Uncertainties in wind speed dependent CO2 transfer velocities due to airflow distortion at anemometer sites on ships

Griessbaum F, Moat BI, Narita Y, Yelland MJ, Klemm O, Uematsu M

Zusammenfassung

Data from platforms, research vessels and merchant ships are used to estimate ocean CO2 uptake via parameterisations of the gas transfer velocity (k) and measurements of the difference between the partial pressures of CO2 in the ocean (pCO(2) (sw)) and atmosphere (pCO(2) (atm)) and of wind speed. Gas transfer velocities estimated using wind speed dependent parameterisations may be in error due to air flow distortion by the ship's hull and superstructure introducing biases into the measured wind speed. The effect of airflow distortion on estimates of the transfer velocity was examined by modelling the airflow around the three-dimensional geometries of the research vessels Hakuho Maru and Mirai, using the Large Eddy Simulation code GERRIS. For airflows within +/-45 degrees of the bow the maximum bias was + 16%. For wind speed of 10 ms(-1) to 15 ms(-1), a + 16% bias in wind speed would cause an overestimate in the calculated value of k of 30% to 50%, depending on which k parameterisation is used. This is due to the propagation of errors when using quadratic or cubic parameterisations. Recommendations for suitable anemometer locations on research vessels are given. The errors in transfer velocity may be much larger for typical merchant ships, as the anemometers are generally not as well-exposed as those on research vessels. Flow distortion may also introduce biases in the wind speed dependent k parameterisations themselves, since these are obtained by relating measurements of the CO2 flux to measurements of the wind speed and the CO2 concentration difference. To investigate this, flow distortion effects were estimated for three different platforms from which wind speed dependent parameterisations are published. The estimates ranged from -4% to +14% and showed that flow distortion may have a significant impact on wind speed dependent parameterisations. However, the wind biases are not large enough to explain the differences at high wind speeds in parameterisations which are based on eddy covariance and deliberate tracer methods.