The modified nutrient quota model can replicate the observed spatial patterns in particulate organic matter stoichiometry. We compare the model particulate flux at 100m to the surface POM C:N:P stoichiometry from the GO-POPCORNv2 database for each observational location where the model grid cell is determined using a least squares calculation24. We note the observational bulk POM contains heterotrophic bacteria and non-sinking detritus that are not in our model. The model broadly captures the latitudinal variations in organic matter stoichiometry along multiple cruise transects (Fig. 1, SFigs. 2–5). The largest mismatches occur where there are biases in the simulated nutrients. The Spearman's rank correlation coefficient (rs) for C:N, N:P, and C:P are 0.34, 0.28, and 0.31, respectively (SFig. 6a-c). Model phytoplankton Fe:C ratios are compared to field observations of individual cell iron to carbon21. The rs for the phytoplankton community Fe:C is 0.28 (SFig. 6d).
Inverse models diagnose elevated C:P ratios (> 140) and N:P ratios (> 20) in the subtropical gyres, and much lower ratios of C:P (< 100) and N:P (< 16) in regions with elevated surface phosphate concentrations (> 0.3 µM)22,25. Combining the regional mean C:P and N:P in export estimated by these two inverse studies gives a mean C:N of 8.5 for the North Pacific gyre, 6.7 for the South Pacific gyre, and 6.0 for the Southern Ocean25,26. Our model captures these large-scale patterns in C:N:P ratios of the sinking export flux (SFig. 6). The inferred global patterns are also broadly in agreement with the stoichiometry of surface POM in the POPCORN database. (Fig. 1, SFigs. 2–5,7)24. Limited observations make evaluating the Si:C ratios more difficult. The model captures the observed patterns of elevated Si:C in iron-limited regions with elevated surface dSi concentrations, and the low Si:C seen under low Si conditions both in situ and in laboratory studies27–29. Thus, our simple approach, dynamically linking phytoplankton stoichiometry to ambient nutrient concentrations, captures observed global-scale patterns in the stoichiometry of exported organic matter.
We compare a variable C:N:P:Fe:Si model simulation (VarAll) with a fixed-ratio model version (FixAll) to investigate how dynamic plankton stoichiometry influences marine biogeochemistry, in terms of the magnitude and spatial patterns of net primary production, sinking carbon export at 100m depth, air-sea CO2 flux, nitrogen fixation, and water column denitrification. Both models are able to replicate observations of surface nutrients, with a better fit for the VarAll simulation. Compared to World Ocean Atlas 2018, the rs for VarAll and surface NO3, PO4, and SiO3 are 0.91, 0.90, and 0.56, respectively. For FixAll, these values are 0.76, 0.91, and 0.32 respectively30. For dissolved iron, we compare primarily to data collected from the GEOTRACES project, supplemented with historical data compilations31–33. The rs for dissolved iron in the top 200m is 0.39 for the VarAll model and 0.31 for the FixAll model.
The models have similar net primary production (NPP), but the FixAll simulation significantly underestimates particulate organic carbon (POC) export and key nitrogen cycle fluxes (N fixation and water-column (WC) denitrification) compared to the fully variable simulation (Figs. 2A, 2B, 3). The VarAll model has a total integrated NPP of 58 PgC/yr, POC Export at 100m of 8.1 PgC/yr, N fixation of 214 TgN/yr, and WC Denitrification of 57 TgN/yr, which are close to or within the range of previous satellite and model-based estimates (NPP = 52–67 PgC/yr, POC Export = 5–10 PgC/yr, N Fixation = 126–223 TgN/yr, WC Denitrification = 56–73 TgN/yr)26,34–38. With fixed stoichiometry, NPP, POC Export, N Fixation, and WC denitrification decrease by 14%, 11%, 29%, and 39%, respectively (FixAll global fluxes: 50 PgC/yr, 7.2 PgC/yr, 153 TgN/yr, 35 TgN/yr).
The global export ratio (e-ratio, sinking POC/NPP) distribution in the VarAll model (SFig. 8) shows higher e-ratios in the high latitudes, > 0.175 in the Southern Ocean, and > 0.25 in the North Atlantic and North Pacific. When compared to the FixAll model, we see up to 0.05 increase in e-ratio in the high latitude North Atlantic, compared to up to a 0.05 decrease in the Southern Ocean, both of which are primarily driven by changes in POC export (Fig. 2C). In the Indian Ocean, which has increases in NPP and POC export throughout the basin, the e-ratio change shows a bimodal pattern, with the Arabian Sea and gyre regions showing increases in e-ratio, while the Bay of Bengal region e-ratio decreases.
The lower POC export in the FixAll model is driven by lower surface nutrient concentrations, but is further enhanced by a community dominated by small phytoplankton in the more oligotrophic regions (SFig. 9)39. When phytoplankton nutrient uptake ratios are fixed, growth-limiting nutrients are exported more efficiently, furthering nutrient stress in surface waters, favoring smaller phytoplankton and leading to decreases in NPP and POC export. Further, with variable stoichiometry, small phytoplankton may out-compete diatoms within the HNLC regions, resulting in increased small phytoplankton biomass and reduced diatom biomass. This leads to a decrease in Southern Ocean POC export without significant changes in NPP. Additionally, diatoms require silicon and are highly sensitive to silicon availability, which causes them to have significant reductions in biomass in low Si areas in the FixAll case. The only regions where diatoms have higher biomass in the FixAll case are upwelling regions where Si is returned to the surface. When diatoms can vary their Si uptake reducing their quotas, their distribution expands to occupy low-Si regions. This suggests that diatom Si:C acclimation is critical for explaining the global distribution of diatoms, preventing Si-limitation of growth over much of the lower latitudes.
Nitrogen fixation and water column denitrification increase under variable stoichiometry (Fig. 3). Nitrogen fixation increases globally, up nearly 300%, but particularly in the oligotrophic gyres where diazotrophs can reduce their quotas and maintain their growth (Fig. 3). With fixed ratios, surface phosphate declines in the North Atlantic and Indian basins and surface dissolved iron declines in the Pacific, increasing P and Fe stress for diazotrophs. Nitrogen fixation rates in the North Atlantic are particularly dependent on phosphate40. WC denitrification decreases 49% with fixed stoichiometry, with little change in the spatial pattern, due to decreases in diatom production and export over the Bay of Bengal and in the eastern equatorial Pacific. The volume of low oxygen (< 30 mmol/m3) waters decreases 42% with fixed stoichiometry.
Both models were run for 300 years with dynamic atmospheric CO2, initiated at 284ppm. Averaged over the final 20 years, the FixAll scenario had 313ppm atmospheric CO2 concentration, while the VarAll scenario had 296ppm atmospheric CO2. This indicates that the FixAll scenario is underestimating the capacity of the ocean carbon inventory when compared to the simulation with fully variable stoichiometry.