1. Introduction

2. "Magic" Experiment

3. Significant Improvement

4. Caveats

5. Challenges

Figure 1  Nino 3.4 SST prediction skill is measured by anomaly correlation for the lead time of 1-6 months.  CFS and CMP14 are the same coupled model system with the only difference in vertical resolution of the atmospheric component with 64 and 28 layers, respectively.  The other acronym labels are the major statistical methods used in Climate Prediction Center for routine ENSO forecasts.

Introduction

(Return to Top)

"Magic" Experiment

• Atmospheric model:  GFS03 of 64 layers, the result of which being compared with that of the same model but with 28 layers  (Figure 2 shows the vertical structure difference between 64 and 28 layers. )

• Ocean model:  GFDL MOM3

• Coupling:  Direct

• Initial condition taken from GDAS and GODAS

(Return to Top)

Significant Improvement

• Simulated SST climatology is reasonably good in the tropics and subtropics, where errors are generally within 1K. (Figure 3)

• Simulated surface momentum flux and sea surface height (SSH) climatology distribution are realistic.

• The model simulates ENSO variability with realistic frequency (period of 3-5 years) and  similar amplitude to that of the observed strong events.

• Simulated two leading EOF modes of SSH are similar to those from oceanic analysis.  The observed feature that the EOF1 and Nino3.4 SSTs are almost in phase and both lead EOF2 by about two to three seasons.

• Correctly simulates the observed ENSO seasonal phase-locking with the peak amplitude occurring near the end of the year.

(Return to Top)

Caveats:

• The ENSO events in the simulation occur more regularly than in observations.

• On average, simulated warm events tend to start about three months earlier and persist longer than observed.

• The amplitude of the simulated east-west gradient of SSH climatology appear to be smaller than that of the observation analysis.

(Return to Top)

Challenges

• Why is the vertical resolution so critical to the CFS prediction skill improvement?  Figure 4 shows comparable result obtained by the coupled ocean (ORCA) and atmosphere (ECHAM) model at T30 and 19 hybrid levels. (Guilyardi and Delecluse 2003)

• What model physics are the causes for the regularity of the simulated ENSO cycle, why the simulated ENSO lasts longer than the observed?

• What is the cause of the warm bias in the south eastern tropical Pacific?

• What is source of the long-term trend in the surface and subsurface temperature?

• How sensitive the simulated ENSO is to the model’s numerics and physics?

(Return to Top)

Reference

Wang, W., S. Saha, H. Pan, S. Nadiga and G. White, 2005: Simulation of ENSO in the new NCEP coupled forecast system model (CFS03).  Mon. Wea. Rev., 133, 1574 - 1593.