Multi-year robotic observations reveal the seasonality of downward carbon export pathways in the Southern Ocean


 At high latitudes, the export of organic matter from the surface to the ocean interior, the biological carbon pump, has conventionally been attributed to the gravitational sinking of particulate organic carbon (POC). Conspicuous deficits in ocean carbon budgets have recently challenged this long-lived paradigm of a sole pathway. Multiple strands of evidence have demonstrated the importance of additional export pathways, including the particle injection pumps (PIPs). Recent model estimates revealed that PIPs have a comparable downward POC flux to the biological gravitational pump (BGP), but with potentially different seasonal signatures. To date, logistical constraints have prevented concomitant and extensive observations of these pumps, and little is known about the seasonality of their fluxes. Here, using year-round robotic observations and recent advances in optical signal analysis, we concurrently investigated the functioning of two PIPs - the mixed layer and eddy subduction pumps - and the BGP in Southern Ocean waters. By comparing three phytoplankton bloom cycles in contrasting environments, we show how physical forcing and phytoplankton phenology influence the magnitude and seasonality of these pumps, with implications for carbon sequestration efficiency.

at the (sub)mesoscale (<100 km) these processes can be equal to, or even greater than the 48 BGP 10,11 . This spatial mismatch illustrates the difficulty of carrying out comprehensive 49 intercomparison of these pumps in the field. The mesopelagic migrant pump 12 , seasonal lipid 50 pump 13 and large-scale subduction 14 pump are also important contributors to the BCP, although 51 assessing their relative contribution is even more challenging 3 and beyond this study. will fundamentally push forward our understanding of ocean biological carbon export and 58 sequestration, and help to close regional ocean carbon budgets. 59 Biogeochemical-Argo (BGC-Argo) floats with multi-year missions and high frequency 60 sampling offer a promising way to jointly investigate the PIPs and the BGP over a broad range 61 of time and space scales. Such platforms have already been successfully used to characterise 62 the MLP at sub-seasonal to seasonal scales 8,9 , the ESP at pan-Antarctic scale 6 , and the 63 seasonality of the BGP 15, 16 . In this work we refine and bring together a range of previously to infer the causal mechanisms that set the magnitude and seasonality of each of these pumps. 75 These mechanisms include bloom timing or seasonality, hereafter called phenology, which 76 influences the concentration, size and composition of the particle assemblage in the upper 77 ocean. 78 79 Characteristics of the particle assemblage in contrasting environments 80 We present here in situ observations collected by a BGC-Argo float (WMO 7900791) that 81 travelled >6000 km across the Pacific sector of the SO (Fig. 1a). This float was deployed in  was profiling (Fig. S4). Bio-minerals act as ballast by increasing particle specific gravity and 110 sinking speeds, and provide protection from remineralisation 26 , potentially explaining the 111 massive invasion of POC to 450 m following the coccolithophore bloom (Fig. 2a).

112
The 2017-18 bloom was less intense with lower phytoplankton biomass. It was characterised 113 by a high Fs / bbs at its peak (apex), when the mixing layer abruptly shoaled from 300 to <100 114 m (Fig. 2a, b). Such high Fs / bbs ratios associated with high-latitude spring blooms, when light 115 is not limiting (Fig. S5), have been attributed to diatom-dominated events 20,21 , which in this 116 case would have benefited from iron inputs by winter deep mixing and melt waters (Fig. 1c, d).   (Fig. S5) and convective mixing commenced again (Fig. S1b, c).

128
The 2018-19 bloom was partially sampled as data were missing for nearly two months early in 129 the productive season. The SCM observed from December to February 2018 (Fig. 2b), 130 relatively small particles in the surface layer (Fig. 2c), and a high iron limitation index (    Fig 3a), the dominance of the Fl spike index was attributed 147 to a diatom bloom that generally precede coccolithophore blooms 30 . Indeed, diatom export 148 events have often been described as an initial pulse of fresh aggregates 18 followed by a second 149 pulse of resting spores, empty frustules and fecal pellets 31 , the latter explaining the significant  suggesting that the ESP also transports large fresh aggregates produced in the Ez, in addition to 197 small suspended particles and dissolved compounds 9 .

198
As for the ESP, the MLP was quantified as the inventory of POC measured in MLP remnant     Large-scale estimates of the PIP contribution to carbon storage require regional high-resolution 235 models able to simulate some of the complex BCP mechanisms we observed 11,36 . However, to    (1) 297 where tres is the residence time of a particle in the sample volume (0.1 s), tsamp is the duration of      The iron stress index was computed following the method in Ryan-Keogh & Thomalla (2020) 55 .

379
The concept of this method is that NPQ variability is linked to iron and light availability and 380 has the potential to provide important diagnostic information on phytoplankton physiology 56 .

381
To remove the effect of in situ light availability on NPQ variability, Ryan-Keogh & Thomalla 382 (2020) 55 proposed to compute NPQ the initial slope of the NPQ-PAR curve. Thereby, NPQ 383 could be used as a proxy for iron limitation, with higher values being associated with greater 384 iron stress. In our study, NPQ as a function of depth was quantified as the difference between 385 the quenching corrected fluorescence profile and the quenched one, normalized by the latter.

386
For each profile, we plotted our iPAR estimates against NPQ values. We then applied a linear