ECMWF Seasonal Forecast Group Meeting: 24 July 2002

1
Forecast Assimilation of DEMETER Coupled Model
Seasonal Predictions
Caio A. S. Coelho e-mail c.a.d.s.coelho_at_reading.
ac.uk Supervisors D. B. Stephenson, F. J.
Doblas-Reyes () Thanks to CAG, S. Pezzulli and
M. Balmaseda () Department of Meteorology,
University of Reading and ECMWF ()
2
Plan of talk

• Issues
• Conceptual framework (Forecast Assimilation)
• DEMETER
• Examples of application 0-d, 1-d, 2-d.
• Conclusions

3
1. Issues

Calibration

• Why do forecasts need it?
• Which are the best ways
• to calibrate?
• How to get good probability
• estimates?

Combination

• Why to combine?
• Should model predictions be
• selected?
• How best to combine?

4
2. Conceptual framework
Forecast Assimilation
Data Assimilation
5
3. Multi-model ensemble approach
Errors
Model formulation
Initial conditions
Multi-model
Ensemble
Solution
http//www.ecmwf.int/research/demeter
6
DEMETER Multi-model ensemble system

• 7 coupled global circulation models

9 member ensembles ERA-40 initial conditions
SST and wind perturbations 4 start dates per
year (Feb, May, Aug and Nov) 6 month hindcasts
Hindcast period 1980-2001 (1959-2001)
7
4. Examples of application

• Niño-3.4 index (0-d)
• Equatorial Pacific SST (1-d)
• South American rainfall (2-d)

8
Example 1 Niño-3.4 forecasts
95 P.I.
Well-calibrated Most observations in the 95
prediction interval (P.I.)
9
ECMWF coupled model ensemble forecasts
m9
DEMETER 5-month lead

• Observations not within the 95 prediction
  interval!
• Coupled model forecasts need calibration

10
Univariate X and Y
Prior
Likelihood
Posterior

Bayes theorem
11
Modelling the likelihood p(XY)
y
12
Combined forecasts
? Note most observations within the 95
prediction interval!
13
All forecasts
Empirical
Coupled
Combined
MAESS 1- MAE/MAE(clim.)100
BSS 1- BS/BS(clim.)100
14
Multivariate X and Y
bias
Prior
Likelihood
Matrices
Posterior
15
Example 2 Equatorial Pacific SST
DEMETER 7 coupled models 6-month lead
BSS 1- BS/BS(clim.)100
SST anomalies Y (C)
Forecast probabilities p
16
Brier Score as a function of longitude
Forecast assimilation reduces (i.e. improves) the
Brier score in the eastern and western
equatorial Pacific
17
Brier Score decomposition
reliability
uncertainty
resolution
18
Reliability as a function of longitude
Reliability as a function of longitude
Forecast assimilation improves reliability in the
western Pacific
19
Resolution as a function of longitude
Forecast assimilation improves resolution in the
eastern Pacific
20
Why South America?
? Seasonal climate potentially predictable
El Niño (DJF)
DEMETER Multi-model
La Niña (DJF)
Source Climate Prediction Center
(http//www.cpc.ncep.noaa.gov)
Correlation DJF rainfall
21
Why South American rainfall?

• Agriculture
• Electricity More than 90 produced by
  hydropower stations
• e.g. Itaipu (Brazil/Paraguay)
• World largest hydropower plant
• Installed power 12600 MW
• 18 generation units (700 MW each)
• 25 electricity consumed in Brazil
• 95 electricity consumed in Paraguay

22
Itaipu
23
Example 3 South American rainfall anomalies
Forecast Assimilation
Obs
Multi-model
DEMETER 3 coupled models (ECMWF, CNRM,
UKMO) 1-month lead Start Nov DJF ENSO
composites 1959-2001 16 El Nino years 13 La
Nina years
r0.51
r0.97
r0.28
r0.82
(mm/day)
24
South American DJF rainfall anomalies
Obs
Multi-model
Forecast Assimilation
r0.59
r-0.09
r0.32
r0.56
(mm/day)
25
South American DJF rainfall anomalies
Obs
Multi-model
Forecast Assimilation
r0.32
r0.04
r0.08
r0.38
(mm/day)
26
Brier Skill Score for S. American rainfall
Forecast assimilation improves the Brier Skill
Score (BSS) in the tropics
27
Reliability component of the BSS
Forecast assimilation improves reliability over
many regions
28
Resolution component of the BSS
Forecast assimilation improves resolution in the
tropics
29
5. Conclusions

• unified framework for the calibration and
  combination of predictions forecast
  assimilation
• improves the skill of probability forecasts
• Example 1 Niño-3.4
• ? improved mean forecast value and
• ? prediction uncertainty estimate
• Example 2 Equatorial Pacific SST
• ? improved reliability (west) and resolution
  (east)
• Example 3 South American rainfall
• ? improved reliability and resolution in the
  tropics ? improved reliability over subtropical
  and central regions

30
More information

• Coelho C.A.S. Forecast Calibration and
  Combination Bayesian Assimilation of Seasonal
  Climate Predictions. PhD Thesis. University of
  Reading (to be submitted)
• Coelho C.A.S., D. B. Stephenson, F. J.
  Doblas-Reyes and M. Balmaseda From Multi-model
  Ensemble Predictions to Well-calibrated
  Probability Forecasts Seasonal Rainfall
  Forecasts over South America 1959-2001. CLIVAR
  Exchanges (submitted).
• Stephenson, D. B., Coelho, C. A. S.,
  Doblas-Reyes, F.J. and Balmaseda, M.
• Forecast Assimilation A Unified Framework for
  the Combination of
• Multi-Model Weather and Climate Predictions.
• Tellus A - DEMETER special issue (in press).
• Coelho C.A.S., S. Pezzulli, M. Balmaseda, F. J.
  Doblas-Reyes and D. B. Stephenson, 2004
  Forecast Calibration and Combination A Simple
  Bayesian Approach for ENSO. Journal of Climate.
  Vol. 17, No. 7, 1504-1516.
• Coelho C.A.S., S. Pezzulli, M. Balmaseda, F. J.
  Doblas-Reyes and D. B. Stephenson, 2003 Skill
  of Coupled Model Seasonal Forecasts A Bayesian
  Assessment of ECMWF ENSO Forecasts. ECMWF
  Technical Memorandum No. 426, 16pp. Available
  at http//www.met.rdg.ac.uk/swr01cac

31
Reliability diagram (Multi-model)
(oi)
o
(pi)
32
Reliability diagram (FA 58-01)
(oi)
o
(pi)
33
Operational Seasonal forecasts for S. America
Coupled models
Europe http//www.ecmwf.int
U.K http//www.metoffice.com
Atmospheric models forced by persisted/forecast
SSTs
U.S.A http//iri.columbia.edu
Brazil http//www.cptec.inpe.br
34
Mean Anomaly Correlation Coefficient
35
Momentum measure of skewness
Measure of asymmetry of the distribution