Upload
akeem-graves
View
22
Download
5
Tags:
Embed Size (px)
DESCRIPTION
National Institute of Meteorological Research. Impact of ProbeX-IOP (KEOP) observations on the predictive skill of heavy rainfall in the middle part of Korea. Hee-Sang Lee and Seung-Woo Lee Forecast Research Laboratory / National Institute of Meteorological Research, KMA. Background. - PowerPoint PPT Presentation
Citation preview
Impact of ProbeX-IOP (KEOP) observations on the Impact of ProbeX-IOP (KEOP) observations on the predictive skill of heavy rainfall in the middle part of predictive skill of heavy rainfall in the middle part of
KoreaKorea
Hee-Sang Lee and Seung-Woo LeeHee-Sang Lee and Seung-Woo Lee
Forecast Research Laboratory / National Institute of Forecast Research Laboratory / National Institute of Meteorological Research, KMAMeteorological Research, KMA
National Institute of Meteorological ResearchNational Institute of Meteorological Research
BackgroundBackground
KMA has been using the NCAR/PSU MM5 as a regional model KMA has been using the NCAR/PSU MM5 as a regional model for over 10 years.for over 10 years.
KMA considers the WRF model as a candidate of the operational KMA considers the WRF model as a candidate of the operational regional model.regional model.
Assessment of WRF model performance for very-short range Assessment of WRF model performance for very-short range forecasting of precipitation is demanded by forecasters.forecasting of precipitation is demanded by forecasters.
12h Precip. ETS for 12h fcst(June, 2007)
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.1mm 5mm 15mm 25mm 50mm
Threshold Value
MM5- 30WRF- 10
12h Precip. Bias Score for 12h fcst(June, 2007)
00.20.40.60.8
11.21.41.61.8
2
0.1mm 5mm 15mm 25mm 50mm
Threshold Value
MM5- 30WRF- 10
MM5-30 vs WRF-10 kmMM5-30 vs WRF-10 km12h Precip. ETS for 24h fcst
(June, 2007)
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.1mm 5mm 15mm 25mm 50mm
Threshold Value
MM5- 30WRF- 10
12h Precip. Bias Score for 24h fcst(June, 2007)
00.20.40.60.8
11.21.41.61.8
2
0.1mm 5mm 15mm 25mm 50mm
Threshold Value
MM5- 30WRF- 10
MM5- 30 km MM5- 30 km MM5- 10 km MM5- 10 km
MM5- 5 km MM5- 5 km WRF 10 km WRF 10 km
AWS observed rainfall AWS observed rainfall
WRF 3.3 km WRF 3.3 km
[2007. 7.3. 21 KST ~ 7. 4. 12 KST][2007. 7.3. 21 KST ~ 7. 4. 12 KST]
[2007. 7. 4. 00 KST ~ 12 KST][2007. 7. 4. 00 KST ~ 12 KST]
Predicted rainfall from two different regional modelsPredicted rainfall from two different regional models
Observations : 4 July 2007Observations : 4 July 2007
No warning by this time in the routine forecasting.
Observations : 4 July 2007Observations : 4 July 2007
12-h rainfall amount 12-h rainfall amount (2007/07/04 00LST ~12LST)(2007/07/04 00LST ~12LST)
IR (2007/07/04 06LST)
SFC (2007/07/04 09LST)
Heavy rainfall event : 09LST 4 July 2007Heavy rainfall event : 09LST 4 July 2007
Mungyung 148.5 Mungyung 148.5 mm/12hmm/12h
Anyang 104 Anyang 104 mm/12hmm/12h
60 min. acc.60 min. acc.
15 min. acc.15 min. acc.
At early morning 4th July, a convective system associated with the Changma front that produced heavy rainfall over the southern part of Korea moved eastward, then local heavy rainfall occurred over the middle part of Korea.
Operational models did not capture this signals over this area.
CAPPI (2007/07/04 06LST)
Observations : 9-12UTC 3 July 2007Observations : 9-12UTC 3 July 2007
111 ASOS111 ASOS
773 AWS data773 AWS data
19 Radiosondes19 Radiosondes
240 AMDAR240 AMDAR
451 AMDAR from Korean Airlines (KAL)451 AMDAR from Korean Airlines (KAL)
10 Wind profiler10 Wind profiler
5 SATEM5 SATEM
Special observations for impact studiesSpecial observations for impact studies
HaenamHaenam
SokchoSokcho
MunsanMunsan
PohangPohang
BaengnyeongdBaengnyeongdoo
IeodoIeodo
GosanGosan
OsanOsan
HuksandoHuksando
Conventional Conventional
KEOPKEOP
Air ForceAir Force
GwangjuGwangju
ProbeXProbeX-2007 IOP-2007 IOP - - Observing period : Observing period :
2007/06/15 ~ 2007/07/152007/06/15 ~ 2007/07/15
- - Increasing time resolutionIncreasing time resolution : : 4 times/day 4 times/day (Baengnyeongdo, Sokcho, (Baengnyeongdo, Sokcho, Huksando, Pohang, Gosan) Huksando, Pohang, Gosan)
- - Increase space resolutionIncrease space resolution : : Additional enhanced Additional enhanced observation (Munsan, observation (Munsan, Haenam, Ieodo)Haenam, Ieodo)
Probex (PRedictability and Probex (PRedictability and
OBservation ExperimentOBservation Experiment
in Korea)in Korea)
MM5-30 & KWRF-10 kmMM5-30 & KWRF-10 km KWRF 3.3 kmKWRF 3.3 km
Physical processesPhysical processes Domain 1 (10 km)Domain 1 (10 km) Domain 2 (3.3 km)Domain 2 (3.3 km) RemarksRemarks
DimensionsDimensions 574 X 514 (with 30 vertical levels)574 X 514 (with 30 vertical levels) 334 X 364 (with 30 vertical 334 X 364 (with 30 vertical levels)levels)
Run time on CRAY-X1ERun time on CRAY-X1E(1024CPUs / 18.4TFLOPS)(1024CPUs / 18.4TFLOPS)
Domain 1 : Domain 1 : 14 min.14 min. with 126 CPUswith 126 CPUsDomain 2 : 5Domain 2 : 50 min.0 min. with 64 CPUswith 64 CPUs
Time interval (Time interval (ΔΔt)t) 60 sec60 sec 20 sec20 sec
Cumulus Cumulus ParameterizationParameterization Kain-Fritsch (new Eta) schemeKain-Fritsch (new Eta) scheme NoneNone
MicrophysicsMicrophysics WSM6 / WSM5 / WSM3 / new EtaWSM6 / WSM5 / WSM3 / new Eta WSM 6-class schemeWSM 6-class scheme
PBLPBL YSU schemeYSU scheme YSU schemeYSU scheme
RadiationRadiation RRTM / Dudhia schemeRRTM / Dudhia scheme RRTM / Dudhia schemeRRTM / Dudhia scheme
Surface-LandSurface-Land Noah LSMNoah LSM Noah LSMNoah LSM
Initial and Boundary dataInitial and Boundary data GDAPST426 hybrid-sigma(0.28125GDAPST426 hybrid-sigma(0.28125oo)) WRF 10 kmWRF 10 km
Model domains and configurationsModel domains and configurations
Verification areaVerification area
0000 0606 1212 1818 2424UTCUTC
Global (T426)
10 days forecast
CYCLE run
COLD run
3 days forecast
10 days forecast
60-h forecast
6-h forecast
60-h forecast
60-h forecast
Nestdown to WRF 3.3 km Nestdown to WRF 3.3 km
Experimental design Experimental design
3DVAR data assimilation3DVAR data assimilation
Nestdown to WRF 3.3 kmNestdown to WRF 3.3 km
3030 3636
Experiment ID Assimilated observation data Remarks( Number of assimilated obs. )
CTL All available observations without KEOP soundings Operational, 1247
ALL All available observations including KEOP soundings 1249
OPR Conventional TEMP soundings 17
TMP Conventional TEMP + KEOP soundings 19
KOP KEOP soundings only 2
PRF Wind profiler data 10
ACS AMDAR data from FSL 240
KAL AMDAR data including KAL reports 451
KON KAL reports only 211
SFC SYNOP, SHIP, BUOY, AWS data 884
SYN SYNOP, SHIP, BUOY 111
AWS AWS 773
SAT SATEM, SATOB, QSCAT 5
Observations for impact studiesObservations for impact studies
OBSOBS CTLCTL ALLALL
OPROPR TMPTMP IOPIOP
103103
T+12 acc rainfallT+12 acc rainfall
Munsan
Munsan
Munsan
The location of rainfall was slightly shifted toward observation when the IOP sounding (even in The location of rainfall was slightly shifted toward observation when the IOP sounding (even in one sounding at Munsan station) data was included.one sounding at Munsan station) data was included.
9898
104104
148.5148.5
100 km100 km
OBSOBS PRFPRF ACSACS
KALKAL SFCSFC SYNSYN AWSAWS
SATSAT
T+12 acc rainfallT+12 acc rainfall
Sounding data shows positive impact on the improvement of rainfall than the surface observation data.Sounding data shows positive impact on the improvement of rainfall than the surface observation data.
The aircraft data from KAL shows most skillful forecasting of precipitation.The aircraft data from KAL shows most skillful forecasting of precipitation.
100 km100 km
Sensitivity to boundary condition from global modelSensitivity to boundary condition from global model
FCST(C24H)FCST(C24H) ANAL_IOPANAL_IOP
OBSOBS ANALANALCTRL (operational)CTRL (operational)
6464
4343
9898
148.5148.5
104104
Since the BCs of WRF-0 are provided by the GDAPS, perfect BCs from global analyses Since the BCs of WRF-0 are provided by the GDAPS, perfect BCs from global analyses lead to an improvement of locations of heavy rainfall.lead to an improvement of locations of heavy rainfall.
100 km100 km
7676
6464
CTRLCTRL COLDCOLD
C12HC12H C24HC24H
Sensitivity to the cycle with WRF-10Sensitivity to the cycle with WRF-10
The cycle plays an important role in the spin-up in precipitation process.The cycle plays an important role in the spin-up in precipitation process.
OBSOBS
100 km100 km
7676
1111
2929
6464
148.5148.5
104104
Sensitivity to microphysics (WRF 10km) with ANAL_BCsSensitivity to microphysics (WRF 10km) with ANAL_BCs
WSM6WSM6OBSOBS WSM3WSM3
WSM5WSM5CTRL (WSM6)CTRL (WSM6)
9696 7474
42427676
148.5148.5
104104
100 km100 km ETA_NEW (Ferrier)ETA_NEW (Ferrier)
9494
Although the simulated rainfall amount was much smaller than the observed one, Although the simulated rainfall amount was much smaller than the observed one, ETA_NEW microphysics does better job in location of main rainfall area over the middle ETA_NEW microphysics does better job in location of main rainfall area over the middle part of Korea.part of Korea.
Sensitivity to microphysics (WRF 3.3km)Sensitivity to microphysics (WRF 3.3km)
OBSOBS
4141
148.5148.5
104104
ETA_NEW (Ferrier)ETA_NEW (Ferrier)
In higher resolution experiment, the magnitude of maximum rainfall is larger than that in In higher resolution experiment, the magnitude of maximum rainfall is larger than that in lower resolution but no difference in phase.lower resolution but no difference in phase.
WSM6WSM6
9797
100 km100 km
WSM3WSM3
128128
WSM5WSM5
109109
9393
The GA is a global optimization approach based on the Darwinian principles of natural selection. This method, developed from the concept of Holland [1975], aims to efficiently seek the extrema of complex function – see Goldberg [1989] for a detailed description.
Genetic Algorithm to optimize WRF-10 modelGenetic Algorithm to optimize WRF-10 model
Start
Initialization
Fitness Evaluation
Selection
Crossover
Mutation
Fitness Evaluation
Terminal condition
End
NO
YES
Variance and length scale of background error (xVariance and length scale of background error (x11, ,
xx22, x, x33, x, x44, x, x55, l, l11, l, l22, l, l33, l, l44, l, l55))
Asymptotic mixing length in PBL(mAsymptotic mixing length in PBL(m11))
Clear air turbulence : 10 – 30 mClear air turbulence : 10 – 30 m
Cyclogenesis in upper troposphere : < 100mCyclogenesis in upper troposphere : < 100m
Closure assumption of KF (mClosure assumption of KF (m22))
In the Kain-Fritsch scheme the closure assumption is that convection In the Kain-Fritsch scheme the closure assumption is that convection consumes at least 90% of the environmental convective available consumes at least 90% of the environmental convective available potential energy (CAPE) over an advective time period ( 30 min ~ 1 potential energy (CAPE) over an advective time period ( 30 min ~ 1
hour) [hour) [Kain et al.Kain et al. 2003]. 2003].
96.002.0 2 m
20010 1 m
0.3,5.0 nn lx
Selection of ChromosomesSelection of Chromosomes
The function to be optimized (i.e., Fitness) is defined by using a QPF skill score, the equitable treat score (ETS) [Schaefer, 1990],
Fitness = ,
i
iETS 100,,2,1 i
RHOF
RHETS
where i is the precipitation threshold in mm. Here, the ETS is defined as:
H : hitR : the expected number of hits in a random forecast F : rain forecast O : rain observation
NFOR /
Fitness functionFitness function
Variance of control variables
var_scaling1 (x1, Ψ) var_scaling2 (x2,χ) var_scaling3 (x3,Tu) var_scaling4 (x4, qRH) var_scaling5 (x5,Pa)
1.32 2.68 1.34 0.96 2.36
Horizontal length scales
len_scaling1 (l1) len_scaling2 (l2) len_scaling3 (l3) len_scaling4 (l4) len_scaling5 (l5)
0.92 2.45 2.50 1.09 0.79
Physical parameters
asymptotic mixing length (m1) (30) reduction rate (m2) (0.95)
132.5 0.36
Evolution of chromosomesEvolution of chromosomes
CTRLCTRL
GAGA
Preliminary results w/ and w/o GA in WRF-10Preliminary results w/ and w/o GA in WRF-10
Overall the tuned WRF by GA works for locations of heavy rainfall.Overall the tuned WRF by GA works for locations of heavy rainfall.
OBSOBS
100 km100 km
148.5148.5
104104 50.950.9
64.564.5
54.154.1
SummarySummary
The assimilation of the intensive observations (KEOP-2007) with the The assimilation of the intensive observations (KEOP-2007) with the high resolution WRF model (3.3 km) and 3DVAR show a positive high resolution WRF model (3.3 km) and 3DVAR show a positive impact on the very-short range forecasting of heavy rainfall over Korea. impact on the very-short range forecasting of heavy rainfall over Korea.
Cycling processes to provide the background in 3DVAR play a crucial Cycling processes to provide the background in 3DVAR play a crucial role in spin-up of precipitation. role in spin-up of precipitation.
Improvement in boundary conditions from global model may lead to Improvement in boundary conditions from global model may lead to improvement in the forecast of heavy rainfall.improvement in the forecast of heavy rainfall.
Cloud microphysics plays an important role in the simulation of the Cloud microphysics plays an important role in the simulation of the heavy rainfall area in this case.heavy rainfall area in this case.