| The southwestern mountainous region in China is a high-risk area for earthquakes,and the occurrence of strong earthquakes and induced landslides are crucial factors that impact the economic development of the region.Therefore,conducting landslide hazard assessment under different earthquake scenarios(pre-earthquake,during-earthquake,and postearthquake)in the earthquake high-risk area of southwestern China,accurately predicting the spatial distribution of earthquake-induced landslides,is an effective way to mitigate the impact of earthquake disasters.It can also provide an important reference for future earthquake prevention and mitigation,land use planning,and major engineering project site selection.Scholars have conducted studies on seismic landslide hazard assessment under different earthquake scenarios from various perspectives.However,most existing research focuses on landslide inversion analysis of historical earthquakes.By contrast,there are relatively few studies on the hazard evaluation of landslides induced by potential earthquakes(preearthquake)and emergency evaluation of co-seismic landslide hazards in earthquake response scenarios(during-earthquake).In the limited existing studies,the permanent displacement model is usually used for regional seismic landslide hazard assessment.The model integrates both the mechanics and kinematics of seismic landslides,allowing for an accurate assessment of the seismic landslide hazard through quantitative analysis of slope instability mechanisms and slip processes.The model’s assessment results heavily rely on the accuracy of rock mass strength parameters,which are challenging to obtain with current technology,especially for complex geo-environments where the spatial heterogeneity of rock mass strength is difficult to quantify accurately.This limitation restricts the application of the model in complex geo-environments.In addition,few studies have taken into account both the effects of rainfall and earthquakes on the distribution of landslides in seismic landslide hazard assessment.Using a single factor is not sufficient to accurately reflect the complex environmental factors that can contribute to landslide occurrence.This thesis firstly develops a quantitative model for rock mass strength spatial heterogeneity in complex geo-environments based on the inversion of historical seismic landslide inventory.This model is subsequently integrated to the permanent displacement model to create a novel model called "Considering Rock Mass Spatial HeterogeneityPermanent Displacement"(CRMSH-PD).The CRMSH-PD model can effectively assess seismic landslide hazards in the complex geoenvironment.Additionally,we propose a landslide hazard assessment method that considers the co-influence of rainfall and earthquake using the CRMSH-PD model.The main contributions of the thesis are as follows:(1)This thesis proposes a method for analyzing spatial heterogeneity of rock mass strength based on historical seismic landslide inversion and develops an empirical model to quantify regional rock mass strength spatial heterogeneity,taking into account the influence of complex geoenvironments.First,this thesis presents a method to address the issue of unreliable non-landslide samples in historical seismic landslide inversion.Specifically,a one-class classifier-based non-landslide sample generation strategy is designed.This strategy divides the earthquake disaster area into regions with different hazard levels using the one-class classifier,and generates non-landslide samples from regions with low and very low hazard levels.This step will guarantee the reliability of non-landslide samples.Based on the generated non-landslide samples,logistic regression is applied to select environmental parameters that are highly related to seismic landslides.These environmental parameters will be applied to construct an empirical model of rock mass strength spatial heterogeneity under the complex geo-environment.Experimental results in the Wenchuan earthquake show that the presented model effectively quantifies regional rock mass strength spatial heterogeneity,and most earthquakeinduced landslides locate in the area with high spatial heterogeneity.In addition,compared to traditional non-landslide sample generation strategies,the designed strategy effectively quantifies the rock mass strength spatial heterogeneity,which is more consistent with the actual seismic landslide distribution.This again verifies the reliability of nonlandslide samples generated by one-class classifier.(2)This thesis proposes a new permanent displacement model called CRMSH-PD that considers the rock mass spatial heterogeneity in the complex geo-environment.First,the principles of the permanent displacement model are outlined,and the correlation between the critical acceleration and the rock mass strength is analyzed.Based on the correlation analysis,the critical acceleration is modified using the empirical model of rock mass spatial heterogeneity.The new critical acceleration is subsequently applied to the permanent displacement model to establish the CRMSH-PD model.Experimental results on Wenchuan and Ludian earthquakes show that,compared to the traditional permanent displacement model,regional seismic landslide hazard distribution produced by CRMSH-PD achieves higher prediction accuracy and is more consistent with the actual earthquake-induced landslide distribution.(3)This thesis proposes a method for assessing landslide hazards under the influence of combined rainfall and earthquake factors.A study is conducted to predict the spatial and temporal distribution of potential seismic landslide hazards under future earthquake scenarios.The rainfall infiltration model is first applied to quantify the wet front depth and calculate the slope saturation.The saturation parameter is then applied to CRMSH-PD to assess seismic landslide hazard.Bomi,a high-risk earthquake area in southwest China,is selected to validate the presented method.Results show a wide range of potential seismic landslide hazard areas in Bomi under rare-and extremely rare-encountered earthquake scenarios in the future,with concentration in the southern part of the region.The potential seismic landslide hazard level in areas with heavy rainfall is significantly increased.The temporal distribution of rainfall reveals that July will be the most exposed month of the year to future earthquakes,with a wide range of potential seismic landslide hazard areas.Conversely,January,February,November,and December of every year are relatively less affected by future earthquakes,with only a few areas being in potential seismic landslide hazard areas.Results in this study provide important support for disaster prevention and mitigation in earthquake high-risk areas.(4)Based on the aforementioned presented methods,we conducted a co-seismic landslide hazard assessment under earthquake emergency scenarios.Using the earthquakes in Lushan on June 1,2022,and Luding on September 5,2022,we rapidly generated a predictive map of the distribution of co-seismic landslides within 2 hours after the earthquake.This map was promptly reported to the Ministry of Emergency Management and the Comprehensive Disaster Mitigation Technology Special Committee of China Disaster Defense Association for use by government emergency departments in disaster analysis and emergency relief efforts.Furthermore,we obtained seismic landslide inventory for severely affected areas using UAV images acquired within 48 hours after the earthquake.This allowed us to visually interpret the results,and verify the validity of the mapping results derived by the presented methods.These findings demonstrate the potential application value of the proposed methods by joint consideration of rainfall and earthquake effects for earthquake emergency scenarios. |
