Study On The Landslide Susceptibility Evaluation Method Based On Multi-source Data And Multi-scale Analysis

Landslide is a large adverse natural geological hazard phenomenon which has the characteristicsof wide distribution, high frequency and destructive. China is one of the countries that are enormously influenced by the landslide. Every year, landslide hazards cause a heavy loss to national construction and people’s life and property. With the rapid development of the economy and society, human’s engineering activities cause an increasing perturbance on the natural environmentthat leads to the increasing frequency and intensity of the landslide hazards. As a method to predict the spatial distribution and the occurrence probability of landslide, landslide susceptible evaluation is widely applied to the analysis and explore of the landslide hazards and actsas an important part in the management of landslide hazards, urban planning and the scientific layout of landslide monitoring work.Landslide susceptible evaluation method can be basically classified into determinacy and indeterminacy methods by the theoretical basis. The determinacy method uses the traditional slope failure mechanics calculation model and combines the basis spatial data to predict the risk of regional landslide. In recently years, with the developing of "3S" (GPS-Global Positioning System、RS-Remote Sensing、 GIS-Geographical Information System), a new way for landslide monitoring, landslide survey and landslide environment factorextraction are developed, which promotes the research and application of the indeterminacy method.Besides, some scholars notice that scale problem is important in the landslide susceptible evaluation and some research works have been conducted. However, Landslide susceptible evaluation method still exists some problems that need to be further studied and discussed. Currently, since there is considerable ambiguity about the properties, characteristics, classification.processing procedure and requirements of the multi-source data which is used in the landslide susceptible evaluation, it’s impossible to make a standardized operation in the data’s collecting and processing. This goes against to sufficient, reasonable and correct using of relevant data and has a bad influenced on the accuracy and reliability of the landslide susceptible evaluation, which lead to a difficult horizontal comparison between the evaluation results. Thedeterminacy method of landslide susceptible evaluation is mainlyused for the landslide spatial prediction of the specific slope unit area or the key area in the large scale. It’s hard to extend the application of this method to a bigger spatial scale.The indeterminacy method of landslide susceptible evaluation has several problems:(1) It generalizes the hazard environment and influent factors, but lacks the analysis of the landslide mechanism and hazard mechanism; (2) Lacking of a deep experiment on the landslide’s environment factors is another problem, most time, it’s not comprehensively considering about the selection of the most representative landslide evaluation factors and the spatial correlation of the evaluation factors; (3) In the choice and the application of the predicted model, less consideration is paid to the scale problem, and researchesdo not distinguish global model from local model. When giving consideration to the scale problem of the landslide susceptible evaluation, the scholars just take the scale as the prerequisite or mechanically use it in the landslide susceptible evaluation, they collect and process the data as the requirement of the different scale, but do not compare with the different scale landslide susceptible evaluation, or notice the different between the results of the different scale landslide susceptible evaluation. When the scale problem and scale concept are takenas screening principles of scale and the data resolution, the scale concept is separated, and the scale problem never taken as a whole phenomenon studied. It means that researchsdo not deeply explore the relationship between the results of the landslide susceptible evaluation and the different scale. Due to the ignoring of the relation among different scale, landslide judgement factors and landslide susceptible evaluation model or model parameter,researches can’t radically reveal how the scale concept influences the result of the landslide susceptible evaluation, nor dig out the key of the problems.Due to the above problems in the research of the landslide susceptible evaluation, this paper researches landslide susceptible evaluation method basing on the multi-source data and multi-scale analysis by combining the landslide susceptible evaluation in Zigui-Badong segment, Three Gorges area. This research work is carried out by integrally applying the relevant theory, technology and method of geology, surveying and mapping science, statistical analysis, and so on. The influence of different scale, landslide judgement factors and the parameter of landslide susceptible evaluation model to the results of landslide susceptible evaluation is deeply discussed, and landslide susceptible evaluation and mapping areachievedin the region scale, local scale and single scale. This research improved the scientificity, accuracy and reliability of the landslide susceptible evaluation.Following with the major research work and result in this paper:(1) On the base of the deep analysis and study of the properties, characteristics, classification and application of the multi-source data involved in the landslide susceptible evaluation, this paper built a uniform multi-source data processing framework, and gave a detailed data processing flow. The multi-source data processing method is extraced 58 initial landslide environment factors in Zigui-Badong in the first area of the three gorges reservoir.(2) Introducing the multi-scale analysis into landslide susceptible evaluation, this paper deeply analyzes and reveals the influence of the scale on the result of the landslide susceptible evaluation, clearly put forward the concepts of local scale, regional scale, monomer scale landslide susceptible evaluation, and deeply discuss the influence of the result of the landslide susceptible evaluation when choosing the different scale, landslide judgement factors or the parameters of the landslide susceptible evaluation model.(3) Geographically weighted idea is proposed for the landslide region segmentation vmethod in the study area, and the coupling model for landslide susceptible evaluationis established on the bases of geographically weighted regression, particle swarm optimization (pso) algorithm and support vector machine (SVM).Through geographical weighted regression coefficient of the study area, the space correlation of landslide evaluation factors between different partitions can be better inhibited. Meanwhilethe influence of landslide evaluation factor inside the partition on landslide tends to be consistent.Thisbetter eliminated the influence of the space correlation of landslide evaluation factor on landslide susceptible evaluation.(4)This paper made the comparison of landslide susceptible evaluation between the region scale and the local scale from several difference angle. Using the specific categories precision analysis, the overall accuracy analysis, and the most commonly used ROC curve analysis, the results between the region scale and the local scale are compared, and the accuracy of region scale obtained by these three methods are 85.75%,87.86% and 0.965, respectively, while the accuracy of local scale are 68.35%,84.44% and 0.944.This proves the importance and the actual result of regional separation in the landslide susceptible evaluation.(5)This paper extend the concept of the landslide susceptible evaluation to the single scale and make it fit to the single landslide.In addition, a susceptible assessment indicator system of single scale is built. On the basis, this paper proposed a susceptible evaluation method in the single scale, which is applied to the susceptible evaluation of Muyubao landslide in the three gorges reservoirs. The result of the evaluation is the same with the stable analysis of Muyubao landslide provided by the three gorges reservoir geological hazard prevention department. On the basis of the accumulative displacement and movement velocity, this paper built the landslide susceptible judgement factors to find more judgement factors combination that caused the unstable block in the Muyubao landslide. Finally, this paper found the potential unstable area, which provided a scientific and reasonable basis of the reasonable layout of the GPS monitoring points and the monitoring data acquisition frequency in Muyubao landslide.(6)This paper discussed the relationship of region scale, local scale and single scale and the landslide susceptible judgement factors, respectively, and the relationship of different spatial scale and the parameter of the landslide susceptible evaluation model. The result proved that the importance ranking of landslide evaluation factors is a gradually "average" process with the increase of spatial scale, thus, the space scale problem is most directly affected the importance ranking of landslide evaluation factors; In the analysis of spatial scales and parameter choice of landslide susceptible evaluation model, with the decrease of spatial scale, the training data within the scope of research with the landslide data in data difference is more apparent, to play an important role in the landslide evaluation factor or landslide evaluation factor combination is more significant, so that model can get better landslidesspace prediction in a shorter period.