Optimization of emergency rescue routes after a violent earthquake

A great earthquake usually causes heavy human casualties and extensive road damage. Rational rescue routing is vital to save lives. Massive loose deposits and unstable slope materials sourced from coseismic landslides severely endanger the safety of rescue workers and rescued persons. However, current rescue routing works concentrate on traffic conditions and neglect the threat of coseismic landslides. Safety and efficiency require to be taken into consideration comprehensively. This work for the first time involves coseismic landslide susceptibility as the safety factor into emergency rescue routing and determines the optimal routes by combining efficiency and safety. Furthermore this work employs the 2015 Mw 7.8 Gorkha earthquake as a benchmark and focuses on the significant scenario of multiple relief centers, multiple destructed residential areas, and limited rescue vehicles to conduct rescue route optimization. Two significant insights are suggested as follows. (1) In terms of about 20% of coseismic landslide data randomly selected (i.e., simulation of limited data collected just after an earthquake), coseismic landslide susceptibility along roads is predicted by LightGBM algorithm. The prediction Accuracy, Precision, Recall, F1-score, and AUC values attain 92.10%, 91.74%, 0.9245, 0.9209, and 0.9721, respectively. (2) The indices of road safety (landslide susceptibility), road accessibility, road length, and road grade are integrated to determine the optimal rescue routes by SPFA and Genetic algorithms. Compared with the optimal routes in terms of only efficiency, the ones coupling safety and efficiency feature lower landslide occurrence probability and smaller cost values. The involvement of landslide threat into rescue routing contributes to finding more rational rescue paths.