This new numerical approach proposes a solution to a fundamental and difficult question in ecology, consisting of the correct geographical representation of multidimensional structures. Firstly, transformation was applied to the original matrix (n sites x q variables) in order to satisfy the condition of multinormality. Then, a hierarchical cluster analysis was used and cacti hierarchical level was Studied and characterised by a certain probability level. For each cut off level an algorithm based on the computation of the Bayesian probabilities produced a smaller matrix (n sites x c groups). These conditional probabilities measure the chance that each site has in belonging to a predefined group of sites. Spatial distributions of these probability values for each group of sites were mapped using kriging interpolation. Finally, the maps were used to define homogenous zones on a single map by superimposing one map on the other. The maximal value of interpolated probability was used as criterion to assign cacti point of the map to the zones predefined by this classification. This method was applied to map demersal fish habitats by using a dataset from bottom trawl surveys in the Bay of Biscay (France) during October 1990. The boundaries between habitats were identified objectively. Then, the indicator species and species assemblages characterising the different habitats were identified by using an indicator value index. This index integrates the specificity and the fidelity quantities calculated for each species in each habitat. The obtained results showed that this method presented a robust tool to describe the habitat of exploited species. The obtained habitats were validated by their correspondence with depth strata, sediment type and also by the biological characteristics of the indicator species. The proposed method is useful in the study of temporal variations of habitats with regards to species assemblages and can also be generalised to other multivariate databases of different descriptors (physical, chemical, biological, etc.).