• European Marine Science
• Publications close
• Other research products close
• European Commission close
• Swiss National Science Foundation close

Abstract 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 ∼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. 80 7

This work was supported by the High Meadows Environmental Institute at Princeton University and the NSF's Southern Ocean Carbon and Climate Observations and Modeling (SOCCOM) Project under the NSF Award PLR-1425989. F.A.H. was supported by NASA Grant 80NSSC19K1115 and by the European Union (ERC, VERTEXSO, 101041743). G.A.M was supported under SOCCOM and UKRI Grant MR/W013835/1. T.L.F was supported by Swiss National Science Foundation (Grant P00P2_198897) and the Swiss National Supercomputing Centre. N.L was supported by the European Union's Horizon 2020 research and innovation program under Grant 820989 (project COMFORT) and no. 862923 (project AtlantECO) as well as the Bretscher Funds. Every austral spring when Antarctic sea ice melts, favorable growing conditions lead to an intense phytoplankton bloom, which supports much of the local marine ecosystem. Recent studies have found that Antarctic sea ice is predictable several years in advance, suggesting that the spring bloom might exhibit similar predictability. Using a suite of perfect model predictability experiments, we find that November net primary production (NPP) is potentially predictable 7 to 10 years in advance in many Southern Ocean regions. Sea ice extent predictability peaks in late winter, followed by absorbed shortwave radiation and NPP with a 2 to 3 months lag. This seasonal progression of predictability supports our hypothesis that sea ice and light limitation control the inherent predictability of the spring bloom. Our results suggest skillful interannual predictions of NPP may be achievable, with implications for managing fisheries and the marine ecosystem, and guiding conservation policy in the Southern Ocean. Peer reviewed

Abstract Background The development of nanoscale secondary ion mass spectrometry (NanoSIMS) has revolutionized the study of biological tissues by enabling, e.g., the visualization and quantification of metabolic processes at subcellular length scales. However, the associated sample preparation methods all result in some degree of tissue morphology distortion and loss of soluble compounds. To overcome these limitations an entirely cryogenic sample preparation and imaging workflow is required. Results Here, we report the development of a CryoNanoSIMS instrument that can perform isotope imaging of both positive and negative secondary ions from flat block-face surfaces of vitrified biological tissues with a mass- and image resolution comparable to that of a conventional NanoSIMS. This capability is illustrated with nitrogen isotope as well as trace element mapping of freshwater hydrozoan Green Hydra tissue following uptake of 15N-enriched ammonium. Conclusion With a cryo-workflow that includes vitrification by high pressure freezing, cryo-planing of the sample surface, and cryo-SEM imaging, the CryoNanoSIMS enables correlative ultrastructure and isotopic or elemental imaging of biological tissues in their most pristine post-mortem state. This opens new horizons in the study of fundamental processes at the tissue- and (sub)cellular level. Teaser CryoNanoSIMS: subcellular mapping of chemical and isotopic compositions of biological tissues in their most pristine post-mortem state.

Snow plays a crucial role in the heat transfer between the ocean and atmosphere in sea ice due to its insulating properties. However, wind-induced transport causes the snow distribution to be inhomogeneous, as snow forms dunes and accumulates mostly around pressure ridges and, leading to a heterogeneous underlying ice growth and melt. While models can help to understand the complex interactions of snow and sea ice, there is currently no 3D snow cover model for sea ice that considers detailed snow cover properties. This study presents the first application of the 3D-snow cover-atmosphere model ALPINE3D with the drifting snow module to Arctic sea ice. The model was calibrated and validated with measurements from the MOSAiC expedition. Wind fields used by the snow drift routine were generated with OpenFOAM which was forced by observations. A sensitivity analysis showed the impact of an increased fluid threshold on snow redistribution. The model performed well in simulating snow transport and mass fluxes, but underestimated erosion and poorly reproduced dune formation due to the missing dynamic mesh. The density was partially reproduced very well by the model, but uncertainties still exist in some cases. Comparing the surface snow density results with 1-D SNOWPACK simulations, ALPINE3D produced smaller differences but larger temporal variation in between setups. The study also investigated details of deposition and erosion using cross sections, showing good agreements of snow height differences between model and observations and revealing spatially high-resolution parameters such as age of deposited snow, density, and thermal conductivity.

AbstractSparse and spatiotemporally highly uneven sampling efforts pose major challenges to obtaining accurate species and biodiversity distributions. Here, we demonstrate how limited surveys can be integrated with global models to uncover hotspots and distributions of marine biodiversity. We test the skill of recent and advanced species distribution model setups to predict the global biodiversity of >560 phytoplankton species from 183,000 samples. Recent setups attain quasi-null skill, while models optimized for sparse data explain up to 91% of directly observed species richness variations. Using a refined spatial cross-validation approach to address data sparsity at multiple temporal resolutions we find that background choices are the most critical step. Predictor variables selected from broad sets of drivers and tuned for each species individually improve the models’ ability in identifying richness hotspots and latitude gradients. Optimal setups identify tropical hotspots, while common ones lead to polar hotspots disjunct from general marine diversity. Our results show that unless great care is taken to validate models, conservation areas in the ocean may be misplaced. Yet a game-changing advance in mapping diversity can be achieved by addressing data-sparse conditions that prevail for >80% of extant marine species.Authorship statementAll authors designed the research and contributed to the writing. D.R. designed the multiscale validation and predictor selection methods, developed the figures with input by M.V. and N.E.Z., performed research, and wrote the first draft.

The start-up, build-up and demise of cold-water coral mounds are governed by environmental changes at global, regional and local scales. Whilst the formation of cold-water coral mounds across the globe is widely documented to follow interglacial-glacial cycles, less is known about their response to local environmental fluctuations during short time intervals. This study investigates the local variations in coral mound build-up along Brittlestar Ridge I (East Melilla Coral Province, Southeast Alboran Sea) by comparing three on-mound gravity cores collected ∼1 km apart, together with five previously described on-mound records, along a longitudinal transect on the ridge crest. Radiocarbon foraminiferal dating associated to U-series coral dating allowed to correlate the different records and to estimate mound aggradation rates, whilst grain-size analysis provided information on bottom current velocities. Prior to a rapid period of coral mound build-up initiated at ∼14.75 ka BP, the three cores present an occurrence of cm-thick bryozoan-dominated intervals nearly entirely consisting of the erect cheleistome Buskea dichotoma. Offsets between benthic foraminiferal and coral ages suggest that older dead allochtonous benthic foraminifera are possibly eroded from neighbouring settings, transported and deposited within the coral framework. In contrast, younger benthic foraminifera would develop on dead coral framework during periods of reef stagnation. The comparison of all cores indicates that mound build-up along the ridge did not follow the same timing during the last ∼16 kyr and that mound aggradation was marked by a staggered dynamic. Both local differences in bottom current velocities and patchiness of other key environmental actors (e.g. substrate availability) are suspected to drive the observed staggered mound build-up. Cold-water coral mound build-up shows important differences during Greenland Interstadial 1 (i.e. the Bølling-Allerød) and the Holocene, hence examplifying how local environmental variability may overprint global and regional climate variability over short time periods. Marine Geology, 457 ISSN:0025-3227

Special attention has been given to chromium (Cr) as a paleoproxy tracing redox cycling throughout Earth's history, due to differences in the solubility of its primary redox species at Earth's surface (Cr(III) and Cr(VI)) and isotope fractionation associated with their interconversion. In turn, chromium's paleoproxy potential has motivated studies of the modern ocean to better understand which processes drive its cycling and to constrain their impact on the Cr isotope composition (δ53Cr) of seawater. Here, we present total dissolved seawater Cr concentrations and δ53Cr along the GEOTRACES GP13 section. This section is a zonal transect extending from Australia in the subtropical southwest Pacific Ocean. Surface signals of local biological Cr cycling are minimal, in agreement with distributions of dissolved major nutrients as well as biologically-controlled trace metals in this low productivity, oligotrophic environment. Depth profiles have Cr concentration minima in surface waters and maxima at depth, and are largely shaped by the advection of nutrient- and Cr-rich subsurface waters rather than vertically-driven processes. Samples close to the sediment–water interface indicate important benthic Cr fluxes across the section. The GP13 transect crosses the hydrothermally-active Kermadec Arc. Hydrothermal fluids (consisting of <15% background seawater) were collected from three venting sites at the Brothers Volcano (along the Kermadec Arc). These fluids yielded near-crustal δ53Cr values (−0.17 to +0.08‰) and elevated [Cr] (7.5–23 nmol kg−1, hydrothermal endmember [Cr] ≈ 8–27 nmol kg−1), indicating that the Kermadec Arc may be an isotopically light Cr source. Dissolved [Fe] enrichments have been reported previously in deep waters (∼1600–3000 m) along the GP13 transect, east of the Kermadec Arc. These same waters show elevated [Cr] compared to Circumpolar Deep Water ([Cr] = 3.88 ± 0.11, δ53Cr = 0.89 ± 0.08, n = 32), with an average [Cr] accumulation of 0.71 ± 0.11 nmol kg−1 (1 SD), and an estimated δ53Cr of +0.46 ± 0.30‰ (2 SD, n = 9) for the accumulated Cr. Comparing high-temperature vent and neutrally buoyant plume data, hydrothermal-sourced Cr is likely negligable compared to Cr contributions from other processes (benthic fluxes, release from particles), and the advection of more Cr-rich Pacific Deep Water. It is unlikely that hydrothermal vents would be a major contributor within the regional or global biogeochemical Cr cycle, even if hydrothermal fluxes change by orders of magnitude, and therefore δ53Cr trends in the paleorecord may be attributable, at least in part, to major changes in other controls on Cr (e.g. widespread anoxia). Geochimica et Cosmochimica Acta, 342 ISSN:1872-9533 ISSN:0016-7037