Introducing thermal sensitivity to devegetation using an AI-based framework for capturing urbanization stress

Despite existing studies on vegetation-temperature relationships, a framework to isolate the thermal response of urban areas attributable solely to devegetation remains limited, restricting its practical use in targeted urban climate management. This study introduces an AI-based framework to pinpoint Thermal Sensitivity to Devegetation (TSD). The input data for the framework consists of the Normalized Difference Vegetation Index (NDVI) and Land Surface Temperature (LST). This framework consists of two levels: (i) Level 1, the principal component analysis, which reduces the dimension of the NDVI and LST images and identifies the principal components, and (ii) Level 2, the self-organizing map, which conducts the cluster analysis using the first components of NDVI and LST at Level 1 and prioritizes the identified clusters based on the simultaneous trend of NDVI decrease and temperature increase. The framework identified TSD clusters in Maragheh, Iran, as a ‘garden city’ threatened by devegetation. The results reveal hotspots in peripheral areas of the city resulting from urbanization. A Silhouette Index of 0.703 confirmed a strong cluster structure. Additionally, the visual compression of TSD clusters using Google Earth images (2007–2024) corroborated severe devegetation in the most thermally vulnerable clusters. The findings provide a spatial basis for prioritizing vegetation conservation and restoration efforts and for supporting climate-responsive urban planning strategies.