Geohazards Susceptibility And Risk Assessment Along The Upper Indus Basin

The China-Pakistan economic corridor(CPEC)project is passing through the Karakorum Highway(KKH)in northern Pakistan,which is one of the most hazardous regions of the world.Among all other hazards,landslides and debris flow are the most common geohazards of this region,causing severe damages every year to lives and infrastructure.Therefore,this work assessed and compared geological hazards(landslides and debris flows)and developed susceptibility maps and risk map by considering two standalone machine learning(ML)and three statistical approaches such as Shannon entropy(SE),logistic regression(LR),artificial neural network(ANN),weights of evidence(Wo E),and frequency ratio(FR)models.With this aim,geological hazards(GHs)inventories,i.e.,landslide and debris flow were prepared using remote sensing(RS)techniques with field observations and historical hazard datasets.In this study,thirteen conditioning factors;slope(degree),distance to faults,geology,elevation,distance to rivers,slope aspect,distance to road,annual mean rainfall,normalized difference vegetation index(NDVI),profile curvature,stream power index(SPI),topographic witness index(TWI),and land cover(LC)were considered.The spatial relationship of these factors with hazards distribution was also analyzed and found that the faults,slope,elevation,geology,LC and rainfall play a key role in the occurrence of geological hazards in the study area.The final susceptibility maps were validated based on ground truth points and plotting area under the receiver operating characteristic(AUROC)curves.The AUROC curves showed that the success rate of the LR,Wo E,FR,and SE models were 85.30%,76.00,74.60%,and 71.40%,respectively while the prediction rate of the models were 83.10%,75.00%,73.50%,and 70.10%,respectively.The success and prediction rates revealed that the logistic regression model performed better than the other three models.The comparison coefficient assessment test for the models'results was performed and found that LR and Wo E are of moderate similarity while a strong similarity was observed between FR and SE models.This test suggests the influence of the same conditioning factors in these models.Furthermore,the percentage of very high,high,moderate and low susceptible areas are 11.9%,9.24%,10.18%,39.14%and 30.25%respectively.The areas that lie in high to very GHs susceptible areas are Jijal,Chilas,NPHM,a part of Astore valley and Hunza-Nagar valley.The Hunza-Nagar valley is further selected for the geo-hazard risk modeling.Initially,the two machine learning techniques(LR,ANN)were used for the geo-hazards susceptibility assessment.The geo-hazards susceptibility maps were developed and analyzed by using the AUROC curve.The ROC curve showed that the success and prediction rate of LR is 82.69%and81.80%while for the ANN model it is 77.90%and 75.40%.Furthermore,the geo-hazard index map of Hunza-Nagar valley was prepared by considering the rainfall density.The settlement was considered as the main element at risk in this region;hence,the vulnerability map was also developed.By using the geo-hazard index map and vulnerability map,the geo-hazard risk map for Hunza-Nagar valley was prepared and divided into four classes;low 37.25%(20.21 km~2),moderate 5.40%(2.93 km~2),high 9.72%(5.27 km~2)and very high 47.64%(25.84 km~2).The geological hazards susceptibility maps(GHSMs)and geo-hazards risk map prepared from the proposed models can be used by the relevant government officials for local scale,as well as the smooth implementation of the China-Pakistan economic corridor project.