CONSTRAINTS FROM ATMOSPHERIC MEASUREMENTS ON THE GLOBAL BUDGET OF

CARBONYL SULFIDE

Parv Suntharalingam

University of East Anglia

N. Krakauer1, A.J. Kettle2, D. J. Jacob3, S. Montzka4 1: City College; 2 : University of East Anglia; 3: Harvard; 4 : NOAA-ESRL

GEOS-Chem Meeting, April 7-10, 2009

WHY INVESTIGATE COS ?

COS

STRATOSPHERE

TROPOSPHERE

WHY INVESTIGATE COS ?

• Source of stratospheric sulfate aerosol; impacts on radiation budget and stratospheric chemistry

COS

Sulfate Aerosol

STRATOSPHERE

TROPOSPHERE

WHY INVESTIGATE COS ?

• Source of stratospheric sulfate aerosol; impacts on radiation budget and stratospheric chemistry

• Plant uptake of COS is proportional to that of CO2.

Can COS measurements provide a constraint on Global Primary Productivity ?

COS CO2

Sulfate Aerosol

STRATOSPHERE

TROPOSPHERE

SEASONAL CYCLES OF COS and CO2 COS measurements : S. Montzka (NOAA-ESRL)

CO2 SH

COS

J F M A M J J A S O N D J F M A M J J A S O N D

Similar seasonal cycles for COS and CO2, especially at forested sites.

CAN COS MEASUREMENTS CONSTRAIN BIOSPHERIC CO2 UPTAKE ?

Significant uncertainties remain on aspects of COS budget

Slide from S. Montzka

SOURCES AND SINKS OF ATMOSPHERIC CARBONYL SULFIDE

COS

Mean atmospheric conc. ~ 500 ppt

SOURCES AND SINKS OF ATMOSPHERIC CARBONYL SULFIDE

SOILS PLANTS ANTHROPOGENIC OCEAN

CS2

DMS

COS

Atmospheric Oxidation, Stratospheric Photolysis

Oxidation

Biomass Burning

SOURCES AND SINKS OF ATMOSPHERIC CARBONYL SULFIDE

SOILS PLANTS ANTHROPOGENIC OCEAN

CS2

DMS

COS

Atmospheric Oxidation, Stratospheric Photolysis

Oxidation

Biomass Burning

?? ?? ?? ??

Uncertainty remains on source/sink fluxes

Rainwater ??

PREVIOUS GLOBAL BUDGET ESTIMATES

OCEAN

ANTH

FIRE

PLANTS

SOILS

ATM LOSS

IMBALANCE IMBALANCE

PRECIP +280

-260

Chin and Davis 1993 Watts 2000

Gg S

/yr

• Large uncertainties

• Net imbalances of opposite sign

• Revision of soil flux term

SOILS

-400

-20

0

0

200

-400

-20

0

0

200

Spatial and Temporal Variation of COS Fluxes Kettle et al. [2002]

+65

Anthropogenic

Plants Soils

Ocean

• Process-based parameterizations of COS sources and sinks (monthly fields; gridded 1ox1o or 4ox5o)

• Budget closed with limits of uncertainties

Gg

S/yr

IMBALANCE Sources = 555 GgS/yr

(170-1010)

Sinks = 490 GgS/yr

(380-600)

OCEAN

ANTH

FIRE

PLANTS

SOILS ATM LOSS

Spatial and Temporal Variation of COS Fluxes Kettle et al. [2002]

+65

Anthropogenic

Plants Soils

Ocean

• Process-based parameterizations of COS sources and sinks (monthly fields; gridded 1ox1o or 4ox5o)

• Budget closed with limits of uncertainties

• Some uncertainties may be underestimated

Gg

S/yr

IMBALANCE Sources = 555 GgS/yr

(170-1010)

Sinks = 490 GgS/yr

(380-600)

OCEAN

ANTH

FIRE

PLANTS

SOILS ATM LOSS

RECENT ESTIMATES OF GLOBAL COS UPTAKE BY PLANTS SUGGEST LARGER SINK

Sandoval- Soto et al. 2005 (Method)

730 -1500

Xu et al. 2004 960 -1490

Montzka et al. 2007 730 - 1500

Units : Gg S/year

Kettle et al. 2002 210-266

Recent estimates are 3-6 times higher than Kettle et al. [2002]

ATMOSPHERIC CONSTRAINTS ON SEASONAL COS FLUXES Suntharalingam et al. [GRL, 2008]

• Matching observed seasonal cycle requires :

- Increase in Plant uptake by 250 Gg S/yr (~100%)

- Decrease in Southern ocean fluxes by 50 Gg S/yr (~ 20%)

• Additional (tropical) source of 250 Gg S/yr required to close budget

Observations (NOAA-GMD)

Kettle2002 Model

‘Increased Plant Uptake’ model

THIS WORK : INVERSE ANALYSIS OF ATMOSPHERIC COS MEASUREMENTS

AIM : Derive optimal estimates of COS sources and sinks consistent with atmospheric measurements and prior flux distributions

APPROACH : Bayesian inverse analysis [Rodgers 2000]

INVERSE ESTIMATES OF COS FLUXES

Prior Fluxes (xa)

Kettle et al. [2002]

Forward Model

GEOS-Chem v. 7-03-06 2ox2.5o; 30 levels

INVERSE MODEL Minimize cost function of Model-Observation concentration mismatch, and deviation from

prior fluxes

J(x) = (y – ym)T Se –1 (y – ym)

+ (x – xa)T Sa –1 (x - xa)

Optimal Flux Estimates (xp)

Modelled Concentrations

Observations

ym

y

FLUX ESTIMATES FROM INVERSE ANALYSIS

Minimize cost function:

J(x) = (x – xa)T Sa –1 (x - xa) + (y – K x)T Sε –1 (y –K x)

Solution:

x = xa + G (y - K xa)

where, G = Sa KT (K Sa KT + Sε) -1

A posteriori errors : S = (KT Sε –1 K + Sa –1) -1

x = state vector (fluxes)

xa = a priori flux estimate

K = Jacobian matrix (model transport)

Sa = Error covariance matrix on fluxes

Sε = Error covariance matrix on concentration error

Rodgers, 2000

THIS ANALYSIS •  Measurements : Monthly means of flask samples at 11 surface sites from NOAA-ESRL network (2001-2007) [Montzka et al 2007]

•  Prior fluxes : Kettle et al. [2002].

•  State vector : 8 COS fluxes (global, annual mean)

ONGOING WORK : Inverse analyses at higher spatial and temporal resolution. E.g., monthly estimates of regional COS fluxes

THIS WORK : INVERSE ANALYSIS OF ATMOSPHERIC COS MEASUREMENTS

AIM : Derive optimal estimates of COS sources and sinks consistent with atmospheric measurements and prior flux distributions

APPROACH : Bayesian inverse analysis [Rodgers 2000]

Soil Plant Anthropogenic Ocean Direct Ocean CS2 Ocean DMS Atmospheric Loss Biomass Burning

X =

COS SURFACE OBSERVATION NETWORK : NOAA-ESRL Flask measurements since 2000 [Montzka et al. 2007]

Barrow (BRW)

Mauna Loa (MLO)

South Pole (SPO)

•

•

• •

• • • •

•

• cgo

mhd

alt brw

kum mlo

smo

spo

nwr hfm lef

From S. Montzka

Measurements used here : Monthly averages from 11 sites (2001-2007)

ERROR COVARIANCE OF THE CONCENTRATION ERROR Specifying Matrix Sε

Elements of matrix Sε account for : 1.  Instrument error in measurements

2.  Representation error

3.  Forward model error (transport, aggregation error, chemistry)

Characterization of Sε in this analysis :

•  Diagonal matrix; errors derived from the residual standard deviation of variability in monthly measurements at each site [‘TransCom’ approach: Gurney et al. 2003]

•  Sensitivity analyses conducted for different weightings of elements of Sε

WORK IN PROGRESS to separately account

for these terms

INVERSE ESTIMATES : Sensitivity Analyses

Cost function:

J(x) = (x – xa)T Sa –1 (x - xa) + (y – K x)T Sε –1 (y –K x)

Prior Uncertainties Sa • ‘Tight priors’ (Kettle 2002) • ‘Loose’ priors (Kettle 2002 x3; and plant uptake uncertainty x6) • Intermediate prior uncertainties (between bounds above)

Prior Fluxes xa • Kettle et al. 2002 • Montzka et al. 2007

Observational error covariance Sε

• ‘TransCom’ weighting of sites • Constant weighting of all sites • Subsets of sites : Land vs. Ocean sites

Sensitivity analyses for combinations of the above : 16 cases

SENSITIVITY ANALYSES: Variation of Cost Function Parameters

MODEL REPRESENTATIONS OF DIFFERENT PRIOR COS FLUXES HAVE SIMILAR ATMOSPHERIC SIGNATURES

2 x Plant uptake

3 x Soil uptake

2 x Ocean direct flux

2 x Ocean DMS oxidation flux

• This results from colocation of fluxes, and similar seasonal variations

• Prior fluxes derived from limited information (e.g., 1 soil study; 1 DMS-to-COS yield study). Need for improved characterization of underlying processes

Prior Fluxes : Kettle et al. [2002]

ERROR CORRELATION MATRIX (Prior fluxes : Kettle et al. 2002)

1.00 -0.81 0.23 -0.13 -0.07 0.09 0.04 0.01

-0.81 1.00 0.22 0.31 -0.01 -0.19 -0.17 0.32

0.23 0.22 1.00 0.29 0.20 -0.13 -0.09 0.05

-0.13 0.31 0.29 1.00 0.15 -0.64 -0.08 0.25

-0.07 -0.01 0.20 0.15 1.00 -0.55 0.24 -0.52

0.09 -0.19 -0.13 -0.64 -0.55 1.00 0.02 -0.16

0.04 -0.17 -0.09 -0.08 0.24 0.02 1.00 0.21

0.01 0.32 0.05 0.25 -0.52 -0.16 0.21 1.00

SOIL

PLANT

ANTH

OCEAN Direct

OCEAN CS2

OCEAN DMS ATM. LOSS

FIRES

SOIL

PLAN

T

ANTH

OCEA

N Di

rect

OCEA

N CS

2

OCEA

N DM

S

ATM.

LOS

S

FIRE

S

Large correlation between Plant and Soil fluxes

Large correlation between Ocean fluxes

SUMMARY OF POSTERIOR FLUX ESTIMATES Comparison to Kettle et al. [2002] Prior

PLANT + SOIL OCEAN

ANTH

ATM LOSS

“FIRES”

Gg S

/yr

• Plant and Soil : Increased uptake (by ~50%)

• Ocean fluxes reduced - close to zero. May be due to aggregation error

• “Fires” : significant increase. May represent unidentified tropical sources

• Ongoing work : flux estimates on regional and monthly basis Kettle 2002 Prior Kettle 2002 Prior

LATITUDINAL DISTRIBUTION OF PRIOR FLUXES [Kettle et al. 2002]

FIRES

PLANT

SOIL

OCEAN

ANTHROPOGENIC

Gg

S / (

degr

ee y

r)

SUMMARY

Preliminary inverse estimates of global COS fluxes indicate:  Increased plant and soil uptake (by ~ 50%)

 Significant decreases in ocean flux (may be due to aggregation error)

 Increased COS fluxes from “Fires” (may be accounting for unrepresented tropical sources)

ONGOING AND FUTURE WORK •  Inverse analyses at higher temporal and spatial resolution.

•  Improved representation of error covariance matrix for concentration error

•  Improved characterization of prior fluxes.

EXTRA

SEASONAL VARIABLITY OF COS FLUXES Flux Distributions : Kettle et al [2002]

Northern Hemisphere Southern Hemisphere OCEAN

ANTH.

PLANT

SOILS FIRE

AGGREGATED FLUXES

• N Hemisphere variability driven by plant uptake and ocean fluxes

• S Hemisphere variability driven by ocean fluxes

TOTAL SURFACE COS FLUX

OCEAN

WHERE ARE THE MISSING FLUXES ?

Largest discrepancies in the tropics

Latitudinal Variation of Annual Mean Mixing Ratios Observations and ‘Increased Sink’ Model (Global Mean Subtracted)

60S 60N 30S 30N 0

Observations

Model

RANGE OF POSTERIOR FLUX ESTIMATES

SOIL

PLANT

ANTH

OCEAN Direct

OCEAN CS2

OCEAN DMS

ATM LOSS

“FIRES”

Gg S

/yr

Kettle 2002 Prior