Landslide Susceptibility And The Impact Of Forest On It In Southwest China

Geological disasters are prone to occur in China,landslide disasters occur most frequently and affect the most widely,and southwest mountainous areas are the most serious.Landslide disasters often occur in the fragile ecological environment.To investigate and prevent hidden hazards of landslides is of great significance for protecting people's lives and protecting ecological security.In recent years,the risk investigation,early warning and ecological control in terms of regional geological hazards have been highlighted in related researches.By combining remote sensing technology,geographic information system,machine learning technology and adopting Bijie city as the test zone,this paper establishes a set of technical methods related to the early risk investigation of the regional landslide disaster.Also,the impacts of forest on landslides are analyzed and forest management suggestions are made from the perspective of ecological control.The main research content and conclusions of this paper include:(1)Regional landslide hazard recognition based on convolutional neural network and attention mechanism can improve the efficiency and accuracy of regional landslide hazard investigation by providing geological disaster experts with target areas.The first regional landslide recognition sample database(including landslide samples,non-landslide samples,landslide boundary documents and relevant DEM data)was built.The database proved to be accurate through the verification by geological experts.It was released to the public through internet in order to facilitate the research on auto landslide detection based on optical remote sensing images.A 3D space-channel attention module is designed in terms of potential landslide hazard design.The unique characteristics of landslide cases under complex backgrounds are emphasized through the direct generation of three-dimensional space and channel attention feature map rather than through the processing and generation of attention maps in space and channels.A set of automatic convolutional neural network recognition solutions is discovered through experiments and successfully applied in project cases.(2)Multiple time phases,wave bands and perspectives In SAR cooperative monitoring method is effective to regional periodic surface deformation monitoring.The In SAR collaborative monitoring method involving multiple time phases,wave bands and perspectives were put forward for the first time in China.The periodic monitoring of regional long term mountainous landslide surface deformation was conducted also for the first time.The radar data of ALOS-2 wave band and the D-In SAR method are used to monitor the abrupt surface deformation in the areas with high vegetation coverage during rainy seasons.The regional peristaltic landslides are subject to long term surface deformation monitoring once every month through the integration of RADARSAT-2 ascending data and Sentinel-1 descending data and by using the MT-In SAR method.The distributed scatter(DS)MT-In SAR landslide deformation analysis method for the southwestern mountainous regions of China is established.Distributed targets are collected through Anderson-Darling(AD)test and then subject to adaptive space filtering,which greatly maintains the coherence of DS and keeps the coherence information for independent strong-scattering targets better.Also,the M-estimation method with ranks is used to eliminate the impacts of non-Gaussian distributed targets and non-steady scattering targets and improve the validity of homogeneous target testing.(3)Regional landslide susceptibility evaluation is made through the recognized potential landslide hazard data.Using MT-In SAR results for process correction can improve the accuracy of landslide susceptibility evaluation.The distribution of the known potential landslide hazards is analyzed.Twelve factors are selected from factors such as geological and environmental conditions as well as mankind's engineering activities for evaluation.The information volume model is adopted to quantify disaster-causing factors.The support vector machine(SVC)model is selected to evaluate landslide susceptibility.Also,a landslide susceptibility zoning map is drawn.The MT-In SAR long term monitoring results are used to re-project the deformation rate measured along the sight line of satellites(V?LOS)onto the slope direction(V?slope).A correction matrix is formed to update the initial landslide susceptibility risk map.The percentage of updated known potential landslide hazard data in vulnerable and highly vulnerable areas increases by 69% from the previous 64% before updating.Updating enables more accurate landslide susceptibility judgments.(4)It is found that forests produce some regulating effect on landslide susceptibility.Landslide influence shuould be considered in forest management.The relationship between different forest space layouts and different levels of landslide susceptibility are studied by establishing an evaluation model and applying the GIS analysis function.Shrubs are less likely to suffer from landslides than woodlands.In contrast with man-made forests,natural forests can control the occurrence of landslide disasters to some extent.In terms of forest age,half-mature forest are not likely to encounter landslides than young,mature,over-mature forest.The dominant specimens in the districts and counties of Bijie are mostly Chinese firs,shallow root coniferous forest represented by pinus yunnanensis and non-wood forests such as walnut,peach and osmanthus.And most of the forests are man-made young ones.So,they involve a high risk of landslides.Landslide factors shall be taken into comprehensive consideration in forest planning and design as well as afforestation.It is necessary to bring the ecological defense function of natural forests into full play,strengthen the protection of the existing man-made forests and gradually repair the natural forests with high slope and broken soil.During the process of forest construction,man-made forests should be planted by following the principle of biomimetic plant community and the combination of trees and shrubs should be highlighted so as to diversify man-made forests.For single forest varieties with high susceptibility,it is essential to optimize the forest structure,add various deep-root shrubs,and improve the shear resistance of soil and the stability of the ecological system.In the planning on planting forests,especially shallow-root coniferous forest and unique special woods in mountainous regions,the areas highly prone to landslides should be avoided and the soil and water conservation should be highlighted.By studying potential landslide hazards and forest in Bijie city,this paper provides multiple technical supports for the investigation,monitoring and early warning of regional geological disasters and facilitates the transformation of geological disaster control from "civil defense" to "technical defense".Particularly,under the macro background of national ecological civilization building in China,this paper discusses the relationship between forest and the susceptibility of potential landslide hazards and provides a reference for decision-making on the ecological control of landslide disasters in the southwestern mountainous regions.