Abstract : This thesis is in the field of spatiotemporal modelling, and our work focuses specifically on the management of territorial statistical information. Today, the availability of large amounts of statistical information by different regional producers (Eurostat, INSEE, the European Environment Agency, the UN, etc..) offers a rich analytical perspective, by the combination of data on various topics (economic, social, environmental), at various levels of study: from the local (municipalities) to global (states). However, it appears that the spatial supports, the definitions, the various classifications and the reliability level of those data are very heterogeneous. This heterogeneity is at the core of our problem. In order to cope with that, that is to say to measure, control and analyse this heterogeneity, we are drawing three proposals allowing for a wiser exploitation of this kind of data. The first proposal aims at taking into account the change of the support through the time, modelling both the evolutive aspect of the territories and their hierarchical organisation. The second proposal deals with the semantic variability of the data values associated to this support, through the use of metadata. A profile of the ISO 19115 standard is defined, in order to ease the edition of those metadata for data producers. The last proposal defines a platform dedicated to spatiotemporal data exploration and comparison. In particular, outliers can be discovered though the use of statistical methods, and their values can be discussed and documented further through the use of metadata showing their origin and how they have been produced.