The Argo array, by offering simultaneously global coverage and high vertical resolution, made the monitoring of inter-annual and decadal variability of the ocean interior possible. In the ARIVO project, we have initiated the production of gridded monthly fields of temperature and salinity using an optimal analysis tool and designed an iterative approach to combine near-real time and delayed mode data. The monthly gridded fields are averaged over time to produce the reference state needed by climatological studies, dynamical modeling and reconstruction methods themselves. Higher degree statistics such as variances and extreme values are later derived. Global indices (heat content, dynamical height) and corresponding error bars are also computed. Sensitivity tests and careful examination of the analysis results have been performed to detect biases and inconsistencies. A strategy is proposed to reduce these errors and produce a coherent statistical representation of the first decade of the third millennium.