Geospatial AI for heritage risk assessment: a machine learning approach to safeguarding cultural landmarks

Traditional heritage risk assessments rely on manual surveys and field inspections, which are often time-consuming and may fail to capture evolving risks comprehensively. In contrast, the emergence of geospatial big data presents new opportunities for leveraging artificial intelligence (AI) in heritage conservation. This study introduces an innovative methodology that integrates remote sensing data and machine learning to assess risks to heritage sites. Cairo, Egypt, with its vast and historically significant urban heritage, serves as the focal point of this analysis, as its cultural heritage (CH) faces increasing threats from rapid urbanization and development pressures. The proposed framework utilizes high-resolution satellite imagery and advanced geospatial analytics to systematically evaluate and prioritize vulnerable heritage sites. The analysis encompasses 9 districts, covering 1476 heritage sites, using a novel risk assessment framework that incorporates four key components: (1) urban development pressure measured through building density at multiple radii (50 m, 100 m, 200 m, 500 m); (2) heritage vulnerability assessment based on site age, area, and cultural significance; (3) isolation risk determined by surrounding building counts; and (4) environmental risk factors including material vulnerability, natural hazard exposure, emergency response limitations, and environmental pollution. Five machine learning models were evaluated using rigorous spatial cross-validation, with building density metrics at the 50 m radius emerging as the strongest independent predictors (r = 0.160, p < 0.001). By combining OpenStreetMap data, satellite imagery, and custom algorithms with methodologically rigorous feature selection to prevent data leakage, the study generates risk scores that facilitate data-driven decision-making for heritage preservation and sustainable tourism development. This study demonstrates how geospatial AI (GeoAI) can support CH preservation, and not just disaster prediction or urban analysis. The findings reveal that heritage risk is primarily driven by immediate urban context rather than site-specific characteristics, offering guidance for policymakers, UNESCO, NGOs, and developers in resource allocation.