Sandra Hervías-Parejo , Anna Traveset , Manuel Nogales , Ruben Heleno , John Llewelyn , Giovanni Strona
{"title":"Sampling biases across interaction types affect the robustness of ecological multilayer networks","authors":"Sandra Hervías-Parejo , Anna Traveset , Manuel Nogales , Ruben Heleno , John Llewelyn , Giovanni Strona","doi":"10.1016/j.ecoinf.2025.103183","DOIUrl":"10.1016/j.ecoinf.2025.103183","url":null,"abstract":"<div><div>Ecological communities rely on complex networks of species interactions. While traditional studies often focus on single interaction types (e.g. plant-pollinator or host-pathogen), there is growing recognition of the need to consider multiple interaction types to accurately model community dynamics. Multilayer networks can be used to model multiple interaction types simultaneously, but building them poses challenges due to the different sampling techniques and expertise needed for documenting different interaction types. This can introduce biases that affect the completeness of data across layers (interaction types). The extent to which such biases affect multilayer network properties remain unclear. Here, we explored this issue using empirical interaction data collected through standardized field sampling in three archipelagos along a latitudinal gradient (the Balearic, Canary, and Galapagos islands). Based on these observations, we compiled three multilayer networks, each incorporating three types of plant-animal interactions: plant-pollinator, plant-herbivore, and plant-seed disperser. We then enhanced these networks by adding interactions from the literature. The observed and enhanced multilayer networks were compared to evaluate how the quantity and bias of missing information affected network properties. In the enhanced networks, the number of herbivore, pollinator and seed disperser interactions exceeded those from the observed networks by, on average, 82 %, 62 % and 96 %, respectively. The species present in the enhanced networks but missing in the observed networks exhibited distinct structural properties. These sampling biases affected both static and dynamic network properties, and the effects varied notably across archipelagos. Observed networks from the Balearic and Canary Islands were less robust to plant removal than their enhanced counterparts, while the opposite was true for the Galapagos. This study, the first to examine the effects of sampling bias on inferred robustness of ecological multilayer networks, reveals that missing data can have complex, hidden effects on modelled network dynamics. Missing data could, therefore, have important implications for predicting and mitigating species loss.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"89 ","pages":"Article 103183"},"PeriodicalIF":5.8,"publicationDate":"2025-05-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143918609","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}
Ruiwu Zhang , Ruru Deng , Jun Ying , Cong Lei , Jiayi Li , Yu Guo , Tongtong Zhao
{"title":"Remote sensing algorithm for dissolved organic carbon in the Laptev Sea: Correction of photobleaching effect using spectral slope","authors":"Ruiwu Zhang , Ruru Deng , Jun Ying , Cong Lei , Jiayi Li , Yu Guo , Tongtong Zhao","doi":"10.1016/j.ecoinf.2025.103177","DOIUrl":"10.1016/j.ecoinf.2025.103177","url":null,"abstract":"<div><div>The absorption coefficient of colored dissolved organic matter (<span><math><msub><mi>α</mi><mi>CDOM</mi></msub></math></span>) is a critical optical parameter for quantifying dissolved organic carbon (DOC). However, photobleaching significantly reduces <span><math><msub><mi>α</mi><mi>CDOM</mi></msub></math></span>, leading to uncertainties in DOC concentration estimation, an issue that has not received widespread attention. Drawing on in situ measurements from the Laptev Sea, this study proposes a method to correct for photobleaching using the spectral slope (S<sub>275–295</sub>). Setting a threshold for S<sub>275–295</sub> identifies areas that are significantly affected by photobleaching. To assess the applicability of this method, a stratified estimation model analyses the relationship between <span><math><msub><mi>α</mi><mi>CDOM</mi></msub></math></span> and DOC concentration before and after correction at different water depths. A remote sensing inversion algorithm for DOC was also developed based on <span><math><msub><mi>α</mi><mi>CDOM</mi></msub></math></span> and remote sensing reflectance data. Results indicate that <span><math><msub><mi>α</mi><mi>CDOM</mi></msub><mfenced><mn>443</mn></mfenced></math></span> effectively characterises DOC concentration across different water depths. After correction, the photobleaching-induced error decreases by approximately 8.04 %, significantly improving the non-linear fitting accuracy of <span><math><msub><mi>α</mi><mi>CDOM</mi></msub><mfenced><mn>443</mn></mfenced></math></span> with DOC concentration in the surface water layer (0-20 m). Results for depths greater than 20 m remain essentially unchanged, although incorporating temperature and salinity improves the linear correlation with DOC, with some uncertainties persisting. The correction method is therefore most applicable to surface waters. Remote sensing results show that this method reduces DOC overestimation in coastal areas by 12 %, improving fitting accuracy and minimising error distribution. This study highlights the impact of photobleaching on DOC estimation and introduces a correction model that enhances the accuracy of remote sensing-based DOC retrieval, thereby supporting marine carbon cycle monitoring</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"89 ","pages":"Article 103177"},"PeriodicalIF":5.8,"publicationDate":"2025-04-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143902479","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}
Laurence A. Clarfeld , Katherina D. Gieder , Robert Abrams , Christopher Bernier , Joseph Cahill , Susan Staats , Scott Wixsom , Therese M. Donovan
{"title":"Two-stage models improve machine learning classifiers in wildlife research: A case study in identifying false positive detections of Ruffed Grouse","authors":"Laurence A. Clarfeld , Katherina D. Gieder , Robert Abrams , Christopher Bernier , Joseph Cahill , Susan Staats , Scott Wixsom , Therese M. Donovan","doi":"10.1016/j.ecoinf.2025.103166","DOIUrl":"10.1016/j.ecoinf.2025.103166","url":null,"abstract":"<div><div>Autonomous recording units are increasingly being used to monitor wildlife on large geographic and temporal scales, paired with machine learning (ML) to automate detection of wildlife. However, false positive detections from ML classifiers can result in erroneous ecological models that can lead to misguided management and conservation actions. We used a two-stage general approach to understand and reduce false positive detections, a technique in which outputs of the primary classification model are passed to a secondary classification model to yield the probability that a detection from the primary model is a true positive detection. This approach is demonstrated on two open-source models that detect Ruffed Grouse (<em>Bonasa umbellus</em>). We analyzed over 9500 h of acoustic data collected in 2022–2023 from the Green Mountain National Forest in Vermont, USA, and found the two models detected different types of acoustic signals associated with differing life history traits. The first model yielded 4106 detections (71.5 % true positives) while the second model yielded 524 detections (17.0 % true positives). Secondary logistic regression models separated true positives and false positives with high accuracy (84.5 % and 89.8 % respectively). Our findings go beyond improving Ruffed Grouse monitoring and conservation efforts to, more broadly, illustrate how two-stage ML approaches can improve the use of model-derived detections in wildlife research.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"89 ","pages":"Article 103166"},"PeriodicalIF":5.8,"publicationDate":"2025-04-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143924257","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}
Fatima Elshukri , Noor Hussam Abusirriya , Nathan Joseph Braganza , Abdulkarim Ahmed , Odi Fawwaz Alrebei
{"title":"Temporal and spatial pattern analysis and forecasting of methane: Satellite image processing","authors":"Fatima Elshukri , Noor Hussam Abusirriya , Nathan Joseph Braganza , Abdulkarim Ahmed , Odi Fawwaz Alrebei","doi":"10.1016/j.ecoinf.2025.103176","DOIUrl":"10.1016/j.ecoinf.2025.103176","url":null,"abstract":"<div><div>Atmospheric dispersion modeling is a critical tool in environmental research, offering insights into spatial and temporal patterns of pollutants. This study introduces an innovative approach leveraging remote sensing technology to analyze and predict methane (CH<sub>4</sub>) levels, specifically focusing on Qatar. Utilizing data from the Sentinel-5P satellite, captured through the Tropospheric Monitoring Instrument (TROPOMI), this research presents a detailed examination of methane concentrations. The methodology includes generating daily, monthly, and yearly average images, alongside Sobel gradient images to enhance the analysis of daily and monthly variations. A thresholding technique is applied to each month's data to identify critical methane concentration levels. Moreover, the study extends to forecasting methane levels for the latter half of 2024 and the entirety of 2025. These predictions are rigorously validated by comparing the predicted methane concentrations with observed data, resulting in a Root Mean Square Error (RMSE) that underscores the model's predictive accuracy. The R-squared (R<sup>2</sup>) value further demonstrates the model's robustness, particularly in scenarios where conventional prediction methods would be hampered by incomplete datasets. This research not only advances the understanding of methane dynamics in arid regions but also illustrates the potential of remote sensing as a cost-effective alternative to traditional data-intensive approaches. The accompanying Python code, detailed in the Appendix, is made publicly available to facilitate further research and application in similar environmental studies.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"89 ","pages":"Article 103176"},"PeriodicalIF":5.8,"publicationDate":"2025-04-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143911987","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}
Daniel Langenkämper , Aksel Alstad Mogstad , Ingunn Nilssen , Tim W. Nattkemper
{"title":"ECO(VI)2SE: Expert-computer vision integration for visual coral status exploration","authors":"Daniel Langenkämper , Aksel Alstad Mogstad , Ingunn Nilssen , Tim W. Nattkemper","doi":"10.1016/j.ecoinf.2025.103162","DOIUrl":"10.1016/j.ecoinf.2025.103162","url":null,"abstract":"<div><div>Cold-water coral reefs and associated habitats are of high ecological relevance and are subject to a diverse spectrum of anthropogenic stressors. Consequently, being able to evaluate the biodiversity and health status of cold-water coral reefs is of high importance. A web application for large-scale assessment of multiple cold-water coral reefs would improve our understanding of the effects of these stressors, and provide an important knowledge base for future planning of human activities in relation to these reefs. In this work, we present a new computational approach to the bottleneck problem of analyzing 77 h of ROV video from cold-water coral reef health status assessments. By combining domain expert knowledge, state-of-the-art deep learning image segmentation and information visualization, we have developed an efficient and sustainable workflow for analyzing visual cold-water coral monitoring data on a continuous basis. The deep learning segmentation network detected and segmented <em>Desmophyllum pertusum</em>, <em>Paragorgia arborea</em>, other gorgonians and sponges from the background in the test set with an intersection over union values of (81.77%, 85.64%, 63.64%, 40.5%, 96.13%) despite fluctuations in water quality and marine snow. Comparisons with manual ROV video evaluations from field personnel showed that the results from the computational approach correlated with the expert-based assessment.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"89 ","pages":"Article 103162"},"PeriodicalIF":5.8,"publicationDate":"2025-04-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143932087","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}
Ho Yi Wan , Michael A. Lommler , Samuel A. Cushman , Jamie S. Sanderlin , Joseph L. Ganey , Andrew J. Sánchez Meador , Paul Beier
{"title":"A multi-level, multi-scale comparison of LiDAR- and LANDSAT-based habitat selection models of Mexican spotted owls in a post-fire landscape","authors":"Ho Yi Wan , Michael A. Lommler , Samuel A. Cushman , Jamie S. Sanderlin , Joseph L. Ganey , Andrew J. Sánchez Meador , Paul Beier","doi":"10.1016/j.ecoinf.2025.103168","DOIUrl":"10.1016/j.ecoinf.2025.103168","url":null,"abstract":"<div><div>The increasing frequency and severity of wildfires pose significant challenges for habitat conservation, particularly in post-fire landscapes. This study evaluates the habitat selection of the Mexican spotted owl (<em>Strix occidentalis lucida</em>) in a post-fire environment using multi-level and multi-scale models derived from LANDSAT and LiDAR data. By focusing on 2nd order (home range selection) and 3rd order (microhabitat selection) habitat use, we assessed the predictive performance and ecological relevance of these datasets. Optimizing predictors across spatial scales revealed that large trees, high canopy cover, and mixed-conifer forests were consistently critical for habitat selection, regardless of the data source. When optimized for spatial scale, LANDSAT- and LiDAR-based models exhibited comparable predictive accuracy (AUC = 0.976 and 0.975, respectively), emphasizing the critical role of scale in model performance. Both models had low out-of-bag (OOB) error rates (0.037 for LANDSAT and 0.050 for LiDAR), indicating high classification reliability. High-severity fire burned 36.6 % of the study area, negatively impacting owl habitat at fine scales around nest and roost sites, whereas a mosaic of burned and unburned patches provided foraging opportunities. Spatial disagreement analysis revealed notable differences in predicted habitat suitability between LANDSAT and LiDAR models, particularly in areas with complex topography and forest composition. These findings underscore the complementary strengths of both datasets, with LiDAR excelling in fine-scale structural detail and LANDSAT providing broad-scale compositional insights. Integrating these technologies offers a scalable and cost-effective framework for monitoring habitat recovery and guiding conservation strategies in fire-affected landscapes.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"89 ","pages":"Article 103168"},"PeriodicalIF":5.8,"publicationDate":"2025-04-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143905938","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}
Ouassine Younes , Conruyt Noël , Kayal Mohsen , A. Martin Philippe , Bigot Lionel , Vignes Lebbe Regine , Moussanif Hajar , Zahir Jihad
{"title":"Deep learning for automated coral reef monitoring a novel system based on YOLOv8 detection and DeepSORT tracking","authors":"Ouassine Younes , Conruyt Noël , Kayal Mohsen , A. Martin Philippe , Bigot Lionel , Vignes Lebbe Regine , Moussanif Hajar , Zahir Jihad","doi":"10.1016/j.ecoinf.2025.103170","DOIUrl":"10.1016/j.ecoinf.2025.103170","url":null,"abstract":"<div><div>Coral reefs are vital for biodiversity, coastal protection, food security, and tourism, yet they face severe threats from anthropogenic activities and climate change, which are leading to their decline. Effective coral reef monitoring is essential for ecological understanding and conservation, but traditional methods are resource-intensive and rely on experts. To address these challenges, we present an automated, deep learning-based monitoring system that integrates YOLOv8, a state-of-the-art object detection algorithm, with DeepSORT, a robust multi-object tracking method, to identify and track coral formations in underwater video footage. Our system was fine-tuned using two curated and annotated datasets: AIMECORAL1 (580 images from the Southwest Indian Ocean) and AIMECORAL2 (282 images from New Caledonia, Pacific Ocean), encompassing diverse coral species and environmental conditions. The system's performance was evaluated using established metrics: object detection precision, Multiple Object Tracking Accuracy (MOTA), Multiple Object Tracking Precision (MOTP), and Identity F1 Score (IDF1). Precision improved from 59.9 % (after fine-tuning on AIMECORAL1) to 84.7 % on the combined datasets. The tracking system achieved a MOTA of 82.63 %, MOTP of 83.28 %, and IDF1 of 70.76 %, demonstrating reliable multi-object tracking in complex underwater environments. We applied our framework to a case study involving video transects from an outer reef site in New Caledonia, comparing data from 2021 and 2022. This automated solution offers a scalable, cost-effective alternative to traditional monitoring methods, supporting seamless, large-scale reef assessment. By leveraging deep learning, our approach enables more efficient data collection, contributing to the protection of these vulnerable ecosystems in the face of increasing environmental pressures.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"89 ","pages":"Article 103170"},"PeriodicalIF":5.8,"publicationDate":"2025-04-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143887135","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}
Franco Ka Wah Leung, Lin Schwarzkopf, Slade Allen-Ankins
{"title":"Advancing invasive species monitoring: A free tool for detecting invasive cane toads using continental-scale data","authors":"Franco Ka Wah Leung, Lin Schwarzkopf, Slade Allen-Ankins","doi":"10.1016/j.ecoinf.2025.103172","DOIUrl":"10.1016/j.ecoinf.2025.103172","url":null,"abstract":"<div><div>Invasive species pose a significant threat to global biodiversity and ecosystem health, necessitating effective monitoring tools for early detection and management. Here, we present the development and assessment of a user-friendly and transferable monitoring tool for the invasive cane toad (<em>Rhinella marina</em>) using passive acoustic monitoring (PAM) and machine learning algorithms. Leveraging a continental-scale PAM dataset (Australian Acoustic Observatory), we trained a cane toad classifier using the BirdNET algorithm, a convolutional neural network architecture capable of identifying acoustic events. We validated thousands of BirdNET predictions across Australia, and our classifier achieved over 90 % accuracy even at many sites outside the areas from which the training data were obtained. Additionally, because cane toads typically call for long periods, we significantly enhanced detection accuracy by incorporating contextual information from time-series data, essentially checking if other calls occurred around each detection (an optimized threshold approach using conditional inference trees). This method substantially reduced false positives and improved overall performance in cane toad detection at sites across Australia. Overall, our method will allow others to develop accurate and precise automated acoustic monitoring tools tailored to their situation, with minimal training data, addressing the critical need for accessible solutions in biodiversity monitoring, control of invasive species and conservation.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"89 ","pages":"Article 103172"},"PeriodicalIF":5.8,"publicationDate":"2025-04-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143895665","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":"Estimating individual tree DBH and biomass of durable Eucalyptus using UAV LiDAR","authors":"Ning Ye, Euan Mason, Cong Xu, Justin Morgenroth","doi":"10.1016/j.ecoinf.2025.103169","DOIUrl":"10.1016/j.ecoinf.2025.103169","url":null,"abstract":"<div><div>Fast-growing eucalyptus species, used as vineyard posts in New Zealand's Marlborough region, offer both durability and potential carbon sequestration benefits. However, the scale of carbon sequestration by these species remains unexplored. This study aimed to estimate individual tree dimensions (diameter at breast height, DBH) and above-ground biomass (AGB) for <em>Eucalyptus globoidea</em> and <em>E. bosistoana</em> using light detection and ranging (LiDAR) data acquired by an unpiloted aerial vehicle (UAV). LiDAR data were captured before destructive sampling, and 96 individual tree LiDAR metrics were extracted. Three machine learning (ML) models, including Partial Least Squares Regression (PLSR), Random Forest, and Extreme Gradient Boosting (XGBoost), were trained. Model performance was evaluated using the root mean square error and coefficient of determination (R<sup>2</sup>). SHapley Additive exPlanations (SHAP) analysis was employed to explain model predictions and evaluate input variables. Results showed that among the ML models, XGBoost and PLSR demonstrated superior performance, with the former yielding the highest R<sup>2</sup> values for AGB (0.903) and the latter getting the highest R<sup>2</sup> values for DBH (0.829). SHAP analysis highlighted that LiDAR height and voxel metrics were the most important factors influencing AGB and DBH predictions. These findings demonstrate that UAV LiDAR can provide efficient and accurate AGB estimates in eucalyptus plantations, supporting the wine industry's carbon neutrality efforts.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"89 ","pages":"Article 103169"},"PeriodicalIF":5.8,"publicationDate":"2025-04-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143899197","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}
Bonyad Ahmadi , Mehdi Gholamalifard , Seyed Mahmoud Ghasempouri , Tiit Kutser
{"title":"Comparative assessment of machine learning algorithms for retrieving colored dissolved organic matter (CDOM) from Sentinel-2/MSI images in the coastal waters of the Persian Gulf","authors":"Bonyad Ahmadi , Mehdi Gholamalifard , Seyed Mahmoud Ghasempouri , Tiit Kutser","doi":"10.1016/j.ecoinf.2025.103171","DOIUrl":"10.1016/j.ecoinf.2025.103171","url":null,"abstract":"<div><div>Colored Dissolved Organic Matter, a pivotal component of aquatic biogeochemical cycles, plays a critical role in regulating water quality and ecosystem functionality. This study provides the first comprehensive assessment of CDOM dynamics in the Persian Gulf's industrialized coastal waters, focusing on the Pars Special Economic Energy Zone (PSEEZ)—a global energy epicenter and the world's largest natural gas reserve. Seasonal field campaigns conducted in 2023 acquired 199 in situ samples stratified across four seasons (Spring: <em>n</em> = 62, Summer: <em>n</em> = 18, Fall: <em>n</em> = 55, Winter: <em>n</em> = 64) using a CTD-integrated Cyclops-7 fluorometer. Sampling intervals were methodologically synchronized with satellite overpasses (±3 h) to minimize temporal discrepancies between ground-truth measurements and remotely sensed data, thereby ensuring spatiotemporal coherence essential for robust algorithm calibration and validation. Contrary to expectations, CDOM concentrations in petrochemical-influenced areas (e.g., stations P7: 0.29 ppb, P13: 0.35 ppb) were markedly lower than in natural mangrove ecosystems (stations N13: 19.61 ppb, NA2: 12.91 ppb), underscoring the antagonistic effects of industrial pollutants on organic matter stability. Initial CDOM retrieval algorithms yielded suboptimal accuracy (MAE = 1.16, RMSLE = 1.2). A regionally tuned band ratio algorithm improved performance by 27 % (MAE = 0.85) and 22 % (RMSLE = 0.94). Machine learning models further enhanced retrievals, with the Mixture Density Network (MDN) emerging as the superior framework. The MDN achieved an RMSLE of 0.47 (17.5 % improvement over MLP, 14.5 % over SVM) and reduced systematic bias (SSPB) by 26.12 units compared to Bayesian Ridge Regression (BRR), outperforming conventional models like SVM (MAE = 0.61, RMSLE = 0.55). While the MDN exhibited marginally higher absolute error (MAE = 0.53) than deterministic models, its probabilistic architecture uniquely addressed the Persian Gulf's optical complexity, characterized by overlapping signals from SGD-driven organics, hydrocarbon plumes, and sediment resuspension. This study establishes MDN as a transformative tool for CDOM retrieval in optically heterogeneous, anthropogenically stressed waters, while advocating for regionally adaptive frameworks to advance precision water quality monitoring in critical marine ecosystems.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"89 ","pages":"Article 103171"},"PeriodicalIF":5.8,"publicationDate":"2025-04-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143899195","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}