Background and aims – The conservation of a rare and endangered plant species requires a clear knowledge of its habitat distribution. Species Distribution Models (SDM) are generally applied to characterize species' suitable habitats and to predict their potential distribution. However, this method is not well suited to rare species that require a fine spatial scale approach. Co-occurring plant species of a targeted species can be used to define suitable habitats taking into account biotic interactions. Our aim was to construct a predictive model of the presence of the rare Limonium girardianum, using the species that are the best indicators of its presence. Methods – In the main sites in the French coastal salt marshes where L. girardianum occurs, we investigated 198 quadrats (1 × 1 m) with and without L. girardianum in three strata along a topographical gradient. In these quadrats, we recorded plant species and measured the maximum water level during flooding, the soil moisture, salinity and texture. We calculated the species indicator value of the presence of L. girardianum. For each quadrat, we added the species indicator values to obtain a score used to model the presence of L. girardianum. We compared the predictive power of the model based on co-occurring plant species with a model based on environmental variables. Key results – The best indicator species of the presence of L. girardianum are Limonium virgatum, Hymenolobus procumbens and Frankenia pulverulenta. The predictive power of the model based on the co-occurring plant species appears to be similar to that of the model based on environmental variables, with about 72% of correct predictions. However, the environmental model shows higher False Positive predictions than the plant species model. Conclusions – The co-occurring plant species can be used to define suitable habitats for L. girardianum. The model based on the co-occurring plant species, that integrates the biotic interactions, would appear to be more efficient to define the habitat where L. girardianum is most likely to be found. Furthermore, as extensive and numerous vegetation databases are available, this simple method could be used to predict the presence of several species with low abundance.

AREA UNDER THE ROC-CURVE (AUC); FUNDAMENTAL NICHE; GENERALIZED LINEAR MODEL (GLM); INDICATOR SPECIES; LIMONIUM GIRARDIANUM; RARE SPECIES; REALIZED NICHE; RECEIVER OPERATING CURVE (ROC); SALT MARSHES; SPECIES DISTRIBUTION MODELS