Ensemble learning framework for landslide susceptibility mapping: Different basic classifier and ensemble strategy

Geoscience Frontiers, Journal Year: 2023, Volume and Issue: 14(6), P. 101645 - 101645

Published: June 7, 2023

The application of ensemble learning models has been continuously improved in recent landslide susceptibility research, but most studies have no unified framework. Moreover, few papers discussed the applicability model mapping at township level. This study aims defining a robust framework that can become benchmark method for future research dealing with comparison different models. For this purpose, present work focuses on three basic classifiers: decision tree (DT), support vector machine (SVM), and multi-layer perceptron neural network (MLPNN) two homogeneous such as random forest (RF) extreme gradient boosting (XGBoost). hierarchical construction deep relied leading technologies (i.e., homogeneous/heterogeneous bagging, boosting, stacking strategy) to provide more accurate effective spatial probability occurrence. selected area is Dazhou town, located Jurassic red-strata Three Gorges Reservoir Area China, which strategic economic currently characterized by widespread risk. Based long-term field investigation, inventory counting thirty-three slow-moving polygons was drawn. results show do not necessarily perform better; instance, Bagging based DT-SVM-MLPNN-XGBoost performed worse than single XGBoost model. Amongst eleven tested models, Stacking RF-XGBoost model, ensemble, showed highest capability predicting landslide-affected areas. Besides, factor behaviors DT, SVM, MLPNN, RF reflected characteristics landslides reservoir area, wherein unfavorable lithological conditions intense human engineering activities water level fluctuation, residential construction, farmland development) are proven be key triggers. presented approach could used occurrence prediction similar regions other fields.

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

---

Integrating Machine Learning Ensembles for Landslide Susceptibility Mapping in Northern Pakistan

Remote Sensing, Journal Year: 2024, Volume and Issue: 16(6), P. 988 - 988

Published: March 12, 2024

Natural disasters, notably landslides, pose significant threats to communities and infrastructure. Landslide susceptibility mapping (LSM) has been globally deemed as an effective tool mitigate such threats. In this regard, study considers the northern region of Pakistan, which is primarily susceptible landslides amid rugged topography, frequent seismic events, seasonal rainfall, carry out LSM. To achieve goal, pioneered fusion baseline models (logistic regression (LR), K-nearest neighbors (KNN), support vector machine (SVM)) with ensembled algorithms (Cascade Generalization (CG), random forest (RF), Light Gradient-Boosting Machine (LightGBM), AdaBoost, Dagging, XGBoost). With a dataset comprising 228 landslide inventory maps, employed classifier correlation-based feature selection (CFS) approach identify twelve most parameters instigating landslides. The evaluated included slope angle, elevation, aspect, geological features, proximity faults, roads, streams, was revealed primary factor influencing distribution, followed by aspect rainfall minute margin. models, validated AUC 0.784, ACC 0.912, K 0.394 for logistic well 0.907, 0.927, 0.620 XGBoost, highlight practical effectiveness potency results superior performance LR among XGBoost ensembles, contributed development precise LSM area. may serve valuable guiding risk-mitigation strategies policies in geohazard-prone regions at national global scales.

Language: Английский

---

Assessment Analysis of Flood Susceptibility in Tropical Desert Area: A Case Study of Yemen

Remote Sensing, Journal Year: 2022, Volume and Issue: 14(16), P. 4050 - 4050

Published: Aug. 19, 2022

Flooding is one of the catastrophic natural hazards worldwide that can easily cause devastating effects on human life and property. Remote sensing devices are becoming increasingly important in monitoring assessing disaster susceptibility hazards. The proposed research work pursues an assessment analysis flood a tropical desert environment: case study Yemen. base data for this were collected organized from meteorological, satellite images, remote data, essential geographic various sources used as input into four machine learning (ML) algorithms. In study, RS (Sentinel-1 images) to detect flooded areas area. We also Sentinel application platform (SNAP 7.0) Sentinel-1 image detecting zones locations. Flood spots discovered verified using Google Earth Landsat press create inventory map Four ML algorithms flash (FFS) Tarim city (Yemen): K-nearest neighbor (KNN), Naïve Bayes (NB), random forests (RF), eXtreme gradient boosting (XGBoost). Twelve conditioning factors prepared, assessed multicollinearity, with inventories parameters run each model. A total 600 non-flood points chosen, where 75% 25% training validation datasets. confusion matrix area under receiver operating characteristic curve (AUROC) validate maps. results obtained reveal all models had high capacity predict floods (AUC > 0.90). Further, terms performance, tree-based ensemble (RF, XGBoost) outperform other algorithms, RF algorithm provides robust performance = 0.982) flood-prone only few adjustments required prior value lies fact being tested first time Yemen assess susceptibility, which be assess, example, earthquakes, landslides, disasters. Furthermore, makes significant contributions effort reduce risk disasters, particularly This will, therefore, help enhance environmental sustainability.

Language: Английский

---

Urban flood risk assessment using AHP and geospatial techniques in swat Pakistan

SN Applied Sciences, Journal Year: 2023, Volume and Issue: 5(8)

Published: July 22, 2023

Abstract The rapid urbanization and changing climate patterns in Swat, Pakistan have increased the vulnerability of urban areas to flood events. Accurate assessment risk is crucial for effective planning disaster management. In current research study hazard index was developed using analytic hierarchy process (AHP) technique combination with geographical information system (GIS) environment Pakistan. integrates various data sources, including topographic maps, land use/land cover information, rainfall data, infrastructure develop a comprehensive model. weights obtained from AHP analysis are combined geospatial geographic generate maps. levels were categorized into five distinct classes: very low, moderate, high, high. Using GIS-AHP approach, higher assigned rainfall, distance river, elevation, slope comparison NDVI, TWI, LULC, curvature, soil type. map then reclassified each parameter. By overlaying these it determined that 5.6% total area classified as high risk, 52% 39.3% moderate 3.1% low risk. model can identify high-risk areas, prioritize mitigation measures, aid

Language: Английский

---

Modeling landslide susceptibility based on convolutional neural network coupling with metaheuristic optimization algorithms

International Journal of Digital Earth, Journal Year: 2023, Volume and Issue: 16(1), P. 3384 - 3416

Published: Aug. 23, 2023

Landslides are one of the most common geological hazards worldwide, especially in Sichuan Province (Southwest China). The current study's main purposes to explore potential applications convolutional neural networks (CNN) hybrid ensemble metaheuristic optimization algorithms, namely beluga whale (BWO) and coati algorithm (COA), for landslide susceptibility mapping (China). For this aim, fourteen conditioning factors were compiled a spatial database. effectiveness development predictive model was quantified using linear support vector machine model. receiver operating characteristic (ROC) curve (AUC), root mean square error, six statistical indices used test compare three resultant models. training dataset, AUC values CNN-COA, CNN-BWO CNN models 0.946, 0.937 0.855, respectively. In terms validation CNN-COA exhibited higher value 0.919, while 0.906 0.805, results indicate that model, followed by offers best overall performance analysis.

Language: Английский

---

Uncertainty Reduction in Flood Susceptibility Mapping Using Random Forest and eXtreme Gradient Boosting Algorithms in Two Tropical Desert Cities, Shibam and Marib, Yemen

Remote Sensing, Journal Year: 2024, Volume and Issue: 16(2), P. 336 - 336

Published: Jan. 15, 2024

Flooding is a natural disaster that coexists with human beings and causes severe loss of life property worldwide. Although numerous studies for flood susceptibility modelling have been introduced, notable gap has the overlooked or reduced consideration uncertainty in accuracy produced maps. Challenges such as limited data, due to confidence bounds, overfitting problem are critical areas improving accurate models. We focus on mapping, mainly when there significant variation predictive relevance predictor factors. It also noted receiver operating characteristic (ROC) curve may not accurately depict sensitivity resulting map overfitting. Therefore, reducing was targeted increase improve processing time prediction. This study created spatial repository test models, containing data from historical flooding twelve topographic geo-environmental conditioning variables. Then, we applied random forest (RF) extreme gradient boosting (XGB) algorithms susceptibility, incorporating variable drop-off empirical loop function. The results showed function crucial method resolve model associated factors methods. approximately 8.42% 9.89% Marib City 9.93% 15.69% Shibam were highly vulnerable floods. Furthermore, this significantly contributes worldwide endeavors focused hazards linked disasters. approaches used can offer valuable insights strategies risks, particularly Yemen.

Language: Английский

---

Flash Flood Susceptibility Modelling Using Soft Computing-Based Approaches: From Bibliometric to Meta-Data Analysis and Future Research Directions

Water, Journal Year: 2024, Volume and Issue: 16(1), P. 173 - 173

Published: Jan. 3, 2024

In recent years, there has been a growing interest in flood susceptibility modeling. this study, we conducted bibliometric analysis followed by meta-data to capture the nature and evolution of literature, intellectual structure networks, emerging themes, knowledge gaps Relevant publications were retrieved from Web Science database identify leading authors, influential journals, trending articles. The results indicated that hybrid models most frequently used prediction models. Results show GIS, machine learning, statistical models, analytical hierarchy process central focuses research area. also revealed slope, elevation, distance river are commonly factors present study discussed importance resolution input data, size representation training sample, other lessons learned, future directions field.

Language: Английский

---

Evaluation and prediction of compound geohazards in highly urbanized regions across China's Greater Bay Area

Journal of Cleaner Production, Journal Year: 2024, Volume and Issue: 449, P. 141641 - 141641

Published: March 16, 2024

Language: Английский

---

Leveraging GIS-based AHP, remote sensing, and machine learning for susceptibility assessment of different flood types in peshawar, Pakistan

Journal of Environmental Management, Journal Year: 2024, Volume and Issue: 371, P. 123094 - 123094

Published: Nov. 2, 2024

Language: Английский

---