Risk Assessment Of Rainfall-induced Landslide Of Xiashu Loess In Ningzhen Region Based On Cloud Model

Landslide is one of the most universal geological disasters in the world.There are many mountainous areas and complex topography in China.Therefore,landslide disasters occur frequently and distribute widely.China is one of the countries with the most serious landslide disasters in the world.Landslides were caused by natural factors which account for 92.3% of the total,and by human factors as well,rainfall being the main natural factor.Ningzhen region,mainly including Nanjing and Zhenjiang,is a rich region in the middle of the lower reaches of the Yangtze River,located in the southwest of Jiangsu Province.Xiashu loess,a kind of cohesive soil,is widely distributed in Ningzhen area,the section plane of which is usually yellow-brown or gray-yellow.It is mainly composed of sand and silty,and is also called expansive soil because of its expansibility.Xiashu loess swells with the infiltration of rainwater after raining,while it shrinks because of the loss of water during the drought.Therefore,it facilitates the formation of preferred plane,which may result in landslides.Combined with the data of historical landslide and rainfall in Ningzhen area,the characteristics of rainfall is analyzed in this paper.By integrating the theoretical knowledge of support vector machine algorithm,genetic algorithm and cloud model,an early warning method for rainfall-induced landslides is put forward based on rainfall series,and the risk,hazard and vulnerability of landslides have been evaluated.The main research contents and results are as follows:(1)Research on the correlation between rainfall-induced landslides and rainfall factors of Xiashu loess in Ningzhen regionBased on the statistical analysis of landslide and rainfall historical data in Ningzhen region,the relationship between landslide and three rainfall factors,namely,rainfall amount,rainfall intensity and rainfall duration,is discussed.By means of SVM algorithm,the historical data of landslides are taken as training and testing samples,and a landslide prediction model is constructed by combining hourly and daily rainfall data.By comparing the accuracy of prediction,it is found that the effective rainfall in the early stage has a greater impact on the induced landslide.According to the characteristics of rainfall in Ningzhen region,a landslide prediction and early warning model based on rainfall sequence is proposed.Firstly,the state machine is used to build a model for the change of rainfall infiltration depth and infiltration volume in the process of precipitation sequence,and the calculation formula of rainfall infiltration and redistribution is given.Thus,the infiltration depth and infiltration amount of rainwater at each time point in the precipitation series are calculated.Then,genetic algorithm is used to screen the change space of the related parameters of the model,and a group of most suitable parameters are selected as a basis for dividing the precipitation sequence into normal ones and those causing landslides.Thus,the correct prediction and early warning of whether the input rainfall sequence can induce landslides is achieved.(2)Research on risk assessment of rainfall-induced landslides based on cloud model(1)Using the traditional analytic hierarchy process of cloud model coupling,the scale of cloud model is calculated when constructing the judgment matrix,so that the three digital features(expectation,entropy and hyper-entropy)of cloud model have been involved in the calculation and the weight of the cloud model can objectively reflect the randomness and fuzziness of evaluation factors.(2)Using cloud model instead of membership function,according to the grading criteria of each landslide risk assessment factor,appropriate cloud model parameters are selected to generate cloud models with different evaluation grades of each factor.The membership degree of a factor value belonging to a certain evaluation level can be obtained from the cloud map,and then combined with the weight of the cloud model,the comprehensive evaluation can be carried out.Finally,the method is validated by the example of "landslide in the preservation area of ancient tomb group in Badou Mountain",which has good effect and is feasible.(3)Research on vulnerability evaluation method of rainfall-induced landslides based on cloud modelAccording to the statistical analysis of landslide hazards in the study area,three main evaluation factors of landslide vulnerability in Ningzhen region are selected: population density,building density and road density.Using the comprehensive evaluation method of cloud model,according to the classification criteria of each factor,appropriate digital characteristics of cloud model are selected to generate the cloud model.Then the evaluation of landslide vulnerability is obtained by combining the weight of the cloud model.Finally,with the formula: risk degree(R)= hazard degree(H)Xvulnerability degree(V),the risk of landslide is quantitatively evaluated.The layers of landslide hazard index and landslide vulnerability index are generated by means of Arc GIS software.After normalization,the two layers are multiplied and the exponential map of landslide risk is obtained.