Accurate early warning systems for landslides are a reliable risk-reduction strategy that may significantly reduce fatalities and economic losses. Several machine learning methods have been examined for this purpose, underlying deep learning (DL) models’ remarkable prediction capabilities. The long short-term memory (LSTM) and gated recurrent unit (GRU) algorithms are the sole DL model studied in the extant comparisons. However, several other DL algorithms are suitable for time series forecasting tasks. In this paper, we assess, compare, and describe seven DL methods for forecasting future landslide displacement: multi-layer perception (MLP), LSTM, GRU, 1D convolutional neural network (1D CNN), 2xLSTM, bidirectional LSTM (bi-LSTM), and an architecture composed of 1D CNN and LSTM (Conv-LSTM). The investigation focuses on four landslides with different geographic locations, geological settings, time step dimensions, and measurement instruments. Two landslides are located in an artificial reservoir context, while the displacement of the other two is influenced just by rainfall. The results reveal that the MLP, GRU, and LSTM models can make reliable predictions in all four scenarios, while the Conv-LSTM model outperforms the others in the Baishuihe landslide, where the landslide is highly seasonal. No evident performance differences were found for landslides inside artificial reservoirs rather than outside. Furthermore, the research shows that MLP is better adapted to forecast the highest displacement peaks, while LSTM and GRU are better suited to model lower displacement peaks. We believe the findings of this research will serve as a precious aid when implementing a DL-based landslide early warning system (LEWS).
Multiple landslide events are common around the globe. They can cause severe damage to both human lives and infrastructures. Although a huge quantity of research has been shaped to address rapid mapping of landslides by optical Earth Observation (EO) data, various gaps and uncertainties are still present when dealing with cloud obscuration and 24/7 operativity. To address the issue, we explore the usage of SAR data over the eastern Iburi sub-prefecture of Hokkaido, Japan. In the area, about 8000 co-seismic landslides were triggered by an Mw 6.6 earthquake on 6 September 2018, at 03.08 local time (JST). In the following study, we modify a Deep Learning (DL) convolutional neural network (CNN) architecture suited for pixel-based classification purposes, the so-called Attention U-Net (Attn-U-Net) and we employ it to evaluate the potential of bi- and tri-temporal SAR amplitude data from the Sentinel-1 satellite and slope angle to map landslides even under thick cloud cover. Four different datasets, composed of two different band combinations per two satellite orbits (ascending and descending) are analyzed. Moreover, the impact of augmentations is evaluated independently for each dataset. The models’ predictions are compared against an accurate landslide inventory obtained by manual mapping on pre-and post-event PlanetScope imagery through F1-score and other common metrics. The best result was yielded by the augmented ascending tri-temporal SAR composite image (61% F1-score). Augmentations have a positive impact on the ascending Sentinel-1 orbit, while metrics decrease when augmentations are applied on descending path. Our findings demonstrate that combining SAR data with other data sources may help to map landslides quickly, even during storms and under deep cloud cover. However, further investigations and improvements are still needed, this being one of the first attempts in which the combination of SAR data and DL algorithms are employed for landslide mapping purposes.
Mapping of landslides over space has seen an increasing attention and good results in the last decade. While current methods are chiefly applied to generate event-inventories, whereas multi-temporal (MT) inventories are rare, even using manual landslide mapping. Here, we present an innovative deep learning strategy which employs transfer learning that allows for the Attention Deep Supervision Multi-Scale U-Net model to be adapted for landslide detection tasks in new areas. The method also provides the flexibility of re-training a pretrained model to detect both rainfall- and earthquake-triggered landslides on new target areas. For the mapping, we used archived Planet Lab remote sensing images spanning a period between 2009 till 2021 with spatial resolution of 3–5 m to systematically generate MT landslide inventories. When we examined all cases, our approach provided an average F1 score of 0.8 indicating that we successfully identified the spatiotemporal occurrences of landslides. To examine the size distribution of mapped landslides we compared the frequency-area distributions of predicted co-seismic landslides with manually mapped products from the literature. Results showed a good match between calculated power-law exponents where the difference ranges between 0.04 and 0.21. Overall, this study showed that the proposed algorithm could be applied to large areas to generate polygon-based MT landslide inventories.
Multiple landslide events occur often across the world which have the potential to cause significant harm to both human life and property. Although a substantial amount of research has been conducted to address mapping of landslides using Earth observation (EO) data, several gaps and uncertainties remain with developing models to be operational at the global scale. The lack of a high-resolution globally distributed and event-diverse dataset for landslide segmentation poses a challenge in developing machine learning models that can accurately and robustly detect landslides in various regions, as the limited representation of landslide and background classes can result in poor generalization performance of the models. To address this issue, we present the High-Resolution Global landslide Detector Database (HR-GLDD), a high-resolution (HR) satellite dataset (PlanetScope, 3 m pixel resolution) for landslide mapping composed of landslide instances from 10 different physiographical regions globally in South and South-East Asia, East Asia, South America, and Central America. The dataset contains five rainfall-triggered and five earthquake-triggered multiple landslide events that occurred in varying geomorphological and topographical regions in the form of standardized image patches containing four PlanetScope image bands (red, green, blue, and NIR) and a binary mask for landslide detection. The HR-GLDD can be accessed through this link: https://doi.org/10.5281/zenodo.7189381 (Meena et al., 2022a, c). HR-GLDD is one of the first datasets for landslide detection generated by high-resolution satellite imagery which can be useful for applications in artificial intelligence for landslide segmentation and detection studies. Five state-of-the-art deep learning models were used to test the transferability and robustness of the HR-GLDD. Moreover, three recent landslide events were used for testing the performance and usability of the dataset to comment on the detection of newly occurring significant landslide events. The deep learning models showed similar results when testing the HR-GLDD at individual test sites, thereby indicating the robustness of the dataset for such purposes. The HR-GLDD is open access and it has the potential to calibrate and develop models to produce reliable inventories using high-resolution satellite imagery after the occurrence of new significant landslide events. The HR-GLDD will be updated regularly by integrating data from new landslide events.
21 April 2022 in Natural Hazards and Earth System Sciences
In the domain of landslide risk science, landslide susceptibility mapping (LSM) is very important, as it helps spatially identify potential landslide-prone regions. This study used a statistical ensemble model (frequency ratio and evidence belief function) and two machine learning (ML) models (random forest and XGBoost; eXtreme Gradient Boosting) for LSM in the province of Belluno (region of Veneto, northeastern Italy). The study investigated the importance of the conditioning factors in predicting landslide occurrences using the mentioned models. In this paper, we evaluated the importance of the conditioning factors in the overall prediction capabilities of the statistical and ML algorithms. By the trial-and-error method, we eliminated the least “important” features by using a common threshold of 0.30 for statistical and 0.03 for ML algorithms. Conclusively, we found that removing the least important features does not impact the overall accuracy of LSM for all three models. Based on the results of our study, the most commonly available features, for example, the topographic features, contributes to comparable results after removing the least important ones, namely the aspect plan and profile curvature, topographic wetness index (TWI), topographic roughness index (TRI), and normalized difference vegetation index (NDVI) in the case of the statistical model and the plan and profile curvature, TWI, and topographic position index (TPI) for ML algorithms. This confirms that the requirement for the important conditioning factor maps can be assessed based on the physiography of the region.