Integrating machine learning and geospatial data analysis for comprehensive flood hazard assessment

Environmental Science and Pollution Research, Год журнала: 2024, Номер 31(35), С. 48497 - 48522

Опубликована: Июль 20, 2024

Flooding is a major natural hazard worldwide, causing catastrophic damage to communities and infrastructure. Due climate change exacerbating extreme weather events robust flood modeling crucial support disaster resilience adaptation. This study uses multi-sourced geospatial datasets develop an advanced machine learning framework for assessment in the Arambag region of West Bengal, India. The inventory was constructed through Sentinel-1 SAR analysis global databases. Fifteen conditioning factors related topography, land cover, soil, rainfall, proximity, demographics were incorporated. Rigorous training testing diverse models, including RF, AdaBoost, rFerns, XGB, DeepBoost, GBM, SDA, BAM, monmlp, MARS algorithms, undertaken categorical mapping. Model optimization achieved statistical feature selection techniques. Accuracy metrics model interpretability methods like SHAP Boruta implemented evaluate predictive performance. According area under receiver operating characteristic curve (AUC), prediction accuracy models performed around > 80%. RF achieves AUC 0.847 at resampling factor 5, indicating strong discriminative AdaBoost also consistently exhibits good ability, with values 0.839 10. indicated precipitation elevation as most significantly contributing area. Most pointed out southern portions highly susceptible areas. On average, from 17.2 18.6% hazards. In analysis, various nature-inspired algorithms identified selected input parameters assessment, i.e., elevation, precipitation, distance rivers, TWI, geomorphology, lithology, TRI, slope, soil type, curvature, NDVI, roads, gMIS. As per analyses, it found that rivers play roles decision-making process assessment. results majority building footprints (15.27%) are high very risk, followed by those low risk (43.80%), (24.30%), moderate (16.63%). Similarly, cropland affected flooding this categorized into five classes: (16.85%), (17.28%), (16.07%), (16.51%), (33.29%). However, interdisciplinary contributes towards hydraulic hydrological management.

Язык: Английский

---

Improving Long-Term Flood Forecasting Accuracy Using Ensemble Deep Learning Models and an Attention Mechanism

Journal of Hydrologic Engineering, Год журнала: 2024, Номер 29(6)

Опубликована: Сен. 14, 2024

Язык: Английский

---

SHAP-Driven Explainable Artificial Intelligence Framework for Wildfire Susceptibility Mapping Using MODIS Active Fire Pixels: An In-Depth Interpretation of Contributing Factors in Izmir, Türkiye

Remote Sensing, Год журнала: 2024, Номер 16(15), С. 2842 - 2842

Опубликована: Авг. 2, 2024

Wildfire susceptibility maps play a crucial role in preemptively identifying regions at risk of future fires and informing decisions related to wildfire management, thereby aiding mitigating the risks potential damage posed by wildfires. This study employs eXplainable Artificial Intelligence (XAI) techniques, particularly SHapley Additive exPlanations (SHAP), map Izmir Province, Türkiye. Incorporating fifteen conditioning factors spanning topography, climate, anthropogenic influences, vegetation characteristics, machine learning (ML) models (Random Forest, XGBoost, LightGBM) were used predict wildfire-prone areas using freely available active fire pixel data (MODIS Active Fire Collection 6 MCD14ML product). The evaluation trained ML showed that Random Forest (RF) model outperformed XGBoost LightGBM, achieving highest test accuracy (95.6%). All classifiers demonstrated strong predictive performance, but RF excelled sensitivity, specificity, precision, F-1 score, making it preferred for generating conducting SHAP analysis. Unlike prevailing approaches focusing solely on global feature importance, this fills critical gap employing summary dependence plots comprehensively assess each factor’s contribution, enhancing explainability reliability results. analysis reveals clear associations between such as wind speed, temperature, NDVI, slope, distance villages with increased susceptibility, while rainfall streams exhibit nuanced effects. spatial distribution classes highlights areas, flat coastal near settlements agricultural lands, emphasizing need enhanced awareness preventive measures. These insights inform targeted management strategies, highlighting importance tailored interventions like firebreaks management. However, challenges remain, including ensuring selected factors’ adequacy across diverse regions, addressing biases from resampling spatially varied data, refining broader applicability.

Язык: Английский

---

A comprehensive framework for assessing the spatial drivers of flood disasters using an optimal Parameter-based geographical Detector–machine learning coupled model

Geoscience Frontiers, Год журнала: 2024, Номер 15(6), С. 101889 - 101889

Опубликована: Июль 11, 2024

Flood disasters pose serious threats to human life and property worldwide. Exploring the spatial drivers of flood on a macroscopic scale is great significance for mitigating their impacts. This study proposes comprehensive framework integrating driving-factor optimization interpretability, while considering heterogeneity. In this framework, Optimal Parameter-based Geographic Detector (OPGD), Recursive Feature Estimation (RFE), Light Gradient Boosting Machine (LGBM) models were utilized construct OPGD–RFE–LGBM coupled model identify essential driving factors simulate distribution disasters. The SHapley Additive ExPlanation (SHAP) interpreter was employed quantitatively explain mechanisms behind Yunnan Province, typical mountainous plateau area in Southwest China, selected implement proposed conduct case study. For purpose, disaster inventory 7332 historical events prepared, 22 potential related precipitation, surface environment, activity initially selected. Results revealed that Province exhibit high heterogeneity, with geomorphic zoning accounting 66.1% variation offers clear advantages over single LGBM identifying analyzing Moreover, simulation performance shows slight improvement (a 6% average decrease RMSE an increase 1% R2) even reduced factor data. Factor explanatory analysis indicated combination sets varied across different subregions; nevertheless, precipitation-related factors, such as precipitation intensity index (SDII), wet days (R10MM), 5-day maximum (RX5day), main controlling provides quantitative analytical at large scales significant offering reference management authorities developing macro-strategies prevention.

Язык: Английский

---

Explainable Sinkhole Susceptibility Mapping Using Machine-Learning-Based SHAP: Quantifying and Comparing the Effects of Contributing Factors in Konya, Türkiye

Applied Sciences, Год журнала: 2025, Номер 15(6), С. 3139 - 3139

Опубликована: Март 13, 2025

Sinkholes, naturally occurring formations in karst regions, represent a significant environmental hazard, threatening infrastructure, agricultural lands, and human safety. In recent years, machine learning (ML) techniques have been extensively employed for sinkhole susceptibility mapping (SSM). However, the lack of explainability inherent these methods remains critical issue decision-makers. this study, Konya Closed Basin was mapped using an interpretable model based on SHapley Additive exPlanations (SHAP). The Random Forest (RF), eXtreme Gradient Boosting (XGBoost), Light Machine (LightGBM) algorithms were employed, interpretability results enhanced through SHAP analysis. Among compared models, RF demonstrated highest performance, achieving accuracy 95.5% AUC score 98.8%, consequently selected development final map. analyses revealed that factors such as proximity to fault lines, mean annual precipitation, bicarbonate concentration difference are most variables influencing formation. Additionally, specific threshold values quantified, effects contributing analyzed detail. This study underscores importance employing eXplainable Artificial Intelligence (XAI) natural hazard modeling, SSM example, thereby providing decision-makers with more reliable comparable risk assessment.

Язык: Английский

---

On Robustness of the Explanatory Power of Machine Learning Models: Insights From a New Explainable AI Approach Using Sensitivity Analysis

Water Resources Research, Год журнала: 2025, Номер 61(3)

Опубликована: Март 1, 2025

Abstract Machine learning (ML) is increasingly considered the solution to environmental problems where limited or no physico‐chemical process understanding exists. But in supporting high‐stakes decisions, ability explain possible solutions key their acceptability and legitimacy, ML can fall short. Here, we develop a method, rooted formal sensitivity analysis , uncover primary drivers behind predictions. Unlike many methods for explainable artificial intelligence (XAI), this method (a) accounts complex multi‐variate distributional properties of data, common systems, (b) offers global assessment input‐output response surface formed by ML, rather than focusing solely on local regions around existing data points, (c) scalable data‐size independent, ensuring computational efficiency with large sets. We apply suite models predicting various water quality variables pilot‐scale experimental pit lake. A critical finding that subtle alterations design some (such as variations random seed, functional class, hyperparameters, splitting) lead different interpretations how outputs depend inputs. Further, from families (decision trees, connectionists, kernels) may focus aspects information provided despite displaying similar predictive power. Overall, our results underscore need assess explanatory robustness advocate using model ensembles gain deeper insights into system improve prediction reliability.

Язык: Английский

---

River ecological flow early warning forecasting using baseflow separation and machine learning in the Jiaojiang River Basin, Southeast China

The Science of The Total Environment, Год журнала: 2023, Номер 882, С. 163571 - 163571

Опубликована: Апрель 21, 2023

Язык: Английский

---

A New Graph-Based Deep Learning Model to Predict Flooding with Validation on a Case Study on the Humber River

Water, Год журнала: 2023, Номер 15(10), С. 1827 - 1827

Опубликована: Май 10, 2023

Floods are one of the most lethal natural disasters. It is crucial to forecast timing and evolution these events create an advanced warning system allow for proper implementation preventive measures. This work introduced a new graph-based forecasting model, namely, graph neural network sample aggregate (GNN-SAGE), estimate river flooding. then validated proposed model in Humber River watershed Ontario, Canada. Using past precipitation stage data from reference neighboring stations, GNN-SAGE could flooding up 24 h ahead, improving its performance by average 18% compared with persistence 9% residual gated convolutional (GNN-ResGated), which were used as baselines. Furthermore, generated smaller errors than those reported current literature. The Shapley additive explanations (SHAP) revealed that prior station was significant factor all prediction intervals, seasonality being more influential longer-range forecasts. findings positioned cutting-edge solution flood valuable resource devising early flood-warning systems.

Язык: Английский

---

Applying a 1D Convolutional Neural Network in Flood Susceptibility Assessments—The Case of the Island of Euboea, Greece

Remote Sensing, Год журнала: 2023, Номер 15(14), С. 3471 - 3471

Опубликована: Июль 10, 2023

The main scope of the study is to evaluate prognostic accuracy a one-dimensional convolutional neural network model (1D-CNN), in flood susceptibility assessment, selected test site on island Euboea, Greece. Logistic regression (LR), Naïve Bayes (NB), gradient boosting (GB), and deep learning (DLNN) are benchmark models used compare their performance with that 1D-CNN model. Remote sensing (RS) techniques collect necessary related data, whereas thirteen flash-flood-related variables were as predictive variables, such elevation, slope, plan curvature, profile topographic wetness index, lithology, silt content, sand clay distance faults, river network. Weight Evidence method was applied calculate correlation among flood-related assign weight value each variable class. Regression analysis multi-collinearity assess collinearity Shapley Additive explanations rank features by importance. evaluation process involved estimating ability all via classification accuracy, sensitivity, specificity, area under success rate curves (AUC). outcomes confirmed provided higher (0.924), followed LR (0.904) DLNN (0.899). Overall, 1D-CNNs can be useful tools for analyzing using remote high predictions.

Язык: Английский

---

Novel optimized deep learning algorithms and explainable artificial intelligence for storm surge susceptibility modeling and management in a flood-prone island

Natural Hazards, Год журнала: 2024, Номер 120(6), С. 5099 - 5128

Опубликована: Фев. 2, 2024

Язык: Английский

---