Susceptibility Assessment of Landslides Triggered by the Lushan Earthquake, April 20, 2013, China

Abstract

The Lushan earthquake ( Ms = 7.0; epicenter located at 30 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">°</sup> 17'N, 102 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">°</sup> 57 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">'</sup> E) occurred on April 20, 2013 and had a focal depth of 12.3 km. The earthquake was triggered by the reactivation of the Longmenshan Fault in Lushan County, Sichuan Province, China. This earthquake caused massive landslides that resulted in tragic loss of life and economic devastation. Strong earthquakes are among the prime triggering factors of landslides. The zone of highest seismic intensity for this earthquake was selected as a case study to assess the susceptibility to earthquake-induced landslides. Visual interpretation of color aerial photographs with 0.4- and 0.6-m spatial resolution and extensive field surveys provided a detailed landslide inventory map that included 226 landslides. Nine primary landslide-related factors were selected as predictor variables, including elevation, slope, aspect, curvature classification, distance from drainages, slope structure, lithology, distance from faults, and peak ground acceleration. The support vector machine (SVM) is a popular learning procedure that is based on statistical learning theory and utilizes a kernel function to map data from the original feature space to a high-dimensional space. Using an SVM, a nonlinear landslide system can be converted into a linear landslide system. Two parameters C and \mbi σ must be carefully predetermined to establish an efficient SVM. Therefore, a genetic algorithm (GA) was adopted to optimize the parameters of the SVM. The proposed GA-SVM model with the highest predictive accuracy and generalization ability was trained and then used to predict landslide susceptibility. The analytical results were validated by comparing them with known landslides using a success rate curve and classification accuracy. The GA-SVM model has an area ratio of 0.9586 and a kappa coefficient of 0.9575 and outperforms the SVM. Approximately, 94.97% of the landslides lie in the very-high-susceptibility region, 2.17% of the landslides lie in the high-susceptibility region, 1.13% of the landslides lie in the moderate-susceptibility region, and 1.73% of the landslides lie in the low- and very-low-susceptibility regions. The experimental results demonstrate that the GA-SVM model provides the best predictive accuracy. The model can effectively assess landslide susceptibility and provides a novel method for landslide prediction.