| Dongchuan District is located in the northern edge of the Yunnan-Guizhou Plateau,which is the overlap between the Sichuan-Yunnan meridional tectonic zone and the northeast China tectonic zone.Influenced by the Xiaojiang fault,the area has a loose geotechnical structure and strong erosion,which,together with the logging and copper refining and over-cultivation,have led to the deterioration of the ecological environment and serious soil erosion,providing favorable conditions for the development of geological hazards such as mudslides and landslides,and posing a serious threat to the safety of life and property of local residents.Therefore,it is of great practical importance to carry out research on early identification of potential landslides and assessment of landslide susceptibility inside this region.In landslide research and disaster monitoring,surface deformation can visually reflect the current stability and movement state of landslides.Although fractometer,GNSS measurement and level measurement are conventional methods for surface deformation monitoring,they can achieve high accuracy monitoring,but they are mostly based on point observation and are severely constrained by field monitoring conditions,which are not suitable for landslide research in wide area space.In recent years,Since it has the advantages of not being affected by the weather and having a broad detection range,synthetic aperture radar(InSAR)has been extensively utilized ground subsidence and landslide identification detection.Based on the advantages of InSAR technology and the urgent need for early identification of landslide hazards in Dongchuan District,this experiment uses SBAS-InSAR to detect surface deformation and identify potential landslides in Dongchuan District,and analyzes the evolutionary characteristics and development patterns of typical potential landslides by combining optical remote sensing images,rainfall data and regional environmental features;takes into account human activities,meteorological and hydrological conditions,and geoenvironmental background affecting landslide hazards in the study area to select susceptibility evaluation factors.The evaluation factors of landslide susceptibility are selected by considering human activities,meteorology and hydrology,geological environment background and other factors affecting the occurrence of landslide hazards in the study area,and combined with the historical landslide inventory data obtained from field mapping,the evaluation system of landslide susceptibility in Dongchuan District is constructed by using logistic regression and support vector machine models respectively,and the ROC is used to compare and assess the accuracy of the proposed models.The evaluation outcomes under the optimal model are then combined with deformation rate variables to improve the initial evaluation outcomes.The specific research contents and conclusions are as follows.(1)The time series outcomes of ground deformation of two tracks in the research region were geted using SBAS-InSAR interferometric processing of 89 ascending and88 descending Sentinel-1A data in Dongchuan District from January 2018 to January2021.During this period,the inverse radar line-of-sight deformation rates of Dongchuan area were-188.1 ? 88.9 mm/year and-163 ? 74.7 mm/year for the ascending and descending orbits data,respectively,with high consistency in spatial distribution,mainly showing that the areas with obvious subsidence centers are concentrated in the areas with complex topography and severe soil erosion.In the high mountain valley areas on both sides of the Xiaojiang River basin,where the topography is complex and soil erosion is serious,and in the towns with small topographic relief and gentle terrain,such as Dongchuan City,Trabucca Town and Redland Town,the ground surface is more stable and the deformation rate is between0-8mm/year.(2)In order to better elaborate the direction and degree of surface deformation and make up for the deformation deficiency of line of sight direction in the application of surface deformation detection in complex mountainous areas,the radar line of sight direction deformation results were decomposed into east-west direction and vertical direction by combining the geometric parameters of elevated track radar shooting.From the east-west deformation results,we can see that the study area is affected by the movement of the fracture zone,with the Xiaojiang fracture as the dividing line,showing different movement characteristics,west of the fracture is mainly westward displacement,east is eastward displacement;combined with the actual surface change conditions in Dongchuan district,the deformation rate threshold is set,and this is used to identify 15,10 and 25 potential landslides in the ascending and descending track LOS direction and vertical direction,respectively.In addition to all the potential landslides detected by the ascending and descending tracks,6potential landslides not identified by a single track were added to the vertical identification results,indicating that the vertical deformation has better landslide identification and detection capability in the study area;in addition,by analyzing the deformation evolution characteristics and triggering mechanism of typical potential landslides,it is found that landslides are seriously affected by external factors,and the deformation is more severe in the months with heavy rainfall or in the areas with frequent human activities.In addition,by analyzing the deformation evolution characteristics and induced mechanism of typical potential landslides,it is found that landslides are seriously influenced by external factors,and the deformation rate is larger and the stability of slopes is poorer in the months with high rainfall or in the areas with frequent human activities.(3)Based on the geological survey of Dongchuan district,the development pattern of potential landslides and analysis of landslide triggering factors,12 influencing factors such as undulation,plane curvature,stratigraphic lithology and average annual rainfall were selected as evaluation analysis indexes,and the landslide susceptibility evaluation results of the study area were obtained based on logistic regression and support vector machine models respectively,and the accuracy of the two models was checked by using ROC curves,and the AUC values of logistic regression and support vector machine were 0.84 and 0.91 respectively.The AUC values of logistic regression and support vector machine were 0.84 and 0.91,respectively.On this basis,the landslide susceptibility classification results under the support vector machine model were combined with the SBAS-InSAR deformation rate to construct an optimization matrix to optimize the original evaluation results,so as to reduce the classification errors caused by incomplete landslide data collection.Finally,the evaluation results before and after optimization were compared and validated by using quantitative statistics and field surveys.The results showed that66,094 misclassified units in the optimized evaluation results had their susceptibility levels updated,and the misclassification errors were corrected to some extent.This indicates that SBAS-InSAR is feasible and reliable for landslide susceptibility evaluation.In summary,the thesis focuses on landslide hazard prevention and control in Dongchuan District,Kunming,and carries out research on the identification of potential landslides and the optimization of landslide susceptibility assessment by SBAS-InSAR.On the other hand,it was demonstrated that the classification errors of the landslide susceptibility assessment results were reduced and the reliability of landslide spatial probability prediction was improved compared with the conventional susceptibility assessment results. |
PDF Full Text Request