Journal of Vegetation Science && (2011) Tracking human impact on current tree species distribution using plant communities Daniel E Silva, Vincent Badeau, Myriam Legay, Emmanuel Corcket & Jean-Luc Dupouey Keywords Fagus sylvatica; Human impact; Niche model; Plant community composition; Potential distribution; Range margin; Silviculture Nomenclature for taxa Flora Europaea (Tutin et al. 1964–1980) Received 4 March 2011 Accepted 12 August 2011 Co-ordinating Editor: Ingolf Ku ¨ hn Dupouey, J.-L. (corresponding author, dupouey@nancy.inra.fr), E Silva, D. (esilva@nancy.inra.fr), Badeau, V. (badeau@nancy.inra.fr) & Legay, M. (myriam. legay@onf.fr): Forest Ecology and Ecophysiology Unit, INRA-Nancy University, 54280, Champenoux, France Corcket, E. (emmanuel.corcket@u-bordeaux1. fr): Bordeaux University, UMR1202 BIOGECO, F-33400 Talence, France Abstract Questions: Can we adequately model beech distribution based on accompany- ing vegetation, and how does this compare with a climate model? To what extent does accompanying vegetation predict larger extension of beech at its SW margin compared with current distribution? Do human factors explain the dis- crepancy between current and potential distribution? Location: Lowland forests at the SW European range margin of beech. Methods: We built a logistic regression model of beech presence based on accompanying understorey vegetation using 66 976 plots in the French National Forest Inventory (NFI) network. Explanatory variables were obtained by multivariate reduction of 252 species in NFI plots. The model was calibrated in regions where human impact on beech has been comparatively low. The probability of beech presence was estimated at its SW margin and compared with current distribution. We tested whether forest management factors could explain beech absence in locations where the flora predicted its presence. Perfor- mance of this model was compared with a classical climatic-envelope model. Results: Modelled potential distribution of beech along its SW margin was lar- ger than observed distribution, suggesting area shrinkage under man’s influ- ence. Sites where beech was predicted as present but was currently absent were significantly less common in high forest stands, and more common in coppices, plantations and private forests. Characteristics of species replacing beech (early- successional, plantation or exotic species) confirmed the role of forest manage- ment and disturbance in beech disappearance. The floristic model provided a finer resolution distribution map than the climatic-envelope model. Conclusions: Plant communities, together with precise tree species maps, allowed us to estimate manipulation of beech by man, and main silvicultural causes of its disappearance. The improvement in quality of current distribution models has important implications for modelling of niches under future climate scenarios. Introduction As climate is considered to be the main factor influencing the distribution of tree species (Woodward 1987; Davis & Shaw 2001; Normand et al. 2009), improving our ability to model tree species climatic niche is necessary to better predict future suitability and possible migration in the con- text of climate change (Pearson & Dawson 2003), espe- cially at their range margins (McKenney et al. 2007). Most models have predicted a reduction at the trailing edge of distribution ranges in the coming decades in response to climate change (Benito-Garzon et al. 2008; Badeau et al. 2010). Niche modelling used in such approaches establishes a statistical relationship between species presence and environmental descriptors; but dispersal limitations can increase discordances between observed distribution and fundamental niche. Svenning & Skov (2004) found that less than 50% of the climatically suitable range of numer- ous tree species was currently occupied when modelled Journal of Vegetation Science Doi: 10.1111/j.1654-1103.2011.01341.x © 2011 International Association for Vegetation Science 1