![Page 1: Application of a serial extended forecast experiment using ...€¦ · forecasting framework (Vitart 2004) ~20 days before and after the maximum convection in the West Pacific Warm](https://reader033.vdocument.in/reader033/viewer/2022050715/5f12b03f71d15e4db53eb400/html5/thumbnails/1.jpg)
Application of a serial extended forecast experiment using the ECMWF model to interpret the predictive skill
of tropical intraseasonal variability
Paula A. Agudelo
Carlos D. Hoyos, Peter J. Webster and Judith Curry
Acknowledgment to Federic Vitart form ECMWF
Georgia Institute of Technology
Agudelo et al. Climate Dynamics 2009
Valsavarenche, June 8-13, 2009
![Page 2: Application of a serial extended forecast experiment using ...€¦ · forecasting framework (Vitart 2004) ~20 days before and after the maximum convection in the West Pacific Warm](https://reader033.vdocument.in/reader033/viewer/2022050715/5f12b03f71d15e4db53eb400/html5/thumbnails/2.jpg)
Intraseasonal Oscillation
Pacific islands
South-East Asian Monsoon
Australian Monsoon
West coast of North America
South America
Variability of rainfall
Not well simulated and not skillfully forecasted
by Numerical models
(Variability in the 25-80 days band)
Africa
Numerical study to investigate the reasons for the poor skill in numerical models reproducing the vertical structure of the ISO during winter and summer (i.e. when and why do models fail?) and its relevance for monsoon systems
![Page 3: Application of a serial extended forecast experiment using ...€¦ · forecasting framework (Vitart 2004) ~20 days before and after the maximum convection in the West Pacific Warm](https://reader033.vdocument.in/reader033/viewer/2022050715/5f12b03f71d15e4db53eb400/html5/thumbnails/3.jpg)
TRANSITION PHASETRANSITION PHASE
SUPPRESED PHASE ACTIVE PHASE
Intraseasonal oscillation
![Page 4: Application of a serial extended forecast experiment using ...€¦ · forecasting framework (Vitart 2004) ~20 days before and after the maximum convection in the West Pacific Warm](https://reader033.vdocument.in/reader033/viewer/2022050715/5f12b03f71d15e4db53eb400/html5/thumbnails/4.jpg)
Series of 30-day forecasting integrations
Summer and Winter
Control forecast and four ensemble members.
Model ExperimentFully-coupled ECMWF model in the monthly forecasting framework (Vitart 2004)
~20 days before and after the maximum convection in the West Pacific Warm Pool or Central Indian Ocean
ensemble members.
Initialized with ERA 40reanalysis
![Page 5: Application of a serial extended forecast experiment using ...€¦ · forecasting framework (Vitart 2004) ~20 days before and after the maximum convection in the West Pacific Warm](https://reader033.vdocument.in/reader033/viewer/2022050715/5f12b03f71d15e4db53eb400/html5/thumbnails/5.jpg)
Zonal WindsVertical Structure1-day to 30-day lead times, starting from 62
different consecutive days
Distribution of the pattern correlation for the control run as a function of forecast lead time and start of the forecast for vertical anomalies of zonal winds
forecast of January 11 1993
Indo-West Pacific region (10°S-10N 60°E-170°W) for the TOGA-COARE case.
There are states of the system for which the skill of the forecast is always low: the forecast scheme is flow dependant
99% statistical significance threshold
Average correlation for the control run (black line) and all ensemble members (color lines).
Skillful up to 11 days
![Page 6: Application of a serial extended forecast experiment using ...€¦ · forecasting framework (Vitart 2004) ~20 days before and after the maximum convection in the West Pacific Warm](https://reader033.vdocument.in/reader033/viewer/2022050715/5f12b03f71d15e4db53eb400/html5/thumbnails/6.jpg)
Zonal WindsVertical Structure1-day to 30-day lead times, starting from 62
different consecutive days
Distribution of the standardized pattern RMS for the control run
The changes in skill from one day to the next appear to depend strongly both on the initial conditions and the atmospheric environment to be forecasted.
forecast of January 11 1993
Standardized pattern root-mean-squared (RMS) error of 5-day moving average anomalies of longitude-height snapshots.
Skillful up to 11 days
environment to be forecasted.
The fact that the skill of all members is similar in all runs suggests that the spread of the ensembles is not large enough to account for all degrees of freedom in the coupled system.
![Page 7: Application of a serial extended forecast experiment using ...€¦ · forecasting framework (Vitart 2004) ~20 days before and after the maximum convection in the West Pacific Warm](https://reader033.vdocument.in/reader033/viewer/2022050715/5f12b03f71d15e4db53eb400/html5/thumbnails/7.jpg)
Specific Humidity
January 11 1993
Vertical Structure
Forecasts of specific humidity, associated with the moist convective anomalies during the ISO, have no skill.
Forecasts of zonal winds which correspond mainly to regional circulation anomalies associated with the ISO, are skillful up to about 11 days.
![Page 8: Application of a serial extended forecast experiment using ...€¦ · forecasting framework (Vitart 2004) ~20 days before and after the maximum convection in the West Pacific Warm](https://reader033.vdocument.in/reader033/viewer/2022050715/5f12b03f71d15e4db53eb400/html5/thumbnails/8.jpg)
January 11 1993
The conditions around this particular date (January 11 1993) appear to be of a transitional to active nature.
ERA-40 anomalies (top) show the growth of the convective event starting in the Indian Ocean.
The control forecast (bottom) suggest a dryer atmosphere than ERA-40 data.
![Page 9: Application of a serial extended forecast experiment using ...€¦ · forecasting framework (Vitart 2004) ~20 days before and after the maximum convection in the West Pacific Warm](https://reader033.vdocument.in/reader033/viewer/2022050715/5f12b03f71d15e4db53eb400/html5/thumbnails/9.jpg)
Mature Convective State
Zonal Winds Specific Humidity
The pattern of the forecasted zonal wind anomalies coincides well with ERA-40 in shape but not in magnitude (weak anomalies).
There is no skill in the specific humidity forecast for the same date and lead time
Weak circulation anomalies in the forecasts are most likely a consequence of the relatively dry atmosphere.
A dryer atmosphere implies less latent heating in the atmospheric column, resulting in weaker circulation.
![Page 10: Application of a serial extended forecast experiment using ...€¦ · forecasting framework (Vitart 2004) ~20 days before and after the maximum convection in the West Pacific Warm](https://reader033.vdocument.in/reader033/viewer/2022050715/5f12b03f71d15e4db53eb400/html5/thumbnails/10.jpg)
VPT profiles for a region in the Equatorial Indian Ocean (10°S-10°N; 70-90°E) from ERA-40 and the forecasts for the same date for different lead times
Higher values of 700-1000hPa virtual potential Higher values of 700-1000hPa virtual potential temperature in the forecasts suggest a more stable atmosphere in the forecasts than the ERA-40 data.
Observed throughout the entire period of the serial runs experiment as confirmed by the
Limits the reproduction of deep convective events associated with the ISO.
Distribution of the differences between the forecast and observed 700-1000hPa VPT for different lead times
![Page 11: Application of a serial extended forecast experiment using ...€¦ · forecasting framework (Vitart 2004) ~20 days before and after the maximum convection in the West Pacific Warm](https://reader033.vdocument.in/reader033/viewer/2022050715/5f12b03f71d15e4db53eb400/html5/thumbnails/11.jpg)
200mb zonal wind
OLR
850mb zonal wind
Ensemble spreadThere is a gap between the RMS and the ensemble spread
Basin-wide amplitude of the forecast ISO relative to the observations.
In other words, the ISO is damped or eroded with in the model.
The model does not maintain the observed amplitude in the intraseasonal band.
Amplitude considerably lower than the observations Amplitude decreases with
the forecasting lead time.
![Page 12: Application of a serial extended forecast experiment using ...€¦ · forecasting framework (Vitart 2004) ~20 days before and after the maximum convection in the West Pacific Warm](https://reader033.vdocument.in/reader033/viewer/2022050715/5f12b03f71d15e4db53eb400/html5/thumbnails/12.jpg)
2004 summer ISO event
Zonal Winds
Summer skill between 5-7 days while in winter it is at least double.
![Page 13: Application of a serial extended forecast experiment using ...€¦ · forecasting framework (Vitart 2004) ~20 days before and after the maximum convection in the West Pacific Warm](https://reader033.vdocument.in/reader033/viewer/2022050715/5f12b03f71d15e4db53eb400/html5/thumbnails/13.jpg)
2002 summer ISO event
Zonal Winds
Skill of the forecast is greater for winter ISO cases than for summer events.
![Page 14: Application of a serial extended forecast experiment using ...€¦ · forecasting framework (Vitart 2004) ~20 days before and after the maximum convection in the West Pacific Warm](https://reader033.vdocument.in/reader033/viewer/2022050715/5f12b03f71d15e4db53eb400/html5/thumbnails/14.jpg)
Meridional Anomalies
Meridional vertical structure of specific humidity
Meridional vertical structure of winds
For specific humidity Correlation and RMS increase with lead time: counter intuitive
![Page 15: Application of a serial extended forecast experiment using ...€¦ · forecasting framework (Vitart 2004) ~20 days before and after the maximum convection in the West Pacific Warm](https://reader033.vdocument.in/reader033/viewer/2022050715/5f12b03f71d15e4db53eb400/html5/thumbnails/15.jpg)
2004 summer ISO event
Match the observations when actual active conditions are present
The model is always forecasting an active-like phase of the monsoon
Artificially increasing the skill in some phases.
conditions are present over India,
![Page 16: Application of a serial extended forecast experiment using ...€¦ · forecasting framework (Vitart 2004) ~20 days before and after the maximum convection in the West Pacific Warm](https://reader033.vdocument.in/reader033/viewer/2022050715/5f12b03f71d15e4db53eb400/html5/thumbnails/16.jpg)
1993 Winter ISO event
Australian Region
![Page 17: Application of a serial extended forecast experiment using ...€¦ · forecasting framework (Vitart 2004) ~20 days before and after the maximum convection in the West Pacific Warm](https://reader033.vdocument.in/reader033/viewer/2022050715/5f12b03f71d15e4db53eb400/html5/thumbnails/17.jpg)
Day 1 Errors
![Page 18: Application of a serial extended forecast experiment using ...€¦ · forecasting framework (Vitart 2004) ~20 days before and after the maximum convection in the West Pacific Warm](https://reader033.vdocument.in/reader033/viewer/2022050715/5f12b03f71d15e4db53eb400/html5/thumbnails/18.jpg)
Day 2 Errors
![Page 19: Application of a serial extended forecast experiment using ...€¦ · forecasting framework (Vitart 2004) ~20 days before and after the maximum convection in the West Pacific Warm](https://reader033.vdocument.in/reader033/viewer/2022050715/5f12b03f71d15e4db53eb400/html5/thumbnails/19.jpg)
Day 3 Errors
![Page 20: Application of a serial extended forecast experiment using ...€¦ · forecasting framework (Vitart 2004) ~20 days before and after the maximum convection in the West Pacific Warm](https://reader033.vdocument.in/reader033/viewer/2022050715/5f12b03f71d15e4db53eb400/html5/thumbnails/20.jpg)
Day 4 Errors
![Page 21: Application of a serial extended forecast experiment using ...€¦ · forecasting framework (Vitart 2004) ~20 days before and after the maximum convection in the West Pacific Warm](https://reader033.vdocument.in/reader033/viewer/2022050715/5f12b03f71d15e4db53eb400/html5/thumbnails/21.jpg)
Day 5 Errors
![Page 22: Application of a serial extended forecast experiment using ...€¦ · forecasting framework (Vitart 2004) ~20 days before and after the maximum convection in the West Pacific Warm](https://reader033.vdocument.in/reader033/viewer/2022050715/5f12b03f71d15e4db53eb400/html5/thumbnails/22.jpg)
Skill of the ECMWF forecast seems to be greater for winter ISO cases than for summer events.
The model tends to simulate more stable atmosphere compared to ERA-40.
Summary
The skill is different for different phases of the oscillation
During summer the model is always forecasting an active-like phase of the monsoon.
The model erodes the intraseasonal signal as the lead time increases
![Page 23: Application of a serial extended forecast experiment using ...€¦ · forecasting framework (Vitart 2004) ~20 days before and after the maximum convection in the West Pacific Warm](https://reader033.vdocument.in/reader033/viewer/2022050715/5f12b03f71d15e4db53eb400/html5/thumbnails/23.jpg)
Thanks
![Page 24: Application of a serial extended forecast experiment using ...€¦ · forecasting framework (Vitart 2004) ~20 days before and after the maximum convection in the West Pacific Warm](https://reader033.vdocument.in/reader033/viewer/2022050715/5f12b03f71d15e4db53eb400/html5/thumbnails/24.jpg)
![Page 25: Application of a serial extended forecast experiment using ...€¦ · forecasting framework (Vitart 2004) ~20 days before and after the maximum convection in the West Pacific Warm](https://reader033.vdocument.in/reader033/viewer/2022050715/5f12b03f71d15e4db53eb400/html5/thumbnails/25.jpg)
![Page 26: Application of a serial extended forecast experiment using ...€¦ · forecasting framework (Vitart 2004) ~20 days before and after the maximum convection in the West Pacific Warm](https://reader033.vdocument.in/reader033/viewer/2022050715/5f12b03f71d15e4db53eb400/html5/thumbnails/26.jpg)
The atmospheric component of the coupled model is the ECMWF atmospheric model Integrated Forecast System (IFS).
The model used in this study has a nominal horizontal resolution of 1.9 x 1.9 degrees (TL95) with 40 levels in the vertical.
The oceanic component is the Max Plank Institute Hamburg Ocean Primitive Equation (HOPE) model which has a zonal resolution of 1.4 degrees, a meridional resolution of 0.3 degrees at the equator and 1.4 at the extratropics, and 29 vertical levels with 10 m resolution in the upper 100 m (Wolff et al. 1997).
The coupling of the components is through the interface Ocean Atmosphere Sea Ice Soil (OASIS) platform (Terray et al. 1995).
Initial conditions for the atmospheric and land surface were obtained from the ERA40.
The ECMWF oceanic data assimilation system provided initial conditions for the oceanic component (Balmaseda et al. 2005).
Each ensemble member was generated by slightly perturbing the initial atmospheric and oceanic conditions in the same way as in the EMCWF Monthly Forecasting System (Vitart 2004, Vialard et al. 2003) using a singular vectors method (Buizza and Palmer 1995).