Machine learning techniques to characterize functional traits of plankton from image data.

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.

Laminarin is a major molecule in the marine carbon cycle.

Marine microalgae sequester as much CO2 into carbohydrates as terrestrial plants. Polymeric carbohydrates (i.e., glycans) provide carbon for heterotrophic organisms and constitute a carbon sink in the global oceans. The quantitative contributions of different algal glycans to cycling and sequestration of carbon remain unknown, partly because of the analytical challenge to quantify glycans in complex biological matrices. Here, we quantified a glycan structural type using a recently developed biocatalytic strategy, which involves laminarinase enzymes that specifically cleave the algal glycan laminarin into readily analyzable fragments. We measured laminarin along transects in the Arctic, Atlantic, and Pacific oceans and during three time series in the North Sea. These data revealed a median of 26 ± 17% laminarin within the particulate organic carbon pool. The observed correlation between chlorophyll and laminarin suggests an annual production of algal laminarin of 12 ± 8 gigatons: that is, approximately three times the annual atmospheric carbon dioxide increase by fossil fuel burning. Moreover, our data revealed that laminarin accounted for up to 50% of organic carbon in sinking diatom-containing particles, thus substantially contributing to carbon export from surface waters. Spatially and temporally variable laminarin concentrations in the sunlit ocean are driven by light availability. Collectively, these observations highlight the prominent ecological role and biogeochemical function of laminarin in oceanic carbon export and energy flow to higher trophic levels.

Environmental Drivers of Free-Living vs. Particle-Attached Bacterial Community Composition in the Mauritania Upwelling System.

Saharan dust input and seasonal upwelling along North-West Africa provide a model system for studying microbial processes related to the export and recycling of nutrients. This study offers the first molecular characterization of prokaryotic particle-attached (PA; >3.0 µm) and free-living (FL; 0.2-3.0 µm) players in this important ecosystem during August 2016. Environmental drivers for alpha-diversity, bacterial community composition, and differences between FL and PA fractions were identified. The ultra-oligotrophic waters off Senegal were dominated by Cyanobacteria while higher relative abundances of Alphaproteobacteria, Bacteroidetes, Verrucomicrobia, and Planctomycetes (known particle-degraders) occurred in the upwelling area. Temperature, proxy for different water masses, was the best predictor for changes in FL communities. PA community variation was best explained by temperature and ammonium. Bray Curtis dissimilarities between FL and PA were generally very high and correlated with temperature and salinity in surface waters. Greatest similarities between FL and PA occurred at the deep chlorophyll maximum, where bacterial substrate availability was likely highest. This indicates that environmental drivers do not only influence changes among FL and PA communities but also differences between them. This could provide an explanation for contradicting results obtained by different studies regarding the dissimilarity/similarity between FL and PA communities and their biogeochemical functions.