This project aims to develop a model of water mass (WM) mixing capable to resolve the key steps of the ocean carbon cycle. Present models have difficulties in resolving mixtures involving more than four WM when assuming fixed Redfield Ratios (RR). A number of methodological improvements are envisaged. These include using physical tracers (such as potential vorticity) or additional chemical ones such as total alcalinity (AT), total inorganic carbon (CT) and C14 in order to enlarge the number of available equations, while at the same time new algorithms are developed to describe and model the spatial variability of the RR. Other improvements based on perturbation methods and with high computational cost will allow a precise evaluation of the uncertanties of the new model. This will be applied to all data available in global data bases of international programmes (GLODAP, CARINA, WOCE). The results of the model will include estimates of AT and CT levels at the pre-industrial age, thus improving determinations of CO2 with anthropic origin. They will also lead to spatial distributions of solved CaCO3 and organic matter remineralization, as well as better estimates of WM ages.