Pub Date : 2025-01-23DOI: 10.1016/j.ecoinf.2025.103024
Guanjun Lin , Hang Zhao , Yufeng Chi
<div><div>Recently, the issue of near-surface ozone pollution has become a growing concern. To effectively manage and control ozone pollution, emerging deep learning (DL) techniques have been applied for future ozone concentration trend prediction, generating promising outcomes. However, existing studies employ various DL models and rely on diverse datasets to predict ozone concentrations. This leads to a lack of comprehensive evaluations of how the architecture and depth of different DL models influence the predictive accuracy of ozone concentration trends when assessed using a unified dataset. This lack of uniformity in evaluations creates a gap in our understanding of the influence of different neural network architectures and depths on ozone concentration predictions. In this work, we aim to address this research gap by conducting a systematic performance evaluation that benchmarks six prominent DL architectures, each with varying depths, to evaluate their effectiveness for predicting ozone concentrations across diverse geographical regions. Our findings indicate that the best-performing DL model in the nationwide prediction task is the one-layer bidirectional long short-term memory (Bi-LSTM) model, which achieves an <span><math><msup><mrow><mi>R</mi></mrow><mrow><mn>2</mn></mrow></msup></math></span> of 0.66, an RMSE of 15.32<span><math><mrow><mi>μ</mi><mi>g</mi><mi>⋅</mi><msup><mrow><mi>m</mi></mrow><mrow><mo>−</mo><mn>3</mn></mrow></msup></mrow></math></span>, and an MAE of 11.51<span><math><mrow><mi>μ</mi><mi>g</mi><mi>⋅</mi><msup><mrow><mi>m</mi></mrow><mrow><mo>−</mo><mn>3</mn></mrow></msup></mrow></math></span>. In contrast, the poorest-performing model in the same prediction task is the one-block transformer-based model, with an <span><math><msup><mrow><mi>R</mi></mrow><mrow><mn>2</mn></mrow></msup></math></span> of 0.57, an RMSE of 17.34<span><math><mrow><mi>μ</mi><mi>g</mi><mi>⋅</mi><msup><mrow><mi>m</mi></mrow><mrow><mo>−</mo><mn>3</mn></mrow></msup></mrow></math></span>, and an MAE of 13.3<span><math><mrow><mi>μ</mi><mi>g</mi><mi>⋅</mi><msup><mrow><mi>m</mi></mrow><mrow><mo>−</mo><mn>3</mn></mrow></msup></mrow></math></span>. Furthermore, fully connected networks (FCNs) demonstrate robust and efficient predictive performance across both nationwide and regional prediction tasks. Notably, our study reveals that no single DL model consistently performs well across all prediction tasks, emphasizing the need for tailored approaches that cater to the specific attributes of each region. Additionally, we observe that DL models with more than two hidden layers frequently suffer from overfitting. Particularly for the Bi-LSTM architecture, as the number of hidden layers increases from 1 to 7, we observe a 12% reduction in <span><math><msup><mrow><mi>R</mi></mrow><mrow><mn>2</mn></mrow></msup></math></span> performance. Our analysis also identifies the most influential meteorological factors among the top-performing DL models, offering insigh
{"title":"A comprehensive evaluation of deep learning approaches for ground-level ozone prediction across different regions","authors":"Guanjun Lin , Hang Zhao , Yufeng Chi","doi":"10.1016/j.ecoinf.2025.103024","DOIUrl":"10.1016/j.ecoinf.2025.103024","url":null,"abstract":"<div><div>Recently, the issue of near-surface ozone pollution has become a growing concern. To effectively manage and control ozone pollution, emerging deep learning (DL) techniques have been applied for future ozone concentration trend prediction, generating promising outcomes. However, existing studies employ various DL models and rely on diverse datasets to predict ozone concentrations. This leads to a lack of comprehensive evaluations of how the architecture and depth of different DL models influence the predictive accuracy of ozone concentration trends when assessed using a unified dataset. This lack of uniformity in evaluations creates a gap in our understanding of the influence of different neural network architectures and depths on ozone concentration predictions. In this work, we aim to address this research gap by conducting a systematic performance evaluation that benchmarks six prominent DL architectures, each with varying depths, to evaluate their effectiveness for predicting ozone concentrations across diverse geographical regions. Our findings indicate that the best-performing DL model in the nationwide prediction task is the one-layer bidirectional long short-term memory (Bi-LSTM) model, which achieves an <span><math><msup><mrow><mi>R</mi></mrow><mrow><mn>2</mn></mrow></msup></math></span> of 0.66, an RMSE of 15.32<span><math><mrow><mi>μ</mi><mi>g</mi><mi>⋅</mi><msup><mrow><mi>m</mi></mrow><mrow><mo>−</mo><mn>3</mn></mrow></msup></mrow></math></span>, and an MAE of 11.51<span><math><mrow><mi>μ</mi><mi>g</mi><mi>⋅</mi><msup><mrow><mi>m</mi></mrow><mrow><mo>−</mo><mn>3</mn></mrow></msup></mrow></math></span>. In contrast, the poorest-performing model in the same prediction task is the one-block transformer-based model, with an <span><math><msup><mrow><mi>R</mi></mrow><mrow><mn>2</mn></mrow></msup></math></span> of 0.57, an RMSE of 17.34<span><math><mrow><mi>μ</mi><mi>g</mi><mi>⋅</mi><msup><mrow><mi>m</mi></mrow><mrow><mo>−</mo><mn>3</mn></mrow></msup></mrow></math></span>, and an MAE of 13.3<span><math><mrow><mi>μ</mi><mi>g</mi><mi>⋅</mi><msup><mrow><mi>m</mi></mrow><mrow><mo>−</mo><mn>3</mn></mrow></msup></mrow></math></span>. Furthermore, fully connected networks (FCNs) demonstrate robust and efficient predictive performance across both nationwide and regional prediction tasks. Notably, our study reveals that no single DL model consistently performs well across all prediction tasks, emphasizing the need for tailored approaches that cater to the specific attributes of each region. Additionally, we observe that DL models with more than two hidden layers frequently suffer from overfitting. Particularly for the Bi-LSTM architecture, as the number of hidden layers increases from 1 to 7, we observe a 12% reduction in <span><math><msup><mrow><mi>R</mi></mrow><mrow><mn>2</mn></mrow></msup></math></span> performance. Our analysis also identifies the most influential meteorological factors among the top-performing DL models, offering insigh","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"86 ","pages":"Article 103024"},"PeriodicalIF":5.8,"publicationDate":"2025-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143101990","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-22DOI: 10.1016/j.ecoinf.2024.102989
Cameron Trotter, Huw J. Griffiths, Rowan J. Whittle
The analysis of image data for benthic biodiversity monitoring is now commonplace within the domain of marine ecology. Whilst advances in imaging technologies have allowed for the collection of vast quantities of data, the curation of this has traditionally been performed manually, resulting in a bottleneck whereby data is collected faster than it can be processed. Recent years have seen marine ecologists turn to the domain of computer vision to help automate this curation process. However, as the knowledge required to build such systems spans both domains, there is a high barrier to entry. To help reduce this barrier, this paper aims to provide an introduction to computer vision-based benthic biodiversity monitoring via a comprehensive literature review. To aid ecologists, key computer vision concepts are described and example use-cases highlighted. The major challenges inherent to benthic imagery for computer vision systems are explored, alongside a discussion of how current systems attempt to mitigate against these. To aid computer scientists wishing to enter the domain, an exploration of currently available open-source benthic datasets is also provided. Recommendations for future research are explored, including a move towards human-centric techniques, committing to ablation studies, reaching community agreement on open-source benchmarking datasets, and an increased use of innovative methods to allow for improved answering of key benthic ecology questions.
{"title":"Surveying the deep: A review of computer vision in the benthos","authors":"Cameron Trotter, Huw J. Griffiths, Rowan J. Whittle","doi":"10.1016/j.ecoinf.2024.102989","DOIUrl":"10.1016/j.ecoinf.2024.102989","url":null,"abstract":"<div><div>The analysis of image data for benthic biodiversity monitoring is now commonplace within the domain of marine ecology. Whilst advances in imaging technologies have allowed for the collection of vast quantities of data, the curation of this has traditionally been performed manually, resulting in a bottleneck whereby data is collected faster than it can be processed. Recent years have seen marine ecologists turn to the domain of computer vision to help automate this curation process. However, as the knowledge required to build such systems spans both domains, there is a high barrier to entry. To help reduce this barrier, this paper aims to provide an introduction to computer vision-based benthic biodiversity monitoring via a comprehensive literature review. To aid ecologists, key computer vision concepts are described and example use-cases highlighted. The major challenges inherent to benthic imagery for computer vision systems are explored, alongside a discussion of how current systems attempt to mitigate against these. To aid computer scientists wishing to enter the domain, an exploration of currently available open-source benthic datasets is also provided. Recommendations for future research are explored, including a move towards human-centric techniques, committing to ablation studies, reaching community agreement on open-source benchmarking datasets, and an increased use of innovative methods to allow for improved answering of key benthic ecology questions.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"86 ","pages":"Article 102989"},"PeriodicalIF":5.8,"publicationDate":"2025-01-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143101993","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-22DOI: 10.1016/j.ecoinf.2025.103010
Alef Iury Siqueira Ferreira , Nádia Felix Felipe da Silva , Fernanda Neiva Mesquita , Thierson Couto Rosa , Stephen L. Buchmann , José Neiva Mesquita-Neto
Buzz-pollinated crops, such as tomatoes, potatoes, kiwifruit, and blueberries, are among the highest-yielding agricultural products. The flowers of these cultivated plants are characterized by having a specialized flower morphology with poricidal anthers that require vibration to achieve a full seed set. At least 446 bee species, in 82 genera, use floral sonication (buzz pollination) to collect pollen grains as food. Identifying and classifying these diverse often look-alike bee species poses a challenge for taxonomists. Automated classification systems, based upon audible bee floral buzzes, have been investigated to meet this need. Recently, convolutional neural network (CNN) models have demonstrated superior performance in recognizing and distinguishing bee-buzzing sounds compared to classical Machine-Learning (ML) classifiers. Nonetheless, the performance of CNNs remains unsatisfactory and can be improved. Therefore, we applied a novel transformer-based neural network architecture for the task of acoustic recognition of blueberry-pollinating bee species. We further compared the performance of the Audio Spectrogram Transformer (AST) model and its variants, including Self-Supervised AST (SSAST) and Masked Autoencoding AST (MAE-AST), to that of strong baseline CNN models based on previous work, at the task of bee species recognition. We also employed data augmentation techniques and evaluated these models with a data set of bee sounds recorded during visits to blueberry flowers in Chile (518 audio samples of 15 bee species). Our results revealed that Transformer-based Neural Networks combined with pre-training and data augmentation outperformed CNN models (maximum F1-score: 64.5% ± 2; Accuracy: 82.2% ± 0.8). These innovative attention-based neural network architectures have demonstrated exceptional performance in assigning bee buzzing sounds to their respective taxonomic categories, outperforming prior deep learning models. However, transformer approaches face challenges related to small dataset size and class imbalance, similar to CNNs and classical ML algorithms. Combining pre-training with data augmentation is crucial to increase the diversity and robustness of training data sets for the acoustic recognition of bee species. We document the potential of transformer architectures to improve the performance of audible bee species identification, offering promising new avenues for bioacoustic research and pollination ecology.
{"title":"Transformer Models improve the acoustic recognition of buzz-pollinating bee species","authors":"Alef Iury Siqueira Ferreira , Nádia Felix Felipe da Silva , Fernanda Neiva Mesquita , Thierson Couto Rosa , Stephen L. Buchmann , José Neiva Mesquita-Neto","doi":"10.1016/j.ecoinf.2025.103010","DOIUrl":"10.1016/j.ecoinf.2025.103010","url":null,"abstract":"<div><div>Buzz-pollinated crops, such as tomatoes, potatoes, kiwifruit, and blueberries, are among the highest-yielding agricultural products. The flowers of these cultivated plants are characterized by having a specialized flower morphology with poricidal anthers that require vibration to achieve a full seed set. At least 446 bee species, in 82 genera, use floral sonication (buzz pollination) to collect pollen grains as food. Identifying and classifying these diverse often look-alike bee species poses a challenge for taxonomists. Automated classification systems, based upon audible bee floral buzzes, have been investigated to meet this need. Recently, convolutional neural network (CNN) models have demonstrated superior performance in recognizing and distinguishing bee-buzzing sounds compared to classical Machine-Learning (ML) classifiers. Nonetheless, the performance of CNNs remains unsatisfactory and can be improved. Therefore, we applied a novel transformer-based neural network architecture for the task of acoustic recognition of blueberry-pollinating bee species. We further compared the performance of the Audio Spectrogram Transformer (AST) model and its variants, including Self-Supervised AST (SSAST) and Masked Autoencoding AST (MAE-AST), to that of strong baseline CNN models based on previous work, at the task of bee species recognition. We also employed data augmentation techniques and evaluated these models with a data set of bee sounds recorded during visits to blueberry flowers in Chile (518 audio samples of 15 bee species). Our results revealed that Transformer-based Neural Networks combined with pre-training and data augmentation outperformed CNN models (maximum F1-score: 64.5% ± 2; Accuracy: 82.2% ± 0.8). These innovative attention-based neural network architectures have demonstrated exceptional performance in assigning bee buzzing sounds to their respective taxonomic categories, outperforming prior deep learning models. However, transformer approaches face challenges related to small dataset size and class imbalance, similar to CNNs and classical ML algorithms. Combining pre-training with data augmentation is crucial to increase the diversity and robustness of training data sets for the acoustic recognition of bee species. We document the potential of transformer architectures to improve the performance of audible bee species identification, offering promising new avenues for bioacoustic research and pollination ecology.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"86 ","pages":"Article 103010"},"PeriodicalIF":5.8,"publicationDate":"2025-01-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143101994","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-22DOI: 10.1016/j.ecoinf.2025.103043
Jun Cao , Tianyu Chen , Jipeng Sun , Jun Zhong , Biao Mu , Xin Wang , Chunyan Wang , Massimiliano Materazzi , Hualun Zhu
This study investigates the potential of fluorescence characteristics of dissolved organic matter (DOM) to identify sediment organic matter (OM) sources in shallow lakes. Spectral analyses were performed on water and alkali extractable organic matter (WEOM and AEOM) from Lake Taihu sediments. The lake was divided into seven distinct regions: R1 and R2 strongly influenced by inflowing rivers, R3 and R4 were characterized by submerged macrophytes, and R5-R7 were dominated by cyanobacterial blooms. The investigation illustrated that the highest values of water and alkali extractable organic carbon (WEOC and AEOC) were found in region R4. Specifically, the Humification Index (HIX) values consistently exceeded 2.0 in the northwest regions, contrasting with values predominantly below 1.0 in most southeastern regions. Moreover, the Fluorescence Index (FI) of WEOM in regions R5, R6 and R7 reached 2.10, markedly higher than the values observed in other regions. The horizontal distribution of the four spectrographic indices of AEOM exhibited partial similarity to the distribution pattern of WEOM. Although the WEOC content marginally trailed AEOC, there was a significant correlation between WEOM and AEOM in three indices including slope ratio (SR), HIX and FI. The identification of sources implied that organic matter in sediments of regions R1 and R2 originated from terrestrial sources, while regions R3 and R4 were largely derived from submerged macrophyte and the regions R5-R7 were notably impacted by cyanobacteria-derived organic matters. Notably, the identification results aligned perfectly with the distribution of inflowing rivers, cyanobacterial blooms and submerged macrophyte coverage within Taihu Lake, underscoring the potential use of dissolved organic matter's spectral characteristics for organic matter source analysis within sediments.
Synopsis
This study identifies distinct sources and spatial distributions of organic matter in Lake Taihu's sediments, using fluorescence characteristics to highlight influences from terrestrial input, submerged macrophytes, and cyanobacterial blooms.
{"title":"Identifying sources of dissolved organic matter in sediments of a shallow lake by fluorescence and ultraviolet spectral characteristics of water and alkali extractable organic matter (WEOM and AEOM)","authors":"Jun Cao , Tianyu Chen , Jipeng Sun , Jun Zhong , Biao Mu , Xin Wang , Chunyan Wang , Massimiliano Materazzi , Hualun Zhu","doi":"10.1016/j.ecoinf.2025.103043","DOIUrl":"10.1016/j.ecoinf.2025.103043","url":null,"abstract":"<div><div>This study investigates the potential of fluorescence characteristics of dissolved organic matter (DOM) to identify sediment organic matter (OM) sources in shallow lakes. Spectral analyses were performed on water and alkali extractable organic matter (WEOM and AEOM) from Lake Taihu sediments. The lake was divided into seven distinct regions: R1 and R2 strongly influenced by inflowing rivers, R3 and R4 were characterized by submerged macrophytes, and R5-R7 were dominated by cyanobacterial blooms. The investigation illustrated that the highest values of water and alkali extractable organic carbon (WEOC and AEOC) were found in region R4. Specifically, the Humification Index (HIX) values consistently exceeded 2.0 in the northwest regions, contrasting with values predominantly below 1.0 in most southeastern regions. Moreover, the Fluorescence Index (FI) of WEOM in regions R5, R6 and R7 reached 2.10, markedly higher than the values observed in other regions. The horizontal distribution of the four spectrographic indices of AEOM exhibited partial similarity to the distribution pattern of WEOM. Although the WEOC content marginally trailed AEOC, there was a significant correlation between WEOM and AEOM in three indices including slope ratio (S<sub>R</sub>), HIX and FI. The identification of sources implied that organic matter in sediments of regions R1 and R2 originated from terrestrial sources, while regions R3 and R4 were largely derived from submerged macrophyte and the regions R5-R7 were notably impacted by cyanobacteria-derived organic matters. Notably, the identification results aligned perfectly with the distribution of inflowing rivers, cyanobacterial blooms and submerged macrophyte coverage within Taihu Lake, underscoring the potential use of dissolved organic matter's spectral characteristics for organic matter source analysis within sediments.</div></div><div><h3>Synopsis</h3><div>This study identifies distinct sources and spatial distributions of organic matter in Lake Taihu's sediments, using fluorescence characteristics to highlight influences from terrestrial input, submerged macrophytes, and cyanobacterial blooms.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"86 ","pages":"Article 103043"},"PeriodicalIF":5.8,"publicationDate":"2025-01-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143102486","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-22DOI: 10.1016/j.ecoinf.2025.103042
Chandra Segaran Thirukanthan , Parashuram Kallem , Idham Sumarto Pratama , Fathurrahman Lananan , Lee Seong Wei , Zulhisyam Abdul Kari , Huan Gao , Mohamad Nor Azra , Wan Izatul Asma Wan Talaat
This comprehensive scientometric analysis, utilizing CiteSpace and data from the Web of Science Core Collection, examines the trajectory of research on Marine Protected Areas (MPAs) in the context of climate change. Analysing 2782 articles and 117,904 cited references, the study observes a significant surge in publications between 2019 and 2023, with Australia, England and Canada as leading contributors. Our findings reveal key conceptual pillars such as ‘marine protected areas’, ‘climate change’, ‘conservation’, ‘management’, and ‘biodiversity’. The research domain is characterized by 10 major co-citation clusters, with a notable focus on “coral reefs”, “temperature-driven coral decline”, and “large MPAs”. The increasing citation frequency during 2020–2023, particularly in clusters related to coral reefs and regional studies, signals a heightened global awareness of MPAs' role in mitigating climate change impacts. This review provides essential insights, informing future directions for both academic research and policymaking in marine conservation amid ongoing climatic changes.
{"title":"Marine protected area and climate change: A mapping review","authors":"Chandra Segaran Thirukanthan , Parashuram Kallem , Idham Sumarto Pratama , Fathurrahman Lananan , Lee Seong Wei , Zulhisyam Abdul Kari , Huan Gao , Mohamad Nor Azra , Wan Izatul Asma Wan Talaat","doi":"10.1016/j.ecoinf.2025.103042","DOIUrl":"10.1016/j.ecoinf.2025.103042","url":null,"abstract":"<div><div>This comprehensive scientometric analysis, utilizing CiteSpace and data from the Web of Science Core Collection, examines the trajectory of research on Marine Protected Areas (MPAs) in the context of climate change. Analysing 2782 articles and 117,904 cited references, the study observes a significant surge in publications between 2019 and 2023, with Australia, England and Canada as leading contributors. Our findings reveal key conceptual pillars such as ‘marine protected areas’, ‘climate change’, ‘conservation’, ‘management’, and ‘biodiversity’. The research domain is characterized by 10 major co-citation clusters, with a notable focus on “coral reefs”, “temperature-driven coral decline”, and “large MPAs”. The increasing citation frequency during 2020–2023, particularly in clusters related to coral reefs and regional studies, signals a heightened global awareness of MPAs' role in mitigating climate change impacts. This review provides essential insights, informing future directions for both academic research and policymaking in marine conservation amid ongoing climatic changes.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"86 ","pages":"Article 103042"},"PeriodicalIF":5.8,"publicationDate":"2025-01-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143102007","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Although differential privacy (DP) is used to obfuscate local information and avoid data leakage, very little research exists on the neural network model performance with applied DP for datasets from online drinking water sensor monitoring. This study aims to examine the accuracy of four different neural network model architectures with DP applications. To compare the performance of the neural network model performance in total 2 215 906 augmented and experimentally obtained sensor readings were obtained from a drinking-water pilot system. Three types of contaminations at three different concentrations were applied as scenarios for anomalies in drinking water monitoring. The results achieved similar accuracy with all model architectures, with the best result showing only a 0.3% reduction in model accuracy compared with a nonprivate neural network model with 94% and 94.7% accuracy, respectively. Thus, differential privacy can be applied in the field of water quality monitoring with a reasonable decrease in the model performance.
{"title":"Application of differential privacy to sensor data in water quality monitoring task","authors":"Audris Arzovs , Sergei Parshutin , Valts Urbanovics , Janis Rubulis , Sandis Dejus","doi":"10.1016/j.ecoinf.2025.103019","DOIUrl":"10.1016/j.ecoinf.2025.103019","url":null,"abstract":"<div><div>Although differential privacy (DP) is used to obfuscate local information and avoid data leakage, very little research exists on the neural network model performance with applied DP for datasets from online drinking water sensor monitoring. This study aims to examine the accuracy of four different neural network model architectures with DP applications. To compare the performance of the neural network model performance in total 2 215 906 augmented and experimentally obtained sensor readings were obtained from a drinking-water pilot system. Three types of contaminations at three different concentrations were applied as scenarios for anomalies in drinking water monitoring. The results achieved similar accuracy with all model architectures, with the best result showing only a 0.3% reduction in model accuracy compared with a nonprivate neural network model with 94% and 94.7% accuracy, respectively. Thus, differential privacy can be applied in the field of water quality monitoring with a reasonable decrease in the model performance.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"86 ","pages":"Article 103019"},"PeriodicalIF":5.8,"publicationDate":"2025-01-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143102488","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-21DOI: 10.1016/j.ecoinf.2025.103018
Xiaoguang Liu , Shiming Yao , Zhongwu Jin , Bing Ding , Lican Ge , Shuo Guan , Weijie Wang
Oxbow lakes in the middle Yangtze River are critical habitats for protected species such as the Yangtze finless porpoise and play a vital role in biodiversity conservation. The impacts of the Three Gorges Dam (TGD) on hydrological processes and water exchange dynamics between these lakes and the Yangtze River were analyzed. Since the TGD began operation in 2003, significant changes in water level fluctuations and their rates of change have reshaped water exchange intensity and ecological balance in the oxbow lakes. A statistical model characterized the probability density distribution of daily water-level change rates, identifying distinct operation-dependent shifts, with the most dynamic changes near the 30 m threshold. An empirical threshold regression model incorporating the Langmuir adsorption formula effectively described the nonlinear relationships among water level, water-level change rate, and water exchange flow, providing a reliable predictive tool. Seasonal and interannual variations in water exchange intensity were quantified across three critical intervals: flood preparation (Interval I), peak fish migration (Interval II), and post-flood recession (Interval III). Findings revealed reduced water exchange during Interval II negatively impacted small fish populations, challenging species such as the Yangtze finless porpoise. Increased water exchange during Interval III improved water quality by reducing nutrient concentrations and enhancing dissolved oxygen levels. Regulation strategies using an exponential function demonstrated the potential to optimize water exchange intensity by controlling water level variation rates. The proposed ecological hydrological regulation framework offers a scientific basis for improving water exchange during key biological periods, ensuring habitat quality and supporting biodiversity. These findings highlight the critical role of hydrological regulation in maintaining the ecological health and functions of oxbow lakes.
{"title":"Modeling and regulation of water exchange between the oxbow lake and the middle Yangtze River","authors":"Xiaoguang Liu , Shiming Yao , Zhongwu Jin , Bing Ding , Lican Ge , Shuo Guan , Weijie Wang","doi":"10.1016/j.ecoinf.2025.103018","DOIUrl":"10.1016/j.ecoinf.2025.103018","url":null,"abstract":"<div><div>Oxbow lakes in the middle Yangtze River are critical habitats for protected species such as the Yangtze finless porpoise and play a vital role in biodiversity conservation. The impacts of the Three Gorges Dam (TGD) on hydrological processes and water exchange dynamics between these lakes and the Yangtze River were analyzed. Since the TGD began operation in 2003, significant changes in water level fluctuations and their rates of change have reshaped water exchange intensity and ecological balance in the oxbow lakes. A statistical model characterized the probability density distribution of daily water-level change rates, identifying distinct operation-dependent shifts, with the most dynamic changes near the 30 m threshold. An empirical threshold regression model incorporating the Langmuir adsorption formula effectively described the nonlinear relationships among water level, water-level change rate, and water exchange flow, providing a reliable predictive tool. Seasonal and interannual variations in water exchange intensity were quantified across three critical intervals: flood preparation (Interval I), peak fish migration (Interval II), and post-flood recession (Interval III). Findings revealed reduced water exchange during Interval II negatively impacted small fish populations, challenging species such as the Yangtze finless porpoise. Increased water exchange during Interval III improved water quality by reducing nutrient concentrations and enhancing dissolved oxygen levels. Regulation strategies using an exponential function demonstrated the potential to optimize water exchange intensity by controlling water level variation rates. The proposed ecological hydrological regulation framework offers a scientific basis for improving water exchange during key biological periods, ensuring habitat quality and supporting biodiversity. These findings highlight the critical role of hydrological regulation in maintaining the ecological health and functions of oxbow lakes.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"86 ","pages":"Article 103018"},"PeriodicalIF":5.8,"publicationDate":"2025-01-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143102484","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-20DOI: 10.1016/j.ecoinf.2025.103037
Mohammad Basyuni , Alfian Mubaraq , Rizka Amelia , Anindya Wirasatriya , Sigit Bayhu Iryanthony , Bejo Slamet , Shofiyah S. Al Mustaniroh , Novia Arinda Pradisty , Frida Sidik , Rizki Hanintyo , Elham Sumarga , Siti H. Larekeng , Severino G. Salmo III , Tadashi Kajita , Hayssam M. Ali , Anjar Dimara Sakti , Virni B. Arifanti
Mangrove forests store higher amounts of organic carbon than other forest types. Despite advancements in remote sensing, accurate mapping of mangrove biomass remains a challenge due to ecosystem complexity and varying forest structures. Although traditional in-situ methodologies have been widely used for carbon stock assessments, numerous studies have demonstrated the effectiveness of remote sensing techniques, including unmanned aerial vehicles (UAVs) for mapping mangrove biomass over larger areas. These techniques are combined with allometric equations and UAV photogrammetry to improve accuracy. This study aimed to spatially estimate aboveground biomass (AGB) in various ecosystems by integrating high-resolution digital surface models and digital terrain models (DTMs) with Lorey's height measurements. Moreover, this study utilized UAV imagery and in-situ measurements to enhance the accuracy of the carbon assessments. The integration of Lorey height into our methodology is essential, as Lorey height, which is the average height of unevenly aged forest stands, is a valuable parameter for mangrove ecosystem management. Accordingly, the correlation between UAV-derived canopy height and field measurements will be improved, resulting in more reliable AGB data in mangrove ecosystems. This study has been conducted in Budeng–Perancak, Bali, Indonesia, including restored mangroves, undisturbed mangroves, Nypa, and ponds. The UAV imagery acquisition is supported by a series of in-situ measurements to obtain field data on the forest structure (for canopy surface model [CSM] cross-validation). The AGB across the mangrove land cover types in Budeng–Perancak ranged from 2 Mg ha−1 to 480 Mg ha−1 (mean: 240 Mg ha−1), with the highest average total AGB in natural mangroves (239 Mg ha−1), followed by restored mangroves (232 Mg ha−1), indicating a successful restoration effort at Budeng–Perancak. UAVs enable detailed data collection at small spatial scales to map mangroves and obtain precise spatial information on mangrove ecosystems. This finding can improve the accuracy of greenhouse gas inventory and carbon storage estimates.
{"title":"Mangrove aboveground biomass estimation using UAV imagery and a constructed height model in Budeng–Perancak, Bali, Indonesia","authors":"Mohammad Basyuni , Alfian Mubaraq , Rizka Amelia , Anindya Wirasatriya , Sigit Bayhu Iryanthony , Bejo Slamet , Shofiyah S. Al Mustaniroh , Novia Arinda Pradisty , Frida Sidik , Rizki Hanintyo , Elham Sumarga , Siti H. Larekeng , Severino G. Salmo III , Tadashi Kajita , Hayssam M. Ali , Anjar Dimara Sakti , Virni B. Arifanti","doi":"10.1016/j.ecoinf.2025.103037","DOIUrl":"10.1016/j.ecoinf.2025.103037","url":null,"abstract":"<div><div>Mangrove forests store higher amounts of organic carbon than other forest types. Despite advancements in remote sensing, accurate mapping of mangrove biomass remains a challenge due to ecosystem complexity and varying forest structures. Although traditional in-situ methodologies have been widely used for carbon stock assessments, numerous studies have demonstrated the effectiveness of remote sensing techniques, including unmanned aerial vehicles (UAVs) for mapping mangrove biomass over larger areas. These techniques are combined with allometric equations and UAV photogrammetry to improve accuracy. This study aimed to spatially estimate aboveground biomass (AGB) in various ecosystems by integrating high-resolution digital surface models and digital terrain models (DTMs) with Lorey's height measurements. Moreover, this study utilized UAV imagery and in-situ measurements to enhance the accuracy of the carbon assessments. The integration of Lorey height into our methodology is essential, as Lorey height, which is the average height of unevenly aged forest stands, is a valuable parameter for mangrove ecosystem management. Accordingly, the correlation between UAV-derived canopy height and field measurements will be improved, resulting in more reliable AGB data in mangrove ecosystems. This study has been conducted in Budeng–Perancak, Bali, Indonesia, including restored mangroves, undisturbed mangroves, Nypa, and ponds. The UAV imagery acquisition is supported by a series of in-situ measurements to obtain field data on the forest structure (for canopy surface model [CSM] cross-validation). The AGB across the mangrove land cover types in Budeng–Perancak ranged from 2 Mg ha<sup>−1</sup> to 480 Mg ha<sup>−1</sup> (mean: 240 Mg ha<sup>−1</sup>), with the highest average total AGB in natural mangroves (239 Mg ha<sup>−1</sup>), followed by restored mangroves (232 Mg ha<sup>−1</sup>), indicating a successful restoration effort at Budeng–Perancak. UAVs enable detailed data collection at small spatial scales to map mangroves and obtain precise spatial information on mangrove ecosystems. This finding can improve the accuracy of greenhouse gas inventory and carbon storage estimates.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"86 ","pages":"Article 103037"},"PeriodicalIF":5.8,"publicationDate":"2025-01-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143101512","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-20DOI: 10.1016/j.ecoinf.2025.103039
Yanxia Wang , Xiaoyu Ni , Xiaoshuang Ma
The occurrence of Ulva prolifera (U. prolifera) can cause significant environmental damage in coastal areas, making its monitoring crucial. Remote sensing technology provides an effective tool for large-scale monitoring of U. prolifera. Most studies rely on optical images to monitor U. prolifera, which are highly dependent on weather conditions. Synthetic Aperture Radar (SAR) can penetrate clouds, rain, and fog, providing clear observations of ocean surfaces in a large scale regardless of time of day. However, current research on SAR data for U. prolifera detection primarily focuses on SAR intensity or amplitude information, while its rich polarimetric data remains underutilized. This paper presents U. prolifera Detection Network (UDNet), an intelligent detection framework based on the DeepLabV3+ deep learning model, leveraging amplitude and polarimetric information from Sentinel-1 dual-polarimetric imageries. To construct the proposed model, 2283 samples were annotated using SAR images of the Yellow Sea, of which 1737 samples were used for training and 546 samples were used for validation and testing. The well-trained model was used to detect U. prolifera in a typical coastal area from 2018 to 2021. The experimental results demonstrate that the proposed UDNet achieves superior performance with an overall accuracy of 0.9859, a mean intersection over union of 0.9198, and an F1 score of 0.9239. Spatio-temporal distribution analyses indicate that the most severe outbreak of U. prolifera in the study area occurred in 2019, with intensive occurrences in June of each year. The outbreak was more severe in the southwest region of the study area than in the northeast. Besides, it was observed that the outbreak area of U. prolifera was larger at night than that during the day, mainly driven by changes in summer temperature. In addition, a larger diurnal temperature difference generally promoted the growth of U. prolifera. These findings are instrumental in formulating management policies and taking actions to control the outbreak of U. prolifera.
{"title":"Identification and spatio-temporal analysis of Ulva prolifera in a typical coastal area using SAR imageries","authors":"Yanxia Wang , Xiaoyu Ni , Xiaoshuang Ma","doi":"10.1016/j.ecoinf.2025.103039","DOIUrl":"10.1016/j.ecoinf.2025.103039","url":null,"abstract":"<div><div>The occurrence of <em>Ulva prolifera</em> (<em>U. prolifera</em>) can cause significant environmental damage in coastal areas, making its monitoring crucial. Remote sensing technology provides an effective tool for large-scale monitoring of <em>U. prolifera</em>. Most studies rely on optical images to monitor <em>U. prolifera</em>, which are highly dependent on weather conditions. Synthetic Aperture Radar (SAR) can penetrate clouds, rain, and fog, providing clear observations of ocean surfaces in a large scale regardless of time of day. However, current research on SAR data for <em>U. prolifera</em> detection primarily focuses on SAR intensity or amplitude information, while its rich polarimetric data remains underutilized. This paper presents <em>U. prolifera</em> Detection Network (UDNet), an intelligent detection framework based on the DeepLabV3+ deep learning model, leveraging amplitude and polarimetric information from Sentinel-1 dual-polarimetric imageries. To construct the proposed model, 2283 samples were annotated using SAR images of the Yellow Sea, of which 1737 samples were used for training and 546 samples were used for validation and testing. The well-trained model was used to detect <em>U. prolifera</em> in a typical coastal area from 2018 to 2021. The experimental results demonstrate that the proposed UDNet achieves superior performance with an overall accuracy of 0.9859, a mean intersection over union of 0.9198, and an F1 score of 0.9239. Spatio-temporal distribution analyses indicate that the most severe outbreak of <em>U. prolifera</em> in the study area occurred in 2019, with intensive occurrences in June of each year. The outbreak was more severe in the southwest region of the study area than in the northeast. Besides, it was observed that the outbreak area of <em>U. prolifera</em> was larger at night than that during the day, mainly driven by changes in summer temperature. In addition, a larger diurnal temperature difference generally promoted the growth of <em>U. prolifera</em>. These findings are instrumental in formulating management policies and taking actions to control the outbreak of <em>U. prolifera</em>.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"86 ","pages":"Article 103039"},"PeriodicalIF":5.8,"publicationDate":"2025-01-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143102009","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-19DOI: 10.1016/j.ecoinf.2025.103023
Qianghao Zeng , Xuehe Lu , Suwan Chen , Xuan Cui , Haidong Zhang , Qian Zhang
Urban vegetation is pivotal in enhancing regional ecological balance and sequestering significant amounts of carbon dioxide (CO2) through photosynthesis, thereby contributing substantially to regional carbon budgets. However, the gross primary productivity (GPP) of urban vegetation remains underexplored due to the absence of robust estimation methodologies, often leading to its exclusion from global and regional carbon budgets. Advances in vegetation indices (VIs) offer promising solutions for improving the accuracy and spatial resolution of urban GPP estimation. In this study, we compared the performance of the enhanced vegetation index (EVI), normalized difference vegetation index (NDVI), near-infrared reflectance of vegetation (NIRv), and kernel normalized difference vegetation index (kNDVI) calculated from Landsat 5/7 images in estimating flux-site-level GPP and incorporated meteorological factors to construct a high-performance VI-GPP model for urban GPP estimation. Our findings demonstrated that the EVI, NIRv, and kNDVI exhibited stronger correlations with GPP dynamics and higher R2 values than did the NDVI in linear VI-GPP relationships across most plant functional types (PFTs). Exceptions were observed in evergreen broadleaf forest (EBF), evergreen needle-leaf forest (ENF), and savanna (SAV), where GPP variations were strongly influenced by temperature, shortwave radiation, and vapor pressure. Incorporating these meteorological factors significantly enhanced GPP estimation accuracy for these PFTs. Among the indices, the NIRv achieved the highest overall model performance, with an R2 of 0.60 and a root-mean-square error (RMSE) of 2.05 g C m−2 d−1 across PFTs. The kNDVI demonstrated unique advantages for specific PFTs, such as deciduous broadleaf forest (DBF) and ENF. Compared with existing VI-GPP relationships created with coarse-spatial-resolution remote sensing data, our model was more suitable for high-spatial-resolution GPP estimation in urban areas. Our results highlight the performance of the NIRv and kNDVI in urban vegetation GPP estimation and provide a solution for estimating fine-resolution GPP to reveal the importance of urban vegetation to regional carbon budgets.
{"title":"Comparing the performance of vegetation indices for improving urban vegetation GPP estimation via eddy covariance flux data and Landsat 5/7 data","authors":"Qianghao Zeng , Xuehe Lu , Suwan Chen , Xuan Cui , Haidong Zhang , Qian Zhang","doi":"10.1016/j.ecoinf.2025.103023","DOIUrl":"10.1016/j.ecoinf.2025.103023","url":null,"abstract":"<div><div>Urban vegetation is pivotal in enhancing regional ecological balance and sequestering significant amounts of carbon dioxide (CO<sub>2</sub>) through photosynthesis, thereby contributing substantially to regional carbon budgets. However, the gross primary productivity (GPP) of urban vegetation remains underexplored due to the absence of robust estimation methodologies, often leading to its exclusion from global and regional carbon budgets. Advances in vegetation indices (VIs) offer promising solutions for improving the accuracy and spatial resolution of urban GPP estimation. In this study, we compared the performance of the enhanced vegetation index (EVI), normalized difference vegetation index (NDVI), near-infrared reflectance of vegetation (NIRv), and kernel normalized difference vegetation index (kNDVI) calculated from Landsat 5/7 images in estimating flux-site-level GPP and incorporated meteorological factors to construct a high-performance VI-GPP model for urban GPP estimation. Our findings demonstrated that the EVI, NIRv, and kNDVI exhibited stronger correlations with GPP dynamics and higher R<sup>2</sup> values than did the NDVI in linear VI-GPP relationships across most plant functional types (PFTs). Exceptions were observed in evergreen broadleaf forest (EBF), evergreen needle-leaf forest (ENF), and savanna (SAV), where GPP variations were strongly influenced by temperature, shortwave radiation, and vapor pressure. Incorporating these meteorological factors significantly enhanced GPP estimation accuracy for these PFTs. Among the indices, the NIRv achieved the highest overall model performance, with an R<sup>2</sup> of 0.60 and a root-mean-square error (RMSE) of 2.05 g C m<sup>−2</sup> d<sup>−1</sup> across PFTs. The kNDVI demonstrated unique advantages for specific PFTs, such as deciduous broadleaf forest (DBF) and ENF. Compared with existing VI-GPP relationships created with coarse-spatial-resolution remote sensing data, our model was more suitable for high-spatial-resolution GPP estimation in urban areas. Our results highlight the performance of the NIRv and kNDVI in urban vegetation GPP estimation and provide a solution for estimating fine-resolution GPP to reveal the importance of urban vegetation to regional carbon budgets.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"86 ","pages":"Article 103023"},"PeriodicalIF":5.8,"publicationDate":"2025-01-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143101511","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：