validation of tes methane with hippo observations for use in adjoint inverse modeling kevin j. wecht...

Validation of TES Methane with HIPPO Observations

For Use in Adjoint Inverse Modeling

Kevin J. Wecht17 June 2010TES Science Team Meeting

Special thanks to:Annmarie Eldering Daniel Jacob

Steve WofsySusan Kulawik Lin Zhang

Eric KortGreg Osterman Christopher Pickett-HeapsVivienne Payne

Adjoint Inverse Modeling of Methane Sources

Adjoint inverse analysis

OPTIMIZATION OF SOURCES

GEOS-Chem CTM

ValidationHIPPO

Methane(Wofsy, Kort)

Apriori Sources

GEOS-Chem Adjoint

EDGAR v4, Kaplan, GFED 2, Yevich and Logan [2003]

HIPPO Methane provides:• Large number of profiles• Wide latitudinal coverage• Remote from sources

(reduces collocation error)

TES Methane (Worden, Kulawik)

HIPPOHIPPO I HIPPO II

January 9-30, 2009 October 22 - November 20, 2009

138 vertical profiles 147 vertical profiles

HIAPER Pole-to-Pole Observation Program

• Five missions:- Jan, Nov 2009- April 2010, June & Aug 2011

• > 700 vertical profiles by finish- 80% surface – 330 hPa- 20% surface – < 200 hPa

• Methane Instrument Properties- Frequency: 1 Hz- Accuracy/Precision: 1.0/0.6 ppb

Flight path

Vertical Profile

Flight path

Vertical Profile

Photos courtesy: S. Wofsy

HIPPO vs. TES by Latitude

RTVM

R [

ppb]

RTVM

R [

ppb]HIPPO I

Jan 9-30,2009

HIPPO IIOct 22 – Nov 20,

2009

TES observation operator is applied to all profiles unless explicitly stated.Positive bias and significant noise, but latitudinal gradient is captured.

TES

HIPPO w/ TES op.HIPPO 250-450 hPa

HIPPO vs. TES by Latitude

RTVM

R [

ppb]

RTVM

R [

ppb]HIPPO I

Jan 9-30,2009

HIPPO IIOct 22 – Nov 20,

2009

HIPPO I: bias = 73.6 ppb, residual error = 41.1 ppb n = 129HIPPO II: bias = 61.0 ppb, residual error = 44.8 ppb n = 147

TES over landTES over ocean

Mean Bias

Distribution of Residual ErrorHIPPO I

Errors are normally distributed, important for derivation of inversion cost function.

HIPPO II also shows Gaussian residual error.

Histograms ofHIPPO – TES difference

All Observations

Land ObservationsOcean Observations

mean = 73.6σ = 41.1n = 129

mean = 78.0σ = 37.3n = 90

mean = 63.7σ = 47.8n = 39

Mean and standard deviation in units [ppb]

Collocation Error

Mean Bias

Quantify observation, representation error. Error and bias differ for HIPPO I and II.

HIP

PO I

HIP

PO II

Residual ErrorTES bias and residual error vs. size of coincidence window

Observation error

[ppb

][p

pb]

Distance [km]125 250 375 500

Distance [km] Distance [km]

40

120

80

0

40

120

80

0

# Observations

24 hour12 hour

Model ComparisonTES and GEOS-Chem along HIPPO I flight track

TES and GEOS-Chem averaged over the Pacific during the HIPPO I period

HIPPO IJan 9-30,

2009

RTVM

R [

ppb]

RTVM

R [

ppb]

Consistency with HIPPO

GEOS-Chem w/ TES op.TES

HIPPO w/ TES op.

Model Comparison

RTVM

R [

ppb]

RTVM

R [

ppb]

Consistency with HIPPO

HIPPO IIOct 22 – Nov 20,

2009

TES and GEOS-Chem along HIPPO II flight track

TES and GEOS-Chem averaged over the Pacific during the HIPPO II period

GEOS-Chem w/ TES op.TES

HIPPO w/ TES op.

Model Comparison – NOAA GMD2008 Annual Average

• GEOS-Chem provides good simulation in the annual average• Seasonality to inter-hem. gradient• Missing northern hemisphere sources? Natural gas production?

[ppb]

GEOS-Chem and GMD surface methane

GE

OS

-Che

m [p

pb]

NOAA GMD [ppb]

[ppb

]Latitude

GEOS-Chem is a good platform inter-comparison of data sets such as TES methane.

[ppb

]

GEOS-ChemNOAA GMD

Model Comparison – TES2009 Annual Average

• TES captures variability in GEOS-Chem, with large residual error.• Interpretation of latitudinal profile complicated by latitudinal structure of DOFS • Model overestimate around equator.

Diff

ere

nce

[p

pb

]R

TV

MR

[p

pb

]

GE

OS

-Che

m w

/ TE

S o

pera

tor

RT

VM

R

GEOS-ChemGEOS-Chem w/ TES op.TES

Model ComparisonTES RTVMR w/

mean TES – GEOS-Chem difference subtracted

GEOS-Chem RTVMR

[ppb]

Difference: GEOS-Chem – TES

[ppb]

TES a priori RTVMR

Quantitative source attribution of differences requires inverse modeling.

GEOS-ChemGEOS-Chem w/ TES op.TES

Adjoint Inverse Modeling of Methane Sources

Adjoint inverse analysis

OPTIMIZATION OF SOURCES

GEOS-Chem CTM

ValidationHIPPO

Methane(Wofsy, Kort)

Apriori Sources

GEOS-Chem Adjoint

EDGAR v4, Kaplan, GFED 2, Yevich and Logan [2003]

TES Methane (Kulawik, Osterman)

4D-Var assimilation of TES data:• Optimize emissions on

– 2x2.5 horizontal grid– monthly temporal res.

• Evaluate results with HIPPO, NOAA GMD

• Incorporation of additionalsatellite products

Thank You!

• TES captures latitudinal gradient in HIPPO data

• TES is biased high but residual instrument error is within self-reported range– 73.6 ± 41.1 ppb during HIPPO I

– 61.0 ± 44.8 ppb during HIPPO II

• Pending prognosis of TES instrument, will validate again with HIPPO IV & V

• Enabling Inverse Modeling:– Time period: TES provides useful information through the end of HIPPO II.

– Quantification of bias, observation error, and representation error.

– Error normally distributed.

– Robust latitudinal gradient with greater coverage than surface stations

• HIPPO and NOAA GMD will be used to evaluate inversion results.

• Future incorporation of additional satellite products.