Study And Application Of Quantitative Assessment Methodology For Landslide Disaster

The construction of villages and towns has become an important part of the process of urbanization in China.Villages and towns are at an important stage of construction in the next decade.And it is also a key stage of disaster reduction in villages and towns.Most villages and towns in our country are located in the mountainous areas,facing the threat of various types of geological disasters,suffered heavy casualties and huge losses of property each year.Among all the geological disasters,the number of landslide disasters is the largest,accounting for more than half of the total number of geological disasters.Scientific accessment and study of landslide in villages and towns can not noly help people to fully understand and grasp the disaster situation,provide theoretical basis for scientific arrangements of disaster prevention and mitigation,but also guide the disaster work of prevention and mitigation of landslides in villages and towns,and play a practical role in the development of safety line for disaster prevention and control in villages and towns and in the planning evacuation sites of disaster.Therefore,this paper studied the quantitative assessment methodology of landslide disaster in villages and towns.And the models were applied to the landslide disaster assessment in the demonstration area in order to verify the feasibility and effectiveness of the assessment methodology and to provide a scientific and advanced Shennongjia forest for the popularization and application of the assessment methodology.The main work of the article and the results are as follows:(1)On the basis of reading a large number of literatures,this paper summarized the methods of classification and forecasting of disaster levels at home and abroad,and clarified the terminology,evaluation indexes,classification methods and forecasting methods of disaster grade.It also summarized the hazard assessment methods and achievements of regional landslide and single landslide at home and abroad and introduced several major landslide hazard assessment models whose existing problems were pointed out.(2)In the fuzzy clustering model,scholars usually directly chose a distance for cluster analysis,without comparison or selection so that the clustering result might not be optimal.In this paper,a classification model of disaster grade of landslide in the Shennongjia forest area was established with the fuzzy clustering method.And Manhattan Distance was determined by the correlation coefficient as the optimal distance in the model.Through the optimized model,the 34 landslides in the Shennongjia forest area were divided into four categories:8.8%of the landslides were classified as extraordinary landslide disasters 11.8%were large landslide disasters,8.8%were general landslide disasters and 70.6%belonged to mild landslide disasters.A fuzzy recognition model of landslide disaster grade was established based on fuzzy close degree.And the disaster grades of the two landslides to be evaluated were identified.According to the principle of proximity,one landslide disaster grade was identified as the general landslide disaster,the other as the large landslide disaster.Then the assessment system of recognition after clustering had been constructed for the landslide disaster grade in villages and towns.(3)Artificial neural network has been widely used in many fields such as pattern recognition,prediction and assessment,especially for BP neural network which is often used to predict disaster assessment,disaster grade and disaster risk.However,it is rare to see the application of RBF neural network in assessment and prediction for landslide disaster grade.Therefore,this paper attempted to build a RBF multi-index prediction model and applied the RBF neural network to the prediction for landslide disaster grade,and compared it with the BP prediction model for landslide disaster grade.(4)Taking the landslides in the villages and towns as the research object,the BP model and RBF model for disaster grade prediction of landslide were constructed with the disaster grade as the forecast target.The BP model and the RBF model were trained in 30 landslides of the Shennongjia forest area.The result showed that the maximum fitting accuracy of BP model is only 70%,and the fitting accuracy of RBF model is up to 100%,which indicated that the learning effect of RBF model was better than BP model.The well-trained network models were used to predict the disaster grade of other 6 landslide samples separately,The BP model could only accurately predict 3 landslides disaster grade,while the RBF model could accurately predict 5 landslides disaster grade.The result showed that the RBF prediction model has a higher prediction accuracy.(5)In the assessment of landslide hazard,most scholars only selected the susceptibility indexes describing the probability of occurrence of landslides,ignoring the hazard indexes which describe the damage of landslides.It led to the assessment results that could only reflect the susceptibility without the damage caused by landslides.In this paper,susceptibility indexes and hazard indexes were both selected as the assessment indexes of landslide hazard.Therefore,a more comprehensive index system was established.(6)In the fuzzy comprehensive assessment model for landslide hazard,ssessing indexes were usually weighted by subjective experience.In this paper,information entropy was used to assign weight to each assessing index.It not only has the advantages of simple operation and higher efficiency of calculation,but also avoids the subjective randomness.The fuzzy comprehensive assessment model based on information entropy was applied to the hazard assessment of the landslides in Shennongjia forest Forest Area and the results showed that the model based on information entropy is more realistic than the traditional model whose weight of assessing indexes were assigned by subjective method.