Ecological Informatics最新文献

{"title":"Mapping hotspots and unveiling drivers of mortality in the endangered Gangetic Dolphin (Platanista gangetica) to mitigate human-mediated conservation conflicts in the Ganga River Basin, India","authors":"Ruchi Badola,&nbsp;Goura Chandra Das,&nbsp;Surya Prasad Sharma,&nbsp;Aftab Alam Usmani,&nbsp;Shivani Barthwal,&nbsp;Srishti Badola,&nbsp;Syed Ainul Hussain","doi":"10.1016/j.ecoinf.2025.103239","DOIUrl":"10.1016/j.ecoinf.2025.103239","url":null,"abstract":"<div><div>Globally, effective conservation of threatened species hinges on robust population monitoring and the identification and minimization of threats that influence population dynamics. In the recent years, human-induced factors have surpassed natural causes as the primary drivers of species decline. Hence, information on human-induced threats of mortality and identifying spatial hotspots are critical for implementing timely and targeted conservation interventions. The Gangetic dolphin (<em>Platanista gangetica</em>) is widely recognized as a flagship species of river ecosystems in the Ganga River Basin, which has suffered population decline due to human-induced threats, including negative human-dolphin interactions. In the present study, we used a combination of a maximum entropy model (MaxEnt) and GIS-based weighted overlay analysis (WOA) to identify Gangetic dolphin mortality hotspots in the Ganga Basin. We compiled 110 records of Gangetic dolphin mortalities (<em>n</em> = 76) and rescues (<em>n</em> = 34) from diverse sources, including media reports, scientific articles, community volunteer network called the <em>Ganga Praharis</em> and basin-wide ecological surveys. The highest number of mortalities was recorded along the Hooghly River (38.2 %), followed by the Ganga mainstem (34.2 %), Sharda canal (9.2 %) and Girwa-Kauriyala rivers (7.5 %). The primary cause of mortality was accidental, such as entanglement in fishing nets, boat collision, dredging of river channels (32.9 %), followed by stranding in canal/barrages/shallow water (14.5 %), consumptive poaching (10.5 %), retaliatory killing (5.3 %), and natural causes (3.9 %). Using a combination of MaxEnt and WOA, we identified 770 km (15.1 %) of the river stretches as mortality hotspots associated with human-induced factors in the Basin. These identified stretches require urgent conservation interventions, such as the establishment of rescue and rehabilitation facilities, improved veterinary response system, and engagement with riverside communities to reduce the human-induced threats to Gangetic dolphin.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"90 ","pages":"Article 103239"},"PeriodicalIF":5.8,"publicationDate":"2025-05-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144213043","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":"Wave-turbulence decomposition and turbulence parameterization in aquatic wave environments using improved Synchrosqueezed Wavelet Transform (iSWT)","authors":"Xingxing Zhao ,&nbsp;Daobin Ji ,&nbsp;Lianghong Long ,&nbsp;Zhongyong Yang ,&nbsp;Zhengjian Yang ,&nbsp;Defu Liu ,&nbsp;Andreas Lorke","doi":"10.1016/j.ecoinf.2025.103241","DOIUrl":"10.1016/j.ecoinf.2025.103241","url":null,"abstract":"<div><div>Accurate quantification of near-surface turbulence is essential for understanding the dynamics of turbulent mixing and mass transport in aquatic systems. However, field measurements of near-surface flow velocities often include contributions from surface gravity waves. For the quantification of turbulence and related transport processes, robust methods are needed to separate wave motion from the turbulent velocity fluctuations. In this study, we evaluated the performance of five different methods for wave-turbulence decomposition in estimating turbulent kinetic energy, Reynolds stress and turbulent kinetic energy dissipation rate. The methods include Ensemble Empirical Modal Decomposition (EEMD), Phase method (PH), Variational Mode Decomposition (VMD), Synchrosqueezed Wavelet Transform (SWT) and improved Synchrosqueezed Wavelet Transform (iSWT). We used these methods for a re-analysis of high-frequency velocity measurements from the water surface of the Kitinen River, Finland. The results show that the different methods remove the wave component to varying degrees, whereas the performance of the VMD method appeared insufficient. The estimated turbulent kinetic energy and Reynolds stresses were generally smaller than 30 % of those calculated from the unprocessed velocity measurements. In terms of energy spectra, the EEMD, PH, SWT and iSWT methods all provide a better removal of wave energy, but the EEMD and SWT methods resulted in substantial energy notches in the wave frequency band, resulting in a significant underestimation of the turbulent velocity fluctuations. In contrast, iSWT achieves the decomposition of wave and turbulence components by applying an optimal decomposition degree index <span><math><msub><mi>p</mi><mi>opt</mi></msub></math></span>, which maximizes the retention of turbulent velocity fluctuations. Application of the inertial dissipation method for estimating dissipation rates of turbulent kinetic energy from the spectra of separated turbulent velocities. The results showed that the iSWT method resulted in the longest inertial subrange, and allowed for most but also has very good robustness spectral fits for dissipation rates ranging from 1.33 × 10⁻⁷ W/kg to 1.06 × 10⁻⁵ W/kg. Using dissipation rate estimates from an advanced methods explicitly considering wave-turbulence interactions as a reference, the iSWT method showed the closest agreement, whereas the dissipation rates estimated from velocities processed by the other four methods were generally lower. The newly proposed method is able to provide accurate estimates of dissipation rates by robustly separating the turbulence from wave-affected velocities compared to the four tested existing methods.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"90 ","pages":"Article 103241"},"PeriodicalIF":5.8,"publicationDate":"2025-05-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144196410","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":"Deep learning-based analysis of insect life stages using a repurposed dataset","authors":"Fatin Faiaz Ahsan ,&nbsp;Melissa L. Thomas ,&nbsp;Hamid Laga ,&nbsp;Ferdous Sohel","doi":"10.1016/j.ecoinf.2025.103202","DOIUrl":"10.1016/j.ecoinf.2025.103202","url":null,"abstract":"<div><div>Insect pests pose a significant risk to agriculture and biosecurity, reducing crop yields and requiring effective management. Accurate identification of early life stages is often required for effective management but is generally reliant on expert evaluation, which is both costly and time-consuming. To address this, we use a deep learning-based approach for insect species and life-stage classification from digital images. We repurposed the IP102 dataset by adding detailed annotations for four life stages — egg, larva, pupa, and adult — alongside the original species categories. Two deep learning models, based on ResNet50 and EfficientNetV2M, were tested for classification accuracy in this dual-layered identification task. Although both models accomplished the task well, the EfficientNetV2M model performed slightly better than the ResNet50, achieving 72.4% precision, 72.1% recall, and an F1-score of 72.0%. Our results demonstrate the potential of deep learning for automated insect species and life-stage classification, providing a high throughput and efficient solution towards agricultural monitoring and pest management.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"90 ","pages":"Article 103202"},"PeriodicalIF":5.8,"publicationDate":"2025-05-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144213030","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":"Functional diversity of benthic macroinvertebrates and fluvial habitat quality: Key biological trait categories","authors":"Gonzalo Sotomayor ,&nbsp;Henrietta Hampel ,&nbsp;Raúl F. Vázquez ,&nbsp;Marie Anne Eurie Forio ,&nbsp;Peter L.M. Goethals","doi":"10.1016/j.ecoinf.2025.103235","DOIUrl":"10.1016/j.ecoinf.2025.103235","url":null,"abstract":"<div><div>Functional diversity (FD) calculations using benthic macroinvertebrates are useful for freshwater ecosystem evaluation. However, it is critical to determine the key traits and their categories that shape a community. This study (i) investigated the effect of fluvial habitat quality (characterised by a fluvial habitat index – FHI) on the trends of individual functional macroinvertebrate categories (FMaCs) and the rRao FD index; and (ii) evaluated the information provided by each FMaC for rRao index calculation along the FHI gradient. Macroinvertebrate samples were collected at 12 locations in Ecuador's Paute River Basin over six years. Families of macroinvertebrates were classified into eight traits and 42 FMaCs. A K-means cluster analysis produced three groups of sampling points based on their FHI values. For each FHI cluster, the percentage of each FMaC within its corresponding trait was calculated. The R² coefficient was computed between the FHI cluster values and the previously obtained FMaC percentages. A second K-means clustering was performed on the R² dataset, resulting in three groups of R² values directly associated with FMaCs. We then assessed the sensitivity of the rRao index to the exclusion of specific trait categories by sequentially removing groups of FMaCs, ordered by decreasing R² importance. This allowed us to evaluate the stability and robustness of functional diversity estimates when less informative traits were removed. Results indicated that certain FMaCs had a greater influence on rRao variation across habitat quality clusters, particularly those related to body form, locomotion, and exoskeleton hardness. In degraded habitats, certain FMaCs contributed little to rRao variation, suggesting limited functional differentiation within the multi-trait functional space and potentially lower monitoring value under such conditions. The most informative traits for rRao index calculation were body form, flexibility, and locomotion. These findings contribute to improved trait-based ecological modelling of macroinvertebrates and offer insights for river managers regarding potential ecohydrological stressors.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"90 ","pages":"Article 103235"},"PeriodicalIF":5.8,"publicationDate":"2025-05-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144196409","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":"MRSEILA: A modified remote sensing ecological index using local adaptability for enhancing ecological environment quality assessment","authors":"Yingzhang Guo ,&nbsp;Mingjin Zhan ,&nbsp;Hanzeyu Xu ,&nbsp;Xiao Li ,&nbsp;Junjun Fang ,&nbsp;Xingchen Zhou ,&nbsp;Dan Lin ,&nbsp;Wenhui Chen","doi":"10.1016/j.ecoinf.2025.103238","DOIUrl":"10.1016/j.ecoinf.2025.103238","url":null,"abstract":"<div><div>The Remote Sensing Ecological Index (RSEI) has been widely applied in ecological environment quality (EEQ) assessments by integrating multiple environmental factors. To enhance RSEI's ability to capture local ecological variations, a locally adapted version (RSEI_LA) was designed and widely adopted using moving windows. However, the randomness in eigenvector directions generated by principal component analysis (PCA) can introduce bias, affecting the accuracy of RSEI_LA's assessment. To enhance the effectiveness of RSEI_LA in EEQ, we propose a modified RSEI_LA model (MRSEI_LA) implemented on the Google Earth Engine (GEE) platform, consisting of three components: (1) optimization of moving window sizes tailored to each target region; (2) automatic recognition and correction of PCA-induced eigenvector direction inconsistencies; and (3) refinement of PCA computation within each circular window to improve the accuracy of EEQ evaluations. We validated MRSEI_LA using Landsat Collection 2 Level-2 surface reflectance data and compared its performance with RSEI_LA across four typical areas in China. The results showed that, compared to RSEI_LA, MRSEI_LA consistently produces aligned eigenvector directions and more accurate EEQ assessments that better reflect actual land surface conditions across all four testing areas, making it an effective tool for regional and large-scale ecological monitoring.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"90 ","pages":"Article 103238"},"PeriodicalIF":5.8,"publicationDate":"2025-05-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144241684","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":"SeagrassFinder: Deep learning for eelgrass detection and coverage estimation in the wild","authors":"Jannik Elsäßer ,&nbsp;Laura Weihl ,&nbsp;Veronika Cheplygina ,&nbsp;Lisbeth Tangaa Nielsen","doi":"10.1016/j.ecoinf.2025.103200","DOIUrl":"10.1016/j.ecoinf.2025.103200","url":null,"abstract":"<div><div>Seagrass meadows play a crucial role in marine ecosystems, providing benefits such as carbon sequestration, water quality improvement, and habitat provision. Monitoring the distribution and abundance of seagrass is essential for environmental impact assessments and conservation efforts. However, the current manual methods of analyzing underwater video data to assess seagrass coverage are time-consuming and subjective. This work explores the use of deep learning models to automate the process of seagrass detection and coverage estimation from underwater video data. We create a new dataset of over 8,300 annotated underwater images, and subsequently evaluate several deep learning architectures, including ResNet, InceptionNetV3, DenseNet, and Vision Transformer for the task of binary classification on the presence and absence of seagrass by transfer learning. The results demonstrate that deep learning models, particularly Vision Transformers, can achieve high performance in predicting eelgrass presence, with AUROC scores exceeding 0.95 on the final test dataset. The application of underwater image enhancement further improved the models’ prediction capabilities. Furthermore, we introduce a novel approach for estimating seagrass coverage from video data, showing promising preliminary results that align with expert manual labels, and indicating potential for consistent and scalable monitoring. The proposed methodology allows for the efficient processing of large volumes of video data, enabling the acquisition of much more detailed information on seagrass distributions in comparison to current manual methods. This information is crucial for environmental impact assessments and monitoring programs, as seagrasses are important indicators of coastal ecosystem health. This project demonstrates the value that deep learning can bring to the field of marine ecology and environmental monitoring.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"90 ","pages":"Article 103200"},"PeriodicalIF":5.8,"publicationDate":"2025-05-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144221558","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":"Cost-effective approaches for microplastic pellets characterization using a machine learning tool","authors":"V.M. Scarrica ,&nbsp;P. Cocozza ,&nbsp;G. Anfuso ,&nbsp;A. Staiano ,&nbsp;G. Bonifazi ,&nbsp;A. Rizzo ,&nbsp;S. Serranti","doi":"10.1016/j.ecoinf.2025.103230","DOIUrl":"10.1016/j.ecoinf.2025.103230","url":null,"abstract":"<div><div>Microplastics, including pellets, are a persistent pollutant on beaches that pose relevant ecological and environmental challenges. Their widespread presence in marine and coastal environments endangers ecosystems, threatens marine life, and risks entering the food chain. Effective microplastic management requires reliable methods for their identification and classification, yet the high cost of required equipment hinders large-scale implementation. Artificial intelligence offers a promising solution for polymer analysis. While machine learning techniques have demonstrated potential in automating microplastic classification, existing approaches often rely on complex models requiring numerous input variables, limiting their practical application. This paper introduces a simplified methodology for pellet polymer classification using a Random Forest model requiring a limited set of variables for training. The approach reduces model complexity while maintaining high classification performance, emphasizing simplicity, speed and efficiency. The method was tested on different pellet samples collected from the coasts of Spain, Portugal and Vulcano Island (Italy). The results highlight the robustness of the proposed model and its suitability to be applied in diverse environmental contexts. By balancing accuracy with computational efficiency, the proposed approach represents a practical tool for pellet classification. This streamlined methodology can offer a significant step forward in microplastic management and pollution mitigation, contributing to the development of cost-effective, scalable solutions for addressing the environmental impacts of microplastics.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"90 ","pages":"Article 103230"},"PeriodicalIF":5.8,"publicationDate":"2025-05-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144196413","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":"Architecture and implementation of ulrb algorithm in R","authors":"Francisco Pascoal ,&nbsp;Rodrigo Costa ,&nbsp;Luís Torgo ,&nbsp;Catarina Magalhães ,&nbsp;Paula Branco","doi":"10.1016/j.ecoinf.2025.103229","DOIUrl":"10.1016/j.ecoinf.2025.103229","url":null,"abstract":"<div><div>Low-abundance microorganisms, often referred to as the “rare biosphere”, play a crucial role in ecosystem resistance and resilience, but remain challenging to study. One of the main difficulties lies in the lack of an appropriate definition of rare taxa. Most studies use relative abundance thresholds (<em>e.g.</em>, 0.1 % relative abundance, per sample) to discern rare from abundant taxa within a microbial community. This is inappropriate because such thresholds are arbitrary and lack biological meaning. To solve this problem, we have proposed the utilization of unsupervised machine learning, through the <em>ulrb</em> (“Unsupervised Learning Definition of the Microbial Rare Biosphere”) algorithm, implemented as an R package (v0.1.8). This algorithm applies the partition around medoids (pam) algorithm to cluster taxa based on their abundance, in a community, for any number of samples. Based on the clusters, <em>ulrb</em> automatically classifies taxa into “rare”, “undetermined” or “abundant”, by default. <em>Ulrb</em> includes functions for all analytical steps necessary to define the rare biosphere. Specifically, we include four groups of functions: 1) process data of the user into the correct format for the <em>ulrb</em> algorithm; 2) cluster taxa into abundance classifications; 3) helper functions to evaluate detailed statistics of the clustering steps; and 4) visualization functions, focused on rank abundance curves and Silhouette scores, for assessment of clustering quality. In addition, <em>ulrb</em> allows the user to change the number of classifications obtained and includes options for detailed reporting. In this article, we describe the <em>ulrb</em> R package architecture, coding organization, and strategy. Furthermore, we use a 16S rRNA gene amplicon sequencing dataset from the Arctic Ocean to provide illustrative examples, with code, on how to use and explore <em>ulrb</em> capabilities. By explaining the architecture and implementation of <em>ulrb</em>, this study allows independent groups to integrate an abundance classification step in their data analysis protocols, instead of relying on taxa labeled by inconsistent or manual strategies.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"90 ","pages":"Article 103229"},"PeriodicalIF":5.8,"publicationDate":"2025-05-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144231658","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":"Flexible hybrid edge computing IoT architecture for low-cost bird songs detection system","authors":"Francisco A. Delgado-Rajó,&nbsp;Carlos M. Travieso-Gonzalez","doi":"10.1016/j.ecoinf.2025.103231","DOIUrl":"10.1016/j.ecoinf.2025.103231","url":null,"abstract":"<div><div>The monitoring of bird populations provides valuable insights into biodiversity variations and their correlation with environmental changes. This study proposes a flexible hybrid edge computing IoT architecture for a low-cost bird song detection system. The system integrates low-power microcomputers, such as Raspberry Pi, equipped with USB microphones, LoRa modules, and Wi-Fi for seamless operation across rural and urban environments. By utilizing deep learning techniques, including convolutional neural networks (CNNs) trained on bird song datasets, the system performs real-time species detection at the edge, minimizing the need for high-bandwidth transmission. Nodes dynamically select communication technologies based on availability, sending data to an IoT analytics platform. Field deployments demonstrate the system's efficiency, interoperability, and adaptability for biodiversity monitoring, particularly in remote areas with limited connectivity. This architecture addresses the challenges of real-time species detection while ensuring low cost, scalability, and energy efficiency. The main advantage is that devices can operate in areas without mobile coverage, as they only transmit the detection signal. This results in significant bandwidth savings, since the processing is carried out at the edge.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"90 ","pages":"Article 103231"},"PeriodicalIF":5.8,"publicationDate":"2025-05-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144178243","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":"Evaluating textural descriptors for automated image classification of stony reefs in turbid temperate waters","authors":"Saulė Medelytė ,&nbsp;Yuri Rzhanov ,&nbsp;Andrius Šiaulys ,&nbsp;Kim Lowell","doi":"10.1016/j.ecoinf.2025.103236","DOIUrl":"10.1016/j.ecoinf.2025.103236","url":null,"abstract":"<div><div>The rise of machine learning (ML) techniques has made automatic image classification increasingly relevant and essential for marine biologists. Despite advancements in computational power and growing interest in the field, underwater image analysis remains a significant challenge, especially in highly turbid environments. This study is the first to assess the potential of texture descriptors for classifying benthic species and habitats using turbid underwater imagery. Underwater images were collected in SE Baltic Sea reefs (4.4–42.2 m depth) using a drop-down camera. A total of sixteen textural descriptors were tested, of which three were selected for the CatBoost ML model image classification task. The model's performance was evaluated using annotated images provided by field experts. Among these, the MRELBP (Median Robust Extended Local Binary Pattern) algorithm achieved the highest overall performance. For individual classes, the best image classification results were achieved for large blue mussels by the LMP (Local Morphological Pattern) algorithm (F1 score: 0.72 ± 0.18) and small blue mussels (F1 score: 0.66 ± 0.13) by MRELBP. For lithological classes, sand was classified with the highest accuracy by MRELBP (F1 score: 0.69 ± 0.23). Model coverage estimates were acceptable in 49 % of the images, with blue mussels being the most suitable for evaluation. The results demonstrate textural descriptors capabilities in classifying real-world underwater images.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"90 ","pages":"Article 103236"},"PeriodicalIF":5.8,"publicationDate":"2025-05-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144196407","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}