Novel Monitoring Technologies - an integrated overview of traditional and emerging methods for monitoring marine biodiversity. Deliverable 3.1_BioEcoOcean

Funded by the European Union under Grant Agreement number 101136748. Views and opinions expressed are, however, those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Agency (REA). Neither the European Union nor the granting authority can be held responsible for them. Deliverable 3.1 Novel Monitoring Technologies - an integrated overview of traditional and emerging methods for monitoring marine biodiversity July 2025 Authors: Lisandro Benedetti-Cecchi, Luca Rindi, Caterina Mintrone, Alexander McGrath, Dolapo Salim Olatoye, Florian Lüskow, Patrick Lehodey, Débora Borges, Bianca Reis, André Valente, Andre Visser, Lina Mtwana Nordlund Ref. Ares(2025)10648931 - 03/12/2025 Deliverable 3.1 Novel Monitoring Technologies 2 Project Information Project full title: Co-Creating Transformative Pathways to Biological and Ecosystem Ocean Observations Project acronym: BioEcoOcean Grant agreement number: 101136748 Project start date and duration: 1st of February 2024, 48 months Project website: https://bioecoocean.org/ Project Coordinator: Lina Mtwana Nordlund Deliverable information Deliverable number: 3.1 Deliverable title: Novel Monitoring Technologies Submission date: 31.07.2025 Level of dissemination: Public Work package and task: WP3 and Task 3.1 Lead Beneficiary: UNIPI Authors: Lisandro Benedetti-Cecchi (UNIPI), Luca Rindi (UNIPI), Caterina Mintrone (UNIPI), Alexander McGrath (UNIPI), Dolapo Salim Olatoye (UNIPI), Florian Lüskow (UU), Patrick Lehodey (MOi), Débora Borges (CIIMAR), Bianca Reis (CIIMAR), André Valente (Air Centre), Andre Visser (DTU), Lina Mtwana Nordlund (UU) Reviewers: Nina Lepola (UU), Ana Lara-Lopez (UNESCO) To be cited as: Benedetti-Cecchi, L., Rindi, L., Mintrone, C., McGrath, A., Olatoye, D. S., Lüskow, F., Lehodey, P., Borges, D., Reis, B., Valente, A., Nordlund, L. M. 2025. Novel Monitoring Technologies - an integrated overview of traditional and emerging methods for monitoring marine biodiversity. BioEcoOcean Deliverable 3.1. http://doi.org/10.5281/zenodo.16630482 Deliverable 3.1 Novel Monitoring Technologies 3 Executive Summary This report provides an integrated overview of traditional and emerging methods for monitoring marine biodiversity, with emphasis on Essential Ocean Variables (EOVs). Traditional visual sampling methods such as underwater visual census (UVC), quadrat surveys, and net tows have long formed the backbone of biodiversity monitoring, offering high taxonomic resolution and compatibility with historical data. Emerging technologies – including environmental DNA (eDNA), imaging combined with Artificial Intelligence (AI) analysis, and remote sensing – promise to boost biodiversity observations by increasing spatial and temporal coverage, automating labour-intensive processes, and reducing environmental impact. Emerging technologies – including environmental DNA (eDNA), imaging combined with Artificial Intelligence (AI) analysis, and remote sensing (satellites and drones) – promise to boost biodiversity observations by increasing spatial and temporal coverage, automating labourintensive processes, and reducing environmental impact. eDNA enables non-invasive detection of cryptic species, while AI image analysis automates species identification and habitat quantification. Remote platforms such as Unmanned Aerial Vehicles (UAVs) and Remotely Operated Vehicles (ROVs) extend survey capabilities into deep or otherwise inaccessible environments. Key findings highlight the strengths and limitations of each method. eDNA excels in sensitivity but faces challenges in assigning species to sampling sites and in abundance estimation. AI imaging supports scalable monitoring, but requires large, annotated training sets. Acoustic methods offer large-scale tracking of biomass via proxies, but they still lack taxonomic precision. Hybrid approaches are emerging as best practice: combining traditional methods with novel tools is necessary for cross-validation and to maximize spatial and temporal coverage. Calibration remains essential, as it requires reference libraries and standardized protocols. The combination of traditional and emerging biodiversity observing technologies aligns with ethical research principles, such as the 3Rs (Replacement, Reduction, Refinement), especially by reducing the impact associated with destructive sampling. We recommend adopting integrated monitoring frameworks that combine in situ, genetic, and remote sensing data collection, tailored to the goals and constraints of each project. Supporting policy development and training efforts will be key to mainstreaming these approaches across the EU and global marine monitoring networks. Deliverable 3.1 Novel Monitoring Technologies 4 Table of Contents Executive Summary ................................................................................................................................................. 3 Abbreviations .......................................................................................................................................................... 6 1. Introduction .................................................................................................................................................... 7 1.1 EOVs in focus ........................................................................................................................................... 8 1.1.1 Macroalgal canopy cover and composition .................................................................................... 8 1.1.2 Seagrass cover and composition ..................................................................................................... 8 1.1.3 Fish abundance and distribution .................................................................................................... 9 1.1.4 Zooplankton biomass and diversity .............................................................................................. 10 2. Methodological overview ............................................................................................................................. 11 2.1 Traditional Sampling Methods .............................................................................................................. 11 2.1.1 Macroalgal canopy cover and composition .................................................................................. 11 2.1.2. Seagrass cover and composition .................................................................................................. 12 2.1.3 Fish abundance and distribution .................................................................................................. 13 2.1.4 Zooplankton biomass and diversity .............................................................................................. 14 2.2 Strengths & Limitations for Traditional Sampling Methods ................................................................. 16 2.2.1 Benthos – marine life found on the bottom – diver-based underwater visual census ................ 16 2.2.2 Reef fish non-destructive sampling methods ............................................................................... 17 2.2.3 Net sampling of zooplankton ........................................................................................................ 18 2.3 Emerging Methods ................................................................................................................................ 19 2.3.1 Environmental DNA (eDNA) .......................................................................................................... 19 2.3.2 Imaging & AI-based analysis ......................................................................................................... 21 2.3.3 Remotely operable platforms and remote sensing ...................................................................... 25 2.3.4 Mapping intertidal macroalgae and canopy floating kelp ............................................................ 26 2.3.5 Passive acoustics for assessing marine forest health and diversity .............................................. 27 2.3.6 Active acoustics and the deep scattering layer ............................................................................ 28 2.4 Strengths & Limitations for Emerging Methods ................................................................................... 28 2.4.1 eDNA ............................................................................................................................................. 28 2.4.2 Imaging & AI techniques ............................................................................................................... 29 2.4.3 Satellite remote sensing ............................................................................................................... 30 2.4.4 UAV-based mapping ..................................................................................................................... 31 2.4.5 Passive acoustic monitoring .......................................................................................................... 32 2.4.6 Active acoustic .............................................................................................................................. 33 3. Case studies ...................................................................................................................................................... 34 3.1 eDNA sampling versus traditional fish surveys in the Baltic Sea ........................................................... 34 Deliverable 3.1 Novel Monitoring Technologies 5 3.2 eDNA versus traditional macroalgal sampling in the Tuscan Archipelago living lab .............................. 35 3.3 AI-based imaging for species recognition in the Tuscan Archipelago living lab ..................................... 36 3.4 Using video to estimate jellyfish abundance in the Baltic Sea ............................................................... 37 3.5 Assessing Vertical Migrant Communities Using Acoustic Backscatter Data .......................................... 37 4. Comparative evaluation .................................................................................................................................... 38 4.1. Summary of strengths and limitations of different methods ................................................................. 38 5. Advancing underwater monitoring in line with the 3rs .................................................................................... 40 6. Integration and hybrid approaches .................................................................................................................. 40 7. Recommendations ............................................................................................................................................ 41 References ............................................................................................................................................................ 43 Deliverable 3.1 Novel Monitoring Technologies 6 Abbreviations Abbreviation Definition AI Artificial Intelligence Air Centre The Atlantic International Research Centre ASV Amplicon Sequence Variant AUV Autonomous Underwater Vehicle CATAMI Collaborative and Automated Tools for Analysis of Marine Imagery CIIMAR Interdisciplinary Centre of Marine and Environmental Research CNN Convolutional Neural Network CPR Continuous Plankton Recorder DSL Deep Scattering Layer DVM Diel vertical migration eDNA Environmental DNA EOV Essential Ocean Variable GZ Gelatinous Zooplankton mAP Mean Average Precision MEDITS International bottom trawl survey in the Mediterranean ML Machine Learning MOi Mercator Ocean International MSFD Marine Strategy Framework Directive PAM Passive Acoustic Monitoring POC Particulate Organic Carbon qPCR Real-time PCR RLS Reef Life Survey RGB Red-Green-Blue ROV Remotely Operated Vehicle SDM Species Distribution Model SSH Sea Surface Height SST Sea Surface Temperature UAV Unmanned Aerial Vehicle UNIPI University of Pisa UU Uppsala University UVC Underwater Visual Census VHR Very High-Resolution VPR Video Plankton Recorder YOLO You Only Look Once Deliverable 3.1 Novel Monitoring Technologies 7 1. Introduction Marine biodiversity contributes key ecosystem functions such as nutrient cycling and carbon sequestration and provides critical services for human wellbeing, but it is increasingly threatened by the cumulative effects of climate change and direct anthropogenic pressure (IPBES 2019; IPCC 2021). Evaluating the response of marine biodiversity to these threats is critical to support informed conservation policies and to assess progress towards national and international conservation targets, including the UN Sustainable Development Goals, EU Biodiversity Strategy, Green Deal, and the Kunming-Montreal Global Biodiversity Framework. Monitoring ocean biodiversity is hampered by the difficulty of conducting underwater observations and the myriad of biodiversity indicators proposed to assess environmental health and the multidimensional nature of biodiversity (Chase et al. 2018). A better understanding of biodiversity patterns and their dynamics requires increased spatial and temporal coverage of observations, harmonization of sampling methods, and better coordination of sampling efforts. Improving our capacity to automate data collection and processing will dramatically increase the amount and quality of information and knowledge available to scientists and decisionmakers. The development of new tools such as environmental DNA (eDNA), sensor networks, underwater imaging through cameras and drones, and the use of artificial intelligence (AI) to extract information from expanding data sets are promising technological advancements to improve non-invasive, costeffective, and high-frequency monitoring of biodiversity. The Essential Ocean Variable (EOV) framework offers a simplified approach to standardize biodiversity observations, reducing the number of measurements to a limited set of socially relevant and feasible variables (Miloslavich et al. 2018; Muller-Karger et al. 2018). For example, EOVs such as coral, mangrove, macroalgal, and seagrass cover and composition can be measured easily and cheaply, informing on habitat state and capturing phenomena such as regime shifts. Microbe, phytoplankton, and zooplankton biomass and diversity provide key links between marine biodiversity and biogeochemical processes, while invertebrate, fish, sea turtle, seabird, and mammal abundance and distribution allow tracking changes in food webs, top predators, and commercial species. The combination of technological innovation and simplified biodiversity frameworks provides an unprecedented momentum to advance monitoring of marine biodiversity globally. In the BioEcoOcean project, there are several living labs - collaborative platforms - focusing on advancing our understanding of biology and ecosystems in the ocean. Each living lab focuses on a set of EOVs and represents different environments. In this deliverable, we are specifically focusing on the following EOVs; Macroalgal canopy cover and composition, Seagrass cover and composition, Fish abundance and distribution and Zooplankton biomass and diversity. Promising technologies are still in the early stages of development and need to be calibrated and standardized by directly comparing their outputs with those of traditional methods for monitoring marine biodiversity. Here, we 1) provide an overview of traditional sampling methods for marine biodiversity, with emphasis on the EOVs addressed by BioEcoOcean living labs, 2) introduce eDNA, imaging, remote sensing and AI-based analysis, 3) provide a comparative evaluation of emerging Deliverable 3.1 Novel Monitoring Technologies 8 technologies and traditional approaches for monitoring biodiversity, 4) discuss hybrid approaches and 5) conclude with some recommendations and way forward. Case studies and applications from living labs are illustrated throughout. 1.1 EOVs in focus 1.1.1 Macroalgal canopy cover and composition Macroalgal canopies create vital underwater forests on temperate rocky reefs, supporting high biodiversity and providing key ecosystem functions and services such as primary production, nursery habitats, food resources, and coastal protection. These forests face global threats like ocean warming and local pressures such as habitat loss, pollution, eutrophication, and invasive species, which together reduce their resilience and increase the risk of regime shifts and population declines. Such changes often involve replacement by less productive, low-diversity algae or barren areas. Macroalgal forests react quickly to environmental stress, serving as early indicators of change, and their wide range allows monitoring of regional trends and species distributions. Macroalgal canopy cover and composition is a reliable, easily measured indicators of ecosystem health and environmental change. Monitoring macroalgae allows scientists to track ecological shifts and broader trends. Healthy macroalgae are vital for biodiversity and sustainable coastal economies. The sub-variables required to assess macroalgal canopy cover and composition include percent cover, stipe density, canopy species diversity and areal extent of occupancy. These sub-variables are indicated in the EOV specification sheet (Macroalgal canopy cover and composition specification sheet), along with ancillary variables and derived products that are necessary to interpret the EOV. Supporting variables include environmental drivers of change such as nutrients, sea surface temperature, salinity, dissolved oxygen and ocean colour, informing on regional productivity. Other EOV related supporting variables include canopy height, photosynthetic biomass plant size classes, photosynthetic efficiency, algal decomposition rate and signs of necrosis on algal thalli. Species composition and abundance of understory assemblages and extent of alternative habitats (e.g., barrens, algal turfs) provide additional measurements to understand the health of macroalgal canopies and the risk of shifting to undesired alternative states. 1.1.2 Seagrass cover and composition Seagrasses are submerged marine plants that form extensive meadows in shallow coastal waters, serving as foundation species for diverse and productive ecosystems (Unsworth et al. 2022). These habitats provide essential ecosystem services: they stabilize sediments, improve water quality, support fisheries, and offer critical nursery and feeding grounds for marine species (Nordlund et al. 2016). Seagrass meadows are also globally significant carbon sinks, capturing and storing “blue carbon” in Deliverable 3.1 Novel Monitoring Technologies 9 both biomass and sediments, which helps mitigate climate change (UNEP, 2020). Despite their importance, seagrasses are declining due to coastal development, nutrient pollution, and climateinduced stressors. Monitoring seagrass cover and species composition is thus essential to detect change, assess ecosystem health, and inform conservation and climate strategies. GOOS defines seagrass cover and composition as an Essential Ocean Variable (EOV), based on three core sub-variables: percent cover, areal extent, and species composition (see Seagrass Cover and Composition EOV Specification Sheet). These are complemented by supporting variables such as water depth, clarity, temperature, salinity, sediment characteristics, and indicators of condition like biomass, productivity, and epiphytic load. Monitoring methods include in situ diver-based surveys, drop cameras, and quadrat sampling for fine-scale data, as well as remote sensing technologies—including drones, satellites, and acoustic tools—for large-scale mapping. New techniques like eDNA are also being explored for species detection. Derived products include estimates of global and regional distribution, diversity metrics, ecosystem resilience, and carbon sequestration potential. These products support applications across marine spatial planning, biodiversity conservation, climate adaptation, and ecosystem service valuation. Standardised methods and integration into platforms like Ocean Biodiversity Information System (OBIS) and the GOOS BioEco Portal ensure that seagrass data are Findable, Accessible, Interoperable, and Reusable (FAIR). 1.1.3 Fish abundance and distribution Fish play a crucial role in maintaining the balance of marine ecosystems, occupying a wide spectrum of trophic levels. They represent a primary source of food and proteins for humans and support the economic well-being of millions of people, generating a global economic value of approximately $150 billion annually (FAO, 2022). However, climate change, overfishing, and habitat destruction are increasingly threatening these resources. Accurate and standardized monitoring of the status and dynamics of fish populations is therefore necessary to guide the sustainable management of these resources and ensure the health of marine ecosystems, food security, and economic stability globally. Fish abundance, length distribution, biomass, and species composition represent the four subvariables proposed by GOOS to estimate this EOV (see Fish abundance and distribution EOV specification sheet). Furthermore, supporting-variables are important to describe the environmental context (e.g., seawater temperature, salinity, habitat and depth), the management of the area, for example the presence of an MPA, and the sampling method and design. Other EOV-related supporting variables include among the others, fish life history traits (such as reproductive strategy), trophic ecology, movement, and behavior. By combining this information, it is possible to derive key indicators of the status of fish communities. These products include abundance, size, and biomass indices, taxonomic and functional diversity, indicators of community structure, foodweb indicators (e.g., the proportion of predatory fish or mean trophic level), and measures of fish production (e.g., spawning biomass or recruitment). Deliverable 3.1 Novel Monitoring Technologies 16 Table 1. Historical developments of zooplankton observation systems. Period System Comment Late 1800s – Early 1900s Net Sampling (e.g., Ring net, Bongo net, WP2) First standardised plankton nets 1931 Continuous Plankton Recorder (CPR) Sir Alister Hardy (UK) https://www.cprsurvey.org/ 1950s–1970s Bottle Sampling (Nansen, Niskin, Van Dorn) Used for collecting water at discrete depths; often paired with fine sieves for zooplankton. 1960s–1980s Pump Sampling Systems 1980s Video Plankton Recorder (VPR) Woods Hole Oceanographic Institution (WHOI); early prototypes for laboratory analyses. 1990 Laser Optical Plankton Counter (LOPC) Brook Ocean Technology (Canada); first commercial version developed. High-frequency sampling of the size spectrum of particles. No species identification. Mid-1990s Video Plankton Recorder (VPR) (Operational) WHOI; used for real-time, high-res imaging of plankton in situ. 1996–2000 ZooScan (Lab-based imaging) French CNRS laboratory allows semi-automated analysis of net/pump samples. 2003–2005 Underwater Vision Profiler (UVP5) Developed by Hydroptic (France). High-resolution imaging system. Archiving photos for further identification and counting. Later, coupled with AI imaging systems. 2010s Integration with Autonomous Platforms Gliders, AUVs, and profiling floats begin to incorporate VPR, LOPC, and UVP. 2020s–present AI and Deep Learning for Image Classification Automated species-level classification of plankton from imaging systems (e.g., EcoTaxa, PlanktonNet). 2.2 Strengths & Limitations for Traditional Sampling Methods 2.2.1 Benthos – marine life found on the bottom – diver-based underwater visual census Strengths • Non-invasive and cost-effective: Ideal for long-term ecological monitoring without damaging habitats. Deliverable 3.1 Novel Monitoring Technologies 17 • High taxonomic resolution (UVC and quadrats): Allows direct identification and size/abundance estimation, especially for fish, macroalgae, and visible invertebrates. • Standardized protocols: Widely used across Europe (e.g., MSFD), facilitating cross-study comparisons. • Useful in structured habitats: Effective in areas like rocky reefs and seagrass beds where habitat complexity supports diverse communities. • Photographic methods enable archiving: Images allow retrospective analysis and reduce observer bias during post-processing. • Applicable across biological systems: Adaptable to fish, macroalgae, benthic invertebrates, and seagrasses. Limitations • Observer bias and training dependency: Accuracy in UVC and quadrat methods depends heavily on diver expertise and consistency. • Limited by environmental conditions: Turbidity and swell can reduce visibility and survey accuracy. • Photographic methods may miss cryptic/understory species: Lower resolution for small or hidden taxa compared to in situ identification. • Spatial and temporal resolution vary: Not all methods capture fine-scale or rapid ecological changes (e.g., seasonality). • Limited depth coverage: SCUBA-based methods are constrained to depths reachable by divers (typically <30 m). • Image analysis is time-consuming: Manual annotation of images requires significant postprocessing effort. 2.2.2 Reef fish non-destructive sampling methods Strengths • UVC is non-destructive: appropriate for the monitoring of MPAs. • Availability of standardized protocols for UVC (e.g., RLS) and capture-based surveys (e.g., ICES 2022). • Video techniques increase capabilities: enabling long-term, continuous monitoring and deephabitat exploration. • Availability of standardized video methods: harmonized protocols for baited video surveys have become available recently, improving data comparability (Langlois et al. 2020). Limitations • Observer bias and skill dependency: UVC accuracy may be influenced by diver expertise and Deliverable 3.1 Novel Monitoring Technologies 18 inter-observer variability. • Frequent underestimation of elusive/mobile species: Species that avoid divers or move rapidly are often missed or miscounted. • Depth limitations for UVC: SCUBA-based surveys are generally restricted to shallow habitats (<30 m). • Manual video analysis is time-consuming: Without automated tools, processing large video datasets is labor-intensive. 2.2.3 Net sampling of zooplankton Strengths • Cost-efficient and accessible: Relatively simple and inexpensive equipment requiring a minimum of infrastructure in the field (in some cases). • Standardised and widely used: Established protocols allow for easy comparison across studies and periods. • Efficient for filtering large volumes: Can sample large volumes of water quickly, which increases the likelihood of capturing rare or patchy species. • Adaptability to environmental requirements: Can be used in various marine settings. • Collection of physical specimens: Samples can be preserved for detailed laboratory analyses. • Assessing trophic community structure: Enables studies on community composition, diversity, and food web dynamics. Limitations • Mesh size bias: Selective for organisms larger than the mesh size (smaller taxa pass through undetected). • Sample damage: Fragile organisms can be destroyed or deformed beyond recognition during towing, especially at high speed. • Net avoidance behaviour: Some zooplankton (mostly macrozooplankton) can detect and avoid approaching nets, leading to an underrepresentation of certain species. • Quantification limitation: Accurately estimating sampled water volume can be difficult, especially in open-ocean and turbulent settings. • Labour-intensive identification: Requires experienced taxonomists to identify organisms, especially for diverse assemblages or cryptic species. • Not-real time: Net sampling does not provide real-time data. • Potential contamination: Cross-contamination between samples or improper cleaning of gear can affect data quality. Deliverable 3.1 Novel Monitoring Technologies 19 2.3 Emerging Methods 2.3.1 Environmental DNA (eDNA) Environmental DNA (eDNA) enables the detection of DNA fragments originating from organisms in aquatic environments. These DNA fragments are typically released into the environment through natural processes such as cell sloughing, the shedding of reproductive structures, or the degradation of tissue (Thomsen and Willerslev 2015; Deiner et al. 2017). By detecting these fragments, eDNA methods provide a highly sensitive tool for identifying the presence of macroalgal, fish, seagrass, and plankton communities without the need for direct visual observation or physical collection of the organisms. This approach offers significant advantages for monitoring hard-to-access or visually cryptic species within aquatic ecosystems. Sampling (water collection, filtration, DNA extraction, primers, sequencing) The eDNA workflow begins with the collection of water samples from environments of interest. The volume of water required for detection is based on the target organism of interest, with rarer or pelagic species (e.g., fish or cephalopods) often requiring greater volumes to be collected. The water is ideally filtered directly in the field, or, if unable to do so, into sterile containers which are then filtered at a later time point (Goldberg et al. 2016). It must be noted that eDNA can also be sampled from sediment, which follows a similar procedure except that sediment replaces the water filters. Following collection, DNA is extracted from the collected filters using specialized extraction kits, such as PowerSoil® Pro or PowerWater® kits (Qiagen), which are selected depending on sample type and desired downstream applications. After extraction, specific primers designed to amplify DNA sequences of organisms of interest, targeting regions such as the rbcL or Cytochrome c oxidase subunit I (COI) genes, are used (see Table 2; Stat et al. 2017). Amplified DNA products are then sequenced using high-throughput sequencing platforms (e.g., Illumina MiSeq), providing detailed information about the diversity and relative abundance of taxa present in the sampled environment. Deliverable 3.1 Novel Monitoring Technologies 20 Table 2. Common primer pairs used to target different organisms of interest. COI indicates the cytochrome-coxidase subunit 1 within the mitochondrial electron transport chain. Data output: presence/absence, abundance proxies, and phenology The primary output from eDNA analysis is the detection of species presence or absence, offering information on whether particular taxa are present within the sampled area (Kelly et al. 2014). Deliverable 3.1 Novel Monitoring Technologies 21 Abundance proxies can also be inferred based on the quantity of DNA sequences obtained for each taxon. Furthermore, incorporating qPCR assays with sequencing data can give clearer information on the ‘absolute’ read numbers of genes of interest (Nappi et al. 2022). While these abundance estimates provide valuable ecological insights, it is important to note that they are influenced by factors such as DNA degradation rates, shedding rates, and various environmental conditions and may not accurately represent the actual biomass of the target organism (Yates et al. 2019). Moreover, eDNA methodologies can capture temporal patterns in species occurrence, including reproductive events. For instance, DNA fragments associated with reproductive phases, such as spore release or gamete production, can be detected, thus providing insights into the phenology of macroalgal species. 2.3.2 Imaging & AI-based analysis Imaging and artificial intelligence-based analysis, particularly those leveraging Machine Learning (ML) techniques like Convolutional Neural Network (CNN), are increasingly demonstrating their abilities and applications in ecological research (Piechaud et al. 2019; Rubbens et al. 2023). These techniques are being used for tasks such as segmentation and classification, which enable the identification of species from images and videos. Such applications enhance the scalability, efficiency, and accuracy of biodiversity monitoring, species recognition analysis, animal behavior tracking, and predictive ecological modelling (Piechaud and Howell 2022; Rubbens et al. 2023). However, these techniques rely on access to sufficient, high-quality datasets to perform effectively. This underscores the need to deploy appropriate data collection protocols or platforms, like underwater cameras, remotely operated vehicles (ROVs), and diver-operated systems for gathering images and other outputs to train algorithms for species detection, pattern recognition, and ecological change monitoring (Bridges et al. 2025). Imaging platforms: underwater cameras, Remotely Operated Vehicles (ROVs), diver-operated systems Accurate image-based ecological analysis begins with the deployment of effective imaging platforms capable of capturing high-resolution visual data in underwater environments (Moghimi and Mohanna 2021; Nawaz et al. 2025). The most common tools include underwater cameras, remotely operated vehicles (ROVs), and diver-operated systems. Each platform offers unique advantages depending on the depth, habitat type, species behavior, and operational conditions. • Underwater cameras are widely used for monitoring benthic habitats, including macroalgal communities and seagrass meadows, as well as fish populations. These cameras can be deployed on tripods, floats, autonomous underwater vehicles (AUVs), or moorings. They are particularly valuable for time-lapse studies, long-term observations, and monitoring the diurnal/nocturnal behavior of marine species. For seagrass meadows, monitoring fish assemblages and the frequency of grazing events within the habitat can provide valuable insights into ecosystem dynamics, trophic interactions, and overall meadow health (Gilby et Deliverable 3.1 Novel Monitoring Technologies 22 al. 2018; Jones et al. 2021). • ROVs are tethered robotic systems operated from the surface and equipped with highdefinition cameras and sensors. They allow access to deep or hazardous environments without risking diver safety. ROVs are ideal for systematic transect imaging, habitat mapping, and exploration of deeper reef systems (Perkins et al. 2022). Their integration with real-time data transmission makes them effective for adaptive data collection and remote monitoring missions. • Diver-operated systems remain crucial in shallow water surveys where precision and flexibility are required. In diver-operated systems, divers are equipped with handheld or body-mounted cameras, allowing for close-range, targeted data collection. These systems are particularly useful for capturing images of targeted species and fine-scale ecological interactions. Leveraging Convolutional Neural Network (CNN) as a Machine Learning (ML) technique Machine Learning (ML) is a subset of Artificial Intelligence, which involves the use of algorithms that are capable of understanding patterns within inputted data and automatically improving from experience without the need for explicit programming (Sarker 2021). However, within ML, there is a more advanced technique known as deep learning, which leverages artificial neural networks to model complex, non-linear relationships in data (enabling tasks like image recognition, natural language processing, etc.). This application is useful when working with unstructured data like images and videos. In ecological research, Convolutional Neural Networks (CNNs) are increasingly applied to automate the process of species identification, coverage estimation through image segmentation, and to monitor behavioral patterns from large volumes of visual data (Mac Aodha et al. 2018; Pichler and Hartig 2023). CNN, as a deep learning technique, is structured to automatically detect spatial hierarchies in images by learning to extract features such as edges, textures, and complex shapes through layers of convolutional filters. Popular CNN model architectures include AlexNet, VGGNet, ResNet, and YOLO, each optimized with different algorithms for different image analysis tasks. Once trained on annotated datasets, these models can generalize well to new, unseen data, making them ideal for large-scale ecological monitoring across space and time. Before CNN can effectively recognize patterns or features in images, it must be trained using annotated data/labelled images. Image annotation Annotation helps define the “ground truth” that CNNs use to learn. This process requires drawing bounding boxes around individual species or objects of interest in the image. These labeled datasets serve as supervised learning targets, enabling CNNs to associate specific visual patterns within the species of interest. There are several software/tools used for image annotation or labelling, most of which are open source (LabelImg, CVAT, LabelMe, Roboflow). CNNs work by learning patterns in visual data, but they require labeled/annotated examples of the data to perform effectively. The annotation serves as supervised learning targets, enabling the network to understand what it should learn from each image to achieve fine-grained classification. These annotated images allow CNNs to learn what Deliverable 3.1 Novel Monitoring Technologies 23 visual characteristics are associated with specific ecological categories or classes (Nawaz et al. 2025). As CNNs learn by recognizing patterns in annotated examples, high-quality annotation is essential because errors or inconsistencies in labeling can lead to poor model performance or misclassification (Moghimi and Mohanna 2021; Perkins et al. 2022). Figure 12. A simple workflow/pipeline for an AI-based species recognition model. Model training To properly train a model, it is essential to select the best model based on the application or use case, because the model can range from segmentation to classification. Therefore, once the dataset has been prepared through image annotation, the CNN model is ready for training (Fig. 2). This is the process where the network learns to identify features in each dataset. During training, CNN makes predictions and compares them to the correct species labels (Fig. 2). Once the training is complete, the model is evaluated using the test dataset split. The model is ready for deployment if high accuracy is achieved (a common method of training is the transfer learning method). To better understand the performance of the model, they are usually evaluated using performance metrics (confusion matrix, precision, recall, F1-score, etc.) (Fig. 3). For instance, the confusion matrix reveals misclassifications and highlights areas needing improvement. These metrics can be combined to further provide a balanced indicator of the overall model's effectiveness. Deliverable 3.1 Novel Monitoring Technologies 24 Figure 3. YOLO training summary showing loss, precision, recall, and mAP metrics over epochs. Automated identification of macroalgae and percentage (%) cover calculation Once trained, CNNs can process large volumes of visual data collected by underwater cameras or remotely operated vehicles (ROVs), enabling rapid and accurate analysis of habitat composition. This includes the identification and quantification of key benthic marine life such as macroalgae species, seagrasses, corals, etc. All of which are components of Essential Ocean Variables (EOVs). By automating these processes, CNNs significantly reduce the time, labor, and cost associated with traditional methods of surveys, which are often resource-intensive, requiring significant time, manual effort, and financial investment, which limits their scalability and broader application (GiménezRomero et al. 2024; Nawaz et al. 2025). Techniques such as image segmentation and masking may be used to quantify percentage coverage. Through segmentation and masking of the area of interest in an image (Fig. 4) and comparing the difference to the total area of the image, the percentage coverage of this isolated area can be calculated. Deliverable 3.1 Novel Monitoring Technologies 25 Figure 4. A depiction of an annotated image and a segmented model prediction output Integrated approaches to zooplankton monitoring Integrating the Underwater Vision Profilers (UVPs) with autonomous platforms and automatic image recognition using AI is a promising avenue for the coming years that should allow the accumulation of very large quantities of data to monitor zooplankton in 3 dimensions and over time. Already, using a few thousand profiles from multiple oceanographic campaigns, global mean biomasses of various groups of mesozooplankton built using AI techniques have been published (Drago et al. 2022; Soviadan et al. 2024). In addition, this technique could be combined with automatic eDNA sampling in the future. 2.3.3 Remotely operable platforms and remote sensing Satellite remote sensing provides unique capabilities for monitoring marine biodiversity. Remote sensing enables regular, synoptic observations across a wide range of spatial and temporal scales from a few meters to ocean basins, and from days to decades. Among available techniques, optical remote sensing in the visible range remains the primary method for biological and biodiversity observations. Although the technology itself is well established, recent advances in spatial and spectral resolutions and data accessibility are expanding its applications. Phytoplankton monitoring has long been a central application of optical remote sensing. Sensors such as NASA’s MODIS and SeaWiFS, and Copernicus Sentinel-3 OLCI, with spatial resolutions between 300 m and 1 km, provide a nearly 30-year continuous global record of ocean color data from which the maps of chlorophyll-a concentration, the proxy for phytoplankton biomass, are produced. Newer hyperspectral missions, such as the recent NASA PACE and the upcoming Copernicus CHIME and NASA SBG, will provide an enhanced spectral resolution, which is widely anticipated to enable new capabilities for observing phytoplankton communities, including their composition and functional types (Groom et al. 2019). Optical satellite data are also instrumental in monitoring foundation species such as submerged and floating macroalgae (e.g., giant kelp, pelagic Sargassum), seagrass meadows, and coral reefs. Mapping these species has relied on medium-resolution data from Landsat (30 m, since 1982) or very high- Deliverable 3.1 Novel Monitoring Technologies 32 • Need for site-specific protocols: Methodologies must often be adapted to specific environments or objectives, making it harder to standardize UAV workflows across diverse coastal ecosystems. 2.4.5 Passive acoustic monitoring Strengths • Non-invasive monitoring: Hydrophones record underwater sounds without disturbing marine life or habitats, making them ideal for sensitive or protected areas (Gibb et al. 2019; Mooney et al. 2020; La Manna et al. 2024; Muñoz-Duque 2024). • Continuous, long-term data collection: PAM enables round-the-clock monitoring, capturing diel, seasonal, and annual trends that periodic surveys may overlook (Van Parijs et al. 2009; Lindseth and Lobel, 2018). • Detection of cryptic and nocturnal species: Many marine organisms are difficult to observe visually, but produce distinctive sounds, allowing detection of species that are hidden, nocturnal, or inhabit deeper waters (Bertucci et al. 2016; Mooney et al. 2020). This strength is particularly valuable in environments with high water turbidity, where traditional visual surveys are limited or ineffective; hydrophones can reliably capture biological sounds regardless of water clarity (Bertucci et al. 2016; Gibb et al. 2019; Mooney et al. 2020). • Broad spatial coverage: Sound travels efficiently underwater, so a single hydrophone can monitor a relatively large area, an advantage in complex habitats like kelp forests (Radford et al. 2010; Dziak et al. 2023). • Cost-effective for extended studies: Once deployed, hydrophones require less human intervention than repeated diver or ROV surveys, reducing operational costs for long-term projects (Van Parijs et al. 2009; Gibb et al. 2019). • Sensitive to ecosystem changes: Shifts in the acoustic community, such as changes in fish chorusing or increases in anthropogenic noise, can indicate alterations in ecosystem health or biodiversity (Merchant et al. 2015; Bertucci et al. 2016). Limitations • Species identification constraints: Not all marine species produce sounds, and some acoustic signals are difficult to attribute to specific taxa without supporting visual or genetic data (Gibb et al. 2019; Mooney et al. 2020). To ameliorate this, integrated cameras can be useful. • Acoustic masking: Dominant sounds (e.g., snapping shrimp, waves, boat engines) can mask quieter biological signals, complicating data interpretation and potentially underestimating diversity (Merchant et al. 2015; Mooney et al. 2020). • Environmental and technical biases: Sound propagation varies with water depth, temperature, substrate, and hydrophone placement, which can bias recordings and complicate comparisons across sites (Radford et al. 2010; Dziak et al. 2023). Deliverable 3.1 Novel Monitoring Technologies 33 • Large data management needs: PAM generates substantial audio data, requiring significant storage, processing power, and expertise for effective analysis (Lindseth and Lobel 2018; Gibb et al. 2019). • Limited to soniferous species: Species that are silent or rarely vocalise remain undetected, potentially biasing biodiversity assessments toward more vocal taxa (Bertucci et al. 2016; Mooney et al. 2020). • Need for validation: Acoustic patterns and indices often require validation with traditional survey methods (e.g., visual censuses, net sampling) to ensure ecological relevance and accuracy (Gibb et al. 2019; Van Parijs et al. 2009). 2.4.6 Active acoustic Strengths • High technological readiness: Acoustic backscatter technology is already highly developed and routinely used on research vessels, with increasing availability on autonomous and remote platforms (Haëntjens et al. 2020). • Established global data pool: A substantial global archive of acoustic backscatter data already exists and continues to grow through ongoing oceanographic research, offering a strong foundation for long-term biodiversity and ecosystem monitoring. • Scalable data integration: Creating pipelines to compile and integrate new observations into global databases would be relatively straightforward, enabling large-scale and consistent analysis across regions. • Functional diversity and ecosystem relevance: Acoustic data can inform understanding of functional diversity and key ecosystem processes, such as the biological carbon pump, linking biodiversity to climate-relevant functions. • Broad application potential: While current uses may be focused on a subset of ecological functions, acoustic backscatter has promising potential for broader applications across biodiversity assessment and marine ecosystem monitoring. Limitations • Proxy-based, not taxon-specific: Acoustic backscatter primarily reflects biomass but not species identity. Differences in organism morphology and material properties mean acoustic signatures vary, leading to uncertainty in species composition. • Inability to resolve community structure alone: Without additional data sources (e.g., eDNA, visual imaging), acoustic backscatter cannot resolve fine-scale taxonomic or trait-based diversity, limiting its effectiveness as a standalone biodiversity tool. • Need for technological and methodological integration: Advancements such as frequencymodulated acoustics, improved classification algorithms, and integration with trait-based ecological models are needed to improve accuracy and ecological relevance. Deliverable 3.1 Novel Monitoring Technologies 34 3. Case studies 3.1 eDNA sampling versus traditional fish surveys in the Baltic Sea In a case study led by Uppsala University (UU) in the Baltic Sea Living Lab, traditional coastal monitoring nets (modified Nordic coastal survey net) were compared with eDNA methods to monitor fish communities in seagrass-dominated ecosystems around the island of Gotland, Sweden. The study, currently being prepared for scientific publication, was conducted at six locations: Ajkesviken, Fårösund, Vägumeviken, Lausviken, Burgsvik, and Klintehamn. At each site, eDNA samples were collected from the same locations where nets were deployed, allowing for direct comparison between the two approaches (Fig. 5). Traditional coastal nets (Nordic coastal survey net), the established monitoring method in the region, offer valuable data continuity and allow for physical measurements such as fish length, weight, and age (through otolith collection). In this study, nets captured 19 species in total, with perch and herring being the most common. However, this method is labour-intensive, requiring two boat trips (at dawn and dusk) to each location, as well as substantial time for net handling and fish processing. Moreover, it is inherently invasive: a significant number of fish are killed during the process. eDNA, by contrast, offers a non-invasive, efficient, and highly sensitive alternative. Across the six sites, eDNA identified more than twice as many fish species, including small-bodied and cryptic species that are typically underrepresented in net catches. Sampling is rapid and minimally disruptive, requiring only a single boat trip per location and no handling of fish. Although eDNA does not provide physical measurements, abundance estimates, and currently lacks historical time series in this region, its ability to expand species detection while reducing impact on wildlife makes it a valuable tool. The main limitations are the relatively high cost and the time required for laboratory analysis. Discussions with local authorities are ongoing, with a strong interest in integrating eDNA into the monitoring framework. A combined approach – reducing the number of nets while incorporating eDNA – is emerging as a promising path forward, enhancing both ecological insight and ethical standards in long-term fish monitoring. Figure 5. Fieldwork around Gotland, looking for suitable fishing grounds and testing environmental DNA (eDNA) sampling protocol (left), Seagrass-dominated fishing ground (middle), Fish catch being measured in the lab (right). Deliverable 3.1 Novel Monitoring Technologies 35 3.2 eDNA versus traditional macroalgal sampling in the Tuscan Archipelago living lab Calibration of emerging technologies such as environmental DNA (eDNA) with traditional visual sampling methods for monitoring the macroalgal canopy cover and composition EOV, is one of the main tasks of the Tuscan Archipelago living lab, led by the University of Pisa (UNIPI). UNIPI has conducted a series of field campaigns to evaluate the efficacy of eDNA sampling compared to traditional methods (photo quadrats) for assessing macroalgal cover and composition in Tuscan Archipelago living lab. During the summer of 2024, surveys were carried out at three islands: Capraia, Pianosa, and Giannutri. A total of 72 1L eDNA samples and 58 tissue samples were collected using a sampling design that matches photoquadrats with water samples for eDNA at multiple sites (Fig. 6) to a sequencing failure, only 32 samples passed quality control, and the results should therefore be considered preliminary. Despite this limitation, the findings are promising. eDNA data showed a positive correlation between algal cover of Cystoseira foeniculacea and the relative abundance of its corresponding Amplicon Sequence Variant (ASV) (β = 0.327, t = 3.62, R² = 0.56, p < 0.01). (Fig. 7). Moreover, on average, eDNA detected approximately 45% more canopy-forming species than traditional visual methods. To address the technical setbacks encountered in 2024, UNIPI is planning an additional series of field campaigns in 2025, with the goal of expanding the barcode reference library and improving sequencing success rates. Furthermore, the detection range of eDNA shall be assessed using the sampling design depicted in Figure 6. Together, these preliminary results highlight the potential of eDNA to detect rare canopy-forming species and to quantify the more abundant ones. Figure 6. Sampling design for the 2025 DNA field campaign. Water samples shall are collected within the macroalgal canopy and at replicate distances from the canopy edge to assess the distance eDNA can be detected from remnant canopies. Deliverable 3.1 Novel Monitoring Technologies 36 Figure 7. Relationship between the relative abundance of Amplicon Sequence Variants (ASVs) and the corresponding percentage cover of Cystoseira foeniculacea across islands (color-coded) in the Tuscan Archipelago Living Lab. 3.3 AI-based imaging for species recognition in the Tuscan Archipelago living lab The University of Pisa (UNIPI) is currently developing a YOLO architecture based on deep learning to support automated assessment of the macroalgal and seagrass canopy cover and composition EOVs in Tuscan Archipelago Living Lab. The model also enables quantitative assessments such as percentage coverage estimation. By developing an automated AI-based biodiversity monitoring technique with considerably high species identification accuracy, this could impact scalable solutions for monitoring diverse coastal ecosystems. Beyond species identification, the project also aims to integrate a percentage cover estimation workflow by combining instance segmentation techniques with pixelbased analysis. This will enable the quantification of species-specific spatial distribution and coverage, providing valuable metrics for monitoring changes in cover of seagrass and macroalgal meadows over time. The development of this AI-powered biodiversity monitoring tool holds the potential to significantly improve the scalability, frequency, and objectivity of coastal habitat assessments, compared to traditional manual surveys. This novel method could offer reliable insights and a costeffective way for tracking habitat changes, detecting early signs of ecological shifts, and informing conservation strategies across the Tuscan Archipelago and similar Mediterranean coastal ecosystems. Deliverable 3.1 Novel Monitoring Technologies 37 3.4 Using video to estimate jellyfish abundance in the Baltic Sea In a recent study led by UU (Lüskow et al. 2025) in the Baltic Sea Living Lab, existing video material, which was originally collected for the study of seagrass-associated fish in the coastal waters of Gotland, was used opportunistically to estimate the abundance of jellyfish. This "by-product" use of the recordings of a GoPro HERO 5 Black camera that showed a considerable number of more than 4000 Aurelia aurita jellyfish, underlines the enormous potential of novel technologies in marine research. In the central Baltic Sea region, where quantitative long-term monitoring programmes for gelatinous zooplankton (GZ) are almost completely missing, innovative approaches such as camera transects offer a valuable, non-invasive, and time-efficient alternative. They enable us to overcome financial and logistical restrictions and historical sampling difficulties. Although the primary collection of data (seagrass-associated fish fauna) served a different purpose, it provided first-time insights into A. aurita aggregation patterns over seagrass-dominated habitats with salinities between 6 and 8. The orientation and inclusion of such new technologies are, therefore, crucial to effectively improve our understanding of long-term pelagic community changes and to enable more comprehensive monitoring of GZ populations. 3.5 Assessing Vertical Migrant Communities Using Acoustic Backscatter Data To assess the feasibility of acoustic backscattering data (historical and current) to quantify the state and functional diversity of the vertical migrant community and its impact on the biological carbon pump. A, database was compiled of recent acoustic profiles (2006 to 2023) covering a combined sampling track of over 800,000 km. To streamline this initial data assembly, only observations collected at a 38 kHz frequency were used. Data was pre-processed to remove noise, deal with missing data, and collate observations into 10 km segments. Depth bins were then classified into background and DSL bins using a column-wise contextual filter. This allowed DLS metrics to be extracted – (a) weighted mean depth, (b) depth and strength of the strongest DLS, (c) depth and strength of all DLSs. Figure 8. The frequency distribution and spatial distribution of the principal DSL. Deliverable 3.1 Novel Monitoring Technologies 38 The principal (i.e., strongest) DSL shows a bimodal frequency distribution (Fig. 8), some in the surface layer (0–200 m), but also a surprisingly large fraction (more than 50%) in the mesopelagic zone (400– 800 m). Figure 9. The frequency distribution and spatial distribution of the multiple DLSs. In addition to the principal DSL, there are often multiple additional DLSs, usually representing other taxa in the mesopelagic community (Fig. 9). 4. Comparative evaluation 4.1. Summary of strengths and limitations of different methods Here, we provide a summary of the relative strengths, limitations, and logistical considerations of traditional and innovative approaches for biodiversity monitoring, focusing on their utility for Essential Ocean Variable (EOV) assessment (Table 3). Traditional methods, while historically rich and taxonomically precise, are often invasive and resource intensive. In contrast, newer technologies such as environmental eDNA, imaging with AI, and remote sensing offer scalable, non-invasive alternatives, though they still require calibration and careful validation. Furthermore, emerging methods offer key advantages for remote or inaccessible environments. For example, eDNA sampling and satellite or UAV imaging can be deployed without requiring divers or vessels, reducing logistical barriers. Imaging systems and eDNA enable rapid assessments, while traditional methods remain superior for highresolution long-term monitoring. Species-specific detection is enhanced by eDNA and AI, whereas habitat-level metrics are best captured by UAVs or remote sensing platforms. The choice of method depends on monitoring goals, spatial scale, and environmental constraints. A hybrid approach that leverages the strengths of each method is increasingly seen as best practice for robust, cost-effective, and ethical monitoring of marine biodiversity. Deliverable 3.1 Novel Monitoring Technologies 39 Table 3. Comparative evaluation of traditional and emerging methods for marine biodiversity monitoring. Criterion Traditional Methods eDNA Imaging / AI Acoustics / Remote Sensing Taxonomic resolution High (if expert available) High (if reference barcodes exist) Medium to high (depends on image quality and training data) Low (typically functional groups or biomass proxies only) Invasiveness Often invasive (e.g., nets, trawls) Non-invasive Non-invasive Non-invasive Abundance estimation Direct (counts, biomass) Indirect (sequence counts as proxies) Possible (via segmentation and pixel quantification) Proxy-based (backscatter strength) Detection of rare/cryptic Variable; often missed Excellent (if DNA shed is present) Moderate (depends on training and resolution) Poor (limited to reflective organisms) Spatial resolution High (fine-scale site-based) Low–medium (due to DNA dispersion) Medium–high (depending on sensor resolution and positioning) High (satellites/UAVs), but lower in species-level detail Temporal resolution Low (periodic surveys) Medium (repeated water sampling possible) High (e.g., continuous camera deployments) High (real-time or repeated satellite/floats) Automation potential Low Medium (automated lab pipelines exist) High (image processing and ML automation) Medium–high (pipelineable via remote platforms) Logistical needs Diver teams, vessels, and lab processing Field kits + lab sequencing Cameras, ROVs/AUVs, annotation tools Ship acoustics, UAVs, satellites, or autonomous floats Calibration required Standardized protocols exist Calibration to biomass needed Ground-truthing against manual ID needed Cross-validation with other datasets is essential Cost (setup & running) High (staffintensive, ship time) Medium (lab-based, but scalable) Medium–high (hardware + training costs) Variable (low for satellites, higher for acoustics/UAVs) Key strengths Detailed ID, longterm comparability Cryptic taxa, early detection Scalable, fast, and objective once trained Broad coverage, functional/biomass data Key limitations Time-intensive, invasive Spatially implicit, not quantitative for abundance Requires large labeled datasets, species bias in training Low taxonomic resolution, signal interference, needs modeling Rapid versus long-term use Excellent for longterm datasets Suitable for repeated, short-term surveys Scalable for highfrequency use Excellent for repeated, realtime monitoring Species versus habitat coverage Good for both, species-specific (if an expert is available) High for species, low for habitat structure High for both, especially with AI segmentation Primarily habitat-level or biomass patterns Remote area accessibility Limited (requires divers, ships) High (lightweight field kits) Medium–high (depends on platform) High (satellites, UAVs, Argo floats) Deliverable 3.1 Novel Monitoring Technologies 40 5. Advancing underwater monitoring in line with the 3s In line with the principles of the 3Rs – Replacement, Reduction, and Refinement – there is a need to transition from traditional fish monitoring methods, such as trawling and netting, to more ethical and innovative approaches. Conventional sampling techniques often result in high levels of bycatch and mortality, including the unintended killing of non-target species, juveniles, and vulnerable populations. These methods conflict with contemporary efforts to minimise animal suffering and are increasingly difficult to justify in the context of both legal obligations and societal expectations regarding animal welfare. Novel, emerging, non-invasive technologies such as environmental DNA (eDNA) sampling, underwater video systems, and acoustic monitoring offer viable alternatives that align closely with the 3Rs. These tools can replace or significantly reduce the need to capture or kill animals by enabling effective biodiversity assessments and fish population monitoring without physical harm. They also contribute to refinement by eliminating or reducing the distress and mortality associated with traditional methods. Adopting these approaches supports a more sustainable and humane model for aquatic research and conservation, consistent with the goals of EU animal welfare legislation and scientific best practice. This is directly relevant to Directive 2010/63/EU of the European Parliament and of the Council, which mandates the implementation of the 3Rs in all scientific procedures involving animals, including those involving fish. However, in certain cases – such as when assessing, for example, fish fitness, analysing bone structure, or examining otoliths for age and growth studies – the physical specimens themselves are still required. In such instances, combining modern non-invasive techniques with targeted, minimally harmful traditional sampling can offer a balanced approach that upholds the 3Rs while meeting essential scientific objectives. 6. Integration and hybrid approaches The integration of diverse monitoring methods represents a significant advancement for marine biodiversity assessments. Hybrid approaches leverage the complementary strengths of various techniques, significantly enhancing both data quality and ecological understanding. For example, combining eDNA with imaging methods, such as underwater photography or video analysis, provides a more comprehensive representation of biodiversity. eDNA captures genetic signatures of organisms, including elusive or cryptic species that might otherwise remain undetected visually, whereas imagery allows direct observation of small-scale variations in species abundance (Alfaro-Cordova et al. 2022). A study conducted in the Gulf of California demonstrated that eDNA and underwater visual census (UVC) each detected distinct subsets of species, and their integration nearly doubled the total number of species detected (Valdivia-Carrillo et al. 2021). Similarly, the combination of drone-based remote sensing and traditional in situ sampling techniques creates opportunities for scaling monitoring efforts while maintaining data accuracy. Drones facilitate rapid, large-scale mapping and assessment, providing spatially extensive data that can guide targeted Deliverable 3.1 Novel Monitoring Technologies 41 in situ sampling (Ventura et al. 2023). Ground-based sampling then validates and refines remote assessments, ensuring accurate interpretation and ecological relevance. A key aspect of successful hybrid approaches is robust calibration, wherein traditional monitoring methods are employed to validate data obtained from eDNA analyses, automated processes, AIdriven approaches, or remote sensing techniques. This step is crucial for ensuring data quality, accuracy, and reliability, facilitating iterative improvement and refinement of machine learning algorithms and predictive models (Chen 2021). Integrating machine learning with eDNA methods, for instance, can effectively detect environmental change patterns at landscape scales (Keck et al. 2023). Consequently, calibrated hybrid approaches significantly enhance the consistency, comparability, and scalability of biodiversity monitoring across various spatial and temporal scales. For long-term monitoring programs, maintaining continuity in traditional methods may be necessary to avoid disrupting valuable time series; however, introducing new technologies as early as possible—ideally with overlapping implementation—can build the foundation for eventual transitions and potentially replace traditional methods in the future. The growing development and usage of portable field-based sequencing platforms, such as Oxford Nanopore's MinION, and their integration with drone-based imagery, field-based imagery, and AIdriven recognition, promise substantial improvements in biodiversity monitoring. Portable sequencing platforms enable rapid, low-cost, and scalable sequencing even in field conditions, making them highly suitable for integration into automated and remote monitoring systems (Srivathsan et al. 2021). Similarly, drones possess considerable yet largely untapped potential for field research, especially when combined with artificial intelligence and additional data sources, greatly enhancing the scope and scale of biodiversity assessments (Surasinghe et al. 2025). Strengthening these integrative approaches would facilitate the establishment of continuous monitoring systems, enabling real-time biodiversity assessments and rapid detection of ecological changes. 7. Recommendations • Combine methods for complementarity. Use hybrid monitoring frameworks that integrate eDNA, AI-based image analysis, and traditional sampling to maximize biodiversity detection and monitoring accuracy. • Prioritize calibration and benchmarking. Side-by-side comparisons are essential to ensure consistency and comparability between emerging and traditional methods. • Build shared, open-access resources. Develop and maintain open-access barcode libraries, annotated image databases, and cloud-based analytical tools to support data standardization and model training. All resources should adhere to FAIR principles (Findable, Accessible, Interoperable, and Reusable) to maximize transparency, usability, and integration across platforms and research communities. • Support capacity building and engagement. Promote training in AI and molecular techniques across institutions and involve citizen scientists using robust and scalable protocols (e.g., RLS). 48 Koh, L. P., & Wich, S. A. 2012. Dawn of drone ecology: Low-cost autonomous aerial vehicles for conservation. Tropical Conservation Science, 5(2), 121–132. https://doi.org/10.1177/194008291200500202 Kulbicki, M. 1998. How the acquired behaviour of commercial reef fishes may influence the results obtained from visual censuses. Journal of Experimental Marine Biology and Ecology, 222(1-2), 11-30. https://doi.org/10.1016/S0022-0981(97)00133-0 La Manna, G., Guala, I., Pansini, A., Stipcich, P., Arrostuto, N., & Ceccherelli, G. 2024. Soundscape analysis can be an effective tool in assessing seagrass restoration early success. Scientific Reports, 14(1), 20910. https://doi.org/10.1038/s41598-024-71975-2 Langlois, T., Goetze, J., Bond, T., Monk, J., Abesamis, R. A., Asher, J., Barrett, N., Bernard, A. T. F., Bouchet, P. J., Birt, M. J., Cappo, M., Currey-Randall, L. M., Driessen, D., Fairclough, D. V., Fullwood, L. A. F., Gibbons, B. A., Harasti, D., Heupel, M. R., Hicks, J., Holmes, T. H., Huveneers, C., Ierodiaconou, D., Jordan, A., Knott, N. A., Lindfield, S., Malcolm, H. A., McLean, D., Meekan, M., Miller, D., Mitchell, P. J., Newman, S. J., Radford, B., Rolim, F. A., Saunders, B. J., Stowar, M., Smith, A. N. H., Travers, M. J., Wakefield, C. B., Whitmarsh, S. K., Williams, J., & Harvey, E. S. 2020. A field and video annotation guide for baited remote underwater stereo‐video surveys of demersal fish assemblages. Methods in Ecology and Evolution, 11(11), 1401-1409 https://doi.org/10.1111/2041-210X.13470 Leray, M., Yang, J. Y., Meyer, C. P., Mills, S. C., Agudelo, N., Ranwez, V., Boehm, J. T., & Machida, R. J. 2013. A new versatile primer set targeting a short fragment of the mitochondrial COI region for metabarcoding metazoan diversity: Application for characterizing coral reef fish gut contents. Frontiers in Zoology, 10(1), 34. https://doi.org/10.1186/1742-9994-10-34 Lindseth, A. V., & Lobel, P. S. 2018. Underwater soundscape monitoring and fish bioacoustics: A review. Fisheries Research, 205, 36-46. 10.3390/fishes3030036 Lüskow, F., Neitzel, P., Lawrence, E. R. & Nordlund, L. M. 2025 Jellyfish and ctenophores around Gotland in the Baltic Sea – Local data contributing to global assessments. J Mar Sci Eng 13 (5): 852. https://doi.org/10.3390/jmse13050852 Mac Aodha, O., Gibb, R., Barlow, K. E., Browning, E., Firman, M., Freeman, R., Harder, B., Kinsey, L., Mead, G. R., Newson, S. E., Pandourski, I., Parsons, S., Russ, J., Szodoray-Paradi, A., Szodoray-Paradi, F., Tilova, E., Girolami, M., Brostow, G., & Jones, K. E. 2018. Bat detective—Deep learning tools for bat acoustic signal detection. PLOS Computational Biology, 14(3), e1005995. https://doi.org/10.1371/JOURNAL.PCBI.1005995 Mallet, D., & Pelletier, D. 2014. Underwater video techniques for observing coastal marine biodiversity: a review of sixty years of publications (1952–2012). Fisheries Research, 154, 44-62. https://doi.org/10.1016/j.fishres.2014.01.019 49 Merchant, N. D., Fristrup, K. M., Johnson, M. P., et al. 2015. Measuring acoustic habitats. Methods in Ecology and Evolution, 6(3), 257-265. http://dx.doi.org/10.1111/2041-210X.12330 Miloslavich, P., Bax, N. J., Simmons, S. E., Klein, E., Appeltans, W., Aburto‐Oropeza, O., ... & Shin, Y. J. 2018. Essential ocean variables for global sustained observations of biodiversity and ecosystem changes. Global change biology, 24(6), 2416-2433. https://doi.org/10.1111/gcb.14108 Moghimi, M. K., & Mohanna, F. 2021. Real-time underwater image enhancement: a systematic review. Journal of Real-Time Image Processing, 18(5), 1509–1525. https://doi.org/10.1007/S11554-020-01052-0 Mooney, T. A., Kaplan, M. B., & Lammers, M. O. 2020. Listening forward: Approaching marine biodiversity assessments using acoustic methods. Royal Society Open Science, 7(8), 201287. 10.1098/rsos.201287 Moore, J.P., Green, H. C., Stewart, D.J., Lüskow, F., Wilder, M. L. 2025 Invasive freshwater jellyfish (Craspedacusta cf. sowerbii) in the Hudson River Basin, NYS: Comparison of detection methods. Hydrobiologia 852 (resubmitted) Muller-Karger, F. E., Miloslavich, P., Bax, N. J., Simmons, S., Costello, M. J., Sousa Pinto, I., et al. 2018. Advancing marine biological observations and data requirements of the complementary essential ocean variables (EOVs) and essential biodiversity variables (EBVs) frameworks. Frontiers in Marine Science, 5, 211 https://doi.org/10.3389/fmars.2018.00211 Muñoz-Duque, S., Fonseca, P.J., Quintella, B., Monteiro, J.G., Fernandez, M., Silva, R., Vieira, M. and Amorim, M.C.P. 2024. Acoustic fish community in the Madeira Archipelago (North Atlantic Ocean): Characterization of sound diversity and daily patterns. Marine Environmental Research, 199,106600. https://doi.org/10.1016/j.marenvres.2024.106600 Nalmpanti, M., Chrysafi, A., Meeuwig, J. J., & Tsikliras, A. C. 2023. Monitoring marine fishes using underwater video techniques in the Mediterranean Sea. Reviews in Fish Biology and Fisheries, 33(4), 1291-1310. https://doi.org/10.1007/s11160-023-09799-y Nappi, J., Goncalves, P., Khan, T., Majzoub, M. E., Grobler, A. S., Marzinelli, E. M., Thomas, T., & Egan, S. 2022. Differential priority effects impact taxonomy and functionality of host‐associated microbiomes. Molecular Ecology, mec.16336. https://doi.org/10.1111/mec.16336 Nawaz, U., Anees-ur-Rahaman, M., & Saeed, Z. 2025. A Survey of Deep Learning Approaches for Monitoring and Classification of Seagrass. Ocean Science Journal, 60(2), 1– 24. https://doi.org/10.1007/s12601-025-00213-1 Ninio, R., Delean, S., Osborne, K., & Sweatman, H. 2003. Estimating cover of benthic organisms from underwater video images: Variability associated with multiple observers. Marine Ecology Progress Series, 265, 107–116. https://doi.org/10.3354/meps265107 50 Nordlund, L. M., Koch, E. W., Barbier, E. B., Creed, J. C. 2016. Seagrass Ecosystem Services and Their Variability across Genera and Geographical Regions. PLoS ONE 11(10): e0163091. https://doi.org/10.1371/journal.pone.0163091 Norouzzadeh, M. S., Nguyen, A., Kosmala, M., Swanson, A., Palmer, M. S., Packer, C., & Clune, J. 2018. Automatically identifying, counting, and describing wild animals in camera-trap images with deep learning. Proceedings of the National Academy of Sciences, 115(25), E5716–E5725. https://doi.org/10.1073/pnas.1719367115 Perkins, N., Zhang, Z., Monk, J., & Barrett, N. 2022. The annotation approach used for marine imagery impacts the detection of temporal trends in seafloor biota. Ecological Indicators, 140, 109029. https://doi.org/10.1016/J.ECOLIND.2022.109029 Pettorelli, N., Laurance, W. F., O’Brien, T. G., Wegmann, M., Nagendra, H., & Turner, W. 2014. Satellite remote sensing for applied ecologists: Opportunities and challenges. Journal of Applied Ecology, 51(4), 839–848. https://doi.org/10.1111/1365-2664.12261 Pichler, M., & Hartig, F. 2023. Machine learning and deep learning—A review for ecologists. Methods in Ecology and Evolution, 14(4), 994–1016. https://doi.org/10.1111/2041-210X.14061 Piechaud, N., Hunt, C., Culverhouse, P. F., Foster, N. L., & Howell, K. L. 2019. Automated identification of benthic epifauna with computer vision. Marine Ecology Progress Series, 615, 15–30. https://doi.org/10.3354/MEPS12925 Piechaud, N., & Howell, K. L. 2022. Fast and accurate mapping of fine scale abundance of a VME in the deep sea with computer vision. Ecological Informatics, 71, 101786. https://doi.org/10.1016/J.ECOINF.2022.101786 Pieretti, N., & Danovaro, R. 2020. Acoustic indexes for marine biodiversity trends and ecosystem health. Philosophical Transactions of the Royal Society B, 375 (1814), 20190447. http://dx.doi.org/10.1098/rstb.2019.0447 Pinti, J., DeVries, T., Norin, T., Serra-Pompei, C., Proud, R., Siegel, D. A., ... & Visser, A. W. 2023. Model estimates of metazoans' contributions to the biological carbon pump. Biogeosciences, 20 (5), 997-1009. https://doi.org/10.5194/bg-20-997-2023 Porter, T. M., & Hajibabaei, M. 2018. Scaling up: A guide to high-throughput genomic approaches for biodiversity analysis. Molecular Ecology, 27 (2), 313–338. https://doi.org/10.1111/mec.14478 MarineGEO. 2020. Protocol: Seagrass Cover and Density. 2020. Tennenbaum Marine Observatories Network, MarineGEO, Smithsonian Institution. https://marinegeo.github.io/assets/modules/seagrassdensity/marinegeo_protocol_seagrass_density.pd Radford, C. A., Stanley, J. A., Tindle, C. T., et al. 2010. Localised coastal habitats have distinct underwater sound signatures. Marine Ecology Progress Series, 401, 21-29. 10.3354/meps08451 51 Rees, H. C., Maddison, B. C., Middleditch, D. J., Patmore, J. R. M., & Gough, K. C. 2014. REVIEW: The detection of aquatic animal species using environmental DNA – a review of eDNA as a survey tool in ecology. Journal of Applied Ecology, 51 (5), 1450–1459. https://doi.org/10.1111/1365-2664.12306 Rossiter, T., Furey, T., McCarthy, T., and Stengel, D. B. 2020a. "Application of multiplatform, multispectral remote sensors for mapping intertidal macroalgae: A comparative approach." Aquatic ConservationMarine and Freshwater Ecosystems 30 (8):1595-1612. https://doi.org/10.1002/aqc.3357 Rossiter, T.,Furey, T., McCarthy, T., and Stengel, D. B. 2020b. "UAV-mounted hyperspectral mapping of intertidal macroalgae." Estuarine, Coastal and Shelf Science 242:106789. https://doi.org/10.1016/j.ecss.2020.106789 Rubbens, P., Brodie, S., Cordier, T., Destro Barcellos, D., Devos, P., Fernandes-Salvador, J. A., Fincham, J. I., Gomes, A., Handegard, N. O., Howell, K., Jamet, C., Kartveit, K. H., Moustahfid, H., Parcerisas, C., Politikos, D., Sauzède, R., Sokolova, M., Uusitalo, L., Van Den Bulcke, L., … Irisson, J. O. 2023. Machine learning in marine ecology: an overview of techniques and applications. ICES Journal of Marine Science, 80(7), 1829– 1853. https://doi.org/10.1093/ICESJMS/FSAD100 Saccomanno, V.R., Bell, T., Pawlak, C., Stanley, C.K., Cavanaugh, K.C., Hohman, R., Klausmeyer, K.R., Cavanaugh, K., Nickels, A., Hewerdine, W., Garza, C., Fleener, G. and Gleason, M. 2023, Using unoccupied aerial vehicles to map and monitor changes in emergent kelp canopy after an ecological regime shift. Remote Sens Ecol Conserv, 9: 62-75. https://doi.org/10.1002/rse2.295 Sarker, I. H. 2021. Machine Learning: Algorithms, Real-World Applications and Research Directions. SN Computer Science, 2, 160. https://doi.org/10.1007/s42979-021-00592-x Shogren, A. J., Tank, J. L., Andruszkiewicz, E., Olds, B., Mahon, A. R., Jerde, C. L., & Bolster, D. 2017. Controls on eDNA movement in streams: Transport, Retention, and Resuspension. Scientific Reports, 7(1). https://doi.org/10.1038/s41598-017-05223-1 Soviadan, Y. D., Dugenne, M., Drago, L., Biard, T., Trudnowska, E., Lombard, F., Romagnan, J.-B., Jamet, J.- L., Kiko, R., Gorsky, G., Stemmann, L., 2024. Combining in situ and ex situ plankton image data to reconstruct zooplankton (>1 mm) volume and mass distribution in the global ocean, Journal of Plankton Research, 46(5): 461–474. https://doi.org/10.1093/plankt/fbae046 Srivathsan, A., Lee, L., Katoh, K., Hartop, E. A., Kutty, S. N., Wong, J., Yeo, D., & Meier, R. 2021. MinION barcodes: biodiversity discovery and identification by everyone, for everyone. bioRxiv. https://doi.org/10.1101/2021.03.09.434692 Stat, M., Huggett, M. J., Bernasconi, R., DiBattista, J. D., Berry, T. E., Newman, S. J., Harvey, E. S., & Bunce, M. 2017. Ecosystem biomonitoring with eDNA: Metabarcoding across the tree of life in a tropical marine environment. Scientific Reports, 7 (1). https://doi.org/10.1038/s41598-017-12501-5 Sueur, J., Farina, A., Gasc, A., et al. 2014. Acoustic indices for biodiversity assessment and landscape investigation. Acta Acustica united with Acustica, 100 (4), 772-781. 10.3813/AAA.918757 52 Surasinghe, T. D., Singh, K. K., & Frazier, A. E. 2025. Harnessing the full potential of drones for fieldwork. BioScience. https://doi.org/10.1093/biosci/biaf015 Sutherland, W. J. 2006. Ecological census techniques: A handbook (2nd ed.). Cambridge University Press. Tabak, M. A., Norouzzadeh, M. S., Wolfson, D. W., Sweeney, S. J., Vercauteren, K. C., Snow, N. P., Halseth, J. M., Di Salvo, P. A., Lewis, J. S., White, M. D., Teton, B., Beasley, J. C., Schlichting, P. E., Boughton, R. K., Wight, B., Newkirk, E. S., Ivan, J. S., Odell, E. A., Brook, R. K., … Miller, R. S. 2019. Machine learning to classify animal species in camera trap images: Applications in ecology. Methods in Ecology and Evolution, 10(4), 585–590. https://doi.org/10.1111/2041-210X.13120 Taddia, Y., Russo, P., Lovo, S., and Pellegrinelli, A. 2020. Multispectral UAV monitoring of submerged seaweed in shallow water. Applied Geomatics 12 (SUPPL 1):19-34. https://doi.org/10.1007/s12518-01900270-x Tait, L., Bind, J., Charan-Dixon, H.,Hawes, I.,Pirker, J., and Schiel, D. 2019. "Unmanned Aerial Vehicles (UAVs) for Monitoring Macroalgal Biodiversity: Comparison of RGB and Multispectral Imaging Sensors for Biodiversity Assessments." Remote Sensing 11 (19):18. https://doi.org/10.3390/rs11192332 Taylor, P. G. 1996. Reproducibility of ancient DNA sequences from extinct Pleistocene fauna. Molecular Biology and Evolution, 13(1), 283–285. https://doi.org/10.1093/oxfordjournals.molbev.a025566 Thomsen, P. F., & Willerslev, E. (2015). Environmental DNA – An emerging tool in conservation for monitoring past and present biodiversity. Biological Conservation, 183, 4–18. https://doi.org/10.1016/j.biocon.2014.11.019 Turner, W., Spector, S., Gardiner, N., Fladeland, M., Sterling, E., & Steininger, M. 2003. Remote sensing for biodiversity science and conservation. Trends in Ecology & Evolution, 18 (6), 306–314. https://doi.org/10.1016/S0169-5347(03)00070-3 UNEP 2020. Policy and management options. Out of the Blue: The Value of Seagrasses to the Environment and to People. UNEP, Nairobi. https://www.unenvironment.org/resources/report/out-bluevalue-seagrasses-environment-and-people Unsworth, R.K.F., Cullen-Unsworth, L., Jones, B.L.H. & Lilley, R.J. 2022. The planetary role of seagrass conservation. Science 377,609-613(2022). https://doi.org/10.1126/science.abq6923 Valdivia-Carrillo, R., Rocha-Olivares, A., & Cupul-Magaña, A. L. 2021. Integrating eDNA metabarcoding and simultaneous underwater visual census to characterize fish communities in coral reefs. Environmental DNA, 3 (1), 70–83. https://doi.org/10.1111/1755-0998.13375 Van der Wal, D., van Dalen, J., Wielemaker-van den Dool, A., Dijkstra, J. T. and Ysebaert, T. 2014. "Biophysical control of intertidal benthic macroalgae revealed by high-frequency multispectral camera images." Journal of Sea Research 90:111-120. https://doi.org/10.1016/j.seares.2014.03.009 53 Van Parijs, S. M., Clark, C. W., Sousa-Lima, R. S., et al. 2009. Management and research applications of real-time and archival passive acoustic sensors over varying temporal and spatial scales. Marine Ecology Progress Series, 395, 21-36. 10.3354/meps08123 Ventura, D., Grosso, L., Pensa, D., Casoli, E., Mancini, G., Valente, T., Scardi, M., & Rakaj, A. 2023. Coastal benthic habitat mapping and monitoring by integrating aerial and water surface low-cost drones. Frontiers in Marine Science, 9. https://doi.org/10.3389/fmars.2022.1096594 Watson, D. L., & Harvey, E. S. 2007. Behaviour of temperate and sub-tropical reef fishes towards a stationary SCUBA diver. Marine and Freshwater Behaviour and Physiology, 40 (2), 85-103. https://doi.org/10.1080/10236240701393263 Weinstein, B. G. 2018. A computer vision for animal ecology. Journal of Animal Ecology, 87 (3), 533–545. https://doi.org/10.1111/1365-2656.12780 Wendy, D., Le Bris, A., Bajjouk, T., Richier, S., Helias, M., Burel, T., Lennon, M., Guyot, A. and Ar Gall, E. 2022. Seaweed Habitats on the Shore: Characterization through Hyperspectral UAV Imagery and Field Sampling. Remote Sensing 14 (13):3124. https://doi.org/10.3390/rs14133124 Wernberg, T., Krumhansl, K., Filbee-Dexter, K., & Pedersen, M. F. 2019. Status and trends for the world’s kelp forests. In World seas: An environmental evaluation (pp. 57-78). Academic Press. 10.1016/B978-012-805052-1.00003-6 Wiborg, K. F. 1976. Quantitative distribution of Zooplankton in the coast and bank areas of western and northwestern Norway during March-June 1959-1966. Williams, B., Lamont, T.A., Chapuis, L., Harding, H.R., May, E.B., Prasetya, M.E., Seraphim, M.J., Jompa, J., Smith, D.J., Janetski, N. and Radford, A.N. 2022. Enhancing automated analysis of marine soundscapes using ecoacoustic indices and machine learning. Ecological Indicators, 140,108986. https://doi.org/10.1016/j.ecolind.2022.108986 Yates, M. C., Fraser, D. J., & Derry, A. M. 2019. Meta‐analysis supports further refinement of eDNA for monitoring aquatic species‐specific abundance in nature. Environmental DNA, 1 (1), 5–13. https://doi.org/10.1002/edn3.7 Disclosure: AI has been used for language editing and to improve clarity in this deliverable, in the text, figure text and tables. 54 Funded by the European Union under Grant Agreement number 101136748. Views and opinions expressed are, however, those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Agency (REA). Neither the European Union nor the granting authority can be held responsible for them.