Techniques for the synoptic analysis, vertical inference, dynamical adjustment, and forecast of altimetric and deeper in situ data are presented as a first step towards the design of continuous assimilation schemes in limited-area oceanic domains. A year-long time series of streamfunction maps, denoted Mark 2, drawn in the POLYMODE area of the western North Atlantic is used as a benchmark for various tests and simulations. An original projection/extension scheme using empirical modes of density and/or pressure anomalies is used to obtain a first guess of the three-dimensional structure of the currents, starting from surface topography only. In the Mark 2 domain, this technique works well, since the first empirical mode is surface intensified and largely dominant. An alternative approach is to incorporate deeper data, e.g., float trajectories below the main thermocline. The first-guess currents are specified as initial and boundary condition of the Harvard 6-level quasi-geostrophic open ocean mode... Abstract Techniques for the synoptic analysis, vertical inference, dynamical adjustment, and forecast of altimetric and deeper in situ data are presented as a first step towards the design of continuous assimilation schemes in limited-area oceanic domains. A year-long time series of streamfunction maps, denoted Mark 2, drawn in the POLYMODE area of the western North Atlantic is used as a benchmark for various tests and simulations. An original projection/extension scheme using empirical modes of density and/or pressure anomalies is used to obtain a first guess of the three-dimensional structure of the currents, starting from surface topography only. In the Mark 2 domain, this technique works well, since the first empirical mode is surface intensified and largely dominant. An alternative approach is to incorporate deeper data, e.g., float trajectories below the main thermocline. The first-guess currents are specified as initial and boundary condition of the Harvard 6-level quasi-geostrophic open ocean mode...