Underutilized Feature Extraction Methods for Burn Severity Mapping: A Comprehensive Evaluation DOI Creative Commons
Linh Nguyen Van, Giha Lee

Remote Sensing, Journal Year: 2024, Volume and Issue: 16(22), P. 4339 - 4339

Published: Nov. 20, 2024

Wildfires increasingly threaten ecosystems and infrastructure, making accurate burn severity mapping (BSM) essential for effective disaster response environmental management. Machine learning (ML) models utilizing satellite-derived vegetation indices are crucial assessing wildfire damage; however, incorporating many can lead to multicollinearity, reducing classification accuracy. While principal component analysis (PCA) is commonly used address this issue, its effectiveness relative other feature extraction (FE) methods in BSM remains underexplored. This study aims enhance ML classifier accuracy by evaluating various FE techniques that mitigate multicollinearity among indices. Using composite index (CBI) data from the 2014 Carlton Complex fire United States as a case study, we extracted 118 seven Landsat-8 spectral bands. We applied compared 13 different techniques—including linear nonlinear such PCA, t-distributed stochastic neighbor embedding (t-SNE), discriminant (LDA), Isomap, uniform manifold approximation projection (UMAP), factor (FA), independent (ICA), multidimensional scaling (MDS), truncated singular value decomposition (TSVD), non-negative matrix factorization (NMF), locally (LLE), (SE), neighborhood components (NCA). The performance of these was benchmarked against six classifiers determine their improving Our results show alternative outperform computational efficiency. Techniques like LDA NCA effectively capture relationships critical BSM. contributes existing literature providing comprehensive comparison methods, highlighting potential benefits underutilized

Language: Английский

A review of hybrid deep learning applications for streamflow forecasting DOI
Kin‐Wang Ng, Yuk Feng Huang, Chai Hoon Koo

et al.

Journal of Hydrology, Journal Year: 2023, Volume and Issue: 625, P. 130141 - 130141

Published: Sept. 12, 2023

Language: Английский

Citations

81
Enhancing wildfire mapping accuracy using mono-temporal Sentinel-2 data: A novel approach through qualitative and quantitative feature selection with explainable AI DOI Creative Commons
Linh Nguyen Van, Vinh Ngoc Tran, Giang V. Nguyen

et al.

Ecological Informatics, Journal Year: 2024, Volume and Issue: 81, P. 102601 - 102601

Published: April 16, 2024

Accurate wildfire severity mapping (WSM) is crucial in environmental damage assessment and recovery strategies. Machine learning (ML) remote sensing technologies are extensively integrated employed as powerful tools for WSM. However, the intricate nature of ML algorithms often leads to 'black box' systems, obscuring decision-making process significantly limiting stakeholders' ability comprehend basis predictions. This opacity hinders efforts enhance performance risks exacerbating overfitting. present study proposes an innovative WSM approach that incorporates qualitative quantitative feature selection techniques within Explainable AI (XAI) framework. The methodology aims precision provide insights into factors contributing model decisions, thereby increasing interpretability predictions streamlining models improve performance. To achieve this objective, we SHapley Additive exPlanations (SHAP)-Forward Stepwise Selection (FSS) method demonstrate its efficacy elucidating impacts predictors on algorithm performance, accuracy, designed Utilizing post-fire imagery from Sentinel-2 (S2), analyzed ten bands generate 225 unique spectral indices utilizing five different calculations: normalized, algebraic sum, difference, ratio, product forms. Combined with original S2 bands, resulted 235 potential classifications. A random forest was subsequently developed using these optimized through extensive hyperparameter tuning, achieving overall accuracy (OA) 0.917 a Kappa statistic 0.896. most influential were identified SHAP values, FSS narrowing them down 12 critical effective WSM, evidenced by stabilized OA values (0.904 0.881, respectively). Further validation ninefold spatial cross-validation technique demonstrated method's consistent across data partitions, ranging 0.705 0.894 0.607 0.867. By providing more accurate comprehensible XAI-based research contributes broader field monitoring disaster response, underscoring analysis models' capabilities.

Language: Английский

Citations

10
A state-of-the-art review of long short-term memory models with applications in hydrology and water resources DOI
Zhong-kai Feng, J. Zhang, Wen-jing Niu

et al.

Applied Soft Computing, Journal Year: 2024, Volume and Issue: unknown, P. 112352 - 112352

Published: Oct. 1, 2024

Language: Английский

Citations

9
Unveiling environmental drivers of soil erosion in South Korea through SHAP-informed machine learning DOI
Linh Nguyen Van, Giang V. Nguyen, Minho Yeon

et al.

Land Use Policy, Journal Year: 2025, Volume and Issue: 155, P. 107592 - 107592

Published: May 9, 2025

Language: Английский

Citations

0
Deep neural network-based discharge prediction for upstream hydrological stations: a comparative study DOI
Xuan-Hien Le, Duc Hai Nguyen, Sungho Jung

et al.

Earth Science Informatics, Journal Year: 2023, Volume and Issue: 16(4), P. 3113 - 3124

Published: Aug. 21, 2023

Language: Английский

Citations

9
Evaluating the Utility of Selected Machine Learning Models for Predicting Stormwater Levels in Small Streams DOI Open Access
Mariusz Starzec, Sabina Kordana

Sustainability, Journal Year: 2024, Volume and Issue: 16(2), P. 783 - 783

Published: Jan. 16, 2024

The consequences of climate change include extreme weather events, such as heavy rainfall. As a result, many places around the world are experiencing an increase in flood risk. aim this research was to assess usefulness selected machine learning models, including artificial neural networks (ANNs) and eXtreme Gradient Boosting (XGBoost) v2.0.3., for predicting peak stormwater levels small stream. innovation results from combination specificity watersheds with techniques use SHapley Additive exPlanations (SHAP) analysis, which enabled identification key factors, rainfall depth meteorological data, significantly affect accuracy forecasts. analysis showed superiority ANN models (R2 = 0.803–0.980, RMSE 1.547–4.596) over XGBoost v2.0.3. 0.796–0.951, 2.304–4.872) terms forecasting effectiveness analyzed In addition, conducting SHAP allowed most crucial factors influencing forecast accuracy. parameters affecting predictions included depth, level, data air temperature dew point last day. Although study focused on specific stream, methodology can be adapted other watersheds. could contribute improving real-time warning systems, enabling local authorities emergency management agencies plan responses threats more accurately timelier manner. Additionally, these help protect infrastructure roads bridges by better potential implementation appropriate preventive measures. Finally, used inform communities about risk recommended precautions, thereby increasing awareness preparedness flash floods.

Language: Английский

Citations

2
Rolling predictive control of tandem multi-canal pools based on water level elasticity intervals: A case study of the South-North water diversion middle route project DOI Creative Commons
Mingrui Chen, Haichen Li, Lingzhong Kong

et al.

Journal of Hydrology Regional Studies, Journal Year: 2024, Volume and Issue: 52, P. 101740 - 101740

Published: March 16, 2024

The Middle Route of the South-North Water Transfer Project (SNWDMRP) is a major water transfer project to optimize spatial allocation resources in China. It difficult for traditional flow adjustment ensure safe operation open-channel projects. In this study, Long Short Term Memory (LSTM) level prediction model combined with elastic interval control method achieve rolling front gates single pool, and then coupled storage compensation algorithm gate group tandem multi-canal pools. applied SNWDMRP, results show that new stable within restricted both multiple canal Moreover, proposed study can make full use capacity satisfy need scheduling scenarios

Language: Английский

Citations

2
A Comparative Analysis of Advanced Machine Learning Techniques for River Streamflow Time-Series Forecasting DOI Open Access
Antoifi Abdoulhalik, Ashraf Ahmed

Sustainability, Journal Year: 2024, Volume and Issue: 16(10), P. 4005 - 4005

Published: May 10, 2024

This study examines the contribution of rainfall data (RF) in improving streamflow-forecasting accuracy advanced machine learning (ML) models Syr Darya River Basin. Different sets scenarios included from different weather stations located various geographical locations with respect to flow monitoring station. Long short-term memory (LSTM)-based were used examine on performance by investigating five whereby RF incorporated depending their positions. Specifically, All-RF scenario all collected at 11 stations; Upstream-RF (Up-RF) and Downstream-RF (Down-RF) only measured upstream downstream streamflow-measuring station; Pearson-RF (P-RF) exhibiting highest level correlation streamflow data, Flow-only (FO) data. The evaluation metrics quantitively assess RMSE, MAE, coefficient determination, R2. Both ML performed best FO scenario, which shows that diversity input features (hydrological meteorological data) did not improve predictive regardless positions stations. results show P-RF yielded better prediction compared other including suggests station tend make a positive model’s forecasting performance. findings evidence suitability simple monolayer LSTM-based networks as for high-performance budget-wise river forecast applications while minimizing processing time.

Language: Английский

Citations

2
Underutilized Feature Extraction Methods for Burn Severity Mapping: A Comprehensive Evaluation DOI Creative Commons
Linh Nguyen Van, Giha Lee

Remote Sensing, Journal Year: 2024, Volume and Issue: 16(22), P. 4339 - 4339

Published: Nov. 20, 2024

Wildfires increasingly threaten ecosystems and infrastructure, making accurate burn severity mapping (BSM) essential for effective disaster response environmental management. Machine learning (ML) models utilizing satellite-derived vegetation indices are crucial assessing wildfire damage; however, incorporating many can lead to multicollinearity, reducing classification accuracy. While principal component analysis (PCA) is commonly used address this issue, its effectiveness relative other feature extraction (FE) methods in BSM remains underexplored. This study aims enhance ML classifier accuracy by evaluating various FE techniques that mitigate multicollinearity among indices. Using composite index (CBI) data from the 2014 Carlton Complex fire United States as a case study, we extracted 118 seven Landsat-8 spectral bands. We applied compared 13 different techniques—including linear nonlinear such PCA, t-distributed stochastic neighbor embedding (t-SNE), discriminant (LDA), Isomap, uniform manifold approximation projection (UMAP), factor (FA), independent (ICA), multidimensional scaling (MDS), truncated singular value decomposition (TSVD), non-negative matrix factorization (NMF), locally (LLE), (SE), neighborhood components (NCA). The performance of these was benchmarked against six classifiers determine their improving Our results show alternative outperform computational efficiency. Techniques like LDA NCA effectively capture relationships critical BSM. contributes existing literature providing comprehensive comparison methods, highlighting potential benefits underutilized

Language: Английский

Citations

1