Soil-based heritages are vital to world cultural heritage, reflecting ancient human activities and the dynamic coupling between historical sediment and contemporary landscape functions (Zhang et al., 2012). Along with climate change, global soil-based heritage sites are exposed to more significant risks (Machat, C., & Ziesemer, J., 2017). Regarding biological factors, vegetation exhibits an annual 25 % increase in underground biomass, creating conflicts between tree root systems and their displacement of soil structures (Grabosky & Gucunski, 2019), which has led to the degradation of the apparent morphology of the soil-based heritages. Nevertheless, trees also provide rainfall interception and soil stabilization (Xiao et al., 2024). The dual impact mechanism and the assessment methodology of trees form a critical foundation for heritage vegetation management, representing a current research gap.
In this study, a novel estimation model for root-system states is developed from a digital perspective using airborne LiDAR and ground-penetrating radar. Above-ground and below-ground components of vegetation on soil-based heritages are linked through pixel clustering, feature extraction, and deep-learning techniques.
We found that: (i) Model outputs match field surveys; (ii) Spatial heterogeneity persists, with the southeast most disturbed by root pressure, underscoring urgent control; (iii) Selective thinning of dominant trees while retaining moderate canopy gaps best lowers risk.
This noval framework offers a new lens for soil-heritage protection, fills the dual-impact assessment gap, and supports fine-scale, multi-source management. Practically, this study informs policy for heritage management and provides planning suggestions aligning conservation with national strategies for Yangzhou ancient city ruins.