Ecological Informatics最新文献

{"title":"Investigating changes of forest aboveground biomass induced by Moso bamboo expansion with terrestrial laser scanner","authors":"Rui Jiang ,&nbsp;Jiayuan Lin ,&nbsp;Xianwei Zhang ,&nbsp;Meiqi Kang","doi":"10.1016/j.ecoinf.2024.102812","DOIUrl":"10.1016/j.ecoinf.2024.102812","url":null,"abstract":"<div><p>As a typical clonal plant, Moso bamboo expands excessively worldwide, causing changes in various aspects of the native forest ecosystem. Among these aspects, aboveground biomass (AGB) is a key indicator characterizing forest productivity and carbon sequestration. However, it is difficult to track AGB changes of a fixed plot in a relatively short period. In this paper, we utilized terrestrial laser scanner (TLS) to investigate AGB changes resulting from the intrusion of Moso bamboo using the space-for-time substitution method. Three sample plots including a China fir stand, a mixed stand and a pure Moso bamboo stand were chosen at an ecotone to represent the different stages of bamboo expansion in Hutou Village, Chongqing, China. Their point clouds were first scanned using TLS, and then segmented into individual plants through refinedly processing the stem intersections. Subsequently, tree and bamboo classification was achieved via combining the structural features, stem texture features, and point distribution features of individual plants. Finally, the compatible biomass models were employed to estimate plant AGBs and analyze the changes. As a result, the overall classification accuracy of trees and bamboos was improved to 92.67 %. The AGB per unit area initially increased and subsequently decreased at three stages of Moso bamboo expansion (5.83 kg/m², 6.04 kg/m² and 5.36 kg/m²), and the AGB differences among individual plants showed the similar tendency. Notably, the average AGB of individual China firs in mixed stand (78.97 kg) was higher than that in the pure stand (70.41 kg), so did the average AGB of individual Moso bamboos (21.22 kg vs 18.70 kg). These results indicated that maintaining a certain degree of tree-bamboo mixture was beneficial for improving the gross forest AGB.</p></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":null,"pages":null},"PeriodicalIF":5.8,"publicationDate":"2024-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S1574954124003546/pdfft?md5=22187fd46ce7caaaea9247414ea7090e&pid=1-s2.0-S1574954124003546-main.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142135938","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}

{"title":"A new approach to estimate total nitrogen concentration in a seasonal lake based on multi-source data methodology","authors":"Xianqiang Xia ,&nbsp;Jiayi Pan ,&nbsp;Jintao Pei","doi":"10.1016/j.ecoinf.2024.102807","DOIUrl":"10.1016/j.ecoinf.2024.102807","url":null,"abstract":"<div><p>Nitrogen, a key limiter in lake eutrophication, presents serious threats to both human health and ecological balance. Despite its non-optically active nature, this study introduces an advanced retrieval approach for total nitrogen, utilizing a synthesis of multi-source data and sophisticated machine learning algorithms to markedly boost estimation precision. This innovative method integrates environmental variables, such as water temperature, depth, and flow rate with spectral reflectance, significantly enhancing the predictive accuracy of our machine learning models with high stability. The models tested, including Support Vector Machine (SVM), Random Forest (RF), Extreme Gradient Boosting (XGB), Multilayer Perceptron (MLP), and Convolutional Neural Networks (CNN), with XGB outperforming others by achieving robust metrics: an R² of 0.78, a Mean Absolute Error (MAE) of 0.21 mg/L, and a Mean Absolute Percentage Error (MAPE) of 16.04 %. Applying the optimized XGB model, we documented fluctuations in nitrogen concentrations within Poyang Lake across different hydrological phases in 2021, revealing the lowest nitrogen levels during the flood season and the highest in low water periods, with high concentrations at the inlets of the North Branch of the Ganjiang River and the Raohe River estuaries. Monte Carlo simulations reveal that the model is not much sensitive to input feature errors, validating its stability. The approach proposed in this study may help more precise total nitrogen retrieval in other similar lake waters.</p></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":null,"pages":null},"PeriodicalIF":5.8,"publicationDate":"2024-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S1574954124003492/pdfft?md5=3fbee532adeef089e743a635d2cf4108&pid=1-s2.0-S1574954124003492-main.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142149318","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}

{"title":"Spatiotemporal changes and driving mechanism of ecosystem carbon sink in karst peak cluster depression basin in Southwest Guangxi based on the interaction of “water-rock-soil-air-biology”","authors":"Song Huang ,&nbsp;Yichao Tian ,&nbsp;Qiang Zhang ,&nbsp;Jin Tao ,&nbsp;Yali Zhang ,&nbsp;Junliang Lin","doi":"10.1016/j.ecoinf.2024.102800","DOIUrl":"10.1016/j.ecoinf.2024.102800","url":null,"abstract":"<div><p>Research on ecosystem carbon sinks is vital for developing policies to reduce emissions and enhance carbon sequestration. Analyzing the temporal and spatial trends of carbon sinks and their driving mechanisms is crucial for guiding policy and measures. Traditional methods for estimating carbon sinks in karst regions focus primarily on vegetation-soil carbon sinks, mainly comprising net primary productivity (NPP) and soil heterotrophic respiration (Rh). However, these methods overlook the unique karst carbon sink (KCS), leading to significant uncertainties in carbon sink evaluations. This study focuses on the karst peak cluster depression basin in southwest Guangxi and proposes a Karst Ecosystem Carbon Sink Assessment System (KECAS) based on the “water-rock-soil-air-biology” coupling framework, considering lithological characteristics. Theil-Sen trend analysis and restricted cubic spline (RCS) regression were used to explore the temporal and spatial variations and driving mechanisms of carbon sinks in the area from 2000 to 2022. The results show the following: (1) The total carbon sink flux (TCSF) in the study area has shown an upward trend over time, with a 23-year average ranging from 538 to 929 g CO₂ m⁻² a⁻¹ and a growth rate of 3.1039a⁻¹. This means TCSF increases by 6.78 g CO₂ m⁻² a⁻¹ each year. (2) Spatially, areas with increasing carbon sinks cover 34.17 % of the study area, mainly in Chongzuo, Nanning, and Wenshan Prefecture. Areas with decreasing carbon sinks cover 11.79 % of the total area, mainly in Chongzuo, with smaller areas in Nanning and Baise. The mean TCSF distribution is higher in the northwest and lower in the southeast. (3) The critical thresholds for influencing factors were identified using RCS regression: precipitation (1430 mm), temperature (17.9 °C), evapotranspiration (855 mm), NDVI (0.63), DEM (637 m), slope (16.3°), nighttime light index (6.21), and population density (421 people/km²). (4) Areas with high carbon sink supply potential, based on natural and human thresholds, account for 25.09 % and 92.97 % of the southwestern Guangxi study area, respectively. When considering both natural and human conditions, high potential areas account for approximately 22.31 % of the total study area. These findings provide a scientific basis for carbon sink protection and management and offer valuable references for government decision-making.</p></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":null,"pages":null},"PeriodicalIF":5.8,"publicationDate":"2024-08-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S157495412400342X/pdfft?md5=76212fd6fa752eb614b26720ded4f333&pid=1-s2.0-S157495412400342X-main.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142135939","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}

{"title":"Seagrass classification using unsupervised curriculum learning (UCL)","authors":"Nosheen Abid ,&nbsp;Md Kislu Noman ,&nbsp;György Kovács ,&nbsp;Syed Mohammed Shamsul Islam ,&nbsp;Tosin Adewumi ,&nbsp;Paul Lavery ,&nbsp;Faisal Shafait ,&nbsp;Marcus Liwicki","doi":"10.1016/j.ecoinf.2024.102804","DOIUrl":"10.1016/j.ecoinf.2024.102804","url":null,"abstract":"<div><p>Seagrass ecosystems are pivotal in marine environments, serving as crucial habitats for diverse marine species and contributing significantly to carbon sequestration. Accurate classification of seagrass species from underwater images is imperative for monitoring and preserving these ecosystems. This paper introduces Unsupervised Curriculum Learning (UCL) to seagrass classification using the <em>DeepSeagrass</em> dataset. UCL progressively learns from simpler to more complex examples, <em>enhancing</em> the model's ability to discern seagrass features in a curriculum-driven manner. Experiments employing state-of-the-art deep learning architectures, convolutional neural networks (CNNs), show that UCL achieved overall 90.12 % precision and 89 % recall, which significantly improves classification accuracy and robustness, outperforming some traditional supervised learning approaches like SimCLR, and unsupervised approaches like Zero-shot CLIP. The methodology of UCL involves four main steps: high-dimensional feature extraction, pseudo-label generation through clustering, reliable sample selection, and fine-tuning the model. The iterative UCL framework refines CNN's learning of underwater images, demonstrating superior accuracy, generalization, and adaptability to unseen seagrass and background samples of undersea images. The findings presented in this paper contribute to the advancement of seagrass classification techniques, providing valuable insights into the conservation and management of marine ecosystems. The code and dataset are made publicly available and can be assessed here: <span>https://github.com/nabid69/Unsupervised-Curriculum-Learning—UCL.</span></p></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":null,"pages":null},"PeriodicalIF":5.8,"publicationDate":"2024-08-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S1574954124003467/pdfft?md5=4f11af008f2fed73b50d654a9fa27b94&pid=1-s2.0-S1574954124003467-main.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142128810","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}

{"title":"Mapping high-resolution XCO2 concentrations in China from 2015 to 2020 based on spatiotemporal ensemble learning model","authors":"Weican Liu ,&nbsp;Rong Li ,&nbsp;Jun Cao ,&nbsp;Congwu Huang ,&nbsp;Fan Zhang ,&nbsp;Meigen Zhang","doi":"10.1016/j.ecoinf.2024.102806","DOIUrl":"10.1016/j.ecoinf.2024.102806","url":null,"abstract":"<div><p>High-resolution column-averaged dry air mole fraction of CO₂ (XCO₂) data is crucial for understanding the spatiotemporal patterns of XCO₂ and for mitigating carbon emissions. Due to the limited scanning range of sensors and strict inversion conditions, satellite-retrieved XCO₂ data are often significantly incomplete. Machine learning models are widely used to fill these gaps in satellite XCO₂ data. However, the limitations of individual machine learning models and the complexity of the spatial distribution of XCO₂ mean that the accuracy of XCO₂ predictions still needs improvement. In this study, a new spatiotemporal stacked ensemble learning model (STEL) was developed by combining random forest (RF), extremely randomized trees (ERT), extreme gradient boosting (XGBoost), optical gradient boosting (LightGBM), and categorical boosting (CatBoost) using the stacking ensemble learning methodology. Considering the spatiotemporal heterogeneity of XCO₂, a novel spatiotemporal weighting feature was constructed as part of the model's input parameters. Finally, the XCO₂ observed by Orbiting Carbon Observatory 2 (OCO-2) was reconstructed using STEL, and a monthly mean XCO₂ dataset covering China from 2015 to 2020 was generated at a spatial resolution of 0.1°. The results show that STEL exhibits superior performance and generalization capabilities compared to individual machine-learning models. R² RMSE and MAPE were 0.9624, 1.0023 ppm, and 0.1583 % on the test set, and 0.8970, 1.4213 ppm, and 0.2475 % for R², RMSE, and MAPE in ground validation, respectively. In 10-fold cross-validation, STEL's RMSE was reduced by 9.52 % compared to the best-performing single model (RF). The spatiotemporal trend of CO₂ in China from 2015 to 2020 was analyzed using STEL XCO₂ data. The results indicate that this dataset accurately reflects the spatiotemporal heterogeneity of XCO₂ distribution at a fine scale. Overall, XCO₂ exhibited a spatiotemporal pattern of “high in the east and low in the west” and “high in spring and low in summer.” Except in summer, high XCO₂ values were mainly distributed in the North China Plain. XCO₂ trends and hotspots showed considerable spatial variation. The Pearl River Delta and Yangtze River Delta urban agglomerations have the fastest XCO₂ growth rates, and the distribution of XCO₂ hotspots is consistent with the distribution of population and economic centers. In the sparsely populated northwest of China, XCO₂ is growing rapidly due to increased thermal power generation and coal mining. XCO₂ hotspots in Northwest China are mainly located in Xinjiang, Ningxia, and Inner Mongolia. The methodology and data presented are useful for further research on carbon emissions, carbon sinks, and climate change.</p></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":null,"pages":null},"PeriodicalIF":5.8,"publicationDate":"2024-08-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S1574954124003480/pdfft?md5=4c160e2eaec9c51f025fa57cafed70e8&pid=1-s2.0-S1574954124003480-main.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142098109","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}

{"title":"Developing a dynamic energy budget model to project potential effects of deep-sea mining plumes on the Atlantic deep-sea mussel, Bathymodiolus azoricus","authors":"Irene Martins ,&nbsp;Alexandra Guerra ,&nbsp;Cândida Gomes Vale ,&nbsp;Cândido Xavier ,&nbsp;Inês Martins ,&nbsp;Marlene Pinheiro ,&nbsp;Teresa Neuparth ,&nbsp;Joana R. Xavier ,&nbsp;Pedro Duarte ,&nbsp;Miguel M. Santos ,&nbsp;Ana Colaço","doi":"10.1016/j.ecoinf.2024.102803","DOIUrl":"10.1016/j.ecoinf.2024.102803","url":null,"abstract":"<div><p>Due to the consistent lack of Environmental Risk Assessment (ERA) for deep-sea mining scenarios, the potential impacts of this industry on marine ecosystems remain largely unknown. In order to fill this gap, a Dynamic Energy Budget (DEB) model was developed to study the consequences of toxic sediment plumes derived from deep-sea mining on the energy budget of the Atlantic deep-sea mussel, <em>Bathymodiolus azoricus.</em> Model calibration was based on environmental conditions observed at the Menez Gwen (MG) vent field (Mid-Atlantic Ridge- MAR), assuming a <em>B. azoricus</em> lifespan of 10 years and a maximum shell length of 119 mm. Scenario simulations were conducted to mimic the effects of increased concentrations of toxic sediment plumes on mussel filtration rates, the absorption of reduced substrates by their endosymbionts, and the energetic costs associated with metal toxicity. Data were sourced from <em>B. azoricus</em> and, when necessary, from proxy species. One disturbance scenario (EF1) incorporated measured rates and realistic parameters, while the other (EF2) was intentionally designed to encompass cumulative effects and uncertainties, representing a potential worst-case scenario. Both disturbance scenarios were initiated at three different timings (0, 1200 and 2400 days) to accommodate the mining effects at different stages of the mussels' life cycle. Results indicate that <em>B. azoricus</em> is significantly impacted by toxic sediment plumes, particularly during earlier life stages, potentially leading to severe growth impairment and mortality. These results were integrated into a food web model of the MG vent field, revealing that disruptions to the energetic balance of the vent mussel have widespread consequences for the entire ecosystem. Overall, we argue that this numerical framework offers a valuable tool for conducting ERA and Environmental Impact Assessments (EIA) in the context of industrial deep-sea mining.</p></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":null,"pages":null},"PeriodicalIF":5.8,"publicationDate":"2024-08-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S1574954124003455/pdfft?md5=09f5045f6e0cae84b0c970b0bb4557a7&pid=1-s2.0-S1574954124003455-main.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142117506","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}

{"title":"Promoting scholarship in the fields of computational ecology and ecological data science","authors":"George Arhonditsis","doi":"10.1016/j.ecoinf.2024.102799","DOIUrl":"10.1016/j.ecoinf.2024.102799","url":null,"abstract":"","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":null,"pages":null},"PeriodicalIF":5.8,"publicationDate":"2024-08-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142194420","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}

{"title":"Camouflage detection: Optimization-based computer vision for Alligator sinensis with low detectability in complex wild environments","authors":"Yantong Liu ,&nbsp;Sai Che ,&nbsp;Liwei Ai ,&nbsp;Chuanxiang Song ,&nbsp;Zheyu Zhang ,&nbsp;Yongkang Zhou ,&nbsp;Xiao Yang ,&nbsp;Chen Xian","doi":"10.1016/j.ecoinf.2024.102802","DOIUrl":"10.1016/j.ecoinf.2024.102802","url":null,"abstract":"<div><p><em>Alligator sinensis</em> is an extremely rare species that possesses excellent camouflage, allowing it to fit perfectly into its natural environment. The use of camouflage makes detection difficult for both humans and automated systems, highlighting the importance of modern technologies for animal monitoring. To address this issue, we present YOLO v8-SIM, an innovative detection technique specifically developed to significantly enhance the identification precision. YOLO v8-SIM utilizes a sophisticated dual-layer attention mechanism, an optimized loss function called inner intersection-over-union (IoU), and a technique called slim-neck cross-layer hopping. The results of our study demonstrate that the model achieves an accuracy rate of 91 %, a recall rate of 89.9 %, and a mean average precision (mAP) of 92.3 % and an IoU threshold of 0.5. In addition, the model operates at a frame rate of 72.21 frames per second (FPS) and excels at accurately recognizing objects that are partially visible or smaller in size. To further improve our initiatives, we suggest creating an open-source collection of data that showcases <em>A. sinensis</em> in its native environment while using camouflage techniques. These developments collectively enhance the ability to detect disguised animals, thereby promoting the monitoring and protection of biodiversity, and supporting ecosystem sustainability.</p></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":null,"pages":null},"PeriodicalIF":5.8,"publicationDate":"2024-08-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S1574954124003443/pdfft?md5=b051ad19e91be804a592cc7522e3fc43&pid=1-s2.0-S1574954124003443-main.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142128809","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}

{"title":"Declining planetary health as a driver of camera-trap studies: Insights from the web of science database","authors":"Thakur Dhakal ,&nbsp;Tae-Su Kim ,&nbsp;Seong-Hyeon Kim ,&nbsp;Shraddha Tiwari ,&nbsp;Seung-Hyun Woo ,&nbsp;Do-Hun Lee ,&nbsp;Gab-Sue Jang","doi":"10.1016/j.ecoinf.2024.102801","DOIUrl":"10.1016/j.ecoinf.2024.102801","url":null,"abstract":"<div><p>Planetary health is crucial to human well-being, ecosystem sustainability, and biodiversity preservation. In this context, camera traps are an effective remote sensing tool for monitoring biodiversity. Given the rising importance of understanding biodiversity patterns and trends, this study examines possible factors influencing camera-trap studies and provides bibliometric insights from 2377 publications indexed in the Web of Science (WoS). To explore the potential drivers of camera-trap research growth, we used a logistic model based on specific variables, including global gross domestic product, temperature growth, a planetary health measure the declining living planet index, and human population growth. The living planet index was identified as a statistically significant driver of camera-trap research growth (<em>p</em>-value &lt;0.01), suggesting that curiosity regarding other living beings influences studies. Through the bibliometric analysis, we observed that camera-trap studies are predominantly conducted in the United States, followed by England and Australia, with a notable upward trend over recent years. These studies align with sustainable development goal 15 (Life on Land) and are primarily classified under the ecology category in WoS. Further, we have visualized the network of co-occurrence of authors and authors' affilation regions, keywords, and keywords plus documents. Overall, this study assesses ecological and conservation informatics and provides a reference to scholars, policymakers, and decision-makers.</p></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":null,"pages":null},"PeriodicalIF":5.8,"publicationDate":"2024-08-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S1574954124003431/pdfft?md5=1b75bbdb65094dbf6e4d475c596af5e8&pid=1-s2.0-S1574954124003431-main.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142098111","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}

{"title":"A joint time and spatial attention-based transformer approach for recognizing the behaviors of wild giant pandas","authors":"Jing Liu ,&nbsp;Jin Hou ,&nbsp;Dan Liu ,&nbsp;Qijun Zhao ,&nbsp;Rui Chen ,&nbsp;Xiaoyuan Chen ,&nbsp;Vanessa Hull ,&nbsp;Jindong Zhang ,&nbsp;Jifeng Ning","doi":"10.1016/j.ecoinf.2024.102797","DOIUrl":"10.1016/j.ecoinf.2024.102797","url":null,"abstract":"<div><p>Wild giant pandas, an endangered species exclusive to China, are a focus of conservation efforts. The behavior of giant pandas reflects their health conditions and activity capabilities, which play an important role in formulating and implementing conservation measures. Researching and developing efficient behavior recognition methods based on deep learning can significantly advance the study of wild giant panda behavior. This study introduces, for the first time, a transformer-based behavior recognition method termed PandaFormer, which employs time-spatial attention to analyze behavioral temporal patterns and estimate activity spaces. The method integrates advanced techniques such as cross-fusion recurrent time encoding and transformer modules, which handle both temporal dynamics and spatial relationships within panda behavior videos. First, we design cross-fusion recurrent time encoding to represent the occurrence time of behaviors effectively. By leveraging the multimodal processing capability of the transformer, we input time and video tokens into the transformer module to explore the relation between behavior and occurrence time. Second, we introduce relative temporal weights between video frames to enable the model to learn sequential relationships. Finally, considering the fixed position of the camera during recording, we propose a spatial attention mechanism based on the estimation of the panda's activity area. To validate the effectiveness of the model, a video dataset of wild giant pandas, encompassing five typical behaviors, was constructed. The proposed method is evaluated on this video-level annotated dataset. It achieves a Top-1 accuracy of 92.25 % and a mean class precision of 91.19 %, surpassing state-of-the-art behavior recognition algorithms by a large margin. Furthermore, the ablation experiments validate the effectiveness of the proposed temporal and spatial attention mechanisms. In conclusion, the proposed method offers an effective way of studying panda behavior and holds potential for application to other wildlife species.</p></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":null,"pages":null},"PeriodicalIF":5.8,"publicationDate":"2024-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S157495412400339X/pdfft?md5=789f7bb46c25667b7b6903e3a1edf5d4&pid=1-s2.0-S157495412400339X-main.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142098112","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}