Article
Forestry
Chao Gao, Honglei Lin, Haiqing Hu
Summary: Forest-fire-risk was predicted in the Heihe area of Heilongjiang Province, China using random forest (RF) and backpropagation neural network (BPNN) algorithms. The driving factors correlated with forest-fire occurrence were analyzed and 11 factors were found to have a significant correlation. The prediction accuracy and goodness of fit of RF and BPNN algorithms were similar, indicating that both methods are suitable for forest-fire occurrence prediction. High-fire-risk zones were mainly located in the northwestern and central parts of the Heihe area.
Article
Computer Science, Information Systems
Rajib Ghosh, Anupam Kumar
Summary: This article proposes a novel hybrid deep learning model that combines CNN and RNN for feature extraction to detect forest fire. The performance of the proposed system has been evaluated on two publicly available datasets, showing very high classification accuracy.
MULTIMEDIA TOOLS AND APPLICATIONS
(2022)
Article
Geography, Physical
Yoojin Kang, Eunna Jang, Jungho Im, Chungeun Kwon
Summary: This study proposed a deep learning-based forest fire detection algorithm that effectively reduced detection latency and false alarms. By combining input features, the research demonstrated that temporal and spatial information contributed to improving the accuracy of machine learning techniques for fire detection.
GISCIENCE & REMOTE SENSING
(2022)
Article
Physics, Multidisciplinary
Zhenwei Guan, Feng Min, Wei He, Wenhua Fang, Tao Lu
Summary: Forest fire detection from videos or images is crucial for forest firefighting. This article proposes a novel neural network approach to balance the complexity of different training samples and extract color feature information of fires. Experimental results demonstrate that this method outperforms the current state-of-the-art methods.
Article
Forestry
Xufeng Lin, Zhongyuan Li, Wenjing Chen, Xueying Sun, Demin Gao
Summary: A forest fire prediction model based on LSTNet is proposed in this study, which utilizes remote sensing satellites and GIS to obtain the influential factors of forest fires and estimates their correlation. The model takes into account the spatial aggregation of forest fires through oversampling methods and proportional stratified sampling. The results show that the LSTNet model has high accuracy (ACC 0.941) and effectively utilizes spatial background information and the periodicity of forest fire factors, providing a novel method for spatial prediction of forest fire susceptibility.
Article
Chemistry, Analytical
Jiarun Huang, Zhili He, Yuwei Guan, Hongguo Zhang
Summary: In this study, a forest fire detection method GXLD based on lightweight YOLOX-L and defogging algorithm is proposed. Experimental results show that the number of parameters of YOLOX-L-Light decreased by 92.6%, and the mAP increased by 1.96%. The mAP of GXLD is 87.47%, which is 2.46% higher than that of YOLOX-L; and the average fps of GXLD is 26.33 when the input image size is 1280 x 720. The research demonstrates that GXLD can detect forest fires in real time accurately even in foggy environments.
Article
Chemistry, Multidisciplinary
Shaoxiong Zheng, Peng Gao, Weixing Wang, Xiangjun Zou
Summary: An improved dynamic convolutional neural network (DCNN) model, named DCN_Fire, was established based on the traditional DCNN model for accurately identifying the risk of a forest fire. Transfer learning and principal component analysis were used to enhance the model's accuracy and speed. The results showed that the improved DCNN model had excellent recognition speed and accuracy, providing a technical reference for preventing and tackling forest fires.
APPLIED SCIENCES-BASEL
(2022)
Article
Ecology
Omer Kantarcioglu, Sultan Kocaman, Konrad Schindler
Summary: This study utilized the Artificial Neural Networks (ANN), a commonly used machine learning technique, to assess forest fire susceptibility in Istanbul Province and Thrace Region, Turkiye. By utilizing freely available Earth Observation datasets and a forest inventory, the ANN model achieved high prediction performance with an AUC value of 0.94 and an F-1 score of 0.80. The proximity to forest roads was identified as the most predictive input variable. The study highlights the potential of data-driven machine learning methods for regional forest fire susceptibility assessments.
ECOLOGICAL INFORMATICS
(2023)
Article
Engineering, Environmental
Hossein Moayedi, Mohammad Ali Salehi Amin Khasmakhi
Summary: This study proposes two hybrid algorithms combining artificial neural networks (ANN) with spatial analysis of forest fire to improve accuracy. By considering multiple factors such as slope aspect, soil type, and rainfall, the study zoned the susceptibility areas for forest fire and demonstrated the spatial interaction between fire and ignition factors using the frequency ratio model.
STOCHASTIC ENVIRONMENTAL RESEARCH AND RISK ASSESSMENT
(2023)
Article
Environmental Sciences
Jiahui Chen, Yi Yang, Ling Peng, Luanjie Chen, Xingtong Ge
Summary: In this paper, a knowledge-graph- and representation-learning-based forest fire prediction method is proposed to address the issues of ignoring complex dependencies and correlations in traditional prediction methods. The results of the experiment show that the method significantly improves prediction accuracy compared to previous methods.
Article
Environmental Sciences
Shaoxiong Zheng, Peng Gao, Yufei Zhou, Zepeng Wu, Liangxiang Wan, Fei Hu, Weixing Wang, Xiangjun Zou, Shihong Chen
Summary: Monitoring and early warning technology for forest fires is crucial. An early warning/monitoring system for forest fires was constructed using deep learning and the internet of things. The system improved forest fire recognition by combining the characteristics of flame, smoke, and area. Complex fire-image features were extracted, and a forest fire risk prediction model was built based on an improved dynamic convolutional neural network. The proposed algorithm provided real-time accurate recognition with 84.37% accuracy, and the monitoring network had a low packet loss rate.
Article
Forestry
Zhengjun Yan, Liming Wang, Kui Qin, Feng Zhou, Jineng Ouyang, Teng Wang, Xinguo Hou, Leping Bu
Summary: This paper applies unsupervised domain adaptation (UDA) to transfer knowledge from a labeled public fire dataset to another unlabeled one, providing a benchmark dataset for fire recognition. Two adaptation networks are experimented on this dataset and achieve impressive performance in fire recognition.
Article
Forestry
Junling Wang, Xijian Fan, Xubing Yang, Tardi Tjahjadi, Yupeng Wang
Summary: This paper proposes a semi-supervised learning-based segmentation network, SemiFSNet, which enhances model performance through occlusion-aware data augmentation, dynamic convolution, and attention mechanism, as well as exploiting rich information from unlabeled data using consistency regularization to effectively monitor forest fires.
Article
Plant Sciences
Shaoxiong Zheng, Peng Gao, Xiangjun Zou, Weixing Wang
Summary: This article introduces an improved forest fire risk identification algorithm that accurately identifies forest fire risk in complex natural environments. The algorithm uses image enhancement, segmentation, and feature extraction to recognize the risk. Experimental results show that it achieves an accuracy of up to 92.73%.
FRONTIERS IN PLANT SCIENCE
(2022)
Article
Environmental Sciences
Tran Xuan Truong, Viet-Ha Nhu, Doan Thi Nam Phuong, Le Thanh Nghi, Nguyen Nhu Hung, Pham Viet Hoa, Dieu Tien Bui
Summary: Frequent forest fires have severe negative impacts on the natural environment, and developing accurate prediction models for forest fire danger is crucial. This research proposes a new modeling approach using TensorFlow deep neural networks and geographic information systems (GIS). The TFDeepNN model shows high predictive performance and outperforms the baseline models, making it a valuable tool for spatially predicting forest fire danger. The forest fire danger map from this study can assist policymakers and authorities in sustainable land-use planning and management.
Article
Environmental Sciences
Huu Duy Nguyen, Dennis Fox, Dinh Kha Dang, Le Tuan Pham, Quan Vu Viet Du, Thi Ha Thanh Nguyen, Thi Ngoc Dang, Van Truong Tran, Phuong Lan Vu, Quoc-Huy Nguyen, Tien Giang Nguyen, Quang-Thanh Bui, Alexandru-Ionut Petrisor
Summary: The study aims to explore future urban flood risk by combining land use change and hydraulic models, aiming to reduce risk under different vulnerability and exposure scenarios. Despite the increase in flood risk due to urbanization, population density, and the number of hospitals in the floodplain, particularly in the coastal region, the area exposed to high and very high risks decreases due to a reduction in poverty rate.
Article
Ecology
Van-Manh Pham, Son Van Nghiem, Cu Van Pham, Mai Phuong Thi Luu, Quang-Thanh Bui
Summary: The study conducted a spatial and temporal analysis on the HuLATIN SMALL LETTER E WITH CIRCUMFLEX AND ACUTE Monuments in Vietnam to understand the conflict between preserving historical landscape and developing urban areas. The results showed that urbanization poses significant risks to the heritage sites, with a notable correlation between urbanization intensity index and risks to Outstanding Universal Values (OUV). The study area serves as an example of global cultural heritages threatened by socio-economic development, and the approach used in this study can be applied to similar areas.
Article
Public, Environmental & Occupational Health
Nguyen Thi Hong Minh, Tran Cao Binh, Trinh Dinh Hai, Nguyen Thuy Duong, Quang-Thanh Bui
Summary: This study utilized data from the 2019 National Oral Health Survey to investigate the impact of sociodemographic features on periodontal conditions through machine learning methods. The results showed that the gradient boosting model had the best predictive performance, with geographic region and consumption of sweet foods/drinks being the most significant factors influencing dental health.
ASIA-PACIFIC JOURNAL OF PUBLIC HEALTH
(2022)
Article
Environmental Sciences
Huu Duy Nguyen, Quoc-Huy Nguyen, Quan Vu Viet Du, Thi Ha Thanh Nguyen, Tien Giang Nguyen, Quang-Thanh Bui
Summary: This research aims to develop a novel hybrid algorithm, DNN-MRFO, combining deep neural network and Manta ray foraging optimization, to generate flood susceptibility map for Quang Ngai province, Vietnam. Comparative analysis with other models shows that combining DNN and MRFO improves flood susceptibility classification precision, providing significant support for policymakers in improving their adaptation strategies.
GEOCARTO INTERNATIONAL
(2022)
Article
Environmental Sciences
Huu Duy Nguyen, Dinh Kha Dang, Quoc-Huy Nguyen, Quang-Thanh Bui, Alexandru-Ionut Petrisor
Summary: The study evaluates the impact of land use and climate change on flood susceptibility in Vietnam using machine learning and Land Change Modeler. The results show that optimizing the Support Vector Machine model can improve flood risk prediction performance. Additionally, the study identifies an increase in areas with high flood susceptibility.
GEOCARTO INTERNATIONAL
(2022)
Article
Environmental Sciences
Huu Duy Nguyen, Vu Dong Pham, Phuong Lan Vu, Thi Ha Thanh Nguyen, Quoc-Huy Nguyen, Tien Giang Nguyen, Dinh Kha Dang, Van Truong Tran, Quang-Thanh Bui, Tuan Anh Lai, Alexandru-Ionut Petrisor
Summary: Agricultural land abandonment due to floods is a global problem. This study assessed the drivers of land abandonment in a river valley in Central Vietnam and found that high-flood zones and community resilience were key factors influencing agricultural land outcomes. The findings highlight the potential for spatial planning to mitigate the adverse effects of flood events and can provide insights for agricultural management in similar regions worldwide.
Article
Forestry
Quoc-Huy Nguyen, Huu-Duy Nguyen, Dinh Tan Le, Quang-Thanh Bui
Summary: This study proposes a hybrid machine learning model based on the LightGBM algorithm and artificial ecosystem-based optimization to improve the accuracy of forest fire susceptibility assessment. The model outperforms other benchmark models and provides insights into the most influential factors of fire hazards.
Article
Environmental Sciences
Quan Vu Viet Du, Huu Duy Nguyen, Viet Thanh Pham, Cao Huan Nguyen, Quoc-Huy Nguyen, Quang-Thanh Bui, Thanh Thuy Doan, Anh Tuan Tran, Alexandru-Ionut Petrisor
Summary: Understanding the negative impacts of climate change and land use/land cover changes on natural hazards, such as landslides, is crucial for sustainable development worldwide. This study proposes a state-of-the-art method using machine learning and remote sensing algorithms to assess the effects of these factors on landslide susceptibility in Vietnam's Tra Khuc river basin. The models developed in this study performed well, with the RBFNN-QSA model having the highest accuracy. The findings indicate significant climate and land use changes, which affect landslide susceptibility, and can inform decision-makers in developing strategies to mitigate landslide damage.
GEOCARTO INTERNATIONAL
(2023)
Article
Geography
Tuan Anh Lai, Ngoc-Thach Nguyen, Quang-Thanh Bui
Summary: Properly choosing hyper-parameters using the Bayesian optimization algorithm improved the performance of machine learning models, specifically LightGBM and XGBoost, for hazard analysis. The models achieved high AUC values and their interpretation using SHAP values enhanced understanding of model workings and feature interactions. Further investigation of optimal hyper-parameter search using novel optimization algorithms is worthwhile for future research.
TRANSACTIONS IN GIS
(2023)
Article
Environmental Sciences
Tuan Linh Giang, Quang Thanh Bui, Thi Dieu Linh Nguyen, Van Bao Dang, Quang Hai Truong, Trong Trinh Phan, Hieu Nguyen, Van Liem Ngo, Van Truong Tran, Muhammad Yasir, Kinh Bac Dang
Summary: Using remote sensing data and GIS tools, the study aims to categorize coastal landscapes based on multi-source data with the help of convolutional-neural-network models. Nine coastal landscapes were identified, including deltas, alluvial, mature and young sand dunes, cliff, lagoon, tectonic, karst, and transitional landscapes. The CvNet models achieved high accuracy in classifying the landscapes along the coasts in Vietnam, except for dalmatian, karst and delta coastal landscapes. The evaluation of additional natural components is necessary and the CvNet model has the ability to update new landscape types at both national and global scales.
JOURNAL OF ENVIRONMENTAL MANAGEMENT
(2023)
Article
Water Resources
Chien Pham Van, Huu Duy Nguyen, Quoc-Huy Nguyen, Quang-Thanh Bui
Summary: The objective of this study was to develop an advanced method using LSTM, SVM, and RF to predict streamflow in the Mekong Delta in Vietnam, which is crucial for food security. Water level and flow data from 2014 to 2018 were used as input for the prediction model. The results showed that the SVM and RF models improved the performance of the LSTM model, with R-2 > 80%. LSTM was found to be a robust technique for characterizing and predicting time series behaviors in hydrology applications.
JOURNAL OF WATER AND CLIMATE CHANGE
(2023)
Article
Transportation
Minh Kieu, Eric Wanjau, Alexis Comber, Kristina Bratkova, Hang Nguyen Thi Thuy, Thanh Bui Quang, Phe Hoang Huu, Nick Malleson
Summary: The dependence on motorbikes has worsened traffic problems in Hanoi. Policymakers are considering a ban on non-electric motorbikes to address congestion and pollution. This paper analyzes survey data to understand residents' perceptions of the potential ban and their future mobility plans.
CASE STUDIES ON TRANSPORT POLICY
(2023)
Article
Geography
Huu Duy Nguyen, Dinh-Kha Dang, Quang-Thanh Bui, Alexandru-Ionut Petrisor
Summary: The main objective of this study was to develop a multi-hazard susceptibility mapping framework by combining flooding and landslides in the North Central region of Vietnam. Support vector machines, random forest, and AdaBoost models were used to accomplish this. The accuracy of the models' predictions was evaluated using various statistical indices, and all models performed well with AUC values over 0.95. The multi-hazard maps can be used as a point of reference for decision makers in land-use planning and infrastructure development to effectively prevent and reduce the frequency of floods and landslides.
TRANSACTIONS IN GIS
(2023)
Article
Geosciences, Multidisciplinary
Tuan Linh Giang, Kinh Bac Dang, Quang Thanh Bui
Summary: This study proposes indicators to identify coastlines and shorelines and trains multiple pixel-based and object-based machine learning models using high-resolution remote sensing images. The U-Net model achieves the best performance in analyzing coastlines and shorelines with linear and continuous structures. The coasts in Danang and Quang Nam provinces have experienced erosion in the last 10 years under the pressure of tourist development. The trained U-Net model can be used to monitor coastline changes worldwide.
VIETNAM JOURNAL OF EARTH SCIENCES
(2023)
Article
Environmental Sciences
Huu Duy Nguyen, Chien Pham Van, Tien Giang Nguyen, Dinh Kha Dang, Thi Thuy Nga Pham, Quoc-Huy Nguyen, Quang-Thanh Bui
Summary: This study aims to develop a low-cost method based on machine learning and remote sensing to map soil salinity in Ben Tre province, located in Vietnam's Mekong River Delta. The results show that six optimization algorithms successfully improved XGR model performance, with the XGR-HHO model being the best performing. The study highlights the effectiveness of using hybrid machine learning and remote sensing in soil salinity monitoring.
ENVIRONMENTAL SCIENCE AND POLLUTION RESEARCH
(2023)