Predicting hot-spots of land use changes in Italy by ensemble forecasting

In the context of environmental change projections at the regional level, not only the climate but also the land use plays a key role. The limited availability of historical information reduces the possibility to calibrate land use change (LUC) models. Even in case of successful calibration, using it both for diagnostic and prognostic studies does not guarantee the reliability of single future simulations. Through ensemble forecasting, useful LUC predictions are evaluable. In this work, after introducing a modified version of the well-assessed CLUE-S model, we present reasonable hot-spots of LUC in Italy for the end of 21st century, derived from the agreement of a 32 simulation ensemble performed alternating two choices for five model configurations or inputs: (1) two different climate projections (reflecting A2 and B2 emission scenarios by IPCC, respectively); (2) two different degrees (slight and strong) of demographic increase; (3) the conservation (or not) of protected areas; (4) the influence (or not) among adjacent land uses in determining their shift; and (5) the importance (or not) of past/recent LUC trends. Results, in terms of LUC hot-spot distribution, were evaluated at administrative, biogeographical, physiographic, and watershed level. The main findings highlighted that some trends of land use substitution will be likely opposite to the past and that a more detailed spatial scale can detect situations neglected by coarser scale evaluations, and due to different transpositions of directives from high-levels to local scales. Biogeographical and physiographic settings seem strongly influencing LUCs, and administrative and catchment units across Italy show very different developments and a highly fragmented territory in terms of LUC hot-spots, all that to be considered in landscape and resource planning.