Optimizing landslide susceptibility mapping using machine learning and geospatial techniques

Ecological Informatics, Journal Year: 2024, Volume and Issue: 81, P. 102583 - 102583

Published: March 30, 2024

Landslides present a substantial risk to human lives, the environment, and infrastructure. Consequently, it is crucial highlight regions prone future landslides by examining correlation between past various geo-environmental factors. This study aims investigate optimal data selection machine learning model, or ensemble technique, for evaluating vulnerability of areas determining most accurate approach. To attain our objectives, we considered two different scenarios selecting landslide-free random points (a slope threshold buffer-based approach) performed comparative analysis five models landslide susceptibility mapping, namely: Support Vector Machine (SVM), Logistic Regression (LR), Linear Discriminant Analysis (LDA), Random Forest (RF), Extreme Gradient Boosting (XGBoost). The area this research an in Polk County Western North Carolina that has experienced fatal landslides, leading casualties significant damage infrastructure, properties, road networks. model construction process involves utilization dataset comprising 1215 historical occurrences non-landslide points. We integrated total fourteen geospatial layers, consisting topographic variables, soil data, geological land cover attributes. use metrics assess models' performance, including accuracy, F1-score, Kappa score, AUC-ROC. In addition, used seeded-cell index (SCAI) evaluate map consistency. using Weighted Average produces outstanding results, with AUC-ROC 99.4% scenario 91.8% scenario. Our findings emphasize impact sampling on performance mapping. Furthermore, optimally identifying landslide-prone hotspots need urgent management planning, demonstrates effectiveness analyzing providing valuable insights informed decision-making disaster reduction initiatives.

Language: Английский

---

Improving pixel-based regional landslide susceptibility mapping

Geoscience Frontiers, Journal Year: 2024, Volume and Issue: 15(4), P. 101782 - 101782

Published: Jan. 12, 2024

Regional landslide susceptibility mapping (LSM) is essential for risk mitigation. While deep learning algorithms are increasingly used in LSM, their extensive parameters and scarce labels (limited records) pose training challenges. In contrast, classical statistical algorithms, with typically fewer parameters, less likely to overfit, easier train, offer greater interpretability. Additionally, integrating physics-based data-driven approaches can potentially improve LSM. This paper makes several contributions enhance the practicality, interpretability, cross-regional generalization ability of regional LSM models: (1) Two new hybrid models, composed modules, proposed compared. Hybrid Model I combines infinite slope stability analysis (ISSA) logistic regression, a algorithm. II integrates ISSA convolutional neural network, representative techniques. The module constructs explanatory factor higher nonlinearity reduces prediction uncertainty caused by incomplete inventory pre-selecting non-landslide samples. captures relation between factors inventory. (2) A step-wise deletion process assess importance identify minimum necessary required maintain satisfactory model performance. (3) Single-pixel local-area samples compared understand effect pixel spatial neighborhood. (4) impact on performance explored. Typical landslide-prone regions Three Gorges Reservoir, China, as study area. results show that, testing region, using account neighborhoods, achieves roughly 4.2% increase AUC. Furthermore, models 30 m resolution land-cover data surpass those 1000 data, showing 5.5% improvement optimal set includes elevation, type, safety factor. These findings reveal key elements offering valuable insights practices.

Language: Английский

---

Assessing the imperative of conditioning factor grading in machine learning-based landslide susceptibility modeling: A critical inquiry

CATENA, Journal Year: 2023, Volume and Issue: 236, P. 107732 - 107732

Published: Dec. 7, 2023

Language: Английский

---

Exploring the uncertainty of landslide susceptibility assessment caused by the number of non–landslides

CATENA, Journal Year: 2023, Volume and Issue: 227, P. 107109 - 107109

Published: March 28, 2023

Language: Английский

---

Comparison of hybrid data-driven and physical models for landslide susceptibility mapping at regional scales

Acta Geotechnica, Journal Year: 2023, Volume and Issue: 18(8), P. 4453 - 4476

Published: March 6, 2023

Language: Английский

---

Tempo-Spatial Landslide Susceptibility Assessment from the Perspective of Human Engineering Activity

Remote Sensing, Journal Year: 2023, Volume and Issue: 15(16), P. 4111 - 4111

Published: Aug. 21, 2023

The expansion of mountainous urban areas and road networks can influence the terrain, vegetation, material characteristics, thereby altering susceptibility landslides. Understanding relationship between human engineering activities landslide occurrence is great significance for both prevention land resource management. In this study, an analysis was conducted on caused by Typhoon Megi in 2016. A representative area along eastern coast China—characterized development, deforestation, severe expansion—was used to analyze spatial distribution For purpose, high-precision Planet optical remote sensing images were obtain inventory related event. main innovative features are as follows: (i) newly developed patch generating land-use simulation (PLUS) model simulated analyzed driving factors land-cover (LULC) from 2010 2060; (ii) stacking strategy combined three strong ensemble models—Random Forest (RF), Extreme Gradient Boosting (XGBoost), Light Machine (LightGBM)—to calculate susceptibility; (iii) distance LULC maps short-term long-term dynamic examine impact susceptibility. results show that maximum built-up 2020 13.433 km2, mainly expanding forest cropland land, with 8.28 km2 5.99 respectively. predicted map 2060 shows a growth 45.88 distributed around government residences relatively flat terrain frequent socio-economic activities. factor contribution has higher than LULC. Stacking RF-XGB-LGBM obtained optimal AUC value 0.915 Furthermore, future network have intensified probability landslides occurring 2015. To our knowledge, first application PLUS models international literature. research serve foundation developing management guidelines reduce risk failures.

Language: Английский

---

Improving the model robustness of flood hazard mapping based on hyperparameter optimization of random forest

Expert Systems with Applications, Journal Year: 2023, Volume and Issue: 241, P. 122682 - 122682

Published: Nov. 24, 2023

Language: Английский

---

Effect of landslide spatial representation and raster resolution on the landslide susceptibility assessment

Environmental Earth Sciences, Journal Year: 2024, Volume and Issue: 83(4)

Published: Feb. 1, 2024

Language: Английский

---

Uncertainties of landslide susceptibility prediction: influences of different study area scales and mapping unit scales

International Journal of Coal Science & Technology, Journal Year: 2024, Volume and Issue: 11(1)

Published: April 5, 2024

Abstract This study aims to investigate the effects of different mapping unit scales and area on uncertainty rules landslide susceptibility prediction (LSP). To illustrate various scales, Ganzhou City in China, its eastern region (Ganzhou East), Ruijin County East were chosen. Different are represented by grid units with spatial resolution 30 60 m, as well slope that extracted multi-scale segmentation method. The 3855 locations 21 typical environmental factors first determined create datasets input-outputs. Then, maps (LSMs) City, produced using a support vector machine (SVM) random forest (RF), respectively. LSMs above three regions then mask from LSM along East. Additionally, at generated accordance. Accuracy indexes (LSIs) distribution used express LSP uncertainties. uncertainties under significantly decrease County, whereas those less affected scales. Of course, attentions should also be paid broader representativeness large areas. accuracy increases about 6%–10% compared m same area's scale. significance exhibits an averaging trend scale small large. importance varies greatly unit, but it tends consistent some extent unit. Graphic abstract

Language: Английский

---

Refined landslide susceptibility mapping in township area using ensemble machine learning method under dataset replenishment strategy

Gondwana Research, Journal Year: 2024, Volume and Issue: 131, P. 20 - 37

Published: March 12, 2024

Language: Английский

---