THERE’S A RECTIFIER IN MY CLOSET: Vertical CO2 Transport and Latitudinal Flux Partitioning

Britton Stephens, National Center for Atmospheric ResearchTransCom Meeting, Purdue 2007

[illustrations from There’s a Nightmare in my Closet by Mercer Mayer]

TransCom3 Modelers:Kevin Robert Gurney, Rachel M. Law, Scott Denning, Peter J. Rayner, David Baker, Philippe Bousquet, Lori Bruhwiler, Yu-Han Chen, Philippe Ciais, Inez Y. Fung, Martin Heimann, Jasmin John, Takashi Maki, Shamil Maksyutov, Philippe Peylin, Michael Prather, Bernard C. Pak, Shoichi Taguchi

Aircraft Data Providers:Pieter P. Tans, Colm Sweeney, Philippe Ciais, Michel Ramonet, Takakiyo Nakazawa, Shuji Aoki, Toshinobu Machida, Gen Inoue, Nikolay Vinnichenko, Jon Lloyd, Armin Jordan, Martin Heimann, Olga Shibistova, Ray L. Langenfelds, L. Paul Steele, Roger J. Francey

Additional Modeling:Wouter Peters, Philippe Ciais, Philippe Bousquet, Lori Bruhwiler

[figure courtesy of Scott Denning]

Seasonal vertical mixing

Annual mean accumulation near surface and depletion aloft

[Denning et al., Nature, 1995]

Observed

Transcom3 neutral biosphere flux response

349.5

350.0

350.5

351.0

351.5

352.0

352.5

353.0

353.5

354.0

-90 -70 -50 -30 -10 10 30 50 70 90

CSU.gurney

GISS.fung

GISS.prather

GISS.prather2

GISS.prather3

J MA-CDTM.maki

MATCH.bruhwiler

MATCH.chen

MATCH.law

NIES.maksyutov

NIRE.taguchi

RPN.yuen

SKYHI.fan

TM2.lsce

TM3.heimann

GCTM.baker

Latitude

ppm

Gurney et al, Nature, 2002

TransCom3 model results show a large transfer of carbon from tropical to northern land regions.

Level 1 (annual mean)Level 2 (seasonal)

Gurney et al, GBC, 2004

Bottom-up estimates have generally failed to find large uptake in northern ecosystems and large net sources in the tropics

Model Model NameNorthern

Total Flux (1)Tropical

Total Flux (1)Northern

Land Flux (1)Tropical

Land Flux (1)

1 CSU -4.4 (0.2) 3.7 (0.6) -3.6 (0.3) 3.3 (0.7)

2 GCTM -3.4 (0.2) 2.3 (0.7) -2.0 (0.3) 2.7 (0.8)

3 UCB -4.4 (0.3) 3.7 (0.6) -3.1 (0.3) 4.0 (0.7)

4 UCI -2.6 (0.3) 0.5 (0.7) -1.5 (0.3) -0.1 (0.8)

5 JMA -1.4 (0.3) -0.5 (0.8) -0.9 (0.4) -0.5 (0.9)

6 MATCH.CCM3 -3.0 (0.2) 2.2 (0.6) -2.1 (0.3) 2.3 (0.7)

7 MATCH.NCEP -4.0 (0.2) 3.2 (0.5) -4.0 (0.3) 3.4 (0.7)

8 MATCH.MACCM2 -3.7 (0.3) 3.1 (0.8) -3.0 (0.3) 2.5 (0.9)

9 NIES -4.0 (0.3) 2.2 (0.6) -3.5 (0.3) 2.7 (0.8)

A NIRE -4.5 (0.3) 1.6 (0.7) -2.8 (0.3) 1.2 (0.8)

B TM2 -1.6 (0.3) -1.4 (0.7) -0.5 (0.3) -1.0 (0.8)

C TM3 -2.4 (0.2) 1.4 (0.6) -2.2 (0.3) 1.0 (0.8)

TransCom 3 Level 2 annual-mean model fluxes (PgCyr-1)

Study N. Total T. Total N. Land T. Land

Jacobson et al., 2006 ('92-'96) -3.9 5.0 -2.9 4.2

Baker et al., 2006 ('91-'00) -3.7 2.7 -2.6 1.9

Gurney et al., 2004 ('92-'96) -3.3 1.8 -2.4 1.8

CarbonTracker, 2007 ('01-'05) -2.8 1.1 -1.8 0.1

Rödenbeck et al., 2003 ('92-'96) -2.3 -0.1 -0.7 -1.0

Rödenbeck et al., 2003 ('96-'99) -2.1 0.3 -0.4 -0.8

Comparison to other studies

TransCom3 predicted rectifier explains most of the variability in estimated fluxes

Impact on predicted fluxes

349.5

350.0

350.5

351.0

351.5

352.0

352.5

353.0

353.5

354.0

-90 -70 -50 -30 -10 10 30 50 70 90

CSU.gurney

GISS.fung

GISS.prather

GISS.prather2

GISS.prather3

J MA-CDTM.maki

MATCH.bruhwiler

MATCH.chen

MATCH.law

NIES.maksyutov

NIRE.taguchi

RPN.yuen

SKYHI.fan

TM2.lsce

TM3.heimann

GCTM.baker

Model Model Name

1 CSU

2 GCTM

3 UCB

4 UCI

5 JMA

6 MATCH.CCM3

7 MATCH.NCEP

8 MATCH.MACCM2

9 NIES

A NIRE

B TM2

C TM3

ppm

pressure

N S N S N S N S

Transcom3 neutral biosphere flux response

Northern Hemisphere sites include Briggsdale, Colorado, USA (CAR); Estevan Point, British Columbia, Canada (ESP); Molokai Island, Hawaii, USA (HAA); Harvard Forest, Massachusetts, USA (HFM); Park Falls, Wisconsin, USA (LEF); Poker Flat, Alaska, USA (PFA); Orleans, France (ORL); Sendai/Fukuoka, Japan (SEN); Surgut, Russia (SUR); and Zotino, Russia (ZOT). Southern Hemisphere sites include Rarotonga, Cook Islands (RTA) and Bass Strait/Cape Grim, Australia (AIA).

Airborne flask sampling locations

Collaborating Institutions: USA: NOAA GMD, CSU, France: LSCE, Japan: Tohoku Univ., NIES, Nagoya Univ., Russia: CAO, SIF, England: Univ. of Leeds, Germany: MPIB, Australia: CSIRO MAR

Airborne flask sampling data

Altitude-time CO2 contour plots for all sampling locations

20

-15

10

-10

10

-10

0

-5

Model-predicted NH Average CO2 Contour Plots

Observed NH Average CO2 Contour Plot

Vertical CO2 profiles for different seasonal intervals

Observed and predicted NH average profiles

• 3 models that most closely reproduce the observed annual-mean vertical CO2 gradients (4, 5, and C):

Northern Land = -1.5 ± 0.6 PgCyr-1

Tropical Land = +0.1 ± 0.8 PgCyr-1

• All model average:

Northern Land = -2.4 ± 1.1 PgCyr-1

Tropical Land = +1.8 ± 1.7 PgCyr-1

Estimated fluxes versus predicted 1 km – 4 km gradients

Observed value

Model Model Name

1 CSU

2 GCTM

3 UCB

4 UCI

5 JMA

6 MATCH.CCM3

7 MATCH.NCEP

8 MATCH.MACCM2

9 NIES

A NIRE

B TM2

C TM3

• Interlaboratory calibration offsets and measurement errors

• Diurnal biases

• Interannual variations and long-term trends

• Flight-day weather bias

• Spatial and Temporal Representativeness

Observational and modeling biases evaluated:

All were found to be small or in the wrong direction to explain the observed annual-mean discrepancies

[Schulz et al., Environ. Sci. Technol. 2004, 38, 3683-3688]

WLEF Diurnal Cycle Observations

Estimated fluxes versus predicted 1 km – 4 km gradients for different seasonal intervals

Observed values

Model Model Name

1 CSU

2 GCTM

3 UCB

4 UCI

5 JMA

6 MATCH.CCM3

7 MATCH.NCEP

8 MATCH.MACCM2

9 NIES

A NIRE

B TM2

C TM3

Should annual-mean or seasonal gradients be used to evaluate model fluxes?

• Annual-mean fluxes are of most interest because they are relevant to annual ecosystem budgeting, to policy makers, and to projections of future greenhouse gas levels.

• No model does well at all times of year, but do not want to reject all models.

• Errors in seasonal timing of fluxes make selection of seasonal criteria problematic.

• Seasonal (rectifier) effects are inherently cumulative, such that a model with large seasonal errors that offset will do better in annual-mean that one with small seasonal errors that compound.

HIPPO ’08-’11 (PIs: Harvard, NCAR, Scripps, and NOAA): A global and seasonal survey of CO2, O2, CH4, CO, N2O, H2, SF6, COS, CFCs, HCFCs, O3, H2O, and hydrocarbons

HIAPER Pole-to-Pole Observations of Atmospheric Tracers

Fossil fuel CO2 gradients over the PacificUCI UCIs

JMA MATCH.CCM3

ppm

pres

sure

pres

sure

S N S N S N

N S

N S

N S

N S

• Models with large tropical sources and large northern uptake are inconsistent with observed annual-mean vertical gradients.

• A global budget with less tropical-to-north carbon transfer is more consistent with bottom-up estimates and does not conflict with independent global 13C and O2 constraints.

• Simply adding airborne data into the inversions will not necessarily lead to more accurate flux estimates

• Models’ seasonal vertical mixing must be improved to produce flux estimates with high confidence

• There is value in leaving some data out of the inversions to look for systematic biases

Conclusions:

And of course, watch out for the next monster. . . .

Representativeness

Northern Hemisphere average CO2

contour plot from observations

Aircraft Flask Sampling Locations and Data Summary.

Site Name Location Lab Time PeriodMean Local Time (± 1)

Briggsdale, Colorado, USA (CAR) 40.9 N 104.8 W 1740 m NOAA GMD Nov. 1992 - Jun. 2006 9:43 (1:57)

Estevan Point, British Columbia, Canada (ESP)

49.58 N 126.37 W 39 m NOAA GMD Nov. 2002 - Jun. 2006 11:22 (2:50)

Molokai Island, Hawaii, USA (HAA) 21.23 N 158.95 W 0 m NOAA GMD May 1999 - Jan. 2006 10:31 (0:56)

Harvard Forest, Massachusetts, USA (HFM)

42.54 N 72.17 W 315 m NOAA GMD Nov. 1999 - Jun. 2006 13:20 (1:59)

Park Falls, Wisconsin, USA (LEF) 45.93 N 90.27 W 472 m NOAA GMD Apr. 1998 - Jun. 2006 13:17 (2:58)

Poker Flat, Alaska, USA (PFA) 65.07 N 147.29 W 210 m NOAA GMD Jun. 1999 - May 2006 12:25 (2:12)

Rarotonga, Cook Islands (RTA) 21.25 S 159.83 W 3 m NOAA GMD Apr. 2000 – Dec. 2005 10:21 (0:31)

Bass Strait / Cape Grim, Australia (AIA)

40.53 S 144.30 E 0 m CSIRO MAR Jun. 1991 - Sep. 2000 13:11 (2:02)

Orleans, France (ORL) 47.80 N 2.50 E 150 m LSCE Apr. 1996 - Jun. 2006 12:51 (2:04)

Sendai, Japan (SEN)1 38.00 N 141.00 E Tohoku Univ. Jan. 1979 - Mar. 2003 NA

Surgut, Russia (SUR) 61.00 N 73.00 E 44 m NIES, CAO, and TU

Jul. 1993 - Jan. 2006 ~ 13:30

Zotino, Russia (ZOT) 60.75 N 89.38 E 191 m MPIB and CSIRO

Jul. 1998 - Nov. 2005 12:10 (2:36)

1Observations over 4 km are from flights between Sendai and Fukuoka, Japan.

annual average summer winter spring fall

model bias model bias model bias model bias model bias

4 0.08 1 0.08 8 0.01 B 0.24 8 -0.03

5 0.10 2 0.44 2 0.24 1 0.59 6 0.09

C 0.14 A -0.48 A 0.38 C 0.66 A -0.10

A 0.25 9 0.61 C -0.44 5 0.69 7 0.10

6 0.31 6 -0.68 7 0.49 4 0.75 9 0.11

B -0.36 C 0.74 6 0.49 2 0.82 2 0.11

2 0.40 4 0.84 9 0.65 9 0.97 C -0.38

1 0.46 7 0.86 5 -0.66 3 0.99 1 0.47

8 0.49 B 0.92 4 -0.69 7 1.07 4 -0.59

9 0.58 8 0.93 1 0.72 8 1.07 5 -0.68

7 0.63 5 1.05 3 1.17 A 1.21 B -1.32

3 1.30 3 1.41 B -1.29 6 1.32 3 1.65

Seasonal gradients ranked by absolute value

[Schulz et al., Environ. Sci. Technol. 2004, 38, 3683-3688]

Diurnal CO2 profile measured over WLEF in 2001 by Michael Jensen using a powered parachute and bag sampling

Diurnal bias analysis

TM5 2000-2005 Predicted Annual-Mean 1 – 4 km Gradients.

SiteAll

DaysSample

DaysSample

Days + 5Sample Days - 5

Sample - All

S+5 - All

S-5 - All

CAR 2.2 1.3 2.3 2.2 -0.9 0.1 0.1

ESP 0.4 0.1 0.2 0.4 -0.3 -0.1 0.1

HAA 0.3 0.4 0.4 0.3 0.1 0.1 0

HFM 1.8 2.2 1.6 1.5 0.4 -0.1 -0.3

LEF 0.8 1 0.5 1.1 0.2 -0.3 0.3

PFA 0.6 0.7 0.5 0.6 0.1 -0.1 0.1

RTA -0.3 -0.4 -0.3 -0.2 -0.1 0 0.1

ORL 0.3 -0.8 0.4 0.7 -1.2 0 0.3

ZOT 1 1.4 1.2 0.7 0.4 0.2 -0.3

N.H. Mean 0.9 0.8 0.9 0.9 -0.2 0.0 0.0

Interannual trend analysis

Site

1 – 4 km CO Gradient

(ppb)

CO Gradient /

10 ppb/ppm

CO Gradient /

20 ppb/ppm

Fossil-fuel CO2 Gradient

(ppm)*

CAR 9.4 0.9 0.5 0.2

ESP 4.0 0.4 0.2 0.3

HAA 8.6 0.9 0.4 0.5

HFM 39.6 4.0 2.0 1.8

LEF 24.2 2.4 1.2 1.2

PFA 6.8 0.7 0.3 0.2

RTA -5.6 -0.6 -0.3 -0.2*Difference between 1 and 4 km averaged across all 12 T3L2 models.

Annual-Mean Observed CO and Modeled Fossil-Fuel CO2 Gradients.