A novel CGBoost deep learning algorithm for coseismic landslide susceptibility prediction

Abstract

The accurate prediction of landslide susceptibility shortly after a violent earthquake is quite vital to the emergency rescue in the 72-h “golden window”. However, the limited quantity of interpreted landslides shortly after a massive earthquake makes landslide susceptibility prediction become a challenge. To address this gap, this work suggests an integrated method of Crossing Graph attention network and xgBoost (CGBoost). This method contains three branches, which extract the interrelations among pixels within a slope unit, the interrelations among various slope units, and the relevance between influencing factors and landslide probability, respectively, and obtain rich and discriminative features by an adaptive fusion mechanism. Thus, the difficulty of susceptibility modeling under a small number of coseismic landslides can be reduced. As a basic module of CGBoost, the proposed Crossing graph attention network (Crossgat) could characterize the spatial heterogeneity within and among slope units to reduce the false alarm in the susceptibility results. Moreover, the rainfall dynamic factors are utilized as prediction indices to improve the susceptibility performance, and the prediction index set is established by terrain, geology, human activity, environment, meteorology, and earthquake factors. CGBoost is applied to predict landslide susceptibility in the Gorkha meizoseismal area. 3.43% of coseismic landslides are randomly selected, of which 70% are used for training, and the others for testing. In the testing set, the values of Overall Accuracy, Precision, Recall, F1-score, and Kappa coefficient of CGBoost attain 0.9800, 0.9577, 0.9999, 0.9784, and 0.9598, respectively. Validated by all the coseismic landslides, CGBoost outperforms the current major landslide susceptibility assessment methods. The suggested CGBoost can be also applied to landslide susceptibility prediction in new earthquakes in the future.

Keywords

Coseismic landslide / Landslide susceptibility prediction / Graph neural network / Deep learning

CRediT authorship contribution statement

Qiyuan Yang: Conceptualization, Methodology, Software, Validation, Formal analysis, Writing—original draft, Writing—review & editing. Xianmin Wang: Formal analysis, Methodology, Writing – original draft, Writing – review & editing. Jing Yin: Visualization, Writing – original draft. Aiheng Du: Writing – original draft. Aomei Zhang: Writing – original draft. Lizhe Wang: Writing – review & editing. Haixiang Guo: Writing – review & editing. Dongdong Li: Writing – review & editing.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

This work is funded by the National Natural Science Foundation of China (42311530065, U21A2013, 71874165), Opening Fund of Key Laboratory of Geological Survey and Evaluation of Ministry of Education (Grant Nos. GLAB2020ZR02, GLAB2022ZR02), State Key Laboratory of Biogeology and Environmental Geology (Grant No. GBL12107), the Fundamental Research Funds for the Central Universities, China University of Geosciences (Wuhan) (CUG2642022006), and Hunan Provincial Natural Science Foundation of China (2021JC0009). The authors would like to acknowledge Xuewen Wang for his suggestions in the manuscript structure. The author also would like to acknowledge the developers of GDAL, PyTorch, and DGL for their open-source projects.

References

[1] S.A. Ali, F. Parvin, Q.B. Pham, K.M. Khedher, M. Dehbozorgi, Y.W. Rabby, D.T. Anh, D.H. Nguyen. An ensemble random forest tree with svm, ann, nbt, and lmt for landslide susceptibility mapping in the rangit river watershed, india. Nat. Hazards, 113 (3) (2022), pp. 1601-1633,

[2] T.I. Allen, D.J. Wald. Evaluation of ground-motion modeling techniques for use in global shakemap — a critique of instrumental ground-motion prediction equations, peak ground motion to macroseismic intensity conversions, and macroseismic intensity predictions in different tecto. U.S. Geol. Surv. Open-File Rep., 2009 (1047) (2009), p. 122

[3] H.A.H. Al-Najjar, B. Pradhan, G. Beydoun, R. Sarkar, H. Park, A. Alamri. A novel method using explainable artificial intelligence (xai)-based shapley additive explanations for spatial landslide prediction using time-series sar dataset. Gondwana Res., 123 (2022), pp. 107-124,

[4] N.A. Asif, Y. Sarker, R.K. Chakrabortty, M.J. Ryan, M.H. Ahamed, D.K. Saha, F.R. Badal, S.K. Das, M.F. Ali, S.I. Moyeen, M.R. Islam, Z. Tasneem. Graph neural network: A comprehensive review on non-euclidean space. IEEE Access, 9 (2021), pp. 60588-60606,

[5] B. Aslam, A. Zafar, U. Khalil. Comparative analysis of multiple conventional neural networks for landslide susceptibility mapping. Nat. Hazards, 115 (8) (2022), pp. 673-707,

[6] Brody, S., Alon, U., Yahav, E., 2021. How attentive are graph attention networks? arXiv preprint arXiv: 210514491.

[7] S. Chen, Z. Miao, L. Wu, A. Zhang, Q. Li, Y. He. A one-class-classifier-based negative data generation method for rapid earthquake-induced landslide susceptibility mapping. Front. Earth Sci., 9 (4) (2021), pp. 1-19,

[8] Corso, G., Cavalleri, L., Beaini, D., Liò, P., Velickovic, P., 2020. Principal neighbourhood aggregation for graph nets. arXiv preprint arXiv: 200405718.

[9] R.K. Dahal, S. Hasegawa. Representative rainfall thresholds for landslides in the nepal himalaya. Geomorphology, 100 (3–4) (2008), pp. 429-443,

[10] A. Dahal, L. Lombardo. Explainable artificial intelligence in geoscience: A glimpse into the future of landslide susceptibility modeling. Comput. Geosci., 176 (2023), Article 105364,

[11] Dhital, M., 2015. Geology of the nepal himalaya. Springer International Publishing Switzerland, March 2015, 1-498.

[12] Z. Fang, Y. Wang, L. Peng, H. Hong. Integration of convolutional neural network and conventional machine learning classifiers for landslide susceptibility mapping. Comput. Geosci., 139 (2020), Article 104470,

[13] S.E. Fick, R.J. Hijmans. Worldclim 2: New 1-km spatial resolution climate surfaces for global land areas. Int. J. Climatol., 37 (12) (2017), pp. 4302-4315,

[14] Q. Ge, H. Sun, Z. Liu, X. Wang. A data-driven intelligent model for landslide displacement prediction. Geol. J., 58 (6) (2023), pp. 2211-2230,

[15] D. Grabowski, I. Laskowicz, A. Małka, J. Rubinkiewicz. Geoenvironmental conditioning of landsliding in river valleys of lowland regions and its significance in landslide susceptibility assessment: A case study in the lower vistula valley, northern poland. Geomorphology, 419 (2022), Article 108490,

[16] Hamilton, W.L., Ying, R., Leskovec, J., 2017. Inductive Representation Learning on Large Graphs. In Proceedings of the 31st Conference on Neural Information Processing Systems (NIPS 2017), Long Beach, CA, USA, 4–9 December 2017.

[17] I. Harris, P.D. Jones, T.J. Osborn, D.H. Lister. Updated high-resolution grids of monthly climatic observations - the cru ts3.10 dataset. Int. J. Climatol., 34 (3) (2014), pp. 623-642,

[18] Q. He, M. Wang, K. Liu. Rapidly assessing earthquake-induced landslide susceptibility on a global scale using random forest. Geomorphology, 391 (2021), Article 107889,

[19] Herath, J.D., Wakodikar, P.P., Yang, P., Yan, G., 2022. Cfgexplainer: Explaining graph neural network-based malware classification from control flow graphs. In the 52nd Annual IEEE/IFIP International Conference on Dependable Systems and Networks (DSN), Baltimore, MD, Jun 27-30, 2022, 172-184.

[20] Y. Hua, X. Wang, Y. Li, P. Xu, W. Xia. Dynamic development of landslide susceptibility based on slope unit and deep neural networks. Landslides, 18 (1) (2021), pp. 281-302,

[21] Jiang, N., Su, F., Xu, J., Ge, Y., Liu, Z., 2020. Fractal characteristics and causes of co-seismic landslides in the nepal earthquake extremely stricken areas. Mt. Res. 38(5), 699-709. doi: 10.16089 /j.cnki.1008-2786.000547.

[22] R.W. Jibson. Regression models for estimating coseismic landslide displacement. Eng. Geol., 91 (2–4) (2007), pp. 209-218,

[23] Kipf, T.N., Welling, M., 2017. Semi-supervised classification with graph convolutional networks. arXiv preprint arXiv: 160902907.

[24] Y. Liu, W. Zhang, Z. Zhang, Q. Xu, W. Li. Risk factor detection and landslide susceptibility mapping using geo-detector and random forest models: The 2018 hokkaido eastern iburi earthquake. Remote Sens., 13 (6) (2021), p. 1157,

[25] J. Lu, W. Li, W. Zhan, Y. Tie. Distribution and Mobility of Coseismic Landslides Triggered by the 2018 Hokkaido Earthquake in Japan. Remote Sens., 14 (16) (2022), p. 3957,

[26] Lundberg, S.M., Lee, S.I., 2017. A unified approach to interpreting model predictions. In the Proceedings of the 31st International Conference on Neural Information Processing Systems, Long Beach, California, USA, December 4–9, 2017, 4765-4774.

[27] S. Ma, C. Xu. Assessment of co-seismic landslide hazard using the newmark model and statistical analyses: A case study of the 2013 lushan, china, mw6.6 earthquake. Nat. Hazards, 96 (1) (2019), pp. 389-412,

[28] B. Pradhan, A. Dikshit, S. Lee, H. Kim. An explainable ai (xai) model for landslide susceptibility modeling. Appl. Soft Comput., 142 (2023), Article 110324,

[29] P. Qi. Algorithm design and simulation of optimal maritime search scheme. Proceedings of the First International Conference on Information Sciences, Machinery, Materials and Energy, 126 (2015), pp. 1787-1790,

[30] K. Roback, M.K. Clark, A.J. West, D. Zekkos, G. Li, S.F. Gallen, D. Chamlagain, J.W. Godt. The size, distribution, and mobility of landslides caused by the 2015 mw7.8 gorkha earthquake, nepal. Geomorphology, 301 (1) (2018), pp. 121-138,

[31] B. Romstad, B. Etzelmüller. Mean-curvature watersheds: A simple method for segmentation of a digital elevation model into terrain units. Geomorphology, 139–140 (2012), pp. 293-302,

[32] S. Saha, A. Arabameri, A. Saha, T. Blaschke, P.T.T. Ngo, V.H. Nhu, S.S. Band. Prediction of landslide susceptibility in rudraprayag, india using novel ensemble of conditional probability and boosted regression tree-based on cross-validation method. Sci. Total Environ., 764 (2021), Article 142928,

[33] V.O. Santos, P.A.C. Rocha, J. Scott, J.V.G. Thé, B. Gharabaghi. A new graph-based deep learning model to predict flooding with validation on a case study on the humber river. Water, 15 (10) (2023), p. 1827,

[34] R.R. Selvaraju, M. Cogswell, A. Das, R. Vedantam, D. Parikh, D. Batra. Grad-cam: Visual explanations from deep networks via gradient-based localization. Int. J. Comput. vis., 128 (2) (2020), pp. 336-359,

[35] H.L. Shrestha, M. Poudel. Landslide susceptibility zonation mapping in post-earthquake scenario in gorkha district. Forestry: Journal of Institute of Forestry, Nepal, 15 (15) (2018), pp. 45-56,

[36] R. Styron, M. Pagani. The gem global active faults database. Earthq. Spectra., 36 (1_suppl) (2020), pp. 160-180,

[37] H. Tanyas, L. Lombardo. Completeness index for earthquake-induced landslide inventories. Eng. Geol., 264 (2020), Article 105331,

[38] K. Ullah, Y. Wang, Z. Fang, M. Rahman. Multi-hazard susceptibility mapping based on convolutional neural networks. Geosci. Front., 13 (5) (2022), Article 101425,

[39] USGS, 2015. Mw 7.8 - 67 km nne of bharatpur, nepal. Earthquake hazards program, usgs. Website: Https://earthquake.Usgs.Gov/earthquakes/eventpage/us20002926/shakemap/intensity. Accessed 25 apirl 2015.

[40] Veličković, P., Casanova, A., Liò, P., Cucurull, G., Romero, A., Bengio, Y., 2018. Graph attention networks. In the Sixth International Conference on Learning Representations, Vancouver Convention Center, Vancouver CANADA, Mon Apr 30th through May 3rd, 2018, 1-12.

[41] H. Wang, L. Zhang, H. Luo, J. He, R.W.M. Cheung. Ai-powered landslide susceptibility assessment in hong kong. Eng. Geol., 288 (2021), Article 106103,

[42] L. Xiang, S. Wang, L. Wang. Response of typical hydrodynamic pressure landslide to reservoir water level fluctuation: Shuping landslide in three gorges reservoir as an example. J. Eng. Geol., 22 (5) (2014), pp. 877-881

[43] Xu, K., Jegelka, S., Hu, W., Leskovec, J., 2019. How powerful are graph neural networks? arXiv preprint arXiv:181000826.

[44] Q. Xu, X. Zhu, W. Li, X. Dong, K. Dai, Y. Jiang, H. Lu, C. Guo. Technical progress of space-air-ground collaborative monitoring of landslide. Acta Geod. Et Cartogr. Sin., 51 (7) (2022), pp. 1416-1436, 10.11947/j.AGCS.2022.20220320

[45] M. Yamada, T. Hayashida, J. Mori, W.D. Mooney. Building damage survey and microtremor measurements for the source region of the 2015 gorkha, nepal, earthquake. Earth Planets Space, 68 (1) (2016), pp. 1-8,

[46] Ying, R., Bourgeois, D., You, J., Zitnik, M, Leskovec, J., 2019. Gnnexplainer: Generating explanations for graph neural networks. In the 33rd Conference on Neural Information Processing Systems (NeurIPS 2019), Vancouver, BC, Canada, December 8-14, 2019, 9240–9251.

[47] H.W. Zeng, Q. Zhu, Y.L. Ding, H. Hu, L. Chen, X. Xie, M. Chen, Y.X. Yao. Graph neural networks with constraints of environmental consistency for landslide susceptibility evaluation. Int. J. Geogr. Inf. Sci., 36 (11) (2022), pp. 2270-2295,

[48] Z. Zhang, J.-A. Fleurisson, F. Pellet. The effects of slope topography on acceleration amplification and interaction between slope topography and seismic input motion. Soil Dyn. Earthq. Eng., 113 (2018), pp. 420-431,

[49] Zhdanov, M., Steinmann, S., Hoffmann, N., 2022. Investigating brain connectivity with graph neural networks and gnnexplainer. In the 26th International Conference on Pattern Recognition / 8th International Workshop on Image Mining - Theory and Applications (IMTA), Montreal, CANADA, Aug 21-25, 2022, 5155-5161.

[50] X. Zhou, H. Wen, Z. Li, H. Zhang, W. Zhang. An interpretable model for the susceptibility of rainfall-induced shallow landslides based on shap and xgboost. Geocarto Int., 37 (26) (2022), pp. 13419-13450,