Conference Agenda

Sustained observations of plankton are critical to understand how environmental drivers and biological interactions shape trophic structure, food web dynamics, and ultimately the distribution and abundance of living marine resources. This study examined image-based marine plankton from surveys in south Florida coastal and shelf waters collected by the Southeast U.S. Marine Biodiversity Observation Network project. A goal is to characterize biogeographic distributions and phenology of phytoplankton and zooplankton. Ten field campaigns aboard the R/V Walton Smith (U. Miami) and R/V Hogarth (Florida Institute of Oceanography) were carried out every six weeks between December of 2022 and July of 2024. Plankton imagery were collected in depth profiles at ~70–90 stations with a Continuous Particle Imaging and Classification System (CPICS) mounted on a CTD rosette. Image segments (Regions of Interest; 11,424) were classified and quantified to estimate concentrations of diatoms, Trichodesmium spp, dinoflagellates, copepods, Rhizaria spp, polychaetes, pteropods, chaetognaths, ostracods, larvaceans, echinoderms, and gelatinous species as organisms per cubic meter. Plankton occurrences were matched to satellite-derived seascapes, which are dynamic biogeographic classifications of water masses based on multiple remote sensing data sources. Four seascape classes dominated the areas sampled: Tropical/Subtropical Transition (TST: 7%), Tropical Seas (TS: 26%), Warm, Blooms, High Nutrients (WBHN: 43%), and Hypersaline Eutrophic (HE: 15%). Results show differentiation in plankton distributions between seascape categories and seasonal variability in species composition within seascapes. Pteropods were notably higher in the oligotrophic TST class. Gelatinous species concentrations were typically lower in TST and other low nutrient, low plankton biomass classes. Copepod concentrations exhibited strong seasonality with abundances up to three orders of magnitude higher during summer months versus winter months. Satellite seascapes can provide a biogeographic framework to evaluate how plankton communities change over time and space, and drive ecosystem dynamics and marine living resources in the Florida Keys and shelf areas.

The Mediterranean marine ecosystems are tremendously impacted by climate change, leading to profound consequences on structure and functioning of living communities and biodiversity loss, starting from primary producers (i.e., phytoplankton). Using satellite-derived surface chlorophyll-a concentration (as a proxy for phytoplankton concentration), various studies have attempted to describe the general seasonal patterns of such organisms at the sea surface. However, the inter-annual variability of the resulting seascape has not been fully addressed, much less along the water column and in relation with climate. Within the ESA 4DMED-Sea project, we explored 26 years (1998-2023) of daily satellite-derived chlorophyll-a images at 4 km of spatial resolution and assess the interannual variability of the Mediterranean pelagic seascape based on the phenology of phytoplankton. By applying a clustering technique, we first confirmed the existence of seven major ecoregions in the Mediterranean Sea, though with different average chlorophyll seasonal cycles for coastal regions. The analysis also shows strong inter-annual variability among clusters, with some regions that are more stable than others, and a cluster with significantly reducing extension through years. We then applied the same clustering methodology up to 150m depth over 6 years (2016-2021) of daily 4D chlorophyll values at 4km resolution. The Mediterranean seascape becomes simpler with depth, revealing higher chlorophyll homogeneity. Finally, we investigated the drivers of observed changes in highest variability areas, focusing on their relationship with temperature trends, circulation patterns, and climatic indices.

Red tides, high-biomass phytoplankton blooms, are noteworthy phenomena and a major source of concern worldwide. Red tides can be harmful to marine fauna due to phycotoxins, mechanical damage, release of ammonia, and/or anoxia. During a red tide, phytoplankton biomass is orders of magnitude higher than during an open ocean bloom, and seawater optical variability is dominated by changes in phytoplankton abundance and composition. As the phytoplankton community is typically dominated by a single taxon, the absorption coefficient of a red tide sample can be merely approximated by the absorption coefficients of pure seawater and of the dominant phytoplankter. Identification of the causative species could therefore be feasible from remote sensing providing that the non-water absorption coefficient can be accurately inversed from the remote-sensing reflectance, and the information contained in the absorption spectrum unambiguously related to the bloom-forming taxon. Here, the objective was to explore the second, absorption-related issue. A unique dataset of 164 hyperspectral absorption measurements was obtained from monospecific culture data, compiling published and new measurements. Using spectral clustering techniques, we assessed the level of taxonomic information amenable to absorption-based analysis. The absorption-based clustering was consistent with phytoplankton taxonomical classes, thus demonstrating the potential of hyperspectral remote sensing to identify the red tide causative phytoplankter at class level, in the absence of field information. In particular, the ability to distinguish dinoflagellates from diatoms, prymnesiophytes, and raphidophytes was demonstrated. This is an important result because Dinophyceae are known to be notoriously challenging to discriminate from other phytoplankton classes. Moreover, several optical clusters were obtained for dinoflagellates, consistently with their pigment composition (e.g. peridin-bearing vs. fucoxanthin-bearing species). Using a single peridinin absorption type as an optical signature of dinoflagellates raises the risk of overlooking important HAB species when trying to identify phytoplankton types from optical observations.

Phytoplankton are dominant marine primary producers, and the structure of phytoplankton assemblages controls food web dynamics, and ecosystem resilience and services. Hence, understanding the environmental determinants that shape phytoplankton assemblage structure is imperative, especially in complex marine domains. This study aimed to assess spatial-temporal variability patterns of phytoplankton assemblages off southwestern Iberia, identify the underlying environmental drivers and predictors, and evaluate the performance of algorithms used to derive phytoplankton composition from space. Physico-chemical variables were acquired from different sources (e.g., satellite remote sensing, models, in situ observations), covering the mixed layer at three stations, along a coastal-offshore transect, over two years (July 2012-July 2014). Phytoplankton composition, derived from microscopic analysis, and specific diagnostic pigment composition (CHEMical TAXonomy analysis, CHEMTAX), was compared with satellite-based algorithms that retrieve phytoplankton size classes and/or specific taxa from abundance-based models or specific spectral features (Copernicus-GlobColour processor and based on inputs from the European Space Agency Climate Change Initiative). Higher mean photosynthetically available radiation in the mixed layer was observed during spring and early summer, whereas increased nutrient supply occurred during winter, and summer periods. The annual cycles of chlorophyll-a concentration ranged from bimodal (coastal) to unimodal further offshore. Phytoplankton abundance was dominated by pico-sized cyanobacteria, but for biomass, diatoms, prasinophyceans and dinoflagellates dominated. The assemblage structure differed between stations and seasons. Upwelling index, nitrite and suspended particulate matter concentrations emerged as the variables that best explained the variability derived from microscopy, whereas considering CHEMTAX-based results, only silicate concentration was identified as a relevant variable. The comparison of data derived from in situ observations and satellite-based algorithms identified the abundance-based model as the best performer for deriving nanophytoplankton, diatoms and dinoflagellates, for most performance metrics. Spectral-based algorithms performed better for retrieving pico- and microphytoplankton. Further calibration and validation are required to refine algorithms at regional scales.

Remote sensing techniques have been employed to elucidate phytoplankton community structure by analyzing spectral data from space, especially when coupled with in situ measurements of photosynthetic pigments. In this study, we introduce a novel ocean color algorithm designed to estimate the relative cell abundance of seven phytoplankton groups and their respective contributions to total chlorophyll a (Chl a) on a global scale. Leveraging machine learning, our algorithm utilizes remotely sensed parameters (including reflectance, backscattering, and attenuation coefficients at various wavelengths, as well as temperature and Chl a) in conjunction with an omics-based biomarker derived from Tara Oceans data. This biomarker targets a single-copy gene called psbO, encoding a component of the photosynthetic machinery present across all phytoplankton, spanning both prokaryotes and eukaryotes. This research delivers a comprehensive global dataset detailing the relative cell abundances of the seven phytoplankton groups and their impacts on total Chl a. These data types offer distinct insights: Chl a serves as a biomass proxy crucial for understanding energy and matter fluxes in ecological and biogeochemical processes, while cell abundance provides crucial information on community assembly processes. Moreover, our methodology allows comparisons with existing approaches, such as pigment-based methods. This integration underscores the potential of remote sensing observations as powerful tools for gathering Essential Biodiversity Variables (EBVs). By expanding our understanding of phytoplankton dynamics on a global scale, this study advances ecological research on the link between biodiversity and ecosystem functions.

Marine ecosystems are supported almost entirely by primary production provided by phytoplankton, which globally perform 50% of the photosynthesis on our planet. Phytoplankton also fix atmospheric carbon dioxide through photosynthesis and transport it to the ocean interior, which is essential for climate regulation. Their ability to capture carbon, however, depends on the availability of scarce nutrients like iron. Climate change and human activities are shifting the oceanic distribution and bioavailability of iron, yet the responses of different phytoplankton groups and overall ocean productivity to these changes remain poorly understood. In this study, we aimed to describe on a global scale the spatio-variability of phytoplankton’s iron nutritional status by integrating omics and satellite observations. First, we examined abundance and expression profiles of genes and transcripts linked to iron-responsive photosynthetic electron transport in metagenomes (n=690) and metatranscriptomes (n=709) from 127 Tara Oceans’ stations. Finally, we trained a random forest model with monthly 4-km satellite observations (Chla, SST, iPAR, Fluorescence in the red), 1-degree resolution monthly composites of biogeochemical model outputs (iron and copper), and WOA2018 compiled measurements (e.g., SiO2, PO43-, and NO3-), to predict the global distribution of these genes/transcripts ratios. The estimated values strongly correlated with in-situ values for metaG (r=0.76 for Fld/Fld+Fd; r=0.52 for PC/CytoC6+PC), and for metaT (r=0.66 for Fld/Fld+Fd; r=0.63 for PC/CytoC6+PC). The mean relative absolute error rates were relatively low for the metaG ratios (MRAE = 14–17%) when compared to the metaT ratios (MRAE = 30–56%). Although we highlight the need for increasing in-situ observations, our workflow provides the foundation for linking genomics and remote sensing to monitor phytoplankton iron nutritional status in the vast global ocean.

Fronts – the interface between water masses – are hotspots for rich and diverse marine life, influencing the foraging distribution of many megafauna. We have analysed a long time-series of Earth observation (EO) data using novel algorithms to characterise the distribution and dynamic of ocean fronts, and used these to investigate links to biodiversity hotspots and to explore key drivers for changes in fronts and these relationships.

FRONTWARD (Fronts for Marine Wildlife Assessment for Renewable Developments) aims to provide evidence to justify the inclusion of frontal locations in marine spatial planning, most pressingly for zones for offshore windfarms. Biodiversity hotspots are identified using a biodiversity index, created using an unprecedented collation of UK at-sea observations of seabirds, fish and cetaceans spanning several decades (1980s-2020s). Generalised additive models (GAMs) reveal the spatial influence of fronts on biodiversity, and provide predictions of biodiversity based on EO-detected front maps. The outcomes from this project will feed into the evidence base for marine conservation, and decisions on siting of future offshore renewable energy projects.

New tools such as optical satellite imagery analysis powered by advances in artificial intelligence, have potential to provide additional broad-scale and near real-time capacities for survey and monitoring marine mammals. While multiple studies demonstrated that large cetaceans are detectable in sub-meter satellite imagery, this work aimed to tackle multiple fundamental challenges of the application such as reaching high fidelity automated detection, performing broad-scale deployment in challenging ocean environments, and testing the shovel-readiness of satellite imagery for conservation monitoring needs with the use case of fishing gear entanglement risk in the California Dungeness commercial crab fishery. Statistical analysis of regional satellite imagery allowed the development of a deep-learning-based detection framework capable of optimizing learning from an originally small dataset. The best architecture generally achieved satisfying performance with an average balanced accuracy reaching up to 99.90% for gray whales. It was also demonstrated that gray-scale imagery can be used to perform detection with a high accuracy of 87.05%, opening a capability to monitor larger spatio-temporal ranges than previously thought. Broad-scale deployment of best-in class machine-learning models over an unprecedented amount of satellite imagery (> 650,000 km2), from December 2009 to March 2023, covering multiple times the entire California coast, resulted in the detection and construction of a satellite imagery database with over 3500 gray whales and 1500 humpback whales as well as opportunistic detections of blue and fin whales. It furthermore provided meaningful data points on the migration routes of gray whales within the Southern California Bight. Through a collaboration between UPC, The Nature Conservancy (TNC) and NOAA, the developed system is currently being deployed in the Channel Islands Sanctuary to inform a vessel speed reduction program and to potentially influence long-term shipping lane design. It is our hope that this approach can be replicated or adapted for other use cases around the world to support conservation policies.