Assessment of climate indicators and morphological features in urban areas within the framework of climate scenarios: A sustainable landscape planning approach

This doctoral research focuses on the historical city of Tobruk, located in northeastern Libya, and aims to evaluate the impacts of desert climate conditions on urban heat island (UHI) effects, vegetation loss, and the vulnerability of cultural landscape assets through an integrated, multi-disciplinary approach. The study combines remote sensing, Geographic Information Systems (GIS), statistical modeling and explainable artificial intelligence techniques to produce both scientifically robust results and practical planning scenarios. Using Landsat and Sentinel-2 satellite imagery spanning the period from 1991 to 2024, key environmental indicators such as Land Surface Temperature (LST), Normalized Difference Vegetation Index (NDVI), and Secondary Urban Heat Island Intensity (SUHII) were calculated. Morphological analyses were conducted using both 2D variables (building density, impervious surfaces) and 3D indicators (building height, Sky View Factor - SVF). Landscape metrics were evaluated within a 300-meter grid framework, and the influential variables affecting LST were identified through machine learning and interpretable modeling approaches. The findings reveal that building volume density, building height, and low SVF values significantly contribute to increased LST. A negative correlation was observed between NDVI and LST, with vegetated areas showing an average of 6–7 °C lower surface temperatures. Nature-based scenarios applied to selected pilot areas led to an 21% reduction in LST and a %28 increase in NDVI. According to SHAP (SHapley Additive exPlanations) analysis, the most influential variables on LST were building volume density (50.9%), NDVI (25.8%), and building height (20.4%). This study proposes a data-driven, interdisciplinary model for sustainable microclimate planning in historic urban environments and demonstrates how remote sensing combined with explainable AI can effectively guide decision-making processes.