An approach for predicting landslide susceptibility and evaluating predisposing factors
Year: 2024
Type: article
Abstract: Effectively leveraging the spatial location information of landslides is crucial for improving the accuracy of deep learning in predicting landslide susceptibility and exploring the impacts of predisposing factors. Current single deep learning models for landslide susceptibility assessment require enhancements in both prediction accuracy and robustness. The prevalent inclusion of noninterrelated positional information among samples leads to reduced prediction accuracy and challenges in quantifying the contributions of various covariates to landslide risk. This study proposes a landslide susceptibility assessment method that integrates ensemble learning with geographically weighted concepts. Using a stacking method, foundational models, namely, 1D convolutional neural network (1D-CNN), recurrent neural network (RNN), and long short-term memory (LSTM) network, are amalgamated to form the CRNN-LSTM ensemble model. Additionally, we construct the Deep learning geographically weighted regression (GW-DNN) model based on the principles of deep learning and geographically weighted regression to quantify the impacts of landslide-predisposing factors. The experimental results demonstrate that the AUC value of CRNN-LSTM for the landslide validation set is 0.961, significantly surpassing the AUC values of other individual deep learning models. The GW-DNN model efficiently establishes a regression relationship for high-risk landslide locations, achieving an R2 value of 0.872.Our devised framework, amalgamating ensemble learning with geographically weighted analysis, attains remarkable accuracy in forecasting landslide susceptibility within the study area. Additionally, it enables meticulous quantitative assessment of predisposing factors' influences on landslides in high-risk zones at individual spatial points (landslide units) . Thus, this framework holds substantial scientific reference value for effective landslide prevention and control strategies.
Language: English
Subfield: Management, Monitoring, Policy and Law
Field: Environmental Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $2,820
Awards: G5137566052, G6393603979
OpenAlex ID: W4405173383