- The influence of managemenT hisTory on spaTial predicTion of Eryngium spinalba - 139 Applied Vegetation Science 11: 139-148, 2008 doi: 10.3170/2007-7-18337, published online 14 December 2007 © IAVS; Opulus Press Uppsala. Abstract Question: Spatial prediction of plant populations is essential for conservation management. This is especially true for rare and/or threatened endemic species, for which knowledge of determinants of distribution is necessary to mitigate threats and counteract decline. We therefore ask if the distribution of an endemic species can be accurately predicted by georeferenced environmental variables or, if anthropogenic variables also need to be taken into account. Location: Alps, Hautes-Alpes, France. Methods: Potential distribution area and abundance of Eryngi- um spinalba were predicted with logistic regression and ordinal logistic regression, respectively, in a 57-km² watershed. Results: Aspect, global solar radiation in March, elevation and grazing pressure were the main predictors of the probability of occurrence of Eryngium spinalba. Taking into account the persistence of agro-pastoral activities by diachronic analysis (Napoleonic cadastral map and orthorectifed photographs) improved predictions from the model and the level of spatial concordance with independent surveys. Conclusions: Niche modelling improved our understanding of the distribution of this threatened species which, in the context of land abandonment, is diminishing as a result of the decline of its favoured habitats. The key role of pastoral activities and historic continuity for its distribution and persistence was clearly demonstrated. Keywords: Calcareous grassland; Conservation management; Endemism; Geographical Information System; Grazing; Land abandonment; Spatial modelling. Abbreviations: AIC = Akaike’s information criterion; AUC = Area under the curve; CBNA = Conservatoire Botanique National Alpin; DEM = Digital elevation model; GLM = Generalized linear model ; LP = Linear predictor; LRM = Lo- gistic regression model ; PET = Potential evapotranspiration; PHDM = Potential habitat distribution map; ROC = Receiver operating characteristic; UGB = Stocking equivalent of one adult bovine. Nomenclature: Kerguélen & Brisse (1994). The infuence of management history on spatial prediction of Eryngium spinalba, an endangered endemic species Marage, Damien 1* ; Garraud, Luc 2 & Rameau, Jean-Claude(†) 1 AgroParisTech, UMR1092 Laboratoire d’étude des Ressources Forêt-Bois (LERFoB), 14 rue Girardet, CS 4216 F-54000 Nancy, France; 2 Conservatoire Botanique National Alpin de Gap-Charance, Domaine de Charance, 05 000 Gap, France; E-mail l.garraud@cbn-alpin.org; * Corresponding author; Fax +33 383396878; E-mail damien.marage@agroparistech.fr Introduction The distribution of plant species in terrestrial eco- systems is infuenced by climatic, physiographic and geological factors which are all part of a natural or an- thropogenic regime that structures these ecosystems in space and time. Distributions are, therefore, the results of complex processes in a hierarchical continuum (Allen & Starr 1982). Many ecological studies have considered plant species distributions both in environmental and geographical space (Brown 1994; Guisan et al. 1998; Zimmermann & Kienast 1999), in which the fundamental niche of a species can be defned as the range of envi- ronmental conditions where a species may persist over time, i.e. complete a full life cycle (Austin & Smith 1989; Austin & Gaywood 1994). Interspecifc interactions such as competition and facilitation may reduce or increase the fundamental niche so that the realised niche (Grubb 1977) differs from the fundamental niche. Static modelling of species distributions (Guisan & Zimmerman 2000) aims to link the species distributional area to the environmental variables that best defne the realized niche. Many such studies have been conducted (Richerson & Lum 1980; Lenihan 1993; Beerling et al. 1995; Huntley et al. 1995; Guisan et al. 1998; Zim- mermann & Kienast 1999; Collingham et al. 2000), and niche modelling is becoming an increasingly important tool for biodiversity management (Lehmann et al. 2002). In Europe, Zimmermann & Kienast (1999) and Guisan et al. (1998) have modelled distributions of Alpine plants using logistic regressions, but niche modelling has not been used in the Mediterranean mountains where anthropogenic infuences are assumed to play a crucial role for plant distributions (Douguedroit 1976; Blondel & Arroson 1995; Reille et al. 1996). Archaeological data and paleo-environmental re- search (Reille et al. 1996; Lopez-Saez et al. 2001) have documented that Mediterranean mountains have been used by humans for a long time (Roberts et al. 2001). For ca. 7000 years, plant communities of these mountains have been profoundly affected by anthropogenic impacts. The intensity and duration of human activities, including