1 Comparative Analysis of Upper Ocean Heat Content Variability from an Ensemble of Operational Ocean Analyses Yan Xue (1), Magdalena A. Balmaseda (2),

1

Comparative Analysis of Upper Ocean Heat Content Variability from an Ensemble of Operational Ocean Analyses

Yan Xue (1), Magdalena A. Balmaseda (2), Tim Boyer(6) ,Nicolas Ferry

(3) , Simon Good (4), Ichiro Ishikawa (5) , Arun Kumar(1) Michele

Rienecker (7), Tony Rosati(8), Yonghong Yin(9)

(1) NOAA/NCEP, 5200 Auth Rd, Camp Springs, MD 20746, USA, [email protected], [email protected]

(2) ECMWF, Shinfield Park, Reading RG2 9AX (UK), [email protected]

(3) Mercator-Océan, 8-10 rue Hermès, 31520 RAMONVILLE ST AGNE (France), [email protected]

(4) Met Office Hadley Centre, FitzRoy Road, Exeter, Devon, EX1 3PB (UK), [email protected]

(5) Japan Meteorological Agency, 1-3-4 Ootemachi, Chiyoda-ku, Tokyo, 100-8122 (Japan), [email protected]

(6) NOAA/ NESDIS/NODC, 1315 East-West Highway, Silver Spring, MD 20910, USA, [email protected]

(7) NASA/GSFC/GMAO, Greenbelt, MD 20771 (USA), [email protected]

(8) NOAA/GFDL, Princeton University, P.O. Box 308, Princeton, NJ 08542. [email protected]

(9) CAWCR , GPO Box 1289, Melbourne, VIC 3001 (Australia), [email protected]

Workshop on Observing System Evaluation and Intercomparisons, June 13-17, 2011, Santa Cruz, CA

2

Focus regionsFocus regionsFocus regions

100E 160W 60WLongitude

50S

0

50N

Latit

ude

EQPAC

TRPAC

NPAC

EQATL

TRATL

NATL

EQIND

GSOP Upper 300m Heat Content Comparison, Reading, 2006(call for validation of climate signals)

12m-rm seasonal anom: TRPAC Averaged temperature over the top 300m

1950 1960 1970 1980 1990 2000Time

-0.4

-0.2

0.0

0.2

0.4

0.6

ukdpukoicfcs2cfas2ecco50y

gfdlsodaecmfaecmfcukgs

ingvmri-eccoSIOcfasamct2

mct3eccoJPLaeccoJPLceccoMITGMAO

sdv ensm = 0.071s/n ensm = 0.532

sdv all = 0.108s/n all = 0.803

spread = 0.134

12m-rm seasonal anom: TRATL Averaged temperature over the top 300m

1950 1960 1970 1980 1990 2000Time

-0.8

-0.6

-0.4

-0.2

0.0

0.2

0.4






sdv all = 0.144s/n all = 1.172

spread = 0.123

12m-rm seasonal anom: TRIND Averaged temperature over the top 300m

1950 1960 1970 1980 1990 2000Time

-0.6

-0.4

-0.2

0.0

0.2

0.4






sdv all = 0.131s/n all = 1.137

spread = 0.115

•The North Pacific does not show a warming trend, but more of a rapid shift in the early 90’s

•Large uncertainty after 2000

•Phase/amplitude of decadal signal is poorly resolved. Outliers.

12m-rm seasonal anom: NPAC Averaged temperature over the top 300m

1950 1960 1970 1980 1990 2000Time

-0.8

-0.6

-0.4

-0.2

0.0

0.2

0.4






sdv all = 0.156s/n all = 1.040

spread = 0.150

12m-rm seasonal anom: NATL Averaged temperature over the top 300m

1950 1960 1970 1980 1990 2000Time

-0.8

-0.6

-0.4

-0.2

0.0

0.2

0.4

0.6






sdv all = 0.206s/n all = 2.028

spread = 0.101

North Pacific North Atlantic

Tropical Pacific Tropical Atlantic Tropical Indian Ocean

Courtesy of Magdalena Balmaseda

3

Name Method& Forcings

In SituData

AltimetryData

Resolution Period Vintage Reference

EN3.v2a Analysis Correction Scheme

No XBTcorrections

No 1°x 1°, 42 LevelsMonthly Temp.

1950-present

2009 Ingleby and Huddleston (2007)

NODC Objective Analysis

No XBT corrections

No 1°x 1°, 16 Levels, 0 to 700mSeasonal Temp.

1955-present

2010 Levitus et al. (2009)

GODAS 3D-VAR

No XBT corrections

NO (Yes in real time)

1°x 1° (1/3° near Eq), 40 LevelsPentad, Monthly

1979-present

2003 Behringer and Xue (2004

ECMWF(S3)

OI

No XBT corrections

Yes 1°x1° (1/3° near Eq), 29 LevelsDaily, Monthly

1959-present

2007 Balmaseda et al. (2008)

JMA 3D-VAR

No XBT corrections

Yes 1°x1° (1/3° near Eq), 50 LevelsPentad, Monthly

1979-present

2009 Usui et al. (2006)

CFSR 3D-VARPartially coupled

No XBT corrections

No (Yes in real time)

1/2°x 1/2° (1/4° near Eq), 40 LevelsDaily, Pentad, Monthly

1979-present

2010 Xue et al. (2010)

GFDL EnKFFully coupled

XBT corrections

Yes 1°x 1° (1/3° near Eq), 50 Levels Daily, Pentad, Monthly

1970-present

2010 Zhang et al. (2009)

GMAO EnOIFully coupled

XBT corrections

No 1/2°x 1/2° (1/4° near Eq), 40 LevelsDaily, Monthly

1980-present

2011 Rienecker at al. (2011)

MERCATOR(PSY2G2)

KF-SEEK No XBT corrections

Yes 2°x 2° (1/2° near Eq), 31 LevelsDaily, Pentad, Monthly

1979-present

2007 Drévillon et al. (2008)

BOM(PEODAS)

EnKF No XBT corretions

No 2°x 1.5 ° (1/2° near Eq.), 25 Levels Daily, Monthly

1980-present

2009 Yin et al. (2010)

OceanObs09 Community White Paper by Xue et al. (2010)(call for operational monitoring of climate signals)

4

In Situ Observations from Saha et al. (2010)

Argo

(2003-09)

Altimetry

(1993-2002)

Pre-Altimetry

(1985-1992)

5

• How well is the mean HC300 analyzed by operational ocean reanalysis (ORA)?

• What are impacts of changes in the ocean observing system on the quality of HC300 analysis?

• How well does operational ORA capture interannual variability, multi-decadal and long term variability in HC300?

• What are prospects for monitoring climate indices using an ensemble of operational ORAs?

• What are roles of HC300 on potential predictability of ENSO, Indian Ocean Dipole and Atlantic Nino?

Questions

6

Mean Heat Content in 1985-2009

7

RMSD from EN3 in Different Ocean Basins

8

HC300 in Equatorial Pacific (2oS-2oN)

1993

1999

2003

9

HC300 in Equatorial Indian Ocean (2oS-2oN)

2003

1997

1997

10

HC300 in Equatorial Atlantic (2oS-2oN)

2005

2005

11

HC300 Anomaly Correlation with EN3

12

HC300 Anomaly Correlation with OI SST

13

E. Pac W. Pac. S.E. Ind. S.W. Ind. E. Atl.

EN3 0.80 0.39 0.16 0.42 0.57

NODC 0.74 0.37 0.16 0.42 0.47

GODAS 0.78 0.39 0.25 0.41 0.46

ECMWF 0.78 0.41 0.32 0.48 0.58

JMA 0.77 0.38 0.36 0.51 0.45

CFSR 0.75 0.37 0.30 0.40 0.35

GFDL 0.76 0.38 0.24 0.45 0.60

GMAO 0.77 0.36 0.29 0.42 0.48

MERCATOR 0.78 0.41 0.39 0.47 0.56

BOM 0.79 0.39 0.34 0.45 0.47

mean 0.77 0.39 0.28 0.44 0.50Spread/mean 5% 11% 29% 11% 18%

Table 3. Anomaly correlation between HC300 and SST in 1982-2009 averaged in various ocean basins drawn as boxes in Fig. 8. Correlations less than 0.4 are in bold.Anomaly Correlation between HC300 and SST

14

HC300 Anomaly Indices for ENSO, IOD and Atlantic Nino

ENSO

IOD

Atlantic Nino

15

Linear Trend of HC300 Anomaly in 1993-2009

16

1995

2004

1999

HC300 Anomaly Indices for Multi-decadal Variability

17

HC300 and HC300 Anomaly Average in 70oS-70oN

El ChichonMt. Pinatubo

2003

18

Impacts of HC300 on SST Anomaly

19

Role of HC300 for Predictability of ENSO

-Equatorial western Pacific HC300a leads NINO3.4 by 10-12 months.

- The lead/lag relationship between HC300a and NINO3.4 is consistently analyzed by all ORAs.

- NODC is too noisy near the equator.

- The peak correlation between HC300a and NINO3.4 in CFSR is lower than others.

20

Role of HC300 for Predictability of Atlantic Nino

-Equatorial eastern Atlantic HC300a leads Atlantic Nino by 4-6 months.

-Equatorial western Atlantic HC300a leads Atlantic Nino by 8-12 months.

-Differences in the lead/lag correlation are quite large.

21

Role of HC300 for Predictability of IOD

- Southeastern Indian Ocean HC300a leads IOD by 2 months.

- Southwestern Indian Ocean HC300a leads IOD by 8-12 months.

- Differences in the lead/lag correlation are quite large.

22

Summary• Consistency in mean HC300 is generally high south of 30oS except

near the Gulf Stream and Kuroshio-Oyashio Extension (KOE), in the tropical Atlantic, and some regional seas such as Gulf of Mexico and South China Sea.

• Consistency, measured by root-mean-square differences from EN3, tends to increase with time, particularly in the tropical Pacific, the tropical Indian Ocean and extra-tropical southern oceans, which are partly due to constraints from a dense observational network from tropical mooring arrays and Argo floats.

• Consistency in the equatorial Pacific HC300 increased significantly after 1993 when the TAO mooring array was fully implemented.

• Consistency in the equatorial Indian Ocean HC300 increased significantly around 2003 when the two RAMA mooring lines were implemented at 80.5o E and 90o E. All model-based HC300 are too cold relative to EN3 in the eastern tropical Indian Ocean before 1997.

• Consistency in the equatorial Atlantic HC300 is much lower than that in the equatorial Pacific and Indian Ocean, but it improved significantly after 2006.

23

Summary• HC300 anomalies (HC300a) associated with ENSO are highly

consistent among ORAs; HC300a associated with IOD are moderately consistent, and model-based analyses are superior to in situ-based analyses in the eastern pole of the IOD; HC300a associated with the Atlantic zonal mode have considerable uncertainties among ORAs, which are comparable to signals.

• Large multi-decadal variability and long-term trends exist in HC300. The consensus among ORAs suggests that the mean HC300 in 70oS-70oN has brief cooling periods during early 1980s and 1992-1993 related to the volcanic eruptions of the El Chichon and Mt. Pinatubo, and a short warming in 1985-1991, and then a continuous warming in 1994-2003, followed by a persistence or weak cooling in 2004-2009.

• Despite of many advances in ORAs in the past decade, uncertainties in HC300 are still large near western boundary currents, in the tropical Atlantic and extra-tropical southern oceans. To improve HC300 analysis in those regions, additional advances, such as improving surface fluxes, model physics, model resolution and data assimilation schemes, will be needed in conjunction with maintaining and enhancing ocean observing systems in those regions.

Documents

1 Comparative Analysis of Upper Ocean Heat Content Variability from an Ensemble of Operational Ocean Analyses Yan Xue (1), Magdalena A. Balmaseda (2),