inter-comparison of surface co fluxes estimated from ... · 0 inter-comparison of surface co 2...

13
Inter-comparison of surface CO 2 fluxes estimated from recent GOSAT X CO2 products using a single inverse modeling scheme Hiroshi Takagi 1 , Robert J. Andres 2 , Dmitry Belikov 3,1 , Andrey Bril 1 , Hartmut Boesch 4 , Andre Butz 5 , Sandrine Guerlet 6 , Sander Houweling 6 , Shamil Maksyutov 1 , Isamu Morino 1 , Tomohiro Oda 7 Christopher W. O'Dell 7 , Sergey Oshchepkov 1 , Robert Parker 4 , Makoto Saito 8 , Osamu Uchino 1 , Tatsuya Yokota 1 , Yukio Yoshida 1 , and Vinu Valsala 9 1 National Institute for Environmental Studies, Tsukuba, Japan 2 Oak Ridge National Laboratory, TN, USA 3 National Institute of Polar Research, Tokyo, Japan 4 University of Leicester, Leicester, UK 5 Karlsruhe Institute of Technology, Leopoldshafen, Germany 6 Netherland Institute for Space Research, Utrecht, Netherland 7 Colorado State University, CO, USA 8 Laboratoire des Sciences du Climat et l’Environment, Gifsur Yvette, France 9 Indian Institute for Tropical Meteorology, Pune, India

Upload: nguyenhanh

Post on 25-Aug-2019

220 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Inter-comparison of surface CO fluxes estimated from ... · 0 Inter-comparison of surface CO 2 fluxes estimated from recent GOSAT X CO2 products using a single inverse modeling scheme

0

Inter-comparison of surface CO2 fluxes estimated from recent GOSAT XCO2 products using a single inverse

modeling scheme Hiroshi Takagi1, Robert J. Andres2, Dmitry Belikov3,1, Andrey Bril1, Hartmut Boesch4, Andre Butz5, Sandrine Guerlet6, Sander Houweling6, Shamil Maksyutov1, Isamu Morino1, Tomohiro Oda7 Christopher W. O'Dell7, Sergey Oshchepkov1, Robert Parker4, Makoto Saito8, Osamu Uchino1, Tatsuya Yokota1, Yukio Yoshida1, and Vinu Valsala9

1 National Institute for Environmental Studies, Tsukuba, Japan 2 Oak Ridge National Laboratory, TN, USA 3 National Institute of Polar Research, Tokyo, Japan 4 University of Leicester, Leicester, UK 5 Karlsruhe Institute of Technology, Leopoldshafen, Germany 6 Netherland Institute for Space Research, Utrecht, Netherland 7 Colorado State University, CO, USA 8 Laboratoire des Sciences du Climat et l’Environment, Gifsur Yvette, France 9 Indian Institute for Tropical Meteorology, Pune, India

Page 2: Inter-comparison of surface CO fluxes estimated from ... · 0 Inter-comparison of surface CO 2 fluxes estimated from recent GOSAT X CO2 products using a single inverse modeling scheme

1

Comparison of GOSAT XCO2 Retrievals with TCCON

NIES L2 v02

ACOS B2.9 RemoTeC UoL FP

PPDF-D

NIES L2 v01

Recent publication by Oshchepkov et al., 2013 JGR Difference between GOSAT and TCCON - NIES v01 → NIES v02 Significant reduction in bias and spread Bias: -8 → -1 ppm SD: 4 → 2 ppm (Less than a few tenth of 1%)

CO2 flux estimation started

GOSAT Inversion Inter-comparison Campaign

Page 3: Inter-comparison of surface CO fluxes estimated from ... · 0 Inter-comparison of surface CO 2 fluxes estimated from recent GOSAT X CO2 products using a single inverse modeling scheme

2

GOSAT Inversion Inter-comparison

Inverse Modeling Schemes

Surface Obs.

NIES L2 v02

ACOS B2.9

PPDF-D

RemoTeC

UoL FP

σ1?

Flux Estimates Concentration Data

Aim: To assess the range of differences and added value of GOSAT XCO2 in flux estimation

A B C

Surface Obs.

NIES L2 v02

ACOS B2.9

PPDF-D

RemoTeC

UoL FP

σ2? B

Surface Obs.

σ3? A B C

Product A

① “Free-Style”

Inversion Comparison (S. Houweling)

② Single

scheme comparison

③ Fixed-data

comparison

Impact of differences in XCO2 retrievals

Impact of differences in inversion schemes

Page 4: Inter-comparison of surface CO fluxes estimated from ... · 0 Inter-comparison of surface CO 2 fluxes estimated from recent GOSAT X CO2 products using a single inverse modeling scheme

3

Location of Surface Observation Sites

NIES L2 v02

ACOS B2.9 RemoTeC

UoL FP PPDF-D

5˚×5˚ monthly mean XCO2 for May 2010 (cells with 3≥ soundings shown, large outliers filtered)

Regions of poor CO2 sampling → GOSAT retrievals nicely complement surface-based data

Page 5: Inter-comparison of surface CO fluxes estimated from ... · 0 Inter-comparison of surface CO 2 fluxes estimated from recent GOSAT X CO2 products using a single inverse modeling scheme

4

Std. deviation of coincident GOSAT XCO2 retrievals (NIES L2 v02, ACOS B2.9, PPDF-D, and UoL)

Jan. 2010 May 2010 Mar. 2010

Jul. 2009 Nov. 2009 Sep. 2009

XCO2 retrieval uncertainty

Coincident XCO2 NIES ACOS PPDF-D UoL Std. Dev. Max. Diff.

Lamont TCCON validation site 381.8 384.2 382.3 383.1 1.1 2.4

NE China 374.6 377.7 382.9 378.4 3.4 8.3

NW Colorado 378.4 381.1 384.9 389.0 4.6 10.6

(ppm)

○: TCCON validation site

Stddev of difference between GOSAT and TCCON reference ppmMAXNH 9.1_ =σ

Page 6: Inter-comparison of surface CO fluxes estimated from ... · 0 Inter-comparison of surface CO 2 fluxes estimated from recent GOSAT X CO2 products using a single inverse modeling scheme

5

Inverse modeling scheme : Fixed-lag Kalman Smoother (Bruhwiler et. al., 2005 ACP)

Observation window size: 3 month Number of source regions: 64 Tracer transport model: NIES08.1 (Belikov et al., 2011 GMD)

Prior flux uncertainty Land : Std. dev. of VISIT model monthly NEE about past 30-year mean Ocean : Std. dev. of assimilated ocean flux about 2001-2009 mean and Takahashi ocean flux climatology

GOSAT XCO2 Std. dev. (5°× 5°)

64 Source Regions

Model-observation mismatch errors GLOBALVIEW : RSD of observations about smooth curve GOSAT XCO2 : Std. dev. of XCO2 values found in a 5°× 5° grid in a month (~2 ppm depending on sounding locations) Minimum errors: GLOBALVIEW→ 0.3 ppm GOSAT XCO2 → 3.0 ppm (2 ppm systematic + 1 ppm XCO2 modeling unc.) (Belikov et al. 2013 AMT)

Prior Flux Uncertainty

Setup of flux estimation scheme (Summarized in Maksyutov et al. 2013 ACPD)

Prior flux data ODIAC/CDIAC monthly, GFED v3.0, Terrestrial ecosys. exchange by VISIT, pCO2-assimilated oceanic exchange

Page 7: Inter-comparison of surface CO fluxes estimated from ... · 0 Inter-comparison of surface CO 2 fluxes estimated from recent GOSAT X CO2 products using a single inverse modeling scheme

6

CO2 data used in inverse modeling (Summarized in Maksyutov et al. 2013 ACPD)

GOSAT XCO2 retrievals 14 months (June 2009 – July 2010)

Surface-based data (GLOBALVIEW-CO2: GV)

Extreme outliers removed via comparison with modeled climatology†

XCO2 gridded to 5°×5° cells and monthly-averaged (grids with N<3 were not considered here)

Curve-fitted values‡ at 220 observation points (monthly means)

→ Input to inverse modeling

‡ Masarie and Tans (1995) † Saito R., et al. (2011)

Combine

Page 8: Inter-comparison of surface CO fluxes estimated from ... · 0 Inter-comparison of surface CO 2 fluxes estimated from recent GOSAT X CO2 products using a single inverse modeling scheme

7

Impact of differences in XCO2 on monthly estimates - July 2009 (NH summer) - Mean of 5 flux estimates (fossil included) SD of 5 estimates

Uncertainty Reduction Rate (mean of 5) Posterior flux uncertainty (mean of 5)

UR(%)= (1-σGV+GOSAT/ σGV only)×100 15% → threshold for identifying clear GOSAT involvement

Amazonia: 5% (non-zero)

15%

Reg. 26

Land Ocean

Land Ocean

Page 9: Inter-comparison of surface CO fluxes estimated from ... · 0 Inter-comparison of surface CO 2 fluxes estimated from recent GOSAT X CO2 products using a single inverse modeling scheme

8

(2.5º ×2.5º resolution, 0.925 σ level )

July 2009

Impact of differences in XCO2 on monthly estimates NH summer

Reg. 26

Max. Diff. = 2.11gC m-2 day-1

(SD of 0.77 gC m-2 day-1)

Largest impact of XCO2 differences on flux estimate among 504 (64*12)estimates:

translates in simulated concentration field

Max. SD of 1.4 ppm in Reg. 26 in Aug. 2009

5 estimated fluxes (Reg. 26) SD of five independent simulations

Flux

(gC

m-2

day

-1)

Page 10: Inter-comparison of surface CO fluxes estimated from ... · 0 Inter-comparison of surface CO 2 fluxes estimated from recent GOSAT X CO2 products using a single inverse modeling scheme

9

Annual mean flux estimates (1 year: Jun. 2009 – May 2010) (unit: GtC yr-1 region-1)

GV only Net: 4.75 Bio: -2.09 Ocn: -2.00

NIES L2 v02 Net: 5.19 Bio: -1.19 Ocn: -2.47

ACOS B2.9 Net: 4.92 Bio: -1.53 Ocn: -2.39

SRON v1 Net: 4.72 Bio: -1.87 Ocn: -2.25

UOL v2 Net: 4.83 Bio: -1.73 Ocn: -2.28

PPDF-D Net: 4.81 Bio: -1.73 Ocn: -2.31

Land Ocean

→ Overall, reduction in land flux and gain in ocean flux

N. ext. tropics→sink Tropics →source (≈ TransCom results Gurney et al. 2002)

(GtC/reg./yr.)

Page 11: Inter-comparison of surface CO fluxes estimated from ... · 0 Inter-comparison of surface CO 2 fluxes estimated from recent GOSAT X CO2 products using a single inverse modeling scheme

10

Impact of differences in XCO2 on annual estimates

Mean departure of regional fluxes from GV-only estimates (GV+GOSAT minus GV-only estimate)

(unit: GtC yr-1 region-1)

Colored values: UR≥15% (5 mean)

Source enhancement: 15, 18, 32 →+1.1 GtC

Sink enhancement: 23, 24, 30 →-0.8 GtC

Ambiguous: 17, 29, 31 (opposite signs)

Page 12: Inter-comparison of surface CO fluxes estimated from ... · 0 Inter-comparison of surface CO 2 fluxes estimated from recent GOSAT X CO2 products using a single inverse modeling scheme

11

Sensitivity to improvement in XCO2 uncertainty

Minimum XCO2 Unc. = 3.0 ppm (same as previous page)

Minimum XCO2 Unc. = 2.0 ppm (1 ppm improvement)

Global total remained nearly the same

36% up

Page 13: Inter-comparison of surface CO fluxes estimated from ... · 0 Inter-comparison of surface CO 2 fluxes estimated from recent GOSAT X CO2 products using a single inverse modeling scheme

12

Summary

- Evaluated the impact of current XCO2 differences on flux estimation

- Annual timescale: - Terrestrial sources enhanced relative to GV-only estimate - Enhancements are sensitive to changes in minimum uncertainty value specified for XCO2 retrievals

- Monthly timescale:

~ 2 gC m-2 day-1 difference found (translates to ~ 1.5 ppm SD) Results to be compared with the “free-style” comparison