Comparative landslide risk assessment of the Beijing-Guangzhou railway in south China using real probability and automated machine learning

Abstract

This study presents an integrated framework combining a real probability-based approach with automated machine learning (AutoML) to quantitatively assess the landslide risk of the southern segment of the Beijing–Guangzhou Railway under heavy rainfall. A total of 6660 landslides triggered by a single rainfall event were analyzed, incorporating twelve key influencing factors, including accumulated rainfall, topography, and geological conditions. A sample database was constructed based on the real probability criterion, and three models, namely XGBoost, Random Forest (RF), and Extra Trees (ET), were efficiently optimized using the FLAML framework. Among them, XGBoost yielded the best performance, achieving the highest AUC and demonstrating the strongest spatial agreement between predicted high-risk zones and observed landslide occurrences. Feature importance analysis identified accumulated rainfall, strata, and elevation as the dominant contributing factors. The results reveal that the overall risk of the railway section during this event was primarily classified as extremely low or very low. Notably, areas experiencing intense precipitation still pose significant risks to railway operations due to potential rapid-onset landslides. This research provides a data-driven, scalable tool for quantitative disaster risk assessment and offers a foundation for the development of a dynamic early warning platform by incorporating real-time rainfall forecasts.

Graphical abstract

Keywords

Automated machine learning / Risk assessment / rainfall-induced landslides / Hazard assessment / Beijing-Guangzhou railway

References

[1] Ahlgren P, Jarneving B, Rousseau R (2003). Requirements for a cocitation similarity measure, with special reference to Pearson’s correlation coefficient.J Am Soc Inf Sci Technol, 54(6): 550–560

[2] Ali Y A, Awwad E M, Al-Razgan M, Maarouf A (2023). Hyperparameter search for machine learning algorithms for optimizing the computational complexity.Processes, 11(2): 349

[3] Areerob K, Nguyen V Q, Li X F, Inadomi S, Shimada T, Kanasaki H, Wang Z J, Suganuma M, Nagatani K, Chun P J, Okatani T (2025). Multimodal artificial intelligence approaches using large language models for expert-level landslide image analysis.Comput-Aided Civil Inf Eng, 40(19): 2900–2921

[4] Breiman L (2001). Random forests.Mach Learn, 45(1): 5–32

[5] Bennett G L, Panici D, Rengers F K, Kean J W, Rathburn S L (2025). Landslide-channel feedbacks amplify channel widening during floods.npj Nat Hazards, 2(1): 7

[6] Chen T, He T, Benesty M, Khotilovich V, Tang Y, Cho H, Chen K, Mitchell R, Cano I, Zhou T (2015). Xgboost: extreme gradient boosting.R Package Version 0.4–2, 1(4): 1–4

[7] Chen Z, Song D Q, Song H Q, Guo Z Z, Juliev M (2023). Editorial: large landslides in the Sichuan-Tibet railway: recognition, mechanism, and mitigation.Front Earth Sci, 11: 1170789

[8] Elsagheer Mohamed S A, Alshalfan K A, Al-Hagery M A, Ben Othman M T (2022). Safe driving distance and speed for collision avoidance in connected vehicles.Sensors, 22(18): 7051

[9] Geurts P, Ernst D, Wehenkel L (2006). Extremely randomized trees.Mach Learn, 63(1): 3–42

[10] Gariano S L, Guzzetti F (2016). Landslides in a changing climate.Earth-Sci Rev, 162: 227–252

[11] Gao H R, Xu C, Xie C C, Ma J X, Xiao Z K (2024). Landslides triggered by the July 2023 extreme rainstorm in the Haihe River Basin, China.Landslides, 21(11): 2885–2890

[12] Gu T F, Duan P, Wang M G, Li J, Zhang Y K (2024). Effects of non-landslide sampling strategies on machine learning models in landslide susceptibility mapping.Sci Rep, 14(1): 7201

[13] Huang Y, Cheng H L (2013). The impact of climate change on coastal geological disasters in southeastern China.Nat Hazards, 65(1): 377–390

[14] Huang F M, Yao C, Liu W P, Li Y J, Liu X W (). Landslide susceptibility assessment in the Nantian area of China: a comparison of frequency ratio model and support vector machine.Geomatics, Nat Hazards Risk, 9(1): 919–938

[15] Huang F M, Cao Z S, Guo J F, Jiang S H, Li S, Guo Z Z (2020). Comparisons of heuristic, general statistical and machine learning models for landslide susceptibility prediction and mapping.CATENA, 191: 104580

[16] Hao Y, Liu C H, Zhang W C, Liu X, Liu G H (2023). Landslide risk evaluation: Rainfall and blast-induced potential soil landsides in an expressway area underneath a railway tunnel, Guangzhou, China.Bull Eng Geol Environ, 82(11): 420

[17] Hao Y N, Qi J, Ma X W, Wu S S, Liu R Y, Zhang X Y (2024). An LLM-based inventory construction framework of urban ground collapse events with spatiotemporal locations.ISPRS Int J Geo-Inf, 13(4): 133

[18] Huang Y D, Xu C, He X L, Cheng J, Huang Y, Wu L Z, Xu X W (2024). Distribution characteristics and cumulative effects of landslides triggered by multiple moderate-magnitude earthquakes: a case study of the comprehensive seismic impact area in Yibin, Sichuan, China.Landslides, 21(12): 2927–2943

[19] Huang Y D, Xu C, He X L, Cheng J, Xu X W, Tian Y Y (2025). Landslides induced by the 2023 Jishishan Ms6.2 earthquake (NW China): spatial distribution characteristics and implication for the seismogenic fault.npj Nat Hazards, 2(1): 14

[20] Jiang W J, Li L, Niu R Q (2025). Impact of non-landslide sample sampling strategies and model selection on landslide susceptibility mapping.Appl Sci, 15(4): 2132

[21] Khan Z H, Altamimi A B (2023). A new traffic system on driver sensitivity and safe distance headway.Appl Sci, 13(20): 11262

[22] Liu K, Wang M, Cao Y X, Zhu W H, Yang G L (2018). Susceptibility of existing and planned Chinese railway system subjected to rainfall-induced multi-hazards.Transp Res Part A: Policy Pract, 117: 214–226

[23] Li R, Huang S Y, Dou H Q (2023). Dynamic risk assessment of landslide hazard for large-scale photovoltaic power plants under extreme rainfall conditions.Water, 15(15): 2832

[24] Liu L L, Zhao S L, Yang C, Zhang W G (2024). Quantifying uncertainty in landslide susceptibility mapping due to sampling randomness.Int J Dis Risk Reduct, 114: 104966

[25] Lin Y C, Hsieh J Y, Shih H S, Wang W H (2025). Strong wind is one of the important factors that trigger landslides.npj Nat Hazards, 2(1): 12

[26] Ma J W, Lei D Z, Ren Z Y, Tan C H, Xia D, Guo H X (2024). Automated machine learning-based landslide susceptibility mapping for the three gorges reservoir area, China.Math Geosci, 56(5): 975–1010

[27] Probst P, Boulesteix A L, Bischl B (2019). Tunability: importance of hyperparameters of machine learning algorithms.J Mach Learn Res, 20(1): 1934–1965

[28] Sassa K, Wang G H, Fukuoka H, Wang F W, Ochiai T, Sugiyama M, Sekiguchi T (2004). Landslide risk evaluation and hazard zoning for rapid and long-travel landslides in urban development areas.Landslides, 1(3): 221–235

[29] Sedgwick P (2012). Pearson’s correlation coefficient.BMJ, 345: e4483

[30] Schratz P, Muenchow J, Iturritxa E, Richter J, Brenning A (2019). Hyperparameter tuning and performance assessment of statistical and machine-learning algorithms using spatial data.Ecol Modell, 406: 109–120

[31] Shao X Y, Ma S Y, Xu C, Zhou Q (2020). Effects of sampling intensity and non-slide/slide sample ratio on the occurrence probability of coseismic landslides.Geomorphology, 363: 107222

[32] Sun D L, Wen H J, Wang D Z, Xu J H (2020). A random forest model of landslide susceptibility mapping based on hyperparameter optimization using Bayes algorithm.Geomorphology, 362: 107201

[33] Sun S Y, Hu J B, Wang R H (2021). Correlation between visibility and traffic safety visual distance in foggy areas during the daytime.Traffic Inj Prev, 22(7): 514–518

[34] Shao X Y, Ma S Y, Xu C (2023a). Hazard assessment modeling and software development of earthquake-triggered landslides in the Sichuan-Yunnan area, China.Geosci Model Dev, 16(17): 5113–5129

[35] Shao X Y, Ma S Y, Xu C, Cheng J, Xu X W (2023b). Seismically-induced landslide probabilistic hazard mapping of Aba Prefecture and Chengdu Plain region, Sichuan Province, China for future seismic scenarios.Geosci Lett, 10(1): 55

[36] Shao X Y, Ma S Y, Xu C, Xie C C, Li T, Huang Y D, Huang Y, Xiao Z K (2024). Landslides triggered by the 2022 Ms. 6.8 Luding strike-slip earthquake: an update.Eng Geol, 335: 107536

[37] Sahin E K (2025). FastLSM-AutoML: fast, reliable, and robust end-to-end AutoML tool for producing a landslide susceptibility map.Stoch Environ Res Risk Assess, 39(10): 4801–4825

[38] Tateishi R, Uriyangqai B, Al-Bilbisi H, Ghar M A, Tsend-Ayush J, Kobayashi T, Kasimu A, Hoan N T, Shalaby A, Alsaaideh B, Enkhzaya T, Gegentana N, Sato H P (2011). Production of global land cover data-GLCNMO.Int J Digital Earth, 4(1): 22–49

[39] Tehrani F S, Calvello M, Liu Z Q, Zhang L M, Lacasse S (2022). Machine learning and landslide studies: recent advances and applications.Nat Hazards, 114(2): 1197–1245

[40] Tang G X, Fang Z C, Wang Y (2023). Global landslide susceptibility prediction based on the automated machine learning (AutoML) framework.Geocarto Int, 38(1): 2236576

[41] Wang S, Li L P, Shi S S, Cheng S, Hu H J, Wen T (2020). Dynamic risk assessment method of collapse in mountain tunnels and application.Geotech Geol Eng, 38(3): 2913–2926

[42] Wang C, Wu Q Y, Weimer M, Zhu E K (2021). FLAML: a fast and lightweight automl library. In: Proceedings of the 4th Conference on Machine Learning and Systems. mlsys. org, 434–447

[43] Xu C, Dai F C, Xu X W, Lee Y H (2012a). (). GIS-based support vector machine modeling of earthquake-triggered landslide susceptibility in the Jianjiang River watershed, China..Geomorphology, 145–146: 70–80

[44] Xu C, Xu X W, Dai F C, Saraf A K (2012b). Comparison of different models for susceptibility mapping of earthquake triggered landslides related with the 2008 Wenchuan earthquake in China.Comput Geosci, 46: 317–329

[45] Xu C, Xu X W, Zhou B G, Shen L L (2019). Probability of coseismic landslides: a new generation of earthquake-triggered landslide hazard model.J Eng Geol, 27(5): 1122–1130

[46] Xie C C, Gao H R, Huang Y D, Xue Z W, Xu C, Dai K B (2025a). Leveraging the DeepSeek large model: a framework for AI-assisted disaster prevention, mitigation, and emergency response systems.Earthquake Res Adv, 5(4): 100378

[47] Xie C C, Xu C, Huang Y D, Liu J L, Jin J L, Xu X W, Cheng J, Wu L Z (2025b). Detailed inventory and initial analysis of landslides triggered by extreme rainfall in the northern Huaiji County, Guangdong Province, China, from June 6 to 9, 2020.Geoenviron Disasters, 12(1): 7

[48] Xie C C, Xu C, Huang Y D, Liu J L, Shao X Y, Xu X W, Gao H R, Ma J X, Xiao Z K (2025c). Advances in the study of natural disasters induced by the “23.7” extreme rainfall event in North China.Nat Hazards Res, 5(1): 1–13

[49] Xie C C, Xu C, Xu X W, Huang Y D, Zhang Z Q, Li H, Zhang H R, Zhu D J (2025d). Comprehensive analysis and assessment of extreme rainfall-induced clustered landslides: a case study of southern Qingyuan city, Guangdong province, China, in June 2020. Georisk: Assess Manag Risk Eng Syst Geohazards, 1–23

[50] Xu C, Lin N (2025). Building a global forum for natural hazard science.npj Nat Hazards, 2(1): 72

[51] Yin L Z, Zhu J, Li W S, Wang J H (2022). Vulnerability analysis of geographical railway network under geological hazard in China.ISPRS Int J Geo-Inf, 11(6): 342

[52] Yang Z Q, Qi W W, Xu C, Shao X Y (2024). Exploring deep learning for landslide mapping: a comprehensive review.China Geol, 7(2): 330–350

[53] Zhao J, Liu K, Wang M (2020). Exposure analysis of Chinese railways to multihazards based on datasets from 2000 to 2016. Geomatics, Nat Hazards Risk, 11(1): 272–287

[54] Zhou X Z, Wen H J, Li Z W, Zhang H, Zhang W G (2022). An interpretable model for the susceptibility of rainfall-induced shallow landslides based on SHAP and XGBoost.Geocarto Int, 37(26): 13419–13450

[55] Zhang S, Dai C, Wang S R, Wang B J, Zhou Y L, Chen Y B, Bate B, Zhan L T, Chen Y M, Tang Y (2023). A catastrophic high-speed train crash caused by a small-scale debris flow in China.Transp Geotech, 42: 101070

[56] Zhu W H, Liu K, Wang S D, Wang M, Liu S L (2024). Estimating the functional impacts associated with rainfall-induced hazards for Chinese railway system: fragility, risk and its uncertainty.Nat Hazards, 120(11): 9585–9607

Rights & permissions

Higher Education Press