Sperimentazione di metodologie avanzate per la determinazione di soglie di contenimento del consumo di suolo nella gestione integrata dei sistemi urbani.

Territorial planning plays an increasingly central role in ensuring sustainable development consistent with European objectives for soil protection and the reduction of settlement pressures. In Italy, strong administrative fragmentation, limited integration of advanced analytical tools, and the complexity of local governance processes hinder the effective interpretation and management of urban transformation dynamics. These structural issues contribute to growing land consumption, settlement dispersion, and the degradation of ecosystem functions, highlighting the need for predictive tools capable of supporting more informed decision-making. This research, developed within the GeoSciences IR project (PNRR - Mission 4), investigates the use of predictive models for analysing soil sealing processes in a selection of municipalities within the Province of Padova, an area characterised by complex urban dynamics and high levels of development pressure. Four quantitative modelling approaches—ARX (Least Squares), Random Forest, Regression Tree, and Neural Network—are compared to assess their performance, capabilities, and limitations in interpreting historical land-consumption trends and simulating future scenarios. By integrating ISPRA historical time series (2001-2023), territorial indicators, and data-driven modelling techniques, the study develops a replicable methodological framework aimed at understanding the relationships between territorial factors and urban transformation processes. The results reveal significant differences among the models: machine learning approaches, particularly Random Forest, demonstrate higher predictive accuracy and a superior ability to capture complex patterns compared to linear models; the ARX model, although simpler, proves effective in identifying aggregate temporal trends. Overall, the findings demonstrate that the integrated use of predictive models can provide valuable support to local administrations, enabling the development of more informed planning strategies aligned with the objectives of the 2030 Agenda for Sustainable Development. The research thus contributes to the advancement of operational tools for sustainable land management, promoting an evidence-based approach grounded in certified data, quantitative analysis, and predictive capabilities applicable to local-scale planning.