Landslides are a common and costly geological hazard, with regular occurrences leading to significant damage and losses. To effectively manage land use and reduce the risk of landslides, it is crucial to conduct susceptibility assessments. To date, many machine-learning methods have been applied to the landslide susceptibility map (LSM). However, as a risk prediction, landslide susceptibility without good interpretability would be a risky approach to apply these methods to real life. This study aimed to assess the LSM in the region of Nayong in Guizhou, China, and conduct a comprehensive assessment and evaluation of landslide susceptibility maps utilizing an explainable artificial intelligence. This study incorporates remote sensing data, field surveys, geographic information system techniques, and interpretable machine-learning techniques to analyze the sensitivity to landslides and to contrast it with other conventional models. As an interpretable machine-learning method, generalized additive models with structured interactions (GAMI-net) could be used to understand how LSM models make decisions. The results showed that the GAMI-net model was valid and had an area under curve (AUC) value of 0.91 on the receiver operating characteristic (ROC) curve, which is better than the values of 0.85 and 0.81 for the random forest and SVM models, respectively. The coal mining, rock desertification, and rainfall greater than 1300 mm were more susceptible to landslides in the study area. Additionally, the pairwise interaction factors, such as rainfall and mining, lithology and rainfall, and rainfall and elevation, also increased the landslide susceptibility. The results showed that interpretable models could accurately predict landslide susceptibility and reveal the causes of landslide occurrence. The GAMI-net-based model exhibited good predictive capability and significantly increased model interpretability to inform landslide management and decision making, which suggests its great potential for application in LSM.

Earthquakes cause a huge number of landslides and alter the regional landslide risk distribution. As a result, after a significant earthquake, the landslide susceptibility maps (LSMs) must be updated. The study goal was to create seismic landslide susceptibility maps containing landslide causative variables which are adaptable to great changes in susceptibility after the Jiuzhaigou earthquake (MS 7.0) and to perform a rapid update of the LSM after the earthquake by means of the distributed scatterer interferometric synthetic aperture radar (DS-InSAR) technique. We selected the territory of Jiuzhaigou County (southwestern China) as the study region. Jiuzhaigou is a world-renowned natural heritage and tourist area of great human and ecological value. For landslide susceptibility mapping, we examined the applicability of three models (logistic regression, support vector machine, and random forest) for landslide susceptibility mapping and offered a strategy for updating seismic landslide susceptibility maps using DS-InSAR. First, using logistic regression, support vector machine, and random forest techniques, susceptibility models of seismic landslides were built for Jiuzhaigou based on twelve contributing variables. Second, we obtained the best model parameters by means of a Bayesian network and network search, while using five-fold cross-validation to validate the optimized model. According to the receiver operating characteristic curve (ROC), the SVM model and RF model had excellent prediction capability and strong robustness over large areas compared with the LR models. Third, the surface deformation in Jiuzhaigou was calculated using DS-InSAR technology, and the deformation data were adopted to update the landslide susceptibility model using the correction matrix. The correction of deformation data resulted in a susceptibility class transition in 4.87 percent of the research region. According to practical examples, this method of correcting LSMs for the continuous monitoring of surface deformation (DS-InSAR) was effective. Finally, we analyze the reasons for the change in the revised LSM and point out the help of ecological restoration in reducing landslide susceptibility. The results show that the integration of InSAR continuous monitoring not only improved the performance of the LSM model but also adapted it to track the evolution of future landslide susceptibility, including seismic and human activities.

Due to the influence of atmospheric phase delays and terrain fluctuation in complex mountainous areas, traditional PS-InSAR technology often fails to select enough measurement points (MPs) and loses effective MPs during phase unwrapping. To solve this problem, this paper proposes an adaptive network construction algorithm, which combines the permanent scatterer (PS) points with the distributed scatterer (DS) points. Firstly, to ensure the extraction quality of the DS points, the covariance matrix of DS points is estimated robustly. Secondly, based on the traditional Delaunay triangulation network, an adaptive network construction method is proposed, which can adaptively increase edge redundancy and network connectivity by considering the edge length, edge coherence, edge number, and spatial distribution. Finally, a total of 31 RADARSAT-2 SAR images that cover the Zongling landslide group in Guizhou Province were used to prove the effectiveness of proposed method. The results show that the quantity of available DS points can be increased by 23.6%, through the robust estimation of the covariance matrix. In addition, it is demonstrated that the proposed network construction algorithm can balance the number, distribution, and quality of edges in the dense and sparse areas of MPs adaptively. This adaptive network construction approach can maintain good connectivity and avoid losing effective MPs to the greatest extent, especially when the scattering points are far away from the reference points. In short, the proposed algorithm improves the number of effective MPs and accuracy of phase unwrapping.