Explainable machine learning links erosion damage to environmental factors on Gansu rammed earth Great Wall
Source abstract
Abstract
The Great Wall, the world’s largest linear cultural heritage, faces rapid erosion due to natural conditions. Erosion damage starting from the wall’s base can lead to structural instability and collapse, posing the most destructive threat. This study conducts the first digital survey of erosion damage to the Ming Great Wall in Gansu using drone. Erosion damage was quantified through kernel density and spatial autocorrelation analysis. An optimal XGBoost machine learning model was developed to assess the contributions of 14 environmental factors. Combining SHAP value decomposition and partial dependence plots (PDP) revealed nonlinear interactions among these factors. Significant spatial aggregation of erosion damage was observed with salinization, topographic fluctuation, and annual average precipitation identified as primary influencing factors. Each factor exhibited either promoting or inhibiting effect. This study enhances understanding of the spatial dynamics driving mechanism of erosion on the Great Wall, providing a scientific basis for its conservation and restoration.
Source: openalex_inverted_index · Quality: baseline
Research context
OpenAlex topics: Urban Heat Island Mitigation · Hygrothermal properties of building materials · Building materials and conservation
SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.
References (66)
Show source-provided references
1. Administration, N.C.H., The Great Wall Resource Survey Report. 2016.
2. Wang, C., Shi, Y. & Liu, K. Study on typical diseases of soil relics and their correlation in arid and semi-arid areas of Northwest China. J. Earthq. Eng. 45, 220–227 (2023).
3. Elert, K., Pardo, E. S. & Rodriguez-Navarro, C. Alkaline activation as an alternative method for the consolidation of earthen architecture. J. Cult. Herit. 16, 461–469 (2015).10.1016/j.culher.2014.09.012 ↗
4. SUN, M. et al. Classification of deteriorations associated with many earthen heritage sites in arid areas of northwest China. J. Eng. Geol. 15, 772–778 (2007).
5. Chen, W. et al. Preventing the secondary salt deterioration in the repaired area of basal erosion of earthen sites based on capillary barrier effect. Constr. Build. Mater. 366, 130168 (2023).10.1016/j.conbuildmat.2022.130168 ↗
6. Chen, H. et al. The effect of SH-lime composite material on capillary water rise resistance in the Great Wall. J. Build. Eng. 85, 108512 (2024).10.1016/j.jobe.2024.108512 ↗
7. Fan, W. et al. Feasibility of protecting earthen sites with sticky rice and lime composite. Constr. Build. Mater. 346, 128449 (2022).10.1016/j.conbuildmat.2022.128449 ↗
8. Han, W. et al. Influence mechanism of salt erosion on the earthen heritage site wall in Pianguan Bastion. Case Stud. Constr. Mater. 17, e01388 (2022).
9. Mosaid, H. et al. Improved soil carbon stock spatial prediction in a Mediterranean soil erosion site through robust machine learning techniques. Environ. Monit. Assess. 196, 130 (2024).10.1007/s10661-024-12294-x ↗
10. Chi, Y. et al. The spatial distribution mechanism of PM2.5 and NO2 on the eastern coast of China. Environ. Pollut. 342, 123122 (2024).10.1016/j.envpol.2023.123122 ↗
11. Wang, M. et al. Data-driven approach to spatiotemporal dynamic risk assessment of urban flooding based on shared socio-economic pathways. Ecol. Indic. 154, 110764 (2023).10.1016/j.ecolind.2023.110764 ↗
12. Guo, Y. et al. Construction of rapid early warning and comprehensive analysis models for urban waterlogging based on AutoML and comparison of the other three machine learning algorithms. J. Hydrol. 605, 127367 (2022).10.1016/j.jhydrol.2021.127367 ↗
13. Ren, Z. et al. A large-scale riverbank erosion risk assessment model integrating multi-source data and explainable artificial intelligence (XAI). Ecol. Indic. 166, 112575 (2024).10.1016/j.ecolind.2024.112575 ↗
14. Zhou, S. et al. Synergistic assessment of multi-scenario urban waterlogging through data-driven decoupling analysis in high-density urban areas: a case study in Shenzhen, China. J. Environ. Manag. 369, 122330 (2024).10.1016/j.jenvman.2024.122330 ↗
15. Kanani-Sadat, Y. et al. A novel explainable PSO-XGBoost model for regional flood frequency analysis at a national scale: exploring spatial heterogeneity in flood drivers. J. Hydrol. 638, 131493 (2024).10.1016/j.jhydrol.2024.131493 ↗
16. Liang, C. et al. Prediction and analysis of atmospheric visibility in five terrain types with artificial intelligence. Heliyon 9, e19281 (2023).10.1016/j.heliyon.2023.e19281 ↗
17. Mangalathu, S., Hwang, S. & Jeon, J. Failure mode and effects analysis of RC members based on machine-learning-based SHapley Additive exPlanations (SHAP) approach. Eng. Struct. 219, 110927 (2020).10.1016/j.engstruct.2020.110927 ↗
18. Hu, Y. et al. Pixel level spatial variability modeling using SHAP reveals the relative importance of factors influencing LST. Environ. Monit. Assess. 195, 407 (2023).10.1007/s10661-023-10950-2 ↗
19. Wang, M. et al. An XGBoost-SHAP approach to quantifying morphological impact on urban flooding susceptibility. Ecol. Indic. 156, 111137 (2023).10.1016/j.ecolind.2023.111137 ↗
20. Jaquin, P. A. et al. The strength of unstabilised rammed earth materials. Géotechnique 59, 487–490 (2009).10.1680/geot.2007.00129 ↗
21. Cui, K. et al. Degradation mechanism of soil strength under the coupling effects of wetting-drying and salinization. Journal of Lanzhou University (Natural Science Edition).J. Lanzhou Univ. (Nat. Sci. Edn.) 53, 582–587 (2017).
22. Chen, W., Jia, B. & Cai, T. Study on freeze-thaw degradation of salt-affected soil sites under snowmelt and rainfall infiltration. J. Geotech. Eng. 44, 334–342 (2022).
23. Du, Y. et al. A model characterizing deterioration at earthen sites of the Ming Great Wall in Qinghai Province, China. Soil Mech. Found. Eng. 53, 426–434 (2017).10.1007/s11204-017-9423-y ↗
24. Ma, D. et al. Effect of wetting-drying cycle on hydraulic and mechanical properties of cemented paste backfill of the recycled solid wastes. Chemosphere 282, 131163 (2021).10.1016/j.chemosphere.2021.131163 ↗
25. Soudani, L. et al. Energy evaluation of rammed earth walls using long term in-situ measurements. Sol. Energy 141, 70–80 (2017).10.1016/j.solener.2016.11.002 ↗
26. Liu, H. et al. Dynamics of global dryland vegetation were more sensitive to soil moisture: Evidence from multiple vegetation indices. Agric. For. Meteorol. 331, 109327 (2023).10.1016/j.agrformet.2023.109327 ↗
27. Wang, J. et al. Analysis of soil erosion characteristics in small watershed of the loess tableland Plateau of China. Ecol. Indic. 137, 108765 (2022).10.1016/j.ecolind.2022.108765 ↗
28. DENG, Y. et al. Effects of land uses on soil physic-chemical properties and erodibility in collapsing-gully alluvial fan of Anxi County, China. J. Integr. Agricult. 15, 1863–1873 (2016).10.1016/S2095-3119(15)61223-0 ↗
29. Wu, L., Liu, X. & Ma, X. Spatiotemporal distribution of rainfall erosivity in the Yanhe River watershed of hilly and gully region, Chinese Loess Plateau. Environ. Earth Sci. 75, 315 (2016).10.1007/s12665-015-5136-6 ↗
30. Duan, Z. et al. Characteristics and mechanisms of soil structure damage under salt weathering. Soil Tillage Res. 238, 106030 (2024).10.1016/j.still.2024.106030 ↗
31. Pu, T. et al. Snowfall-related deterioration behavior of the Ming Great Wall in the eastern Qinghai-Tibet Plateau. Nat. Hazards 84, 1539–1550 (2016).10.1007/s11069-016-2497-4 ↗
32. Chen, X., Zhang, N. & Wu, F. Impacts of rainfall and land use on urban surface runoff: a case study of area surrounding the North Moat in Beijing, China. J. Nat. Resour. 29, 1391–1402 (2014).
33. Chi, Y. et al. Quantification of uncertainty in short-term tropospheric column density risks for a wide range of carbon monoxide. J. Environ. Manag. 370, 122725 (2024).10.1016/j.jenvman.2024.122725 ↗
34. Wang, M. et al. Service facilities in heritage tourism: identification and planning based on space syntax. Information (Basel) 12, 504 (2021).10.3390/info12120504 ↗
35. Li, C. et al. Identifying factors influencing the spatial distribution of minority cultural heritage in Southwest China. Herit. Sci. 12, 117 (2024).10.1186/s40494-024-01230-4 ↗
36. Kontorovich, A. & Kpotufe, S. Nearest-neighbor methods: a modern perspective. in Machine Learning for Data Science Handbook: Data Mining and Knowledge Discovery Handbook, L. Rokach, O. Maimon and E. Shmueli, L. Rokach, O. Maimon and E. Shmueli, Editors. 2023, Springer International Publishing: Cham. 75-92.10.1007/978-3-031-24628-9_5 ↗
37. Lyons, M. B. et al. A comparison of resampling methods for remote sensing classification and accuracy assessment. Remote Sens. Environ. 208, 145–153 (2018).10.1016/j.rse.2018.02.026 ↗
38. Alkharusi, H. Categorical variables in regression analysis: a comparison of dummy and effect coding. Int. J. Educ. 4, 202–210 (2012).10.5296/ije.v4i2.1962 ↗
39. Lones, M. A. Avoiding common machine learning pitfalls. Patterns 5, 101046 (2024).10.1016/j.patter.2024.101046 ↗
40. Kong, F. et al. Spatio-temporal evolution of water erosion in the western Songnen Plain: analysis of its response to land use dynamics and climate change. Soil Tillage Res. 245, 106299 (2025).10.1016/j.still.2024.106299 ↗
41. Shu, H. et al. The coupling effects of wet-dry and freeze–thaw cycles on the mechanical properties of saline soil synergistically solidified with sulfur-free lignin, basalt fiber and hydrophobic polymer. CATENA 238, 107832 (2024).10.1016/j.catena.2024.107832 ↗
42. Jorenush, M. H. & Sepaskhah, A. R. Modelling capillary rise and soil salinity for shallow saline water table under irrigated and non-irrigated conditions. Agric. Water Manag. 61, 125–141 (2003).10.1016/S0378-3774(02)00176-2 ↗
43. Monger, C. et al. Soil classification as a tool for contributing to sustainability at the landscape scale and forecasting impacts of management practices in agriculture and forestry. Soil Tillage Res. 244, 106216 (2024).10.1016/j.still.2024.106216 ↗
44. Liu, B. et al. Mechanisms and feedbacks for evapotranspiration-induced salt accumulation and precipitation in an arid wetland of China. J. Hydrol. 568, 403–415 (2019).10.1016/j.jhydrol.2018.11.004 ↗
45. Mao, W. et al. Modelling the salt accumulation and leaching processes in arid agricultural areas with a new mass balance model. J. Hydrol. 591, 125329 (2020).10.1016/j.jhydrol.2020.125329 ↗
46. Hollander, M., Wolfe, D. A., & Chicken, E. Density Estimation. 2015. 609-628.10.1002/9781119196037.ch12 ↗
47. Wu, X. & Huang, X. Screening of urban environmental vulnerability indicators based on coefficient of variation and anti-image correlation matrix method. Ecol. Indic. 150, 110196 (2023).10.1016/j.ecolind.2023.110196 ↗
48. Hemphill, J. Interpreting the magnitude of correlation coefficients. Am. Psychol. 58, 78–9 (2003).10.1037/0003-066X.58.1.78 ↗
49. Wu, Z. et al. Nonlinear forces in urban thermal environment using Bayesian optimization-based ensemble learning. Sci. Total Environ. 838, 156348 (2022).10.1016/j.scitotenv.2022.156348 ↗
50. Li, Z. et al. Investigating the nonlinear relationship between surface solar radiation and its influencing factors in North China Plain using interpretable machine learning. Atmos. Res. 280, 106406 (2022).10.1016/j.atmosres.2022.106406 ↗
51. Tang, H., Shi, Y. & Liu, K. Study on the damage characteristics of the Ming Great Wall Ruins along the Silk Road. J. Earthq. Eng. 40, 1278–1286 (2018).
52. Sun, M. & Liu, J. Spatial distribution characteristics of animal diseases in the Beacon towers of the Great Wall in Gansu based on spatial analysis. J. Lanzhou Univ. (Nat. Sci. Ed.) 57, 783–790 (2021).
53. JOHN, K. et al. Using machine learning algorithms to estimate soil organic carbon variability with environmental variables and soil nutrient indicators in an alluvial soil. in Land, (2020).10.3390/land9120487 ↗
54. Bui, Q. B. et al. Durability of rammed earth walls exposed for 20 years to natural weathering. Build. Environ. 44, 912–919 (2009).10.1016/j.buildenv.2008.07.001 ↗
55. Zhang, Q. et al. Salt distribution of earthen heritage site wall and its mechanism in northern China. J. Build. Eng. 76, 107154 (2023).10.1016/j.jobe.2023.107154 ↗
56. Li, J. et al. Cracks evolution and micro mechanism of compacted clay under wet-dry cycles and wet-dry-freeze-thaw cycles. Cold Reg. Sci. Technol. 214, 103944 (2023).10.1016/j.coldregions.2023.103944 ↗
57. Lu, N. & Likos, W. J. Rate of capillary rise in soil. J. Geotech. Geoenviron. Eng. 130, 646–650 (2004).10.1061/(ASCE)1090-0241(2004)130:6(646) ↗
58. Wang, C., Zhang, P., & Xie, C. Study on the effect of different concentrations of saline solutions and salinized soil on capillary water rise. Water Sav. Irrigation 26–28 (2014).
59. Li, L. et al. Preservation of earthen heritage sites on the Silk Road, northwest China from the impact of the environment. Environ. Earth Sci. 64, 1625–1639 (2011).10.1007/s12665-010-0829-3 ↗
60. Cui, K. et al. Effects of salinized deterioration and aeolian ullage on soils in undercutting areas of earthen ruins in arid regions (III): capillary process. Chin. J. Geotech. Eng. 44, 2043–2051 (2022).
61. Chen, W., Wang, K. & Qi, Q. Salinity migration and distribution patterns in northwest soil relics under alternating precipitation and evaporation. J. Geo-Eng. 45, 2209–2220 (2024).
62. Dong, J., Lv, H. & Xu, G. Water state of soil experienced wetting–drying cycles. Indian Geotech. J. 50, 1041–1047 (2020).10.1007/s40098-020-00429-z ↗
63. Rokach, L. & Maimon, O. Decision Trees. 165–192 (2005).10.1007/0-387-25465-X_9 ↗
64. Breiman, L. Random forests. Mach. Learn. 45, 5–32 (2001).10.1023/A:1010933404324 ↗
65. Ke, G. et al. LightGBM: a highly efficient gradient boosting decision tree. in Neural Information Processing Systems (2017).
66. Chen, T. & Guestrin, C. XGBoost: a scalable tree boosting system. in Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. 2016. San Francisco, California, USA: Association for Computing Machinery10.1145/2939672.2939785 ↗