Transcript FluxVar_CarboOcean_Fi+

CarboOcean final meeting, Os, Norway, 5-9 October 2009
Session on
Simulating variability of air-sea CO2 fluxes
R. Matear (CSIRO, Hobart, Australia) :
Impact of Historical Climate Change on the Southern Ocean Carbon Cycle
J. Orr (LSCE, Gif-sur-Yvette, France)
Effects of forcing and resolution on simulated variability of air-sea CO2 fluxes
Funding: EU (GOSAC,
NOCES), NASA, DOE, Swiss
NSF, CSIRO
CarboOcean final meeting, Os, Norway, 5-9 October 2009
Effects of forcing and resolution on
simulated variability of air-sea CO2 fluxes
J. Orr (LSCE)
Contributors:
LSCE - J. Simeon, M. Gehlen, L. Bopp
LEGI (Grenoble) – C. Dufour, B. Barnier, J. LeSommer, J.-M. Molines
Funding: EU (GOSAC,
NOCES), NASA, DOE, Swiss
NSF, CSIRO
Outline
• Hints from North Atlantic (Raynaud et al., 2006)
• Hints from transient tracer simulaitons (Lachkar et al., 2007)
• Forcing
• Resolution
BATS: Sea-air CO2 flux anomalies (12-mo running mean)
Raynaud et al., 2006 (Ocean Science, 2, 43-60)
Why general underprediction?
 “Data” errors?
 Low horizontal resolution
(near west. boundary)
 Weak Forcing (atm.
reanalysis)
o from above
o affects lateral lags
NCEP underestimates real wind speed variability
North Atlantic

Interannual var. in wind
speed:
Raynaud et al. (2006, Ocean Science)
NCEP < (1/3) ERA40


NCEP wind speeds lower
than WOCE ship track
winds
NCEP atm. transport
variability only half that
observed (Waliser et al., 1999)
Smith et al. (2001, J. Climate)
HOT: Sea-air CO2 flux anomalies (12-mo running mean)
Raynaud et al., 2006 (Ocean Science, 2, 43-60)
LSCE testing importance of resolving eddies (global model):
Non-eddying 2°
Improvements:
Eddying ½°
• Mixed layer depth
CFC-11 burden (integrated vertically & zonally)
Zonal Integral of CFC-11 (Mmol degree-1)
• CFC-11 inventory
non-eddying
 non-eddying + GM
 eddying
 eddying + GM
Data*
de Boyer Montégut (2004, JGR)
*Lachkar et al (2007, Ocean Science)
Southern Ocean carbon sink – different stories
• Le Quéré et al. (2007): slower than expected
[coarse-resolution model, NCEP forcing]
• Matear and McNeil (2008): not slower
[another coarse-res. model, NCEP forcing]
• Sarmiento et al. (2009): slower
[4 coarse-res. models, NCEP forcing]
• Bopp (2009): [coarse-res. model]
– slower with NCEP forcing;
– not slower with ERA40
Changes in observed T across ACC reveal
fingerprint of anthropogenic climate change
52 447 Argo Profiles
Mean for neutral
densities 26.9 to 27.7
Boening et al. (2008, Nature Geoscience)
Observed T trend on density surfaces
• Bin by dynamic height
(0.09 levels)
• Average
• Remap onto mean bin
latitudes
Boening et al. (2008, Nature Geoscience)
Observed trends on depth surfaces
Temperature
Boening et al. (2008, Nature Geoscience)
Salinity
In forcing ocean GCM’s, there is much room
for artistry … and error
• Atmospheric
surface variables
• Bulk formulas
Large uncertainties
L. Brodeau, B. Barnier, T. Penduff, J.-M. Molines (2009) An ERA40based atmospheric forcing for simulations and reanalyses of the
global ocean circulation between 1958 to present, submitted.
Building adequate forcing requires huge effort
Example from high-res. ocean modeling consortium
(DRAKKAR DFS3, DFS4):
•
Strategy to blend
– corrected ERA40 surface atmospheric state fields (wind, air
temperature, humidity) with
– satellite products (ISCCP for radiation, CMAP for precipitation)
processed by Large & Yeager (2004) for CORE data set.
•
Procedure:
– Replace CORE’s NCEP with ERA40 (surface T, humidity, wind)
• Extend ERA40 until 2004 with ECWMF operational product
• Correct major ERA40 flaws (biases, inter-annual discontinuities)
– Adjust CORE shortwave radiation and precipitation products
– Quantify changes in forcing with a series of 1958-2004 interannual
2° (ORCA2) simulations
 assess impact of every forcing variable on the model solution.
DFS4.1 forcing in 2° model (NEMO/ORCA2)
S trend
Depth (m)
Density (σ)
T trend
LSCE simulations (J. Simeon et al.) with LEGI forcing (DRAKKAR DFS4.1)
NCEP-2 forcing in 2° model (NEMO/ORCA2)
Depth (m)
Density (σ)
T trend
LSCE simulations (J. Simeon et al.) with NCEP-2 forcing
S trend
Different forcing results in different air-sea fluxes
of natural CO2 during pre-satellite era
Southern Ocean (south of 45°S)
NEMO/ORCA2 model
Ocean efflux
Preliminary comparison of resolution (2°
vs. 0.5°) + many other differences:
Conclusions
• Different forcing fields – strengthen ties to evolving
developments of ocean circulation modeling community
• Different resolutions – ibid
• Different models – need more concerted evaluation,
comparison & strategy
• Different BGC components – minimum complexity to
properly simulate interannual variability & trends?
Conclusions:
• Arctic surface [CO32-]: high in summer, low in winter
(as elsewhere: Bering Sea, Norwegian Sea, Southern Ocean)
• High summertime [CO32-] from
• Biologically driven increase (from DIC drawdown)
overwhelms
• Physically driven decrease (freshening, i.e., dilution)
• Opposite trend in models with excessive fresh-water input
• Chukchi Sea surface water:
– observed seasonal amplitude (≥12 μmol kg-1)
(equivalent to past 30+ years of transient change)
– That annual cycle + Beringia 2005 summer data, yields
Wintertime Ωa < 1 already by 1990 (pCO2 atm = 354 ppmv),
i.e., 30 years sooner than summertime observations