caio a. s. coelho centro de previs ã o de tempo e estudos clim á ticos (cptec)
DESCRIPTION
The EUROBRISA operational system. Caio A. S. Coelho Centro de Previs ã o de Tempo e Estudos Clim á ticos (CPTEC) Instituto Nacional de Pesquisas Espaciais (INPE) [email protected]. PLAN OF TALK 1. Introduction 2. EUROBRISA integrated forecasting system 3. Forecasts for 2007-2008 - PowerPoint PPT PresentationTRANSCRIPT
Caio A. S. CoelhoCentro de Previsão de Tempo e Estudos Climáticos (CPTEC)
Instituto Nacional de Pesquisas Espaciais (INPE)[email protected]
1st EUROBRISA workshop, Paraty, 17-19 March 2008
PLAN OF TALK1. Introduction2. EUROBRISA integrated forecasting system3. Forecasts for 2007-20084. Skill of the hindcasts 5. Summary
The EUROBRISA operational system
1. Seasonal climate forecasts
•
Current forecast approaches• Empirical/statistical models• Dynamical atmospheric models• Dynamical coupled (ocean-atmosphere) models
Forecasts of climate conditions for the next 3-6 months
• • •• • •Nov Dec Jan Feb MarApr May
DJF
0 1 2 3 4 5 61-month lead for DJF
Empirical modelPredictors: Atlantic e Pacific SSTPredictand: Precipitation
Hindcast period: 1987-2001
2. EUROBRISA integrated forecasting system for South America
Integrated
U.K.UKMO
InternationalECMWF System 3
CountryCoupled model
Combined and calibrated coupled + empirical precip. forecastsHybrid multi-model probabilistic system
The Empirical model
Y|Z ~ N (M (Z - Zo),T)
TYZ
1ZZYZYY
o
1ZZYZ
SSSST
MZYZM
SSM
Y: DJF precipitation
Z: October sea surface temp. (SST)
Model uses first three leading Maximum CovarianceAnalysis (MCA) modes of the matrix YT Z.
vnZ
qnY
:
:
qqT :
Y Z
Coelho et al. (2006)
Data sources:• SST: Reynolds OI v2
Reynolds et al. (2002)
• Precipitation: GPCP v2Adler et al. (2003)
5
First mode (71%)
Second mode (7.7%)
Observed Oct 2007 SST
DJF 2007 forecast Corr. DJF
Empirical forecast: DJF 2007/08
Issued: November 2007
Conceptual framework
)y(p
)x(p)x|y(p)y|x(p
i
iiiii
Data Assimilation “Forecast Assimilation”
)x(p
)y(p)y|x(p)x|y(p
f
fffff
Stephenson et al. (2005)
7
),(~ CYNY b
1
111
)(
)()(
))((
SGCGCGL
CLGICGSGD
YYGXLYY
TT
T
obba
)),((~| SYYGNYX o
Prior:
Likelihood:
Posterior:
1 YYXY SSGGYXGYo
TYYXX GGSSS
),(~| DYNXY a
Calibration and combination procedure: Forecast Assimilation
qq:D
qn:Y
pn:X
qq:C q1:Yb
pp:S qn:Ya
Matrices
Forecast assimilation uses the first three MCA modes of the matrix YT X.
X: forecasts (coupled + empir.)Y: DJF precipitation
)(
)()|()|(
Xp
YpYXpXYp
Stephenson et al. (2005)
8Web site launched in Oct 2007: http://www6.cptec.inpe.br/eurobrisa/
Real time and verification products
1-month lead precip. forecastsEUROSIP: ECMWF UKMO Meteo-FranceEmpirical (SST based)Integrated (Combined)
10
Empirical Integrated
Examples of forecast productsProbability of most likely precip. tercile:
DJF 2007/08
Issued: Nov 2007
ECMWF UKMO
EUROBRISA integrated forecast for AMJ 2007
Issued: March 2007
Gerrity score(tercile
categories)
Prob. of most likelyprecip. tercile (%)
Hindcasts: 1987-2001
Observed precip.tercile
Obs. SST anomaly Feb 2007
EUROBRISA integrated forecast for MJJ 2007
Hindcasts: 1987-2001
Prob. of most likelyprecip. tercile (%)
Observed precip.tercile
Issued: April 2007
Obs. SST anomaly Mar 2007
Gerrity score(tercile
categories)
EUROBRISA integrated forecast for JJA 2007
Issued: May 2007
Obs. SST anomaly Apr 2007
Prob. of most likelyprecip. tercile (%)
Observed precip.tercile
Gerrity score(tercile
categories)
Hindcasts: 1987-2001
EUROBRISA integrated forecast for JAS 2007
Issued: Jun 2007
Hindcasts: 1987-2001
Gerrity score(tercile
categories)
Prob. of most likelyprecip. tercile (%)
Obs. SST anomaly May 2007
Observed precip.tercile
Hindcasts: 1987-2001
EUROBRISA integrated forecast for ASO 2007
Obs. SST anomaly Jun 2007
Issued: Jul 2007
Gerrity score(tercile
categories)
Prob. of most likelyprecip. tercile (%)
Observed precip.tercile
EUROBRISA integrated forecast for SON 2007
Obs. SST anomaly Jul 2007
Issued: Aug 2007
Hindcasts: 1987-2001
Prob. of most likelyprecip. tercile (%)
Gerrity score(tercile
categories)
Observed precip.tercile
EUROBRISA integrated forecast for OND 2007
Issued: Sep 2007
Obs. SST anomaly Aug 2007
Hindcasts: 1987-2001
Prob. of most likelyprecip. tercile (%)
Observed precip.tercile
Gerrity score(tercile
categories)
EUROBRISA forecastsfor NDJ 2007/08
Issued: Oct 2007
Obs. SST anomaly Sep 2007
Integrated UKMOEmpirical
Prob. of most likely precip. tercile (%)
ECMWF
EUROBRISA forecastsfor DJF 2007/08
Issued: Nov 2007
Obs. SST anomaly Oct 2007
Integrated UKMOEmpirical
Prob. of most likely precip. tercile (%)
ECMWF
EUROBRISA forecastsfor JFM 2008
Issued: Dec 2007
Obs. SST anomaly Nov 2007
Integrated UKMOEmpirical
Prob. of most likely precip. tercile (%)
ECMWF
EUROBRISA forecastsfor FMA 2008
Issued: Jan 2008
Obs. SST anomaly Dec 2007
Integrated UKMOEmpirical
Prob. of most likely precip. tercile (%)
ECMWF
EUROBRISA forecastsfor MAM 2008
Issued: Feb 2008
Obs. SST anomaly Jan 2008
Integrated UKMOEmpirical
Prob. of most likely precip. tercile (%)
ECMWF
27
Empirical IntegratedCorrelation btw. obs. and fcst. DJF precip. anom.
UKMOECMWF
Examples of verification products
• Hindcast period: 1987-2001• Coupled models with I.C. 1st Nov (1-month lead for DJF)• Empirical model uses Oct SST as predictor for DJF precip.• Integrated forecasts (coupled + empirical) with forecast assimilation
Best skill in tropical and southeast South America
28
Empirical Integrated
Brier Skill Score (pos. or neg. anomaly): DJF precipitation
UKMOECMWF
• Hindcast period: 1987-2001• Coupled models with I.C. 1st Nov (1-month lead for DJF)• Empirical model uses Oct SST as predictor for DJF precip.• Integrated forecasts (coupled + empirical) with forecast assimilation
limcBS
BS1BSS
n2
k kk 1
1BS (p o )
n
29
Empirical Integrated
Reliability diagram (pos. or neg. anomaly): DJF precipitation
UKMOECMWF
• Hindcast period: 1987-2001• Coupled models with I.C. 1st Nov (1-month lead for DJF)• Empirical model uses Oct SST as predictor for DJF precip.• Integrated forecasts (coupled + empirical) with forecast assimilation
30
Empirical Integrated
ROC curve (pos. or neg. anomaly): DJF precipitation
UKMOECMWF
• Hindcast period: 1987-2001• Coupled models with I.C. 1st Nov (1-month lead for DJF)• Empirical model uses Oct SST as predictor for DJF precip.• Integrated forecasts (coupled + empirical) with forecast assimilation
31
Empirical Integrated
ROC skill score (pos. or neg. anomaly): DJF precipitation
UKMOECMWF
• Hindcast period: 1987-2001• Coupled models with I.C. 1st Nov (1-month lead for DJF)• Empirical model uses Oct SST as predictor for DJF precip.• Integrated forecasts (coupled + empirical) with forecast assimilation
1A2ROCSS A is the area under the ROC curve
32
Empirical IntegratedUKMOECMWF
• Hindcast period: 1987-2001• Coupled models with I.C. 1st Nov (1-month lead for DJF)• Empirical model uses Oct SST as predictor for DJF precip.• Integrated forecasts (coupled + empirical) with forecast assimilation
limcRPS
RPS1RPSS 3K;BSRPS
K
1mm
Ranked probability skill score (tercile categories): DJF precipitation
33
Empirical Integrated
Gerrity score (tercile categories): DJF precipitation
UKMOECMWF
• Hindcast period: 1987-2001• Coupled models with I.C. 1st Nov (1-month lead for DJF)• Empirical model uses Oct SST as predictor for DJF precip.• Integrated forecasts (coupled + empirical) with forecast assimilation
5. Summary•EUROBRISA integrated forecasting system: First operational hybrid (empirical-dynamical) probabilistic seasonal forecasting system for South America
•Current operational system: SST-based empirical model + two dynamical coupled models (ECMWF and UKMO)
•Good performance in 2007 over regions where forecasts have historically moderate to good skill
•Web products include a range of forecast and verification products for the EUROBRISA integrated forecasting system in addition to Meteo-France coupled model forecasts
•Additional information at http://www6.cptec.inpe.br/eurobrisa and in Coelho et al.(2007)-CLIVAR Exchanges No 43 (Volume 12 No 4)
35
•Adler, R.F., G.J. Huffman, A. Chang, R. Ferraro, P. Xie, J. Janowiak, B. Rudolf, U. Schneider, S. Curtis, D. Bolvin, A. Gruber, J. Susskind, P. Arkin (2003), The Version 2 Global Precipitation Climatology Project (GPCP) Monthly Precipitation Analysis (1979-Present). J. Hydrometeor., 4,1147-1167.•Coelho C.A.S., D. B. Stephenson, F. J. Doblas-Reyes, M. Balmaseda and R. Graham, 2007: Integrated seasonal climate forecasts for South America. CLIVAR Exchanges. No.43. Vol. 12, No. 4, 13-19.• Coelho C.A.S., D. B. Stephenson, F. J. Doblas-Reyes and M. Balmaseda, 2005: From multi-model ensemble predictions to well-calibrated probability forecasts: Seasonal rainfall forecasts over South America 1959-2001 CLIVAR Exchanges. No.32. Vol. 10, No. 1, 14-20.•Coelho C.A.S., D. B. Stephenson, M. Balmaseda, F. J. Doblas-Reyes and G. J. van Oldenborgh, 2006: “Towards an integrated seasonal forecasting system for South America”.J. Climate., Vol. 19, 3704-3721.•Reynolds, R. W., N. A. Rayner, T. M. Smith, D. C. Stokes and W. Wang (2002), An improved in situ and satellite SST analysis for climate. J. Climate, 15, 1609-1625.•Stephenson, D. B., C.A.S. Coelho, F. J. Doblas-Reyes, and M. Balmaseda, 2005:“Forecast Assimilation: A Unified Framework for the Combination of Multi-Model Weather and Climate Predictions.” Tellus A, Vol. 57, 253-264.
References