Predicting potential shifts in Fagus sylvatica range in response to climate change: comparing and integrating multiple models based on different structural concepts to reduce predictions uncertainty

International audienceClimate change effects on tree function, structure and growth have been observed in various forests throughout Europe and France in the last decades. Forest productivity and survival are limited by a variety of environmental abiotic factors such as soil nutrients, temperature, atmospheric CO2 concentration, and water balance. Differential impacts of global climatic changes on forests, as well as complex interactions with the carbon and water balance, and vegetation feedbacks on climate makes it hard to estimate future vegetation patterns. Obtaining reliable predictions of species range shifts under climate change is a crucial challenge for ecologists and stakeholders. However, it is evident that no single modelling procedure can provide appropriate predictions in the future. Model comparison is a powerful tool to evaluate models in a more conceptual way by comparing and interpreting predictions on the basis of the underlying model approaches and assumptions. This work aims at understanding future spatial distribution, physiological response and potential adaptation of 7 species in France under future climate, by using and comparing simulations stemming from different model concepts. Here, we present only the results for Temperate broadleaf species (Fagus sylvatica and Quercus robur) and compare predictions of range shifts under climate change scenarios for 2050 derived from three niche-based models (BIOMOD, NBM and STASH) with those derived from a mechanistic tree growth process-based model (CASTANEA), a phenology-based model (PHENOFIT) and three dynamic global vegetation model DGVM (LPJ, ORCHIDEE and IBIS). A contrasted pattern emerged from our comparisons: although all models project significant losses ofT emperate broadleaf species range in France in 2050, significant divergences between models appear: niche-based models tend to predict a stronger level of range loss in the plains across France than the other models, and they project less range loss in mountains than in lowlands. LPJ projects the smallest loss of range in plains and increased presence of beech in mountains. In general, LPJ, Orchidee, Castanea and PhenoFit are more conservative in the North, East and North-East of France than in the South, West and South West. This result likely arises because niche-based models do not take complex interactions between physiological processes and climate, phenotypic plasticity, and local adaptation into account. Nevertheless, there is much greater uncertainty in projected tree response to climate change than previous studies have suggested. This means that forest management must focus on dealing with an uncertain future, i.e., by increasing the resilience of forest ecosystems