Biological production in two contrasted regions of the Mediterranean Sea during the 1 oligotrophic period : An estimate based on the diel cycle of optical properties measured by 2 BGC-Argo profiling floats 3

Biological production in two contrasted regions of the Mediterranean Sea during the 1 oligotrophic period: An estimate based on the diel cycle of optical properties measured by 2 BGC-Argo profiling floats 3 Marie Barbieux1, Julia Uitz1, Alexandre Mignot2, Collin Roesler3, Hervé Claustre1, Bernard 4 Gentili1, Vincent Taillandier1, Fabrizio D'Ortenzio1, Hubert Loisel4, Antoine Poteau1, Edouard 5 Leymarie1, Christophe Penkerc’h1, Catherine Schmechtig5, Annick Bricaud1 6 1CNRS and Sorbonne Université, Laboratoire d’Océanographie de Villefranche, LOV, 06230 Villefranche-sur7 Mer, France 8 2Mercator Océan, 31520 Ramonville-Saint-Agne, France 9 3Bowdoin College, Earth and Oceanographic Science, Brunswick, Maine 04011, USA 10 4Université Littoral Côte d’Opale, Université Lille, CNRS, Laboratoire d’Océanologie et de Géosciences, 59000 11 Lille, France 12 5OSU Ecce Terra, UMS 3455, CNRS and Sorbonne Université, Paris 6, 4 place Jussieu, 75252 Paris CEDEX 05, 13 France 14

chlorophyll maximum (SCM) layers, using an existing approach applied to diel cycle 23 measurements of the particulate beam attenuation (cp) and backscattering (bbp) coefficients. The 24 diel cycle of cp provided a robust proxy for quantifying biological production in both systems; 25 that of bbp was comparatively less robust. Derived primary production estimates vary by a factor 26 of 2 depending upon the choice of the bio-optical relationship that converts the measured optical 27 coefficient to POC, which is thus a critical step to constrain. Our results indicate a substantial 28 contribution to the water column production of the SCM layer (16-42%), that varies largely 29 with the considered system. In the Ligurian Sea, the SCM is a seasonal feature that behaves as 30 a subsurface biomass maximum (SBM) with the ability to respond to episodic abiotic forcing 31 by increasing production. In contrast, in the Ionian Sea, the SCM is permanent, primarily 32 induced by phytoplankton photoacclimation and contributes moderately to water column 33 production. These results clearly demonstrate the strong potential for transmissometers 34 deployed on BGC-Argo profiling floats to quantify non-intrusively in situ biological production 35 of organic carbon in the water column of stratified oligotrophic systems with recurring or 36 permanent SCMs, which are widespread features in the global ocean. 37

Introduction 39
Primary production is an essential process in the global ocean carbon cycle (Field et al. 40 1998). As a major driver of the biological carbon pump, this biogeochemical process plays a 41 critical role in the regulation of the Earth's climate (e.g. Sarmiento & Siegenthaler 1992;42 Falkowski 2012). Hence, quantifying primary production as a function of time and space in the 43 ocean stands as a major challenge in the context of climate change. The balance between gross 44 primary production and community respiration in the ocean determines the trophic status of 45 marine systems, i.e. whether the system acts as a source or a sink of carbon (Williams 1993). 46 This balance depends on the considered region and varies substantially according to spatial and 47 temporal scales (Geider et al. 1997

BGC-Argo multi-profiling floats and data processing 177
We deployed BGC-Argo floats programmed for "multi-profile" sampling in each of these 178 two regions (Fig. 1) The BGC-Argo floats used in this study are "PROVOR CTS-4" (nke Instrumentation, 189 Inc.). They were both equipped with the following sensors and derived data products: (1) a 190 CTD sensor for depth, temperature and salinity; (2) a "remA" combo sensor that couples a 191  Argo fluorescence data from both the Ligurian and Ionian regions, consistently  213 with the processing performed at the Coriolis Data Center. 214 For the particulate backscattering coefficient (bbp), we followed the BGC-Argo 215 calibration and quality control procedure of Schmechtig et al. (2016). The backscattering 216 coefficient at 700 nm (m -1 ) is retrieved following Eq. (1): 217 (1) 218 where = 1.076 is the empirical weighting function that converts particulate volume 219 scattering function at 124 o to total backscattering coefficient (Sullivan et al. 2013); !" is the 220 raw observations from the backscattering meter (digital counts); !" (digital counts) and 221 !" (m -1 sr -1 count -1 ) are the calibration coefficients provided by the manufacturer; and 222 is the contribution to the Volume Scattering Function (VSF) by the pure seawater at the 223 700 nm measurement wavelength that is a function of temperature and salinity (Zhang et al. 224 2009). 225 The calibration procedure applied to the particulate beam attenuation coefficient (cp) is 226 similar to that described in Mignot et al. (2014). The beam transmission, T (%), is transformed 227 into the beam attenuation coefficient, c (m -1 ), using the relationship: 228 where x is the transmissometer pathlength (25 cm). The beam attenuation coefficient c is the 230 sum of the absorption and scattering by seawater and its particulate and dissolved constituents. 231 At 660 nm, the contribution of CDOM (cCDOM) can be considered negligible in oligotrophic 232 waters because, although its absorption in the blue is comparable to that of particulate material 233 (Organelli et al. 2014), the cCDOM spectrum decays exponentially towards near zero in the red 234 12 followed by a second profile at solar noon (tn), a third at sunset (tss) and a last night profile at 260 approximately midnight (tm). For this float, the sampling cycle is repeated each day. We also consider the relative daily variation ∆ B " and ∆ B !" (expressed as % change) for 271 each float and each day of observation, from sunrise to sunrise as follows: 272 with " ( -. ) and !" ( -. ) being the values of cp and bbp at sunrise and " ( -./# ) and 275  The time-rate-of-change in depth-resolved POC biomass, b(z,t), can be described by a 331 partial differential equation: 332 where µ(z,t) is the particle photosynthetic growth rate and ( , ) the particle loss rate at depth 334 The daily (24-hour) depth-integrated gross production of POC, P (in units of gC m -2 d -1 ), 361 is defined as: 362 363 with tsr the time of sunrise on day 1 and tsr+1 the time of sunrise the following day. Equation (5)  364 can be used to express P as a function of l, b(z,t), and the rate of change of b(z,t): 365 which yields: 367 where the gross production P is calculated as the sum of the net daily changes in POC biomass 369 plus POC losses, assuming a constant rate (l) during daytime and nighttime. 370 Finally, using the trapezoidal rule, Eq. (11) simplifies into We first provide an overview of the biogeochemical and bio-optical characteristics 411 measured by the two BGC-Argo profiling floats in the Ligurian and Ionian Seas. We then assess 412 the usefulness of the diel cycle of the bbp coefficient for deriving community production, in 413 comparison to the cp-derived estimates as a reference, and discuss the cp-derived estimates. 414 Finally, we examine the community production estimates in both study regions, with an 415 emphasis on the SCM layer and its biogeochemical significance. 416

Biogeochemical and bio-optical context in the study regions 417
Both study regions are characterized by either seasonal or persistent oligotrophy, with 418 mean surface Chl values ranging within 0.08-0.22 mg m -3 (Fig. 3) In the surface layer of the Ligurian Sea, the diel cycles of cp and bbp exhibit, respectively, 481 mean relative daily variation ( ∆ E ) of 12.7% and 2.3%, and a range in relative daily variations 482 ( ∆ E ) of 256.7% and 28.5% (Table 3). These values are of the same order of magnitude as those 483 reported by Kheireddine & Antoine (2014), acquired from the BOUSSOLE surface mooring in 484 the same area and during the oligotrophic season (from -5% to 25% for cp and from -2% to 10% 485 for bbp). Interestingly, the diel cycle of the cp coefficient appears systematically more 486 pronounced than that of bbp, with larger values of ∆ E and ∆ E , regardless of the considered 487 region and layer of the water column (Table 3). 488 To first order, the variability in the bbp and cp coefficients is determined by the variability 489 in particle concentration, which underpins their robustness as POC proxies in open-ocean 490 conditions and explains their coherent evolution on a monthly timescale (Figs. 3-4). 491 Nevertheless, to second order, these coefficients vary differentially with the size and 492 composition of the particle pool. In particular, phytoplankton make a larger contribution to cp 493 than bbp, in part due to their strong absorption efficiency. In addition, bbp is more sensitive to used interchangeably with cp for assessing daily changes in POC or community production, but 506 perhaps provides additional information on the particulate matter and its production rates. Our 507 results support these previous findings, not only for the surface layer of the Ligurian Sea, but 508 also for the whole water column of both the Ligurian and Ionian regions. 509 We now consider the integrated euphotic zone gross community production estimates 510 derived from the bio-optical diel cycle-based method (Fig. 6). We compare the cp-and bbp-511 based estimates with primary production estimates computed with the model of Morel (1991). a factor of ten, with respective mean values of 0.11±0.28 gC m -2 d -1 and 1.18 ±1.13 gC m -2 d -1 514 in the Ligurian Sea, and 0.04±0.04 gC m -2 d -1 and 0.46 ±0.11 gC m -2 d -1 in the Ionian Sea. In 515 addition, the bbp-derived production is much lower than the primary production computed with 516 the model of Morel (1991), which has mean values of 0.91±0.14 gC m -2 d -1 in the Ligurian Sea 517 and 0.31 ±0.04 gC m -2 d -1 in the Ionian Sea. The significantly lower community production 518 rates are a direct effect of the dampened relative daily amplitude of the bbp diel cycle (Table 3) The cp-derived estimates of gross community production, integrated within the euphotic 534 layer, compare favorably with those found in the literature for similar Mediterranean areas (see 535   Table 4 and references therein). The cp-based estimates show a 2.5-fold difference between the 536 Ligurian Sea and the Ionian Sea (mean of 1.18 gC m -2 d -1 and 0.46 gC m -2 d -1 , respectively; 537 Table 6). In comparison, water column-integrated primary production values, either inferred 538 from satellite observations and biogeochemical models or measured in situ, vary within the 539 range 0.13-1 gC m -2 d -1 and 0.14-0.69 gC m -2 d -1 for the Western (or Ligurian) and Eastern (or 540 Ionian) region, respectively (Table 4). As expected, our cp-based community production rates 541 are larger than published primary production rates. The present cp-derived values also compare 542 favorably with gross community production estimates inferred from a similar approach applied 543 to bio-optical measurements from the BOUSSOLE mooring in the Ligurian Sea The empirical relationships linking the cp (or bbp) coefficient to POC are known to exhibit 548 regional and seasonal variability in response to changes in the composition of the particle 549 assemblage and associated changes in particle size, shape and type, i.e. biogenic or mineral 550  (Table 5). That said, although the absolute magnitudes vary depending upon proxy choice, the 563 differences observed between locations is robust. 564 The use of the single relationship established from Mediterranean waters (Oubelkheir et al. 565 2005) appears as a reasonable choice for the study region. Yet, if more relevant bio-optical 566 proxy relationships are available, such as one that accounts for spatial and seasonal variations, 567 and even applicable to different layers of the water column, that would certainly reduce the 568 uncertainty in the rate estimation. Although this is beyond the scope of the present study, we 569 recognize that such investigations should be conducted in the future in order to refine optics-570 based biomass (POC) and community production estimates. 571

Regional and vertical variability of production 572
The temporal evolution of the cp-derived POC biomass integrated within the three distinct 573 layers of the water column is presented for the two study regions in Fig. 7. The integrated POC 574 concentration values follow similar temporal trends as reported for cp (Figs. 3-4). In the 575 Ligurian Sea, the euphotic layer-integrated POC varies between 1.5 and 6.0 gC m -2 (mean of 576 3.7±1.1 gC m -2 ; Fig. 7a and Table 6). There was a decrease from late May to mid-July (6.0 to 577 1.5 gC m -2 ) followed by a moderate peak (3.9 gC m -2 ) between mid-July and mid-August (as 578 bounded by the dashed lines in Fig. 5). The cp-based community production did exhibit large 579 variability over the time period (Fig. 7b and Table 6), but interestingly, the moderate POC peak 580 observed in the core of the oligotrophic season (between mid-July and mid-August) is 581 associated with the maximum production rate of the time series (4.3 gC m -2 d -1 ). 582 In the Ionian Sea, the POC biomass integrated within the euphotic zone is much lower 583 than in the Ligurian Sea and remains more stable over the time period (1.9±0.24 gC m -2 ; Fig.  584 7c and Table 6). As with POC, the community production is much lower in the Ionian Sea than 585 in the Ligurian Sea, but still exhibits substantial variability with values ranging within 0.06-0.68 gC m -2 d -1 (Fig. 7d) The gross community production estimates integrated over different layers of the water 592 column reveal distinct patterns. In the Ligurian Sea, both the euphotic and SCM layers show 593 large production rates (0.96±1.3 gC m -2 d -1 ), with production in the SCM layer frequently 594 equaling or overtaking on the production in the euphotic layer (Fig. 7b). This is particularly 595 striking in late July, when the production peak is actually associated with a large enhancement 596 of the production in the SCM layer (4.9 gC m -2 d -1 ). In contrast, the surface layer shows reduced 597 production rates (0.29±0.33 gC m -2 d -1 ), a pattern also observed in the Ionian Sea (0.11±0.04 598 gC m -2 d -1 ). In the Ionian Sea, the production is maximal in the euphotic zone, and very variable 599 and occasionally larger in the SCM layer (0.14±0.39 gC m -2 d -1 ; Fig. 7d). The bio-optical diel 600 cycle-based method produces several occurrences of negative values in the SCM layer, 601 indicating that the 1D assumption is occasionally not satisfied in the lower part of the euphotic 602 layer. This could arise when physical processes that transport particles are larger than local 603 growth and loss of POC. 604 Our results support the hypothesis raised in previous studies (e.g. Mignot et al. 2014; 605 Barbieux et al. 2019) that, in the Ligurian temperate-like system, the SCM, which is in fact a 606 SBM, may be highly productive. Conversely, in the Ionian region, which shows similarities 607 with subtropical stratified oligotrophic systems, the SCM primarily reflects photoacclimation 608 and is less productive. Beyond these mean regional trends, both SCM systems exhibit some 609 temporal variability in production, a somewhat unexpected pattern at the core of the presumably 610 stable oligotrophic season. 611

Production in the SCM layer in relation with the biotic and abiotic context 612
Here we investigate the temporal variability in the SCM layer production and attempt to 613 interpret the observed patterns in the context of biological and abiotic conditions. 614

615
The pigment data collected during the BOUSSOLE and PEACETIME cruises 616 concomitantly with the deployments of the fLig and fIon floats, respectively, are used as proxies 617 for phytoplankton community structure (Fig. 8). In the Ligurian Sea, nanophytoplankton 618 (mainly prymnesiophytes) appear as dominant contributors to the phytoplankton assemblage 619 both in the surface layer (48±8%; Fig. 8b) and SCM layer (54±10%). Picophytoplankton 620 (prokaryotes and small chlorophytes) and microphytoplankton (diatoms and dinoflagellates) 621 are present in moderate proportions, with 30±11% and 22±5% in the upper layer, and 19±7% 622 and 27±9% in the SCM layer, respectively (Figs. 8a and 8c). No marked shift in the community 623 composition is observed during the timeseries, although occasional increase in the contribution 624 of microphytoplankton is observed in the SCM layer, with no clear temporal trend (Fig. 8a and  625 Appendix B). In the Ionian Sea, the surface layer displays large contribution of 626 nanophytoplankton (56±2%; Fig. 8e) and, to a lesser extent, picophytoplankton (29±3%; Fig.  627 8d). However, the SCM level is characterized by an enhanced contribution of 628 microphytoplankton (diatoms) to the algal assemblage (49±5%; Fig. 8f Chl and cp / Chl ratios lie in the fact that they are sensitive to different particle size ranges 655 (Roesler and Boss 2008) and, thus, when they are not correlated, one can qualitatively discern 656 differing dynamics across the phytoplankton size spectrum. 657 The bbp / cp ratio is very different between the Ligurian and Ionian Seas, with significantly 658 lower values in the Ligurian Sea (0.0068±0.0009, and 0.0095±0.0009; Fig. 9). These ratios the Ionian Sea SCM, which tends towards nonalgal particles. In the Ligurian Sea, the bbp / cp 661 ratio remains <0.0087 and reaches a minimum of 0.0055 over the period coinciding with the 662 production event from mid-July to mid-August (Fig. 9a), consistent with phytoplankton 663 dominance. In contrast, in the Ionian Sea SCM, the bbp / cp ratio increases from 0.0085 in late 664 May, peaking at nearly 0.012 in early August, and then decreasing back to 0.0085 in September 665 (Fig. 9b). The tendency towards a ratio of 0.01 (or 1%) in the core of the oligotrophic season, The Ligurian Sea exhibits enhanced community production during the period from mid-710 July to mid-August 2014, which is associated with a comparatively moderate increase in the 711 biomass indicators (Figs. 3-4) and cp-derived POC (Fig. 7a). During this time period, the depth 712 of the SCM shoals by 25 m. This change occurs concurrently with a slight shoaling of the 713 density isopycnals (Figs. 3a-c), and a doubling (from 0.5 to 1 mol quanta m -2 d -1 ) in the daily 714 PAR within the SCM layer (Fig. 10a). Therefore, we suggest that the observed production 715 episode may result from physical forcing that induces an upwelling of the water mass, thereby 716 resulting in an alleviation of the light/nutrient limitation and an adequate balance between light 717 and nutrient availability in the SCM layer. This SCM production episode is associated with a 718 moderate phytoplankton biomass (0.8 Chl mg m -3 ), dominated by a nanoplankton community. 719 It coincides with an increase in the cp / Chl and bbp / Chl ratios, which we attribute to a boost in 720 the carbon-to-Chl ratio resulting from production in enhanced light conditions. Because it 721 appears to result from changes in light conditions, we may attribute this production event to 722 photosynthetic (not community) growth. 723 In the Ionian Sea, the depth of the SCM follows the depth of the isopycnal 28.85 during 724 the period from late to May to mid-August 2017 (Figs. 3d-f). In mid-August, the SCM reaches 725 its deepest point (~125 m) concurrent with deepening isopycnals, decreased PAR levels within 726 the SCM layer (Fig. 10b) and minimum values of Chl, cp and bbp. Afterwards, the SCM depth 727 decouples from the position of the isopycnals (Fig. 3d-f)

Contribution of the SCM to the water column production 741
In order to assess the relative contribution of the SCM layer to the production occurring 742 in the whole water column, we compare the cp-based estimates integrated within the productive 743 layer (0-1.5 Zeu) and SCM layers. Our results suggest that, for these oligotrophic systems, the 744 production integrated within the SCM layer represents a substantial fraction (FSCM) of the gross 745 community production integrated within the productive layer. This is particularly the case for 746 the Ligurian Sea where FSCM reaches ~42%, and to a lesser extent for the Ionian Sea with Our study emphases the promising potential of BGC-Argo profiling floats for providing 802 a non-intrusive, high-frequency assessment of POC production within the whole water column, 803 which is critical in particular for applications to stratified oligotrophic environments with 804 recurring or permanent SCMs. The present results, based on data from two Mediterranean BGC-Argo floats in the broad, remote subtropical gyres. In such systems, biological production 807 is not constant but, instead, shows high temporal heterogeneity (Karl et al. 2003;Claustre et al. 808 2008) that may be missed by traditional sampling, leading to a potential underestimate of the 809 biogeochemical impact of these systems in global carbon budgets. Implementing such a BGC- the BOUSSOLE and PEACETIME cruises, as well as David Antoine, PI of the BOUSSOLE 849 project, and Cécile Guieu and Karine Desboeufs, PIs of the PEACETIME project. We thank 850 the International Argo Program and Coriolis project, which contributed to making the data 851 freely and publicly available. Marin Cornec is also warmly thanked for useful discussion 852 regarding biological production in SCM systems. We finally wish to thank the two anonymous 853 Reviewers and the co-Editor-in-Chief for their useful comments and suggestions.