Evaluating urban fabric transformations using GeoAI

Urban growth has reshaped land use patterns globally, demanding robust and scalable methodologies to monitor its long-term dynamics. This study proposes a GeoAI-based framework that integrates Random Forest (RF) classification with spatial indicators to analyze urban fabric transformations in Ravenna, northern Italy, from 2000 to 2024. Using Landsat 5 and Landsat 9 multispectral imagery processed in the Google Earth Engine (GEE) cloud computing platform, six Land Use and Land Cover (LULC) classes were mapped with high accuracy. The RF classifier achieved an overall accuracy of 86.2% in 2024, confirming its suitability for complex urban environments. The classified maps were imported into a GIS environment to extract built-up surfaces and compute spatial indicators, including Urban Density (UD), Urban Dispersion Index (UDI), Annual Growth Rate (AGR), and Urban Expansion Index (UEI). Results reveal a moderate densification in Ravenna’s urban core alongside an increase in dispersed residential nuclei, confirming a dual trend of consolidation and sprawl. The indicator values align with northern Italian urbanization trends reported in the literature. This approach demonstrates how combining supervised classification with spatial metrics can provide deeper insights into urban growth, supporting more informed planning and policy-making. The framework is scalable, reproducible, and adaptable to different urban contexts.