Our work takes part of a generic approach proposal to efficiently deal with urban data within a GIS. Currently, the major shortcomings of a GIS are in the spatiotemporal data design and in the incomplete coverage of the temporal dimension and multi-representation of data. For that, we propose a GIS based on a conceptual model specific to spatiotemporal data modeling MADS, and on an ODBMS for storing and manipulating data. MADS (Modeling of Application Data with Spatio-temporal features) (Parent et al. 1998) is a modeling tool dedicated to spatiotemporal data. It is based on the Entity-Relation formalism, and allows the representation of real world entities as objects and associations and supports orthogonality between the structural, spatial and temporal dimensions. MADS has interesting features like multiple data representations, inheritance, effective temporal and historical data manipulation, and uniform handling of spatiotemporal data. It has intuitive and very readable visual notations to represent spatial and temporal concepts clearly. However to take advantage of the semantics offered by the MADS model and to achieve its coupling with a GIS, we perform its transformation into an object model. For that, a key step in creating structures (generic classes) simulating the MADS concepts in the object language is necessary. This step is followed by the definition of mapping rules (algorithm) that allow the translation of specific conceptual schemas defined by MADS into the corresponding object language. In this paper, we present the general procedure for implementing MADS into an object paradigm. We then detail the transition rules to express the thematic, spatial, temporal and multi-representation characteristics of MADS schemas into an object language. The transformation rules of concepts like multi association, spatial aggregation relationship, transition relation and generation relation are presented. In fact, the orthogonality of the structural dimensions (spatial, temporal and multi-representation) on one hand and of the structural concepts (entity, attribute, association) of the MADS model on the other hand, allowed us to conduct separate studies of transformation of these dimensions in each concepts. The concepts offered by the programming languages models (on which the ODBMS is based) are less expressive than those of MADS. So semantic verification methods are implemented automatically, on different levels, during the transformation to retain the semantics imposed by the designer when creating conceptual MADS schemas. We propose the design of spatiotemporal applications to be made according to MADS schemas whose formal specifications then undergo a series of systematic transformations in operative object programs. The scripts generation, the compliance with OGC specifications (Herring et al. 2006) and the Object Database creation are viewed as sequences of the transformations.