CNN-BiLSTM: A Novel Deep Learning Model for Near-Real-Time Daily Wildfire Spread Prediction

Remote Sensing, Journal Year: 2024, Volume and Issue: 16(8), P. 1467 - 1467

Published: April 20, 2024

Wildfires significantly threaten ecosystems and human lives, necessitating effective prediction models for the management of this destructive phenomenon. This study integrates Convolutional Neural Network (CNN) Bidirectional Long Short-Term Memory (BiLSTM) modules to develop a novel deep learning model called CNN-BiLSTM near-real-time wildfire spread capture spatial temporal patterns. uses Visible Infrared Imaging Radiometer Suite (VIIRS) active fire product wide range environmental variables, including topography, land cover, temperature, NDVI, wind informaiton, precipitation, soil moisture, runoff train model. A comprehensive exploration parameter configurations settings was conducted optimize model’s performance. The evaluation results their comparison with benchmark models, such as (LSTM) CNN-LSTM demonstrate effectiveness IoU F1 Score 0.58 0.73 validation training sets, respectively. innovative approach offers promising avenue enhancing efforts through its capacity prediction, marking significant step forward in mitigating impact wildfires.

Language: Английский

---

Trending and emerging prospects of physics-based and ML-based wildfire spread models: a comprehensive review

Journal of Forestry Research, Journal Year: 2024, Volume and Issue: 35(1)

Published: Sept. 27, 2024

Language: Английский

---

Enhancing wildfire mapping accuracy using mono-temporal Sentinel-2 data: A novel approach through qualitative and quantitative feature selection with explainable AI

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

---

CVTNet: A Fusion of Convolutional Neural Networks and Vision Transformer for Wetland Mapping Using Sentinel-1 and Sentinel-2 Satellite Data

Remote Sensing, Journal Year: 2024, Volume and Issue: 16(13), P. 2427 - 2427

Published: July 2, 2024

Wetland mapping is a critical component of environmental monitoring, requiring advanced techniques to accurately represent the complex land cover patterns and subtle class differences innate in these ecosystems. This study aims address challenges by proposing CVTNet, novel deep learning (DL) model that integrates convolutional neural networks (CNNs) vision transformer (ViT) architectures. CVTNet uses channel attention (CA) spatial (SA) mechanisms enhance feature extraction from Sentinel-1 (S1) Sentinel-2 (S2) satellite data. The primary goal this achieve balanced trade-off between Precision Recall, which essential for accurate wetland mapping. class-specific analysis demonstrated CVTNet’s proficiency across diverse classes, including pasture, shrubland, urban, bog, fen, water. Comparative showed outperforms contemporary algorithms such as Random Forest (RF), ViT, multi-layer perceptron mixer (MLP-mixer), hybrid spectral net (HybridSN) classifiers. Additionally, mechanism (AM) sensitivity highlighted crucial role CA, SA, ViT focusing model’s on regions, thereby improving regions. Despite at boundaries, particularly bog misclassifications swamp pixels, presents solution

Language: Английский

---

Machine learning methods for wildfire risk assessment

Earth Science Informatics, Journal Year: 2025, Volume and Issue: 18(1)

Published: Jan. 1, 2025

Language: Английский

---

Application of Explainable Artificial Intelligence in Predicting Wildfire Spread: An ASPP-Enabled CNN Approach

IEEE Geoscience and Remote Sensing Letters, Journal Year: 2024, Volume and Issue: 21, P. 1 - 5

Published: Jan. 1, 2024

Language: Английский

---

BD-SKUNet: Selective-Kernel UNets for Building Damage Assessment in High-Resolution Satellite Images

Remote Sensing, Journal Year: 2023, Volume and Issue: 16(1), P. 182 - 182

Published: Dec. 31, 2023

When natural disasters occur, timely and accurate building damage assessment maps are vital for disaster management responders to organize their resources efficiently. Pairs of pre- post-disaster remote sensing imagery have been recognized as invaluable data sources that provide useful information identification. Recently, deep learning-based semantic segmentation models widely successfully applied tasks. In this study, a two-stage, dual-branch, UNet architecture, with shared weights between two branches, is proposed address the inaccuracies in footprint localization per-building level classification. A newly introduced selective kernel module improves performance model by enhancing extracted features applying adaptive receptive field variations. The xBD dataset used train, validate, test based on evaluation metrics such F1-score Intersection over Union (IoU). Overall, experiments comparisons demonstrate superior model. addition, results further confirmed evaluating geographical transferability completely unseen from new region (Bam city earthquake 2003).

Language: Английский

---

Performance evaluation of convolutional neural network and vision transformer models for groundwater potential mapping

Journal of Hydrology, Journal Year: 2025, Volume and Issue: unknown, P. 132840 - 132840

Published: Feb. 1, 2025

Language: Английский

---

A Comprehensive Feature Extraction Network for Deep-Learning-Based Wildfire Detection in Remote Sensing Imagery

Applied Sciences, Journal Year: 2025, Volume and Issue: 15(7), P. 3699 - 3699

Published: March 27, 2025

As global climate change escalates, wildfires have emerged as a critical form of natural disaster, presenting substantial risks to ecosystems, public safety, and economic development. While satellite remote sensing has been extensively utilized for wildfire monitoring, current methodologies face limitations in addressing complex backgrounds environmental variations. These techniques usually depend on set thresholds or the extraction local features, which can lead incorrect positives overlooked detections. Consequently, existing methods inadequately capture comprehensive characteristics fire points. To mitigate these challenges, this study proposes deep-learning-based point detection method that integrates Swin Transformer BiLSTM multi-dimensional features associated with This research represents inaugural application context detection, leveraging its self-attention mechanism discern dependencies information within environments. By amalgamating at various levels, proposed significantly improves accuracy robustness detection. Experimental findings demonstrate surpasses traditional models such DenseNet, SimpleCNN, Multi-Layer Perceptron (MLP) across multiple performance metrics, including accuracy, recall, F1 score.

Language: Английский

---

Modeling the spread of forest fires through cellular automata by leveraging deep learning to derive transition rules

Ecological Informatics, Journal Year: 2025, Volume and Issue: unknown, P. 103150 - 103150

Published: April 1, 2025

Language: Английский

---