49 B18 4
Annuals of Disas. Prev. Res. Inst. Kyoto Univ., No.49B, 2006
(GPV)
* *
*
51 GPV
GPV
GPV
: GPV RSM
1.
(1996)
1 2
1
GPV(Grid Point
Value)
GPV
GPV
GPV
(1997)
(2005) (1997)
GPV
京都大学防災研究所年報 第 49 号 B 平成 18 年 4 月
Annuals of Disas. Prev. Res. Inst., Kyoto Univ., No. 49 B, 2006
Table 1 the summary of RSM
items contentsAnalysis time 00, 12 UTCForcat ranges 51 hours
Horizontal grid system Lambert projectionNumber of grid points 325 × 257
Grid spacing 20kmVertical levels 40 levels
( )
( ) ( ) ( )
( ) 5
GPV
GPV
GPV
GPV
2.
2.1 GPV
GPV
(2001) GPV
GSM(Global Spectrum Model)
RSM(Regional Spectrum Model) MSM(Meso Spec-
trum Model) 2
RSM GPV RSM
Table 1 2001 3 2005 5
4 RSM-GPV
Siroyama dambasin
Kurobe dammbasin
Biwako basinSameura dam
basin
Matsubara dambasin
the location ofbasins
Fig. 1 the mesh data and location of each basin
2.2
2001 4 0.025 0.03125
( 2.5km ) 2003
5 1 1
2003 6 30 1
1
2.3
(W07-52M)
(W01-07P) (G04-56M)
( ) (
) 100m
Fig. 2 correlation of mesh data and statical value
Fig.1
GPV
GPV
Fig.2
2002
0.99
3. GPV
3.1RSM 00UTC 12UTC 12
51
1 1 (3 5 )
(6 8 ) (9 11 ) (12 2 )
GPV
GPV
2.5km GPV
20km
00UTC 12UTC 48
4 12 2 24
GPV 00UTC
12UTC 48 4 12
2 24 2001
6 2005 5 4
CC(Correlation Coefficient)
RMSE(Root Mean Square Error)
(BIAS) RMSE
(COV
) COV GPV
RMSE
COV
n k
GPVk
RAPk 1 2
1 90 4
n 720 CC RMSE BIAS
COV
μR =( n∑
k=1
RAPk
)/n · · · · · · · · · · · · · · · · (1)
μG =( n∑
k=1
GPVk
)/n · · · · · · · · · · · · · · · · (2)
CC =
n∑k=1
(GPVk − μG)(RAPk − μR)/
(√√√√ n∑k=1
(GPVk − μG) ×
√√√√ n∑k=1
(RAPk − μR))
· · · · · · · · · · · · (3)
RMSE =
√√√√ n∑k=1
(RAPk − GPVk)2
n· · · · · · · · (4)
BIAS =μG
μR· · · · · · · · · · · · · · · · · · · · · · · · · · · · · (5)
COV =RMSE
μR· · · · · · · · · · · · · · · · · · · · · · · · · (6)
48 4
12
CC 12
12
0.2
12 24
48
CC BIAS
RMSE COV
Fig. 3 The distribution of the evaluation index
(spring)
RMSE
BIAS
12
12
Fig.3 CC
0.7 RMSE 0.5
BIAS
1
BIAS
BIAS
1
BIAS 1
BIAS
BIAS 1 RMSE
0.8 COV
CC BIAS
RMSE COV
Fig. 4 the distribution of the evaluation index (sum-
mer)
Fig.4 CC 0.7
0.5 BIAS
1 1
CC 0.5 0.7
BIAS 1 RMSE 0.6
0.7 0.8
CC BIAS RMSE
RMSE CC 0.7
COV
2.0 2.5
RMSE
BIAS 1
1 RMSE 0.8 5
Fig.5 CC
CC BIAS
RMSE COV
Fig. 5 the distribution of the evaluation index (au-
tumn)
RMSE BIAS
RMSE BIAS
RMSE BIAS
Fig.6 CC 0.6
BIAS
CC
RMSE
CC BIAS
RMSE COV
Fig. 6 the distribution of the evaluation index (win-
ter)
RMSE COV
COV
GPV
RSM
5
3.2( ) ( )
( ) ( ) (
) 5
3
CC
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 3 6 91215182124273033363942454851
CC
time
3h-integration siroyama CC time-series
mamjja
sondjf
RMSE
00.20.40.60.8
11.21.41.61.8
2
0 3 6 91215182124273033363942454851
RM
SE
time
3h-integration siroyama RMSE time-series
mamjja
sondjf
scattering diagram
0123456789
10
0 1 2 3 4 5 6 7 8 9 10
GPV
RAP
siroyama djf 3h-18 scatter
15h-18h18h-21h
the result of theadoption of RCC
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 3 6 91215182124273033363942454851
RC
C
time
3h-integration siroyama RCC time-series
mamjja
sondjf
Fig. 7 siroyama dam basin (precipitation of 3 hours
average)
Fig.7 mam jja son djf
Fig.7 6 12
CC
0.4 0.5
RMSE 0.2
CC
GPV
18 21 15 18
4
3mm/h
15 18
15 18 4
18 21 CC
3
CC RMSE
RCC
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 3 6 91215182124273033363942454851
RC
C
time
3h-integration siroyama RCC time-series
mamjja
sondjf
RMSE
00.20.40.60.8
11.21.41.61.8
2
0 3 6 91215182124273033363942454851
RM
SE
time
3h-integration siroyama RMSE time-series
mamjja
sondjf
BIAS
0.5
1
1.5
2
2.5
3
0 3 6 91215182124273033363942454851
BIA
S
time
3h-integration siroyama BIAS time-series
mamjja
sondjf
Fig. 8 siroyama dam basin (precipitation of 3 hours
average)
CC RMSE
(RCC ) RCC
GPV
RCC
Fig.7
CC
RCC
RCC RMSE BIAS
2001 6 2005
5 4 3 3
5
RCC
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 3 6 91215182124273033363942454851
RC
C
time
3h-integration kurobe RCC time-series
mamjja
sondjf
RMSE
0.6
0.8
1
1.2
1.4
1.6
1.8
2
0 3 6 91215182124273033363942454851
RM
SE
time
3h-integration kurobe RMSE time-series
mamjja
sondjf
BIAS
00.5
11.5
22.5
33.5
44.5
55.5
66.5
0 3 6 91215182124273033363942454851
BIA
S
time
3h-integration kurobe BIAS time-series
mamjja
sondjf
Fig. 9 kurobe dam basin (precipitation of 3 hours av-
erage)
RCC
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 3 6 91215182124273033363942454851
RC
C
time
3h-integration biwako RCC time-series
mamjja
sondjf
RMSE
0
0.2
0.4
0.6
0.8
1
1.2
1.4
0 3 6 91215182124273033363942454851
RM
SE
time
3h-integration biwako RMSE time-series
mamjja
sondjf
BIAS
0.5
1
1.5
2
2.5
3
0 3 6 91215182124273033363942454851
BIA
S
time
3h-integration biwako BIAS time-series
mamjja
sondjf
Fig. 10 biwako dam basin (precipitation of 3 hours
average)
Fig.8
RMSE
RCC 0.5 0.6
RMSE RCC
Fig.9
RCC 0.6 0.7
BIAS RMSE
RCC
RCC
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 3 6 91215182124273033363942454851
RC
C
time
3h-integration sameura RCC time-series
mamjja
sondjf
RMSE
00.20.40.60.8
11.21.41.61.8
22.22.42.62.8
3
0 3 6 91215182124273033363942454851
RM
SE
time
3h-integration sameura RMSE time-series
mamjja
sondjf
BIAS
0
0.5
1
1.5
0 3 6 91215182124273033363942454851
BIA
S
time
3h-integration sameura BIAS time-series
mamjja
sondjf
Fig. 11 sameura dam basin (precipitation of 3 hours
average)
RCC
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 3 6 91215182124273033363942454851
RC
C
time
3h-integration matsubara RCC time-series
mamjja
sondjf
RMSE
00.20.40.60.8
11.21.41.61.8
2
0 3 6 91215182124273033363942454851
RM
SE
time
3h-integration matsubara RMSE time-series
mamjja
sondjf
BIAS
0
0.5
1
1.5
2
2.5
0 3 6 91215182124273033363942454851
BIA
S
time
3h-integration matsubara BIAS time-series
mamjja
sondjf
Fig. 12 matsubara dam basin (precipitation of 3
hours average)
BIAS
RCC 42 0.6 0.4
Fig.10
3 RCC0.8
RMSE CC
BIAS 1
Fig.11
BIAS RCC
BIAS
1 RCC
Fig.12
RMSE 1.0
RCC
3 RCC RMSE BIAS
BIAS
RMSE
RCC
RCC
RCC 0.6
GPV
4.
GPV
GPV 1
3 GPV
BIAS 1
BIAS 1
BIAS
BIAS
BIAS CC 1
BIAS
BIAS
12
BIAS
48
12 BIAS
BIAS 1.2 0.8
GPV
1
GPV
GPV
1 GPV
9
3 5 7 3
5 7
GPV
12
12
BIAS
( )
BIAS
Fig.13 Fig.15 Fig.17
Fig.13 ( ) RCC, CC,
RMSE
top3 3
GPV
12
ave9 9
raw
3 12
RCC raw
0.65 0.4
0.7
CC 0.1 RMSE
top3 12
Fig.14
raw top3 top5 12
7mm/h
top3 top5
5mm/h
3mm/h
top3 top5 raw
RMSE top3 raw
RCC
0.4
0.5
0.6
0.7
0.8
0.9
1
0 12 24 36 48
RC
C
time
sameura RCC mam 12h time-series
rawtop3top5top7ave9
CC
0.4
0.5
0.6
0.7
0.8
0.9
1
0 12 24 36 48
CC
time
sameura CC mam 12h time-series
rawtop3top5top7ave9
RMSE
0.5
0.6
0.7
0.8
0.9
1
0 12 24 36 48
RM
SE
time
sameura RMSE mam 12h time-series
rawtop3top5top7ave9
Fig. 13 RCC ,CC and RMSE of Sameura dam basin
(spring)
0123456789
1011121314
0 1 2 3 4 5 6 7 8 9 1011121314
GPV
RAP
sameura mam 12h-48 scatter
rawtop3top5
0
1
2
3
4
5
6
7
0 1 2 3 4 5 6 7
GPV
RAP
sameura mam 12h-48 scatter
rawtop3top5
Fig. 14 Sameura dam basin (spring), the relation of
the past record and forecast of precipitation
Fig.15 ( ) RCC, CC,
RMSE
RCC CC RCC 0.2
CC 0.1 RMSE
raw top5
12 Fig.16
top5 raw
0mm/h
2mm/h
RCC
RMSE
Fig.17 ( ) RCC, CC,
RMSE
CC
top5 top7
Fig.18
RCC
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 12 24 36 48
RC
C
time
sameura RCC jja 12h time-series
rawtop3top5top7ave9
CC
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 12 24 36 48
CC
time
sameura CC jja 12h time-series
rawtop3top5top7ave9
RMSE
11.11.21.31.41.51.61.71.81.9
22.12.22.32.42.5
0 12 24 36 48
RM
SE
time
sameura RMSE jja 12h time-series
rawtop3top5top7ave9
Fig. 15 RCC, CC and RMSE of Sameura dam basin
(summer)
0123456789
101112131415161718
0 1 2 3 4 5 6 7 8 9101112131415161718
GPV
RAP
sameura jja 12h-12 scatter
rawtop3
0
1
2
3
4
5
0 1 2 3 4 5G
PV
RAP
sameura jja 12h-12 scatter
rawtop5
Fig. 16 Sameura dam basin (summer), the relation of
the past record and forecast of precipitation
raw CC0.2
top5
BIAS
GPV
BIAS
( ) ( )
RCC
0.5
0.6
0.7
0.8
0.9
1
0 12 24 36 48
RC
C
time
sameura RCC son 12h time-series
rawtop3top5top7ave9
CC
00.10.20.30.40.50.60.70.80.9
1
0 12 24 36 48
CC
time
sameura CC son 12h time-series
rawtop3top5top7ave9
RMSE
1.51.61.71.81.9
22.12.22.32.42.5
0 12 24 36 48
RM
SE
time
sameura RMSE son 12h time-series
rawtop3top5top7ave9
Fig. 17 RCC, CC and RMSE of Sameura dam basin
(autumn)
02468
10121416182022242628
0 2 4 6 8 10121416182022242628
GPV
RAP
sameura son 12h-12 scatter
rawtop3
0123456789
1011121314
0 1 2 3 4 5 6 7 8 9 1011121314
GPV
RAP
sameura son 12h-12 scatter
rawtop5
Fig. 18 Sameura dam basin (autumn), the relation of
the past record and forecast of precipitation
Fig.19 Fig.20 (
) 12 RCC
12 RCC 0.1
0.3 bot5 0.8
12
12 Fig.21
12 bot5
1 1
12 12 raw
bot5 raw
1 1
12 RCC raw
bot5 ( )
bot5 raw RCC
CC RMSE
12
12 Fig.22
3mm/h
RCC
0.6
0.7
0.8
0.9
1
0 12 24 36 48
RC
C
time
biwako RCC djf 12h time-series
rawave9bot7bot5bot3
CC
0.6
0.7
0.8
0.9
1
0 12 24 36 48
CC
time
biwako CC djf 12h time-series
rawave9bot7bot5bot3
RMSE
00.10.20.30.40.50.60.70.80.9
1
0 12 24 36 48
RM
SE
time
biwako RMSE djf 12h time-series
rawave9bot7bot5bot3
Fig. 19 RCC, CC and RMSE of Biwako basin (win-
ter)
RCC
0.5
0.6
0.7
0.8
0.9
1
0 12 24 36 48
RC
C
time
kurobe RCC djf 12h time-series
rawave9bot7bot5bot3
CC
0.5
0.6
0.7
0.8
0.9
1
0 12 24 36 48
CC
time
kurobe CC djf 12h time-series
rawave9bot7bot5bot3
RMSE
00.10.20.30.40.50.60.70.80.9
11.11.2
0 12 24 36 48
RM
SE
time
kurobe RMSE djf 12h time-series
rawave9bot7bot5bot3
Fig. 20 RCC, CC and RMSE of Kurobe dam basin
(winter)
bot5
GPV
0h-12h precipitaionof 12 hours average
0
1
2
3
4
5
0 1 2 3 4 5
GPV
RAP
biwako djf 12h-12 scatter
rawbot5
36h-48hpreticipation of 12
hours average
0
1
2
3
4
5
0 1 2 3 4 5G
PV
RAP
biwako djf 12h-48 scatter
rawbot5
Fig. 21 Biwako basin (winter), the relation of the
past record and forecast of precipitation
0h-12h preticipationof 12 hours average
0
1
2
3
4
5
6
7
0 1 2 3 4 5 6 7
GPV
RAP
kurobe djf 12h-12 scatter
rawbot5
36h-48hpreticipation of 12
hours average
0
1
2
3
4
5
6
7
0 1 2 3 4 5 6 7
GPV
RAP
kurobe djf 12h-48 scatter
rawbot5
Fig. 22 Kurobe dam basin (winter), the relation of
the past record and forecast of precipitation
GPV
BIAS 1
GPV
( )
GPV
BIAS 1
GPV
Biwako(autumn)
0123456789
1011
0 1 2 3 4 5 6 7 8 9 10 11
GPV
RAP
biwako son 3h-9 scatter
1st latest4th latest
Sameura(winter)
0123456789
10
0 1 2 3 4 5 6 7 8 9 10
GPV
RAP
sameura djf 3h-3 scatter
1st latest4th latest
Fig. 23 the case of good CC and RMSE of the fore-
cast value of early initial time compared with
that of the latest forecast value
5. GPV
5.1GPV
12
12 GPV
GPV
4 5
3 12
RCC RMSE
(
0h 3h) ( 3h 6h) ( 3h 6h)
Fig.24 ( )
CC RMSE (1st latest )
2 (2nd latest
) RCC
4mm/h
1 3mm/h
2nd latest
2mm/h
( ) 14mm/h
6mm/h
2nd latest 2mm/h
( ) 11mm/h
3mm/h 2nd
Biwakobasin(autumn)
0123456789
10
0 1 2 3 4 5 6 7 8 9 10
GPV
RAP
biwako son 3h-3 scatter
1st latest2nd latest5th latest
Sameura dambasin(autumn)
02468
10121416182022242628
0 2 4 6 8 10121416182022242628
GPV
RAP
sameura son 3h-6 scatter
1st latest2nd latest4th latest
Matsubara dambasin(autumn)
0123456789
10111213
0 1 2 3 4 5 6 7 8 9 10 11 12 13
GPV
RAP
matsubara son 3h-6 scatter
1st latest2nd latest4th latest
Fig. 24 the comparison of the time of good RCC and
the time of good CC and RMSE
latest
5.2
RCC CC RMSE
GPV
2
GPV
Table 2 2
RCC RCC
1-2 1st latest 2nd latest
RCC 1st latest
0.01 0.02 0.03
0.02
1st latest
1st latest
1st latest
RCC
RCC
RCC 0.03
Fig.25 ( ,6h-9h)
2 RCC
( 1st latest+3rd
latest)
3mm/h
2
1mm/h
2mm/h
1mm/h
RCC
CC RMSE
CC 0.6244 0.5729 RMSE 0.56
0.59
RCC
Fig.26 ( ,0h-3h)
0mm/h
10mm/h 4mm/h
1 2mm/h
1mm/h
CC RMSE
Table 2 the time that the average of two forecast value is better than the latest forecast value
Siroyama Kurobe Biwako Sameura Matsubaraspring 0h-3hspring 3h-6h 1-2 1-3spring 6h-9h 1-2 1-2 1-2spring 9-12h 1-3 1-2summer 0h-3h 1-2 1-4summer 3h-6h 1-2 1-2 1-3 1-2summer 6h-9h 1-2 1-2 1-3 1-2summer 9-12h 1-2 1-2 1-3 1-2autumn 0h-3h 3-5 1-3 1-2 1-4autumn 3h-6h 1-2 1-3 1-3 1-2 1-2autumn 6h-9h 1-2 1-4 1-2 1-2 1-3autumn 9-12h 1-3 1-2 1-2 1-3winter 0h-3h 1-3 1-2winter 3h-6h 1-2 1-2winter 6h-9h 1-2winter 9-12h 1-2 1-2 1-2 1-3
Fig.27 ( ,3h-6h)
1mm/h
1mm/h
CC
Fig.28 ( ,9h-12h)
5mm/h
0mm/h
6mm/h
1 2mm/h
CC RMSE
Fig.29 ( ,6h-9h)
1mm/h
6mm/h
1mm/h
1mm/h
1mm/h
CC RMSE
CC RMSE
RCC
0
1
2
3
4
5
6
7
8
9
0 1 2 3 4 5 6 7 8 9
GPV
RAP
biwako jja 3h-9 scatter
latest,shiftlatest
comb,shiftcomb
0
1
2
3
0 1 2 3
GPV
RAP
biwako jja 3h-9 scatter
latest,shiftcomb,shift
Fig. 25 Biwako basin(summer,6h-9h), the validation
of the latest forecast value and the average
of two forecast value
0123456789
10
0 1 2 3 4 5 6 7 8 9 10
GPV
RAP
biwako son 3h-3 scatter
latestcomb
latest,shiftcomb,shift
0
1
2
3
4
0 1 2 3 4
GPV
RAP
biwako son 3h-3 scatter
latest,shiftcomb,shift
Fig. 26 Biwako basin(autumn,0h-3h), the validation
of the latest forecast value and the average
of two forecast value
2 RCC
RCC
RCC
0
1
2
3
4
5
6
7
8
0 1 2 3 4 5 6 7 8
GPV
RAP
kurobe son 3h-6 scatter
latestcomb
latest,shiftcomb,shift
0
1
2
3
0 1 2 3
GPV
RAP
kurobe son 3h-6 scatter
latest,shiftcomb,shift
Fig. 27 Kurobe dam basin(autumn,3h-6h), the val-
idation of the latest forecast value and the
average of two forecast value
0123456789
101112131415
0 1 2 3 4 5 6 7 8 9 101112131415
GPV
RAP
sameura son 3h-12 scatter
latestcomb
latest,shiftcomb,shift
0
1
2
3
0 1 2 3
GPV
RAP
sameura son 3h-12 scatter
latestcomb
latest,shiftcomb,shift
Fig. 28 Sameura dam basin(autumn,9h-12h), the
validation of the latest forecast value and the
average of two forecast value
0123456789
10
0 1 2 3 4 5 6 7 8 9 10
GPV
RAP
matsubara son 3h-9 scatter
latestcomb
latest,shiftcomb,shift
0
1
2
3
4
5
6
0 1 2 3 4 5 6
GPV
RAP
matsubara son 3h-9 scatter
latest,shiftcomb,shift
Fig. 29 Matsubara dam basin(autumn,6h-9h), the
validation of the latest forecast value and the
average of two forecast value
CC RMSE
6.
GPV
•
•
2
GPV
GPV 2
1
GPV
GPV
2 GPV 12
51
GPV
GPV
: - -.
:
, 1996.
, , , :
, ,
Vol.268, pp.83-90, 1997.3.
.
Validation of JMA numerical prediction data (GPV) by statistical analysis
Kenji YAMADA*, Shuichi IKEBUCHI, Kenji TANAKA and Kazuyoshi SOUMA* Graduate School of Engineering, Kyoto University
Synopsis
It is important to predict rainfall with high accuracy in dam basins because rainfall prediction isnecessary to control and operate dams properly. Major methods for prediction are kinematic or physical.Japan meteorological agency (JMA) numerical forecasting is one of physical prediction methods.Gridpoint value (GPV) is output of JMA numerical forecasting. Its resolution is insufficient to reproducephenomena unique to mountainous regions. Therefore, downscaling by another high-resolution rainfallforecasting model is a major method to advance accuracy in a lot of researches. In this study, it isexamined how accurate GPV is, and formulated how to take advantage of GPV efficiently.
Keywords : GPV,rainfall prediction,RSM,dam
,316,56-60 2005
Tarek Merabtene
(MSM)GPV
. ,47,91-96 2003
: GPV
, ,273,40-46 1997.5.
