Paleosalinity: confidence limits and future applications

Abstract

Error propagation is investigated through the only commonly applied technique for reconstruction of paleosalinity, based on oxygen isotope residuals. Although this technique is known to be fairly inaccurate, there remained a need for thorough assessment of confidence limits in a variety of global settings. Here, it is found that a theoretical minimum error bar of between ±0.3‰ and 0.7‰ applies to paleosalinity estimates, and that real-life practical limits are rarely better than ±0.6‰. Best results are expected in the latitudinal ranges >50°N and >70°S, taking care to avoid the sea–ice margin. Confidence intervals at low and mid-latitudes (including marginal seas) are very wide, at least ±1.8‰ for the Mediterranean and ±2.2‰ for the equatorial Atlantic. It appears that oxygen isotope distributions are best used in their own right to investigate circulation and advective processes in coupled ocean–atmosphere–cryosphere models, instead of calibrating them to paleosalinity to define restoring boundary conditions. However, isotope residual-based paleosalinity maps from large numbers of (regionally averaged) studies in the above-mentioned latitudinal bands may help in defining initial conditions and first-order validation arguments for models.