Search

In marine ecosystems, most physiological, ecological, or physical processes are size dependent. These include metabolic rates, the uptake of carbon and other nutrients, swimming and sinking velocities, and trophic interactions, which eventually determine the stocks of commercial species, as well as biogeochemical cycles and carbon sequestration. As such, broad-scale observations of plankton size distribution are important indicators of the general functioning and state of pelagic ecosystems under anthropogenic pressures. Here, we present the first global datasets of the Pelagic Size Structure database (PSSdb), generated from plankton imaging devices. This release includes the bulk particle normalized biovolume size spectrum (NBSS) and the bulk particle size distribution (PSD), along with their related parameters (slope, intercept, and R2) measured within the epipelagic layer (0–200 m) by three imaging sensors: the Imaging FlowCytobot (IFCB), the Underwater Vision Profiler (UVP), and benchtop scanners. Collectively, these instruments effectively image organisms and detrital material in the 7–10 000 µm size range. A total of 92 472 IFCB samples, 3068 UVP profiles, and 2411 scans passed our quality control and were standardized to produce consistent instrument-specific size spectra averaged to 1° × 1° latitude and longitude and by year and month. Our instrument-specific datasets span most major ocean basins, except for the IFCB datasets we have ingested, which were exclusively collected in northern latitudes, and cover decadal time periods (2013–2022 for IFCB, 2008–2021 for UVP, and 1996–2022 for scanners), allowing for a further assessment of the pelagic size spectrum in space and time. The datasets that constitute PSSdb's first release are available at https://doi.org/10.5281/zenodo.11050013 (Dugenne et al., 2024b). In addition, future updates to these data products can be accessed at https://doi.org/10.5281/zenodo.7998799.

Automated camera-based sensors are widely used in vessel-based research to monitor plankton and marine particles. However, current methods suffer from the costly and time-consuming requirement of annotating data for fully supervised learning, especially in plankton grouping tasks characterized by long-tailed datasets. In response, we propose a novel self-supervised learning (SSL) framework that significantly reduces reliance on expensive human annotations by leveraging crucial metadata such as water depth and location. The method comprises three major steps: self-supervised training, innovative sampling, and final classification. It identifies key sample subsets from an unlabelled dataset using hierarchical clustering approach and incorporates an innovative balancing representative subsampling strategy that addresses the challenge of dataset imbalance and enhances generalisability across diverse plankton classes. Our approach prioritises discerning representation features observed in images that exhibit correlations with the patterns found in their associated metadata. Furthermore, our method introduce a novel grouping based on visual perspective selection method, enabling the identification of balanced subset views that depart from traditional class-based categorisation. Our experimental results showcase a significant enhancement in image classification accuracy, with a 23% improvement over methods that do not utilise metadata, and attains a macro F1-score of 54% for 10 populated species from a severely long-tailed dataset. This is achieved with a mere 0.3% of the entire dataset used for annotation.

The effect of mesoscale features on the distribution of planktonic organisms are well documented. Yet, the interaction between these spatial features and the temporal scale, which can result in sudden increases of the planktonic biomass, is less known and not described at high resolution. A permanent mesoscale front in the Ligurian Sea (north-western Mediterranean) was repeatedly sampled between January and June 2021 using a SeaExplorer glider equipped with an Underwater Vision Profiler 6 (UVP6), a versatile in situ imager. Both plankton and particle distributions were resolved throughout the spring bloom to assess whether the front was a location of increased zooplankton concentration and whether it constrained particle distribution. Over the 5 months, the glider performed more than 5000 dives and the UVP6 collected 1.1 million images. We focused our analysis on shallow (300 m) transects, which gave a horizontal resolution of 900 m. About 13,000 images of planktonic organisms were retained. Ordination methods applied to particles and plankton concentrations revealed strong temporal variations during the bloom, with a succession of various zooplankton communities. Changes in particle abundance and size could be explained by changes in the plankton community. The front had a strong influence on particle distribution, while the signal was not as clear for plankton, probably because of the relatively small number of imaged organisms. This work confirms the need to sample both plankton and particles at fine scale to understand their interactions, a task for which automated in situ imaging is particularly adapted.

Machine learning covers a large set of algorithms that can be trained to identify patterns in data. Thanks to the increase in the amount of data and computing power available, it has become pervasive across scientific disciplines. We first highlight why machine learning is needed in marine ecology. Then we provide a quick primer on machine learning techniques and vocabulary. We built a database of & SIM;1000 publications that implement such techniques to analyse marine ecology data. For various data types (images, optical spectra, acoustics, omics, geolocations, biogeochemical profiles, and satellite imagery), we present a historical perspective on applications that proved influential, can serve as templates for new work, or represent the diversity of approaches. Then, we illustrate how machine learning can be used to better understand ecological systems, by combining various sources of marine data. Through this coverage of the literature, we demonstrate an increase in the proportion of marine ecology studies that use machine learning, the pervasiveness of images as a data source, the dominance of machine learning for classification-type problems, and a shift towards deep learning for all data types. This overview is meant to guide researchers who wish to apply machine learning methods to their marine datasets.

Plankton imaging instruments are increasingly used to record species occurrences, and they are also able to repeatedly measure ecological traits. However, due to the extensive variety of instruments and the different formats of the data output, there are currently no guidelines and best practices available to store all the relevant data and information in a standard format. Overcoming this challenge will allow for the integration and exchange of these datasets, enabling end users to analyse and visualise them more effectively. To make these data as FAIR (Findable, Accessible, Interoperable, and Reusable) as possible and to share them with international biodiversity data portals, such as the European Marine Observation and Data Network (EMODnet Biology) and the international Ocean Biodiversity Information System (OBIS) Network, like EurOBIS (the European node of OBIS), best practices for the management of plankton imaging data are needed. Thus, the goal of this document is to provide recommendations to plankton imaging users on how to format their data following the OBIS-ENV-DATA format, a Darwin Core based approach to standardise biodiversity data, for submission to these international data portals. These best practices and recommendations are created by an expert working group in the framework of the JERICO-S3 project and by intensive interactions and feedback from the global marine plankton and OBIS community. This document provides (1) an introduction of the current landscape of plankton imaging instruments and the processing of their images, (2) a description of the data standards and format used in biodiversity and guidelines on how to use these, (3) a workflow from instrument to EMODnet Biology, and (4) a discussion on the data management issues identified. With the best practices presented here, it is possible to report a detailed taxonomic characterisation of plankton observations as well as quantitative information that is useful for ecological studies. Thi

The TechOceanS project is developing new remote ocean sensing technology supporting wider ocean measurement and a drive to net zero. The project will deliver 5 new sensor classes for biogeochemistry, biology and ecosystems addressing 10 of 19 EOVs, 31 of 73 subvariables, 6 of 9 MSFD targets together with microplastics and a range of biotoxins and contaminants. It will also develop a new image processing workflow for extracting EOVs (9) and MSFD (6) and litter measurements from images. These innovations concentrate on key capability gaps in ocean observing from non-ship systems with a focus on low-cost per measurement through minimised instrument and deployment costs. This paper gives a brief overview of the technologies, and were possible, because of progress or protection of intellectual property, details of our approaches and early results.

Zooplankton plays a major role in ocean food webs and biogeochemical cycles, and provides major ecosystem services as a main driver of the biological carbon pump and in sustaining fish communities. Zooplankton is also sensitive to its environment and reacts to its changes. To better understand the importance of zooplankton, and to inform prognostic models that try to represent them, spatially-resolved biomass estimates of key plankton taxa are desirable. In this study we predict, for the first time, the global biomass distribution of 19 zooplankton taxa (1-50 mm Equivalent Spherical Diameter) using observations with the Underwater Vision Profiler 5, a quantitative in situ imaging instrument. After classification of 466,872 organisms from more than 3,549 profiles (0-500 m) obtained between 2008 and 2019 throughout the globe, we estimated their individual biovolumes and converted them to biomass using taxa-specific conversion factors. We then associated these biomass estimates with climatologies of environmental variables (temperature, salinity, oxygen, etc.), to build habitat models using boosted regression trees. The results reveal maximal zooplankton biomass values around 60°N and 55°S as well as minimal values around the oceanic gyres. An increased zooplankton biomass is also predicted for the equator. Global integrated biomass (0-500 m) was estimated at 0.403 PgC. It was largely dominated by Copepoda (35.7%, mostly in polar regions), followed by Eumalacostraca (26.6%) Rhizaria (16.4%, mostly in the intertropical convergence zone). The machine learning approach used here is sensitive to the size of the training set and generates reliable predictions for abundant groups such as Copepoda (R2 ≈ 20-66%) but not for rare ones (Ctenophora, Cnidaria, R2 < 5%). Still, this study offers a first protocol to estimate global, spatially resolved zooplankton biomass and community composition from in situ imaging observations of individual organisms. The underlying dataset cover

A major challenge in characterizing plankton communities is the collection, identification and quantification of samples in a time-efficient way. The classical manual microscopy counts are gradually being replaced by high throughput imaging and nucleic acid sequencing. DNA sequencing allows deep taxonomic resolution (including cryptic species) as well as high detection power (detecting rare species), while RNA provides insights on function and potential activity. However, these methods are affected by database limitations, PCR bias, and copy number variability across taxa. Recent developments in high-throughput imaging applied in situ or on collected samples (high-throughput microscopy, Underwater Vision Profiler, FlowCam, ZooScan, etc) has enabled a rapid enumeration of morphologically-distinguished plankton populations, estimates of biovolume/biomass, and provides additional valuable phenotypic information. Although machine learning classifiers generate encouraging results to classify marine plankton images in a time efficient way, there is still a need for large training datasets of manually annotated images. Here we provide workflow examples that couple nucleic acid sequencing with high-throughput imaging for a more complete and robust analysis of microbial communities. We also describe the publicly available and collaborative web application EcoTaxa, which offers tools for the rapid validation of plankton by specialists with the help of automatic recognition algorithms. Finally, we describe how the field is moving with citizen science programs, unmanned autonomous platforms with in situ sensors, and sequencing and digitalization of historical plankton samples.

© The Author(s), 2022. This article is distributed under the terms of the Creative Commons Attribution License. The definitive version was published in Orenstein, E., Ayata, S., Maps, F., Becker, É., Benedetti, F., Biard, T., Garidel‐Thoron, T., Ellen, J., Ferrario, F., Giering, S., Guy‐Haim, T., Hoebeke, L., Iversen, M., Kiørboe, T., Lalonde, J., Lana, A., Laviale, M., Lombard, F., Lorimer, T., Martini, S., Meyer, A., Möller, K.O., Niehoff, B., Ohman, M.D., Pradalier, C., Romagnan, J.-B., Schröder, S.-M., Sonnet, V., Sosik, H.M., Stemmann, L.S., Stock, M., Terbiyik-Kurt, T., Valcárcel-Pérez, N., Vilgrain, L., Wacquet, G., Waite, A.M., & Irisson, J. Machine learning techniques to characterize functional traits of plankton from image data. Limnology and Oceanography, 67(8), (2022): 1647-1669, https://doi.org/10.1002/lno.12101., Plankton imaging systems supported by automated classification and analysis have improved ecologists' ability to observe aquatic ecosystems. Today, we are on the cusp of reliably tracking plankton populations with a suite of lab-based and in situ tools, collecting imaging data at unprecedentedly fine spatial and temporal scales. But these data have potential well beyond examining the abundances of different taxa; the individual images themselves contain a wealth of information on functional traits. Here, we outline traits that could be measured from image data, suggest machine learning and computer vision approaches to extract functional trait information from the images, and discuss promising avenues for novel studies. The approaches we discuss are data agnostic and are broadly applicable to imagery of other aquatic or terrestrial organisms., SDA acknowledges funding from CNRS for her sabbatical in 2018–2020. Additional support was provided by the Institut des Sciences du Calcul et des Données (ISCD) of Sorbonne Université (SU) through the support of the sponsored junior team FORMAL (From ObseRving to Modeling oceAn Life), especially through the post-doctoral contract of EO. JOI acknowledges funding from the Belmont Forum, grant ANR-18-BELM-0003-01. French co-authors also wish to thank public taxpayers who fund their salaries. This work is a contribution to the scientific program of Québec Océan and the Takuvik Joint International Laboratory (UMI3376; CNRS - Université Laval). FM was supported by an NSERC Discovery Grant (RGPIN-2014-05433). MS is supported by the Research Foundation - Flanders (FWO17/PDO/067). FB received support from ETH Zürich. MDO is supported by the Gordon and Betty Moore Foundation and the U.S. National Science Foundation. ECB is supported by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) under the grant agreement no. 88882.438735/2019-01. TB is supported by the French National Research Agency (ANR-19-CE01-0006). NVP is supported by the Spanish State Research Agency, Ministry of Science and Innovation (PTA2016-12822-I). FL is supported by the Institut Universitaire de France (IUF). HMS was supported by the Simons Foundation (561126) and the U.S. National Science Foundation (CCF-1539256, OCE-1655686). Emily Peacock is gratefully acknowledged for expert annotation of IFCB images. LS was supported by the Chair VISION from CNRS/Sorbonne Université.

In the Gulf of Lions, small pelagic fish have shown reduced body size and body condition after 2007 that would result from changes in zooplankton community. We therefore examined zooplankton density, body size, and taxonomic composition at the closest long-term monitoring station (1995–2019): the coastal Point-B. To cover a broader spectrum of zooplankton community, samples obtained from two nets, the WP2 (200 µm mesh size) and the Regent (690 µm), were analysed with the imaging Zooscan method. One important result was the high stability through time of the zooplankton community. No long-term monotonous trends in density, size, and taxonomic composition were detected. Interannual variations in zooplankton size and density were not significantly correlated to any environmental variable, suggesting the possible importance of biotic interactions. Still, an increase in temperature was followed by a sharp decrease of zooplankton density in 2015, after which only gelatinous groups recovered. No change in the zooplankton community was detected around 2007 to support bottom-up control on small pelagic fish. Whether this derives from different local processes between the Gulf of Lions and the Ligurian Sea cannot be excluded, highlighting the need for simultaneous monitoring of different ecosystem compartments to fully understand the impact of climate change.

As the basis of oceanic food webs and a key component of the biological carbon pump, planktonic organisms play major roles in the oceans. Their study benefited from the development of in situ imaging instruments, which provide higher spatio-temporal resolution than previous tools. But these instruments collect huge quantities of images, the vast majority of which are of marine snow particles or imaging artifacts. Among them, the In Situ Ichthyoplankton Imaging System (ISIIS) samples the largest water volumes (> 100 L s-1) and thus produces particularly large datasets. To extract manageable amounts of ecological information from in situ images, we propose to focus on planktonic organisms early in the data processing pipeline: at the segmentation stage. We compared three segmentation methods, particularly for smaller targets, in which plankton represents less than 1% of the objects: (i) a traditional thresholding over the background, (ii) an object detector based on maximally stable extremal regions (MSER), and (iii) a content-aware object detector, based on a Convolutional Neural Network (CNN). These methods were assessed on a subset of ISIIS data collected in the Mediterranean Sea, from which a ground truth dataset of > 3,000 manually delineated organisms is extracted. The naive thresholding method captured 97.3% of those but produced ~340,000 segments, 99.1% of which were therefore not plankton (i.e. recall = 97.3%, precision = 0.9%). Combining thresholding with a CNN missed a few more planktonic organisms (recall = 91.8%) but the number of segments decreased 18-fold (precision increased to 16.3%). The MSER detector produced four times fewer segments than thresholding (precision = 3.5%), missed more organisms (recall = 85.4%), but was considerably faster. Because naive thresholding produces ~525,000 objects from 1 minute of ISIIS deployment, the more advanced segmentation methods significantly improve ISIIS data handling and ease the subsequent taxonomic classifica

Plankton imaging systems supported by automated classification and analysis have improved ecologists' ability to observe aquatic ecosystems. Today, we are on the cusp of reliably tracking plankton populations with a suite of lab-based and in situ tools, collecting imaging data at unprecedentedly fine spatial and temporal scales. But these data have potential well beyond examining the abundances of different taxa; the individual images themselves contain a wealth of information on functional traits. Here, we outline traits that could be measured from image data, suggest machine learning and computer vision approaches to extract functional trait information from the images, and discuss promising avenues for novel studies. The approaches we discuss are data agnostic and are broadly applicable to imagery of other aquatic or terrestrial organisms.

Marine particles of different nature are found throughout the global ocean. The term "marine particles" describes detritus aggregates and fecal pellets as well as bacterioplankton, phytoplankton, zooplankton and nekton. Here, we present a global particle size distribution dataset obtained with several Underwater Vision Profiler 5 (UVP5) camera systems. Overall, within the 64 mu m to about 50 mm size range covered by the UVP5, detrital particles are the most abundant component of all marine particles; thus, measurements of the particle size distribution with the UVP5 can yield important information on detrital particle dynamics. During deployment, which is possible down to 6000 m depth, the UVP5 images a volume of about 1 L at a frequency of 6 to 20 Hz. Each image is segmented in real time, and size measurements of particles are automatically stored. All UVP5 units used to generate the dataset presented here were inter-calibrated using a UVP5 high-definition unit as reference. Our consistent particle size distribution dataset contains 8805 vertical profiles collected between 19 June 2008 and 23 November 2020. All major ocean basins, as well as the Mediterranean Sea and the Baltic Sea, were sampled. A total of 19 % of all profiles had a maximum sampling depth shallower than 200 dbar, 38 % sampled at least the upper 1000 dbar depth range and 11 % went down to at least 3000 dbar depth. First analysis of the particle size distribution dataset shows that particle abundance is found to be high at high latitudes and in coastal areas where surface productivity or continental inputs are elevated. The lowest values are found in the deep ocean and in the oceanic gyres. Our dataset should be valuable for more in-depth studies that focus on the analysis of regional, temporal and global patterns of particle size distribution and flux as well as for the development and adjustment of regional and global biogeochemical models. The marine particle size distribution dataset (Kiko et al

Plankton imaging systems supported by automated classification and analysis have improved ecologists' ability to observe aquatic ecosystems. Today, we are on the cusp of reliably tracking plankton populations with a suite of lab-based and in situ tools, collecting imaging data at unprecedentedly fine spatial and temporal scales. But these data have potential well beyond examining the abundances of different taxa; the individual images themselves contain a wealth of information on functional traits. Here, we outline traits that could be measured from image data, suggest machine learning and computer vision approaches to extract functional trait information from the images, and discuss promising avenues for novel studies. The approaches we discuss are data agnostic and are broadly applicable to imagery of other aquatic or terrestrial organisms.