In a climate change context, forest managers must consider long-term changes when faced with the choice of tree species during the management process. This consideration is intended to avoid the potential vulnerability of tree species in view of global warming. Traditional decision-making tools (i.e. forest site classifications and silvicultural guides) do not take long-term environmental changes into account. In this context, species distribution models (SDM) are becoming an important decision support tools for forest managers. The performances of these methods have been successfully tested at national scale with spatial resolutions mainly ranging from 1km2 to 100 km2. However, management decision-making requires information at a fine spatial resolution (<1km2) to capture environmental variability at the forest stand scale. The aim of this study was to evaluate the efficiency of distribution models calibrated at national scale using climatic and soil predictors to assess tree species suitable areas at the scale of forest management. Distribution models calibrated at the scale of France for eight broad-leaves and coniferous species of high economic importance were applied to 682 plots carried out on a systematic sampling grid of 50 x 50 to 250 m x 250 m in three lowland and mountain forests of northern France.
Results/Conclusions
Models including climate coupled to soil acidity and nitrogen availability can significantly explain the distribution of the eight studied species at the local scale with a prediction success reaching 85%. The local performance of SDM, evaluated by the area under the curve (AUC) calculated at the forest scale, revealed good to excellent levels (from 0.70 to 0.91). An exception was the pioneer coniferous species, Pinus sylvestris which showed a lower performance (mean AUC of 0.52). Spatial variation of tree species distribution was better predicted for species having a niche highly controlled by soil conditions (e.g. nutrient-demanding species as Fraxinus excelsior or acidophilus species as Castanea sativa with AUC of 0.85 and 0.91, respectively). The best quality of predictions was reached in mountain context where both climatic and soil conditions are highly variable (mean AUC of lowland and mountain forests were 0.70 and 0.78, respectively).
Our results highlight that within a favourable climatic envelope, studied models allowed to identify fine-scale environmental variations useful when the goal is to map tree species suitable or unsuitable areas at the forest scale. This approach, taking into account both soil and climate conditions, allows to formally integrate future environmental changes in forest management decision-making tools.