to October，the first Ietters of Region First Second Third Fourth the names of months whose frequency deviations from the means are remarkably JOAs 15。1 一〇．3 2．0 1．7 positive are given in capital letters and thQse JAs 22．4 8．8 －4．6 －2。1 with large negative values in small letters。 JAo 8．9 9．6 －2．4 0．0 The characteristics of the regions given Jo －4．1 9。8 －2．9 0．8 in Fig．34and Tables6and7are as follows： Oj 4．0 －11．7 －1．7 －0．2 （a） 」OAs：・This region coptains the Sjoa －12．1 0．2 －0．5 －0．6 Amami Islands and the sea from southeast SAoj －1．5 2．8 10．2 －0．5 of the Kyushu District to south of the Shi－ Sj －3．1 －4．1 7．7 －0。2 koku District． The perio（i of the season of Sa －3．6 －2．6 3．5 7．3 frequent typhoon visits is long in this region． Aoj －2．7 3．3 5．0 －6．2 This is the only region where the number

Table7． Annual frequency anomalies of regional typhoon visits． No typhoons reach Japan from January to March．

『Reg ion Apr． May June July Aug． Sept． Oct． Nov． Dec．

JOAs 一Q．0 2．Q 1．9 8．4 7．8 3．5 7．9 4．7 0．6 JAs 0．4 1．9 4．2 17．7 14．1 －3．7 4．4 6．2 1．5 JAo 0．7 1．3』 3．9 ！0．7 6．1 －1。4 －3．2 0．6 0．7 Jo －O．1 一1．7 －O．2 4．3 －1．4 －2．2 －9．5 －2．3 －0．5 Oj 0．2 1．7 0．9 －3．1 －1．6 －1．5 11．7 1．0 0．3 Sjoa －0．3 －1．4 －3．3 －7。0 －6．1 －1．3 －6．5 －2．3 －0．5 SAoj －0．3 －1．6 0．6 －3．1 4．2 8，2 －3．6 －1。8 －0．5 Sj －0．3 6 1．9 －6．6 －0．1 6．0 1．4 －2．1 －0．5 Sa －O．3 －0．8 －0．0 －0．9 －6．4 6，6 0．9 －1．3 －0．5 Aoj －0．3 －1．8 1．3 －4．6 5．3 0，6 －4．8 －2．3 －0。5

mean value O．3 1．8 6．7 12．6 27．7 25、4 13．8 2．3 0．5 98 T．Aoki VoL36，No．2

40 20 〆 0 A 図 J J A S O N D 7．

40 0 笥一

20

O A 図 J A O N D ＼、 ＼ A 凹 」 J A S O N DS Sloα

40

20 40 O 20 Sj A図JJASOND o A凹J」AS。ND、Jo ムol Sα 〃 40 SAoj ’ ＼Oj、 oσ 0 〆、 A凹JJASOND 04 40 40 20 20 一一7JAo O 0 ♂ JOAs A図JJASONDA M J J A S O A凹JJASOND o ＿1 丈・飴 40 20

o JAo。。 A 図 J J A S O N D ／＼

40

20

40 0 A M J J A S O N D 20 O A図JJASOND Fig．34． Regional（iivision of Japan by typhoon visit。 The mean frequencies of亡yp互oon visi亡s to regions are also shown．

of typhoon visits for each month from July the 且rst eigenvector has a maximum positive to October is above the monthly average． value and that of the secon（1eigenvector also The first eigenvector contributes greatly to has a large positive value．The amplitude the annual variation in this region． coefacient is negative for the thir（i eigen－ （b） JAs：This region is the sea between vectOr． the Okinawa Islands and the Sakishima Is－ （c） JAo：The Sakishima Islands， the 1ands． Typhoons frequently appear in this southern Kyusllu District，the Osumi Islands region in July and August．The number of and the sea southwest of the Osumi Islands typhoon visits for both months is the highest are included in this region． The number of in Japan，while the frequency is the lowest typhoon visits in this region is highest in in September．The amplitude coe伍cient of August． The frequency in July is the same 1985 AClimatologica1StudyofTyphoon 99

as in September，which is notably above highest in Japan． The amplitude coe伍cient average．The frequency in October is some－ of the thir（i eigenvetor is positive an（l large． what low．The amplitude coe伍clents of the （h）Sj：、The westem Kanto District and first and the second eigenvectors are both the sea southeast of the Hokkaido District positive． The former is slightly smaller than are in this region． The frequency of typhoon the latter． visits is high in September and the pro丘1e （（i） Jo：This region covers the Kyushu of the annual variation lies between the Sjoa District，the midwestem Chugoku District the SAoj．The frequency is，however，some－ and the sea to the west and north of these what higher in October than average． The Districts． The monthly frequency of typhoon amplitude coe伍cient of the third eigenvector visits falls somewhat below the average，ex－ is positive，while that of the second eigen－ cept in July． Another feature is that the vector is a large negative value． mmber of typhoon visits is relatively small （i） Sa：This region covers the TGkai in October． The amplitude coe伍cient cor－ District and the eastem Kii Peninsula． There respon（iing to the secon（i eigenvector takes is a prominent sharp peak i血September an（i the highest positive value， but that of the frequency in August is relatively low．The first eigenvector is negative． amplitude coe伍cient of the fourth eigenvector （e） Oj：The eastem Kanto District and has a large positive value． the wide sea area extending from the south （j） Aoj：This region is restricted to the of the Kii Peninsula to the east of the To－ westem part of the Kinki District．The hoku District make up this region． The frequency is low in July and October．The annual variation of the frequency of typhoon proHle is similar to that of both the Sjoa an（i visits fails to show a peak value in August． SAoj regions in this regard，except that the The frequency is almost constant from Au－ frequency of typhoon visits in August is gust to October．It is important to note that higher than in September．The amplitude the frequency in October is the highest in coe伍cient of the fourth eigenvector has a Japan．Alargenegativevalueoftheampli． large negatiVe ValUe， tude coefacient of the second eigenvector may account for this characteristic。 3．4． S郡η切εαr〃 （f）Sjoa：This region is composed of The following became clear from in－ the northern and central parts of the Chubu vestigations on typhoon visit to Japan． District，the Tohoku District，the Hokkaido The averaged frequency of typhoon vis－ District and the seas west of these Districts its to Japan is 9．3 a year in the 70－year and near the Hokkaido District，excluding period from l913to l982．Since the annual the sea north of Sado Island．The number mean frequency of typhoon formation is27．1， of typhoon visits is below average throughout one－third of the total typhoons apProach the year． The annual variation has a peak Japan． Typhoons・visit Japan most frequently’ in September． The frequency’is especially in August and the months with the next low in July and October．The amplitude most frequent visits are September，July and coefacient of the 丘rst eigenvector is the October in that order． We have had no negatiVe maXimUm。 typhoons from January to March in Japan． （g） SAoj：This region covers the Shi－ Typhoons pass most frequently through koku District，the westem Kii Peninsula，the the sea south of the Okinawa Islands． The eastern Chugoku District and the sea north area with the minimum number of typhoon of Sado Island． The profile of annual varia． visits in Japan is the sea northeast of the tion is similar to that of the region Sjoa Hokkaido District． described above．The number of the visits Around the Nansei Islands，the frequency is，on the whole，higher than in the Sjoa of typhoon visits is high in the open sea region． The frequency in September is the between the islands and is relatively Iow 100 T．Aoki Vol．36，No．2 near the islands． There are two elongated the proHle of the annual variations of typhoon regions especially prone to typhoon visits， visits vary in phase． The second eigenvector that is，from the sea south of the Shikoku describes the inversely proportional relation－ District to Wakasa Bay and from the Izu ship between th3number of typhoon visits Islands to the sea east of the Kanto District in October and that in July or August． The through the southern part of the Kanto Dis－ third eigenvector shows regional differences trict．The number of typhoon visits is com－ in frequency in September． The fourth paratively low in the area from the Kii Pen－ eigenvetor represents a negative correlation insula to the Chubu District and the southern between the frequency in July or September Tohoku District。 and that in August． The frequency is high in the southem Regional division of Japan is propose（1 Kyushu District and aroun（i the Nansei Is－ by using amplitude coe伍cients corresponding 1ands in July． In August， in addition to to the first fOur eigenvectors． Ten regions these regions， the frequency is high near are obtained． Broadly speaking，these regions the Shikoku District and in the sDuthern can be divided into four groups，two groups Kanto District．Typhoons visit the westem comprising the Pacific Ocean side and the part of Japan frequently in September， but Japan Sea side． The Kinki District and few typhoons approach the area around the the Shikoku District form another boundary． Nansei Islands． In October，the fre（luency is high in the sea southeast of the Pacific coast． 4． Relationship between Typhoon and Sea Few typhoons approached Japan from Surface Temperature the latter half of the l920’s to the l930’s． The frequency was high from the latter half Relationships between the anmal fre－ of the 1940’s to the early 1950’s and r also quencies of typhoon formation or typhoon around l960．The number of typhoon visits visits for the perio（1from l953 to l982 an（1 to Japan per year （iecrease（1 from the l960ラs sea surface temperature（hereafter SST）are to the middle of the l970夕s． Recently，the investigated by correlation analysis． In this frequency has increased somewhat．There part of the study夕three－month average（1 are periodicities of 5to6 years an（i of ap－ SST，namely，winter（January to March）， proximately40years as well as2to 2．5 spring（April to June），summer（July to SeP－ years in the power spectrum densities of tember）and autumn（October to December） typhoon visits． are used unless specially noticed． The frequency distributions of regional typhoon visits both in frequent and infre－ ・4．1． Corrε」α蕊oπわc孟ωeεηF「7ε（μeπc〃of quent years of typhoon formation are widely T〃phooπFor1犯α孟εoπαnαSεαS祝rfαcε different． In years of frequent typhoon for－ Tε7πPε7α勘7θ mation，・the area of maximum fre（1uency of 4．1．1EJ亙魏oE∂副S侃4F脚㍑2n6yo∫ typhoon visits is north of normal around the Tニソ／）hooη Foη冗‘zオioη southeast of the Nansei Islands． Another The major episodes of positive SST anom－ maximum is situated over the coast from alies along the equator covering the entire the Tokai District to the Kanto District． eastem and central Pacific，which are termed The Iatter area became the area of minimum EI Nifio events，occur about once every four frequency of typhoon visits in years of in－ years on the average． Bjerknes （1966，1969） frequent typhoon formation． showed that伽ctuations in SST，associated According to principal component analysis with EI Nifio events，occur in conjunction of the profile of annual variations in regional with variation of the Southem Oscillation． typhoon visits to Japan，the 盒rst four eigen－ The primary manifestation of the Southem・ vectors account for 95．7％ of the total Oscmation is a seesaw in sea level pressure variance． The first eigeDvector shows that between the southeast Paci丘c and the region 1985 ，A Climatological Study of Typhoon 101

stretching across the Indian Ocean from MB SOl northem Australia to Africa． The warm episodes in equatorial Paci且c 0 SST are associated with large positive pre－ cipitation anomalies in the central Paci丘c → and negative precipitation anomalies over 。C 一〇．5 much of Indonesia and In（iia（Rasmusson an（l SST Carpenter，1982，1983）． They tend to be ac－ O．O companied by Paci且c／North American pattem of mid－tropospheric geopotential height which O．5 is characterize（i by below－normal height in the North Pacific and the southeastern Unite（1 States，and above－normal height over westem Canada（Wallace and Gutzler，19811Horel and Wallace，1981）．Tanaka（！982）pointed out that the・EI Nifio years of l965and 1976 wereyearsofweakmonsoon． 〔一2》 （一1） （O》 （1） 〔2） Great concem for the relationship be－ Fig．35． Fluctuations of Southern Oscillation tween the frequency of typhoon formation ln（玉ex （toP）， equatorial sea surface and EI Ni五〇events has arisen from the temperature（middle）and frequency results of such studies． In order to clarify of typhoon formation（bottom）dur－ ing the且ve years before and after this relationship，we examined a total of six El Ni釘o events。Warm episode years events from l953，1957，1965，1969，1972 an（1 and preceding an（1following years 1976（Rasmusson and Carpenter，1982）．The are referred to as WEY（O），WEY warm episode years and the preceding and （一1）and WEY（1），respectively． following years will be referred to as WEY （0），WEY（一1）and WEY（1），respectively． The prospects of utilizing information on To obtain a picture of fluctuations during Pacific SST to improve the quality of Iong－ the five years before and after EI Nifio range forecasting of typhoon activities w皿 events， we constructed composites of the be investigate（1in the following section． yearly averaged Southern Oscillation In（lex 4．1．2 S召αsoηα」『V‘z万α蕗on o∫Co7701αあoη based on the Tahiti Darwin pressure differ－ P副27ns∫07F729祝召鰐o∫丁妙hoon ence（top of Fig．35）， the yearly averaged Fo”欄オ10η equatorial SST anomaly in the region of5。N Correlat1on coefncients between the an－ －50S and160。W－100。W（middle of Fig． nual fre（luency of typhoon formation and SST 35）， an（i the annual frequ6ncy of typhoon at each gri（i point at5。latitude and longitude formation（bottom of Fig．35）． interv，als for every season from winter to The most prominent feature of the com－ summer of the same year are computed in posites in Fig．35is that the frequency of the first step． The correlations are also typhoon formation shows a maximum in WEY calculated for the period from winter to au－ （2）and a minimum in WEY（0）．An apparent tumn for the two preceding years． In this correlation is also found in the sense that as way correlations are obtained for eleven the e（luatorial SST increases and the Southem seasons at every grid point． Oscillation In（1ex decreases，the frequency of In Order tO eXamine SeaSOn－tO－SeaSOn Var－ typhoon formation decreases，and vice versa。 iation of correlation between the frequency Although there are other factors such as of typhoon formation and SST，the degree of mid－1atitude circulation features，a significant correlation must be de丘ned first．In this relationship appears to exist between warm case，the number of samples totals30（years）， episodes and frequency of typhoon formation． meaning the degree of freedom is28．The 102 T．Aoki Voi．36、No、2

30

20 ZOFく」崖亡OO」O出配O国O O O O POSITIVE

WIN SPR SUM AUT WIN SPR SUM AUT WIN SPR SU卜4 2YEARS BEFORE PRECEDING YEAR SA図E YEAR

NEGATIVE 一10 Fig．36． Seasonal variations in degree of correlation between frequency of typhoon formation and sea surface temperature in the North Pacific．SPR，SUM， AUT and WIN indicate spring，summer，autumn and winter，respectively．

correlation coe伍cient corresponding to a correlation． The negative correlation be－ significant level of5％is±0．36． The degree tween the frequency of typhoon formation of correlation is defined as the sum of the and SST is not so signincant．There is，how－ positive or negative correlation coefncients ever，a period when the degree of nega－ for the grid points which have an absolute tive correlation becomes larger，from the value of correlation coe伍cient is higher spring to the autumn of two years before than O．36．When the degree of correlation typhcon formation，an（l also in the autumn shows a high positive or negative value，the of the preceding year in addition to every correlation between the frequency of typhoon seasonofthesameyear， formation and SST of the season concemed 4．1．3Z）1sオ肋痂ono∫Co7761面on is high． CO頭6伽オS∫07．F脚πon6ッ0∫ Figure36shows seasonal variation in T二y／）hoon Fo7ηzαあon the degree of correlation，which suggests the A typical example of the distribution of following interesting features． correlation coe伍cients between the frequency’ It should be note（1 that the （1egree of of typhoon formation and SST is shown in correlation between the frequency of typhoon Fig．37． This figure presents the correla－ formation and SST in the same year is lower tion pattern for each significant level of cor－ than is expected． It is the Iagged correla－ relation coe伍cients excee（ling O．1％，1％ an（i tion of two years before typhoon formation 5％ by making use of areas shaded differ－ that shows a high degree of correlation． ently according to density． Positive correlation shows large values The North Pacific is divided into two during the period lfrom the autumn of two regions showing Positive an（i negative years before typhoon formation to the winter correlations for the autumn of two years of the preceding year． Few positive corre－ before typhoon formation as shown in Fig． 1ations are found from the autumn of the 37． Similar results were also presented in preceding year to the winter of the same papers by Wang（1982），Xie6砲1．（1983）and year。 Subsequently，the degree of correla－ Aoki and Yoshino（1984）． tion becomes somewhat large in the typhoon The banded region of positive correla－ season of the year concerned． tion stretches from east to west in the south－ In general，the degree of negative corre－ eastem part of the North Paci丘c，namely lation is much smaller than that of positive the eastem Equatorial Pacific，where the pos 1985 AClimatologica1StudyofTyphoon 103

N o クo 曳 500 唱 ゐ 400 傷 300 の ⑳ 碗も （7 藝麟灘

20。 o 、曳 ＠ loo EQ．難凝・ I oo S 120。E 1400 i600 1800 1600 140。 1200 lOO。 800W 國一・．36一一・。45圏一・．46一一α56囮寺・．36一＋・鴻5國尋・．46一令・．56團＋・．57一

Fig． 37． Distribution of correlation coe伍cients between frequency of typhoon fOrmatiOn and Seα surface temperature（autumn two years before typhoon formation）。

N 、 500 唱 ％ ⑭ ぬ『レ 400 ノ ⑰ 300 ． の ⑫ o ’ 『㌧

200 “ 儀 の I oo！讐， EQ．鰍凝・ ［oo S 1200E I400 1600 1800 1600 1400 1200 Iooo 800W

Fig． 38。Regions of significant positive correlation between frequency of typhoon formation an（1sea surface temperature．Shading indicates areas where the correlation coe伍一 cient excee（1s the criterion of degree of correlation，namely a significant Ievel of 5％at one or more times in the two preceding years before typhoon formation． itive correlation region is wide and the cor－ Plain this fact， the correlation coe伍cients relation coe伍cients are signi丘cant at a O．1％ are examined for all gri（1points，for the two Ievel at．many grid points．The maximum preceding years before typhoon formation． value of the correlation coefacient is O．63 at Figure38shows that the gri（l points of the grid point of10。N and105。W。The neg－ signi昼cant positive correlation，namely those ative correlation region in the northwestem exceeding the criterion of degree of correla－ part of the North Pacific is small，and the tion，are locate（i in the southeastern part of maximum value of the correlation coe伍cient the North Pacific，in most cases the region is lOW。 10。N and 10。S．Other regions sometimes The distribution of the correlation coef五一 contain’gri（1 points of significant positive cients throughout the entire perio（i display correlation，but the correlate（l region， if it characteristics as shown in Fig．37．To ex一 exists，is very small and the occurrence of 104 T．Aoki Vol．36，No．2

N 馳 ク 曳 500 唱 ぬ『9一 400

300 繊嚢 ◎ 1…嬢 『へ

200 o

I oo 縄 ⑧ ＠ EQ にク ひ ⑧ 讐凝・．、 ＠ I oo S 1200E 1400 1600 1800 1600 1400 1200 Iooo 80●W Fig．39． Regions of significant negative correlation between frequency of typhoon formation and sea surface temperature．Shading indicates areas where the correlatioh coef一 丘cient exceeds the criterion of degree of correlation，namely a significant level of 5％at one or more times in the two preceding years before typhoon formation．

100N 匡αの E Q ⑫鯵 100S U配OL田mω窪（国＞偶

l OON Σ⊃の E Q 〃 10。S 物

iOON

↑⊃く E Q，

lO。S l OON ⑫ 亡（国＞OZ『O魁Q切肛巳≡ EQ ） lOOS

IOON 匡氏ω E う Q ＠勘 l OOS の 1800 1600 1400 1200 Iooo 800W 囮＋α36一＋・・45團←α46一＋・，56國燗一

Fig。40，Distributions of correlation coef五cients between frequency of typhoon formation and sea surface temperature in the eastem Equatorial Pacific。SPR，SUM，AUT and WIN indicate spring，summer，autumn and winter，respectively． 1985 AClimatologicalStudyofTyphoon 105

significant positive correlation is temporary． small and short． The gri（i points of significant negative It seems that higher positive coefacients correlation are located in the northwestem both apPear and disapPear earlier in the part of the North Pacific，as shown in Fig． eastem areas．As shown in Fig．40，the area 39． Throughout the entire period the area of higher level coe伍cients at100。W in the of the negatively correlated region is small spring two years before typhoon formation compared with the positive region．There reaches to l70。W in theautumn，andanother are，however，some grid points southeast of area of higher level coe缶cients which ap－ the Kuril Islands and southwest of Midway pears near800W in the summer two years Island which show a high degree of negative before exists at1500W in the winter of the correlation． preceding year． These results seem to point Figure 40 presents the distribution of to the existence of intermittent westward correlation coemcients in the area between propagation of the significant correlation area． lOoN and10。S latitude and l80。and80。W The speed of these shifts is （1ifncult to longitude， as showing the representative calculate accurately，but is estimated roughlyヤ variation of the correlation in the southeastern at70。longitude／2seasons，i．e．about l knot． part of the North Pacific． Positively correlated areas，where the 4．2 00rreJα渉εoπわαルcα2Fregμεπ6〃of correlation coe伍cients are signi丘cant at a5％ T〃phooπ 1Vεsε∫s置o JαPαπαηo『Sεα 1eve1，exten（i over a large area during the S配rfαcεTe〃ψεrα加re season between the summer two years be－ 4．2．！ S2αsoη‘zlγα7宛々07z o∫Co7721α存oη fore typhoon formation and the winter of Pα惚7ns∫07’F吻麗π6フ0∫乃加00π the preceding year． There are some high 四s1オs correlation areas exceeding a signi且cant level Figure41shows the seasonal variation of l．0％．However，the signi丘cant correla－ of correlation between the frequency of ty－ tion（iisapPears in the summer an（i autumn phoon visits and SST in the North Pacific． of the preceding year，and afterward，small The following features of correlation become negatively correlated regions appear． These clear from this figure． negatively correlated regions remain from The degree of correlation is low in the the winter to the summer of the same year． same year，as in the case of typhoon formation． However，correlation coe伍cients are much The degrees of positive correlation are smaller than positive ones， an（i the regions high for the period from the autumn two and periods of signi丘cant correlation are years before to the spring of the preceding

20 Z9トく」国匡匡OO」O国U亡OUO 0 0 0 POSITIVE

WIN SPR SU卜4 AUT WlN SPR SU卜4 AUT WIN SPR SU図 2YEARSBEFORE PRECEDlNG YEAR SAME YEAR

NEGATIVE

－IO

Fig．41． Seasonal variations in degree of correlation between frequency of typhoon visits to Japan and sea surface temperature in the North Paciflc，SPR，SUM，AUT and WIN indicate spring，summer，autumn and winter，respectively・ 106 T．Aoki Vol．36，No．2

year．A peak occures in the winter of the tive correlation in the northwestem part of the preceding year．The maximum level of pos－ North Pacific except for the area slightly itive correlation is lower than that for typhoon northwest of the Hawaiian Islands． formation．However，Fig．41shows that the It is of interest to describe here the dis乙 graph of seasonal visits has a wide base， tribution of correlation coe伍cients between which suggests that high correlation contin－ the frequency of typhoon visits and SST in ues for a longer period． the summer of the preceding year．Figure43 There are no remarkable characteristics shows the pattems of significant positive for negative correlation．However，the de－ correlation over the ocean south an（l southeast grees of negative correlation are large for of Japan．The maximum correlation coefn－ the summer two years before typhoon forma－ cient is O．66，with a significant level exceed－ tion and also for the spring of the prece〔1ing ing O。1％． This is one of the characteristic year」 features of typhoon visits，but there are no such 4．2．2 Pどsオ7必z漉on o∫Co77θ1αオZoπ positive correlations for typhoon formation． Co6．旋1副s∫07F脚㍑2n6ッo∫ Grid points exceeding a significant level Ty加oon’V翻s of5％at one or more times in each of the Figure42shows the distribution of cor－ preceding two years are shown in Figs．44 relation coe伍cients between the frequency and 45．There are many grid points of of typhoon visits and SST in the vレinter of signi貴cant positive correlation of the eastern the prece（1ing year，when the degree of cor－ Equatorial Pacific，similar to the case of．ty－ relation is highest． phoon formation shown in Fig．38．There Areas’of positive correlation are widely are other areas of positive correlation over distribute（l in the southeastern part of the the ocean from south of Alaska to off the North Paci盒c，as shown in Fig．42．The area westem coasts of North America． This is small and the correlation is Iower in the signi且cant correlation persists for more than banded region of positive correlation in the one year，during the period from the spring eastern Equatorial Pacific than that of typhoon two years before typhoon visit to the spring formation shown in Fig．37． Positively cor－ of the preceding year． The locations of grid related areas also exist off the west coast of points of high correlation coe伍cients，how－ North America．There are no regions of nega一 ever，do vary slightly． The positively corre一

N 岱々 500 2ノ 唱

グ グ 驚 ゐ 400 ぴ》

300 o ’ ◎ 『“ 20。 ② o 10。 為 ㊧ ㊥ EQ 〆肱 鋤態・．． ⑫② Ioo S 120。E 1400 160。 ［800 1600 1400 1200 lOO。 800W 團一・。36一一・．45囮＋・。36一＋・45圃＋・46～＋・．56

Fig．42． Distribution of correlation coe伍cients between frequency of typhoon visits to Japan and sea surface temperature（winter of preceding year）． 1985 A Climatological Study of Typhoon 107 含々 N ノ、。 ＼ クρ 、 50● 2ノ 唱 ゐぴツ 400

3QO 『亀 o 〃 200 o 塾踏 Q

100 ノ、

EQ 。」簿 ⑳

10。 無ヤ・ S 1200E 1400 1600 1800 1600 1400 1200 looo 800W

團一・，36一一・。45囮†・．56一卓・．45圃＋・。46一＋・・56屋翻＋・・57－

：Fig． 43。 Distribution of correlation coe伍cients between frequency of typhoon visits to Japan and sea surface temperature（summer of preceding year）．

5 QNo ’ 、 ・② ’⑰” 櫓 ノ』 《 400 修 300 o 『“ 200 ◎ ＠ l oo 為 EQ．墾凝・．、 蓼 I oo S 1200E I400 160。 180。 1600 140。 120。 100。 800W Fig．44。 Regions of significant positive correlation between frequency of typhoon visits to Japan and sea surface temperature． Shading indicates areas where the correlation c・e伍cientexceedsthe“riteri・n・fdegreeofcorrelati・ゴ・namelyasigniHcantlevel、 of5％at one or more times in the two preceding years before typhoon visits． 倉み N ノ 、 500 2〆 蟻 ＠ ゐ8勝 40。

300 ◎ o ’ 『“ 200 4 ⑧ l oo 嘱 EQ 簸o。9『 も亀 ⑱ l oo S 1200E 1400 1600 1800 1600 140。 i200 looo 80。W Fig．45。 Regions of significant negative correlation between frequency of typhoon visits to Japan and sea surface temperature． Shading in（iicates areas where the correlation coef丑cient excee（玉s the criterion of degree of correlation，namely a significant level of5％at one or more times in the two preceding years before typhoon visits。 108 T．Aoki Vol．36，No．2

1ate（i area south and southeast of Japan is 40． There are no seasons in which corre－ temporary，apPearing only in the summer of 1ated areas widen as do those for typhoon the preceding year． formation．However，many grid points are Gri（i points of negative correlation exist highly correlated even at the beginning or mostly in the northwestem part of the North at the end of the perio（1． The areas of signif－ Pacific as shown in Fig．45．The area of iCant COrrelatiOn mOVe WeStWard at a Similar negative correlation is relatively small speed to that of typhoon formation． throughout the period． Signi且cant negative 4．3． 皿乱」‘εpZe Rεgrεssεo泥且πα」〃8εs correlations are foun（i in the spring an（1 the summer，for the two preceding years before As describe（i in sections4．1an（14．2，it typhoon formation． becomes evident that both the frequencies of Figure 46 presents the distribution of typhoon formation and typhoon visits are col・relation coemcients in the eastern Equa－ highly correlated with the SST of the two torial Pacific over the same area as in Fig． preceding years， In this part of the study，

ゆON 匡Lの E Q 〃 ⑳ lOOS ⑰

田匡OL田口 ω配（U＞ q lO◎N ⑰ Σ⊃ω E Q ⑰ 職 lOOS ⑰ 〃 lOON ⑨ 卜⊃く E Q 碓

iOOS

100N ⑫ 匡《国＞ OZ一〇UO国配色Z一） 匡氏ω EQ ⑲

IQOS 愈

lOON ⑦ E Q の IOOS

1800 1600 1400 1200 looo 800W

囮寺α36一＋・・45國手・．46一＋α56

Fig．46。 Distributions of correlation coef五cients between frequency of typhoon visits to Japan and sea surface temperature in the eastem Equatorial Paci丘c．SPR，SUM， AUT and WIN indicate spring，summer，autumn and winter，respectively． 1985 A Climatological Study of Typhoon 1G9

multiple regression analysis is performed・to is the number of indepen（ient variables em－ explore the possib皿ty of Iong－range fore－ ployed in the equation． SST are chosen as casting of the frequencies of typhoon forma－ in（iependent variables and the frequencies of tion an（1typhoon visits using SST of the typhoon formation and typhoon visits are de－ North Pacific． pen（1ent variables． 4。3．1Poss狛161no12カ6no16nオ7α7宛δ12s The皿mber of independent variables Let｝7bethedependentvariablean（IX should not excee（1 the number of observa＿ be the independent variable in the multiple tions，i．e．30（years）in this case． This is in regression analysis． An equation is derived accordance with statistical restrictions．The which can be used to estimate the dependent following conditions are applied to the selec－ variable as a linear combination of indepen（i－ tion of possible in（1ependent variables＝ ent variables： （a） The correlation coemcient is signifi－ cant at a5％ 1evel． γ＝αo十α1×1十α2×2十…十砺X勘 （b）There exist more than four highly』 whereαis the regression coefncient and を correlate（i grid points・in adjacent positions．

Table8、 List of possible independent variables for the equation of annual frequency of typhoon formation．Averaged sea surface temperature on the four grid points is used in the development of the equation．

GRID POINTS SEASON YEAR XXXXXXXXXXXX 30。N，1250E3300N，1300E；25。N，1250E320。N，125。E SUMMER 2YEARS BEFORE 30。N，135。E；30。N，140QE；25QN，135QE325QN，140。E AUTUMN PECEDING YEAR リリ 400N，150。E340。N，155。E340。N，1600E；350N，150。E SUMMER 2YEARS BEFORE 20。N，145。E；20。N，1500E；150N，1450E315。N，150。E SPRING 2YEARS BEFORE ら 25QN，165。E；25QN，170。E；25。N，175。E；25。N，180QE AUTUMN 2YEARS BEFORE 5。N，180。E l E Q，180。E l E Q，175。W三E Q，170。W AUTUMN 2YEARS BEFORE マ 5。S，160。Wl5。S，155。W；10。S，1600Wl10。S，155。W SUMMER PRECEDING YEAR 5。N，155。Wl E Q，155。Wl E Q，150。Wl E Q，145。W WINTER PRECEDING YEAR サ 5。N，145。Wl5。N，140。Wl5。N，135。Wl5QN，1300W AUTUMN 2YEARS BEFORE 100N，120。Wl lO。N，115。Wl10。N，110。W510。N，105。W WINTER PRECEDING YEAR 50N，120。Wl5。N，1150Wl E Q，120。W3E Q，115。W AUTUMN 2YEARS BEFORE ユ 5。S，115。Wl5。S，110。Wl5。S，105。Wl5。S，100。W AUTUMN 2YEARS BEFORE

Table9． List of possible in（1ependent variables for the equation of annual frequency of typhoon visits to Japan。 Averaged sea surface temperature on the four grid points is used in the development of the equation．

GRID POINTS SEASON YEAR XXXXXXXXXX エ エ え250N，135。E320。N，135。E320。N，1400E；15。N，1400E SUMMER PRECEDING YEAR 345678901225。N，1400E325。N，1450E；25。N，150。E三25。N，155。E SUMMER PRECEDING YEAR 40。N，160。E三40。N，165。E340。N，1700E；・35。N，160。E SUMMER 2YEARS BEFORE 50。N，150QW l50。N，1450Wl500N，140。W；500N，1350W SPRING 2YEARS BEFORE 50。N，1300W三45。N，130。W三40。N，1300Wl40。N，125。W AUTUMN 2YEARS BEFORE E Q，180。E l E Q，1750Wl E Q，170。W；E Q，165。W AUTUMN 2YEARS BEFORE E Q，160。W∋E Q，155。Wl E Q，150。Wl E Q，145。W WINTER PRECEDING YEAR 5。S，160。W三50S，155。Wl5。S，150。W三10。S，160。W WINTER PRECEDING YEAR 5。N，1200Wl E Q，120。Wl E Q，115。W35。S，115。W AUTUMN 2YEARS BEFORE 50N，80。Wl E Q，85。W；5。S，85。Wl IO。S，80。W SPRING 2YEARS BEFORE 110 T．Aoki Vol．36，No．2

（c）When conditions（a）and（b）are correlations between in（1ependent variables． satisfied in an area through several successive Tables10and ll show the correlation coef－ seasons，the season having the highest cor－ ficients among possible independent variables relation is selected． which are used in this analysis．The char－ After this selection，four grid points are acteristics of the inter－correlations between chosen in order of correlation for each area possible independent variables for typhoon and the average（1SST on the four grid points formation are obtained as fo110ws． is used as the possible in（1ependent variable． The variables X6，Xg，XIl and X12are Thus，as shown in Tables8and9，twelve SST in the southeastem part of the North an（1 ten possible independent variables are Pacific in the autumn two years before lobtained for the frequencies of typhoon for－ typhoon formation when the degree of posi－ mat圭on and typhoon visits，respectively． tive correlation is the highest． There are Generally speak：ing，there are some inter一 high correlations between these four variables．

Table10． Simple correlation coef五cients between the dependent variable （annual frequency of typhoon formation，TYF） an（i possible independent variables and inter－correlation coe伍cients among possible indepen（lent variables．

TYF XI X2P X3 ×4 ×5 X6 ×7 X8 Xg Xlo Xll X12 TYF 1．00 Xl －0．51 1．00 X2 －0．51 0．41 1．00 ×3 －0．63 0．51 0．48 1．00

×4 －0．52 0．61 0．41 0．40 1．00

×5 －0。53 0．38 0．43 0．37 0．56 1．00 ×6 0．56 －0．39 －0．05 －0．42 －0．34 －0．30 1．OG ×7 0．56 －0．16 －0．10 －0．35 0．11 －0．24 O．53 1．00 ×8 0．53 －0．57 －0．04 －0．47 －0．46 －0．44 0。84 0．45 1．00 Xg 0．56 －0．63 －0．19 －0．41 －0．52 －O．36 0．83 0．32 0．77 1．00 Xlo 0．65 －0．59 －0．29 －0．39 －O．48 －0．38 0．54 0。41 0．67 0。69 1．00 Xll 0．59 －O．59 －0．05 －0．44 －O．48 －0．31 0．83 0．45 0．83 0．84 0．62 1。OO X12 0．49 －0．52 0．09 －O．34 －0．36 －0。18 0．77 0。48 0．75 O．79 0．60 0．90 1．00

Table ll． Simple correlation coef五cients between the depen（ient variable （annual frequency of typhoon visits to Japan，TYV）and possible indepen（ient variables and inter－correla－ tion coef五cients among possible independent variables．

TYV X13 Xユ4 Xl5 ×16 X17 ×18 Xlg X20 ×21 ×22 TYV 1．00 X13 0．50 1．00 ×14 0．63 0．80 1．00 ×15 －0．49 0．03 －0．06 1．00 ×16 0．43 0．12 0．28 －O．20 10∩VOOOO0FD－⊥りQ400086457月1 ×17 0．62 0．24 0．33 －0．54 1．QO Xl8 0．47 0．28 0．47 －0．50 O．57 1．00 ×19 0。55 0．24 0．45 －0．57 0．56 0．87 1．00 ×20 0．59 0．21 0．45 －0．53 0．62 0．76 0．82 1．00 ×21 0。44 0．27 0．34 －0．55 0．48 0．87 0．84 0．76 1．00 ×22 0．50 0．26 0．22 －0．44 0．36 0．58 0．63 0．57 0．73 1．00 1985 AClimatologica1Stu（1yofTyphoon 111

AsimilarsituationholdstrueforvariablesX8 The variable X3，which has second highest and Xlo． correlation，is SST in the ocean southeast of The v景riable X1，which represents SST the Kuril Islands for the summer two years in the East China Sea in the summer two before． The simple correlation coefncient of years before typhoon formation，has a signif－ this variable is －0．63． icant negative correlation with variables X8， For typhoon visits，the highest simple Xg，Xlo and X11，which are SST in the south－ correlation coef潰cient of O．63 is obtained by eastern part of the North Paci丘c in the the variable X14，SST in the ocean near an（1 winter of the preceding year and in the west of the Marcus Island for the summer 、autumn two years before typhoon formation． of the preceding year．The variable X17， Inter－correlations between the possible which has the secon（1highest correlation，is independent variables for typhoon visits have SST off the northwest coast of the United the fo110wing characteristics． States for the autumn two years before High correlations exist within a set of typhoon formation． The simple correlation two groups of variables，X13and X14，0r X18， coefncient is O．62． XIg，X20and X21，which are close in time and In the case of typhoon visits， negative space．The variable X22，SST from off the correlation is not remarkable as state（1pre－ Peruvian coast to Central America，is highly viously，and only one variable which shows correlated with variables X18，XIg，X20and negative correlation is selected as the pos－ X21though there are some differences in sea－ sible independent variable X15．The simple son．This may have been partly due to the correlation coefHcient・・has a comparatively westward movement of highly correlated low・value of・一〇．49． areas mentioned in previous sections． To predict frequencies of typhoon forma－ The variable X17，SST off the northwest tion and typhoon visits， equations are devel－ £oast of the United States in the autumn two oped by apPlying a stepwise regression pro－ years before typhoon：formation， has a high gram using the possible independent variables correlation with variables X18，Xlg and X20， mentioned above．The equations obtained which are SST in the southeastem part of for the frequency of typhoon formation are the North Pacific for the same season． The as follows： variable X15，SST of the ocean southeast of γ∫＝41．50－2．970×2－5．040×4 the Kuril Islands for the summer two years before typhoon formation，is negatively cor－ 十4．963×7十2．277×10 related with variables from X17to X21，SST and for the frequency o：f typhoon visits， 〈）ff the west coast of the Unite（l States an（i the ocean southeast of the North Pacific． γ”＝一125．00十4．270×14－0．9370×15 There is a time lag of one or two seasons． 十1．393×17十〇．526×22． 4．3．2 ∠）ω210ヵ規2nオo∫ハ4㍑1オiヵlo1～6976ss／on Eg躍10ns There are a number of variables which Simple correlation coefncients between have high simple correlation coefncients，but dependent variables and possible independent are not selected as variables for the equa一・ variables are analyze（1．The simple correla－ tions． At least one variable is selected for tion coe伍cients are listed in the leftmost use in the equation from among the highly column in the Correlation matrix shown in inter－correlated variables． Tables IO and l1． The variable X3，which is SST of the The variable Xlo，which has the highest ocean southeast of the Kuril Islands for the simple correlation coe伍cient of O．65 for summer two years before typhoon formation， typhoon formation，is SST in the southeast－ has a high simple correlation coe伍cient for ern part of the North Paci丘c in the winter the丘equency of typhoon formation． This of the year preceding typhoon formation． variable was entered into the equation as a 112 T．Aoki VoL36，No．2 variable at the（1evelopment of the equation Table12． Percent of re（luction for the pre（lictio豆 in the second step． However，when other equation of typhoon formation。 variables were ad（1ed to the equation in the indepen（ient variable P6rcent・freduction following step，a combination of variables 13．2 which exclude（l X3became more effective， XXXX 19．3 an（i the variable X3was removed from the マ 27．4 equation． 13．0 The relative importance of the in（lepe血（1－ ent variables in each equation is examined coe伍cient of determination 72．9％ by percentage of reduction as a measure of contribution to the dependent variable． The coe伍cient of determination（1～2）for the mul－ Table 13． Percent of reduction for the pre（1iction tiple regression equation is given by equation of typhoon visits to Japan． in（iependent variable percentofreduction 1～2二Σ（s娚／sッ）αゐ 乞 XXXX 4 30．7 where1～is the multiple correlation coefacient， 5 12．1 s師is the standar（l deviation of the independ－ 7 16．0 ent variable X乞，sッis the standar（1devaition 9．2 2 of the（1epen（1ent variable 】r，α乞is the regres－ coe租cient of sion coefncient，andηis the simple correla－ determ1natiOn 68．0％ tion coemcient between the dependent an（l independent variables． Thus，the percentage of reduction of the indepen（ient variable X乞 Table 14．Multiple correlation coe伍cients and standard errors of estimation for the is estimated from（sの乞／sッ）α乞η． equation based on development data・ As shown in Table12，the contributions to the （1epen（lent variable are of little differ－ typhoon typ典9・n formation VISlt ence in typhoon formation． On the other multiple correlation hand，Table 13shows that variable X14， coefHcient 0．854 0．825 which is SST in the ocean southeast of Japan for the summer of the preceding year， standar（1error 2．9 1．9 is found to contribute significant information to the frequency of typhoon visits to Japan。 of typhoon formation and typhoon visits of The performance analysis of the regres－ the fo110wing year． sion equations on depen（ient data is given in Table l4．An F－test is used for assessing 4．4 S麗ηηηα｝・〃 statistical significance of the sets of in（1epen（1－ The fre（luency of typhoon formation ent variables． The ratios of mean vari－ shows a minimum during EI Nifio events an（1 ances are calculated from the sample variance a maximum two years after．There is an analysis an（l compared to the limiting value apParent correlation in the sense that as sea of the significant level of1％．The computed． surface temperature in the eastern Equa．一一 variance Tatios are l6．78 an（i l3．27 for 】r∫ toI’ial Pacific increases，fre（luency of typhoon． and，y．夕respectively． The significance test formation decreases，an（i vice versa． is passe（1because these values are lal’ger than The highest lagged correlation exists the limiting value of 3．86。 between frequency’of typhoon formation Thus5the regression equations、are found andseasurfacetemperatureofthe autumn effective． As independent variables in the two years before typhoon formation． There regression equations are SST of the two are positively correlate（l regions in the pre¢eding years，these regression e（luations southeastern part of the North Pacific can be used・for prediction of the frequencies an（i negatiVely COrrelated regiOnS in the 1985 AClimatologicalStudyofTyphoon 113 northwestem part． A positive correlation is months from July to October are the typhoon especially remarkable in the eas亡ern Equa－ season in Japan夕while no typhoons visit torial Pacific． Japan from January to March． The correlation between frequency of Typhoons most frequently pass through typhoon visits and sea surface temperature the sea south of the Okinawa Islands． The in the North Pacific is similar to that for typhoon season opens around the Nansei Is－ typhoon formation． Correlation is high with 1ands and in the southem Kyushu District in sea surface temperature from the autumn July。 In August，the frequency also in－ two years before typhoon for耳nation to the creases near the Shikoku District and．in the preCeding year． southem Kanto District． Typhoons visit the The number of typhoon visits shows a westem part of Japan frequently in September， significantpositive correlation with sea surface but a few typhoons approadh the area around temperature in the ocean south and sou㌻heast， the Nansei Islands。Seゆtember is the month Of Japan in the SUmmer Of the preCeding in which the frequency of typhoon visits is year． There is also high positive correlation at a maximum in the northem part of Japan． with sea surface temperature off the westem In October， typhoon visits frequently occur ・coast of North America．However，this is in the sea southeast of the Pacific coast． not the case with frequency of typhoon for－ A schematic map of fre（luent typhoon mεltion． formation is shown in Fig．47，which gives a Multiple regression equations are deve1－ summary of the foregoing chapters． Ioped by using sea surface temperatures in In the case of frequent typhoon forma－ the North Pacific as independent variables tion the frequency shows a marked increase for both frequencies of typhoon formation north of l50N，while typhoons usually form a．nd typhoon visits．Multiple correlation coef一 most frequently in the ocean east of the 且cients of O．854an（l O．825are obtained， re一 Philippine Islands between10。N and15。N． ・spectively．These equations seem to be useful The Aleutian low is active and the polar for long－range forecasting， vortex north of Siberia is weak．The zonal index which is a measure of the strength of middle latitude westerlies is high．These 5． Conclusions factors provide favorable conditions for a In this study，the author has clarme（1the northward shift of both ITCZ and the posi－ ・characteristics of typhoon formation an（i tion of typhoon formation．This is confirmed typhoon visit to Japan from a climatological by the fact that the area of increased clou（1－ point of view．A summary of the results is iness extends to the subtropics in the west－ presented in the fo110wing paragraphs， ern North Paciflc． The average annual frequency of typhoon Because of the convergence of the tracks formation is27per year in the westem of typhoons from the sea south of the Shi－ North Pacific，one－third of the total typhoons koku District an（1those coming north from formed visiting Japan． Frequencies of the sea south of the Izu Islands， the fre－ typhoon formation and typhoon visits to Japan quency of typhoon visits increases considerably vary widely from year to year． The most on the coast from the Tokai District to the frequent formation was observe（l in the mid－ Kanto District in years of frequent typhoon dle of the l960’s．Typhoons visited most formation． frequently（luring the period from the latter Negative anomalies of SST exist in the half of the l940ヲs to the early l950’s and ocean south of Japan，resulting from decrease（1 ，aroun（i l960． insolation due to heavy cloudiness and a com－ For amual variation of monthly typhoon bination of upwelling、and vertical mixing of aCtiVitieS，maXimUm frequenCy OccurS in ocean．water induced by typhoons， August and minimum in February．The four It should be noted that maximum fre一 ll4 T．Aoki VoL36，No．2

800N ／”” ．9⑲ POSITiVEANOMALIES鶴く⊃一 一 〇F500MBHEIGHT蕊

脅 瓜 〆 ，。 600N 〆 グ ゐ ノ噸…1…………………箋……肇li……室き……1…茎……………嚢……………………1…………i…iliiiiiiilil謙iiiillMAL，El 鵠一

400N リィ OF500MBHEIGHT

INCREASE OF CLOUDINESS 20。N

TYPHOON

EQ 、鵡怒ぐ，．灘難羅…叢灘羅響

i200E 150。E i800 150。W i200W 900W Fig．47．Schematic map of frequent typhoon formation．

80。N 一※アー 窃 P i；i NEGATIVE ANOMAUESミミ……ミ3 ・：泓※ 0F500MB HEIGHT 脅 o グ ρ。 タ 60。N 誤 〃一 ノ 、 軸 、 溌 ，4 POSITIVE ANOMAUES 傍 OF500MB HEIGHT 40。N ：NEGATIVE ANOMAUES F500MB HEIGHT

RM SST 20。N 0 りわ COLD SST （AUTUMN OF TWO YEARS BEFORE） 樽 TYPHOONFORMATION EQ lNCREASE OF CLOUDINESS 」覧 、 曜‘ク 1200E 1500E 180。 150。W 120。W 90。W Fig．48。 Schematic map of infrequent typhoon formation．

quency of typhoon formation occurs in the The polar vortex is intense，but both two years following an EI Ni五〇event． This the Aleutian low and the subtropical an－ is consistent with the positive correlation ticyclone near Japan are weak：．This circula－ between frequency of typhoon formation and tion pattem corresponds to a low index with SST in the autumn two years before typhoon positiveanomaliesof500mb geopotentialin high formation in the eastem Equatorial Pacific． latitudes and negative ones in middle latitudes． The characteristics of infrequent typhoon In the case of a low index pattem，the formation are shown in Fig．48diagonally southwafd（iisplacement of the mid（ile latitu（le oPPosite to those of frequent ones． westerlies leads to similar shifts of the sub一 1985 AClimatologicalStudyofTyphoon 115

tropical anticyclone，ITCZ and the positions of em North Pacific（the PC Method）．Geophys． typhoon formations．The areas of maximum Mag．，38，17－28． cloudiness are observed near Karimantan an（1 （1981）：Empirical Orthogonal Function Ana1－ ysis of the Track of Tropical Storms and． to the south of New Guinea． Typhoons and Its Application to Forecasting． Thus，positive SST anomalies are ex－ Pap．Meteor．Geophys．，32，57－64． tensive in the subtropics as a result of abun＿ （1983）：Geographical and Long－Term Varia－ dant insolation and a calm sea，cause（1by tions in the Westem North Paci丘c Typhoons． decreases in cloudiness and typhoon forma－ Pap．Meteor。Geophys，，34，143－150．（in Japa－ nese with English abstract） tion． and S．Nomoto（1977）：A Statistical Predic－ In accordance with lagged correlation， tion of Tropical Cyclone Position Base（10n SST in the autumn two years before typhoon Persistence and Climatological Factor（the PC formation in the eastem Equatorial Pacific is Method）．J。Meteor．Soc．Japan，55，147－150． below－normal for the case of infrequent and M．M．Yoshino（1984）＝ Relationship be－ tween the Frequency of Typhoon Formations typhoon formation．The frequency of typhoon and Sea Surface Temperature．J．Meteor．Soc． formation shows a minimum in EI Ni五〇 Japan， 62，172－176． years。 Arakawa，H。（1947）： The Areas of Frequent This significant relationship between the Typhoon Landfalls． Kenkyu S6kuho，20，12－ 15． （in Japanese） frequency of typhoon formation and SST （1961）： Topographic Effect of Japanese suggests that the variation of SST has an Archipelago upon the Typhoon Movements． effect on the frequency of typhoon formation Tenki，8，387－388．（in Japanese） through large－scale general circulation of the and K．Tutumi（1944）： Monthly Atlas of atmosphere over a long－term time scale．The Tracks of570typhoon l891－1940．J．Meteor． Soc．Japan，22，268－274． （in Japanese） mechanism of the relationship mentioned in Asakura，T．，M。Matsushita and K．Takahashi this study requires further study． （1975）： The Frequency and the Date of Typhoon Visits．Saigai No Kenkyu，8，54－63． 。46伽oω1248・翻onた一The author wishes （in Japanese） to express his sincere gratitude to Professor Bell，G。J。（1963）：Predicting the Movement of Tropical Cyclones in the Region of the China M．M．Yoshino of the University of Tsukuba， Sea． Proceedings of the Inter－Regional Sem－ for his continuing suggestion an（1 encourage－ inar on Tropical Cyclones in Tokyo，Japan ment throughout the present study．The Meteor．Agency，Tokyo，195－198． author is also grateful to Professors T． Bjerknes，J．（1966）：APossibleresponseofthe Kawamura and T．Nishizawa of the Univer－ atmospheric Hadley Circulation to EquatoriaL Anomalies of Ocean Temperature． Tellus，18， sity of Tsukuba for their helpful advice． 820－829． Frequent and stimulating discussion with Drs． （1969）：Atmospheric Teleconnections from S．Nomoto and H．Tamiya of the Meteoro－ the Equatorial Paci批．Mon．Wea．Rev．，9了， 10gical Research Institute are gratefully ack－ 163－172． nowledge（1．The members of the Forecast Chen，L．S．and Y．H．Ding （1979）＝ Introduction to Typhoon in the Westem Paci且c。Science Research Division of the Meteorogical Re． Publishing Company， Beijing， 10－30。 （in search Institute must be thanked for their Chinese） valuable discussion． Daidoji，S．（1948）：AStudyontheLong－Range The computations in the present study Forecasting of Typhoon． Kenkyu Sokuho，35， were made with the use of the HITAC M－ 31－33． （in Japanese） Ding，Y。H，an（1E。R．Reiter（1983）： Large Scale 200H computer at the Meteorological Research Circulation Affecting the Variability in the Institute． Frequency of Tropical Cyclone Formation over the Northwest Paci最c． Acta Oceanologica 1～4272η62s Sinica，5，561－574． （in Chinese） Aoki，T．（1979）：AStatisticaIPredictionofthe Dunn，G．E．（1951）：Tropical Cyclones． In Com－ Tropical Cyclone Position Based on Persist－ pendium of Meteorology（T．F．Malone，ed．）， ence andP Climαtological Foctor in the West一 Amer．Meteor。Soc．，Boston，887－90ユ． 116 T．Aoki Vol．36，No．2

Gadgil，S．and N．V．Joshi （1983）： Climatic Clus－ Daily Mean Temperature Variations． Geog， ters of the Indian Region． J．ClimatoL，3， Rev．Japan，56，1－16．（in Japanese with Eng－ 47－63． 1ish abstract） Gentry，R．C．（1963）： Historical Survey an（i Cli－ Krishnamurti，T．N．（1971）：Tropical East－West matology of Hurricanes and Tropical Storms． Circulations during the Northem Summer。」． Proceedings of the Inter－Regional Seminar on Atmos．Sci．，28，13424347． Tropical Cyclones in Tokyo，Japan Meteor． ，H．L．Pan，R．Pasch and D．Subrahmanyam Agency，Tokyo，17－29． （1983）：Ihterannual Variab皿ty of the Tropical Gray，W．M．（1968）：Global View of the Origin Motion Field．Florida State University Report of Tropical Disturbances and Storms。Mon． No．83－3，Tallahassee，378pp． Wea．Rev．，96，669－700． Matsuoka，T．（1971）： Relation between Long （！975）： Tropical Cyclone Genesis． Dept。of Period Circulation Anomalies and Typhoon Atmos．Sci．Paper No．234，CQlorado State Formation．Geophys．Mag．，35，153－164． Univ。，Ft．Collins，121pp． Mcbri（1e，J．L．and T．D．Keenan （1982）： Clima－ （1977）： Tropical Cyclone Genesis in the tology of Tropical Cyclone Genesis in the Westem North Paci且c．．J．Meteor．Soc．Japan， Australian Region． J．ClimatL，2，13－33． 55 465－482． Murakami，T．（1953）： On the Birth Place and ン （1979）：Hurricanes l Their Formation，Struc－ Recurving Point of Typhoons． 」。Meteor． ture and Likely Role in the Tropical Circula－ Res．，5，607－612． （in Japanese） tion．In Meteorology over the Tropical Oceans Namias，J．and R．M．Bom （1970）l Temporal （D．B。Shaw，ed．），Roy．Meteor．Soc．，Brack－ Coherence in North Pacific Sea－Surface Tem－ ne11ン 155－218． perature Patterns． J．Geophys．Res．，75，5952－ Hayami，K．and M．Ouchi（1968a）： Long－Term 5955． Variations of Typhoon Tracks． Kisho Kenkyu Neyama，Y．（1963）：A Study of the Long－Range Note，9了，277－281． （in Japanese） Forecasting of Typhoons on Summer． Tenki， and （1968b）：Long－Term Variations of 10，151－158． （in Japanese） Typhoon Fomlations （1）l Relationship with Nomoto，S．and Y．Okamura（1976）：Movement of the Sea Surface Temperature． Kisho Kenkyu Typhoon an（l Its Forecasting． Kisho Kenkyu Note，97，282－285． （in Japanese） Note， 129， 503－548． （in Japanese） Hayashi，Y．（1977）： Space－Time Power Spectral ，M．Chiba and E．Yajima（1983）：An Applica－ Analysis Using the Maximum Entropy Method． tion of Eigenvector Analysis for RegionaI 」．皇Meteor。Soc．Japan，55，415－420・ Division of Climatic Elements．Tenki，30，385－ He，J．H．（1978）： Relationship between Typhoon 392。 （in Japanese） ・Activity in the Westem Paci且c and the Sea Nyoumura，Y．（1979）：Statistics of Amual Fre－ Surface Temperature，Procee（lings of the1978 quency and Minimum pressure of Typhoon． Symposium on the Air－Sea Interaction and Wea．Service．Bu11．，Japan Meteor．Agency， Long－Range Forecasting of the Long Spells of 46，263－270． （in Japanese） Dry and Rainy Weather，Science Publishing （1981）： On the Secular Variation of Typhoon Company，Beijing，72－86。（in Chinese） Existence and Typhoon Pressure and Decrease Hiyama，K．（1972）： Some Features of Recent in the Number of Intense Typhoons Reaching Climatic Change in the World，with Some the Vicinity of Japan。 J．Meteor．Res．，33， Detail for Japan．Manual of Seasonal Weather 303－312． （in Japanese） Forecast呈ng in Japan，Japan Meteor．Agency， （1982）：On the Relation between Tropical Tokyo，47－58， （in Japanese with English Storm Formations an（1Sea Surface Tempera． abstract） ture in the Westem North Paci丘c Ocean．J． Hore1，J。D．and J。M。Wallace（1981）：Planetary、 Meteor．Res．，34，181－184。（in Japanese） Scale Atmospheric Phenomena Associated with and S．Miyazawa （1980）：Some Statistical the Southem Oscillation。Mon．Wea．Rev．， Results of Tropical Cyclones in the Western 109，813－829． Pacific Ocean． J．Meteor．Res。，32，107－133． Hudimoto，S，and Y．Neyama ’（1962）： On the （inJapanese） Similarity between Typhoon Tracks in Spring Otani，T，（1940）：Bofu－u。Iwanami Shoten，Tokyo， and Those in Autumn．Tenki，9，153－159．（in 158pp． （in Japanese） Japanese） Ouchi，M。（1970）：Long－Term Variations of Ty－ Jordan，C・L and T．C。Ho （1962）：Variations in phoon Formations（2）3Relationship with the the Annual Frequency of Tropical Cyclones， sea surface con（iition． Kisho Kenkyu Note， 1886－1958．Mon．Wea．Rev．，90，157－164． 105，344－348．（inJapanese） Kato．H。（1983）：Regionality of Climate in Hok－ Pan，Y．H『．（1982）： The Effect of tlle Thermal kaido Characterize（1by Sunshine Duration an（1 State of Equatorial Eastern Pacific on the 1985 AClimatologicalStudyofTyphoon ll7

Frequency of Typhoon over Westem Paci丘c． abstract） Acta Meteor。Sinica，40，24－34．（in Chinese Tanaka，M．（1980）：Role of the Circulation at the with English abstract） l50mb Level in the Winter and Summer Mon． Ramage，C．S．（1974）：Monsoonal Influences on the soon in the Asian and Australia Regions． Annual Variation of Tropical Cyclone Develop－ Climatological Notes，Univ．of Tsukuba，26， ment over the Indian and PaciHc Oceans．Mon． 1－134． Wea。Rev．，102，745－753． （1982）： Interannual Fluctuations of the Trop－ lRasmusson，E。M。 and T．H。Carpenter（1982）： ical Easterly Jet and the Summer Monsoon Variations in Tropical Sea Surface Tempera－ in the Asian Region．J．Meteor。Soc．Japan． ture and Surface Wind Fields Associated with 60，865－875． the Southem Oscillation／EI Ni五〇．Mon．Wea． Tao，S．Y．and K．Q．Dong （1963）： Relationship Rev．，110，354－384． between the Frequency of Typhoons in the and （1983）：The Relationship between Westem Paci丘c and Atmospheric Circulation． Eastem Equatorial Pacific Sea Surface Tem－ In Studies on the Weather Systems of Chinese peratures an（i Rainfall over India and Sri Subtropics in Summer（S．Y．Tao，ed．），Science Lanka．Mon．Wea。Rev。，川，517－528， Publishing Company，Beijing，2－19．（in Chinese） Sa（玉ler，」．C。， L Oda and B．J．Kilonsky （1976）： Tashiro，S． （1981）： Long－Range Forecasting of Pacific Ocean Clou（iiness from Satellite Ob－ Typhoon． Manual of One－Month Weather servations。Dept．of Meteor．，Unlv．of Hawaii， Forecasting in Japan，Japan Meteor．Agency， UHMET76－Ol，137pp． Tokyo，225－238．（in Japanese） Sakai，H．（1952）： On the Path of Typhoon． 」． Tokita，M．（1972）：Long－Range Forecasting for Meteor。Res。，4，169－171．（in Japanese） Tropical Cyclone．Manual of Seasonal Fore－ Shideler，D．H． and J．C．Sadler （1979）： Pacific casting in Japan， Japan Meteor．Agency， Ocean Cloudiness from Satellite Observations． Tokyo，401－420． （in Japanese with English Dept．of Meteor．，Univ。of Hawaii，UHMET abstract） 79－13，56 pp． Wallace，」．M．and D．S．Gutzler（1981）：Telecon－ Shimizu，N．（1971）：Areal Distribution of Genera－ nections in the Geopotential Height Fiel（1dur－ tion and Passage of Typhoons．J．Meteor。 ing the Nofthem Hemisphere Winter．Mon． Res．，23，205－215． （in Japanese） Wea。Rev．，109，784－812． and K．Shirai（1974）： Areal and Monthly Wang，J。Y．（1982）：Statistical Relationship be－ Distribution of Travelling Directions of Ty－ tween the Frequency of Typhoon over North－ phoons．J。Meteor．Res．，26，535－548。（in Japa－ west．Paci行c and Field of Monthly Mean Sea nese） Surface Temperature． Acta Oceanologica Sikka，D．R，and K．D．Prasad （1981）：Empirical Sinica∫ 1，40－46。 Orthogonal Analysis of700mb Geopotentials Wendland，W．M．（1977）：Tropical Storm Fre－ during the Summer Monsoon over Asia．J・ quencies Related to Sea Surface Temperatures． ClimatoL伊1，367－379． J．AppL Meteor．，16，477－481， Suzuki，Y．and M．Miyahira （1982）： On the Xie，S。M．，T．Aoki and M．M．Yoshino （1983）： Secular Variation of Intense Typhoons Reach－ Relation between the Number of Typhoon ing the Vicinity of Okinawa District． J． Visit in East Asia and Sea Surface Tempera－ Meteor．Res．，34，307－317．（in Japanese） ture in the North Pacific．Tenki，30，495－502． Tachi，T． （1961）： Some Statistical Studies on （in Japanese） Tropical Cyclones in the Westem Paci丘c Ocean Xu，N．Y．and Z．Y．Lu（1982）：The7－year Cli－ （first report）．Tenki，8，337－344．（in Japanese） matic Oscillation of Typhoon Frequency over （1968）：Some Statistical Studies on Tropical the Northwest Pacific Ocean in August and its Cyclones （fifth report， on Landing Japan）。 Mechanism． Proceedings of the 1980sym－ Tenki，15，97－101．（in Japanese） posium on Hydrometeology of the Chinese Takahashi，K．（1948）：Typhoon in Japan．Geophys． Society of Oceanology and Limnology，Science Mag．，1了，1－35． Publishing Company， Beijing， 155－165。 （in （1962）：AClassStormsAttackedonJapan Chinese with English abstract） during the Past300Years．Tenki，9，277－283． Yoshino，M．M．and K．Kai（1974）：Pressure Pat－ （inJapanese） tem Calendar of East Asia，1941－1970and Its （1982）： Climatic Variations and Disasters． Climatological Summary．Climatological Notes， Chiri，27身No．12，44－50． （in Japanese） Univ．of Tsukubaン16，1－71． Tanabe，S． （1959）： Seanonal Variations in the ，S．M．Xie and T．Aoki（1983）：Interamual Frequency of Typhoons Relate（i to the General Variation of Typhoon Visit in East Asia and Circulation of Northem Hemisphere．J．Meteor． Its Relation to the Sea Surface Temperature Res。，11，561－566．（in Japanese with English of the North Pacific． Bu1L Envir，Res．Center， 118 T．Aoki yア01．36， No．2

Univ．of Tsukuba，了，63－68．（in Japanese）

台風の発生と日本への台風の襲来に関する気候学的研究

青 木 孝＊

北太平洋西部における台風の発生および日本への台風の襲来について，地域分布や年変化，経年変化などの気候 学的特徴を明らかにした。台風の発生については1953－1982年の30年間を解析の対象とした。日本への台風の襲来 は，さらに長い期間の資料を収集して，1913－1982年の70年間について解析した。台風が多く発生したときと少な いときの両者について，発生場所や500mb高度場，雲量，海面水温を比較した。また台風が日本へ多く襲来した ’年と少ない年における台風ゐ襲来数の分布の違いを調べた。日本各地の台風襲来数の年変化型の地域差を主成分分 析で明らかにし，得られた固有ベクトルに対応する振幅係数を使って日本の地域区分を行った。 次に，東部赤道太平洋における海面水温の異常現象であるエル・二一ニヨと台風の発生数との関係を見いだすと ともに，北太平洋の海面水温と1953－1982年の30年間における台風の発生数および日本への台風の襲来数との相関 関係を解析した。大きな相関係数が得られた海面水温，すなわち台風が発生する前年と2年前の北太平洋の海面水温 を予測因子として重回帰分析を行ったところ，北太平洋の海面水温が，台風の発生数や日本への台風の襲来数を長 期予報するための資料とLて役立つことがわかった。

＊現在 気象庁長期予報課