on the other side of the air/sea interface on the other side of the air/sea interface a title...

On the Other Side of the On the Other Side of the Air/Sea InterfaceAir/Sea Interface

A title conceived before I knew thatManuel and Corinne were also giving talks

Andy Jacobson, Nicolas Gruber, Manuel Gloor, Jorge L. Sarmiento,

Christopher L. Sabine, and Richard A. Feely

13 May 2003TransCom meeting in Jena

Special thanks to Robert M. Key and Kitack Lee for providing data.

Road MapRoad Map

TransientFootprints

AnthropogenicInversion

EqulibriumFootprints

PreindustrialInversion

Contemporary inversion,∆pCO2, gas exchange

AtmosphericFootprints

JointInversion

Recent Carbon SurveyRecent Carbon SurveyNumber of Observations ~ 60000

C* of Gruber, Sarmiento, and Stocker (1996) to estimate anthropogenic DIC.

Innumerable data authors, but represented by Feely, Sabine, Lee, Key.

The C* MethodThe C* Method

DIC C* = DIC - ∆Cbio

Salinity (psu) Salinity (psu)

Source: Gruber, Sarmiento, and Stocker (1996) GBC 10(4)

∆∆CCantant

∆∆CCgasexgasex

Dissolved Inorganic Carbon (DIC) and C* on AAIW Density Surface

∆Cbio from soft-tissue (AOU or P) and carbonate (Alk) changes

∆Cgasex = sC* - ∆Cant

Anthropogenic Anthropogenic CarbonCarbon

Inventory Inventory

Dye29 RegionsDye29 Regions

Conformable to the 11 TransCom3 oceanregions; allows direct comparison

Dye Flux PatternsDye Flux Patterns

Takahashi et al.(2002) CO2 flux pattern ( ∫∫dxdy = 1 )

daSilva et al. heat flux pattern ( ∫∫dxdy = 1 )

Eastern Tropical South Pacific

• Uniform• Heat Flux• CO2 Flux• “Forward” (OCMIP2 biotic)

Spatial distribution of the unit fluxwithin each region.

Sensitivity Analysis:Sensitivity Analysis:Transport and Surface Flux PatternTransport and Surface Flux Pattern

OGCMConfiguration

Takahashi CO2 flux pattern

Uniform flux pattern

Heat flux pattern

“Forward”pattern

Ai low, Kv low“LL”

Ai high, Kv high“HH”

Ai low, Kv high South“LHS”

ECMWF, ndp, 4pt salinity rest,

Ai low, Kv med-HiS (2000m) “PSS”

Ai low, Kv med-HiS

(2000m),4pt salinity rest ”RDS”

MOM3 Southern Ocean MOM3 Southern Ocean CFCs and RadiocarbonCFCs and Radiocarbon

ObservationsObservations

““Standard” Standard” configurationconfiguration Courtesty

of KatsumiMatsumoto

Modeled Southern Ocean Modeled Southern Ocean Anthropogenic COAnthropogenic CO22 and Radiocarbon and Radiocarbon

Courtesy of Katsumi MatsumotoNote: values not yet final

Modeled Southern Ocean Modeled Southern Ocean CFCs and RadiocarbonCFCs and Radiocarbon

Courtesy of Katsumi MatsumotoNote: values not yet final

Anthropogenic Basis FunctionsAnthropogenic Basis Functions1765 - 2000: aggregated and vertically integrated

MOM3 rds MOM3 pss

PacificTropical

(16-19)

PacificSubpolar

(22-24)

Results for Anthropogenic Carbon InversionsResults for Anthropogenic Carbon Inversions

First, the diagnostics--can we trust the inversions?

1. Simulations with synthetic data show that this inversion is very stable. Retrieved fluxes differ by O(10-3) PgCyr-1reg-1 from known “true” values. This is as expected from the small condition number of the A matrices, and is a manifestation of the uncorrelated footprints.

2. 2 looks pretty good. E.g., 53174 with 60246 DOFs (ratio 0.88)

3. Analysis of residuals is ongoing. Outliers are frightening.

Global Ocean Uptake of Anthropogenic COGlobal Ocean Uptake of Anthropogenic CO22 PgC/yr into Ocean, in 1995PgC/yr into Ocean, in 1995

OGCMConfiguration

Takahashi CO2 flux pattern

Uniform flux pattern

Heat flux pattern

“Forward”pattern

Ai low, Kv low“LL”

1.93 1.99

Ai high, Kv high“HH”

2.31 2.27 2.34

Ai low, Kv high South“LHS”

2.05 2.10

ECMWF, ndp, 4pt salinity rest,

Ai low, Kv med-HiS (2000m) “PSS”

2.27 2.35 2.33

Ai low, Kv med-HiS (2000m),

4pt salinity rest ”RDS”2.19 2.23 2.17 2.23

Across-model Mean Sink: 2.2 ± 0.3 PgC/yr (1995)

Ocean Sink Estimates Ocean Sink Estimates (after LeQuéré et al., 2003)(after LeQuéré et al., 2003)

Recent Recent DecadesDecades

(after LeQuéré et al., 2003)(after LeQuéré et al., 2003)

New inversion estimates:

1985: -1.7 ± 0.21990: -1.9 ± 0.21995: -2.2 ± 0.32000: -2.5 ± 0.3

Uncertainties for current results are two standard deviations across all 14 model runs.

Current Estimate

LL HH

PSS RDS

Circulation SensitivityCirculation SensitivityAnthropogenic Carbon Flux into Ocean (mol C m-2 yr-1, for “forward” pattern)

Simple Pycnocline Depth ModelSimple Pycnocline Depth Model

equatorial upwellingequatorial upwelling

NADW NADW formationformation

dense waterdense water

light waterlight waterDD

NorthSouth Equator

Pycnocline depth D sets the NADW formation rate. Our models are all configured to have about the same pycnocline depth.

Return flow is a via diffuse equatorial upwelling, and thus a condition is set upon the magnitude of vertical diffusivity Kv.

Gnanadesikan (1999)Gnanadesikan (1999)

Return flow is a balance of upwelling both at the equator and in the Southern Ocean. This balance is set by along-isopycnal diffusivity Ai and vertical diffusivity Kv.

Recall that pycnocline depth--and thus NADW formation rate--isheld (nearly) constant

equatorial upwellingequatorial upwelling

NADW NADW formationformation

SouthernSouthernOceanOcean

upwellingupwelling

dense waterdense water

light waterlight waterDD

NorthSouth Equator

Watermass Transformation Rates Watermass Transformation Rates

Circulation Model

SouthernOcean

Upwelling(Sv)

EquatorialUpwelling

(Sv)

NADWFormation

(Sv)

Ai low, Kv low“LL”

7.63 3.67 11.3

Ai high, Kv high“HH”

-12.5 24.5 12.0

Ai low, Kv high South“LHS”

8.59 2.81 11.4

ECMWF, ndp, 4pt salinity rest, Ai low, Kv med-

HiS (2000m) “PSS”

14.8 -2.0 12.8

Ai low, Kv med-HiS (2000m),

4pt salinity rest ”RDS”

6.81 5.39 12.2

Pathway of Return Flow

Transformation rates diagnosed from models by analyzing the meridional transport of light (0 < 27.4) waters.

LL

LHS

HH

PSS

RDS

Uni Heat Tak Fwd

Global Anthropogenic Carbon Flux Global Anthropogenic Carbon Flux vs.vs.

NADW Formation RateNADW Formation Rate

NADW Formation Rate (Sv)

An

thro

Flu

x (

Pg

C/y

r)

Transport sensitivity: range of 0.4 PgC/yr

Pattern sensitivity: range of 0.08 PgC/yr

Southern Ocean

Low Latitudes

LL

LHS

HH

PSS

RDS

Uni Heat Tak Fwd

Regional Fluxes vs. Regional Transformation RatesRegional Fluxes vs. Regional Transformation Rates

Inter-Pentadal Variability?Inter-Pentadal Variability?

DO NOT CITE

This interpretation, while inconsistent with prior assumptions, suggests that an interpentadal signal is present in the data.

Forward and Inverse Anthropogenic DICForward and Inverse Anthropogenic DICSame five models in all cases

Preindustrial InversionPreindustrial Inversion

Evidence for SeasonalEvidence for SeasonalRectifierRectifier

DYE29_LL

RECT29_LL

PreindustrialPreindustrialCarbon FluxCarbon Flux

InversionsInversions

Zonally-Integrated Preindustrial FluxZonally-Integrated Preindustrial Flux

Pre

ind

ust

rial Fl

ux (

mol C

deg

-1 y

r-1)

DYE29 unregularized Gloor 13 region

Zonally-Integrated Preindustrial FluxZonally-Integrated Preindustrial Flux

Pre

ind

ust

rial Fl

ux (

mol C

deg

-1 y

r-1)

DYE29 unregularized Gloor 13 regionDYE29 SVD: 23 retained

DYE29 SVD: 15 retained

Zonally-Integrated Preindustrial FluxZonally-Integrated Preindustrial Flux

Pre

ind

ust

rial Fl

ux (

mol C

deg

-1 y

r-1)

DYE29 unregularized Gloor 13 regionDYE29 SVD: 23 retained

DYE29 SVD: 15 retainedDYE29 aggregated to Gloor regions (22)

Covariate Data ErrorsCovariate Data Errors

Recall cost function:

Biases manifested as off-diagonal covariances in C.

N.B. ij = ji and

ii = i2.

Multivariate normal PDF

Minimizing J still maximizes the likelihood(but you have to do it numerically).

Ongoing ProjectsOngoing Projects

• Characterization of C* biases (Sarah Fletcher, Andy)

• Transport sensitivity: OCMIP group to produce Green’s functions (Sarah Fletcher). Also use MOM4 and HIM at Princeton.

• Time-dependent anthropogenic inverse, possibly with bomb radiocarbon and CFC constraints (Andy)

• Can our estimates tell us anything about gas exchange parameterizations? (Andy, Manuel)

• Joint atmosphere-ocean inverse.

fin