Is cellular automata algorithm able to predict the future dynamical shifts of tree species in Italy under climate change scenarios? A methodological approach

In this paper is presented a methodological approach which integrates statistic modelling and 2-D cellular automata (CA) in order to describe tree species shifts responding to the climate changes foreseen for Italy in the 21st century. Five Italian tree species populations of Abies alba, Pinus sylvestris, Fagus sylvatica, Acer campestris and Quercus sober and their actual potential distributions (PDs) - represented by Importance Value (IV), have been considered. Environmental and climatic relationships have been modelled through application of a new statistical methodology called extreme discretization, where the PD of a species was considered as a random field. The IV-based PD has been spatialized through a probability function pi(A,S), which represented the spatio-temporal relationships between IV values and climatic (A) and geomorphological (S) variables. For each tree species pi = (A,S) has been estimated and inserted as rule in the 2-D cellular automata. The latter, acting by a Moore neighbouring, took in consideration also the suitability map for tree species, which has been obtained by land cover map. Two time frames (2050 and 2080) and two climatic scenarios (A2 and B1) have been considered. Results described a general reduction of the IV values and their distribution for A. alba, P. sylvestris and F. sylvatica, in both climatic scenarios, whereas an increase of IVs and distribution for Q. suber and only a slight increment of distribution for A. campestris was mainly observed under the B1 scenario, but not for the more limiting A2 scenario. Convergent results have been obtained with respect to other simulation systems concerning the shift of tree species responding to different climatic change scenarios but lacking of the description of dynamical paths. Our approach seems natural and practical to describe such phenomena. The transition rules for the CA and the parameters taken into account for the construction of the probabilistic models can be surely improved to obtain a more realistic pattern of tree species shifts. Future efforts should be made to take in account the inter-specific relationships inside the Italian forest ecosystems, in order to also consider the competiveness for resources that exert some effects on the plant distribution both in time and space.