Rapid urbanization and climate change are intensifying heat stress in cities, yet assessments often overlook the physical diversity of urban form across different climatic contexts. This study develops an Urban Heat Vulnerability (UHV) index, with application to a semi-arid city (Irbid, Jordan), while maintaining transferability across different climate zones. The objectives are to: (i) assess exposure, sensitivity, and adaptive capacity across Local Climate Zones (LCZs), (ii) compare indicator weighting using the Entropy Weight Method and Random Forest Regression (RFR), and (iii) identify vulnerable LCZ classes, and quantify the influence of LCZs parameters on heat vulnerability. A GIS-based LCZ map at 250 & times; 250 m resolution was created using urban canopy parameters and integrated with thermal, demographic, morphological, and accessibility indicators to build three component indices. Indicator weights were derived using Entropy and RFR, and the weights were combined within a regularized multiplicative model to compute the UHV index. ANOVA was applied to evaluate differences among LCZs classes and agreement between weighting methods. Results show that exposure is highest in compact and bare LCZs. At the same time, sensitivity and adaptive capacity are strongly influenced by urban morphology, with LCZ type explaining a substantial proportion of their variation. Overall, UHV is greatest in compact mid- and lowrise LCZs and lowest in vegetated classes. Geographically Weighted Regression (GWR) further reveals spatial non-stationarity in these relationships, providing localized insights for targeted mitigation. The proposed LCZbased hybrid weighting framework provides a transferable and scalable tool for urban heat risk assessment, supporting climate-adaptive planning across diverse urban environments.