Ecological Informatics最新文献

{"title":"Improved lightweight DeepLabV3+ for bare rock extraction from high-resolution UAV imagery","authors":"Pengde Lai ,&nbsp;Chao Lv ,&nbsp;Lv Zhou ,&nbsp;Shengxiong Yang ,&nbsp;Jiao Xu ,&nbsp;Qiulin Dong ,&nbsp;Meilin He","doi":"10.1016/j.ecoinf.2025.103204","DOIUrl":"10.1016/j.ecoinf.2025.103204","url":null,"abstract":"<div><div>Bare rock information extraction in karst regions is crucial for geological hazard monitoring and ecological assessment. However, in sparsely vegetated areas, bare rock exhibits similar spectral characteristics to surrounding land cover, and the boundaries are often indistinct, making it challenging for traditional classification methods to distinguish these transitional zones accurately. To address these challenges, this study proposes a bare rock extraction method based on an improved lightweight DeepLabV3+ model. MobileNetV2 is used as the backbone network, and the Channel Attention Module (CAM) and Spatial Attention Module (SAM) are introduced to enhance feature extraction capability. Results show the following: (1) When MobileNetV2 is used as the backbone of DeepLabV3+, the Accuracy, F1 score, and MIoU reach 97.39 %, 78.91 %, and 82.11 %, respectively, outperforming VGG16, Xception, SqueezeNet, and traditional segmentation models. (2) Applying the lightweight DeepLabV3+ model to bare rock identification in orthophoto imagery of the study area results in a bare rock rate error of approximately 5 %, demonstrating the practical applicability of the model. (3) After the introduction of the attention mechanism, the model's Recall, F1 score, and MIoU increased by 14.00 %, 8.37 %, and 5.62 %, respectively, remarkably enhancing identification completeness and boundary accuracy. Meanwhile, the improved model had a parameter count of 6.98 M and a computational complexity of 7.24G, achieving enhanced accuracy while maintaining computational efficiency. The research results can provide accurate bare rock information to support geological hazard monitoring and early warning, and offer new technical solutions for ecological restoration and risk assessment. (Data sets and code links: <span><span>https://figshare.com/articles/dataset/Bare_rock_dataset/28143443?file=53186633</span><svg><path></path></svg></span>).</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"89 ","pages":"Article 103204"},"PeriodicalIF":5.8,"publicationDate":"2025-05-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144089607","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Marine soundscape forecasting: A deep learning-based approach","authors":"Shashidhar Siddagangaiah","doi":"10.1016/j.ecoinf.2025.103189","DOIUrl":"10.1016/j.ecoinf.2025.103189","url":null,"abstract":"&lt;div&gt;&lt;div&gt;Advancements in autonomous monitoring technology over the past decade have led to the widespread use of marine soundscape monitoring to assess marine environments. These environments are rapidly changing and exhibit complex temporal patterns and trends across different frequencies, influenced by biotic and abiotic factors as well as extreme events. This variability introduces a high degree of unpredictability. Despite the rapid development of anomaly detection algorithms and deep-learning models for forecasting, their application to marine soundscapes remains unexplored. This study investigates the use of the unsupervised learning-based isolation forest (iForest) technique to detect anomalous events in marine soundscapes that cause sudden changes in sound levels. Additionally, it evaluates the potential of deep-learning models for estimating trends and forecasting soundscapes while identifying the factors that influence their accuracy. To address these questions, I used marine passive acoustic monitoring data collected from the Taiwan Strait in 2017. The iForest method identified a higher number of anomalies (∼17) in the lower frequency range (10–500 Hz) with a precision of 75 %, primarily due to typhoons, cold bursts, and flooding. In contrast, precision was around 50 % in the mid (500–3000 Hz) and high (3000–24,000 Hz) frequency ranges, where most anomalies resulted from sudden changes in the acoustic behaviors of fish and shrimp, respectively. To analyze trends in marine soundscapes at different temporal scales—annual, seasonal, and diurnal—the anomaly-informed NeuralProphet model was employed. Results showed that NeuralProphet effectively captured annual and seasonal trend changes compared to the traditional singular spectrum analysis method. Beyond NeuralProphet, I also tested two recently developed state-of-the-art forecasting models—time-series dense encoder (TiDE) and neural hierarchical interpolation for time series (NHiTS)—to predict marine soundscapes. In the seven-day ahead seasonal forecasting task, the NHiTS model outperformed both TiDE and NeuralProphet. The deep-learning forecasting models produced more accurate predictions in the mid (500–3000 Hz) (MAE ∼0.4–1) and high (3000–24,000 Hz) (MAE ∼1.5–3) frequency ranges, where seasonal acoustic activity from fish and shrimp strongly influenced sound levels. In contrast, forecast accuracy declined in the lower frequency range (10–500 Hz) (MAE ∼4–8), where sound levels are more stochastic due to anthropogenic and meteorological influences. The findings of this study highlight the potential of deep-learning models for forecasting and trend estimation in marine soundscapes. These models not only improve our understanding of the conditions under which trends change but also enhance our ability to anticipate anomalies and forecast failures. This capability could provide researchers and policymakers with a powerful tool for monitoring transitions and deviations across different tem","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"89 ","pages":"Article 103189"},"PeriodicalIF":5.8,"publicationDate":"2025-05-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144098352","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Bayesian networks for causal analysis in socioecological systems","authors":"Rafael Cabañas ,&nbsp;Ana D. Maldonado ,&nbsp;María Morales ,&nbsp;Pedro A. Aguilera ,&nbsp;Antonio Salmerón","doi":"10.1016/j.ecoinf.2025.103173","DOIUrl":"10.1016/j.ecoinf.2025.103173","url":null,"abstract":"<div><div>Analyzing the influence of socioeconomy on land use is an important task, as socioeconomic factors can drive changes in land use that may ultimately affect human well-being. Recognizing the key factors that induce these changes may help policymakers design more effective strategies for addressing socioeconomic alterations on land-use planning, anticipate potential challenges, and mitigate negative impacts on both the environment and society. While probabilistic graphical models have been employed for this purpose in the past, this paper proposes the application of counterfactual reasoning to enhance the analysis by quantifying the degrees of necessity and sufficiency of various socioeconomic factors influencing land uses and population growth. Specifically, we present a case study using non-experimental data from southern Spain. For this, we propose the use of structural causal models, which are kind probabilistic models for causal analysis that simplify this kind of reasoning due to their graphical representation. They can be regarded as extensions of the so-called Bayesian networks, a well known modeling tool commonly used in environmental and ecological problems. This proposed approach is particularly effective for the identification of social and ecological variables that can be used in environmental monitoring and planning, offering key advantages including enhanced interpretability, and ease of adoption by environmental researchers. Our study reveals that immigration is both necessary and sufficient for population growth. In addition, built-up areas and herbaceous crops are favored by non-mountainous terrain and by high population density, whereas natural areas and mixed crops are supported by mountainous terrain and by low population density.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"89 ","pages":"Article 103173"},"PeriodicalIF":5.8,"publicationDate":"2025-05-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144083781","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Strategic landfill site selection for sustainable waste management in Phu Yen Province, Vietnam using geospatial technologies","authors":"Diem-My Thi Nguyen ,&nbsp;Dorian Tosi Robinson ,&nbsp;Christian Zurbrügg ,&nbsp;Thi Hanh Tien Nguyen ,&nbsp;Huu-Lieu Dang ,&nbsp;Van-Manh Pham","doi":"10.1016/j.ecoinf.2025.103198","DOIUrl":"10.1016/j.ecoinf.2025.103198","url":null,"abstract":"<div><div>Solid waste management is a growing global challenge, especially in developing countries such as Vietnam, where rapid urbanisation and inadequate infrastructure intensify environmental and public health risks. Landfilling is one of the most environmentally harmful waste disposal methods. However, it remains widely used in many countries because of its cost-effectiveness. Proper disposal of solid waste is a significant priority for reducing environmental pollution. Selecting suitable landfill sites requires consideration not only of physical and environmental aspects but also of economic and social factors. In Phu Yen Province, located in south central Vietnam, solid waste management faces growing challenges in solid waste management. Limited landfill infrastructure and poor operational standards are already impacting public health and the environment. Moreover, with existing landfills approaching the end of their usable lifespans, identifying new, appropriate sites has become an urgent priority. This study introduces a novel approach that integrates a geographic information system (GIS)-based multi-criteria decision analysis (MCDA) with a fuzzy analytic hierarchy process (Fuzzy AHP) to enhance landfill site suitability assessments. This study's approach enables a holistic evaluation of economic, environmental, topographical, and social factors, thereby ensuring a more comprehensive decision-making process. The findings reveal that 45 % of the study area is very highly or highly potential for landfill sites, 28 % is of medium potential, 27 % is of low or very low potential, and 25.7 % of the existing landfill locations pose significant environmental and human health risks. A spatial distribution map obtained from a comprehensive analysis incorporating economic, social, environmental, and topographical factors helped identify potential future sites for solid waste disposal in Phu Yen Province. The methodology demonstrated in this study is highly transferable and can be applied to other low- and middle-income countries that face similar waste management challenges.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"89 ","pages":"Article 103198"},"PeriodicalIF":5.8,"publicationDate":"2025-05-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144068987","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Spatiotemporal patterns of desertification sensitivity and influencing factors across the Western Inner Mongolia Plateau, China","authors":"Yang Chen ,&nbsp;Long Ma ,&nbsp;Xixi Wang ,&nbsp;Tingxi Liu ,&nbsp;Zixu Qiao","doi":"10.1016/j.ecoinf.2025.103190","DOIUrl":"10.1016/j.ecoinf.2025.103190","url":null,"abstract":"<div><div>Desertification remains a critical global ecological and environmental challenge that threatens sustainable development. Although our understanding of desertification dynamics and their underlying drivers has improved, continued research is needed due to the region-specific nature of these processes. This study focuses on the Western Inner Mongolia Plateau in China as a case study to examine the evolution of desertification and its driving factors using a multifaceted approach, including the Mediterranean Desertification and Land Use (MEDALUS) model. Results show that the desertification sensitivity index (DSI) across the plateau ranged from 1.12 in prairie regions to 1.87 in desert areas, with a spatial gradient decreasing from west to east. Overall, the DSI exhibited a declining trend over the study period, though some areas showed localized degradation. Between 2001 and 2020, the DSI decreased across approximately 64 % of the plateau, with approximately 23 % (primarily desert regions) experiencing a significant reduction. In contrast, 36 % of the area, particularly the southeastern grasslands, saw an increase in DSI. Among the examined factors, seven—precipitation, normalized difference vegetation index (NDVI), leaf area index(LAI), drought resistance, erosion protection, fire risk, and land-use intensity—demonstrated high explanatory power greater than 0.6, highlighting their significant positive or negative impact on desertification. Additional factors such as temperature, sunshine duration, and potential evapotranspiration also influenced desertification, albeit to a lesser extent. Notably, interactions among these variables played a crucial role in shaping desertification trends. Addressing desertification, therefore, requires integrated strategies that account for the complex interplay of soil, climate, vegetation, and land management.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"89 ","pages":"Article 103190"},"PeriodicalIF":5.8,"publicationDate":"2025-05-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144098354","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Integrating UAV and Landsat data: A two-scale approach to topsoil moisture mapping in coastal wetlands","authors":"Ricardo Martínez Prentice ,&nbsp;Miguel Villoslada ,&nbsp;Raymond D. Ward ,&nbsp;Kalev Sepp","doi":"10.1016/j.ecoinf.2025.103197","DOIUrl":"10.1016/j.ecoinf.2025.103197","url":null,"abstract":"<div><div>Surface soil moisture (SSM) is a key variable influencing ecosystem dynamics, particularly in wetland systems, highlighting its importance for management. This study integrates UAV-derived high-resolution SSM maps with Landsat-based predictions to enable multiscale SSM monitoring in wetland ecosystems. UAV multispectral and thermal imagery were used to estimate the Temperature Vegetation Dryness Index (TVDI), which was calibrated with in-situ measurements of volumetric water content percentage (VWC%) to produce fine-scale SSM maps. These maps were aggregated to train and test XGBoost models using Landsat-derived predictors.</div><div>While UAV data captured fine-scale SSM variability, Landsat-based predictions provided consistency at lower spatial scales (30 m of spatial resolution from Collection-2 Level-2), with RMSE values below 10 %. Among all surveyed periods, August yielded the most reliable results. During this month—the warmest and most hydrologically dynamic—TVDI and Land Surface Temperature (LST) emerged as the strongest predictors. This also demonstrates that XGBoost model to better represent the full range of moisture conditions.</div><div>This framework addresses challenges like cloud cover in high-latitude regions and offers scalable solutions for SSM monitoring. Results contribute to the understanding of essential climate variables and support the restoration and management of coastal meadows. By bridging UAV and satellite observations, this approach provides a reliable and scalable tool for SSM assessment across diverse ecosystems. Future efforts should prioritize surveys during ecologically responsive periods, such as August, and explore the methodology's applicability in other wetland systems and long-term monitoring schemes.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"89 ","pages":"Article 103197"},"PeriodicalIF":5.8,"publicationDate":"2025-05-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143941296","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"BattyCoda: A novel open-source software for bat call annotation and classification","authors":"Gabriela C. Nunez-Mir ,&nbsp;Kevin M. Boergens ,&nbsp;Jessica C. Montoya ,&nbsp;Hannah ter Hofstede ,&nbsp;Angeles Salles","doi":"10.1016/j.ecoinf.2025.103195","DOIUrl":"10.1016/j.ecoinf.2025.103195","url":null,"abstract":"<div><div>The field of acoustic communication needs tools that facilitate the annotation and labeling of animal calls. Bat acoustic libraries gathered over the past few decades have primarily focused on compiling echolocation calls, which have been leveraged to develop machine learning algorithms capable of classifying bat species. However, because these classification methods require large training datasets, they have not yet been generalized to classify types of bat communication calls. Communication call repertoires in bats are wide, and distinct syllables occur with varying frequency, with some call types being recorded only rarely. Furthermore, collecting communication calls poses greater technical challenges, making these calls more difficult to capture reliably. Here, we present BattyCoda, an open-access, customizable tool to categorize and label bat communication call types within the repertoire of a species using small training datasets (tens to hundreds of labeled calls). In this work, we compiled an initial training dataset of 11 types of big brown bat (<em>Eptesicus fuscus</em>) calls, tested the performance of various candidate classifiers, and assessed the final classifier's training sample size sensitivity. We found that the best performing classifier achieved a balanced accuracy of ∼50 %, with common call types achieving classification accuracies over 70 %. Our tool can greatly facilitate annotating bat calls in recordings by providing accurate labels for common call types, while also assisting researchers in categorizing rarer communication calls. BattyCoda has the potential to build research capacity in the field of acoustic communication by expanding the availability of libraries including a wider range of bat calls and species, thereby enabling the exploration of new hypotheses.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"89 ","pages":"Article 103195"},"PeriodicalIF":5.8,"publicationDate":"2025-05-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143934854","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"River total dissolved gas prediction using a hybrid greedy-stepwise feature selection and bidirectional long short-term memory model","authors":"Khabat Khosravi ,&nbsp;Salim Heddam ,&nbsp;Changhyun Jun ,&nbsp;Sayed M. Bateni ,&nbsp;Dongkyun Kim ,&nbsp;Essam Heggy","doi":"10.1016/j.ecoinf.2025.103191","DOIUrl":"10.1016/j.ecoinf.2025.103191","url":null,"abstract":"<div><div>The supersaturation of total dissolved gas (TDG) in rivers serves as a critical indicator of water quality downstream of high dams. This study models TDG levels at two monitoring stations in the Columbia and Snake River Basins (USA), where high TDG concentrations were recorded. Hourly data on water temperature, barometric pressure, dam spill, sensor depth, and discharge serve as input variables for deep-learning models. Several models are developed and tested, including long short-term memory (LSTM), bidirectional LSTM (BiLSTM), gated recurrent unit (GRU), and an alternating model tree (AMT) hybridized with iterative absolute error regression (IAER) and iterative classifier optimizer (ICO) algorithms. A greedy stepwise feature selection technique (GSFST) is employed to identify the optimal model inputs. Each model is trained and evaluated at one station and validated at the second station to assess transferability and generalization capability. Model performance was compared using multiple quantitative and qualitative metrics, including the Nash–Sutcliffe Efficiency and uncertainty coefficient. Additionally, Friedman and Wilcoxon signed-rank tests confirmed statistically significant differences between models. Dam spills emerged as the most influential predictor of TDG levels at both sites. The GSFST selected the optimal input combination, including dam spill, water temperature, barometric pressure, and sensor depth. Among all models, GSFST-BiLSTM achieved the highest predictive accuracy, with Nash–Sutcliffe values of 0.95 (testing) and 0.90 (validation) and uncertainty coefficients of 5.2 % and 7.0 %, respectively. These findings demonstrate that GSFST-BiLSTM provides a robust and transferable framework for TDG prediction, with the potential for broader application pending further validation.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"90 ","pages":"Article 103191"},"PeriodicalIF":5.8,"publicationDate":"2025-05-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144124806","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Forest aboveground biomass retrieval integrating ICESat-2, Landsat-8, and environmental factors","authors":"Sunjie Ma ,&nbsp;Jisheng Xia ,&nbsp;Chun Wang ,&nbsp;Zhifang Zhao ,&nbsp;Fuyan Zou ,&nbsp;Maolin Zhang ,&nbsp;Guize Luan ,&nbsp;Ci Li ,&nbsp;Xi Tu ,&nbsp;Letian Li","doi":"10.1016/j.ecoinf.2025.103194","DOIUrl":"10.1016/j.ecoinf.2025.103194","url":null,"abstract":"<div><div>The synergistic integration of optical imSagery and LiDAR data provides a comprehensive spatial framework for the precise estimation of aboveground biomass (AGB). However, the technical pathway for AGB estimation in complex mountainous regions using multi-source heterogeneous data, including active and passive remote sensing and environmental data, requires further validation. This study proposes a novel framework for high-resolution AGB retrieval by integrating ICESat-2 LiDAR and Landsat-8 data, along with meteorological and topographic factors. AGB estimates were derived from ICESat-2 footprints using second-class forest survey data from the Jinsha River Basin, China. Relationships between canopy metrics and AGB were analyzed across beam types using LASSO and random forest (RF) models. The optimized RF model was then used to generate wall-to-wall AGB maps incorporating Landsat-8, meteorological, and topographic variables. The Nighttime-Strong beam achieved the highest AGB retrieval accuracy (R² = 0.71), followed by the Nighttime-Weak beam (R² = 0.69), all beams combined (R² = 0.68), the Daytime-Strong beam (R² = 0.68), and the Daytime-Weak beam (R² = 0.55); the LASSO model outperformed the RF model. In the AGB retrieval model using canopy metrics, mean canopy height, relative canopy height, canopy coverage, and canopy quadratic mean were strong predictors (correlation coefficients of 0.67, 0.65, 0.63, and 0.62, respectively). Adding meteorological and topographic data substantially improved wall-to-wall AGB mapping, with topography having a greater impact than meteorology. In conclusion, AGB retrieval accuracy can be markedly improved by using ICESat-2 Nighttime-Strong beams combined with meteorological and topographic datasets. This study proposes a more precise and effective methodology for forest monitoring in complex environments.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"89 ","pages":"Article 103194"},"PeriodicalIF":5.8,"publicationDate":"2025-05-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143932086","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Exploiting facial side similarities to improve AI-driven sea turtle photo-identification systems","authors":"Lukáš Adam ,&nbsp;Kostas Papafitsoros ,&nbsp;Claire Jean ,&nbsp;ALan F. Rees ,&nbsp;Vojtěch Čermák","doi":"10.1016/j.ecoinf.2025.103158","DOIUrl":"10.1016/j.ecoinf.2025.103158","url":null,"abstract":"<div><div>Animal photo-identification (photo-ID), the process of identifying individual animals from images, has proven to be a valuable tool for various studies on sea turtles, increasing the knowledge of their ecology and informing conservation efforts. Photo-ID in sea turtles is predominantly based on the geometric patterns of the scales of their two head sides, which are unique to every individual and different from side to side. As such, both manual and automated photo-ID techniques are traditionally performed under a side-specific setting. There, an image showing a single profile of an unknown individual is compared only to images showing the same side of previously identified individuals. In this paper, we show for the first time an inherent visual similarity between left and right facial profiles of the same individuals in three sea turtle species. We do so by employing two state-of-the-art automated neural network-based photo-ID methods, one local feature-based and one deep embedding-based, designed to rank profiles based on their similarities. Both methods rank the similarity of the left and right profiles of the same individual higher than those of different individuals. These similarities are detectable even when images are taken years apart under diverse conditions. We further show that the exploitation of this similarity results in improved accuracies when compared to the traditional side-specific photo-ID setting. Our results indicate two concrete guidelines for improving automated sea turtle photo-ID workflows. When trying to match a photo of a given profile, searches should not be restricted only to photos of the same profile. As the first method of choice, a deep embedding model finely-trained using a photo-database of the focal sea turtle population should be used. In the absence of such training database, a neural network-based local feature method is preferable, but in that case searches should be performed with both the original query image and its horizontally flipped version.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"89 ","pages":"Article 103158"},"PeriodicalIF":5.8,"publicationDate":"2025-05-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143935028","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}