Atmosphere-based “top-down” emission estimates of HFC-134a and HCFC-22 from the U.S. over multiple years L. Hu1,2, S.A. Montzka1, J. Miller1,3, A.E. Andrews 1, B.R. Miller1,3, S. Lehman4, K. Thoning1, C. Sweeney1,3, H. Chen5,3, K. Masarie1, L. Bruhwiler1, M.L. Fischer6, S.C. Biraud6, M.S. Torn6, E. Saikawa7, S. Miller8, M. Mountain9, T. Nehrkorn9, J. Eluszkiewicz9, J.W. Elkins1, and P. Tans1

1 NOAA/ESRL/GMD, Boulder, Colorado, USA 2 NAS/NRC, Washington D.C., USA. 3 CIRES, U of Colorado, Boulder, Colorado, USA 4 INSTARR, U of Colorado, Boulder, Colorado, USA 5 CIO, University of Groningen, Groningen, The Netherlands 6 Lawrence Berkeley National Laboratory, California, USA 7 Environmental Sciences, Rollins School of Public Health, Emory University, Atlanta, GA, USA 8 Harvard University, Cambridge, MA, USA 9 AER, Lexington, MA, USA

Thanks to: A. Jacobson and S. Basu for inversion discussion, D. Godwin for providing EPA estimates and C. Siso, B. Hall and others involved with sampling, analysis, logistics, and program management. Funding in part from: NOAA’s Climate Program Office and its Atmospheric Chemistry, Carbon Cycle, and Climate Program, National Research Council.

Goal: • To derive reliable “atmosphere-based” national emission estimates of ozone-depleting substances (ODSs) and greenhouse gases (GHGs) • To assess inventory-based “bottom-up” estimates

US Emission magnitudes? Seasonality & Inter-annual variability? Emission trends?

Key Questions:

HFC-134a: • A potent GHG • Mainly used in mobile air conditioning to replace CFC-12

HCFC-22: • An ODS and potent GHG • Mainly used in commercial and residential air conditioning • US production and consumption currently declining

Today, regional inverse modeling of HFC-134a and HCFC-22.

Study Period: 2008 – 2012 Network change: • Fewer sites in 2008 • Reduced sampling frequency in 2012

North American Halocarbon Flask Sampling Network

Model Domain

Surface sites (daily flasks)

Surface sites (weekly flasks)

Biweekly aircraft profiling sites

x x

Aircraft campaigns

Total : ~20000 measurements Data in inversion : ~10000 independent observations Surface and aircraft (< 1 km a.g.l.) data collected at day-time

Daily-flask site (WGC)

Biweekly aircraft profiling site (CMA)

Weekly-flask site (HFM)

enhance-ments

KUM BRW

Inversion Method

Observed enhancement

Footprint (sensitivity of observed

enhancement to upstream fluxes )

Flux = x + Model-data mismatch

errors

Solve for: monthly 1o × 1o scaling factors for a prior emission field using a Bayesian inversion

HYSPLIT- NAM12 (2008 – 2012) STILT-WRF (2010)

Synthetic-data inversion Objective:

Design:

“True” emission x Footprints Pseudo observations

(pseudo enhancements)

Prior emissions + Pseudo

observations Footprints

+

Posterior emissions

To test the credibility of our inversion system to derive national fluxes, given our sampling network

A forward calculation:

An inverse calculation:

Actual sampling times and locations as real measurements

Posterior emissions = “True”

emissions ?

Gaussian Noise

+ Different emission magnitudes and distributions than “true” emission

“Truth” & Priors “Truth” & Priors & Posteriors

0

20

40

60

80

100

2008 2009 2010 2011 2012 2013

HCFC

-22(

Gg/

y)

0

20

40

60

80

100

2008 2009 2010 2011 2012 2013

HCFC

-22(

Gg/

y)

0

20

40

60

80

100

2008 2009 2010 2011 2012 2013

0

20

40

60

80

100

2008 2009 2010 2011 2012 2013

HFC-

134A

(Gg/

y)

0

20

40

60

80

100

2008 2009 2010 2011 2012 2013

HFC-

134A

(Gg/

y)

0

20

40

60

80

100

2008 2009 2010 2011 2012 2013

HFC-

134A

(Gg/

y)

0

20

40

60

80

100

2008 2009 2010 2011 2012 2013

HFC-

134A

(Gg/

y)

Real-Data Inversion: HFC-134a and HCFC-22 (Multiple Priors)

Prior = EDGARv4.2

Prior = EDGARv4.1

Prior = Population-Based

Prior = Population-Based Prior = Saikawa et al (2012)

HFC-134A

HCFC-22

Real-Data Inversion: HFC-134a and HCFC-22 (Multiple Priors & Backgrounds)

bkg = 10th prtile, HYSPLIT-NAM12

bkg = “Curtain” + Air back-trajectory HYSPLIT-NAM12

* Multiple a-seasonal priors

* Inversion results:

bkg = 10th prtile, STILT-WRF

Real-Data Inversion: HFC-134a and HCFC-22 (Multiple Priors & Backgrounds & Transports)

Evaluating derived fluxes HFC-134a

Independent Data

0

20

40

60

80

100

2000 2002 2004 2006 2008 2010 2012

HC

FC-2

2 (G

g/y)

0

20

40

60

80

100

2000 2002 2004 2006 2008 2010 2012

HC

FC-2

2 (G

g/y)

400

450

500

550

600

650

700

0

20

40

60

80

100

2000 2002 2004 2006 2008 2010 2012

HC

FC-2

2 (G

g/y)

0

20

40

60

80

100

2000 2002 2004 2006 2008 2010 2012

HFC

-134

A (G

g/y)

0

20

40

60

80

100

2000 2002 2004 2006 2008 2010 2012

HFC

-134

A (G

g/y)

400

450

500

550

600

650

700

0

20

40

60

80

100

2000 2002 2004 2006 2008 2010 2012

HFC

-134

A (G

g/y)

Comparison with other national estimates

“Bottom-up”: U.S. EPA ( ) EDGARv4.2 ( )

“Top-down”:

This study ( )

Millet 2009 ( ) Stohl 2009 ( x )

Saikawa 2012 ( ) Miller 2012 ( ) Montzka 2013 ( )

HFC-134a

HCFC-22

Barletta 2011 ( )

Tran

spor

tatio

n (T

g C

yr-1

) R

esid

entia

l and

C

omm

erci

al (T

g C

yr-1

)

Fossil Fuel CO2 Emission (US EIA)

Conclusions • Synthetic-data inversion: Given our sampling network, derived fluxes

using our inversion system are fairly insensitive to priors on a national scale.

• Real-data inversion: - Derived national emissions of HFC-134a and HCFC-22 are fairly

insensitive to priors, backgrounds and transports (within 1sd = ±20%).

- Seasonally varying emissions: winter emissions are 20 – 50 % lower than summer emissions for both gases.

- Comparing to US EPA national emission estimates: - HFC-134a: comparable. - HCFC-22: ~10 – 50% lower; a relatively more rapid decline.

• Future work: apply to other gases (e.g. other HFCs, HCFCs, N2O, CH4)

Evaluating derived fluxes

AMT AMT

HFC-134a

Prior (RMSE = 11.8)

Posterior (RMSE = 3.5)

Simulation (post)

1:1 line Best Fit 1:1 line

Best Fit

Simulation (prior)

Simulation (post)