Salinity model errors induced by wind stress uncertainties in the Macaronesian region

Motivated by the imminent launches of two Sea Surface Salinity (SSS) satellite missions, this paper investigates salinity error covariances due to wind stress uncertainties in a regional simulation of the eastern North-Atlantic Ocean from 2000 to 2006. This work follows the study by Mourre et al. (2008), which concluded that the wind stress forcing was the main source of SSS model error in this region. The present study is based on ensemble simulations that start from the same initial conditions but are driven by different wind stresses. The error modes of the wind stress forcing field are approximated by the variability modes of the wind stress. sults indicate that the surface salinity ensemble spread has an average value over the modelling domain of around 0.1 psu. Under the influence of model dynamics, the largest spreads are found in areas of significant SSS advection. The surface salinity spread also experiences a significant temporal variability at seasonal and interannual scales. The periods of error growth have a duration around 3–4 months. Horizontal SSS error correlation distances are inhomogeneous and anisotropic due to ocean dynamics. They are larger than 100 km almost everywhere in the domain, and can extend over 400 km along the path of the main surface currents. In addition, salinity and temperature anomalies induced by the wind stress forcing are not confined to the surface layers, but propagate into the ocean interior under the influence of vertical dynamical processes. Vertical salinity error correlation distances from the surface are locally larger than the annual maximum mixed layer depth. In areas where subduction occurs, the vertical correlation radii exceed 500 m. ing to these results, it is expected that satellite SSS data assimilation in this region will be challenging because (i) the magnitude of the SSS model error is of the same order as the accuracy of the observations, and (ii) the horizontal SSS error correlation scales are comparable to the spatial resolution of satellite SSS products. On the positive side, the temporal resolution of satellite SSS data is finer than the time scale of SSS model error growth.