A novel rockfall hazard assessment using laser scanning data and 3D modelling in GIS

• Published in: Catena (Giessen) Vol. 172; pp. 435 - 450
• Main Authors: Fanos, Ali Mutar, Pradhan, Biswajeet
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.01.2019
• Subjects: 3D modelling; AHP; algorithms; artificial intelligence; geographic information systems; GIS; Hazard; hazard characterization; kinetic energy; landscapes; LiDAR; Malaysia; risk; Rockfall; rockfalls; transportation; urban areas; Malaysia; GIS; AHP; 3D modelling; Hazard; Rockfall; LiDAR
• ISSN: 0341-8162; 1872-6887
• DOI: 10.1016/j.catena.2018.09.012

Rockfall hazards occur widely in regions with steep terrain such as Kinta Valley, Malaysia. Rockfalls threaten urban areas and the transportation corridors that pass through such areas. This paper proposes a comprehensive rockfall hazard assessment strategy based on high-resolution laser scanning data (LiDAR), both airborne and terrestrial. It provides (1) rockfall source identification by developing a hybrid model based on a bagging neural network (BBNN), which is compared with various machine learning algorithms and ensemble models (bagging, boosting, voting) and a Gaussian mixture model; (2) 3D modelling of rockfall kinematic processes (trajectory distribution, frequency, velocity, kinetic energy, bounce height, impact location); and (3) hazard zonation based on spatial modelling in combination with an analytical hierarchy process (AHP) in a geographic information system (GIS). In addition, mitigation measures are suggested based on the modelling results. The proposed methodology was validated in three study areas to test the applicability and generalisability of the methods. The results show that the proposed hybrid model can accurately identify rockfall source areas at the regional scale. It achieved a 97% training accuracy and 5-fold cross-validation area under curve (AUC) value of 0.96. The mechanical parameters of the developed 3D model were calibrated with an accuracy of 97%, 93% and 95% for Gunung Lang, Gua Tambun and Gunung Rapat areas, respectively. In addition, the proposed spatial model effectively delineates areas at risk of rockfalls. This method provides a comprehensive understanding of rockfall hazards that can assist authorities to develop proper management and protection of urban areas and transportation corridors. [Display omitted] •Airborne and terrestrial laser data were employed to derive high-resolution DTM.•Gaussian Mixture Model (GMM) was used to calculate slope angle thresholds.•Different machine learning algorithms were applied and tested.•A hybrid model was proposed to identify the source of rockfall.•3D rockfall modelling was performed to derive rockfall trajectories.