1. Operational MJO prediction at ECMWF
Frédéric Vitart
With contributions from
A. Beljaar, P. Bechtold and M . Miller
European Centre for Medium-Range Weather Forecasts

2. Outline MJO prediction using the ECMWF monthly forecasting system
2. Skill of the monthly forecasting system to predict the MJO
Some sensitivity experiments
Impact of changes in physic parameterization
Conclusion

3. Forecasting systems at ECMWF
Product
Tool
ECMWF:
Weather and Climate
Dynamical Forecasts
Forecasting at ECMWF
Medium-Range
Forecasts
Day 1-15
Monthly
Forecast
Day 1-32
Seasonal
Forecasts
Month 1-7 (13)
Here is an overview of the ECMWF products:
Medium-Range Forecasts:
These are global atmosphere forecasts and ocean wave forecasts, they are produced Twice daily and they range from 1 days up to 15 days. Medium range forecasts are created by using a sophisticated numerical model that simulated the atmospheric circulation. The atmospheric model is initialized with the status of art data assimilation system.
Monthly forecasts are produced Weekly, every Thursday, and they extend up to 31 days. Beyond 15 days the interactions between atmospheric circulation and ocean circulation play an important role. As a consequence of that the Monthly forecasting system consists of an Atmosphere-ocean coupled model.
Seasonal forecasts are produced Monthly, every 15th , and they extend up to seven months. Those forecast are also created by an Atmosphere-ocean coupled model. Since March 2007 we also produce a limited ensemble of seasonal forecast that are extended up to 13 months.
All products available for commercial use through the ECMWF Catalogue
Atmospheric model
Ocean model
Atmospheric model
Ocean model
Atmospheric model

4. The ECMWF monthly forecasting system
Coupled ocean-atmosphere integrations: a 51-member ensemble is integrated for 32 days every Thursday
Atmospheric component: IFS with the latest operational cycle and with a T159L62 resolution (about 125 km)
Oceanic component: HOPE (from Max Plank Institute) with a zonal resolution of 1.4 degrees and 29 vertical levels
Coupling: OASIS (CERFACS). Coupling every ocean time step (1 hour)
Calibration: a 5-member ensemble is integrated at the same day and same month as the real-time time forecast over the past 12 years.
The forecasts beyond 15 days are computed by a coupled ocean-atmosphere system. The system is run many times for the month ahead to build up a picture of the likelihood of different weather types affecting the Europe and the entire globe.
Every Thursday an ensemble of 51 forecasts is launched. The initial conditions of each forecast are perturbed to represents the uncertainties of the initial conditions. Each forecast evolves independently for 32 days.
The monthly forecast products is based on probabilities and ensemble means computed by using those 51 members.
The atmospheric component is the same used for the medium range but with a coarser horizontal resolution (~120Km vs 50Km )
The oceanic component has enough horizontal resolution to resolve the ocean baroclinic waves and processes confined at the Equator. (meridional resolution is variable with a max of 0.3 degree at the equator)
Exchange of fluxes between atmosphere and ocean are computed every hours.

5. Prediction of the Madden Julian Oscillation (MJO) VP200
Monthly Forecast starting on 7 December 2007
VP200

6. Prediction of the Madden Julian Oscillation (MJO)
Monthly Forecast starting on 7 December 2007

7. MJO metric (similar to Wheeler and Hendon (2004))
Forecast anomalies of OLR, U850, & U200 are obtained by subtracting the hindcast climatology of the respective fields from the forecast.
The forecast anomalies are then averaged over latitudes (-15, 15).
The forecast anomalies of the 3 fields are divided by the respective global standard deviations calculated from the NCEP reanalyses.
The resulting anomalies are projected onto the first two EOF patterns, the latter being computed by Matt Wheeler from NCEP reanalyses.
The two projection coefficient time series are further divided by the standard deviations of first two PC (also from NCEP reanalyses), to obtain the RMM12.
To remove the interannual variability, a 120-day running mean of RMM12 is subtracted.

8. Operational MJO Prediction

9. + 12-year climatology + 18-year climatology
Planned unified 32-day VAREPS/monthly system
Present TL159 monthly system:
Initial condition
Day 32
Coupled forecast at TL159 (~125 km)
+ 12-year climatology
Future 32-day VAREPS/monthly system:
EPS Integration at T399 (~50km)
Coupled forecast at TL255 (~80km)
Initial condition
Day 10
Day 32
Heat flux, Wind stress, P-E
The next upgrade of the monthly system will be a unification of the medium -range and monthly forecast.
Currently the monthly forecast runs from the initial conditions defined by the operational analysis at a relatively low resolution (~120km).
In future the system will start from the day 10 of the ensemble prediction system. The ensemble prediction system run at higher resolution than the current monthly forecast (50Km vs 120Km) so the 10 days forecast will benefit from the higher resolution.
From days 10 the coupled system will be run at horizontal resolution of about 80km .
The test version of such system has been set up and a number of forecast covering the period have been performed.
SST anomalies are persisted instead of the full SST field.
This suite also produces hindcasts once a week. So far, 5 sets hindcasts (5 ensemble members, 18 years) have been produced.
The cost of the forecast increases with the cube of the horizontal resolution.
Ocean only integration
+ 18-year climatology

10. Skill of the MJO Prediction
EXPERIMENT: A 5-member ensemble has been run every day from 15 December 1992 to 31st January 1993.
The MJO diagnostic is based on the same method as in the real-time forecasts, except for:
Velocity potential anomalies at 200 hPa instead of U200
Fields are averaged between 10N and 10S instead of 15N and 15S.

11. Skill of the MJO Prediction
ERA40: 15/12/92-31/01/93
Velocity Potential 200 hPa
U850
OLR

12. Skill of the MJO Prediction

13. Current operational version of the monthly forecasting system (CY32R2)
Skill of the MJO Prediction
Current operational version of the monthly forecasting system (CY32R2)
Anomaly correlation
Amplitude
PC1
PC2

14. Impact of ocean-atmosphere coupling
Cycle CY32R2
PC1
PC2
VAREPS
Coupled
Ocean ML

15. Warm ocean (skin ) layer model to represent diurnal time scale
Similarity temperature profile is assumed with depth scale d representing penetration depth of solar radiation:
(surface/skin
temperature)
(deep/bulk
temperature)
Prognostic equation for temperature difference between skin and bulk ocean temperature:
Energy forcing from surface and solar absorption
Wind and stability dependent mixing term (reduces temperature difference to 0 in strong winds)
Zeng and Beljaars (2005), Geophys. Res. Lett. 32, L14605

16. Tsk difference between 20 UTC and 12 UTC averaged from 20 to 22 May 1988
Model Tsk difference; forecast from UTC
Wind speed (m/s)
Tsk difference retrieved from GOES-8 (Wu et al. 1999; BAMS, 80, )

17. Impact of the Ocean Slab Model
PC1
PC2
Coupled+slab model
Mixed-layer model
Coupled

18. Evolution of SST anomalies. 5-day means Control
Coupled
Coupled + Slab
Analysis
pentad2-pentad1
pentad3-pentad2
pentad4-pentad3
pentad5-pentad4
pentad6-pentad5

19. Impact of changes in model parameterization
OLR- Forecast range: day 15
ERA40
28R3
29R1
31R1
32R2
32R3
29/12
05/01
12/01
days
20/01
28/01
04/04
12/02
10/04
04/05
09/06
06/07
11/07

20. Amplitude of the MJO PC1 PC2
28R3
29R1
30R2
31r1
32r2
32r3

21. MJO Propagation
PC1
PC2
28R3
29R1
30R2
31r1
32r2
32r3

22. CY32R3: velocity Potential at 200hPa
Starting date: 29/12/1992

23. Recent or planned changes to the ECMWF physical parametrizations
Improved treatment of ice sedimentation, auto- conversion to snow in cloud scheme and super-saturation with respect to ice (CY31R1)
Change in the short-wave radiation code (more wavelengths) + McICA
MODIS albedo + revised cloud optical properties (CY32R2)
New formulation of convective entrainment
Variable relaxation timescale for closure
Reduction in background free atmosphere vertical diffusion (CY32R3)

24. Impact on relative humidity (RH) climatology
31r1 – 30r1 annual mean difference
Largest changes in the tropical upper troposphere

25. Convection changes to operational massflux scheme
New formulation of convective entrainment:
Previously linked to moisture convergence
– Now more dependent on the relative humidity of the environment
New formulation of relaxation timescale used in massflux closure:
Previously only varied with horizontal resolution – Now a variable that is dependent on the convective turnover timescale i.e. variable in both space and time also
Impact of these changes is large including a major increase in tropical variability. Midlatitude synoptic variability is also increased and is more realistic.

26. Composite vertical profile for west pac, JJA
Minimum cloud top heights distributions (CloudSat)
precipitating clouds
Of note:
Trimodality (quadra-modal) heights
precipitating clouds are deeper than non precipitating clouds
non- precip
clouds

27. Average tropical cloud profiles
In the Tropics now a tri-modal cloud distribution becomes apparent, with a strong increase in mid-level clouds. CloudSat data will be used to verify these results.

28. Forecast Biases and Convection Precipitation for DJF against GPCP for different cycles: from 15 year 5 months integrations for
CY31R1 Sep 2006
CY32R2 June 2007
CY32R3 Nov 2007

29. Tropical variability in wave-number frequency space of OLR (DJFM)
Obs (NOAA)
32R1
32R3

30. MJO Prediction
Current system
CY32R2
VarEPS/Monthly +
CY32R3

31. Conclusion
The current monthly forecasting system has some skill to predict the amplitude of the MJO up to about 14 days. However the amplitude of the MJO is significantly reduced after a few days of forecasts, but this will change after the new model operational implementation on 3 November.
Sensitivity experiments:
Strong impact of ocean-atmosphere coupling
No significant impact when increasing the horizontal resolution to about 80 kilometres
Strong sensitivity to changes in the model parameterization, and in particular to the mass flux scheme.