Bydil-opac.bosai.go.jp/publication/nrcdp/nrcdp_report/PDF/...wind,considering the deve1opment and...
Transcript of Bydil-opac.bosai.go.jp/publication/nrcdp/nrcdp_report/PDF/...wind,considering the deve1opment and...
Report ofthe National ResearchCenter for Disaster Prevention,No.30,March1983
551,577.21
Eva1uation of Upper Winds as a Translati㎝Vector
for the Short-Term For㏄ast of Pr㏄ipitation Echoes
By
Himshi Uyeda‡and Tsuruhei Yagi‡
ル加刎1地∫θακんα〃召γ〃〃∫狐伽〃θ㈱ガo〃,ノ;妙舳
Abstmct
Observation of precipitation echoes with conventiona1weather radar of3.2
cm wavelength was carried out from September7to October6,ユ981at the
center of the Kanto P1ain in Japan.For studies on the short・term forecast of
precipitation echoes,the evaluation of upper winds as a trans1ation vector was
made by simple trans1ation with recorded echo data for1or2hours,The winds
used for the translation were those at800,700,600,500and400mb and the mean
wind from850to400mb,provided from rawin data at Tateno.
The resu1t of the forecasts was eva1uated by Critica1Success Index(CSI)、
Overa11,the mean wind gave the1argest CSI score for translation.Next to this,
the CSI score for translation from the wind at700mb was1arger than that from
the other upper winds.The analysis demonstrated,therefore,as a who1e that the
mean wind wou1d be practical for use as a trans1ation vector for the short-term
forecast of precipitation echoes.
In addition to CSI ana1ysis,meteorological ana1ysis was made for two cases.
The possibi1ity of the app1ication of Doppler radar to the present method of
short-term forecast of precipitation echoes is also discussed.
1. Introduction
Aioi in!971and Nagasaki in1982are two examp1es of heavy rainfa11which is
not infrequent in Japan,causi㎎serious1oss of life,crops and property.Weather radar
has been p1aying an important.role in the watching of such heavy rainfa11.Recent1y,
severa1studies of short-term weather forecasts with radar records have been carried out.
The information of short-term forecasts will be va1uab1e for disaster prevention in1arge
cities and pub1ic faci1ities1ike dams and airports.To protect these important1ocations,
*Climatic Anoma1y Laboratory,First Research Division
一1一
Report of the Nationa1Research Center for Disaster Prevention,No.30,March1983
a medium radar range of about100km might be practical,considering the scale and1ife-
time of most cases of heavy rainfa1l,
Up to date,several methods of forecasting from radar echo pattems with the use
of digital echo data have been deve1oped.Austin and Be11on(1974)proposed a technique
to trans1ate digita1radar echoes by Crosscorre1ation Method(CCM).Be11on and Austin
(1978)made an evaIuation of rea1-time operation of a short-term precipitation forecasting
procedure(known as SHARP)with actua11y observed data.E1vander(1976)evaluated
three different methods of short-term precipitation forecasting by using the same data.
In Japan,Tatehira and Makino(1974)made an eva1uation of the forecast of digita1echo
pattems with an echo composite map from severa1adjacent radars.They extrapolated
the echoes from the wind at700mb inc1uding an adjustment by echo deve1opment.
Tatehiraθチα/.(1976)examined the forecasting technique of trans1ation from700mb
wind,considering the deve1opment and dissipation derived from two photographs at an
interva1of1hour.These studies have shown that a short-term forecasting procedure
using CCM or tI.ans1ation method from a certain wind could be practicab1e.
Any method of short-term forecasting first needs to determine a trans1ation
vector.For CCM,both current map and past map stored from the previous hour are
necessary;therefore,a certain store time is required to determine the disp1acement
vector.In the case of a l or2hour short-term forecast,the disp1acement vector shou1d
be determined as quick1y as possible.Not on1y is this a difficu1t requirement,but a1so
a proper disp1acement vector may not be determined when1ocal1y irregu1ar movements
of wide1y spread precipitation echoes are prevai1ing.Therefore,in this paper we use a
certain wind vector instead of a disp1acement vector for translation.For this purpose,
Dopp1er radar is considered to be usefu1and convenient to determine the translation
vector quick1y and simp1y by measuring upper winds at rea1time when the precipitation
echo ho1ds its activity for2or3hours.
The technique of measuring the wind by a sing1e Dopp1er radar has been
described by Lhermitte and At1as(1961),Browni㎎and Wex1er(1968)and othersl The
present authors tried rea1-time measurement of horizonta1winds using fan-beam Dopp1er
radar(Uyeda and Yagi,1980a,1980b).
In the present investigation,the radar observations were made in the rainy
periods in September and October in1981and1982,using conventiona1radar to measure
the intensity of ref1ectivity.For a trans1ation vector,the authors used upper winds from
一2一
Evaluation ofリpper Winds as a Trans1ation Vector-Uyeda and Yagi
rawin data,regarding them as upper winds measured by a sing1e Dopp1er radar.Aoyagi
(1982)has observed that upper winds measured by a sing1e Dopper radar coincided with
rawin winds.In this paper,the eva1uation of upper winds as a trans1ation vector is
described as the resu1t of analysis for the1981observation data,and the possibi1ity of the
app1ication of Dopper radar to the short-term forecast of precipitation is discussed.
2. Methods of observation and analysis
2.1 Radar data acquisition and digitization
Observation of radar echo for the study of short-term forecast of digitized echo
pattem was carried out in a rainy period from September7to October6.1981in the
center of the Kanto Plain.The observation range was l00km in radius,as shown by the
circ1e in Fig.1.Mountainous regions over600m above sea1evel are hatched.A mobi1e
weather radar set was set up on the bank of the River Tone.The height of the bank was
about20m above the neighbouring ground.The wave1ength of the radar was3.2cm;
other specifications of the radar are1isted in Tab1e1.
The program of radar observation was schedu1ed from0200to0600JST,from
0800to1200JST,from1400to1800JST and from2000to2400JST.These observation
series were chosen to ut1ize the rawin data of0300.0900.1500and2100JST which would
provide a translation vector for radar echo pattems.After first precipitation echo was
identified,photographs of isoecho(ISO)contours on a p1an position indicator(PPI)were
taken by35mm cinecamera every lO minutes.The range of the PPI scope was lOO km
in radius,and the e1evation was set at an ang1e of3。.In addition to this PPI recording,
ISO contours of a range e1evation indicator(REI)were photographed at interva1s of30
minuteS.
After the observation period was comp1eted,off-1ine ana1ysis was made of the
obtained data.The intensity of PPI echoses on the screen of a cinefilm analyzer was
digitized in a2km×2km mesh by hand.The digitized1eve1s were IS01,2,3,4,5and
6of the isoecho device of the radar set used.Ground echoes were eliminated from the
digitization by the comparison of PPI echoes to the gromd echoes taken on days without
precipitation.
一3一
Report of the’National Research Center for Disaster Prevention,No.30,Marchユ983
37o
MT.FU」1
籔 くMT,AKAg! . }
% i、
/’ .ノ
_, 、.^一 r「]’ 、
㌧. 、一.・/一 、 一’’’’’、’く MT TSuKu8ム
、・、εADAR血 ○、.
へ. 十
’ 、\.・舳 KANTO PLAlN /.、・一t
-r、’.、?、.。へ・∫叶J
TOKYO} ’、。、.、一 {・一一\、 鶉
‘ 、. :わ ’\
TOKYO
BAY
1O O km
36oN
喬35oN 亀
紐139oE Q 140oE
35oN
0 30kn1
」]」
Fig.1一 Schematic map showing the1ocation of a mobi1e weather radar set in Goka Vi11age
(●)and the location of the rawin observation(Aero1ogical Observatory,Japan
Meteorologica1Agency)at Tateno(十).The circle lOO km in radius shows the radar
observation range.Mountainous regions over600m above sea1eve1are hatched.
Tab1e1 Specifications of the mobi1e weather radar set.
TERM
FREQUENCYPEAK POWER OUTPUT
PULSE WIDTH
PULSE REPETITION RATE
REFLECTORBEAM WIDTH
INDICATOR
ISO ECHO
PERFORMANCE
9375MHz
40kw
1μs
260pPs
ユ.2mφPARABOLA 2口
PPI,REI,A/R
<100km
一4一
E・・1・・ti…fUpP・・Wi・d・…T・…1・ti・・V・・t・・一U。・d・・。dY.gi
2.2 Echo translation by wimd vector
These digitized echo patterns were simply trans1ated by computer using the
upper winds observed by the Aero1ogica1Observatory at Tateno. The evo1ution of
precipitation echoes was not considered in the trans1ation for one or two hours.The
observatory is1ocated32km from the radar site atユ0ぴ、.as shown in Fig.1.The upper
winds used for the trans1ation were the winds at800,700,600,500and400mb and a mean
wind(Vm).The mean wind Vm is defined as the vector mean of winds at ten points from
850to400mb at interva1s of50mb.Initia1echo pattems app1ied in the trans1ation using
these winds were the records observed on the hour or30minutes after the hour.The
trans1ation1engths were lO,20,30and60minutes and,if possib1e,the1engths were
extended up to gO,120and180minutes.The REI echoes in these trans1ation spans were
examined to check the vertica1structure of precipitation.
For verification of forecasts from radar echoes,a we11known parameter was
chosen.That parameter is the Critica1Success Index(CSI)(Donaldsonθ≠α/.,1975),
defined as
X CSI=100 X+Y+Z
where X,Y and Z have meanings exp1ained in Tab1e2. Fig.2wi11aid better
interpretation of the tab1e.The CSI score was ca1cu1ated for each trans1ation from upper
winds.The1eve1of intensity of the echoes trans1ated were over IS01(corresponding to
l mm/h)and over IS02(corresponding to2mm/h),since the area of intensity over IS0
3was very sma1L
Tab1e2 Definition of parameters used in the calcu1ation of CSI,
adapted from Dona1dson〃α/.(1975).X,hits;Y,missed;Z,
fa1se aIarms,as schematica1ly i11ustrated in Fig.2.
FORECASTEDOBSERVED VALUE EQUAL TO VALUE LESS THAN
OR LARGER THAN THRESHOLDTHRESHOLD
VALUE EQUAL TOOR LARGER THANTHRESHOLD X Y
VALUE LESS THANTHRESHOLD Z
一5一
Report of the National Research Center for Disaster Prevention,No.一30,March1983
Fig.2 Conceptiona1illustration of ovgr1apPing
of two echoes.The area inside the so1id
line shows the observed echo and the
area of broken1ine shows the forecasted
echo.PartXisthe“over1apped”area,
partsYare“obsemedbutnotforecast-ed”,andthepartsZare“forecastedbutnot observed”、
OBSERVED
一一一 FORECASTED
2.3 Echo tmns1ation by displacement v㏄tor
A disp1acement vector between a present PPI echo and the one30minutes
previous was a1so ca1cu1ated for the translation of a present echo into the future.The
common method of obtaining the disp1acement vector was described by Austin and
Be11on(1974).In the present paper,many CSI scoresbetween two PPI echo pattemswere
first ca1cu1ated by computer by trans1ating the past echo in many directions and
distances aromd the present echo.Then,for future trans1ation,one single disp1acement
vector was determined so as to give the1argest CSI score.The interval between the two
echoes was30minutes.This method derives the same result as the Crosscorre1ation
Method for the echo area of IS01.
Using the disp1acement vector,short-term forecasts of precipitation echoes
were made in the same mamer as short-term forecasts using the wind vector,for
comparison.
2.4 DIeteoro1ogical analysis
Meteoro1ogica1ana1ysis was made for typica1cases observed from September
7to October6,198ユ.
In order to examine the relation between the types of precipitation and the
va1idity of trans1ation vectors of echo pattems,surface weather ana1ysis using
Automated Meteorological Data Acquisition System(AMeDAS)data was made for
every case in the observation period.A1so,time cross-sections were made for typica1
CaSeS.
一6一
Eva1↓lation of Upper Winds as a Trans1ation Vector-Uyeda and Yagi
3. Resu1ts
3.1 Averaged CSI scores
The movements of precipitation echoes were observed in the period from
September7to October6.1981in the Kanto Plain.The observation data were c1assified
into eight cases,as listed in Table3,in terms of rawin observation time of0300.0900.
1500and2100JST.In Cases1,2and3,observed rains were associated with a cyc1one.
Especia11y in Case2,heavy rainfa11was observed in a sma11area according to the
AMeDAS data ana1ysis.In Case4,the echo observed by the radar was ofconvective type
and its intensity was high,but no heavy rains were observed at the ground.In Cases5,
6,7and8,observed rains were associated with Typnoon No.8122(E1sie)approaching the
radar observation range.Especia11y in Case8,the duration of the precipitation echo was
1ong enough for trans1ation of over2hours.
On each case,the echo areas over the1eve1of IS01on the hour and30minutes
after the hour were trans1ated for forecast1ength as1ong as possib1e by upper winds at
800,700,600,500and400mb and the mean wind,Vm(850to400mb),at Tateno.The
averaged CSI scores of the trans1ations are shown in Tab1e4.Two echo patterns of Case
2were trans1ated up to gO minutes and three echo pattems of Case8were trans1ated up
to120minutes.Case3and Case4were e1iminated from the averaging in the tab1e
because the precipitation,echoes cou1d not be trans1ated more than60’minutes.
Table3 List of casens of radar observation data.The observation
time of rawin data used for the trans1ation is shown in the
RAWIN TIME co1umn,The disturbance associated with
the precipitation echoes observed is shown in the RE-
MARKS co1umn.
CASE NO. RADAR OBSERVATION TIME RAWIN TIME REMARKSユ 2000-2400JST,SEP.25.!981 2100JST CYCLONE2 0000-0600JST,SEP.26,198ユ 0300JST CYCLONE3 1640一ユ800JST,SEP.26,198ユ 1500JST CYCLONE4 1400-1630JST,SEP.28.1981 1500JST FRONT5 0805-1200JST,OCT.ユ,1981 0900JST TYPHOON6 1200-1800JST,OCT.1,ユ98! 1500JST TYPHOON7 2000-2400JST,OCT.1.1981 2100JST TYPHOON8 0000-0600JST,OCT.2.1981 0300JST TYPHOON
一7
Report of the Nationa1Research Center for Disaster Prevention,No.30,March1983
Tab1e4 Averaged CSI score for trans1ation from winds at
800,700,600,500,and400mb and the mean wind Vm.
30min. 60min、 90min. 120min.
800mb 22.7 8.0 2.6 0.3
700mb 23.3 1!.7 8.3 5.1
600mb 21.6 8,O 3.2 0.9
500mb 19.8 5.4 1.3 0.1
400mb 19.9 4.6 2.1 0.2
Vm 27,3 15.5 11.2 10.9
Eventua11y,the number of trans1ations were14,14,6and3for trans1ation1ength up to
30,60,90and120minutes,respective1y,
In Case2and Case8,the echo duration was1onger and precipitation amount
was greater than in the other cases.The resu1ts of trans1ation on the two cases are
precise1y described in the fo11owing sections and,in addition,the resu1ts of meteoro-
1ogica1ana1ysis for these cases are given.
3.2 Detai1ed description of Case2
Ear1y on September26.1981,Kanagawa Prefecture and Chiba Prefecture,the
southem part of the Kanto P1ain,had heavy rainfa11of about60to70mm in6hours.The
radar echo corresponding to this rain was obtained,and designated as Case2in Tab1e
3.An example of the PPI echo and the REI echo at0030JST is shown in Photo.1.The
height of the echo top was8km,the strongest echo was IS05in the intensity1eve1which
corresponded to16mm/h,and the echo moved toward the east-northeast.The ech0
features in the photograph show that the echo was wide1y spread and was of convective
form.
The changes in CSI score for forecast length up-to gO minutes starting at0030
and0010JST are shown in Fig.3(a)and(b).The echo pattem translated was of the leve1
over the intensity of IS01.In addition to the changes in CSI score o〃∫〃∫each upper
wind and the mean wind,the change in CSI score in the case of a disp1acement vector is
Eva1uation of Upper Winds as a Trans1ation Vector-Uyeda and Yagi
O030JST SEP.26.1981
(N)
Oo
PPI
170o
IS01
REI
Photo.1
(170◎) ISO
An examp1e of the PPI echo and REI echo at O030JST,September26.
1981.The REI is the vertica1cross-section in the direction of170.on
the PPI scope.
一9一
Report of the National Research Center for Disaster Prevention,No.30,March1983
ω
O
70
60
50
40
30
20
10
O030」ST,SEP.26.1981
(lSO l)
70
60
50
、、 40’、\
・、“ ω・“ O\’、 へ 30
、
㌧ 、 、 、 、ミーOPT1MuM 、 、 }・
“’㍉\ 瓜一 20 ・1・・、ぐ、PASTごvT一㍉
4。。“、。。。溢、’・榊・・ 1・
400mb\、 ’込、 \\\・一一.二三三…ミ葦
Fig.3
30 60
FORECAST LE NGTH
(a)
90 0
0100」ST SEP.26 1981
(lSO l)
、
・、一 ム、
“ 、、も 、、“ 、 “ 、○OPTlMUM
州 ’\
へ旦全・H \ ●■一一.:● 、、 、
{\\・ 一Vm\ 500mb込εOOm6ろ、、一、 b
1・・一... \・。600mドー、、
伽}ら\こ・1\≦讐二b
\ヒ==二。,;;一
(min) 30 60 90FORECAST LENGTH (min)
(b)
Change in CSI scores for forecast1engths up to90minutes:(a)starting at O030JST,
and(b)starting at0100JST. The thresho1d of echo intensity is IS01for the
ca1cu1ation of CSI score.The trans1ation vectors are the wind ve1ocities at800,700,
600,500and400mb and the ve1ocity ofthe mean wind,Vm(from850to400mb).TheCSI score for trans1ation from a disp1acement vector ca1cu1ated between past and
present PPI echoes is shown by PAST-30.The possib1e maximum CSI score by the
simp1e trans1ation method is shown by OPTIMUM.
shown as PAST-30in both figures.The disp1acement vector was determined from two
PPI echoes:present and30minutes previous.Furthermore,the change in possib1e
maximum score by the simp1e trans1ation method was shown as OPTIMUM for
comparison.
As is we11known,any CSI score decreases with the forecast length as a who1e.
Among them,trans1ation from mean wind Vm shows the1argest CSI score,being c1ose
to the OPTIMUM.Trans1ation from PAST-30a1so presents a relatively good score.On
the other hand,trans1ation from the wind at700mb shows a sma11CSI score,not as
genera11y expected.Fig.4shows the resu1t of forecasts of the echoes over the intensity
of IS02starting at O030and0100JST.The tota1tendency is a1most the same as
mentioned above and i1lustrated in Fig.3.
一10
Evaluati㎝of Upper Winds as a Translation Vector-Uyeda and Yagi
ω
O
70
60
50
40
30
20
10
O030」ST,SEP.26. 198170
(IS02) 60
、、
い、\ 、、 ・、\ .■一一Vm
トー、、二・
\ \\ 、 . や、
1 \、 、ぐ800mb 、、500m{\ \・、、
帖OOmb\、 \\700mb“ \、 、苧600mb ・\、
ω
O
50
40
30
20
10
30 60 90 0FORECAST LENGTH (m i n)
0100」ST,SEP.26, 1981
(lS02)
、
“’ぺ、
1,1、
、““、、 ←Vm
、’1\
、ぺ\.
、\、. \く800mb 、、、’・ ・、
、 、、、、“500mb
’・400mト、、
700mb 」 ’ト“600mb
(a)
30FORECAST
(b)
60 90LENGTH(min)
Fig.4 Same as Fig.3,but the threshold of echo intensity is IS02.
0300」STSEP.26.1981 L14o 150oE 〃 160oE
/し 呂、O,2
4
34oN η 5
O.N
々’016
’70
他
2’
2
11
8o
、一
1{ 〆 、ム1
1.
17 16
い
亭1約、。と。、。。洲7
■
,。 21も 7・一 窃 、、19
。,8・、。,、グ、、
2= 23
40oN
、020
3ooN
2多 鼻~5 ・14
η♂ 争
?6g
1200 130oE 140.E 150oE
Fig.5 Surface weather chart at0300JST,September26.1981(○:
c1ear,①:fine,◎:c1oudy,●1rain,●二:shower,◎:fog)・
一11一
Report of the Nationa1Research Center for Disaster Prevention,No.30,March1983
」棚O一ユコOO」5T 、一’■、・舳.1㌣∵1ノ∴1’\しノ、へ、、一二’・。(’目ユ・(、ド」・∵1・.㌧1.ドドエ㌧’』、.、.・・ノ..,ノ1… 川 、一r~・;∴パ∴∴1∴∴1パてllゴll.、,∴。...、、、、
1‘1Eヨー・^
/三ノ1口。二;ケ/・旭 旧洲_一’;’1’㍉
lOn…いヨ〕o……{10〔[10……
∴。…(:ll∵喜
‘
○盲。。 ヨ≡“
0 50一…1‘OoE
::蝋鴉丁一、、 、,/■一「\ 一」一工
阯1’El
一
;;1=
■一ユ E oj。皇・ノ..・’’I1〕・1’.へ、・’㌧/!阯・;ニゾ..ダー一、二㌧市ノー.い.1
\し、∴1;l1砧11岬[皇11㌔≡∵’
’
’
’
」■二’
二、∴\ ll、:[
、
一5……・
.1.∵、、三二j
{・‘1…’]5“ ’
コ;.一
0 50一…舳 Q 削・・ …
0200-oコoo」昌丁 、、一!一’’う“、SEP’6’1舶1・〔㌧.、、・・1/1。ド1、一、、、、∴、ン目、1∵∵・ゲ「!∴三・二一!咀;・㌧、;;、.一.、・ノ.・■!一:.’0。へ、1.一!」f■一㌻…1ゾ..ダーべ、、ノー、・一、
;1::
’∵∴∴1二逗;豪1’盟、ノ’。 5【…
1、{
コ6
し、
コヨ“ヨ;・}
0 50一〔!・、1血・ ○ 帖杵 …
莇’一、’一
ゴ.
、、ベニ洩、
;・/一ノパ1・》”、、
(’; ・1.
旧Ψ ;
1眺 ㊦
SEP 三苫. 1901
肺ぺ、’’’
・ 1眺 Q
、凶誼、一
… 1、…一1
舳’v-
L1;一{..,辻.
〃奪一
趾 加o
Fig.6 Hourly amounts of precipitation from2100JST,September25to0300JST,September26.1981and6hour amount of precipitation from2100JST,September25
to0300JST,September26,198]
一12一
Evaluation of Upper Winds as a Translation Vector-Uyeda and Yagi
Next to the CSI ana1ysis,the surface weather data and rawin data were
ana1yzed as fo11ows.Fig.5shows the synoptic chart of0300JST,September26.The
warm front of a cyc1one is1ocated south of the Kanto area.Fig.6shows the hourly
amounts and6・hour amount of precipitation from2100JST,September25to0300JST,
September26from AMeDAS data.The figure shows that heavy rainfa11is distributed
in the southem part of the radar observation range;the heavy rainfal1area moved from
west to east. According to echo tracing,the movement of the precipitation area
coincided with the movement of the radar echo.Fig.7shows the time cross・section of
wind at Tateno from2100JST,September24to2100JST,September26.The wind at
700mb at0300JST,September26is seen to be different from the adjacent wind.Fig.
8shows the cange in the surface wind fields.A sma11cyc1onic convergence zone moved
to the east,corresponding to the movement of the precipitation area in Fig.6,Fig.9
showsthe changein temperature distributions.It can be seenthat the20℃isotherm inthe
southern part migrated1oca11y eastward,corresponding to the movement of the rainfa11
area,a1though the temperature situation scarce1y changed as a who1e、
コ・TA二㌧〉〉一、L虻V㌧lmb〕
ノノ〉》レ虻虻し火
・・ノノレ》しk虹㌧L “」 “」 “」
5。。yy〉〉/ 》》
舳ノノノノ〉》yy〉 ㌧
η。ノノノ 〉ノノノノ
。。ξと。 09 21 09
SEP25 21」STSEP.24.1981
Fig.7 Time cross-section of winds at Tateno from2100JST,
September24to2100JST,September26.1981.
一13一
Report of the National Research Center for Disaster Prevention,No.30,March1983
帥\・1一
SEPユ川1・へ /1 (
一、くノ’
舳 v■’
SEP26・1981、〔、
!.三(
く 1ソ
く く’
.、.へ
Y’
戸」1 州
・、γパ∵
Fig.8 Change in surface wind fie1ds
from2400JST,September25 to0400JST,September26. 1981.
r1与,
亜・パv…
、.1’旧ヅπ 1・1.・
/ 、、、」、.
一 1目t
がc/珊コ
舳咀・Q
㌣ノ、り舳
た 別t 19 鴉.■ 1, 、
1“一・;昌5 ’害・ 写1 11J
二鮒τ 一ノ1・帆、 1・1.・
眺’“ \
19’二六、ご
1軸・・lllQ
70t
1眺
○ヤ
二言㍑,、。、 1へ
!㌧シ;宜\㌻ll々㌧…洲…・ll㌻、許㌶努
\ づ
1加・・≡16Q lω・・
Fig.9 Change in surface tempera-
ture fie1ds from 2400JST,
September25to0400JST, September26.1981.
一14一
Evaluation of Upper Winds as a Trans1ation Vector-Uyeda and Yagi
3.3 Detai1ed description of Ca.se8
Typhoon No.8122(EIsie)was about200km south of the radar observation site
at about0300JST,October2,and her outer rain band was catched-by the radar as Case
8.The amount of precipitation during the radar observation times from0000to0600JST
was!O to20mm in the radar observation range.Photo.2shows an examp1e of PPI echo
and REI echo at0200JST,October2.The echo feature in the photograph shows that the
echo was stratiforml The direction of echo mevement was north-northeast.
The changes in CSI score for forecast1ength up to!20minutes starting at0200
JST are shown in Fig.10.The echo pattem trans1ated was of the1evel over the intensity
of IS01.The legend for Fig.10is the same as for Fig.3.Any CSI score decreases with
the forecast length as a who1e in the same mamer as seen in Fig.3.Trans1ation from
mean wind Vm shows the1argest CSI score,being c1ose to the OPTIMUM.Trans1ation
from PAST-30a1so presents a re1atively good score.Trans1ation from the wind at700
mb shows a1arger CSI score than from the winds at the other heights in this case.Fig.
11shows the forecasts from echoes over the intensity of IS02starting at0200JST.The
tota1tendency of the forecasts for IS02echoes is simi1ar to that for IS0!in Fig.10.
70
Fig.10
60
50
40
ω
0 30
20
10
0200」ST,OCT.2.1981
1 .,’一9 (1S01)
弍 ひ1 \・、 、
い、 \\、、、・・…一碧\。。。。1・。・
、“ \ 、 1\ 、、 ’、 、 、、、 ’ 、
、・\ ㌧一一一一一一一一つ \\ 、・\
・、 、 \
500mbヘミ、 \ \ “Vm
、・ ・ \ x ’\ \ \ \ 、 、 、 、 、 \や800mb\ ぐ、一 ・
、、ミ、600mb{、 “700mb x
’、\ 、.
400mbづ・\ \、、
、・\ \、
0 30 60 90 120 FORECAST LENGTH (min)
Change in CSI score for forecast length up to120minutes starting at0200JST,
October2.1981.The thresho1d of echo intensity is IS01for the calcu1ation of
CSI score.The legend is the same as for Fig.3.
一15一
Report of theNational Research Center for Disaster Prevention,No.30,March1983
0200」ST OCT.2 1981 ,
(N)
Oo
PPI
MTl A■くAG1
;詑
1SO1
136o
RE1(1360) 1SO l
Photo.2 An example of the PPI echo and REI echo at0200JST,October2.
1981.The REI is the vertica1cross-section in the direction of136.on
the PPI scope.
一16一
Evaluatioh of Upper Wihds as a Translation VectoトUyeda and Yagi
70
Fig.11
ω
U
60
50
40
30
20
10
、
、’、、
、 、 、 、
“
、 、・、・、、 、\
\ \\\
、 、、 \ 、・ 、 、、 、 、 、 、 、 、
0200」ST,OCT,2. 1981
(lS02)
\ ‘Vm
、 、 700mb\ 、、
、 、 、
、 \ 、F600mb 、 、 ・
\\\\、 “ 、 、 、 、 、 ’\、ミOOmb\ 、 、 、 、
、 ・. 、、 、
400mbへ \、\ \・\ “500mb、、ミ、 』、_㌧
0 30 60 90 120
FORECAST LENGTH (min)
Same as Fig.10,but the threshold of echo intensity is IS02。
0300」ST,OCT.2.1981
、。、(姦ノ
=0N 、
ゼ
L
、い一・∵ ■
差銘炉・・帖鳩
勺.、蓑蘭翰N、、、、一
2噛
帝
・・♂ 魂
941/
1000 タ
旧昨 140㌔
Fig・12Surfaceweatherchartat0300JST,October2.1981(○:c1ear,①:fine,◎:c1oudy, ●:rain,●二:shower).
Next to CSI ana1ysis,the surface weather data and rawin data were ana1yzed
as fo11ows.Fig.12shows the synoptic chart for0300JST,October2.The typhoon’s area
of influence had extended to the Kanto Plain.Fig.!3shows the hour1y amounts of
一17一
Report of the National Research Center for Disaster Prevention,No.30,March1983
!一い1:1・
秘≡
ロヨ・正 更⊃ 1・O・・
/,、、=∵
\・二、 言”
1∵、Fig.13 Hourly amounts of precipitation at0200and0300JST,October2.1981・
TATENO
300
{mb〕
4oo
500
600
700
800
900
1OOO
㌧
也
帖
》
伽虹
し “」
忙
㌧
」
㌧
》
し
ノ
ノ
ノ
ノ
〉
》/ 〉
〉
〉〉
ノ
レ
〉
〉
〉〉
し
し
㌧
“ノ \!
L ノノy》yノく に
〃_ノレ
21
0CT.209 21 0CT.1
09 21」ST SEP.30 1981
Fig.14 Time cross-section of wind at Tateno from2100JST,September30to2100JST, October2.1981.
precipitation at0200and0300JST,October2.The amount of precipitation was not so
great,since the typhoon did not1and in the Kanto P1ain and1eft rapid1y after a change
of course to the northeast.Fig.14shows the time cross-section of wind at Tateno from
2100JST,September30to2100JST,October2.The wind direction at low1eve1be1ow
800mb at0300JST,October2was NNE,associated with the typhoon.
一18一
Evaluation of Upper Winds as a Translation Vector-Uyeda and Yagi
(b)
Fig.15
(a)
{N,
50
‘N j
5◎
‘W〕 ◎
100km{5〕
{W,
100km(S〕
50 100km‘E)
身・一j.
,
RADAR
囲1二
囲 一...
甘 、三一1・
国・’二=箏1
O
{二.1
口
囚CS二:16.0
0200」ST.SEP26
口:08SERVED BuT NOT F團:FORECAS正D8uT NO■l OVERLAPPED
km
0 50
08SERVED BuT NOT FORECASTEO
FORECAS TED8uT NOT08SERVED
100km{E〕
RADAR
騨 團
口
囮 B;:・’,
口
・。1し’1∴
1.∴ 洛函CSl=5.6
0200」SτSEP26
口:OBSERVED BUT NOT F図:FORECASTED BUT NO□:OVERLAPPED
km
OBSE RVED BUT NOT FORECASTED
FORECASTED BUT NOT OBSERVED
mustration of two examples of the fit between forecast echoes and present echoes.
(a):the echo observed at0200JST and the echo trans1ated from the echo observed
at0100JST for60minutes by the mean wind Vm.(b):the echo observed at0200JST
and the echo translated from the echo observed at0100JST for60minutes by the
wind of700mb.The overlapped parts are b1acked,corresponding to X in Fig.2一
19
Report of the National Research Center for Disaster Prevention,No.30,March1983
4. Discussion
4.1 Meanillg of CSI score
In order to investigate the proper trans1ation vector for short-term forecasts of
precipitation echoes,the meaning of CSI scores in the present ana1ysis ml」st be
considered beforehand.Fig.15i11ustrates two examples of how forecast echoes and
present echoes fit.The S.E,sector of the radar observation range was chosen for
il1ustration.The over1apped echo areas are b1acked,corresponding to area X in Fig.2.
The upPer figure(a)shows the fit between the echo observed at0200JST and the echo
trans1ated from the echo observed at0100JST for60minutes from mean wind Vm;the
1ower figure(b)shows the fit between the echo observed at0200JST and the echo
trans1ated from the echo observed at0100JST for60minutes from the wind at700mb,
The CSI score in the case of trans1ation from Vm is16.0and the main part of the echo
coincides we11,according to ana1ysis by echo tracing.On the other hand,the CSI score
W
(m’SeC〕
TATENO 03」ST,SEP26.198115
300rnb
10
350400
600 500OO
Vm850
1000
l0 5 S O 5 10 15 20 25 ‘m∫5ec〕
5 700
10
15
E
S
Fig・16Wind hodograph at0300JST,September26.1981at Tateno.The mean wind is shown by arrow designated Vm一
一20一
Eva1uation(〕f Upper Winds as a Translati㎝Vector-Uyeda and Yagi
in the case of trans1ation from the700mb wind is5.6.The b1acked echo area is
over1appedbydifferentechoes:echoAofFig.15(a)shouldoverlaptoechoB,aswas
ana1yzed from the successive tracing of PPI echoes recorded at interva1s of ten minutes.
Thus,the comparison of CSI scores gives an interpretation of how we11trans1ations fit.
In genera1,a smal1mesh gives a sma11CSI score,as suggested by Tatehira〃
α/.(1976).The authours adopted a sma11mesh of2km×2km for convenience to
eva1uate upper winds as a proper trans1ation vector.For the puspose of present paper,
it is not important whether the abso1ute va1ue of CSI scores is1arge or sma11.The
important thing is to compare the re1ative va1ue of CSI scores for the se1ection of a
proper trans1ation vector.In operationa1forecasts,severa1kinds of mesh size are used
according to the operational purpose.
4.2 Proper wind as translation▽ector
The trans1ation vectors eva1uated in the anaIysis were the winds at heights of
800,700,600,500and400mb and the mean wind from850to400mb at Tateno.The
se1ection of these winds was reasonab1e since most precipitation echoes were vertica11y
distributed from the ground to a height of400mb according to REI observation records.
A proper trans1ation vector can be derived from comparison of the averaged
CSI scores in Tab1e4.The wind at700mb was the best as a trans1ation vector among
the upper winds,which has a1ready been we11discussed(e.g.Tatehiraθ≠α1.(1976)).
However,the CSI score for translation from the mean wind was1arger than from the
wind at700mb as a who1e.Hence,it can be said as the resu1t of the present ana1ysis that
the mean wind is of practica1use as the trans1ation vector on average for short-term
forecasts of precipitation echoes,In particu1ar,the mean wind gives a better CSI score
when a wind at specific height is greatly different from the winds at other heights,as
briefly discussed be1ow,
In Case2,the CSI score for trans1ation from the wind at700mb was
unexpected1y sma11er than from other upper winds.The reason for this dissimilarity can
be exp1ained as fo11ows.Fig.16shows the wind hodogragh at0300JST,September26
at Tateno.The mean wind is shown by the arrow designated Vm in the figure.The wind
directi㎝at700mb is NW,being remarkab1y different from the other upper winds and
the mean wind Vm.This noticeab1e aspect is a1so seen in the time cross-section of winds
(Fig.7).The deviation in the wind direction at700mb from the direction of the mean
一21一
Report of the National Research Center for Disaster Prevention,No.30,March1983
wind wou1d produce the sma11CSI score.
The above discussion indicates that a translation vector for the short-term
forecast shou1d be chosen carefu11y since a wind at a certain height does not necessari1y
represent the driving force of echo movements.
4.3 Usage of Dopp1er mdar for short-term for㏄a.sts
In the present paper,the wind information from rawin data at1imited times was
made use of forα力o∫左力6オo ana1ysis,but for operationa1short-term forecasts,rea1-time
wind information is necessary for trans1ating echoes,Doppler radar is considered to be
usefu1and profitab1e for measuring the upper winds on a rea1-time basis.
Techniques of measuring wind by a sing1e Dopp1er radar have been reported by
Lhermitte and At1as(ユ961),Browi㎎and Wex1er(1968)and others.The present authors
tried the rea1-time measurement of horizonta1winds using fan-beam Dopp1er radar
(Uyeda and Yagi,1980a,1980b). The technique required in the present trans1ation
method is the so-caned Velocity-Azimuth Disp1ay(VAD).The VAD technique emp1oys
an azimutha11y scanning beam at an intermediate e1evation angle,providing a profi1e of
horizonta1winds above a Dopp1er radar set when the radar is surrounded by scatterers
with uniform motion(Battan,1973).
An operationa1experiment with short-term forecasts using Dopp1er radar
shou1d be undertaken.
5. Comc1usion
Observation of precipitation echo mevemonts was p1amed for the eva1uation of
upper winds as a trans1ation vector,which is necessary for short-term forecasts.The
observation in1981was carried out with conventiona1radar from September7to
October6,in the Kanto P1ain.Digita1echo pattems were simp1y trans1ated by computer
using the upPer wind data provided by the Aero1ogica1Observatory at Tateno.The
rawin winds were regarded as upper winds measured by a sing1e Dopp1er radar.
The resu1ts of trans1ation were eva1uated by Critica1Success Index(CSI).The
CSI score for trans1ation from the mean wind was the1argest in the present data.Next to
this,the wind at700mb gavebetter CSI scores for trans1ation than the other upper winds.
However,the CSI score for trans1ation from the wind at700mb was unexpected1y sma11
一22一
Ev早1uati㎝of Upper Winds as a Trans1ation Vector-Uyeda and Yagi
in a certain examp1e.The reason for this was that the wind at700mb was noticeab1y
different from other upper winds:this wind was not1ike1y to represent the driving force
of echo movements.Therefore,the upper wind as a trans1ation vector for short-term
forecasts shou1d be chosen carefu11y.The genera1conc1usion of the present analysis is
that the mean wind is of practica1use as the the trans1ation vector.
The app1ication of Dopp1er radar to the rea1-time measurement of the upper
wind was discussed for possible operationa1short-term forecasts.
Acknowledgements
The authors appreciate the assistance at the radar site during the observation,
afforded by the Edo River Improvement Office,Ministry of Construction.The authors
a1so express their thanks to the Aero1ogical Observatory,Japan Meteoro1ogica1Agency
for the presentation of the rawin data.
This work was part1y supported by the Special Coordination Fund for
Promoting Science and Techno1ogy.
References
1)Aoyagi J.(1982):Signa1processing of Dopp1er weather radar and wind measurement with
it.Preprints(SANE82),丁加”∫”〃θoグE/κ”0”6∫舳∂C0刎刎”此”0”E”9加θ”∫0グ
ノ;αカα〃,15-22(in Japanese).
2)Austin,G.L.and A.Be1lon(1974):The use of digita1weather radar records for short・term
precipitation forecasting. Q〃α〃.∫.R.〃θチ.∫06.,100,658-664.
3)Battan,L J.(1973):Radar observation of the atmosphere.丁加σ〃加〆∫めoゾC〃ωgo,324
pp,cf.pp.145-154.
4)Be11on,A.and G.L.Austin(1978):The eva1uation of two years of rea1-time operation of a
short-term precipitation forecasting procedure(SHARP).∫.A力〃〃〃肋oγo/o馴.17.1778
-1787.
5)Browning,K.A.and R.Wexler(1968):The determination of kinematic properties of a wind
field using Doppler radar.ノ.λ〃伽∂〃θねoγolo8ツ,7,105-113.
6)Donaldson,R.J.,R.M.Dyer andM.J.Kraus(1975):Anobjectiveeva1uationoftechniques
for predicting severe weather events.1〕κ卯加な,9肋Co似Sωθκムoω/∫foプ舳∫,λ舳θ〃.
〃肋oプ.∫oo.,321-326.
7)E1vander,R.C、(1976):An eva1uation of the re1ative performance of three weather radar
echo forecasting techniques. Pκρガ〃な,17肋Rα6o1γ 〃θねoグ Coηグ,■4〃〃γ.〃θ加oγ.∫oc.,526
-532.
8)Lhermitte,R.M.andD.At1as(1961):Precipitationmotionby pu1seDopp1er.〃ψγ〃∫,9肋
肋α伽γ肋6〃Co〃,λ舳グ〃肋oプ.∫06.,218-223.
一23一
Report of the National Research Center for Di晶ster Prevention,No.30,Marchユ983
9)Tatehira,R.and Y.Makino(1974):Use ofdigitized echo pattem for rainfa1l forcecasting.
∫〃θ加oプo/oφω/Rθ∫θαγ肋,26,188-199(in Japanese,with Eng1ish abstract).
10)Tatehira,R.,H,Sato and Y.Makino(ユ976)=Short-term forecasting of digitized echo
pattern.∫”θねoκo/og北α/R8∫ωκん,28,61-70(in Japanese,with Eng1ish abstract)、
11)Uyeda,H.and T.Yagi(1980a):Measurement of the horizonta1wind ve1ecity using a fan-
beam Dopp1er radar、 地∫ωκ乃Noチθ∫ぴ肋εN”ゴo〃α/地∫〃κ々Cθ””力7D兆α∫伽
P〃〃〃ゴo〃.42,24pp.(in Japanese,with English Abstract).
12)Uyeda,H and T.Yagi(1980b)=Some considerations on the Dopp1er spectrum of the
horizontal wind ve1ocity as measured by a fan-beam radar.Rψo〃ぴ伽N”o〃α/肋∫ωκ乃
Cθ〃〃力7Dムα∫加γP7ωθ〃τ6o〃,24,33-43(in Japanese,with English abstract).
(Manuscript received November27.1982)
降水エコーの短時間予測のための.上層風
の補外ベクトルとしての評価研究
上田博・八木鶴平
国立防災科学技術センター
降水エコーの短時問予測の研究の一環として,降水エコーの観測を,波長3.2cmの通常気
象レーダーを用いて,関東平野の中心部で1981年9月7日から10月6日まで行った.降水
エコーの短時間予測の補外ベクトルとしての上層風の評価を行うために,レーダーエコーの
記録を1,2時問先まで単純補外した。補外に用いた上層風は,館野に位置する高層気象台で
観測されているレーウィンのデータによる800,700,600,500及び400mbの風と850mbか
ら400mbまでの平均風である.
予測結果はCSIの値によって評価した.全体として,平均風による補外がCSIの最大値を
示した.これに次いで,700mbの風による補外が他の高度の風による補外よりも大きなCSI
の値を示した.この解析から,全体的には,平均風を降水エコーの短時問予測のための補外
ベクトルとして用いるのが実用的であると結論された.
CSIの解析に加え,二つの例について気象解析を行なった.また,降水エコーの短時問予
測にドップラーレーダーを適用する可能な手法について討論した.
一24一