**2. Materials and Methods**

### *2.1. Spatial Data and Climatic Scenarios*

Forest inventory plots represent one of the main input data for SDM procedures, given their ability to provide tree-level information which allow a refinement of modelling steps. Among the 263 tree species detected in the framework of the last available national forest inventory (INFC 2005) 19 forest tree species were considered in this study and selected as the most interesting and relevant for Italy under economic, ecological, and aesthetic aspects. Their ecological requirements were previously studied by Pecchi et al. [25]. INFC 2005 was based on a three-phase sampling procedure resulting in a total of 7272 sampling plots, spatially distributed according to a probabilistic sampling scheme [55] and with associated data for 230,874 trees measured in the field [56]. In this framework, statistical inferences on the realized ecological niche of the 19 considered tree species was possible due to the probabilistic sampling scheme.

In order to derive the climatic niche of target species and to project its spatial distribution into the future conditions, current climate data (1981–2010 normal period) were firstly retrieved from the downscaled E-OBS climatological maps. This dataset is available for the whole Italy at 1 km of spatial resolution as a result of a downscaling procedure [46,57]. Such data were then used to generate the set of 19 Worldclim's bioclimatic variables to be used as predictors in SDM. This set is format by a series of biological important variables that better describe the annual and seasonality trends and the extreme and limiting factors [58]. These variables are generated using dismo, a package available for R statistical language [59] using the bioclim function. This step was done to compare the current climate condition with six GCMs we downloaded from the WorldClim website with 30 arc-sec of spatial resolution. The selected GCMs are those elaborated by the fourth version of Community Climate System (CCSM) here and for the following models CC, the Hadley Centre Global Environment Model version 2 family (HADGEM2 2-AO, 2-CC, 2-ES), respectively, HD, HE, and HG, the Max Planck Institute for Meteorology Earth System Model (MPI-ESM-LR) hereafter MP and the Meteorological Research Institute climate model (MRI-CGCM3) MG. To avoid potential biases that originated from different climate data sources (i.e., WorldClim portal and E-OBS data), the WorldClim future projections were recalculated as anomalies from the 1961–1990 climatic normal period, currently distributed as WorldClim version 1.4 [60,61]. Once anomalies were calculated, these were added to the same climatic normal period we obtained from E-OBS for Italy (1961–1990), using spatial reprojection to realign the two grids. An additional climate dataset was then added to this study and provided by the Institute of Bio-Economy (IBE) of Italian National Research Council (CNR), representing the RCM we used in this study. The RCM model is here represented by the output of COSMO-CLM climate model hereafter, COSMO, the climate version of operational weather forecast model COSMO-LM, developed by the

German weather service [62]. This RCM was selected for its acknowledged ability to characterize the Italian climate conditions [63]. All climatic scenarios were referred to RCP 4.5 of AR5 for 2050s.
