Systematic errors in global air-sea carbon fluxes caused by temporal averaging of sea-level pressure

Helen Kettle and Chris Merchant, University of Edinburgh

[ Atmospheric Chemistry and Physics. Vol. 5: p1459-1466]

Long-term time averaging of meteorological data, such as wind speed and air pressure, can cause large errors in air-sea carbon flux estimates. Other researchers have already shown that time averaging of wind speed data creates large errors in flux due to the non-linear dependence of the gas transfer velocity on wind speed (Bates and Merlivat, 2001). Here we show that long-term time averaging of the partial pressure of CO2 in the atmosphere (pCO2air) also causes significant errors in flux. pCO2air changes over time according to changes in air pressure (P) such that

pCO2air = m (P - SVP)

where m is the atmospheric molar fraction of CO2 in air and SVP is the saturation vapour pressure of water just above the sea surface. In general, wind speed is negatively correlated with air pressure so that low pressure systems which cause a drop in pCO2air, are associated with high winds. The simultaneous drop in pCO2air and increase in wind speed thus leads to enhanced outgassing/reduction of uptake of CO2 from the ocean. We calculated the errors in global carbon flux estimates caused by using monthly or climatological pressure data to calculate pCO2air (and thus ignoring the covariance of wind and pressure) over the period 1990-1999, using Wanninkhof's (1992; W92) and Wanninkhof and McGillis' (1999; WM99) parameterisations for gas transfer velocity. The results showed that on average, compared with estimates made using 6 hourly pressure data, the global oceanic sink is systematically overestimated by 7% (W92) and 10% (WM99) when monthly mean pressure is used, and 9% (W92) and 12% (WM99) when climatological pressure is used because the covariation of wind and pressure is ignored.

Figure 1 below shows how the errors in flux caused by time averaging air pressure (and hence pCO2air) vary seasonally with larger errors in winter. Figure 2 shows the spatial pattern in the flux errors with the largest errors in regions that have high wind-pressure covariation. Figure 3 shows this more explicitly - the large amount of scatter in this plot is due to the wide range of oceanic pCO2 and sea surface temperature and salinity over regions with the same wind-pressure covariance.

	Fig. 1: Net air-sea carbon flux over time. Red and black lines represent WM99 and W92 gas transfer parameterisations respectively.
	Fig. 2: First 2 plots show mean errors (mol C m-2 yr-1) in flux fields caused by using monthly pressure averaging (1990-1999) for W92 and WM99 gas transfer parameterisations. Third plot shows the wind-pressure covariance field (mb m/s). White cells indicate sea ice.
	Fig. 3: Relation between flux error caused by monthly averaging of pressure (ie the difference between the 6 hourly and monthly flux fields) averaged over 1990-1999, and the wind-pressure covariance for W92 and WM99.

References