Upper Tropical Pacific Variation during recent ENSO:

An Evaluation of Global Ocean Data Assimilation System (GODAS)

4th Annual CTB SAB MeetingESSIC, UMD

11 September 2008

Dongxiao Zhang Yan Xue

Michael J. McPhaden David Behringer

JISAO/UW and NOAA/PMEL NOAA/NCEP

2005-08 NINO3.4 Predictions

• Like most other models, CFS had difficulty to predict the 2006 El Nino, even initialized in summer 2006.

• CFS tends to overshoot in prediction of either warm (Dec. 2006) or cold (Jan. 2008) events.

Objective

Evaluate the extent to which systematic errors in GODAS may limit skill in NOAA Climate Forecast System (CFS) forecasts

Compare GODAS analyses with ocean temperature, salinity, and velocity observations, emphasizing the period 2005-08

Approach

Data Products• GODAS

Model: GFDL MOM v3, 75°S-65°N

Horizontal resolution: 1° x 1° resolution (enhanced to 1/3° in latitude within 10 °of the equator)

Vertical resolution: 40 levels with 10-m resolution in upper 200m

Forcing: NCEP atmospheric reanalysis-2 wind stress; SST relaxed to weekly Reynolds analysis; SSS relaxed to annual mean climatology

Data assimilation: Temperature profiles from TAO, TRITON, ARGO, and XBT; “synthetic” subsurface salinity based on mean T/S

Available: 1980-present

• Near real-time PMEL temperature/salinity analysisData: Argo, TAO/TRITON (Temp & Salinity)

Resolution: 10 days, 0.5° lat x 0.5° long

Available: 2005-present

Analysis on density surfaces

Arrows show Subtropical Cell pathways in the thermocline

PMEL Argo-TAO/TRITON Analysis(Week Centered on 26 Dec 2007)

Text

Sal. 115m

SSS

Potential Temp 115m

U, V at 115m (ref. 1000db)

20°C isotherm standard deviation (2005-2008)PMEL GODAS

PMEL (seasonal cycle removed) GODAS (seasonal cycle removed)

• Z20 anomalies associated with ENSO development are reasonably simulated in GODAS, with higher amplitude than in observation.

• Z20 seasonal cycle amplitude in GODAS is however biased too high.

20oC isotherm anomalies 5°S-5°N Average (2005-08)

PMEL GODAS

Z20 anomalies associated with ENSO development are reasonably reproduced, though GODAS anomalies tend to be larger.

SSS standard deviation (2005-2008)

PMEL GODAS

PMEL (seasonal cycle removed) GODAS (seasonal cycle removed)

SSS variability is underestimated in GODAS.

Surface Salinity and Mixed Layer Depth, 5oS-5oN Average PMELGODAS

SSS

GODAS-PMEL

SSS

MLD

SSS Diff.

MLD MLD Diff.

White contour shows the 29°C isotherm, indicating the edge of warm poolLarge bias in MLD in western Pacific, presumably caused by salinity bias.

5oS-5oN Averaged Anomalies PMEL 20°C isothermZonal Wind Stress OI sst

•Slow eastward propagating of both downwelling Kelvin wave and SST anomalies during 2006 El Nino, after westerly anomalies in western Pacific, are characters of slow coupled waves.

•No propagating signature is observed in Z20 or SST during 2007-08 La Nina. Persistent shallow thermocline anomaly indicates entrainment cooling.

SSS at 165°E, 0°N vs. Nino3.4

Heat Budget in Nino3.4 Area

• Large dT/dt anomalies are forced by surface heat flux, which is however in response to damp anomalies from entrainment.

• Zonal advection is large, and tends to follow large flux anomalies.• Dominance of zonal advection and surface heat flux during summer 2006

suggests a role of slow coupled waves (Hirst 1986) in El Nino development.• Persistent entrainment cooling is important during 2007-08 La Nina.

Meridional Structure Along 140°W

1950-1999

Temperature

Salinity

Density surfacekg m

Shallow Overturning in the Subtropical Cells

U, V at 115m (ref. 1000db)

Persistent Entrainment Cooling vs. Subtropical Cell (STC)

Interior STC transport convergence is estimated from equatorward thermocline transports, marked by arrows, at 9°S and 9°N.

Entrainment anomalies follow STC convergence transport, which advects subducted subtropical cold waters equatorward to feed equatorial upwelling.

U, V at 115m (ref. 1000db)

•Stronger STC since 2007•Stronger equatorial upwelling•More entrainment cooling

PMEL

Summary• PMEL has developed a near real-time gridded T, S, and

geostrophic current product to validate GODAS over 2005-08.

• All terms in mixed layer heat budget equation, except vdT/dy, are important in controlling SST tendency term.

• The observed STC variability is well reproduced in GODAS, and contributes to the persistent entrainment cooling during 2007-08 La Nina.

• The equatorial Pacific heat content, or Z20, associated with the development of ENSO cycle agree reasonably well in GODAS and PMEL analyses, though GODAS variability tends to be larger.

• GODAS surface salinity variability is too weak and mixed layers are too deep in the western Pacific warm pool, with potential adverse affects on zonal currents, advection and ENSO SST development when used to initialize CFS.

Questions§ Why does the annual cycle variability in GODAS upper ocean

temperatures tend to be higher than observed? Does it contribute to the overshoot of CFS prediction during either warm or cold events?

§ What accounts for the systematic error in warm pool mixed layer currents and how do those errors affect temperature advection and ENSO SST evolution?

§ GODAS will benefit from assimilation of ocean salinity measurements. What is the optimal method for reducing systematic errors in analyzed fields?

Zonal velocity (cm/s) at 10m in TAO and GODAS

0°, 165°E

0°, 170°W

Surface current at 170°W is better simulated in GODAS than in western Pacific at 165°E, where a systematic bias is apparent.