Debris-flow Susceptibility Assessment In Strong Seismic Zone

The Longxi River Basin in Dujiangyan City,Sichuan Province,which is located in strong seismic zone of Wenchuan earthquake,was strongly disturbed by the earthquake.Secondary geological disasters such as collapse and landslide became potential sources of debris flows.Debris flows broke out during the“8.13”heavy rainfall.Based on field surveys,literature surveys,and indoor calculus,the review and interpretation of post-earthquake geological disasters,source reserve calculation,factor sensitivity analysis,and index collinearity diagnosis were conducted in the Longxi River Basin,and the evaluation system for debris flow hazards in strong earthquake regions were established.The information entropy model(IE)and logistic regression model(LR)models were selected for debris flow susceptibility mapping and 41 sets of models were performed.The best IE model and LR model showed 12%and 13%predicted success rates,respectively,according to the receiver's work characteristic curve(ROC)test results.Finally,the evaluation results of two model were compared and cross-checked.(1)More than 3,300 collapse-accumulating deposits in the Longxi River Basin after the earthquake were checked and interpreted,of which 2,943 deposits can provide the potential source for debris flow.The total reserves of provenances and dynamic reserves of sources of material are respectively 113.87 million m~3 and 43.45million m~3.The interpretation delineated 31 valley-type debris flow basins and 9slope-type mudslide points.(2)Through the factor sensitivity analysis and the index collinearity diagnosis,8evaluation factors were initially reserved,and the strongest response range of each index to the development of debris flow was obtained.In the modeling process,the indicators are screened for the third time,and the order of the dominant factors for the debris flow was determined as:stratum lithology,distance from the seismogenic fault,dynamic reserve of the source of material,monthly rainfall in the rainy season,and slope.Two conclusions can be drawn from the differences in the predicted effects of the models.In the two situations in which the same indicator is used for independent evaluation and combined with other indicators to participate in evaluation,the role has different weights,which reflects the coordination and containment relationship between the indicators.Due to the lack of analysis of the disasters and the geological conditions of the disaster,adding a non-master factor will reduce the prediction success rate of the model.(3)The results of hazard zoning indicated that more than 90%of the actual debris flow occurred in the extremely high risk zone(A)and the high risk zone(B).The total area of the AB subarea of the IE and LR models accounted for 56.9%and57.1%of the study area respectively,and the their distribution range was the same.The A zone and B zone are concentrated within 4km from the Beichuan-Yingxiu fault zone.The south side of Longxi River is more dangerous than the east side,which is consistent with the distribution of real debris flow.(4)The results of the comparative analysis of the two models show that the IE model is easy to operate,emphasizing its own mathematical logic,and it performs well in the analysis of the relationship between disaster and geological environmental conditions with a reliable the evaluation results.However,it is weak in quantitative judgments such as the proportion of subdivisions,neglecting the mutual influence of indicators and failing to predict the probability of the sample.The LR model performed well in qualitative and quantitative analysis of the relationship between disasters and geological environmental conditions.Based on a certain number of data samples,the model can predict sample probability,and increasing the number of samples can improve the reliability and accuracy of the prediction results.