T3 evolution

Level 1:

Annual mean control inversion (Nature paper)

Annual mean flux sensitivity and model-to-model (Tellus)

Annual mean data sensitivity (Tellus)

Level 2:

Cyclostationary control (in preparation – GBC)

Cyclo sensitivity, model-to-model?

IAV in progress (Baker, Law, Gurney)

Level 3 in progress (Houweling)

Offshoots

Isotope

networks

Satellite

etc

Kevin Gurney, Scott Denning, Rachel Law*, Peter Rayner*, Bernard Pak†

Colorado State University, *CSIRO Australia, †UC Irvine

Philippe Bousquet, Lori Bruhwiler, Yu-Han Chen, Philippe Ciais, Songmiao Fan, Inez Y. Fung, Emanuel Gloor, Martin Heimann, Kaz Higuchi, Jasmin John,

Shamil Maksyutov, Philippe Peylin, Michael Prather, Jorge Sarmiento, Shoichi Taguchi, Taro Takahashi, Takashi Maki, Ken Yuen

May 2003

TransCom control results for the estimation of seasonal carbon sources and sinks

T3 regions

Inversion set-up

Flux Priors

Land: CASA NEP (presub) +

Ocean: Takahashi (presub)

composite inventory from Level 1 (distributed over particular months)

Prior Flux Uncertainties

Land: Level 1 (annual GSNF) + +

Ocean: 2.0 * Level 1

30% of CASA |resp|

30% of CASA |NPP|

Set-up continued

CO2 Observations

• Monthly mean for 1992-1996 period (filled data) from GV2000

• 75 stations (dropped DAA) meeting 70% criteria

“Data Uncertainties”

Setting the floor:

If then

1992 19961979 1996

1979 1996

3.6

am m

a

RSDRSD

RSDP

122

1 0.25m

m

appm

A

min

0.25m

A

Presub results: surface maps

Strong northern seasonality for MATCH variants, NIES, NIRE,

and TM3

• MATCH:NCEP and NIES show widespread Winter max

Weak northern seasonality for CSU, JMA, UCI, and TM2

GCTM has average Winter max but strong Summer minimum

Presub results: at stations

Aggregated model mean results

Uncertainties

Greatest uncertainty reduction in the northern extratropics and Southern Ocean

Model spread is greatest in the tropics but varies in the north – greatest at height of growing season

Model spread goes up where there is data and vice-versa

Fluxes

Northern Land: less emission during March, April, and September and greater uptake in July relative to prior

Oceans: greater seasonality……may be ‘contamination’ from land

Disaggregated model mean results

Land

Boreal NA: deviations in April, June, and August….uptake occurring later in the growing season

Europe: greater net uptake - June through September

Boreal Asia: Reduction in early Spring emissions. Less net uptake in June, more in July

Temperate NA: lessened seasonal amplitude. Significant deviations in Spring and late Fall

Temperate Asia: Large phase disagreement. CASA error or real advance?

Disaggregated continued

Ocean

Heightened seasonality in northern and tropical ocean regions

North Pacific and North Atlantic: out of phase with prior

Southern Ocean winter uptake lessened compared to prior

Tropical Oceans: net uptake as an annual mean and large negative fluxes centered on July to September months

Sensitivity to prior flux uncertainty

Observationally constrained regions show little change

Boreal NA, Temperate NA, Boreal Asia, Temperate Asia, Europe, Northern Ocean, Southern Ocean, South Indian

Tropics (land mainly) are generally sensitive to the prior flux uncertainty

Lower uncertainty (4x lower) on Northern Ocean, N Atlantic, and N. Pacific

Northern Land:

Combined flux difference (Gt C/year)

Predicted CO2

Difference between model mean predicted CO2 and observed

(by station, by month)

X2 per station

Two or more months with a 2 exceeding 4:

CRI, GMI, IZO, UTA, CARR (5000m), PRS, HUN

Driven by Feb value (14.7)

which results from small

Station sensitivity

Most pronounced changes in tropical regions – primarily the removal of GMI and CRI

Some stations in the southern high latitudes (SPO, SYO, MAA, PSA, HBA, Bass Strait) had 2 of < 0.25 implying uncertainties that may be 2x too loose

Tightening uncertainty on these 6 stations (factor of 2) imparts small changes in the South Indian Ocean

Comparison to annual mean inversion

-1.0

-0.8

-0.6

-0.4

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

Annual Inversion

Cyclostionary Inversion

-3.5

-2.5

-1.5

-0.5

0.5

1.5

NorthernLand

TropicalLand

SouthernLand

NorthernOcean

TropicalOcean

SouthernOcean

Total Land TotalOcean

GlobalTotal

Posterior Northern Flux Amplitude

R2 = 0.75

Posterior Northern Flux

R2 = 0.6

Posterior Northern Flux continued

R2 = 0.82

When is the net uptake?

Wild speculation on Biogeochemistry

Greater net uptake in Boreal Asia and Europe at the height of the growing season……..

greater NPP or lessened respiration?

The use of air temperature versus soil to drive respiration in CASA? Would explain some of the Winter flux differences

NPP could be wrong in CASA

Phase difference in Temperate Asia

Aggregation error (big region spanning large latitudinal gradient)

Real Springward shift in growing season?

Delay in the onset of the growing season in Boreal regions

Earlier Spring thaw causing respiration to precede photosynthesis

Could heightened ocean seasonality be real?

Conclusions Amplitude of N background flux response is inversely related to N

posterior flux amplitude. This is directly related to N land mean uptake and strength of rectifier

Land posterior flux: deviations from the prior occur in Summer and Winter

Europe: greater uptake in growing season

Temperate Asia: ~2 month uptake timing discrepency

Spring timing in Coreal land

Ocean posterior flux

Heightened seasonality in Northern oceans

Southern Ocean discrepency with prior is primarily during Austral Winter

Northern extratropics are insensitive to prior flux information

Northern extratropics are insensitive to high 2 per station instances

Good agreement with annual mean inversion

Phase and amplitude differences from CASA may be biogeochemically interpretable

PTP versus rectifier