Upload
neve-robbins
View
86
Download
11
Tags:
Embed Size (px)
DESCRIPTION
WRF-based short-range forecast system of the Korea Air Force : v erification of prediction skill in 2009 summer. Ui -Yong Byun , Song-You Hong, Hyeyum Shin Deparment of Atmospheric Science, Yonsei Univ. Ji -Woo Lee, Jae- Ik Song, Sook -Jung Ham, Jwa-Kyum Kim, Hyung -Woo Kim - PowerPoint PPT Presentation
Citation preview
WRF-based short-range forecast system of the Korea
Air Force : verification of prediction
skill in 2009 summer
Ui-Yong Byun, Song-You Hong, Hyeyum ShinDeparment of Atmospheric Science, Yonsei Univ.
Ji-Woo Lee, Jae-Ik Song, Sook-Jung Ham, Jwa-Kyum Kim, Hyung-Woo Kim73rd Weather Group, Republic of Korea Air Forece
Numerical Modeling Laboratory, Department of Atmospheric Sciences, Yonsei University
Introduction
Configuration of KAF-WRF
Configuration of verification system
Results
Further study
Summary
Outline
Numerical Modeling Laboratory, Department of Atmospheric Sciences, Yonsei University
WRF model is designed for both research and operational appli-cations.
Research of extreme weather in Korea peninsula using WRF model Lee et al, 2005 : Orographic effect for a heavy rainfall Lim et al, 2007 : Heavy snowfall over the Ho-Nam province
WRF model operation in forecast institution of Korea Jo et al, 2005 : KWRF construction and test run in KMA The 73rd Weather Group (73WG) of Republic of Korea Air Force (ROKAF) oper-
ates the KAF-WRF model based on Weather Research and Forecasting (WRF) model since 2007.
In this study, KAF-WRF model results in 2009 summer are evalu-ated using quantitative verification system.
Introduction
Numerical Modeling Laboratory, Department of Atmospheric Sciences, Yonsei University
Configuration of KAF-WRF
250 x 250211 x 211230 x 170Grids
24hr84hr
KAF-WRF V07 (based on WRFv2.2)Model version
31 LayerVertical Layer
None
Noah LSMLSM
RRTM(LW), Dudhia Scheme (SW)Radiation
YSU PBLPBL
Kain-FritcshCumulus
WSM6Microphysics
84hrFCST
2 km6 km18 kmResolution
DM 3DM 2DM 1
250 x 250211 x 211230 x 170Grids
24hr84hr
KAF-WRF V09 (based on WRFv3.1)Model version
31 LayerVertical Layer
None
Noah LSMLSM
RRTM-G(LW), Goddard SW(SW)Radiation
YSU PBLPBL
Kain-FritcshCumulus
WDM6Microphysics
84hrFCST
2 km6 km18 kmResolution
DM 3DM 2DM 1
WSM3
Operation model Experimental model
+ Ocean Mixed Layer+ MODIS Land use data
Numerical Modeling Laboratory, Department of Atmospheric Sciences, Yonsei University
Domain 1 Domain 2
18 km 6 km12 00 24 36 48 60 72 84
4 times/day84 hour fcst.
Numerical Modeling Laboratory, Department of Atmospheric Sciences, Yonsei University
Numerical Modeling Laboratory, Department of Atmospheric Sciences, Yonsei University
Configuration of verification system (domain 1)
00 12 24 36 48
1-day00 UTC :
1-day12 UTC :
Making difference data Monthly mean data
Field figure & score – SLP., 500hPa GPH., Temp., wind
2-day00 UTC :
Numerical Modeling Laboratory, Department of Atmospheric Sciences, Yonsei University
Configuration of verification system (domain 2)
Extracting precipitation field from model output
Changing the precipitation data from field to point
Extracting 1hr precipitation from AWS
data
Making 6hr precipitation data
Making skill scoreUsing contingency table
Model output process AWS data process
Numerical Modeling Laboratory, Department of Atmospheric Sciences, Yonsei University
Verification Model : KAF-WRF V07, V09 Period : 2009. JJA Parameter
Domain 1 : SLP., 500hPa GPH, Temperature, Wind
Domain 2 : 6 hour accumulated precipitation Statistics
Domain 1 : RMSE, Bias score Domain 2 : Skill score
Result
Numerical Modeling Laboratory, Department of Atmospheric Sciences, Yonsei University
Result - domain 1
2009. 07. 24hr fcst
9.438
7.768
500hPa GPH.SLP
KAF-WRF V07
KAF-WRF V09
1.593
1.515
Numerical Modeling Laboratory, Department of Atmospheric Sciences, Yonsei University
Result - domain 1
2009. 07. 24hr fcst
KAF-WRF V07
KAF-WRF V09
0.725
0.718
U : 2.837V : 2.821
U : 2.781V : 2.744
500hPa Temp. 500hPa Wind
Numerical Modeling Laboratory, Department of Atmospheric Sciences, Yonsei University
• Bias : Sea level pressure
• RMSE : Sea level pressure
06 12 18 24 30 36 42 480
0.5
1
1.5
2
2.5
3
3.5
4
KW07 JunKW07 JulKW07 AugKW09 JunKW09 JulKW09 Aug
06 12 18 24 30 36 42 48
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
KW07 JunKW07 JulKW07 AugKW09 JunKW09 JulKW09 Aug
06 12 18 24 30 36 42 480
1
2
3
4
KW07 JunKW09 Jun
06 12 18 24 30 36 42 480
1
2
3
4
OPR JulEXP Jul
06 12 18 24 30 36 42 480
0.51
1.52
2.53
3.54
OPR AugEXP Aug
Numerical Modeling Laboratory, Department of Atmospheric Sciences, Yonsei University
06 12 18 24 30 36 42 480
2
4
6
8
10
12
14
16
18
KW07 JunKW07 JulKW07 AugKW09 JunKW09 JulKW09 Aug
06 12 18 24 30 36 42 480
0.2
0.4
0.6
0.8
1
1.2
KW07 JunKW07 JulKW07 AugKW09 JunKW09 JulKW09 Aug
06 12 18 24 30 36 42 48
-4
-3
-2
-1
0
1
2
3
KW07 JunKW07 JulKW07 AugKW09 JunKW09 JulKW09 Aug
06 12 18 24 30 36 42 480
0.05
0.1
0.15
0.2
0.25
0.3
0.35
KW07 JunKW07 JulKW07 AugKW09 JunKW09 JulKW09 Aug
• Bias : 500hPa Geopotential Height
• RMSE : 500hPa Geopotential Height
• Bias : 500hPa Temperature
• RMSE : 500hPa Temperature
Numerical Modeling Laboratory, Department of Atmospheric Sciences, Yonsei University
• Bias : 500hPa u-wind
• RMSE : 500hPa u-wind
• Bias : 500hPa v-wind
• RMSE : 500hPa v-wind
06 12 18 24 30 36 42 48
-0.25
-0.2
-0.15
-0.1
-0.05
0
0.05
KW07 JunKW07 JulKW07 AugKW09 JunKW09 JulKW09 Aug
06 12 18 24 30 36 42 48
-0.3
-0.25
-0.2
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
0.2
KW07 JunKW07 JulKW07 AugKW09 JunKW09 JulKW09 Aug
06 12 18 24 30 36 42 480
0.5
1
1.5
2
2.5
3
3.5
4
4.5
KW07 JunKW07 JulKW07 AugKW09 JunKW09 JulKW09 Aug
06 12 18 24 30 36 42 480
0.51
1.52
2.53
3.54
4.55
KW07 JunKW07 JulKW07 AugKW09 JunKW09 JulKW09 Aug
Numerical Modeling Laboratory, Department of Atmospheric Sciences, Yonsei University
Contingency table
POD = H / (M + H) ; Probability of Detection FAR = F / (H + F) ; False Alarm Ratio Bias = (H + F) / (H + M) ; Bias Score ETS = (H – E) / (H + M + F – E)
( E = (H + F) x (H + M) / (H + M + F + C) ); Equitable Threat Score
Result - domain 2
Forecast
Yes No
Observati
on
Yes H M Observation yes
No F C Observation no
Forecast yes Forecast no
Numerical Modeling Laboratory, Department of Atmospheric Sciences, Yonsei University
Precipitation analysis
Result - domain 2
AWS KAF-WRF V09
2009. 07. 1-month precipitation
Numerical Modeling Laboratory, Department of Atmospheric Sciences, Yonsei University
’09. June 12hr fcst precipitation (6 hour accumulated)
0.5 1 3 5 7 10 15 20 250
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
KW07KW09
0.5 1 3 5 7 10 15 20 250
0.1
0.2
0.3
0.4
0.5
0.6
KW07KW09
0.5 1 3 5 7 10 15 20 250
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
KW07KW09
0.5 1 3 5 7 10 15 20 250
0.050.1
0.150.2
0.250.3
0.350.4
0.450.5
KW07KW09
• POD
•Bias
• FAR
• ETS
Found a problem with weak precipitation
Numerical Modeling Laboratory, Department of Atmospheric Sciences, Yonsei University
OBS : TMPA & FNL WSM6 exp. WDM6 exp.
Hong et al., 2010
Int : 2008.02.23 00 UTC, 36hr fcst, 6hr precip.
Nc Nr
A : B :
B
A
Numerical Modeling Laboratory, Department of Atmospheric Sciences, Yonsei University
Purpose of further study Finding the cause of low accuracy of KAF-WRF V09 on
weak precipitation. Improvement of accuracy of weak precipitation
Possibility 1 : Microphysics Microphysics is changed from V07 to V09
Further Study (1)
Domain 1 Domain 2
KAF-WRF V07 (OPR)
WSM6 WSM6
KAF-WRF V09 (EXP) WSM3 WDM6
Numerical Modeling Laboratory, Department of Atmospheric Sciences, Yonsei University
Int : 2010.08.24 00 UTC, Valid : 15 UTC, 3hr precip.
KAF-WRF V07WSM6-WSM6
KAF-WRF V09WSM3-WDM6
KAF-WRF V09WDM6-WDM6
KAF-WRF V09WSM6-WDM6
OPR EXP
Numerical Modeling Laboratory, Department of Atmospheric Sciences, Yonsei University
Possibility 2: Error of YSU PBL
Some error of YSU PBL was corrected in updated WRF model (ver. 3.2.1). Minor bug fixes for PBL Prandtl number calculation in stable and unstable
condition.
WRF model that based on WRF ver. 3.2.1 and that has same physics setting with ‘KAF-WRF V09’ , is defined ‘KAF-WRF V10’.
Select case - 2009. 07. 09. precipitation Initial time : 2009. 07. 08. 12 UTC
: 2009. 07. 09. 00 UTC : 2009. 07. 09. 12 UTC
Compare the verification score; KAF-WRF V07, V09, V10
Further Study (2)
Numerical Modeling Laboratory, Department of Atmospheric Sciences, Yonsei University
12hr fcst precipitation (6 hour accumulated)
• POD
•Bias
• FAR
• ETS0.5 1 3 5 7 10 15 20 25
00.10.20.30.40.50.60.70.80.9
1
V07V09V10
0.5 1 3 5 7 10 15 20 250
0.05
0.1
0.15
0.2
0.25
0.3
V07V09V10
0.5 1 3 5 7 10 15 20 250
0.2
0.4
0.6
0.8
1
1.2
V07V09V10
0.5 1 3 5 7 10 15 20 250
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
V07V09V10
Numerical Modeling Laboratory, Department of Atmospheric Sciences, Yonsei University
Summary
Quantitative verification system is constructed
RMSE and Bias score of SLP, 500hPa Geopotential Height, temper-ature and wind of the KAF-WRF V09 shows better performance than V07.
Verification result of precipitation shows different patterns de-pending on precipitation intensity Score of V07 is better than V09 in weak precipitation intensity (less than 3 mm/6hour)
Score of V09 is better than V07 in heavy precipitation intensity (more than 10 mm/6hour)
Accuracy of light-precipitation prediction is possible to increase adapting microphysics change and PBL debug.
ROKAF has plan that is changed EXP model instead of OPR model
Thank you
Numerical Modeling Laboratory, Department of Atmospheric Sciences, Yonsei University
Minor bug fixes for PBL Prandtl number calculation in stable and unstable condition.
PR = 1 + (PR_0 – 1) x exp(PR_fac)PR_0 = (ph_h/ph_m + prfac)
prfac = conpr / ph_m / (1 + 4 x karman * wstar3 / ust3) prfac = conpr / ph_m / (1 + 4 x karman * wstar3 / ust3)^h1 (h1 = 0.33333335) PR = momentum diffusivity(Km) / heat, moisture diffusivity(Kh)
(0.25 <= PR <= 4.0)
It means ‘prfac’ of new ver. has larger values in same con-dition. Also, ‘PR_0’ and ‘PR’ has larger values. In boundary layer, Km Kh ; Kh ↓ In free atmosphere and stable condition, Kh Km ; Km ↑
Numerical Modeling Laboratory, Department of Atmospheric Sciences, Yonsei University
KAF-WRF V07 WSM6 KAF-WRF V09 WSM3
KAF-WRF V09 WDM6KAF-WRF V09 WSM6
2010.08.24 00 UTC, 48 hour precip.
Numerical Modeling Laboratory, Department of Atmospheric Sciences, Yonsei University
KAF-WRF V07 KAF-WRF V09 KAF-WRF test
Int : 2009.07.09 00 UTC, Valid : 12 UTC, 6hr precip.