How useful is Argo for investigation of interannual variability in the North Atlantic heat content?

Rachel Hadfield, Neil Wells and Simon Josey

NOC
Empress Dock
Southampton
SO14 3ZH
UK

The usefulness of the Argo dataset for estimating the annual cycle and interannual variability in the North Atlantic Ocean Heat Content (OHC) is investigated. The work carried out involved subsampling a model temperature field to the Argo sampling density and then using an optimal interpolation (OI) scheme to obtain a gridded temperature field. OHC was calculated from the subsampled, interpolated and full model temperature fields. Comparison between the two OHCs indicates errors.

It was found that away from regions of high mesoscale activity, complex bathymetry and low sampling density, the change in mixed-layer OHC was accurate to about 2 Wm-2. However, within the Gulf Stream region, north of 50°N and south of 25°N, errors were in excess of 15 Wm-2. In these areas Argo cannot be used to investigate interannual variability in OHC since the errors are the same magnitude as expected variability. An experiment was undertaken to determine the critical number of floats required to reduce errors to less than 10 Wm-2 at each 2° grid point. The relationship between RMS differences and buoy numbers was found to be complicated, with dependence on length scales used in the OI and the distribution of the float profiles. An additional experiment, which tested the sensitivity of the interpolated field to the correlation parameters revealed the result was insensitive to selected parameters in areas of low errors. This suggests that in such areas we can be confident in the results even though the statistics of the field are essentially unknown.