Research On Risk Assessment Of Geohazards In Xiuyan Based On Remote Sensing Interpretation And Machine Learning

In recent years,geohazards had frequently occurred in China due to the effects of extreme weather events.Although local governments had implemented many prevention and control measures,hidden dangers and potential risks of geohazard persist,its risk assessment is always the basic work which the geohazard domain attaches great importance to.The thesis focuses on the Xiuyan Manchu Autonomous County area of Liaoning province,a typical geohazard-prone region in northeast China.The geohazard development information was identified,and its characteristic distribution was mastered based on remote sensing image.The geohazards related,induced factors and the elements at risk in the study area were summarized and used to establish a series of geohazards assessment index system.A comprehensive risk assessment of geohazards in the study area was carried out using machine learning models.The results are as follows:(1)A total of 265 geohazards were delineated.Basic information such as lithology,tectonics,and vegetation were identified.The data precision was so high that it benefitted the accurate construction of the index system and the effective input of model data.(2)The characteristics of geological environment,such as high topography and low vegetation coverage,were the main geohazard related geological conditions in the study area.The precipitation conditions were the primary factor controlling the outbreak of geohazards,which combined engineering activities such as buildings and roads,constituted the geohazard induced conditions.The population,GDP and arable land were the main elements at risk in the study area.(3)The convolutional neural network(CNN)model was the dominant model for assessing the susceptibility and probability of geohazards in the study area.It was more accurate,stable and reasonable.The extremely high susceptibility and probability area under the model was the smallest,with the highest geohazard density,and the division of the area was clear.The main susceptibility index was elevation,while the main probability index was heavy rainfall.(4)The results of gradient boosting regression(GBR)model reveal that the population density was the primary index to assess geohazards vulnerability in the study area,and it had an obvious correlation with geohazards vulnerability.It was a direct cause of high vulnerability in the western,northern and some central-southern parts of the study area.(5)Both the extremely high vulnerability and probability area in study area was very few.Therefore,the extremely high risk area is small.Sanjiazi,Shimiaozi and other towns in the study area were extremely high risk area for geohazards,where geohazards were concentrated and belong to the key prevention and control objects in the study area.It was suggested that measures,such as blocking,guiding and biological combination,be taken to prevent and control geohazards.(6)The number of geohazards in the study area was gradually reduced after the third period of geohazards investigation,and the decline rate of geohazards in the extremely high risk area was the largest,accounting for 85% of the total reduction of geohazards.The vegetation area had greatly restored,and the sustainable development situation in the study area had also improved.It shows that the risk assessment method of geohazards can effectively guide the work of geohazards prevention and control in the study area according to the priorities.The thesis includes 38 pictures,19 tables and 90 references.