Susceptibility Assessment Of Landslides Induced By Earthquake In Wenchuan County Western Sichuan Plateau (WSP) Using Random Forest (RF) And Artificial Neural Network (ANN)

Earthquake-induced landslide is one of the most dangerous secondary disasters in mountainous areas worldwide.It is crucial to demarcate landslide-prone zones to plan land management,development,and urbanization in mountainous regions.Clarifying geomorphological and geological controls on earthquake-induced landslides is difficult.Because an earthquake can transiently release powerful energy and commonly exert a first-order control on the landslide distribution and abundance,the effects of geomorphology and geology on coseismic landslides are easy to be weakened by the seismic factors,particularly in those areas where the valley and the seismogenic fault are aligned.Also,the Identification of landslides and preparation of landslide susceptibility maps are crucial steps in landslide susceptibility assessment.Therefore,an optimal landslide susceptibility model is presented that can produce accurate landslide susceptibility maps and assess landslide susceptibility.The devastating Wenchuan earthquake occurred in the western part of Sichuan province in China on 12 May 2008.Several catastrophic debris flows occurred in the years following the earthquake,demonstrating many of the postseismic landslide problems.The study area is Wenchuan county in Western Sichuan Plateau,located in the North-western Sichuan Plain with 4,084 Km2 in the Longmenshan seismic belt,one of the most critical active continental earthquake regions in the world,on the eastern edge of the Tibetan Plateau.A series of causal factors control a landslide;hence precisely identifying the causative elements is crucial for reliable LSM.This work identified the eight most often utilized causative factors for earthquake-induced in Wenchuan by analyzing publications from 2005 to 2016.We built a landslide inventory using historical records.We selected Eight causative factors from a library of factors and then performed a landslide susceptibility assessment(LSA)based on RF and ANN models.These parameters include slope Aspect,NDVI,Lithology,Distance to the river,Land Use,Soil,Peak ground acceleration(PGA)and Slope Angle.The main goals of this research work are(1)To construct ANN and RF models in the study area for seismic induced landslides and(2)To evaluate the spatial speculation capability and forecast precision of the landslide susceptibility models.(3)To determine other influential factors and develop an occurrence probability model of earthquakeinduced landslides.The prediction ability of the above two LSM models was assessed using the area under curve(AUC)value of the receiver operating characteristics(ROC)curve.And the new proposed benchmark model and existing soft computing models were also evaluated using five statistical measures: sensitivity(SE),specificity(SP),accuracy(AC),mean absolute error(MAE),and root mean square error(RMSE).True Positive(TP),False Positive(FP),True Negative(TN),and False Negative(FN)were used.To compute sensitivity,specificity,and accuracy based on four possible outcomes.The numbers of landslide cells classified as landslide and non-landslides are denoted by TP and FP,respectively,precision,recall ratio,accuracy and specificity.The machine learning algorithms performed very well based on LSMs and AUC values.But RF outperformed when it came to accuracy and prediction.The results showed a difference between the performances of the two models,with the RF model(AUC = 0.966)outperforming,which is better than the ANN model(AUC = 0.914).Moreover,compared with the ANN model,the RF model showed a higher coincidence degree between the areas in the high and the very low susceptibility classes.Finally,the results provided:(1)A theoretical framework for machine learning applications(e.g.,RF and ANN)could be a reliable and low-cost tool to assess landslide susceptibility.(2)This result will provide a valuable description of earthquake-induced landslides in the study area that can be used to anticipate the features of landslides triggered in future in the study area.(3)This result will help decision-makers take appropriate disaster prevention and mitigation measures to protect residents in these areas.(4)The resultant landslide susceptibility maps would be helpful for regional planning and reconstruction in this earthquake-prone area.It will also play a vital role in proper anthropogonic activities,resources management,and infrastructural development.