Planktonic foraminiferal species identification is central to many paleoceanographic studies, from selecting species for geochemical research to elucidating the biotic dynamics of microfossil communities relevant to physical oceanographic processes and interconnected phenomena such as climate change. However, few resources exist to train students in the difficult task of discerning amongst closely related species, resulting in diverging taxonomic schools that differ in species concepts and boundaries. This problem is exacerbated by the limited number of taxonomic experts. Here we document our initial progress toward removing these confounding and/or rate-limiting factors by generating the first extensive image library of modern planktonic foraminifera, providing digital taxonomic training tools and resources, and automating species-level taxonomic identification of planktonic foraminifera via machine learning using convolution neural networks. Experts identified 34,640 images of modern (extant) planktonic foraminifera to the species level. These images are served as species exemplars through the online portal Endless Forams (endlessforams.org) and a taxonomic training portal hosted on the citizen science platform Zooniverse (zooniverse.org/projects/ahsiang/ endless-forams/). A supervised machine learning classifier was then trained with~27,000 images of these identified planktonic foraminifera. The best-performing model provided the correct species name for an image in the validation set 87.4% of the time and included the correct name in its top three guesses 97.7% of the time. Together, these resources provide a rigorous set of training tools in modern planktonic foraminiferal taxonomy and a means of rapidly generating assemblage data via machine learning in future studies for applications such as paleotemperature reconstruction.

The production of organic carbon in the ocean's surface and its subsequent downward export transfers carbon dioxide to the deep ocean. This CO 2 drawdown is countered by the biological precipitation of carbonate, followed by sinking of particulate inorganic carbon, which is a source of carbon dioxide to the surface ocean, and hence the atmosphere over 100-1,000 year timescales 1 . The net transfer of CO 2 to the deep ocean is therefore dependent on the relative amount of organic and inorganic carbon in sinking particles 2 . In the Southern Ocean, iron fertilization has been shown to increase the export of organic carbon 3-5 , but it is unclear to what degree this e ect is compensated by the export of inorganic carbon. Here we assess the composition of sinking particles collected from sediment traps located in the Polar Frontal Zone of the Southern Ocean. We find that in high-nutrient, low-chlorophyll regions that are characterized by naturally high iron concentrations, fluxes of both organic and inorganic carbon are higher than in regions with no iron fertilization. However, the excess flux of inorganic carbon is greater than that of organic carbon. We estimate that the production and flux of carbonate in naturally iron-fertilized waters reduces the overall amount of CO 2 transferred to the deep ocean by 6-32%, compared to 1-4% at the non-fertilized site. We suggest that an increased export of organic carbon, stimulated by iron availability in the glacial sub-Antarctic oceans, may have been accompanied by a strengthened carbonate counter pump.The biological carbon pump is the downward flux of particulate organic carbon (POC) from the surface to the deep ocean 6 . The fraction of settling POC that is not remineralized in the winter mixed layer sinks to depth, driving a reduction in surface ocean partial pressure of carbon dioxide (pCO 2 ) that is compensated by oceanic uptake of atmospheric CO 2 (ref. 7). The iron hypothesis 8 suggests that increased iron supply to the nutrient-rich, but irondeficient Southern Ocean, contributed towards the termination of glacial periods by enhancing phytoplankton growth and the biological carbon pump. Recent studies of natural systems 3,4 and artificial 5 iron fertilization experiments support this idea by demonstrating enhanced POC flux well below the mixed layer into the deep ocean 3-5 .Counteracting the organic carbon pump in terms of its influence on air-sea CO 2 exchange is the carbonate counter pump 9 . The precipitation of CaCO 3 shells by mainly coccolithophores, foraminifers (both calcite) and pteropods (aragonite), and the resulting particulate inorganic carbon (PIC) flux from the surface ocean, causes an increase in surface ocean pCO 2 (ref. 10) on timescales of the order of 100-1,000 years 1 . As a result of these opposing effects, particle flux studies addressing deep-ocean CO 2 sequestration need to discriminate between organic (soft-tissue) and inorganic carbon, with a strong focus on a relationship formalized as the rain ratio (POC:PIC; ref. 2). Despite its obviou...

Abstract. Planktic foraminifera are heterotrophic mesozooplankton of global marine abundance. The position of planktic foraminifers in the marine food web is different compared to other protozoans and ranges above the base of heterotrophic consumers. Being secondary producers with an omnivorous diet, which ranges from algae to small metazoans, planktic foraminifers are not limited to a single food source, and are assumed to occur at a balanced abundance displaying the overall marine biological productivity at a regional scale. With a new non-destructive protocol developed from the bicinchoninic acid (BCA) method and nano-photospectrometry, we have analysed the protein-biomass, along with test size and weight, of 754 individual planktic foraminifers from 21 different species and morphotypes. From additional CHN analysis, it can be assumed that proteinbiomass equals carbon-biomass. Accordingly, the average individual planktic foraminifer protein-and carbonbiomass amounts to 0.845 µg. Samples include symbiont bearing and symbiont-barren species from the sea surface down to 2500 m water depth. Conversion factors between individual biomass and assemblage-biomass are calculated for test sizes between 72 and 845 µm (minimum test diameter). Assemblage-biomass data presented here include 1128 sites and water depth intervals. The regional coverage of data includes the North Atlantic, Arabian Sea, Red Sea, and Caribbean as well as literature data from the eastern and western North Pacific, and covers a wide range of oligotrophic to eutrophic waters over six orders of magnitude of plankticforaminifer assemblage-biomass (PFAB). A first order estimate of the average global planktic foraminifer biomass production (> 125 µm) ranges from 8.2-32.7 Tg C yr −1 (i.e. 0.008-0.033 Gt C yr −1 ), and might be more than three times as high including neanic and juvenile individuals adding up to 25-100 Tg C yr −1 . However, this is a first estimate of regional PFAB extrapolated to the global scale, and future estimates based on larger data sets might considerably deviate from the one presented here. This paper is supported by, and a contribution to the Marine Ecosystem Data project (MAREDAT).