comparing parameterizations of gas transfer velocity and their

Susanne Ufermann, Comparing parameterizations of gas transfer velocity, Liège, 2-6 May 200537th International Liège Colloquium on Ocean Dynamics, Gas Transfer at Water Surfaces

Comparing parameterizations of gas transfer velocity

and their effect on the global marine CO2 budget

Susanne Ufermannand David K. Woolf

Centre for the observation of Air-Sea Interaction & fluXes (CASIX)National Oceanography Centre

Southampton, UK

email: [email protected]: +44 (0)2380 592726


Acknowledgements

Data sources:• ECMWF ERA-40

http://www.ecmwf.int/products/data

• KNMI (Sofia Caires, ERA-40 u* and Hs)http://www.knmi/nl/onderzk/oceano/publ/caires/NPcor.

• REMSS (QuikSCAT, AMSR)http://www.remss.com

• NOAA's Climate Diagnostics Centre (Reynolds OI)http://www.cdc.noaa.gov/cdc/data.reynolds_sst.html

• Ifremer (ERS)http://www.ifremer.fr/cersat/english

• Taro Takahashi et al. (∆pCO2)http://islscp2.sesda.com

• Scott Doney & Rik Wanninkhof, Christine Gommenginger


Factors affecting air-sea

gas transfer

From http://www.uea.ac.uk/env/solas -Wade McGillis

Wind

Transfer Velocity

k

Air-Sea Gas FluxF = k∆C


Overview

• Parameterizations of gas transfer velocity– Traditional u10-based parameterizations– “New” parameterizations

• Glover et al. (2002)• Woolf (2005)• ALT1 & ALT2

• Satellite data– Data characteristics– Corrections

• Global & regional flux results• Conclusions


Parameterizations of gas transfer velocity

u10-based parameterizationsLiss & Merlivat, '86 •

3 parts: k ~ u

Wanninkhof, '92 •k ~ u2

Wanninkhof & McGillis, '99 •k ~ u3

Nightingale et al., '00 •k ~ u2


Parameterizations of gas transfer velocity

“New” parameterizationsGlover et al., ‘02

k ~ 1/σKu - 1/σC

Woolf, ‘05:non-breaking

k0 ~ u*breaking

k1 ~ u*Hstotal

k = k0 + k1


Mean square slope algorithm

Total mean square slope yields

k = 1.49·10-3 · 0.38/σKu + 1·10-6

from Bock et al., JGR 1999.


Altimeter-based algorithms• Based on Woolf (2005)

k = kJ + kb

non-breakingcontribution

whitecappingcontribution

• kJ = a/σKu + b from Bock et al. (1999)

• kb = u* Hs / vw

• ALT1: kJ : kb = 3:1 dominant direct transfer

• ALT2: kJ : kb = 1:3 dominant whitecapping


Satellite data

1991-1999monthly

1995 monthly

1x1ERS u10

1982-2002monthly

1982-2002monthly

2003-2004daily, 3-day,monthly

2003-2004daily, 3-day,monthly

2003-2004daily, 3-day, monthly

2000-2004daily,monthly

2000-2004daily,monthly

1x1

HadISST

1982-2002monthly

1982-2002monthly

1.5x1.5

1982-2002 monthly

1982-2002 monthly

1x1ERA40 u*

1982-2002monthly

1.5x1.5

1982-2002monthly

1x1ERA40u10





4x5





1x1





0.25x0.25

AMSRu10

2000-2004daily, monthly

1x1

2000-2004 daily, monthly

0.25x0.25

QSCATu10

1.5x1.51x14x51x10.25x0.251x1

ERA40 HsAMSR SSTReynolds SST

model re-analyses

passive microwave

scatterometers


Satellite data

• u10, SST– Model reanalysis (ERA-40): 1982 – 2001– Passive microwave (AMSR-E): 2003 –

2004– Scatterometer (QuikSCAT): 2000 – 2004

• u*, Hs, σKu– Altimeter (TOPEX): 1992 –

2004gridded to 1.5ºx1.5º


Time series

global flux, 1982 – 2001assuming no trend in ∆pCO2

SST: Reynolds IOwind: ERA-40, monthly, R corr.SSS: NADC∆pCO2: Takahashi et al., 02k: Wanninkhof, 92


Air-sea flux of CO2 in 2001Jan Feb Mar Apr

May Jun Jul Aug

Sep Oct Nov Dec

1.5

0

-1.5

[mol/(m2 month)]


Satellite data - corrections

Wind distributionsFlux differences resulting from Rayleighdistribution of wind speed vs. realistic distribution (calculated from 12hr QuikSCAT data, 2000-2004)

R = un/un with n = 2,3

Global flux: ∆F = 0.2 – 0.6 Gt C/yr


Satellite data - corrections

Skin effectFlux differences resulting from skin effect (sst – 0.17°C for u10 > 6 m/s) mainly affecting pCO2w, but also solubility and transfer velocity.

Global flux: ∆F = 0.3 – 0.6 Gt C/yr


Wind speed ≠ wind speedScatterometer(QuikSCAT)

Model re-analysis(ERA-40)


Wind speed ≠ wind speedRQSCAT – RAMSR

[% RQSCAT]30

-30

0

QuikSCAT vsAMSR-E wind

speed1:1fit


Results: u10-basedk parameterizations

-2.2-2.1-2.7-2.6WG’99-1.5-1.6-1.8-1.9W’92

-1.3-1.5-1.5NG’00-0.9-0.9-1.0LM’86

AMSR-EECMWF ERA40

QuikSCATERS1 scatt.

∆F (k parameterization): 300% (150% W’92/WG’99)∆F (data source): 29%


Results: Global fluxes

1.0118.4Altimeter 1a/σo

Ku +c + b u*H/υ

1.6618.4Altimeter 2x(a/σo

Ku +c) + y b u*H/υ

2.1317.7Wanninkhof & McGillis (u3)

1.6018.2Wanninkhof ‘92 (u2)

Net Sink[Gt C/yr]

Mean Transfer Velocity [cm/h]

kJ:kb=3:1

kJ:kb=1:3


Transfer velocityNorth Atlantic summer (Apr-Sep)W’92

k = 13.3cm/h

ALT1k = 15.3cm/h

ALT2k = 11.1 cm/h

WG’99k = 11.2 cm/h


Transfer velocityNorth Atlantic winter (Oct-Mar)

W’92k = 22.2

cm/h

ALT1k =19.9 cm/h

ALT2k = 22.6 cm/h

WG’99k = 25.0 cm/h


Transfer velocitySummer N Atlantic

(Apr – Sept)ALT2 – W’92

Winter N Atlantic(Oct – Mar)ALT2 – W’92

10

0

-1010

0

-10

cm/h

cm/h


Inter-annual variability

Sensitivity of whitecapping(u*H/υ) in the theNorth Atlantic to the NAO Index.


Results: Regional fluxesSummer North Atlanticsea-air flux [Gt C/y]W’92 -0.19 WG’99 -0.19

ALT1-0.24 ALT2 -0.20

W’92 -0.31 WG’99 -0.37

ALT1 -0.36 ALT2 -0.41

Winter North Atlanticsea-air flux


Summary

• Global mean transfer velocities of new parameterizations are similar to conventional ones, global fluxes vary significantly

• regionally bubble mediated part of the transfer contributes significantly to net flux (e.g., winter NA)

• significant inter-annual variability


Conclusions• measuring wind speed is tricky• flux calculations need documentation

(wrt data sources and processing) & error bars - ~30% between scatterometers and passive microwave or model re-analysis

• k parameterizations which go beyond simple wind speed dependence show promise but require validation

• revised regional & global CO2 fluxes

comparing parameterizations of gas transfer velocity and their

Documents