Number of Casts

Ocean Reanalyses: Prospects for Climate Studies

James A. Carton (University of Maryland)

Thanks: Gennady Chepurin, Anthony Santorelli, You-Soon Chang

# pubs refering to 'ocean reanalysis'

0 50 100 150 200 250

1 9 85

1 9 87

1 9 89

1 9 91

1 9 93

1 9 95

1 9 97

1 9 99

2 0 01

2 0 03

2 0 05

2 0 07

C talks

r-Rosati ‘89

SODA Ocean Obs

‘99

Ocean Reanalyses -- Carton 2

Some motivating questions

• What climate signals can we detect?

– Where and when?

– How large?

– What level of diagnostic analysis is possible?

• How biased are the results?

– Are the signals we see real?

• How do we evaluate the error (and bias) in our analyses?

• What comes next?

Profile Obs Coverage

1930-1939

1960-1969

Ocean Reanalyses -- Carton 4

OSD cast data with time at NODC (picture from NODC)

0 10,000 20,000 30,000 40,000 50,000 60,000 70,000

1900 1905

1910 1915

1920 1925

1930 1935

1940 1945

1950 1955

1960 1965

1970 1975

1980 1985

1990

Year

N u m b e r o f C a s ts

GODAR as of WOD01 (2001): 1,050,509 casts

NODC (1991): 783,912 casts

Ocean Reanalyses -- Carton 5

Growth of ARGO since 2003

(pictures from ARGO website and

S. Wilson)

Ocean Reanalyses -- Carton 6

In-situ SST observation coverage

SST obs

Remotely sensed SST since 1981

Source: John Maurer, UC Boulder http://cires.colorado.edu/~maurerj/class/SST_presentation.htm

Ocean Reanalyses -- Carton 8

Most of this talk will focus on the time period 1960-2001 corresponding to

ERA40. At the end I will consider the full 20 ^th century.

I’ll begin by looking at ocean heat

content, essentially the vertical integral

of temperature. Then I’ll look at water

masses.

Global Heat Content 0/700m

-4 -2 0 2 4

1960 1970 1980 1990 2000

H e a t C o n te n t ( x 1 0 8 J m -2 )

Trend: 0.77x10

Jm

/10yr

Problem of time-dependent bias in the profile data

Levitus et al., 2005

Why would the ocean warm up for a decade, and then cool off again??

‘no-model’ analysis of

Ocean Reanalyses -- Carton 10

Assimilation/synthesis methodologies

• Sequential filters

• Smoothers

friction stress

U p U

U x k t f

U ∇ + +

−

=

∇

⋅ +

∂ +

∂

ρ r

r ) r

r

ns observatio x

background x

x Hx

R x

Hx x

x B

x x

x J

o b

o T

o b

T b

:

; :

) (

) (

) (

) (

)

( = −

− + −

−

ECMWF Training manual ECMWF Training manual

J=J[X(t)]

‘physically

consistent’

Eight examples

Objective Analysis 1962-2001

UK-OI

sequential 1962-2001

ECMWF

Sequential 1962-1998

UK-FOAM

Bell. (2000), Bell et al. (2004)

Sequential 1958-2005

SODA

Carton and Giese (2007)

Objective analysis 1955-2003

LEVITUS

Levitus et al. (2005)

Objective analysis 1945-2005

ISHII

Ishii et al. (2006)

Sequential 1962-2001

INGV Davey (2005)

Sequential 1979-2005

GODAS Behringer (2005)

Sequential, Coupled Sequential 1955-1999

1980-2005 GFDL 1,2

Sun et al. (2007)

4DVar 1950-1999

GECCO

Köhl et al. (2006)

Sequential 1962-2001

CERFACS Davey (2005)

Analysis procedure Time Span

Analysis

Ocean Reanalyses -- Carton 12

Global heat content

Global Heat Content 0/700m

-4 -2 0 2 4

1960 1970 1980 1990 2000

H e a t C o n te n t ( x 1 0 8 J m -2 )

CERFACS GFDL

ISHII MEAN

Sato

Trend: 0.77x10⁸ Jm^-2/10yr

Aerosol forcing

Bathythemograph fall rate corrections

L09

W08 XBT

(from Sippican)

Ocean Reanalyses -- Carton 14

Global heat content after obs correction

Assim experiment using L09

Assim experiment using W08

Impact of bias correction on mean

tropical circulation

Ocean Reanalyses -- Carton 16

Heat Content by decade

Vertical/Time Structure

1960-1969 1970-1979 1980-1989 1990-1999

Correlation with Pacific Decadal Oscillation

Colors – heat content Contours - SST

North Pacific Heat Content 0/700m

-4 -2 0 2 4 6

1960 1970 1980 1990 2000

H e a t C o n te n t ( x 1 0

J m

)

Much of the decadal variability is

correlated with PDO

Ocean Reanalyses -- Carton 18

Decadal N. Pacific density variations

Depth of sigma 25.5 surface

(Miller and Schneider, 2000)

Heat Content by Decade: Indian Sector

1960s 1970s 1980s 1990s

Ocean Reanalyses -- Carton 20

Quick Look at Upper Ocean Water Masses

Examples of the upper ocean response to freshwater events:

• HOT

• Bermuda

• Great Salinity Anomalies

SST

-SST

during winter

McPhaden and Zhang (2002)

Change in depth (meters) of the 24.5 σ

surface averaged 1990-1999 minus

1970-1977

Ocean Reanalyses -- Carton 22

Response of the North Pacific to Heavy precipitation (’95-’97)

Hawaii Ocean Time series

Lukas (2001)

Salinity Anomalies at HOT showing penetration of near-surface freshwater

Heavy rainfall

Time

Depth

Salinity Precip

PV variability at Bermuda

Joyce and Robbins (1996)

Normalized PV from the Analyses

Ocean Reanalyses -- Carton 24 Dickson et al (1988) revised:

Ellett and Blindheim (1992, Fig.

6)

Great Salinity Anomalies

Annually averaged upper ocean salinity (0-500m) in the Norwegian Basin (0-5oE, 63oN-69oN) for the seven analyses spanning the time period. ECMWF becomes quite fresh after 1990.

0/250m Salinity changes within the southern Labrador Sea (53oW-59oW, 50oN-56oN).

ECMWF gets extremely fresh Lab Sea GSAs appear in 5 of the analyses

Ocean Reanalyses -- Carton 26

Spatial structure of 0/500 salinity’

0/250m Salinity changes within the southern Labrador Sea (53oW-59oW, 50oN-56oN).

20 ^th Century Reanalysis:

ENSO

From: Giese et al. (2009)

reconstructed SST simulated SST reanalysis SST (blue)

Ocean Reanalyses -- Carton 28

From: Giese et al. (2009)

Ocean Reanalyses -- Carton 30

From: Giese et al. (2009)

What have we learned?

• What climate signals can we detect?

– See above.

• How biased are the results?

– Observation bias is certainly present, but seems mainly to afflict basin- or global integral quantities. Model bias (including bias in

meteorological forcing) does not seem insurmountable. But we still aren’t sure just how large it is.

• Are the signals we see real? Are we learning new things?

– Yes. Beginning to. This is a new tool and the community is just getting used to it.

• How do we evaluate the error (and bias) in our analyses?

– Unbiased data sets like ctd/osd are uniquely valuable for this.

• What next?

– Ensemble methods seem like a logical next step.

– Analyses really should be done in the coupled system.

– Impact of circulation on ecosystem models.