This paper describes the geo-spatial multidimensional database architecture we built to construct a regional air quality decision-support system. The system retrieves data from three main categories of system sources: a numerical atmospheric model, in -situ detectors and satellites. We worked on an idealized case study for which we established spatial multidimensional OLAP cubes on specific measures and dimensions. This sche ma aims to provide original an alysis like deviations of the concentration simulated by the atmospheric model pe r region, per date and pe r pollutant. Cartographic representation of the results is then achieved by using the SOLAP approach. Thematic maps are generated as a response to a particular query that the end user may submit. A range of options are offered to explore these charts in both space and time. Cartographic contours are not based on pre-defined images but are dynamically elaborated from the numerical coordinates stored in the spatial dimension. Qualities of the design and implementation are finally assessed according to recognized criteria established by the OLAP/SOLAP community as scalabi lity, time response a nd data exploration intuitiveness.