Integration techniques are used to combine Earth Observation (EO) datasets to study the Water Cycle (WC). By merging several datasets, they reduce uncertainty and introduce coherency among them. Several EO integration methods are presented and compared: The Optimal Selection (OS) simply choses the best individual datasets. Simple Weighting (SW) is a weighted sum of the datasets to reduce uncertainties. Three other techniques introduce a closure-constraint on the WC budget: (1) The SW plus Post-Filtering (PF) is very efficient but it is applied at the basin-scale only, and lacks in spatial information. (2) By using a spatial interpolation scheme, the INTegration (INT) solution allows obtaining a pixel-scale database, but only for the common period of the all the water components. (3) A simple CALibration (CAL) of the EO datasets is therefore introduced to reproduce the INT results over the longer temporal extent of the EO datasets, but its closure constraint is relaxed. Results are presented over the Mediterranean region, one of the more complex environnements and a hot-spot for climate change. We extended previous techniques to close simultaneously the terrestrial, oceanic and atmospheric WC budgets. We also introduce temporal and spatial multi-scaling constraints. The evaluation is performed for precipitation and evapotranspiration: in addition to better close the WC budget, the integrated database is also closer to in situ measurements. The resulting integrated database provides new estimates for the WC components: stock and flux annual-means are re-evaluated, and we now estimate the Bosporus net-flow mean value at 129 ± 60 mm.yr −1 for the 2004–2009 period. This new EO-based database describing the terrestrial, oceanic and atmospheric WC over the Mediterranean is now proposed to the scientific community.