Predictability of ENSO-Monsoon Relationship in NCEP CFS

1. Predictability of ENSO-Monsoon Relationship in NCEP CFS Emilia K. Jin Center for Ocean-Land-Atmosphere studies (COLA) George Mason University (GMU) Thanks to J. Kinter, B. Kirtman, J. Shukla, and B. Wang* COLA/GMU, *IPRC/Univ. of Hawaii NOAA 32th Climate Diagnostics and Prediction Workshop (CDPW), 22-26 Oct COAPS/FSU, Tallahassee, FL

2. Contents • ENSO-monsoon relationship in NCEP/CFS forecasts • The role of perfect ocean forcing in coupled systems: CGCM vs. “Pacemaker” • The role of air-sea interaction on ENSO-monsoon relationship • Shortcoming in “Pacemaker”: Decadal change of ENSO-Indian monsoon relationship

3. JJA Forecast Skill of Rainfall with respect to Lead Month Temporal correlation with respect to lead month 1st month 3rd month 5th month 9th month Area mean (60-140E, 30S-30N) For retrospective forecasts, reconstructed data with respect to lead time (monthly forecast composite) is used. Correlation Forecast lead month • Retrospective forecast (Courtesy ofNCEP)

4. Relationship between NINO3.4 and Monsoon Indices Lead-lag correlation with respect to lead month WNPSM index Extended IMR index N34 lead N34 lag N34 lead N34 lag COR(OBS,CFS) Observation 1st month forecast 8th month forecast EIMR WNPSMI 1st month 8th month 0.48 0.20 0.48 0.12 • Western North Pacific Summer Monsoon Index (Wang and Fan, 1999) • WNPSMI : U850(5ºN–15ºN, 100ºE–130ºE) minus U850(20ºN–30ºN, 110ºE–140ºE) • Extended Indian Monsoon Rainfall Index (Wu and Kirtman 2004) • EIMR: Rainfall (5ºN–25ºN, 60ºE–100ºE) • Green line denotes 95% significant level.

5. Relationship between NINO3.4 and Monsoon Indices WNPSM index Extended IMR index Observation 1st month forecast 8th month forecast CFS long run • NCEP/CFS 52-year long run (Courtesy ofKathy Pegion)

6. Regressed field of 1st SEOF of 850 hPa zonal wind Observation 1 Shading: 500 hPa vertical pressure velocity Contour: 850 hPa winds 1 Shading: Rainfall(CMAP and PREC/L) Contour: SST COR (PC, NINO3.4) = 0.85 • From the summer of Year 0, referred to as JJA(0), to the spring of the following year, called MAM(1), a covariance matrix was constructed using four consecutive seasonal mean anomalies for each year. • SEOF (Wang and An 2005) of 850 hPa zonal wind over 40E-160E, 40S-40N • High-pass filter ofeight years • The seasonally evolving patterns of the leading mode concur with ENSO’s turnabout from a warming to a cooling phase (Wang et al. 2007).

7. Regressed field of 1st SEOF of 850 hPa zonal wind Observation 1 Shading: 500 hPa vertical pressure velocity Contour: 850 hPa winds 1 Shading: Rainfall(CMAP and PREC/L) Contour: SST COR (1st PC timeseries of SEOF, N34) Correlation N34 lead N34 lag

8. Impact of the Model Systematic Errors on Forecasts Pattern Cor. of EOF Eigenvector Patternl correlation of eigenvector with observation Pattern correlation of eigenvector with free long run Correlation • With respect to the increase of lead month, forecast monsoon mode associated with ENSO is much similar to that of long run, while far from the observed feature. Forecast lead month COR (1st PC timeseries of SEOF, N34) Correlation Observation 1st month forecast 8th month forecast CFS long run N34 lead N34 lag

9. In CFS coupled GCM, what is responsible to drop the predictability of ENSO – monsoon relationship? • Ocean forcing? • Atmospheric response? • Air-sea interaction? …..

10. “Pacemaker” Experiments • The challenge is to design numerical experiments that reproduce theimportant aspects of this air-sea coupling while maintaining the flexibility to attempt to simulatethe observed climate of the 20th century. • “Pacemaker”: tropical Pacific SST is prescribed from observations, but coupled air-seafeedbacks are maintained in the other ocean basins (e.g. Lau and Nath, 2003). • Anecdotalevidence indicates that pacemaker experiments reproduce the timing of the forced response to ElNiño and the Southern Oscillation (ENSO), but also much of the co-variability that is missingwhen global SST is prescribed. • In this study, we use NCEP/GFS T62 L64 AGCM mainly.

11. Pacemaker region Outside the pacemakerregion To handle model drift To merge the pacemaker and non-pacemaker regions Mixed-layer depth “Pacemaker” Experimental Design In this study, the deep tropical eastern Pacific where coupled ocean-atmospheredynamics produces the ENSO interannual variability, is prescribed by observed SST. 165E-290E, 10S-10N No blending Slab ocean mixed-layer Weak damping of 15W/m2/K to observed climatology Zonal mean monthly Levitus climatology

12. Model and Experimental Design Pacemaker CGCM Atmosphere (GFS T62L64) Atmosphere (GFS T62L64) Local air-sea interaction Fully coupled system Ocean (Full dynamics) SST SST heat flux, wind stress, fresh water flux -γTclim Observed SST heat flux Slab ocean (No dynamics and advection) Mixed layer model + AGCM (1950-2004, 4runs) CGCM (52 yrs)

13. Lead-lag correlation with Nino3.4 Index WNPSMI EIMR 1st PC timeseries of SEOF ISMI Observation PACE CFS N34 lead N34 lag N34 lead N34 lag ISMI: U850(5ºN–15ºN, 40ºE–80ºE) minus U850(20ºN–30ºN, 70ºE–90ºE) Ensemble spread of 4 members of Pacemaker exp.

14. ENSO Characteristics in CFS CGCM NINO3.4 Index during 1950-2005 (a) Observation (b) CFS CGCM (52 year long run)

15. Lead-lag correlation with Nino3.4 Index WNPSMI EIMR 1st PC timeseries of SEOF ISMI Observation PACE CFS N34 lead N34 lag N34 lead N34 lag ISMI: U850(5ºN–15ºN, 40ºE–80ºE) minus U850(20ºN–30ºN, 70ºE–90ºE) Ensemble spread of 4 members of Pacemaker exp.

16. ENSO Characteristics in CFS CGCM Regression of DJF NINO3.4 Index to SST anomalies (a) Observation (b) CFS long run • In CGCM, ENSO SST anomalies show westward penetration with narrow band comparing to the observed.

17. JJA Regression map of 1st SEOF of 850 hPa zonal wind Difference from Obs. Total field 850 hPa zonal wind and SST 850 hPa zonal wind and rainfall Obs. Pace Pace-Obs. CFS CFS-Obs. Contour: zonal wind Shading: SST Contour: zonal wind Shading: rainfall

18. Model and Experimental Design Control Pacemaker Atmosphere (GFS T62L64) Atmosphere (GFS T62L64) No air-sea interaction Local air-sea interaction SST -γTclim Observed SST Observed SST heat flux Slab ocean (No dynamics and advection) Climatology SST Mixed layer model + AGCM (1950-2004, 4runs) AGCM (1950-2004, 4runs)

19. Leag-lag correlation with Nino3.4 Index WNPSMI EIMR 1st PC timeseries of SEOF ISMI Observation PACE Control N34 lead N34 lag N34 lead N34 lag ISMI: U850(5ºN–15ºN, 40ºE–80ºE) minus U850(20ºN–30ºN, 70ºE–90ºE) Ensemble spread of 4 members of Pacemaker exp.

20. JJA Regression map of 1st SEOF of 850 hPa zonal wind Difference from Obs. Total field 850 hPa zonal wind and rainfall 850 hPa zonal wind and SST Obs Pace-Obs. Pace Ctl-Obs. Ctl Contour: zonal wind Shading: SST Contour: zonal wind Shading: rainfall

21. 1st S-EOF modes: Observation 1956-76 1977-2004

22. Lead-lag Correlation between NINO3.4 and Monsoon indices 56-76 77-04 Decadal change of ENSO-Monsoon relationship based on SEOF analysis (Wang et al. 2007) • Remote El Niño/La Niña forcing is the major factor that affects A-AM variability. • The mismatch between NINO3.4 SST and the evolution of the two major A-AM circulation anomalies suggests that El Niño cannot solely force these anomalies. 2. The monsoon-warm pool ocean interaction is also regards as a cause (a positive feedback between moist atmospheric Rossby waves and the underlying SST dipole anomalies) • The enhanced ENSO variability in the recent period has increased the strength of the monsoon-warm pool interaction and the Indian Ocean dipole SST anomalies, which has strengthened the summer westerly monsoon across South Asia, thus weakening the negative linkage between the Indian summer monsoon rainfall and the eastern Pacific SST anomaly.  However, in pacemaker, the strengthen of the Indian Ocean dipole SST anomalies is not shown due to fixed mixed-layer depth and SST climatology.

23. Change of Lead-lag Correlation (Extended IMR, NINO3.4) 20-year Moving Window during 1950-2004 OBS (IMR) (HadSST and CMAP) Lag correlation with respect to 20-yr moving window during 55 years

24. Summary • In CFS CGCM, the predictability of lead-lag ENSO-monsoon relationship drops with respect to lead month due to systematic errors of ENSO and its response. • To improve the predictability, “pacemaker” experiment is designed and conducted to reproduce theimportant aspects of air-sea coupling while maintaining the flexibility to attempt to simulatethe observed climate. • Surprisingly, “pacemaker” mimics the realistic ENSO-monsoon relationship compared to other experiments including control and coupled (CGCM). • However, the recent change of ENSO-Indian monsoon relationship is missed in “pacemaker”, possibly associated with the Indian Ocean dynamics, while the decadal change of western North Pacific summer monsoon is well related with that of eastern tropical Pacific SST anomalies. • To find out the cause of this discrepancy, supplementary “pacemaker” experiments can be performed based on this shortcoming.

25. Thank You ! Emilia K. Jin kjin@cola.iges.org

26. Change of DJF Simultaneous Correlation 20-year Moving Window during 1950-2004 Observation PACE CONTROL Ensemble spread of Pace Ensemble spread of Control • Shading denotes ensemble spread among 4 members. Note that correlation for ensemble mean is not the average of correlations for four members.