| Rainfall-induced accumulation Landslides represent one of the most catastrophic natural disasters in mountainous areas of the world,such as Italy,Japan,the Himalayas,and southeast China,with complex mechanisms,widespread distribution,and frequent clusters.Regional rainfall-induced accumulation landslide susceptibility and hazard assessment can predict the time and space of landslide occurrence and answer the question"When and where will landslide occur in a heavy storm?"Simultaneously,it is crucial to carry out numerical simulation risk quantitative evaluation of single landslides based on the dynamic assessment of regional hazards to reliably represent the actual risk of landslides,which is crucial for accurate landslide early warning systems and effective risk management.At present,in the research of rainfall-induced accumulation landslide hazard and risk assessment,there are still difficult problems that hinder the application of multi-drive model:Modeling methods and scale uncertainty for regional landslide susceptibility evaluation.The physical mechanism-driven model is mainly based on the grid units infinite slope model,which is difficult to select slope units with physical and mechanical properties efficiently and reasonably.The rainfall threshold model coupled with data-driven method cannot calculate the continuous spatiotemporal probability value between 0 and 1.There are also many problems to be solved in the risk assessment of single landslide numerical simulation,such as the standard of landslide disaster action intensity and the quantitative calculation method of landslide vulnerability.These constraints seriously affect the accuracy of rainfall-induced accumulation landslide risk assessment,and reduce the engineering practicability.In view of the above issues,this study takes rainfall-induced accumulation landslides in the southeastern hilly area as an example.Based on a series of theories and technologies such as data-driven model,TRIGRS physical mechanism-driven model,multi-scale segmentation method,rainfall spatiotemporal probability model,Massflow depth integration model,and finite element numerical simulation method,three stages of landslide research were carried out:from"Research on uncertainty in regional landslide susceptibility assessment based on data-driven model"to"Regional rainfall-induced accumulation landslides hazard assessment based on multi-driven models",and then to"Risk assessment of single landslide numerical simulation that effectively integrates the hazard of regional rainfall-induced accumulation landslides",the assessment system of rainfall-induced accumulation landslide region and single in different scale cooperative hazard and single landslide risk are constructed.The main conclusions are as follows:(1)Explore the uncertainties of landslide susceptibility prediction under the coupled conditions of influences of different data-driven models and the nonlinear connections between landslides and environmental factors.Accordingly,a total of20-types of different coupling modeling conditions are proposed for landslide susceptibility prediction with five different connection methods(probability statistics(PS),frequency ratio(FR),information value(IV),index of entropy(IOE)and weight of evidence(WOE))and four different data-based models including logistic regression(LR),back propagation neural networks(BPNN),support vector machines(SVM)and random forest(RF).Meanwhile,four single LR,BPNN,SVM and RF models with the original data as input variables are also proposed,a total of24 types of modelling conditions for LSP are obtained based on the above 20 types of coupling conditions and 4 types of single models.the results of WOE-RF model have the lowest uncertainties and the predicted landslide susceptibility indexes are more consistent with the actual Landslides distribution characteristics.(2)Investigate the effects of different mapping unit scales and study area scales on the uncertainty rules of landslide susceptibility prediction.To illustrate various study area scales,Ganzhou City in China,its eastern region(Ganzhou East),and Ruijin County in Ganzhou East were chosen.Different mapping unit scales are represented by grid units with spatial resolution of 30 and 60 m,as well as slope units that were extracted by multi-scale segmentation method.On the premise that the regional landslide data and related environmental factors are valid,the data-driven models are efficient for the prediction of rainfall-induced accumulation landslides in large-scale region,and the results can better reflect the general rule,especially when the prediction units are expressed by slope units divided by multi-scale segmentation algorithm.(3)A continuous probability model is proposed to calculate the time probability value of rainfall-induced landslides.The combination of the landslide susceptibility prediction based on data-driven model and the rainfall time continuous probability value realize the spatial-temporal landslide hazard warning,and solves the problem of low identification degree of conventional critical rainfall threshold of landslide induced by heavy rainfall.The critical rainfall parameters(EE and D)and the corresponding rainfall hazard probability(P_R)under the double logarithmic coordinates of critical rainfall thresholds at each level are calculated.Combined with sigmoid function and using Python to fit the nonlinear equation to obtain the landslide probability under the conditions of effective rainfall and rainfall duration in each early stage.(4)Based on slope units,a physical mechanism-driven model for regional rainfall-induced accumulation landslides hazard assessment has been constructed and programmed using Python.The physical mechanism-driven model includes three integral components:Firstly,the infinite slope model based on Moore-Coulomb theory is used for geotechnical modeling.Secondly,the hydrological model in TRIGRS is used to calculate the groundwater level within slope units.Lastly,the surface runoff model.(5)Through regional rainfall-induced accumulation landslide hazard assessment research,slope units in very high level of hazard can be identified,and calculations hazard probability of single landslides under different rainfall return periods(10 years/20 years/50 years)are calculated from the perspective of mathematical modeling and numerical simulation.Based on the principle of depth integrated continuum mechanics,a fluid friction model is implemented in Massflow to calculate the motion process of the source region from starting to sliding through the slip plane in three-dimensional terrain,and to predict the landslide velocity,accumulation thickness and flow path.(6)A risk assessment system has been established,integrating numerical simulation,hazard analysis,and vulnerability assessment of single landslide.Hazard assessment is conducted by assessing the instability probability and threat range of single landslide.By quantifying the landslide disaster intensity,vulnerability of hazard-bearing bodies,and fragility of landslide disasters,the risks of the economic loss and personnel casualties for the landslide in Shuikou Village,Huaixi Town,are estimated for 10-year,20-year,and 50-year rainfall return periods,respectively. |
