GEOSPATIAL TEMPORAL FRAMEWORK ON LANDSLIDES MITIGATION STRATEGIES FOR PIPELINES

This research has proposed a newer method of improving landslide susceptibility development and utilization. A 50-year return period of five years intervals of susceptibility maps was proposed to monitor the degree of deterioration of the slope
surfaces caused by the landslide. The susceptibility mapping was developed using data mining techniques and remote sensing data. These improvements in landslide susceptibility mapping were used to establish a landslide mitigation strategies framework for pipelines. The proposed framework is expected to help prevent the continued pipeline failures caused by landslides. Support Vector Machines (SVM) and
Artificial Neural Network (ANN) were used to develop the prediction models and conduct the temporal analysis of the landslides. Eight statistical indices, which include Root Mean Square Error (RSME), F-Measure, Sensitivity, Specificity, Absolute Mean Error (AME), Area Under the receiver operator curve (AUC), Accuracy (ACC), and Kappa, were used to validate the predictions. AUC values of 0.879 were obtained for the susceptibility models developed from the SVM algorithms, indicating outstanding predictive performance.