| Debris flow is one of the most frequent geological disasters,widely developed in mountainous areas,volcanoes,semi-arid and sub-polar regions.The strong fault activity,steep terrain,broken landforms and low vegetation coverage in northwest yunnan provide favorable natural conditions for the birth of debris flow disaster,making northwest yunnan one of the most seriously affected areas in the world.Debris flow disaster is preventable,controllable and treatable.It is better to contain it at the source than to spend a lot of manpower and material resources to rebuild after the disaster.Debris flow vulnerability assessment is the primary task to mitigate the risk of debris flow disaster.The investigation and assessment of debris flow disaster are of great significance to the prevention and mitigation of debris flow geological disaster,which is directly related to the development of local economy and social stability.This paper takes northwest yunnan as the research area,mainly elaborates the selection of influencing factors by using geographical detector,and evaluates the susceptibility of debris flow geological disasters in northwest yunnan based on full-factor logistic regression model,regional characteristic logistic regression model and bayesian hierarchical model.The following is the main research content of this paper:(1)Use geographical detector to screen the influencing factors of debris flow in the study area.In this paper,the geological disaster of debris flow in northwest yunnan is taken as the research object,the spatial variability of debris flow in northwest yunnan is preliminarily explored with geographical detector,the influence degree of 12 primary evaluation factors on debris flow is revealed,and the first five index factors that have the largest influence on debris flow in northwest yunnan are selected.In northwest yunnan,the top five influencing factors for the distribution of debris flow points are "maximum daily rainfall","river density","distance from fault zone","maximum continuous monthly rainfall",and "distance from road".(2)Establishment of evaluation model.Based on the method of systematic sampling,this paper selects the training sample set to construct the bayesian hierarchical model,full-factor logistic regression model and regional characteristic factor logistic regression model,and obtains the prior knowledge.In this aspect,the rules of sample selection,the establishment of training sample data set and validating sample data set are described.In the process of building bayesian hierarchical model,the first step is to build an initial model.In this paper,K function is used to build bayesian spatial adjacency matrix with surrounding K(K=20 in this paper)for model training and learning.In the second step,the structure of the model is adjusted according to the super prior knowledge and prior overall distribution information to make the evaluation result of the model more accurate and more robust.(3)Establish a cross-validation sample set comparison evaluation model based on random samplingIn order to verify the performance of the bayesian hierarchical debris flow vulnerability evaluation model,the results of the model were cross-compared and verified in this study,and were compared and analyzed with the results of the full-factor logistic regression model and the regional characteristic logistic regression model.The results of cross validation show that the performance of bayesian hierarchical model is high and stable.Compared with the full-factor logistic regression model and the regionally featured logistic regression model,the bayesian hierarchical model has better evaluation performance,better robustness and better anti-interference,in terms of both the evaluation performance and the evaluated spatial distribution of debris flow geological disaster susceptibility.The results of the spatial distribution of the vulnerability of debris flow geological hazards calculated by the bayesian hierarchical model are in line with the actual situation,which also shows that the bayesian hierarchical model can be applied to the evaluation of debris flow geological hazards in a large scale.(4)Obtained the distribution map of debris flow susceptibility in northwest yunnan. |
