1 / 24

Caio A. S. Coelho Centro de Previs ã o de Tempo e Estudos Clim á ticos (CPTEC)

Towards a new 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) caio.coelho@cptec.inpe.br. PLAN OF TALK 1. Current operational system 2. Investigation on identified issues

xuan
Download Presentation

Caio A. S. Coelho Centro de Previs ã o de Tempo e Estudos Clim á ticos (CPTEC)

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Towards a new EUROBRISA operational system Caio A. S. Coelho Centro de Previsão de Tempo e Estudos Climáticos (CPTEC) Instituto Nacional de Pesquisas Espaciais (INPE) caio.coelho@cptec.inpe.br PLAN OF TALK 1. Current operational system 2. Investigation on identified issues 3. Advances on use of upper level circulation 4. Applications: river flow and dengue risk transmission prediction 5. Future plans 6. Summary 2nd EUROBRISA workshop, Dartmoor, Devon, 21-24 July 2009

  2. Coupled model Country ECMWF System 3 International UKMO (GloSea 3) U.K. EUROBRISA integrated forecasting system for South America • Combined and calibrated coupled + empirical precip. forecasts • Hybrid multi-model probabilistic system Integrated forecast Empirical model Predictors: Atlantic and Pacific SST Predictand: Precipitation Coelho et al. (2006) J. Climate, 19, 3704-3721 Hindcast period: 1987-2001

  3. Obs. SST anomaly Jun 2009 Most recentEUROBRISA integrated forecast for ASO 2009 Issued: Jul 2009 Prob. of most likely precip. tercile (%) UKMO Integrated Empirical ECMWF

  4. Calibration and combination procedure: Forecast Assimilation Stephenson et al. (2005) Tellus, 57A, 253-264 X: precip. fcsts (coupled + empir.) Y: DJF precipitation Prior: Likelihood: Matrices Posterior: Forecast assimilation uses the first three MCA modes of the matrix YT X.

  5. Calibration and combination procedure: Forecast Assimilation X: precip. fcsts (coupled + empir.) Y: DJF precipitation Stephenson et al. (2005) Tellus, 57A, 253-264 If prior param.: Matrices FA becomes: Posterior:

  6. Why is skill negative for some grid points? Correlation skill: Integrated forecast (precipitation) Issued: Nov Valid: DJF (1987-2001) Multivariate regression (MCA on YT X: 3 modes) Principal component regression at each grid point (EOF on X: 1 mode)

  7. How stable are cross-validated predictorsand regression parameter estimates? Predictor: First PC of X • Stable • ENSO 1988 1990 1992 1994 1996 1998 2000 Grid point with pos. corr. skill Grid point with neg. corr. skill Robust Y Y First PC of X First PC of X Param. estimates sensitive to removal of indiv. data points

  8. How influential is each data point? 2p/n 1 col. matrix (1st PC of X) H is the hat matrix Leverage: diag(H) n =15 data points p =1 PC Leverage is a function of the predictor alone, and measures the potential for a data point to affect the model parameter estimates

  9. Can precipitation forecasts over the Pacifichelp improve forecasts over land? Source: Franco Molteni (ECMWF)

  10. Correlation skill: Integrated forecast Issued: Nov Valid: DJF (1987-2001) South America domain (270o, 300o, 60oS, 15 oN) South America+Pacific domain (100o, 300o, 60oS, 15 oN) • Use of precip. fcsts over Pac. does help improve fcst. skill in S. America

  11. Can skill be improved by adding more models to the system? Correlation skill: Integrated forecast (precipitation) Issued: Nov Valid: DJF 1987-2001 1981-2005 South America + Pacific domain: ECMWF, UKMO, MF, CPTEC and empirical (diff. hind. periods) South America domain: ECMWF, UKMO and empirical (limited to common hindcast period) • Adding more models does help improve skill in S. America

  12. Can model predicted circulation variables help improve precip. forecast skill? • Use calibration procedure to explore atmospheric teleconnections

  13. Rationale for the use of circulation patterns as predictor for seasonal precip. Precip. is influenced by atmospheric circulation patterns On seasonal timescales the frequency of occurrence of such patterns is influenced by anomalous patterns of sea surface temperatures (particularly in the tropics) The link between tropical SSTs and global circulation patterns involves the generation of quasi-stationary upper level wave trains from tropical diabatic heat sources to remote regions (e.g. ENSO teleconnections to South America) If upper level circulation is well simulated by seasonal climate models, it may then be possible to use upper level circulationpredictions to produce precip. predictions for South America (i.e. precip. downscaling from upper level circulation )

  14. How well do coupled seasonal forecast models simulate upper level circulation? Correlation skill: 1-month lead forecasts for DJF ECMWF UK Met Office (GloSea 3) Pert. stream func. (’) Veloc. Poten. () Obs NCEP/NCAR Reanalysis Kalnay et al. (1996) BAMS, 77(3), 437-471 • Generally good skill in the tropics Hindcasts: 1987-2005

  15. Stephenson et al. (2005), Tellus A . Vol. 57, 253-264. Downscaling procedure: Forecast Assimilation Matrices Y: DJF precipitation X: 1-month lead 200 hPa ( ’,) pred. for DJF (ECMWF + UKMO) Correlation skill: 1-month lead precipitation forecasts for DJF ECMWF UKMO Forecast Assimilation ’ (’,) • Downscaled forecasts obtained with forecast assimilation have improved skill in North and Southeast South America compared to individual model predictions Forecast assimilation uses first three leading MCA modes of the matrix YT X.

  16. How does this compare with circulation-based and SST-based empirical predictions? Correlation skill: 1-month lead precipitation forecasts for DJF Fcst Assim. Emp: Circ-based Emp: SST-based Predictor: 1-month lead  ’ pred. for DJF (ECMWF + UKMO) Predictor: Obs SST in previous Oct. Predictor: Obs  ’ in previous Oct

  17. Seasonal forecast applications:

  18. obs fcts Flow prediction: Paraná river Issued: Nov Valid: Dec • Flow (ONS) 1982-2003: F • Oct SST (Reynolds et al. 2002): PC1, PC2 • Precip. GPCP (Adler et al. 2003): P • Integrated precip. forecasts (EUROBRISA) ECMWF, UKMO, MF, CPTEC: Pr Corr: 0.02 Issued: Nov Valid: Jan Corr: 0.38 Corr: 0.43 Issued: Nov Valid: Feb

  19. Dengue risk transmission index predictions ECMWF System 3: Anderson et al. (2007) ECMWF Tech. Memo, 503, pp 56 NCEP/NCAR Reanalysis: Kalnay et al. (1996) BAMS, 77(3), 437-471 Work by: Caio Coelho Rachel Lowe Nicolas Degallier • Dengue risk trans. model: • Degalier et al. (2005) • Environ, Risques & Santé • 4 (2), 1-5 • Favier et al. (2006) • Trop. Med. and Int. Health • 11 (3), 332–340 Bias corr. T RH (climat.) T, RH Fcst. risk Sim. risk Morse et al. (2005) Tellus, 57A(3), 464-475 Hindcast period: 1981-2005 0 to 5 month lead predictions; 11 ensemble members

  20. Dengue risk transmission index (R) Source: Nicolas Degallier (IRD) m: Environmental (climatic) capacity to sustain the development of the vector (optimum disease reproduction rate) n: climatic capacity to ensure transmission of the pathogen (larva/hab. ~ vector density capable of sustaining stable transmission) Both m and n are modelled as function of T and RH if n=m (R=0) if n>m (R>0) favorable conditions for transmission if n<m (R<0) unfavorable conditions for transmission 50<R<100: endemic risk R>100: epidemic risk

  21. Skill assessment: Dengue risk transmission index prediction issued in Nov. (Gerrity score: terc. cat.) Valid: Dec Valid: Jan Valid: Nov 1-month lead 0-month lead 2-month lead Valid: Feb Valid: Mar Valid: Apr 3-month lead 4-month lead 5-month lead

  22. Example: Dengue risk transmission index prediction issued in Nov 1997, valid for Apr 1998 5-month lead fcst Obs Corr. skill Brasília Salvador Nov Dec Jan Feb Mar Apr Nov Dec Jan Feb Mar Apr

  23. Future plans • Investigate alternative methods of dimensionality reduction for the multivariate regression in the FA procedure • Implement new version of EUROBRISA forecasting system - able to accommodate models with different hindcast periods - incorporate Meteo-France System 3 and CPTEC forecasts - how to proceed with UK Met Office GloSea 4 • Research on seasonal forecast applications (agriculture, hydropower and health) • Implement new approaches to visualise forecasts • Produce joint EUROBRISA publication

  24. Summary • Early stage of El Niño: EUROBRISA forecast for ASO 2009 is for below normal precip. in N South America and above normal precip. in SE South America • Use of precip. forecasts over Pacific improves robustness of predictors and forecast skill over South America • Adding more models to the integrated system helps improve forecast skill • Coupled model upper level circulation predictions can be successfully used for producing skilful precip. forecasts for South America • Preliminary results on application are encouraging for further developing research using seasonal forecasts • New web link http://eurobrisa.cptec.inpe.br

More Related