Modeling Multidimensional Spatio-temporal Data Warehouses in a Context of Evolving Specifications

In this paper, we study some problems linked to the integration of data in a spatio-temporal data warehouse. In many cases, the specifications of the data sets have evolved over time, especially when the observed period is large. Under those circumstances, data sources have temporal, spatial and semantic heterogeneity. In order to explore and analyse spatio-temporal data sets in a SOLAP (Spatial On Line Analytical Processing) application, we propose two approaches to model heterogeneous data in multidimensional structures. The first solution consists in a unique temporally integrated cube with all the data of all epochs. The second solution consists in creating a specific cube (data mart) for each specific view that users want to analyse. The final objective is to support geographic knowledge discovery through data exploration of detailed data for an epoch and of integrated comparable data for timevariant studies. Using a practical example in the field of forestry, we evaluate the implementation of these two models.