In recent years, countries and regions affected by major natural disasters have been increasingly reported, the occurrences of which have led to a great number of casualties and serious economic losses. In 2008, a devastating earthquake of magnitude 8.0 occurred in Wenchuan, Sichuan Province, China. This triggered a series of disaster chains such as an earthquake–landslides–reservoir disaster chain and an earthquake–landslide–debris flow disaster chain [1
]. The total number of casualties was 69,225, and approximately one quarter of those casualties related to secondary landslides caused by the earthquake [3
]. In 2013, a disastrous earthquake of magnitude 7.0 took place in Lushan, Sichuan Province, China. The event induced many landslides and caused a significant number of casualties [4
]. In 2017, an earthquake of magnitude 7.0 occurred in Sichuan Province, China. The earthquake triggered numerous geological disasters such as landslides and debris flows, which severely destroyed the natural landscape of the Sparkling Lake and the Panda Lake Waterfall [5
]. Frequent disaster events have shown that for most major disasters, secondary disasters are always induced; ultimately combining as a considerable destructive power. The phenomenon of secondary disasters being caused by some kind of primary disaster is considered to be a disaster chain [7
]. More importantly, the casualties and property damage related to disaster chains are deemed to be greater than those resulting from the primary source disasters themselves [8
]. Therefore, disaster chain risk assessment has become one of the urgent core issues to be addressed in current international research.
There are currently three disaster chain risk assessment methods. The first is the probabilistic analysis method based on data. Examples include that presented by Gupta [9
], which involves a probabilistic risk assessment method for structural systems based on the Bayesian network framework under multiple hazards, and that presented by Korswagen [10
], who explored a probabilistic risk assessment framework toward the structural damage of masonry housing induced by earthquakes and secondary floods. The second method is based on complex networks, for example, presented by Zheng [11
], who adopted a complex network model to build a natural hazard network to research disaster chain mechanisms. Liu [12
] also put forward a disaster chain risk evaluation model on the basis of complex networks. The third method is based on remote sensing, such as that undertaken by Meena [13
], who applied remote sensing and geographic information systems (GIS) to map earthquake-induced landslide susceptibility zones. A further example is that of Sharma [14
], who used space remote sensing techniques to analyze time and space satellite images of landslide induced by earthquake. Therefore, earthquake disaster chain risk evaluation, on the basis of natural hazard risk formation theory, is currently rare.
Many different assessment models can be used for disaster hazard and risk research. Artificial neural network models, logistic regression models, and Bayesian network (BN) models are often used for probability analysis of earthquake-induced secondary disasters [15
]. Finite element modeling, pseudo-static analysis, and the Newmark model are frequently applied to depict the slope behavior of disaster events during earthquakes [18
]. Among them, a BN model provides a useful way to deal with complicated problems as it can combine probabilistic inference methods with a graphical representation that reveals causal relationships between different network nodes, and thus offers a network structure for handling uncertainty and complexity [21
]. Song [17
] proposed a hybrid method based on a BN to assess the susceptibility of an earthquake-induced landslide. Vogel [24
] applied the various learning algorithm of the BN to assess the hazards of earthquakes, floods, and landslides. Ozdemir [25
] analyzed the relationships between landslide distributions with 19 landslide related parameters by a Bayesian model. The Newmark model has been used more widely than pseudo-static analysis and finite element modeling in specific slope analyses [26
], and earthquake-induced secondary disaster analyses [28
] are considered to yield much more helpful information. Caccavale [30
] proposed an integrated approach to assess earthquake-induced landslide hazards based on the Newmark method and frequency–magnitude curves. Liu [31
] put forward Newmark’s sliding rigid-block model to calculate cumulative displacements and to identify potentially unstable areas. Chousianitis [32
] developed an empirical estimator of co-seismic landslide displacements based on the Newmark model to assess the hazards related to earthquake-induced landslides. Del Gaudio [33
] used the Newmark model to analyze seismic hazards in landslide-prone regions.
The author’s previous work was mainly to construct the hazard assessment model of earthquake–collapse–landslide–debris flow disaster chain based on a BN from the perspective of every single disaster according to the hazard formation mechanism, which included chain probability and hazard intensity, and were obtained from inference of BN [34
]. There are still some deficiencies in the previous works, first, the risk of the earthquake disaster chain was not considered, and the hazard discussed in the previous works was one factor in risk assessment elements. Second, BN cannot adequately reflect the nature of the impact of earthquake on secondary disasters, and cannot well calculate the hazard intensity of earthquake disaster. Finally, the hazard of earthquake disaster chain was analyzed from a single disaster perspective. Based on the above deficiencies, this study put forward a volcanic earthquake–collapse–landslide disaster chain risk evaluation model on the basis of a BN model and Newmark model. The risk evaluation model was constructed according to natural disaster risk formation theory from the perspective of the overall disaster chain. The BN can be used to analyze chain probability between the adjacent disaster events, and the hazard intensity of earthquake on secondary disasters can be analyzed according to the permanent displacement calculated in the Newmark model as the permanent displacement obtained from the Newmark model can better describe the impacts of earthquakes on secondary disasters. To validate the effectiveness of the assessment method, the earthquake–collapse–landslide disaster chain induced by the Changbai Mountain volcano eruption was used as a case study. The earthquake disaster chain risk assessment model and case study presented in the study aim to offer a framework and tool for seismic disaster chain risk identification on the basis of natural hazard risk theory, which considers the chain probability of the disaster environment, hazard intensity of the hazard factor, and the vulnerability of the disaster body.
Most major natural hazards can trigger a series of catastrophically secondary disasters, either simultaneously or sequentially, due to the disaster chain characteristics of temporal inducibility and spatial sprawl. The risk assessment of disaster chains is more complicated than that of individual disasters because the primary disaster can trigger a series of secondary disasters, and it is very difficult to recognize the interactions and chain mechanism involved. Despite more attention having been paid to the relationships between different disaster events, there is currently still no uniform conceptual model that can be used to evaluate the earthquake disaster chain risk. Compared with individual disasters, there are more hazard factors in a disaster chain. Furthermore, the vulnerability of a disaster body is changed by multi-hazard factors, where not only are there more disaster events, but the spatial scope of damage is also greater in a multi-hazard environment. According to major disasters of the past, there is therefore a real need for a new method for earthquake disaster chain risk evaluation. In this research, a new earthquake disaster chain risk evaluation conceptual model that coupled a BN and Newmark models was proposed on the basis of natural hazard risk formation theory. The new method combined these models on the basis of Excel and ArcGIS software, which is useful for the identification of quantitative risk parameters including chain probability, hazard intensity, and vulnerability. The chain probability of an earthquake disaster chain was obtained from the BN model, the hazard intensity of the earthquake disaster chain was calculated by the Newmark model. The joint method successfully highlights the comprehensive recognition of the disaster chain formation mechanism and quantitatively assesses disaster chain risk.
Using this joint method, the risk assessment results for the Changbai Mountain volcano earthquake disaster chain were obtained. The risk map showed that the high- and medium-risk zones were predominantly located within a 10 km radius of Tianchi, whereas the other regions of the study area primarily contained very low- or low-risk values. The verification results showed that the area under the ROC curve was 0.817, a value reasonably in agreement with both the AUROC and with the value obtained in the validation procedure on the test subset, thus suggesting that the simulated results based on this new method were coincident with disaster events of the past. The earthquake disaster chain risk assessment model proposed in this study provides a reference for the prevention and mitigation of disaster chains in mountainous area.
Although the joint method presented good performance for the earthquake disaster chain risk evaluation in the Changbai Mountain region, obstacles still exist for the assessment of disaster chain risk. The chain formation process from the primary to the secondary disasters is complicated, particularly in the vulnerability changes of the disaster body due to being repeatedly damaged by the same (or new) hazard factors. Therefore, a lack of consideration of the vulnerability changes for earthquake disaster chain risk assessment is a significant defect, and represents an issue for further study.