Geospatial Assessment and Mapping Landslide Susceptibility for the Garo Hills Division, Meghalaya, India

Geological Journal, Год журнала: 2025, Номер 60(5), С. 1184 - 1201

Опубликована: Фев. 19, 2025

ABSTRACT Creating accurate and effective Landslide Susceptibility (LS) maps can aid disaster prevention mitigation efforts provide sufficient public safety. The primary aim of this study is to develop an LS map for the Garo Hills region in Meghalaya, India, using weight evidence (WoE), frequency ratio (FR), Shannon entropy (SE) methods. A comprehensive landslide inventory catalogued 98 events from 2000 2023 analysis, nine key geographical environmental parameters were prepared. Conducted multicollinearity correlation analysis identify mitigate collinearity issues between factors. model's performance was analysed through area under curve (AUC) value receiver operating characteristic (ROC) curves three recent landslides. results showed that FR method achieved highest accuracy, with successive rate (SRC) AUC predictive (PRC) values 0.860 0.940, respectively, classified susceptibility at sites as high, moderate, low. WoE effectively identified landslides site high very zones, achieving SRC PRC 0.844 0.915, respectively. SE robust predicting landslide‐prone areas, comparable other methods (0.913), though its (0.771) lower. Developed revealed zones account approximately 10% 3% area, predominantly near roads, steep slopes, higher elevations. information valuable civilians government authorities involved hazard monitoring management.

Язык: Английский

---

A new procedure for optimizing neural network using stochastic algorithms in predicting and assessing landslide risk in East Azerbaijan

Stochastic Environmental Research and Risk Assessment, Год журнала: 2024, Номер unknown

Опубликована: Март 21, 2024

Язык: Английский

---

Ensemble learning landslide susceptibility assessment with optimized non-landslide samples selection

Geomatics Natural Hazards and Risk, Год журнала: 2024, Номер 15(1)

Опубликована: Июль 17, 2024

Non-landslide samples influence the outcomes of landslide susceptibility assessment. Existing studies did not fully consider equilibrium between and non-landslide in similar environments, resulting poor reliability This study proposed a optimization method with constraint disaster-pregnant environment similarity to construct balanced samples. We employed heterogeneous stacking blending ensemble learning models generate focused on Bailong River Basin complex frequent landslides as area. First, we extracted 12 influencing factors based multiple sources analyzed spatial distribution patterns landslides. Second, constructed environments assessment units obtained from curvature watershed selected an equal amount both every different environment. Finally, three classic neural network models, namely multilayer perceptron, convolutional network, gated recurrent unit were used base for assess susceptibility. The findings suggested that results optimized more reliable, especially improved prediction sample-sparse regions. this demonstrated highest area under curve 0.88 testing dataset, outperforming models. issue unreliable regions within can be effectively addressed by considering sampling environments.

Язык: Английский

---

Integrated Machine Learning Approaches for Landslide Susceptibility Mapping Along the Pakistan–China Karakoram Highway

Land, Год журнала: 2025, Номер 14(1), С. 172 - 172

Опубликована: Янв. 15, 2025

The effectiveness of data-driven landslide susceptibility mapping relies on data integrity and advanced geospatial analysis; however, selecting the most suitable method identifying key regional factors remains a challenging task. To address this, this study assessed performance six machine learning models, including Convolutional Neural Networks (CNNs), Random Forest (RF), Categorical Boosting (CatBoost), their CNN-based hybrid models (CNN+RF CNN+CatBoost), Stacking Ensemble (SE) combining CNN, RF, CatBoost in along Karakoram Highway northern Pakistan. Twelve were examined, categorized into Topography/Geomorphology, Land Cover/Vegetation, Geology, Hydrology, Anthropogenic Influence. A detailed inventory 272 occurrences was compiled to train models. proposed stacking ensemble improve modeling, with achieving an AUC 0.91. Hybrid modeling enhances accuracy, CNN–RF boosting RF’s from 0.85 0.89 CNN–CatBoost increasing CatBoost’s 0.87 0.90. Chi-square (χ2) values (9.8–21.2) p-values (<0.005) confirm statistical significance across This identifies approximately 20.70% area as high very risk, SE model excelling detecting high-risk zones. Key influencing showed slight variations while multicollinearity among variables remained minimal. approach reduces uncertainties, prediction supports decision-makers implementing effective mitigation strategies.

Язык: Английский

---

Advancements in mapping landslide susceptibility in Bafoussam and its surroundings area using multi-criteria decision analysis, statistical methods, and machine learning models

Journal of African Earth Sciences, Год журнала: 2024, Номер 213, С. 105237 - 105237

Опубликована: Март 22, 2024

Язык: Английский

---

Stacking Ensemble Technique Using Optimized Machine Learning Models with Boruta–XGBoost Feature Selection for Landslide Susceptibility Mapping: A Case of Kermanshah Province, Iran

Information, Год журнала: 2024, Номер 15(11), С. 689 - 689

Опубликована: Ноя. 2, 2024

Landslides cause significant human and financial losses in different regions of the world. A high-accuracy landslide susceptibility map (LSM) is required to reduce adverse effects landslides. Machine learning (ML) a robust tool for LSM creation. ML models require large amounts data predict landslides accurately. This study has developed stacking ensemble technique based on optimization enhance accuracy an while considering small datasets. The Boruta–XGBoost feature selection was used determine optimal combination features. Then, intelligent accurate analysis performed prepare using dynamic hybrid approach Adaptive Fuzzy Inference System (ANFIS), Extreme Learning (ELM), Support Vector Regression (SVR), new algorithms (Ladybug Beetle Optimization [LBO] Electric Eel Foraging [EEFO]). After model optimization, weight combine outputs increase reliability LSM. combinations were optimized LBO EEFO. Root Mean Square Error (RMSE) Area Under Receiver Operating Characteristic Curve (AUC-ROC) parameters assess performance these models. dataset from Kermanshah province, Iran, 17 influencing factors evaluate proposed approach. Landslide inventory 116 points, combined Voronoi entropy method applied non-landslide point sampling. results showed higher with EEFO AUC-ROC values 94.81% 94.84% RMSE 0.3146 0.3142, respectively. can help managers planners reliable LSMs and, as result, associated events.

Язык: Английский

---

Integrating Knowledge Graph and Machine Learning Methods for Landslide Susceptibility Assessment

Remote Sensing, Год журнала: 2024, Номер 16(13), С. 2399 - 2399

Опубликована: Июнь 29, 2024

The suddenness of landslide disasters often causes significant loss life and property. Accurate assessment disaster susceptibility is great significance in enhancing the ability accurate prevention. To address problems strong subjectivity selection indicators low efficiency process caused by insufficient application a priori knowledge assessment, this paper, we propose novel framework combing domain graph machine learning algorithms. Firstly, combine unstructured data, extract based on Unified Structure Generation for Universal Information Extraction Pre-trained model (UIE) fine-tuned with small amount labeled data to construct graph. We use Paired Relation Vectors (PairRE) characterize graph, then target area characterization factor recommendation calculating spatial correlation, attribute similarity, Term Frequency–Inverse Document Frequency (TF-IDF) metrics. select optimal feature combination among six typical (ML) models interpretable mapping. Experimental validation analysis are carried out three gorges (TGA), results show effectiveness factors recommended learning, overall accuracy after adding associated reaching 87.2%. methodology proposed research better contribution data-driven susceptibility.

Язык: Английский

---

A Graph–Transformer Method for Landslide Susceptibility Mapping

IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, Год журнала: 2024, Номер 17, С. 14556 - 14574

Опубликована: Янв. 1, 2024

Landslide susceptibility mapping (LSM) is of great significance for regional land resource planning and disaster prevention reduction. The machine learning (ML) method has been widely used in the field LSM. However, existing LSM model fails to consider correlation between landslide disaster-prone environment (DPE) lacks global information, resulting a high false alarm rate Therefore, we propose an with GraphTransformer that considers DPE characteristics information. Firstly, analysis importance are employed on nine contributing factors (LCFs), dataset generated by combining remote sensing image interpretation verification. Secondly, graph constrained similarity relationship constructed realize DPE. Then, Transformer module introduced construct Graph-Transformer Finally, analyzed, accuracy proposed compared evaluated. experimental results show effectively improves models weakens influence environmental differences models. Compared convolutional network (GCN), sample aggregate (GraphSAGE), attention (GAT) models, AUC value more than 2.05% higher under relationship. In addition, 8.8% traditional ML conclusion, our framework can get better evaluation most methods, its provide effective ways key technical support investigation control

Язык: Английский

---

Risk Assessment of Multi-Hazards in Hangzhou: A Socioeconomic and Risk Mapping Approach Using the CatBoost-SHAP Model

International Journal of Disaster Risk Science, Год журнала: 2024, Номер 15(4), С. 640 - 656

Опубликована: Авг. 1, 2024

Abstract As the global push for sustainable urban development progresses, this study, set against backdrop of Hangzhou City, one China’s megacities, addressed conflict between expansion and occurrence geological hazards. Focusing on predominant hazards troubling Hangzhou—urban road collapse, land subsidence, karst collapse—we introduced a Categorical Boosting-SHapley Additive exPlanations (CatBoost-SHAP) model. This model not only demonstrates strong performance in predicting selected typical hazards, with area under curve (AUC) values reaching 0.92, 0.94, respectively, but also, through incorporation explainable SHAP, visually presents prediction process, interrelations evaluation factors, weight each factor. Additionally, study undertook multi-hazard evaluation, producing susceptibility zoning map multiple while performing tailored analysis by integrating economic population density factors Hangzhou. research enables decision makers to transcend “black box” limitations machine learning, facilitating informed making strategic resource allocation scheduling based demographic area. approach holds potential offer valuable insights cities worldwide.

Язык: Английский

---

A Novel Strategy Coupling Optimised Sampling with Heterogeneous Ensemble Machine-Learning to Predict Landslide Susceptibility

Remote Sensing, Год журнала: 2024, Номер 16(19), С. 3663 - 3663

Опубликована: Окт. 1, 2024

The accuracy of data-driven landslide susceptibility prediction depends heavily on the quality non-landslide samples and selection machine-learning algorithms. Current methods rely artificial prior knowledge to obtain negative from landslide-free regions or outside buffer zones randomly quickly but often ignore reliability samples, which will pose a serious risk including potential landslides lead erroneous outcomes in training data. Furthermore, diverse models exhibit distinct classification capabilities, applying single model can readily result over-fitting dataset introduce uncertainties predictions. To address these problems, taking Chenxi County, hilly mountainous area southern China, as an example, this research proposes strategy-coupling optimised sampling with heterogeneous ensemble machine learning enhance prediction. Initially, 21 impact factors were derived five aspects: geology, hydrology, topography, meteorology, human activities, geographical environment. Then, screened through correlation analysis collinearity diagnosis. Afterwards, (OS) method was utilised select by fusing certainty factor values basis environmental similarity statistical model. Subsequently, adopted historical combined create datasets. Finally, baseline (support vector machine, random forest, back propagation neural network) stacking employed predict susceptibility. findings indicated that OS method, considering achieved higher-quality than currently widely used methods. outperformed those three models. Notably, hybrid OS–Stacking is most promising, up 97.1%. integrated strategy significantly improves makes it reliable effective for assessing regional geohazard risk.

Язык: Английский

---