[en] The presented methodological study illustrates a Geo-statistical approach suitable for radiological evaluation in nuclear premises. The waste characterization is mainly focused on floor concrete surfaces. By modeling the spatial continuity of activities, Geo-statistics provide sound methods to estimate and map radiological activities, together with their uncertainty. The multivariate approach allows the integration of numerous surface radiation measurements in order to improve the estimation of activity levels from concrete samples. This way, a sequential and iterative investigation strategy proves to be relevant to fulfill the different evaluation objectives. Waste characterization is performed on risk maps rather than on direct interpolation maps (due to bias of the selection on kriging results). The use of several estimation supports (punctual, 1 m², room) allows a relevant radiological waste categorization thanks to cost-benefit analysis according to the risk of exceeding a given activity threshold. Global results, mainly total activity, are similarly quantified to precociously lead the waste management for the dismantling and decommissioning project. This paper recalled the geo-statistics principles and demonstrated how this methodology provides innovative tools for the radiological evaluation of contaminated premises. The relevance of this approach relies on the presence of a spatial continuity for radiological contamination. In this case, geo-statistics provides reliable activity estimates, uncertainty quantification and risk analysis, which are essential decision-making tools for decommissioning and dismantling projects of nuclear installations. Waste characterization is then performed taking all relevant information into account: historical knowledge, surface measurements and samples. Thanks to the multivariate processing, the different investigation stages can be rationalized as regards quantity and positioning. Waste characterization is finally obtained through the analysis of probability maps of exceeding activity levels. The estimation support must be taken into account to discriminate punctual issues, such as hot spot identification, and waste production issues on larger areas. The main goal of this data processing remains an easier radiological waste management and the best waste categorization with acceptable investigation costs. Ongoing research deals with the implementation of the Geo-statistical methodology on nuclear systems and equipments and on former radiological waste storages. (authors)