Can We Reconstruct Mean and Eddy Fluxes from Argo Floats?

Abstract

The capacity of deep velocity estimates provided by the Argo float network to reconstruct both mean and eddying quantities, such as the heat flux, is addressed using an idealized eddy resolving numerical model, designed to be representative of the Southern Ocean. The model is seeded with 450 "virtual" Argo floats, which are then advected by the model fields for 10 years. The role of temporal sampling, network density and length of the float experiment are then systematically investigated by comparing the reconstructed velocity, eddy kinetic energy and heat-flux from the virtual Argo floats with the "true" values from the from the model output. We find that although errors in all three quantities decrease with increasing temporal sampling rate, number of floats and time span, the error approaches an asymptotic limit. Thus, as these parameters exceed this limit, only marginal reductions in the error are observed. The parameters of the real Argo network, when scaled to match those of the virtual Argo network, generally fall near to, or within, the asymptotic region. Using the numerical model, a method for the calculation of cross-stream heat-fluxes is demonstrated. This methodology is then applied to 5 years of Argo derived velocities using the ANDRO dataset of Ollitrault \& Rannou (2013) in order to estimate the eddy heat flux at 1000m depth across the Polar Front in the Southern Ocean. The heat-flux is concentrated in regions downstream of large bathymetric features, consistent with the results of previous studies. 2$\pm$0.5TW of heat transport across the Polar Front at this depth is found, with more than 90\% of that total concentrated in less than 20\% of the total longitudes spanned by the front. Finally, the implications of this work for monitoring the ocean climate are discussed.