Dynamic evaluation and prediction of solar radiation effects on large-scale semi-open built heritage surfaces

Abstract

Large-scale semi-open heritage sites are highly vulnerable to surface deterioration driven by the interplay between intrinsic rock properties and external micro-environmental factors. Among them, solar-radiation-driven thermal stress (SRT) exerts a more prolonged impact than strong winds or heavy rainfall. This study develops a high-precision, multi-dimensional dynamic evaluation and prediction framework for large-scale assessment of SRT on heritage surfaces. The results show that: On-site monitoring revealed strong agreement between simulations and measurements in both overall patterns and local variations, confirming the framework's reliability across temporal and spatial scales. The XGB-based SRT prediction model achieved excellent accuracy (R² = 0.985, RMSE = 0.121, MAE = 0.071). Orientation and seasonality play significant roles, with annual cumulative solar radiation on southwest side surfaces exceeding that on northeast side surfaces by more than threefold, and monthly variations being especially pronounced on southeast and southwest sides. Sensitivity analysis identified solar radiation as the dominant driver of thermal stress variation, responsible for over 50 % of the variation in most cases, and for nearly 60 % on the southeast and southwest surfaces. These findings provide critical insights into the spatiotemporal dynamics of SRT on heritage surfaces and offer valuable guidance for conservation and restoration strategies.