Evaluation of neural network models for landslide susceptibility assessment

International Journal of Digital Earth, Journal Year: 2022, Volume and Issue: 15(1), P. 934 - 953

Published: June 19, 2022

Identifying and assessing the disaster risk of landslide-prone regions is very critical for prevention mitigation. Owning to their special advantages, neural network algorithms have been widely used landslide susceptibility mapping (LSM) in recent decades. In present study, three advanced models popularly relevant studies, i.e. artificial (ANN), one dimensional convolutional (1D CNN) recurrent (RNN), were evaluated compared LSM practice over Qingchuan County, Sichuan province, China. Extensive experimental results demonstrated satisfactory performances these accurately predicting susceptible regions. Specifically, ANN 1D CNN yielded quite consistent but slightly differed from those RNN model spatially. Nevertheless, accuracy evaluations revealed that outperformed other two both qualitatively quantitatively its complexity was relatively high. Experiments concerning training hyper-parameters on performance suggested small batch size values with Tanh activation function SGD optimizer are essential improve LSM, which may provide a thread help who apply efficiency.

Language: Английский

---

Unraveling the evolution of landslide susceptibility: a systematic review of 30-years of strategic themes and trends

Geocarto International, Journal Year: 2023, Volume and Issue: 38(1)

Published: Sept. 5, 2023

:Landslide susceptibility mapping (LSM) research is critical for preventing and mitigating regional landslide disasters. Despite its importance, few researchers have systematically analyzed the key areas of LSM's development. SciMAT, a scientometric tool, offers possibility graphically displaying hotspot themes their evolutionary trends. In this study, We searched Web Science core collection database literature on LSM published from 1993 to 2022 with search term "TI=(landslide susceptibility)". The type language were limited "article" "English". After removing duplicate irrelevant data, total 1661 papers obtained. To analyze retrieved literature, we employed bibliometric VOSviewer SciMAT. Innovatively, conducted cluster analysis thematic evolution using which revealed popular trends in susceptibility. results showed an upward trend publications over past 30 years. Landslide modeling methods, geological information, landslide-triggering factors topics interest. methods primary knowledge path, related appearing most frequently as essential nodes map. There notable widespread towards utilizing machine learning deep techniques achieve precise risk zonation research. application artificial intelligence (AI)-based has gained significant popularity due consistently high accuracy rates, often surpassing 90 percent, evidenced numerous studies. Particularly recent years, advent big data era, Convolutional Neural Network (CNN)-based approaches emerged dominant theme, showcasing exceptional fitting capabilities robust predictive performance. study provides valuable references scholars identify gaps highlight directions, inform policy decision-making

Language: Английский

---

Landslide mapping based on a hybrid CNN-transformer network and deep transfer learning using remote sensing images with topographic and spectral features

International Journal of Applied Earth Observation and Geoinformation, Journal Year: 2023, Volume and Issue: 126, P. 103612 - 103612

Published: Dec. 16, 2023

Landslides frequently cause serious property damage and casualties. Therefore, it is crucial to have rapid accurate landslide mapping (LM) support post-earthquake assessment emergency rescue efforts. Many studies been conducted in recent years on the application of automatic LM methods using remote sensing images (RSIs). However, existing face challenges accurately distinguishing landslides due problems large differences features scales among landslides, as well similarities different ground objects optical RSIs. Here, we propose a semantic segmentation model called SCDUNet++, which combines advantages convolutional neural network (CNN) transformer enhance discrimination extraction features. Then, constructed multi-channel dataset Luding Jiuzhaigou earthquake areas Sentinel-2 NASADEM data. We evaluated performance SCDUNet++ this dataset. The results showed that can extract fuse spectral topographic information more effectively. Compared with other state-of-the-art models, achieved highest IoU F1 score all four test areas. In addition, models significant improvements area after knowledge transfer fine-tuning. direct prediction, eight namely DeepLabv3+, Segformer, TransUNet, SwinUNet, STUNet, UNet, UNet++, demonstrated ranging from 8.33% 27.5% 6.58% 23.67% implementing deep learning (DTL). This finding highlights practicality DTL for cross-domain data-poor

Language: Английский

---

Influence of anthropogenic activities on landslide susceptibility: A case study in Solan district, Himachal Pradesh, India

Journal of Mountain Science, Journal Year: 2023, Volume and Issue: 20(2), P. 429 - 447

Published: Feb. 1, 2023

Language: Английский

---

Comparison of bias-corrected multisatellite precipitation products by deep learning framework

International Journal of Applied Earth Observation and Geoinformation, Journal Year: 2023, Volume and Issue: 116, P. 103177 - 103177

Published: Jan. 3, 2023

Despite satellite-based precipitation products (SPPs) providing a worldwide span with high spatial and temporal resolution, their efficiency in disaster risk forecasting, hydrological, watershed management remains challenge due to the significant dependence of rainfall on spatiotemporal pattern geographical features each area. This research proposes an effective deep learning-based solution that combines convolutional neural network benefit encoder-decoder architecture eliminate pixel-by-pixel bias enhance accuracy daily SPPs. work uses five gridded products, four which are (TRMM, CMORPH, CHIRPS, PERSIANN-CDR) one is gauge-based (APHRODITE). The Lancang-Mekong River Basin (LMRB), international basin, was chosen as region because its diverse climate spread spanning six countries. According results analyses, TRMM product exhibits better performance than other three learning model proved efficacy by successfully reducing spatial–temporal gap between SPPs APHRODITE. In addition, ADJ-TRMM performed best corrected items, followed ADJ-CDR ADJ-CHIRPS products. study's findings indicate SPP has advantages disadvantages across LMRB. aftermath discontinuation APHRODITE 2015, we believe framework will be for generating more up-to-date dependable dataset LMRB research.

Language: Английский

---