Seasonal Predictability Seasonal Predictability in East Asian Region in East Asian Region T argeted argeted T raining raining A ctivity: ctivity: S easonal easonal P redictability in redictability in T ropical ropical R egions: egions: R esearch and esearch and A pplications pplications 『 『 East Asian Group East Asian Group 』 』 Juhyun Park (Republic of Korea) Juhyun Park (Republic of Korea) Yanju Liu (China), qiaoping Li Yanju Liu (China), qiaoping Li (China) (China) N. Jyothi (India), A. P. N. Jyothi (India), A. P. Dimri (India) Dimri (India)
T argeted T raining A ctivity: S easonal P redictability in T ropical R egions: R esearch and A pplications. Seasonal Predictability in East Asian Region. - PowerPoint PPT Presentation
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Seasonal Predictability Seasonal Predictability in East Asian Region in East Asian Region
3. Deterministic Forecast skill in MME3. Deterministic Forecast skill in MME
Correlation skill between Observation and Multi-Model Correlation skill between Observation and Multi-Model
MSLP MSLP JJAJJA
DJFDJF
PRCP PRCP JJAJJA
DJFDJF
TA2M TA2M JJAJJA
DJFDJF
3. Deterministic Forecast skill in MME3. Deterministic Forecast skill in MME
Indian ocean index Indian ocean index JJAJJA
DJFDJF
The Indian Ocean The Indian Ocean RegionRegion
Lon. : 40 E ~ 110 ELon. : 40 E ~ 110 E
Lat. : 15 S ~ 10 NLat. : 15 S ~ 10 N
Var. : SST anomalyVar. : SST anomaly
Black line : Black line : Observation indexObservation index
Red line : the index in Red line : the index in DMMEDMME
3. Deterministic Forecast skill in MME3. Deterministic Forecast skill in MME
East Asia Summer Monsoon index East Asia Summer Monsoon index
The East Asia RegionThe East Asia Region
Black line : Observation Black line : Observation indexindex
Green line : the index in Green line : the index in DMMEDMME
)500:(
):(sin
45sin
)125,20(25.0
)125,40(50.0
)125,60(25.0
heightlgeopotentahPaZZZZ
latitudeZZ
ENZ
ENZ
ENZEASM
s
s
s
s
Correlation between EASM Index and Precipitation Correlation between EASM Index and Precipitation
Observation(left) MME(right)
Observation(left) MME(right)
Precipitation in strong monsoon year(1997)Precipitation in strong monsoon year(1997)
Observation(left) MME(right)
Precipitation in weak monsoon year(1998)Precipitation in weak monsoon year(1998)
4. Probabilistic Forecast skill in DMME4. Probabilistic Forecast skill in DMME
First Step : Climatological Probability Distribution Function
Climatological PDF
0 Xc-Xc
Second Step : Probability Forecast for particular time
- Normalizing all the forecast value )1,0(~),( NxZ
• In the 3 category case, -Xc and Xc make the below area separate 1/3 value each.
Ensemble PDF of particular time
BN NN AN
Below normal (BN)
Near normal (NN)
Above normal (AN)
Non-parametric approach
Parametric approach
dotstotal
dotsred
N
nP AAN
),(1
XcAP mGP
/ For Above normal case /
m : Ensemble mean value for particular year
4. Probabilistic Forecast skill in DMME4. Probabilistic Forecast skill in DMME
TS2M PRCP
1982 winter mean
1985 winter mean
Reliability diagramReliability diagram: graphically represent the performance of probability forecasts of
dichotomous events for each category
The plot of observed relative frequency as a function of forecast probability :
The 1:1 diagonal perfect reliability line : A summary of the frequency of use of each forecast value
4. Probabilistic Forecast skill in DMME4. Probabilistic Forecast skill in DMME
qqf )(
qagainstqfofvaluethe )(
Black, dashed line : Each Black, dashed line : Each modelmodel
Red, solid line : DMMERed, solid line : DMME
Above Normal Category ( Global Above Normal Category ( Global Region )Region )
Precipitation / JJAPrecipitation / JJA
4. Probabilistic Forecast skill in DMME4. Probabilistic Forecast skill in DMME
Brier Score (B)Brier Score (B)
Brier skill score (BSS)Brier skill score (BSS)
n : the number of realizations of the forecasts over
which the validation is preformed
For each realization i ,
pi : forecast probability of the occurrence of the event
vi : a value equal to 1 or 0
depending on the event occurred / not.
refB
BBSS 1
Bref : a reference forecast
(taken to be the low-skill climatological forecasts)
BSS = 1 : a perfect forecast system
BSS = 0 (negative) : performs like (poorer than) the reference system
n
iii vp
nB
1
2)(1
4. Probabilistic Forecast skill in DMME4. Probabilistic Forecast skill in DMME
CERCERFF
ECMECMWW
INGINGVV
LODLODYY
MAXMAXPP
METMETFF
UKMUKMOO
Above Normal Category Above Normal Category
Precipitation / JJAPrecipitation / JJA
DMMDMMEE
4. Probabilistic Forecast skill in DMME4. Probabilistic Forecast skill in DMME
L
Cr
roro
orophorpfroV tt
,),min(
)1()()1()(),min(
Economic valueEconomic value
Observation (real event)
Yes No
Forecast(Action)
YesHit (h)
Cost (C)False (f)Cost (C)
NoMiss (m)Loss (L)
Correct reject
0
),(min, oL
CE
L
CoE
EE
EEV
mcliperf
perfimlc
fcstimcl
V = 1 : a perfect forecast system
V = 0 : performs like the reference system
* pt : threshold probability
4. Probabilistic Forecast skill in DMME4. Probabilistic Forecast skill in DMME
CERCERFF
ECMECMWW
INGINGVV
LODLODYY
MAXMAXPP
METMETFF
UKMUKMOO
Above Normal Category Above Normal Category
Precipitation / JJAPrecipitation / JJA
DMMDMMEE
The Multi-model shows the better predictability than the single model following this study.
But, the forecast skill is different about the variables and the target region. This is the same results as in the deterministic forecast.
Probability forecasts show more information for users about future climate than deterministic forecast. Because this contains the uncertainty in the forecast problem.