Temperature is a better predictor of stable carbon isotopic compositions in marine particulates than dissolved CO₂ concentration

Abstract

The stable carbon isotopic composition of marine particulate organic matter (δ¹³C_POM) varies with source and environmental conditions. Dissolved carbon dioxide (CO₂) concentration is thought to influence δ¹³C_POM more than temperature, but this relationship is poorly constrained in marginal seas. Here we present δ¹³C_POM, hydrographic and carbonate system variables at the deep chlorophyll maxima of the southern Yellow Sea in late summer 2017. We find δ¹³C_POM values varied between stable and cyclonic gyre regions, but indicated autochthonous organic matter production and were more strongly correlated with temperature than dissolved CO₂ concentration throughout. We find that the relationship between temperature and δ¹³C_POM was independent of CO₂ concentration, whereas the relationship between δ¹³C_POM and CO₂ concentration was dependent on temperature also being correlated with CO₂ concentration. We suggest that temperature is the primary determinant of marine δ¹³C_POM due to temperature-dependent metabolism in phytoplankton, irrespective of inorganic carbon acquisition mode.

Introduction

Suspended particulate organic matter (POM) in seawater is a key component of the biological pump which transfers the upper ocean photosynthesized organic matter to the deep sea^1,2. Stable carbon isotopic composition (δ¹³C) of the bulk POM (δ¹³C_POM) has long been used to evaluate proportions of marine-derived versus terrestrial POM fractions in coastal oceans^3,4,5, which are disproportionately important to ocean carbon cycles and budgets^6,7,8. However, large variations of δ¹³C (‒9 to ‒35‰) naturally occur in marine phytoplankton (δ¹³C_phyto) in the modern ocean^9,10,11. These values severely overlap with δ¹³C of terrestrial plants-derived OM (‒37 to ‒20‰ of C3 plants¹² and ‒14 to ‒11‰ of C4 plants¹³). The widely varied δ¹³C_phyto are increasingly understood to be a function of environmental variables (pH, temperature, inorganic carbon concentration, and its δ¹³C value, nutrient availability, light intensity, and day length) under which the OM was synthesized^11,14,15 and the phytoplankton cell physiology, including cell size, growth rate^16,17,18,19, and inorganic carbon acquisition mode (passive diffusion versus active uptake or CO₂ concentrating mechanisms, i.e., CCMs)^20,21,22,23. The widespread CCMs in marine phytoplankton are processes of increasing the CO₂ level at the site of Rubisco through transporters and carbonic anhydrases (CAs), which accelerate the otherwise slow interconversion between HCO₃^‒ and CO₂^24,25,26. The possible influences of biochemical CCMs involving C4-type photosynthesis on δ¹³C_phyto have been suggested in some species of marine phytoplankton^27,28 but not yet proven^29,30. Not all of the factors are necessarily relevant under natural conditions in the field, and only a few are likely to exert primary control on δ¹³C_phyto at most times and location³¹. Although the exact biological processes determining the wide δ¹³C_phyto ranges are still under debate, in the inferred absence of a major contribution of terrestrial OM to the oceanic POM, the changes of [CO₂aq] have been proposed as the principal determinant of δ¹³C_POM variability over relatively large areas across the global ocean, where inverse correlations between δ¹³C_POM and [CO₂aq] were commonly observed^32,33,34. On the other hand, due mainly to the well-defined relationship between CO₂ solubility and temperature³⁵, strong positive correlations of temperature and oceanic δ¹³C_POM^36,37 have long been considered to be resulting from the temperature effects on CO₂ availability^{38,39,40,41,42}.

Hence, a better understanding of the scales and mechanisms of the marine endmember δ¹³C_POM variations is an essential prerequisite for the reliable and effective use of this proxy. However, limited observations result in poorly constrained marine end member δ¹³C_POM in marginal seas, particularly in the southern Yellow Sea (SYS). The SYS is known as a large dynamic marine ecosystem surrounded by China and South Korea^43,44,45 (Fig. 1a). The deep chlorophyll maxima (DCM) layer in the SYS^46,47, as the general case in the global coastal oceans, forms a substantial part of the annual primary production and export productivity^47,48,49. The DCM layer, identified by a pronounced peak in the vertical profiles of chlorophyll (Chl) fluorescence, is a distinct feature of stratified waters that commonly formed between the nutrient-poor, light-replete upper mixed layer and a nutrient-rich, light-limited bottom mixed layer^50,51,52. Specifically, during stratified seasons in the SYS, the production within the surface mixed layer is limited as nutrients are depleted due to the consumption by prior spring bloom^53,54. By contrast, the primary producers at the DCM layer can access nutrients from the bottom water, which is enriched in nutrients due to regeneration processes. For the refinement of marine endmember δ¹³C_POM in the SYS, it is of particular interest, therefore, to investigate the δ¹³C_POM at the DCM layer of the SYS. In addition, investigating δ¹³C_POM in such a dynamic oceanic regime is helpful to better understand the environmental regulation of the oceanic δ¹³C_POM. In fact, previous studies on the mechanisms of δ¹³C_POM variations have mostly focused on the surface ocean^33,55,56,57. Investigation on the δ¹³C_POM was rarely carried out in the subsurface water⁵⁸, such as the DCM layer⁵⁹. Recently, Close and Henderson⁶⁰ compiled data from published sources reporting δ¹³C_POM values for the upper 250 m of open ocean. They demonstrated that δ¹³C_POM values in the lower euphotic zone, on average, are 1.4‰ lower than those in the upper euphotic zone. This pattern may suggest that the marine endmember δ¹³C_POM, or at least δ¹³C_POM at the DCM layer, cannot be simply represented by surface δ¹³C_POM.

Fig. 1: Study area and spatial distribution of the hydrological and biogeochemical parameters.

a Sample location in the southern Yellow Sea (SYS). Circles represent the hydrological stations, crosses are stations with distinct deep chlorophyll maxima (DCM) layers, and red pins indicate stations where [CO_2aq] in the DCM layer is concurrently obtained. The solid gray line is the 50-m isobath. δ¹³C_POM in two DCM stations (H01 and H24) will not be discussed further in this study, since the seawater sample at the DCM layer of H01 was not collected, and the filter at H24 contains less than 50 μgC, which will introduce error for the δ¹³C_POM measurement. b Vertical distribution of temperature, salinity, and chlorophyll fluorescence in four latitudinal transects of middle SYS as marked in (a). The dark stars indicate the location where suspended particles were collected. The strong stratification and position of the Yellow Sea Cold Water Mass (YSCWM) can be clearly seen from the structures of temperature and salinity. The upwelling fronts in the western boundary of the YSCWM area can be recognized from temperature profiles. Chl fluorescence profiles clearly demonstrate the existence of the DCM layer. Source data for (b) is available in Supplementary Data 2. c Horizontal distribution of temperature in both surface and DCM layer, and elemental and stable carbon isotopic composition of particulate organic matter in the DCM layer. Note that the DCM is the depth at which the data was collected. The polygon marked by a white line represents the location of the cyclonic gyre.

Here, we concurrently measure Chl a, particulate organic carbon (POC) and nitrogen (PN) concentrations and δ¹³C_POM along with hydrological parameters (e.g., temperature and salinity), and dissolved inorganic carbon (DIC) and total alkalinity (TAlk) (to calculate [CO₂aq]) at DCM layers of the SYS. Our data confirm that the POM at the DCM layer is predominantly from marine production, of which the δ¹³C_POM are widely varied, and that the importance of different phytoplankton communities in two of the contrasting oceanographic areas within the SYS in shaping the δ¹³C_POM-temperature relationships is evident. Moreover, we find that the previously hypothesized inverse relationship between δ¹³C_POM and [CO₂aq] occurs only when [CO₂aq] is temperature-dependent. When [CO₂aq] is deviating from the temperature-constrained value, then δ¹³C_POM is less related to [CO₂aq] but still strongly correlated with temperature. This is apparent that δ¹³C_POM is associated more with temperature than with CO₂aq level in which δ¹³C_POM is produced. A revisit of studies concluding a general inverse relationship between δ¹³C_POM and [CO₂aq] strongly supports this view. We provide a mechanistic explanation that reconciles the apparent differences in the trends of δ¹³C_POM-temperature relationships (positive and negative) and reveals that differential temperature scaling on processes regulating δ¹³C_POM values provides a mean to distinguish between the relative importance of factors on δ¹³C_POM variations. In addition, our findings do not support a direct effect of [CO₂aq] on δ¹³C_POM variations but point to a shift in the regulation of CCMs activity and, consequently, the predominance of passive diffusion or efficient CCMs. Given that the [CO₂aq] in large areas of the present-day ocean is higher than the critical 10 μmol/L^14,16,18,61, the increasing CO₂ in the future may not be very effective on carbon assimilation and, therefore, δ¹³C_POM in the oceanic ecosystem, especially in high latitude areas and subsurface water (e.g., DCM layer).

Results and discussion

Dynamic oceanographic conditions

The hydrological profiles clearly show the presence of a bottom cold water dome at the central region of the SYS (Fig. 1a, b), the so-called Yellow Sea Cold Water Mass (YSCWM), which is a remnant of previous winter water that developed due to stratification from late spring to autumn⁶². The YSCWM acts as a large reservoir of both nutrients and [CO₂aq] due to the remineralization of eutrophication produced OM during spring bloom^63,64. The DCM layer in most locations is positioned within or at the bottom of the thermocline (Fig. 1b). Temperature of seawater in the DCM layer ranges broadly from 8.6 to 25.1 °C (on average 15.8 ± 4.7 °C; Supplementary Data 1), and it is distinctly lower than the surface water (22.9‒27.1 °C, on average 25.0 ± 0.9 °C) (Fig. 1c). The location of DCM in the SYS was also found within the euphotic zone and at the upper part of nitracline⁴⁷. According to the published nutrient data from the same cruise in the present study⁶⁵, the dissolved inorganic nitrogen (DIN = NO₃^‒ + NO₂^‒ + NH₄⁺, 0.2‒13.7 μmol kg^‒1) and soluble reactive phosphorus concentrations (P, 0.1‒1.0 μmol kg^‒1), essential macronutrients for phytoplankton growth, are relatively high at the DCM layer of the SYS (Supplementary Fig. 1). Even though the DIN/P ratio in the seawater varies in a large range (1‒21, on average 10 ± 5)⁶⁵, the regeneration and assimilation ratio of these two nutrients at the DCM layer of the SYS is constant and close to the traditional Redfield ratio (16)⁶⁶ (Supplementary Fig. 1). This divergence results from the fact that mixing, mainly upward supply from deep water^63,64, rather than consumption in situ is the dominant source of nutrient at the DCM layer of the SYS⁶⁷. The DCM layer is absent in the well-mixed shallow coastal area (Fig. 1a, b) and the well-stratified southern SYS (except at station H35 in L32), wherein the presence of buoyant Yangtze River plume strengthened the stratification (Fig. 1c). For the former, the Chl fluorescence shows a nearly constant value from surface to bottom (Fig. 1b), while for the latter, maximum of Chl fluorescence presents in the upper water instead of the subsurface.

Strong upwelling can be clearly identified from the isotherms uplifting toward the coast in the western boundary of the YSCWM, and it can reach the surface, accounting for the surface cold water patches (Fig. 1b, c) and the ventilation of the thermocline there⁶⁸. Uplift of nutrient-rich water from the bottom reservoir was also observed in the western boundary of the YSCWM⁴⁵. The shallow coastal water and the deep YSCWM area are therefore separated by the upwelling fronts (Fig. 1b, c). The horizontal distributions of temperature in surface water and the DCM layer indicate the presence of a temperature front at ~50 m isobath in the western boundary of YSCWM (Fig. 1a, c). A summertime cyclonic gyre has been identified along the temperature front in the subsurface of the SYS in both observational and numerical studies⁶⁸. Indeed, the DCM stations are divided into three groups based on the relationships of δ¹³C_POM and temperature along with the hydrological conditions (see subsection Temperature Other than [CO₂aq] Dependent δ¹³C_POM for details). Specifically, these three groups belong to the stable region, which includes stations with relatively stable hydrological structure, cyclonic gyre which is formed by stations in the western boundary of YSCWM (blue polygon), and the remaining that consists of eight isolated dynamic stations (yellow shaded circles (Fig. 1a). Three of these eight stations are in the coastal upwelling region (H16, H17, H22) and three in the deep central area of SYS (H09, H10, HS2), where downwelling was reported⁶².

Marine phytoplankton source of POM in the DCM of the SYS

POC and PN are chemically dominating POM. POC here is operationally defined as all combustible, non-carbonate that can be collected on a filter whose pore size is 0.7 μm but can also effectively retain most particles >0.2 μm²,⁶⁹. Hence, POM in our samples potentially contains phytoplankton, zooplankton, heterotrophic bacteria, and terrestrial components. The concentration of POC and PN in the DCM layer of the SYS varies from 27.7 to 150.0 μg L^‒1 (on average 60.1 ± 25.0 μg L^‒1) and 5.5 to 26.4 μg L^‒1 (12.1 ± 4.8 μg L^‒1) (Supplementary Data 1), respectively. The spatial distribution of POC and PN concentrations resemble each other, with higher values on the onshore side of the temperature front than that on the offshore side (Fig. 1c). Significant covariations between these two variables are observed, but with differences in regression from the stable region to cyclonic gyre (Fig. 2a). The nearly zero intercepts on PN axis suggest that the measured nitrogen in the suspended POM is mainly in the organic form. This facilitates the molar C/N ratio to distinguish sources of POM from terrestrial plants (>12)⁷⁰ to marine plankton (6‒9)⁶⁵, and bacteria (2.6‒4.3)⁷¹. The low C/N, ranging from 4.5 to 7.7 (5.8 ± 0.8; Supplementary Data 1 and Fig. 1c) observed in the DCM of the SYS, indicates a non-terrestrial source. Most samples show a C/N ratio close to and below the Redfield ratio (Fig. 2a), indicating that the POM is predominantly derived from the marine plankton assemblages^72,73.

Fig. 2: Marine phytoplankton source of suspended particulate organic matter.

a Relationships of POC vs. PN concentrations in three groups of DCM stations. Note that almost all of the data points fall below the line of the Redfield ratio⁶⁶. b The covariation of POC and Chl a concentrations also indicates the predominant source of marine plankton in the POM at the DCM layer of SYS. Open diamonds in b were not included in regressions. The numbers within parentheses of the fitted equation represent standard errors.

The relationships between concentrations of POC and Chl a also differ between the stable region and cyclonic gyre (Fig. 2b). Both regressions show large correlation coefficients indicating that a large part of the POM in the DCM layer of the SYS is associated with Chl a and so with living phytoplankton^74,75,76. The intercept on the POC axis is assumed to represent the non-photosynthetic component of the POM, which includes both living (heterotrophic) and non-living detritus; this fraction is generally high in both stable region and cyclonic gyre (Fig. 2b). Indeed, scanning electron microscope (SEM) images of two stations (H12 and H15) representative of the stable region and cyclonic gyre show that many of the visible phytoplankton cells (main diatom) are broken (Supplementary Figs. 2 and 3). This may explain the considerable amount of detritus from phytoplankton. The Redfield-like C/N indicates most detritus is autochthonous and without degradation, which in general produces a high C/N due to preferential loss of nitrogen (N) over carbon (C) during OM degradation^77,78. Therefore, the measured POM is mainly of mixed phytoplankton assemblages. Moreover, the average phytoplankton assemblage-specific C/Chl a ratios inferred from the slopes of linear regressions between POC and Chl a are close to the ratio normally found in phytoplankton in nutrient-rich waters, c. 30^79,80. In fact, the phytoplankton C/Chl a ratio has been reported in a large range (6‒ > 400) due to influences of environmental factors such as nutrients and light^76,79,81. The larger C/Chl a ratio in the cyclonic gyre (35.9) than in the stable region (28.4) indicates a more efficient photosynthetic activity of phytoplankton inhabiting the DCM layer in the cyclonic gyre⁸⁰. This is consistent with a higher NH₄⁺/DIN ratio (Supplementary Fig. 4), as cells grown on NH₄⁺ have a higher affinity for CO₂ and increased photosynthetic CO₂-use efficiency than those grown on NO₃^‒; this NH₄⁺-induced high CO₂ affinity has been thought to be related to induction/activation of CCMs^26,82.

Overall, the stable region and cyclonic gyre are distinguished from one another in terms of two physiological indicators of phytoplankton assemblages: C/N and C/Chl a. This difference is likely a result of variations in phytoplankton community composition associated with distinct hydrological dynamics. It has been reported that picophytoplankton, composed of Synechococcus and picoeukaryotes, dominated the phytoplankton community in both cell abundance and carbon biomass (>90%) at the DCM layer of the central SYS (east of 122.5°E)⁴⁷, which belongs to the stable region in the present study, but less was known on the phytoplankton species in the cyclonic gyre. Nonetheless, studies in the SYS found convergence effects of temperature fronts on nutrients⁸³ and biology, such as phytoplankton, zooplankton, bacterioplankton, and anchovy eggs^45,46, suggesting a biological enhancement and potentially high biodiversity in the cyclonic gyre⁴⁴. The ancillary data from SEM imaging indicate that large-celled phytoplankton (mainly diatoms) were clearly visible in the cyclonic gyre (Supplementary Fig. 3), but a very limited amount was observed in the stable region (Supplementary Fig. 2). This abundant diatom along the front current in the cyclonic gyre is consistent with observations and simulations that revealed the predominance of diatom in meso- and submesoscale structures such as fronts⁸⁴. Furthermore, the larger amount of diatom likely accounts for the lower C/N in the cyclonic gyre (4.5‒6.2; 5.4 ± 0.6 on average) than that in the stable region (4.5‒6.8; 6.0 ± 0.6 on average) (Supplementary Data 1 and Figs. 1c and 2a), because diatom is capable to store large amounts of N^85,86 and may produce a low C/N ratio in NO₃^‒-replete environments⁸⁷. Note that although the SEM images show that it is difficult to observe picophytoplankton in both stable regions and cyclonic gyre, this cannot faithfully indicate a relatively low concentration of this microbe present in the filters. This is because the soft and smaller-sized biology do not tend to preserve well during SEM processing (e.g., freeze-drying and sputter coating)⁸⁸. In contrast, diatoms are protected from heavy alterations during SEM processing by the hard, siliceous cell wall, and are usually still identifiable under SEM observation⁸⁹. Therefore, we tend to suggest that the picophytoplankton (composed of Synechococcus and picoeukaryotes)⁴⁷ likely dominates in the stable region, while a greater proportion of large-celled phytoplankton (mainly diatom) presents in the cyclonic gyre at the DCM layer of SYS. The low C/N in the picophytoplankton-dominated stable region is clearly different from the reported constant and high C/N (c. 9) of picophytoplankton living in N-depleted environments⁸⁷. Indeed, the biomass C/N has been proposed as an indicator of nutrient supply to phytoplankton^90,91,92, and the low C/N (< 6.63 of Redfield ratio⁶⁶) observed in the DCM of SYS is indicative of N enrichment. This is well consistent with high stoichiometry of DIN/P by vertical supply from deep water (>16), the main nutrient source of phytoplankton at the DCM layer of SYS, as we discussed earlier in the subsection Dynamic Oceanographic Conditions.

Temperature other than [CO₂aq] dependent δ¹³C_POM

In the inferred absence of major contributions of terrestrial OM, δ¹³C_POM are representative of δ¹³C_phyto of locally predominant phytoplankton^{33,56,57,93,94}, which fuel the entire ocean food web, because heterotrophic processing and decomposition of OM have no or very limited effects on carbon isotopic fractionation^78,95. Analogous to the large gradient of seawater temperature, δ¹³C_POM at the DCM layer varies widely from ‒29.9 to ‒19.8‰ (average ‒26.0 ± 2.4) (Supplementary Data 1 and Fig. 1c). Around 10‰ variability of δ¹³C_POM in such a dynamic marginal sea is quite large, sharing approximately 40% of the 27‰ variation in plankton-derived POM observed globally in the modern ocean^9,10,11. A significant linear relationship between δ¹³C_POM and temperature is observed for the whole data set (R² = 0.39; p < 0.0001; n = 38), and the correlation can be improved further when samples in the stable region and cyclonic gyres are separated (Fig. 3a).

Fig. 3: Association of δ¹³C_POM with temperature and [CO₂aq].

a and b are for POM at the DCM layers of the southern Yellow Sea, and b is only for stations where concurrent measurement of [CO₂aq] is available (red pins in Fig. 1a). c, d are same plots for compiled data of POM in the surface water of SW Indian Ocean and Southern Ocean; data in c, d are derived from Francois et al.⁹³ and Kennedy and Robertson⁹⁴, respectively. [CO₂aq] data derived from Kennedy and Robertson²¹ were converted from μmol kg^‒1 to μmol L^‒1 by multiplying 0.001×density (kg m^‒3). The apparent trend of δ¹³C_POM with [CO₂aq] in b may not reflect the availability of more data in the temperature dataset, since [CO₂aq] wasn’t concurrently determined with temperature and δ¹³C_POM for each station in the DCM of SYS. However, data in the warm SW Indian Ocean (dark gray in c) apparently show the presence of a robust linear relationship between δ¹³C_POM and temperature under a nearly constant [CO₂aq]. The numbers within parentheses of the fitted equation represent standard errors.

Contrary to the hypothesis that the strong positive correlation between δ¹³C_POM and temperature is resulting from a well-defined solubility relationship between [CO₂aq] and temperature^38,39,40,41, we find that the concurrently obtained [CO₂aq] (14.3‒31.9 μmol L^‒1, average: 22.7 ± 4.3 μmol L^‒1; Supplementary Data 1) doesn’t show significant correlation with seawater temperature (Fig. 3b). Rather, temperature sets an upper limit of [CO₂aq] with only five samples in the stable region being [CO₂aq]-saturated with respect to its potential as constrained by temperature (Fig. 3b). This is reasonable because [CO₂aq] can change due to both solubility differences (temperature) and biological acitivity⁹⁴. The hypothesized inverse relationship between δ¹³C_POM and [CO₂aq] does present in these five samples whose [CO₂aq] follows the temperature-constrained line (Fig. 3b). Meanwhile, the regression line between δ¹³C_POM and temperature for these five samples (Fig. 3b) is similar to that for the overall stable region as shown in Fig. 3a. However, δ¹³C_POM is independent of the external [CO₂aq] for the remaining subset when [CO₂aq] deviates from the temperature-constrained line (Fig. 3b). Thus, it is apparent that the δ¹³C_POM at the DCM layer of SYS is more closely associated with temperature than with CO₂ levels.

The most striking aspects of our data set are that the previously hypothesized inverse relationship between δ¹³C_POM and [CO₂aq] occurs only when [CO₂aq] is controlled by temperature, but the robust linear correlation of δ¹³C_POM and temperature present independently of [CO₂aq] (Fig. 3a, b). Hence, which one, temperature or [CO₂aq], is the direct factor controlling δ¹³C_POM variation needs to be reconsidered. We are attracted to evaluate whether a temperature-dependent [CO₂aq] is accompanied in previous studies which concluded inverse relationships between δ¹³C_POM and [CO₂aq] and also evaluate whether the δ¹³C_POM is significantly correlated with temperature. We only consider studies in which δ¹³C_POM and [CO₂aq] were concurrently obtained on the same water samples other than calculating [CO₂aq] from pCO₂ in the atmosphere by assuming air-sea equilibrium because the assumption of air–sea equilibrium is not always true for large areas of modern ocean^10,93. Francois et al.⁹³ is one of the few earlier field researches in which [CO₂aq] was obtained along with δ¹³C_POM on the same water samples. Data in Francois et al.⁹³ showed that although there is a general negative relationship between δ¹³C_POM and [CO₂aq] in the southwestern (SW) Indian Ocean, these two variables are weakly correlated when [CO₂aq] is nearly constant and low (9.8‒11.7 μmol L^‒1; 10.9 ± 0.5 μmol L^‒1 on average), corresponding to a temperature of > c. 12 °C (Fig. 3c).

To test the temperature effect on [CO₂aq] and δ¹³C_POM, we divided their data into two groups based on temperature: <12 °C (n = 38; green circles in Fig. 3c) and >12 °C (n = 15; dark gray circles in Fig. 3c). They clearly display different regressions of [CO₂aq] against temperature (Fig. 3c). For samples with temperature <12 °C which shows a significant inverse relationship between δ¹³C_POM and [CO₂aq] (R² = 0.68, p < 0.0001), there is indeed a temperature-dependent [CO₂aq] (R² = 0.95, p < 0.0001) and also a robust correlation of δ¹³C_POM vs. temperature (R² = 0.69, p < 0.0001) (Fig. 3c). More strikingly but consistent with our data is that as for the temperature >12 °C subset, δ¹³C_POM are weakly correlated to [CO₂aq] (R² = 0.13, p = 0.1947) but significantly and strongly covaries with temperature (R² = 0.81, p < 0.0001; Fig. 3c), and the [CO₂aq] vs. temperature are weakly correlated (R² = 0.35, p = 0.0206; Fig. 3c). The presence of such a robust linear relationship between δ¹³C_POM and temperature for data grouped by nearly-constant [CO₂aq] strongly supports temperature having an impact on δ¹³C_POM values that is independent of the previously hypothesized covariation of [CO₂aq] and temperature. This idea is also supported by the culture experiment of different marine species that revealed an increasingly more negative δ¹³C_phyto with decreasing temperature under a constant [CO₂aq]⁹⁶. Notably, the warm (>12 °C) SW Indian Ocean displays a positive correlation between δ¹³C_POM and [CO₂aq] other than a previously hypothesized inverse relationship; inverse relationships are observed in the DCM of SYS (Fig. 3a) and in the cold (<12 °C) SW Indian Ocean (Fig. 3c). Meanwhile, contrary to the positive relationships between δ¹³C_POM and temperature (Fig. 3a, c), a negative covariation is observed in the relatively warm SW Indian Ocean (Fig. 3c). In fact, a similar but weak negative relationship also occurred in the POM at the DCM layers of the East China Sea (ECS), where the δ¹³C_POM is nearly constant (‒23.1 ± 1.5‰) and the seawater temperature is relatively high (19.1‒28.2 °C)⁵⁹ with [CO₂aq] relatively stable and <12 μmol/L⁹⁷. This contrast will enrich our understanding of the biogeochemical mechanisms involved in the temperature-dependent δ¹³C_POM variations, and this will be discussed below.

The more sensitive and specific thermal response than [CO₂aq] in δ¹³C_POM variations is further confirmed by the compiled data set of surface POM from the Southern Ocean⁹⁴, which shows a generally weak covariation between the simultaneously obtained δ¹³C_POM and [CO₂aq] for the whole data set (Fig. 3d). The authors⁹⁴ suggested that the large deviations from the hypothesized inverse relation of δ¹³C_POM and [CO₂aq] are due to the complex hydrological conditions and the associated biological differences, that are more important than [CO₂aq] in controlling the δ¹³C_POM variations. In this case, we find that temperature can better explain their hypothesis because the plot of δ¹³C_POM against temperature clearly distinguishes their data set into three hydrologically distinct areas from south to north (Fig. 3d), indicating temperature-induced relative isolation of environments and locally adapted primary producers⁹⁸. This is consistent with the fact that the ecology in the Southern Ocean is highly region-specific; the number of distinct biome zones is divided by fronts⁹⁹. This is also to some extent similar to the case in the DCM of SYS, where two different biomes (i. e. stable region and cyclonic gyre) are distinguished by the different δ¹³C_POM-temperature relationships.

Mechanistic explanations of temperature-dependent δ¹³C_POM

Despite the significance of temperature for many of the patterns and processes in oceanic environments^100,101, we have only a limited understanding of its impact on the stable carbon isotopic composition of marine plankton-produced POM, i.e., δ¹³C_POM. The mechanistic models describing carbon isotopic fractionation during photosynthesis would provide a process-based understanding of community-level δ¹³C_POM variations. The widely applied models^32,93,102 were originally from terrestrial C3 plants formulated by Farquhar et al.¹⁰³, which imply that pure passive diffusion of CO₂aq is the primary mechanism for inorganic carbon transport across the cell membrane. Given the evidence of the prevalence of CCMs in marine phytoplankton and that there is the entry of both CO₂aq and HCO₃^‒ in CCMs in most algae^25,26, we employ the model that permits the simultaneous uptake of both CO₂aq and HCO₃^‒ irrespective of the mechanisms by which inorganic carbon enters the cell^14,104:

$${\delta }^{13}{C}_{{{{{\rm{phyto}}}}}}={\delta }^{13}{C}_{{{{{\rm{CO}}}}}_{2}}-a{\varepsilon }_{{{{{\rm{CO}}}}}_{2}-{HC}{O}_{3}}-{\varepsilon }_{f}+{\varepsilon }_{f}\frac{\mu }{{F}_{t}}$$

(1)

δ¹³C_CO2 is the δ¹³C of dissolved CO₂ in the external medium, a likely source of inorganic carbon taken up by phytoplankton. a is the fractional contribution of HCO₃^‒ to total inorganic carbon uptake (F_t), and ${\varepsilon_{{{{{{\rm{CO}}}}}}_2-{{{{{\rm{HCO}}}}}}_3}}$ is the equilibrium discrimination between CO₂aq and HCO₃^‒ (${\varepsilon_{{{{{\rm{CO}}}}}_2-{{{{{\rm{HCO}}}}}_3}}}$ = 24.12–9866/T_k)¹⁰⁵. ε_f refers to the intrinsic fractionation of Rubisco, which is independent of temperature and varies among species^106,107. This CO₂aq can be from direct uptake of environmental CO₂aq (passive and/or active), or CA-catalyzed conversion of HCO₃^‒ to CO₂aq, both ways share similar δ¹³C_CO2 values. This is because there is a fractionation during the dehydration of HCO₃^‒ by CA¹⁰⁸, which is similar to the fractionation during isotopic equilibration between extracellular HCO₃^‒ and CO₂aq¹⁰⁵.

Accordingly, the considerable differences in responses of δ¹³C_POM to changes in temperature (0.23‒1.14‰ per °C; Fig. 3a, c, d) among these different oceanic regimes presumably reflect distinct temperature effects on δ¹³${{{{{\rm{C}}}}}}_{{{{{\rm{CO}}}}}_2}$ and/or the overall fractionation (ε_p) during photosynthesis relative to CO₂ (ε_p ≈ δ¹³${{{{{\rm{C}}}}}_{{{{{\rm{CO}}}}}_2}}$‒δ¹³C_POM), which includes C demand/supply represented by the ratio of growth rate relative to gross C uptake (μ/F_t), and the relative contribution of HCO₃^‒ to the gross flux of inorganic carbon into the cell. The robust linear correlation of δ¹³C_POM vs. temperature (Fig. 3a, c, d) implies a constant ε_f, or the same dominant species of phytoplankton across the temperature gradient within each of these five ecosystems since the variation in this term would result in variability in the intercept (but the intercept of δ¹³C_POM vs. temperature is not solely influenced by ε_f). This is consistent with our analysis of the homogeneous POM sources in the stable region, as well as the cyclonic gyre of the SYS, as discussed above. In addition, it is known that diatom dominates in the cold SW Indian Ocean⁹³ and Southern Ocean⁹⁴, while small-celled phytoplankton such as cyanobacteria, marine prochlorophytes, Emiliania huxleyi predominates in the oligotrophic warm SW Indian Ocean⁹³.

The ${\delta }^{13}{{{{{{\rm{C}}}}}}}_{{{{{{{\rm{CO}}}}}}}_{2}}$ is commonly calculated from δ¹³C of total DIC (δ¹³C_DIC) based on the relationship describing the isotopic fractionation between HCO₃^‒ (the main form of DIC in seawater) and CO₂aq as a function of temperature (${\varepsilon }_{{{{{{{\rm{CO}}}}}}}_{2}-{{{{{{\rm{HCO}}}}}}}_{3}}=24.12-9866/{T}_{k}$)¹⁰⁵. This calculation neglects the small isotopic fractionation between CO₃^2‒ and HCO₃^‒ (i.e., δ¹³C_DIC = δ¹³C_HCO3) and assumes isotopic equilibrium between HCO₃^‒ and CO₂aq⁹³. Unfortunately, the in situ δ¹³C_DIC was not measured in both the DCM of SYS and Southern Ocean⁹⁴. Similar studies that lack measurement of δ¹³C_DIC usually assume a constant δ¹³C_DIC or δ¹³C_CO2⁹⁴, which would correspondingly result in a temperature-dependent ${\delta }^{13}{{{{{{\rm{C}}}}}}}_{{{{{{{\rm{CO}}}}}}}_{2}}$ of c. 0.12¹⁰⁵ or 0 ‰ °C^‒1. However, the error associated with this approach is difficult to evaluate, and therefore, we are not going to discuss further the temperature effects on ${\delta }^{13}{{{{{{\rm{C}}}}}}}_{{{{{{{\rm{CO}}}}}}}_{2}}$ and, consequently, the overall fractionation during carbon fixation in these two areas. Instead, with the measurement of δ¹³C_DIC by the authors⁹³, we assess the temperature effects on ${\delta }^{13}{{{{{{\rm{C}}}}}}}_{{{{{{{\rm{CO}}}}}}}_{2}}$ and consequently ε_p in the SW Indian Ocean. The results show increased ${\delta }^{13}{{{{{{\rm{C}}}}}}}_{{{{{{{\rm{CO}}}}}}}_{2}}$ of 0.20 and 0.08 ‰ °C^‒1 with increasing temperatures in the cold and warm areas, respectively (Supplementary Fig. 5). The large deviation from 0.12‰ °C^‒1 associated with a constant δ¹³C_DIC is resulting from the fact that δ¹³C_DIC in these areas also covary with temperature (0.08 and ‒0.03 ‰ °C^‒1, respectively) (Supplementary Fig. 5). Clearly, the response of δ¹³C_POM to temperature in both warm (0.45 ‰ °C^‒1) and cold areas (‒0.44 ‰ °C^‒1) (Fig. 3c) cannot be solely accounted for by their temperature-dependent ${\delta }^{13}{{{{{{\rm{C}}}}}}}_{{{{{{{\rm{CO}}}}}}}_{2}}$. This indicates the important role of temperature in the overall isotopic fractionation, and this is confirmed by the strong relationship of ε_p and temperature (Supplementary Fig. 6). The relative importance of ${\delta }^{13}{{{{{{\rm{C}}}}}}}_{{{{{{{\rm{CO}}}}}}}_{2}}$ and ε_p on the variations of δ¹³C_POM are further estimated to be c. 43.9% (18.7%) and 56.1% (81.3%), respectively in the cold (warm) area.

Growth rate (μ), a potentially key variable in the fractionation processes^15,17,18,22, can be evaluated from POC in all five ecosystems, as POC is proportional to growth rate and primary production³³. The ratio of POC/[CO₂aq], therefore, provides a proxy variable for phytoplankton C demand/supply ratios (μ/F_t). During pure passive diffusion and even with some amount of CCMs, the gross flux of inorganic carbon into the cell is proposed to be proportional to [CO₂aq] in the external medium^{14,18,102,109}. With the exception of the warm SW Indian Ocean, the temperature effects on δ¹³C_POM through associated changes in growth rate are generally weak, as evidenced by the large scatter in the δ¹³C_POM (and temperature) vs. POC relationships (Fig. 4a–c). Despite this, tight relationships between δ¹³C_POM and POC/[CO₂aq] are observed in both the Synechococcus sp.-dominated stable region of SYS and the diatom-dominated cold SW Indian Ocean. These results underscore the comparable influences of both growth rate and C supply on δ¹³C_POM in these two regions. However, the chemostat culture studies¹⁸ showed that the fractionation of Synechococcus sp. has a much less sensitive relationship (essentially flat, with a small ε_p value) with the ratio of μ/[CO₂aq]. This difference may be attributed to the higher rates of CCMs in N-limited chemostat, fueled by elevated ATP/e^‒ ratios^15,82, than in the N-enriched, together with a CCM-unfavorable relatively low light intensity, in the DCM of SYS. Because the direct uptake of HCO₃^‒ is the major mechanism of CCMs in cyanobacteria¹¹⁰, when HCO₃^‒ is the major carbon source entering Synechococcus sp. cell, the fractionation is expected to remain constant and small^14,111. This explanation is consistent with Brandenburg et al.¹⁹, who recently found that the ε_p-μ/[CO₂aq] relationships vary significantly between species and groups but also as a result of differences in culture conditions. In addition, the presence of a significant linear correlation between δ¹³C_POM of mixed phytoplankton assemblages and a proxy variable for C demand/supply (Fig. 4a, b) implies a nearly constant physiological property such as cell size, variation of which confounds the linear relationship between these two variables^18,22. Consistently, constant C/Chl a ratios, a physiological proxy of the average phytoplankton^76,79, are observed in both stable regions of SYS (Fig. 2b) and the SW Indian Ocean (Supplementary Fig. 7). Moreover, a greater proportion of large-celled phytoplankton tends to have higher C demand and thus magnifies the temperature effects on δ¹³C_POM. This may be partly responsible for the larger coefficients in the cyclonic gyre (0.62‰°C^‒1) relative to the stable region (0.23‰ °C^‒1) (Fig. 3a). By contrast, the large variability of POC vs. Chl a in the Southern Ocean (R² = 0.02, p = 0.6009; Supplementary Fig. 7) provides evidence for the heterogeneity of phytoplankton physiology, which likely caused the absence of a high correlation between δ¹³C_POM and POC/[CO₂aq] (Fig. 4c). Besides, this phytoplankton diversity can potentially lead to changes in the proportional relationship between POC and growth rate and therefore decreases the reliability of POC/[CO₂aq] to reflect the actual C demand/supply.

Fig. 4: Mechanisms of temperature-dependent δ¹³C_POM.

a Correlation of δ¹³C_POM vs. POC concentrations and [CO₂aq]-normalized POC, a proxy variable for carbon (C) demand/supply ratios, as well as [CO₂aq]-normalized POC vs. temperature, POC vs. temperature in the DCM of SYS. Due to the small number of [CO₂aq] data available, the specific relationship of δ¹³C_POM and temperature with POC/[CO₂aq] is not evaluated in the cyclonic gyre of SYS. b, c Same plots for POM in the surface water of SW Indian Ocean and Southern Ocean for comparison; data in b, c are derived from Francois et al.⁹³ and Kennedy and Robertson⁹⁴, respectively. Open squares in a were not included in the regression. The numbers within parentheses of the fitted equation represent standard errors.

Despite the unavailability of temperature effects on ${\delta }^{13}{{{{{{\rm{C}}}}}}}_{{{{{{{\rm{CO}}}}}}}_{2}}$, the relatively large effect of temperature on δ¹³C_POM in the Southern Ocean (1.14‰ °C^‒1; Fig. 3d) indicates that there is a temperature-dependence of the uptake ratio of HCO₃^‒ to gross C because the ${\varepsilon }_{{{{{{{\rm{CO}}}}}}}_{2}-{{{{{{\rm{HCO}}}}}}}_{3}}$ is nearly constant (‒11.2 to ‒11.7‰) within the small temperature gradient (2.5‒6.0 °C, Fig. 3d). Specifically, the model equation indicates that this uptake ratio should be a positive linear function of temperature. This assumes that the above model equation captures all of the dominant processes influencing δ¹³C_POM signatures. Although the effect of temperature changes on the relative contribution of HCO₃^‒ to total carbon uptake in marine phytoplankton has not been studied well, there are reports showing that this uptake ratio generally increased with decreasing [CO₂aq] in the growth medium, yet strongly differed between species^21,112,113. The temperature dependence of [CO₂aq] in the Southern Ocean (63.6‒59.4°S) (Fig. 3d) well supports the idea that temperature positively covaries with the uptake ratio of HCO₃^‒ to gross C in diatom in this region. This, in turn, indicates that the observed inverse correlation of δ¹³C_POM with [CO₂aq] in the Southern Ocean (Fig. 3c) may be caused by the gradual induction of HCO₃^‒ uptake upon an increase in temperature or a decline in CO₂aq supply. Based on the limited data, no decision can be made about the temperature effects on the uptake ratio of HCO₃^‒ to gross C in the DCM layers of SYS and the surface water in the SW Indian Ocean.

Generally, below optimum growth temperatures, growth rates of phytoplankton tend to be positively correlated with temperature^101,114. This trend is due to the enhancement of enzymatic activities of Rubisco and other enzymes related to photosynthesis and depends on the assumption that the ecosystems are of high nutrient availability^92,100,115. Apparently, the negative linear relationship between δ¹³C_POM (and ε_p) and temperature in the warm (>12 °C) surface water of SW Indian Ocean (Fig. 3c) is mainly resulting from the decrease of growth rate (and C demand/supply) with increasing temperature (Fig. 4b). The robust inverse relationship between growth rate and temperature can be explained by enzyme kinetics during carbon fixation as described by Michaelis-Menten equation ($\mu =\frac{{V}_{{\max }}.C}{{K}_{0.5C}+C}$), in which both maximum C fixation capacity (V_max) and half-saturation constants (K_0.5C) increase with temperature⁹². In the N-starved warm SW Indian Ocean, phytoplankton cells may downregulate the cellular abundance of Rubisco and hence suppress the response of the V_max to temperature increase^15,22. As an alternative strategy to sustain the photosynthesis rate, the CCMs in phytoplankton are therefore essential in the N-limited conditions^15,22, and any changes of growth rate might occur primarily through changes in the CO₂ affinity of Rubisco (i. e. K_0.5C)¹¹⁶. This will lead to a robust inverse correlation of POC and temperature in the N-depleted warm SW Indian Ocean (Fig. 4b). By contrast, in the N-replete systems such as the DCM of SYS and cold SW Indian Ocean, increased cellular Rubisco content may be a viable mechanism to increase V_max^22,116. In addition, the [CO₂aq] in the warm SW Indian Ocean is close to the critical [CO₂aq], that is, roughly 10 μmol L^‒1, below which the uncatalyzed passive diffusive flux of CO₂ is limited and therefore active CCMs is induced^14,16,18,61. Moreover, the correlation between δ¹³C_POM (ε_p) and POC is better than either δ¹³C_POM (ε_p) vs. POC/[CO₂aq] (Fig. 4b and Supplementary Fig. 6) or δ¹³C_POM (ε_p) vs. [CO₂aq] (Fig. 3c and Supplementary Fig. 6), indicating CO₂-insensitive photosynthesis due likely to a great involvement of active uptake of inorganic carbon in the warm SW Indian Ocean. All these results agree that the transport of DIC into the phytoplankton cell is primarily by CCMs (or efficient CCMs) in the warm SW Indian Ocean that are probably induced by low [CO₂aq] and N-depletion^{26,110,117,118}. For phytoplankton possessing efficient CCMs, low temperature reduces the energetic requirement for the CCMs due to a decrease of the CO₂ requirement of Rubisco reflected by K_0.5C (or increasing CO₂ affinity of Rubisco) and CO₂ leakage¹¹⁹. This may produce an increasing efficiency of CCMs with decreasing temperature and, therefore, a negative relationship of δ¹³C_POM vs. temperature. By contrast, for marine ecosystems where carbon transport into the cell is dominated by passive diffusion, the decrease of K_0.5C at low temperatures would reduce the necessity for CCMs to maintain the supply of CO₂aq to the active site of Rubisco⁸². This downregulation of CCMs with decreasing temperature will result in a positive trend between δ¹³C_POM and temperature¹¹⁸. The linear correlation of δ¹³C_POM (mainly ε_p) vs. temperature, to some extent, reflects the fine-scale tuning of CCMs due to temperature variations.

In fact, a prominent maximum δ¹³C_POM in the small-celled phytoplankton dominated and NO₃^‒ starving warm SW Indian Ocean is coinciding with a maximum in POC (Fig. 4b) and Chl a (Supplementary Fig. 7), but with a minimum in [CO₂aq] (Fig. 3c). In addition, by performing a separate linear regression of POC vs. Chl a for stations with NO₃^‒ of 0 μmol L^‒1 in the warm SW Indian Ocean (red color in Supplementary Fig. 7b), we find that the POC is highly correlated with Chl a (R² = 0.97, p < 0.0001; n = 7), accompanied by a significantly high C/Chl a (647) and a very small amount of non-photosynthetic component in POC (9.5 μg L^‒1) (Supplementary Fig. 7b). This also suggests a phenomenon of phytoplankton bloom termination due to NO₃^‒ exhaustion. All these results are indicative of phytoplankton bloom termination due to NO₃^‒ depletion^31,113. The shift from passive diffusion or inefficient CCMs to efficient CCMs, especially during bloom, is deemed the most effective strategy to overcome carbon limitation and guarantees efficient carbon fixation^20,24,120. The capability of adjusting the CCMs activity and efficiency to the actual demand in marine phytoplankton may, in part, explain the high productivity related ¹³C enrichment in the coastal upwelling stations of the SYS (H16: ‒20.7‰ and H17: ‒19.8‰; Fig. 1c), which are deviating from the regressions of δ¹³C_POM against temperature in both stable region and cyclonic gyre (Fig. 3a). Additionally, due to increasing drain of available CO₂ in ambient water, isotopic disequilibrium between HCO₃^‒ and CO₂aq would lead to the assimilation of inorganic carbon whose δ¹³C has been shifted to that of marine HCO₃^‒, thus producing a more positive δ¹³C_POM^121,122. By comparison, a slow growth rate would thus favor the full expression of isotopic discrimination by Rubisco^119,121.

Despite the primary role of [CO₂aq] in regulating δ¹³C of aquatic POM has been questioned, less study provides a parameter that shows systematic effects at least as strong as [CO₂aq]. Based on partial correlation analysis, Wang et al.¹²³ proposed that pH rather than [CO₂aq] determines the carbon isotopic fractionation during photosynthesis in lake phytoplankton. However, we find through reanalysis that the correlation of carbon isotope fractionation (${\varepsilon }_{{{{{{{\rm{HCO}}}}}}}_{3}-{{{{{\rm{POM}}}}}}}$ and ${\varepsilon }_{{{{{{{\rm{CO}}}}}}}_{2}-{{{{{\rm{POM}}}}}}}$) to temperature in their data is even higher than to pH (Supplementary Fig. 8), which is supported by our partial correlation analysis (Supplementary Table 1). In addition, the correlation of δ¹³C_POM to temperature is higher than both ${\varepsilon }_{{{{{{{\rm{HCO}}}}}}}_{3}-{{{{{\rm{POM}}}}}}}$ and ${\varepsilon }_{{{{{{{\rm{CO}}}}}}}_{2}-{{{{{\rm{POM}}}}}}}$ to temperature (Supplementary Fig. 8 and Supplementary Table 1), showing the advantage of temperature in directly regulating the δ¹³C_POM signals. This may result from the temperature effects on inorganic carbon isotopic values¹⁰⁵. These results further confirm our idea that the effect of temperature on phytoplankton-produced δ¹³C_POM variations is stronger than factors that receive far greater attention, e.g., [CO₂aq] and pH. Similarly, recent studies provide evidence that ocean warming rather than acidification controls the coccolithophore calcification¹²⁴ and enzyme (e.g., CAs) activities in some marine phytoplankton¹²⁵.

In summary, both new data from the DCM layer of the SYS and a reanalysis of selected high-latitude δ¹³C_POM–[CO₂aq] and δ¹³C_POM–temperature paired data indicate that temperature other than [CO₂aq] more strongly determines the oceanic δ¹³C_POM, measured on POM of mixed phytoplankton assemblages. The similarity from very different oceanic settings could be taken as evidence for a common driving force of temperature for determining marine plankton-produced δ¹³C_POM in natural environments. We find that the predominance of passive diffusion and CCMs, respectively, determines the observed positive and negative δ¹³C_POM-temperature relationships. We, therefore, propose that the role of [CO₂aq] in δ¹³C_POM variations is mainly to regulate CCMs activity, while temperature-dependent metabolism strongly mediates δ¹³C_POM irrespective of inorganic carbon acquisition modes. This perspective is helpful to better understand and predict the effect of global warming and ocean acidification on marine ecosystem functioning. It is mechanistically reasonable that temperature imposes such a fundamental and inevitable constraint on the community level δ¹³C_POM variations, considering that temperature has multiple interactive effects with other properties such as [CO₂aq], pH and nutrient concentrations, and on phytoplankton processes such as growth, physiology, and photosynthesis^98,100,101. As a consequence, compared to [CO₂aq], the temperature can reduce such phytoplankton-driven variability in deviating their linear relationships with δ¹³C of the bulk organic fraction of marine plankton in natural assemblages. However, we are not suggesting that δ¹³C_POM can be used to reconstruct seawater temperature, because this prediction must assume that the relationship between δ¹³C_POM and temperature is constant and predictable. Instead, we tend to propose that the strong linear relationship between δ¹³C_POM and temperature is a phenomenological representation of the impact of temperature on δ¹³C_POM, which encompasses broad physiological differences among taxa adapted to diverse thermal environments¹⁰¹. In addition, differential temperature scaling on processes regulating δ¹³C_POM values provides a mean to distinguish between the relative importance of variation in C demand/supply, ${\delta }^{13}{{{{{{\rm{C}}}}}}}_{{{{{{{\rm{CO}}}}}}}_{2}}$, and ${a\varepsilon }_{{{{{{{\rm{CO}}}}}}}_{2}-{{{{{{\rm{HCO}}}}}}}_{3}}$ on δ¹³C_POM (and ε_p). Based on the large deviations from the hypothesized inverse relationships of δ¹³C_POM and [CO₂aq], pioneers^10,93,94 realized that factors linked to intracellular physiological processes could be more important than [CO₂aq] in establishing the marine plankton produced δ¹³C_POM, but the potentially responsible parameters have not been revealed yet. We provide an alternate perspective that temperature principally links intracellular physiological and photosynthetic processes, resulting in a sensitive and specific δ¹³C_POM response, which can obscure the simple relationship between δ¹³C_POM and [CO₂aq]. This brings a promising approach to the decades-old problem in marine biogeochemistry.

Future studies with simultaneous measurements of in situ δ¹³C_DIC and taxon identification on the same water samples of POM, as well as the improved knowledge of the intrinsic fractionation of Rubisco in various species of phytoplankton, are helpful for an in-depth understanding of the mechanisms driving δ¹³C_POM variability at species and community level in natural environments. Better estimation of ${\delta }^{13}{{{{{{\rm{C}}}}}}}_{{{{{{{\rm{CO}}}}}}}_{2}}$ by considering isotopic disequilibrium between HCO₃^‒ and CO₂aq could also greatly improve the distinction of relative impact between [CO₂aq] and processes regulating ε_p on δ¹³C_POM variations. It is also important to study the δ¹³C_POM of classified phytoplankton by separating different contributions through a variety of complementary techniques, such as genetic/molecular tools or cell sorting, or size-class sorting using filters with different pore sizes or microscopy. In addition, molecular biomarkers known to be derived from restricted taxonomic sources may also help minimize issues associated with secondary biological processing or terrestrial inputs and therefore improve the likelihood that the measured isotopic signature reflects the fractionation due to photosynthesis by marine phytoplankton rather than some other processes.

Methods

Hydrographic data collection

This investigation was conducted on the Chinese side of the southern Yellow Sea (SYS) (Fig. 1a) between 29 August and 3 September 2017 on the research vessel Dongfanghong 2. Hydrological data (temperature and salinity) and Chl fluorescence concentration were measured in situ by a calibrated conductivity-temperature-depth rosette (Sea-Bird SBE 911+) fitted with detectors for Chl fluorescence. The DCM layers were identified from the Chl fluorescence profiles. Seawater samples for suspended POM, DIC, and TAlk were collected from the DCM layers through Niskin bottles mounted on the conductivity–temperature–depth rosette.

POM measurement

Suspended particles for organic carbon and nitrogen, as well as for Chl a measurement, were obtained by immediately filtering seawater through 0.7 μm Whatman glass fiber filters (GF/F) under an ultimate pressure of 0.08 MPa to avoid the rupturing of phytoplankton cells. The volume of filtered seawater, 0.95–10.95 L for each filter, depended on the particle contents. In many cases, the amounts of particles on the filters were so high that seawater could not flow through. At least two filters were collected at each site for separate downstream analysis (Chl a and POC/PN/δ¹³C_POM). At some stations, filters with different diameters (25 mm and 47 mm) were used, according to the seawater amount shared onboard. All filters were pre-combusted at 450 °C for 4 h in a muffle furnace to remove the background carbon and wrapped in aluminum foil. After filtration, these filters were folded in half without rinsing, and wrapped again in aluminum foil. They were immediately stored at ‒20 °C in a freezer onboard until shore analysis at the State Key Laboratory of Marine Environmental Science, Xiamen University, where filters with suspended particles, except those for Chl a measurement, were firstly freeze-dried by a CHRIST Alpha 1–4 LDplus freezer-drier interfaced to a Vacuubrand RZ 6 vacuum pump.

For Chl a analysis, filters of 25 mm diameter were used. The Chl a retained on the filters was determined by the non-acidification method¹²⁶ using a Turner Trilogy fluorometer available in the Center of Major Experiment and Technology of Xiamen University. The fluorometer was previously calibrated with a Chl a standard (Sigma-C6144) by using 90% acetone; the concentration of the standard solution was determined spectrophotometrically (Ultraviolet Spectrometer, model 722) using an extinction coefficient of 87.67 l g^‒1 cm^‒1 at 664 nm against a 90% acetone blank. The filter was removed from the aluminum foil, and the whole filter was placed in the 5 ml centrifuge tube with a screw cap. 3.5 ml of 90% acetone was then added into the centrifuge tube by careful pipetting. These centrifuge tubes were then wrapped with aluminum foil to avoid light and placed in the freezer at ‒20 °C for 16–20 h. Before analysis, the sample extract was allowed to stand at room temperature for 30 min, and the fluorometer was allowed to warm up for 15 mins. The sample extracts were transferred from centrifuge tubes to the fluorometer cuvette, and the fluorescence of the sample extract was finally measured against a 90% acetone blank. The Chl a concentration for each sample was calculated based on the previously calibrated relationship of Chl a concentration and fluorescence. The detection limit for Chl a is 0.02 μg L^‒1.

For POC, PN, and δ¹³C_POM analysis, both 25 mm (n = 28) and 47 mm (n = 10) filters were employed. These filters were first removed from the aluminum foil. Half of the 47 mm or full 25 mm filter was then placed into the culture dishes, and several drops of 1 N HCl were added to each dish to cover the filter. They were allowed to react for 16 h to remove inorganic carbon (mainly carbonate). The carbonate-free samples were dried in an oven at 50 °C to remove the excess HCl. Then a half (i.e., a quarter of the 47 mm filter) or full (25 mm filter) of the carbonate-free filter, corresponding to 0.75–4.35 L filtered seawater, was then punched and packed in tin capsules. The POC, PN, and δ¹³C_POM were determined through an elemental analyzer (Elementar Analysensysteme GmbH) interfaced with a PDZ Europa 20–20 isotope ratio mass spectrometer at the Stable Isotope Facility of the University of California Davis in the USA. The stable carbon isotope ratios were presented as per mil deviation from standard-VPDB and expressed as $\delta {\,}^{13}{{{{{\rm{C}}}}}}=\left(\frac{{({\,}^{13}{{{{{\rm{C}}}}}}/{\,}^{12}{{{{{\rm{C}}}}}})}_{{{{{\rm{sample}}}}}}}{{({\,}^{13}{{{{{\rm{C}}}}}}/{\,}^{12}{{{{{\rm{C}}}}}})}_{{{{{\rm{standard}}}}}}}-1\right)\times 1000$ . During analysis, samples are interspersed with several replicates of four different laboratory standards (Glutamic Acid, Bovine Liver, Enriched Alanine, and Nylon 6). These laboratory standards have been previously calibrated against international reference materials, including IAEA-600, USGS-40, USGS-41, USGS-42, USGS-43, USGS-61, USGS-64, and USGS-65. A sample’s provisional isotope ratio is measured relative to a standard gas peak analyzed with each sample. These provisional values are finalized by correcting the values for the entire batch based on the known values of the included laboratory standards (https://stableisotopefacility.ucdavis.edu/carbon-and-nitrogen-solids). The standard deviation of δ¹³C_POM was 0.2‰, and the precision decreases when samples contain <100 μgC. Among 38 particulate samples measured in the present study, fourteen contain <100 μgC, but six of these fourteen show >90 μgC, all >67 μgC. Molar C/N ratios of POM were calculated from the measured POC and PN.

DIC and TAlk analysis

Seawater samples for DIC and TAlk analyses were stored in 60 mL borosilicate glass bottles and 140 mL high-density polyethylene bottles, respectively, and were then immediately mixed with 50 μL saturated HgCl₂. During DIC sample collection, seawater was overflowed by a full bottom volume, while TAlk samples were collected after several swing-washings without overflowing. These samples were sealed with screw caps and stored at room temperature before further analysis in the platform of Marine Carbon Chemistry in Shandong University. This sampling protocol is slightly different from Dickson et al.¹²⁷ guides but has received a favorable evaluation by Huang et al.¹²⁸. It was concluded that there were no differences between the results obtained using our procedure and those derived using the widely adopted sampling procedure recommended by Dickson et al.¹²⁷. The data quality of surface DIC and TAlk of this cruise has been evaluated by comparison between the calculated sea surface fugacity of CO₂ (fCO₂) and the underway measurement of sea surface partial pressure of CO₂ (pCO₂, nearly the same as the fCO₂ in usual sea surface settings)¹²⁹; this was discussed below. These seawater samples were firstly allowed to static sedimentation to obtain the supernatant for DIC and TAlk measurement. For DIC analysis, a commercial analytical system (model AS-C3, Apollo SciTech Inc., USA) was used. Briefly, a DIC sample of 0.5–0.9 mL was injected using a Kloehn^® digital syringe pump and acidified with 1.5 mL of 10% analytical-grade phosphoric acid. The extracted gas was quantitatively collected in internal pipes for final determination using an infrared CO₂ detector (model Li-7000, Li-Cor Inc., USA). For TAlk analysis, a water sample of 15–25 mL was determined at 25 °C by the Gran acidimetric titration using a semi-automated titrator (AS-ALK2, Apollo SciTech Inc., USA). During analysis, we used the certified reference materials from Dickson’s lab (Batch#156) (https://www.ncei.noaa.gov/access/ocean-carbon-acidification-data-system/oceans/Dickson_CRM/batches.html) for quality assurance, achieving a precision of ±2 μmol kg^‒1.

The [CO₂aq] was calculated based on pressure, temperature, salinity, and measured DIC and TAlk, using the program CO2SYS.XLS (version 24). The dissociation constants of carbonic acid and for HSO₄^‒ were used by Millero et al.¹³⁰ and Dickson¹³¹, respectively, in the calculation. Phosphate and silicate concentrations required by the program were from measurements that reported by Zheng and Zhai⁶⁵. The calculated [CO₂aq] was finally converted from a unit of μmol kg^‒1 into μmol L^‒1 by multiplying 0.001×density (kg m^‒3). We also calculated [CO₂aq]₀ by the same program, except that the phosphate and silicate concentrations were replaced by zero. The dissolved CO₂ concentrations calculated from these two methods show significant linear correlation (R² = 1.00, p < 0.0001; n = 21; Supplementary Fig. 9). By applying this regression equation, we derived the [CO₂aq] of sample HS1, of which the nutrient concentration was not measured.

To assess the quality of the calculated data set, we conducted the underway measurement of pCO₂ and also collected surface seawater for fCO₂ calculation using the same program as [CO₂aq] calculation mentioned above; this surface carbonate data set has been published by Wang and Zhai¹²⁹. The calculated fCO₂ was greater than the field-measured pCO₂ by 14 ± 28 μatm (n = 40) on average (mean ± S.D.), indicating less than 5% error relative to the 400 μatm, the approximate air-equilibrated pCO₂ level during our surveys¹²⁹. Note that this is for the surface water, and the error for the DCM sample is expected to be smaller due to the decrease of associated systematic error in subsurface water¹²⁹.

SEM imaging

Two filters (0.7 μm/25 mm) representative of the stable region (H12) and cyclonic gyre (H15) were selected to be imaged in the Field Emission SEM (FEI Quanta 650 FEG) in the Center of Major Experiment and Technology of Xiamen University. Particles were evenly distributed on the surface of the filters. Before SEM imaging, the selected filters were first cut in half and trimmed to fit the sample stub. The conductive tap was then used to mount the filter on the sample stub and to create a contact between these two. After sputtering with gold for 120 s, the sample stub was inserted into the sample holder in the vacuum chamber. SEM images were recorded at magnifications ranging from 127× to 74,059×.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.