Geologic controls on phytoplankton elemental composition

Significance

The elemental composition of marine phytoplankton reflects their quality as a food source and regulates the flow of carbon, oxygen, and nutrients between the ocean, atmosphere, and rock reservoirs. Despite this importance, no systematic estimate exists of the geologic history of phytoplankton elemental composition, which is often regarded to have been constant through time, limiting our understanding of evolutionary patterns and global biogeochemistry. Here, using a biogeochemical model forced by geologic and evolutionary events (e.g., the colonization of the continents by plants), we find that the nutrient content of phytoplankton has increased over the last 550 million y of Earth history. This increase in the organic matter nutrient content has likely been important for the evolution of marine fauna.

Abstract

Planktonic organic matter forms the base of the marine food web, and its nutrient content (C:N:P_org) governs material and energy fluxes in the ocean. Over Earth history, C:N:P_org had a crucial role in marine metazoan evolution and global biogeochemical dynamics, but the geologic history of C:N:P_org is unknown, and it is often regarded constant at the “Redfield” ratio of ∼106:16:1. We calculated C:N:P_org through Phanerozoic time by including nutrient- and temperature-dependent C:N:P_org parameterizations in a model of the long-timescale biogeochemical cycles. We infer a decrease from high Paleozoic C:P_org and N:P_org to present-day ratios, which stems from a decrease in the global average temperature and an increase in seawater phosphate availability. These changes in the phytoplankton’s growth environment were driven by various Phanerozoic events: specifically, the middle to late Paleozoic expansion of land plants and the Triassic breakup of the supercontinent Pangaea, which increased continental weatherability and the fluxes of weathering-derived phosphate to the oceans. The resulting increase in the nutrient content of planktonic organic matter likely impacted the evolution of marine fauna and global biogeochemistry.

The elemental composition of marine primary producers (phytoplankton) is a key governing factor in the evolution of marine fauna and biogeochemical cycling on geologic (million-year and longer) timescales (1–3). As phytoplankton form the base of the marine food web, the molar ratio by which they incorporate essential nutrients, namely nitrogen and phosphorus, relative to carbon (C:N:P_org) reflects the quality of the phytoplankton as a food source. Low-quality food, which has high C:N:P_org, may limit the growth and reproduction of herbivores, affecting the dynamics and demographics of marine organisms at higher trophic levels (4–8). In contrast, high-quality organic matter with lower C:N:P_org allows the consumers to expend less energy on carbon respiration, leaving more energy to evolve traits that facilitate motility, grazing, and reproduction (1, 3, 8). Furthermore, together with the total availability of the limiting nutrient, C:N:P_org governs the amount of organic carbon produced in the photic layer by marine primary producers and the amount of dioxygen (O₂) consumed in the deep ocean by respiration. The dynamics of primary production at the ocean surface and respiration in the ocean interior govern the degree of ocean anoxia, and ultimately the burial of organic carbon in sedimentary rocks, which is responsible for the maintenance of O₂ in Earth’s atmosphere over geologic time (9–11).

Despite its evolutionary and biogeochemical importance, the geologic history of C:N:P_org is poorly constrained, and this ratio is often regarded to have remained constant at its present-day value throughout Phanerozoic Earth history. A part of the reason for this knowledge gap is the absence of a direct proxy for paleo-C:N:P_org. Preferential remineralization of phosphorus and nitrogen during burial in sediments alters the primary composition of fresh organic matter and precludes the use of C:N:P_org in sedimentary rocks as such a proxy (1, 12, 13).

Present-Day Variability in C:N:P_org

In the present ocean, C:N:P_org is often considered to have an average value of 106:16:1 (molar ratio), i.e., the “Redfield ratio” (14, 15). However, substantial spatiotemporal C:N:P_org variations have been shown to exist, some of which represent variations in the phytoplankton community composition (16–18). For example, C:P_org and N:P_org in the cold, nutrient-rich, high-latitude oceans are ∼78:1 and ∼13:1, and this may reflect the regional dominance of diatoms, which display lower-than-Redfield N:P_org and C:P_org (19–21). In contrast, the warm, nutrient-depleted (oligotrophic) midlatitude oceans are characterized by higher-than-Redfield C:P_org and N:P_org of ∼195:1 and ∼28:1, and this may reflect the regional dominance of cyanobacteria, which display higher-than-Redfield elemental ratios under nutrient-replete conditions (17, 22). Marine C:N:P_org may further deviate from the taxon-specific cellular stoichiometry in response to environmental stresses (23, 24). For example, C:P_org of some phytoplankton increases under phosphate scarcity, due to substitution of nonphosphorus membrane lipids for phospholipids (25). Irrespective of their exact cause, these large-scale variations in C:N:P_org seem to depend on the abiotic environment, that is, on the temperature and phosphate concentration of seawater (17, 26–28). Thus, a change in these environmental parameters in response to geologic drivers is expected to have altered the average C:N:P_org of the planktonic biomass over Earth history (2).

Phanerozoic Global Average Temperature and Surface–Ocean Phosphate Concentration.

Phytoplankton C:N:P_org ratios are thought to depend mostly on the temperature and/or phosphate concentration of the phytoplankton’s growth environment (24, 26, 27). Below we review the Phanerozoic trajectories of these variables in response to tectonic and evolutionary forcings, as predicted by a model for the long-term cycles of phosphate, carbon, O₂, and calcium, which is fully described and discussed elsewhere (29). Briefly, the model temperature is mostly affected by volcanic outgassing rates and continental weatherability, which control atmospheric CO₂ concentrations (SI Appendix, SI Text and Fig. S1). The model surface–ocean phosphate concentration mostly reflects a balance between a phosphate source from continental weathering and phosphate removal by burial in sedimentary rocks, the latter of which depends on the concentration of O₂ in seawater (SI Appendix, SI Text).

Paleozoic global average temperatures and surface–ocean phosphate concentrations were likely higher and lower than present, respectively (Fig. 1). The model seawater phosphate concentrations are in agreement with the few existing observational constraints and lower than those in most published long-term Earth-system models (SI Appendix, SI Text and Fig. S2). Both the high Paleozoic temperature and the low surface–ocean phosphate concentrations are an outcome of relatively low continental weatherability in the absence of land plants (29). The extensive root systems of vascular land plants increase fluid infiltration, the surface area of rock exposed to weathering, and the concentration of organic acids in soils, all of which increase continental weatherability (30, 31). Low continental weatherability means that a balance between the long-term carbon sources (outgassing of CO₂) and sinks (the temperature-dependent rate of continental and seafloor weathering) is achieved at a higher temperature (32, 33). Alongside the effect of continental weatherability on the global average temperature, the importance of seafloor weathering (relative to continental weathering) as a source of alkalinity to the ocean is high when continental weatherability is low. Continental weathering is a source of both alkalinity and phosphate, whereas under the oxidizing conditions characteristic of most of Phanerozoic time, seafloor weathering is a source of alkalinity only and possibly even a small phosphate sink (34). Thus, intervals of low continental weatherability, like the early Paleozoic, may have been characterized by lower rates of weathering-derived phosphate delivery to the ocean and generally lower phosphate concentrations (29). This effect may have been compensated by lower-than-present early Paleozoic deep-ocean O₂ concentrations (35, 36). Anoxic basalt weathering has been suggested to be a phosphate source similar to continental weathering (37), and in this case, early Paleozoic seawater phosphate concentrations may have been close to today’s, but the global average temperature would have still been high relative to today (SI Appendix, SI Text and Fig. S3).

Fig. 1.

Global temperature and surface–ocean phosphate concentration throughout the Phanerozoic (adapted from ref. 29). (A and B) The evolution of (A) temperature (K) and (B) surface–ocean phosphate concentration (µM P) obtained in ∼10⁶ default model simulations (5th to 95th percentiles of the results) where input parameters and time-dependent forcings were drawn from probability distributions that represent uncertainty in their values. Adapted from 29.

Several tectonic and evolutionary events shaped the Phanerozoic trajectories of global average temperature and seawater phosphate concentrations. The expansion of land plants (∼400 to 350 Ma) enhanced continental weatherability, leading to a decrease in the global average temperature and an increase in the rate of phosphate supply to the ocean from continental weathering. Assembly of the supercontinent Pangaea in the late Paleozoic and its tenure into the middle Mesozoic likely resulted in higher global average temperatures and lower surface–ocean phosphate concentrations—an outcome of the difficulty of delivering moisture to the continental interior and the resulting low continental weatherability. At the onset of the breakup of Pangaea (∼237 Ma) surface–ocean phosphate concentrations rose and global temperature decreased toward their present-day values (Fig. 1).

Here, we explore the effects of these Phanerozoic variations in the marine environment on the evolution of phytoplankton elemental composition. In the model described above, we implemented phosphate- and temperature-dependent C:N:P_org parameterizations (Materials and Methods) adopted from empirical global-scale relationships (26, 27) and meta-analysis of laboratory culture data (24). Our results suggest a decrease from high Paleozoic C:P_org and N:P_org to present-day Redfield ratios, which is driven by geologic and biological controls on Earth’s surface temperature and the seawater phosphate concentration.

Results and Discussion

The Evolution of Phytoplankton C:N:P_org.

The overall evolution of Phanerozoic C:N:P_org is a consequence of evolving global temperatures and seawater phosphate concentrations. The effect of these environmental parameters on model phytoplankton C:N:P_org is tested using three different empirical relationships (Materials and Methods): 1) a field-based phosphate–C:N:P_org relationship (26), 2) a field-based temperature–C:N:P_org relationship (27), and 3) a laboratory meta-analysis–based temperature–and phosphate–C:N:P_org relationships (24). Driven by the Phanerozoic evolution of seawater phosphate concentrations and global average temperatures (Fig. 1), all three parameterizations yield higher-than-Redfield C:P_org and N:P_org during the early Paleozoic (Fig. 2 A and B). For example, the temperature-dependent field-based parameterization 2 yields early Paleozoic C:P_org between ∼130 and 230 and N:P_org between ∼20 and 30 (95% of model results). If lower deep-ocean O₂ concentrations resulted in a phosphate source from seafloor weathering (e.g., ref. 37), then Paleozoic phosphate concentrations higher than our default scenario would result in lower C:P_org and N:P_org, but only in the phosphate-dependent parameterization 1. Paleozoic temperatures were likely higher irrespective of the role of seafloor weathering in the phosphate cycle, and so the two temperature-dependent parameterizations 2 and 3 consistently yield higher C:P_org and N:P_org (SI Appendix, Fig. S3). The high C:P_org values that we predict in the early Paleozoic are in broad agreement with estimates of Precambrian C:P_org of 300 to 400 (38), which were calculated on the basis of relations between cyanobacteria-dominated biomass C:N:P_org and ambient phosphate concentrations (39).

Fig. 2.

The evolution of phytoplankton C:N:P_org. (A and B) C:P_org (A) and N:P_org (B) (molar ratio) calculated from ∼10⁶ model simulations (5th to 95th percentiles of the results) where we draw ∼37 input parameters and 11 time-dependent forcings from probability distributions that represent uncertainty in their values. The elemental ratios were calculated using three parameterizations for the dependence of C:N:P_org on environmental parameters (Materials and Methods). Boxplots show the C:P_org and N:P_org of modern major phytoplankton groups (boxes, 25th, 50th, and 75th percentiles; whiskers, 99.3% of the data) growing under nutrient-replete conditions, plotted against their time of peak occurrence in the fossil record. (C) Occurrences of the different phytoplankton groups in the geologic record (Materials and Methods).

With all three parameterizations, C:P_org and N:P_org decreased when green algae colonized the continents (i.e., the evolution of land plants) and enhanced continental weatherability at ∼400 to 350 Ma. During the tenure of the supercontinent Pangaea (∼300 to 237 Ma), our model predicts an increase in C:P_org and N:P_org in response to the decrease in continental weatherability and the associated increase in the global average temperature and decrease in the seawater phosphate concentration (Fig. 1). As Pangaea broke apart, our model C:P_org and N:P_org decline to present-day Redfield values (Fig. 2 A and B). The ratios calculated on the basis of laboratory studies of phytoplankton growing under different degrees of phosphate and temperature limitation (parameterization 3) (24) are consistently highest. Field-based empirical relationships between C:N:P_org and sea-surface temperature (parameterization 2) (27) or seawater phosphate concentrations (parameterization 1) (26) yield intermediate and lowest C:N:P_org ratios, respectively. The Phanerozoic C:P_org and N:P_org median values and 95% confidence intervals calculated with all three parameterizations are provided in Zenodo (40).

We compared our calculated C:P_org and N:P_org with observed C:P_org and N:P_org of modern representatives of the major phytoplankton groups. The phytoplankton group-specific C:P_org and N:P_org values are derived from laboratory strains growing under nutrient-replete conditions (18), which show that red algae tend to thrive under nutrient-rich conditions and display lower C:P_org and N:P_org than their green and cyanobacterial counterparts (19–21). The group-specific C:P_org and N:P_org values are plotted against the group’s time of peak appearance in the fossil record (Fig. 2). Although molecular clock analyses suggest that the three principal red-taxa phytoplankton groups (dinoflagellates, coccolithophores, and diatoms) evolved before the Paleozoic (41, 42), the fossil record indicates that these groups rose to ecological prominence only during the Mesozoic (Fig. 2C). Prior to the Mesozoic, evidence suggests that the green superfamily and cyanobacteria dominated marine primary production (43–47).

Model C:P_org and N:P_org at the time of the major phytoplankton groups’ peak appearance are similar to C:P_org and N:P_org in modern representatives of those groups, grown under nutrient-replete conditions (Fig. 2). Application of relationships from the modern ocean to the geologic past carries uncertainty. However, the favorable comparison of our results with the microfossil record suggests that nutrient preferences and elemental compositions of modern phytoplankton groups may be conserved traits that reflect their inherent physiology (21, 48), rather than evolutionary adaptations to the modern marine nutrient regime. If correct, this means, for example, that cyanobacteria have not evolved to thrive at the low nutrient levels in the oligotrophic gyres, in response to the colonization of high-nutrient waters by modern phytoplankton. Instead, cyanobacteria have always been well suited to the low nutrient levels prevalent during their emergence, and they currently occupy the environments remaining to them—oligotrophic waters unsuited for growth of the more nutrient-rich modern taxa. The conclusion that the elemental composition of phytoplankton is constrained by genetics and evolutionary history is supported by the observation that cyanobacterial strains that have been cultivated under nutrient-replete conditions for almost 40 y still retain a high C:N:P_org (21, 22). Given the short generation time of cyanobacteria (approximately two divisions per day), this duration of cultivation is one to two orders of magnitude longer than that required for adaptive evolution (49).

Higher early Paleozoic C:P_org and N:P_org values are also consistent with the low suggested “energetics” (biomass, metabolic rates, and physical activity, such as motility and predation) of marine fauna of that time (1, 45, 50–54). On a high C:P_org diet (i.e., low-quality food), the energy spent on carbon respiration to access nutrients limits the energy available to evolve traits that facilitate motility, grazing, and reproduction (1, 3, 8). Most Paleozoic fauna fed close to the sediment–water interface and appear to have been characterized by low metabolic and diversification rates (52). Predation and bioturbation, which require substantial amounts of energy (for more advanced nervous and muscular systems), appear to have become more common through time, especially during the Meso-Cenozoic (54, 55), perhaps indicating that food quality increased (i.e., C:P_org and N:P_org decreased) around that time.

Biogeochemical Implications of Higher Paleozoic C:P_org.

Phytoplankton elemental composition has been frequently assumed to be constant at the Redfield ratio in studies of the long-term biogeochemical cycles. However, our study suggests higher-than-Redfield C:P_org and N:P_org during the early Paleozoic, driven by lower phosphate concentrations and a higher temperature. Persistently high Paleozoic C:P_org (Fig. 3A), produced by plasticity in the elemental composition of phytoplankton biomass, would have had significant consequences for marine biogeochemical cycling. For example, early Paleozoic carbon-based new production and burial of organic carbon are higher at elevated C:P_org relative to the case with a Redfield C:P_org (Fig. 3 B and C), and this leads to higher atmospheric pO₂ (Fig. 3E). Although O₂ production rates in the surface ocean are expected to be higher at elevated C:P_org (relative to a Redfield C:P_org), so is export of organic carbon from the surface to the deep ocean. The associated consumption of O₂ in the deep ocean results in lower deep-ocean O₂ concentrations (Fig. 3F), likely leading to more widespread anoxia. This is in accordance with proxy-based suggestions of deep-ocean anoxia persisting into the middle Paleozoic (35, 56, 57). In contrast to a pronounced effect on the organic carbon and O₂ cycles, our model predicts that the long-term inorganic carbon cycle is not affected by the introduction of flexible C:N:P_org (Fig. 3 G and H, all lines overlap). Atmospheric pCO₂ and the global average surface temperature are insensitive to C:N:P_org because the enhanced burial of organic carbon at high C:P_org is compensated by larger oxidative weathering fluxes of sedimentary organic matter. The increase in organic carbon weathering arises from the increase in atmospheric pO₂, which is a consequence of the enhanced organic carbon burial (Fig. 3D).

Fig. 3.

Effects of the proposed Phanerozoic evolution of C:N:P_org on the carbon and O₂ cycles. (A–H) Comparison between (A) C:P_org (molar ratio), (B) carbon-based primary productivity ($\times {10}^3$ Tmol C × y^–1), (C) marine organic carbon burial (Tmol C × y^–1), (D) weathering of organic matter in sedimentary rocks (Tmol C × y^–1), (E) atmospheric pO₂ (atm), (F) the deep-ocean O₂ concentration (µM), (G) atmospheric pCO₂ ($\times {10}^3$ ppm), and (H) temperature (K), for C:N:P_org that is fixed at the Redfield ratio (dashed black line) or variable according to the three parameterizations presented in Fig. 2.

The compensating effects of organic carbon burial and oxidative weathering on the inorganic carbon cycle and temperature, described above, operate on the timescale of formation, burial, uplift, and weathering of sedimentary rocks (i.e., tens of millions of years or more). On shorter timescales, before the sedimentary organic matter pool has fully adjusted to a change in the surface environment, we find that flexible C:P_org stabilizes the organic and inorganic carbon cycles and the O₂ cycle, similar to the findings of previous studies (26, 58). Upon perturbation of phosphate supply rates, e.g., by continental flooding or the development of anoxic basins, a rigid coupling (i.e., fixed C:P_org) leads to a stronger response of organic carbon burial, CO₂ drawdown, and O₂ production and release to the atmosphere. During a phosphate enrichment event (Fig. 4A), when C:P_org is fixed at the Redfield ratio (Fig. 4B), more carbon is fixed by marine primary producers, and as a result organic carbon burial and pO₂ transiently reach higher values than in the case of flexible C:P_org (Fig. 4 C–E). The higher rates of organic carbon burial shuttle carbon from the ocean–atmosphere into marine sediments, and pCO₂ drops substantially (Fig. 4F). The pCO₂ drop is significantly larger when C:P_org is fixed at the Redfield ratio, resulting in global cooling.

Fig. 4.

Dynamic response of model variables to an increase in the phosphate delivery rate at a constant (Redfield) and variable C:N:P_org. (A–F) Surface phosphate concentration (forcing function) (A), C:P_org of the primary producers (mol C ×mol P^–1) (B), carbon-based primary productivity (C), burial of organic carbon (D), atmospheric pO₂ (E), and atmospheric pCO₂ (F). All fluxes and pools are normalized to present-day values and are unitless.

Conclusions

We constrained the elemental composition of marine biomass through Phanerozoic time by including phosphate- and temperature-dependent parameterizations of C:N:P_org in a long-term biogeochemical model. We infer a decrease from high Paleozoic C:P_org and N:P_org to present-day Redfield ratios. The decrease in C:P_org and N:P_org is a result of a decrease in surface temperature and an increase in phosphate availability in the ocean, driven mainly by the emergence and expansion of land plants in the middle to late Paleozoic and by the Triassic breakup of the supercontinent Pangaea. Our results are consistent with the evolutionary trajectory of phytoplankton and with an increase in the energetics of marine fauna suggested by the fossil record. Our time-dependent C:N:P_org at the times of peak appearance of the major phytoplankton groups is consistent with that of modern representatives of those groups. This suggests that the elemental composition of these phytoplankton groups is a conserved trait, which is constrained by genetics and evolutionary history, and not an acquired trait. These ideas may be taken a (speculative) step farther. Although uncertainty exists in the reasons and exact timing of evolution and emergence to prominence of the different phytoplankton groups, we raise the possibility that the more nutrient-rich red-lineage phytoplankton were driven to dominance by the shifts in nutrient availability and climate that we describe above. This suggests a link between evolution and tectonics.

We further explored the biogeochemical consequences of flexible elemental ratios on the carbon and O₂ cycles. Our results suggest that the higher-than-Redfield C:P_org likely prevalent during the early Phanerozoic led to higher marine primary productivity and organic carbon burial, relative to the case of a time-invariant C:P_org. This simultaneously leads to higher atmospheric O₂ levels and more extensive anoxia in the ocean interior, relative to the case of a constant, Redfield C:P_org. Finally, on short timescales, we find that flexible C:P_org weakens the coupling between the phosphorous, carbon, and O₂ cycles and dampens perturbations in the carbon and O₂ cycles driven by changes in the rate of phosphate delivery to the ocean (Fig. 4).

In summary, it seems that flexibility in the elemental composition of phytoplankton biomass has led to a Phanerozoic decrease in C:P_org and N:P_org in response to tectonic and evolutionary drivers. This long-term decrease in C:P_org and N:P_org affected biogeochemical cycling and the stability of Earth’s atmospheric composition and climate over Phanerozoic time.

Materials and Methods

Experiments and Numerical Solution.

To explore the possibility that the elemental composition of planktonic organic matter varied through time, we implemented nutrient- and temperature-dependent C:N:P_org parameterizations in a model for the long-term cycles of phosphate, carbon, O₂, and calcium (the model development, validation, and the numerical procedure are fully described in ref. 29). We tested the evolution of C:N:P_org in four experiments. In one, C:N:P_org was fixed at the Redfield ratio of 106:16:1. In three others, C:N:P_org depended on environmental parameters, as described below. In each experiment we performed ∼10⁶ simulations, where we randomly drew 36 to 37 parameters and 11 time-dependent forcings from distributions that represent the uncertainty in their values (34 parameters are associated with the original model described in ref. 29, and 2 to 3 additional parameters are associated with each C:N:P_org parameterization; Table 1).

Table 1.

C:N:P_org parameterizations employed in this study

Symbol	Range	Distribution*	Ref.
${\rho }_{\mathrm{1}}$	${\rho }_{\mathrm{1}}=\left(\frac{{\rho }_{\mathrm{2}}-\mathrm{0}.\mathrm{0094}}{-\mathrm{0}.\mathrm{5}}\right)$		(26)
${\rho }_{\mathrm{2}}$	0.0048 to 0.006	Uniform	(26)
C:Norg	1.961 (0.256)	Lognormal	(18)
${\rho }_{\mathrm{3}}$	0.037 (0.010)	Normal	(27)
${\rho }_{\mathrm{4}}$	0.032 (0.008)	Normal	(27)
${\rho }_{\mathrm{5}}$	0.29 (0.8)	Normal	(24)
${\rho }_{\mathrm{6}}$	–7.8 (0.8)	Normal	(24)
C:${\text{P}}_{\text{org}}^0$	106
C:${\text{N}}_{\text{org}}^0$	16

*For normal and lognormal distributions, numbers outside and inside the parentheses represent the mean and SD, respectively.

Flexible elemental ratios in marine organic matter over Phanerozoic time are calculated in our model using three different relationships between C:N:P_org and the abiotic environment. The C:N:P_org parameterization 1 is based on an empirical global relationship between C:N:P_org and surface phosphate concentrations in the ocean (26): ${(\text{P}:\text{C})}_{\text{org}}=\left({\rho }_1\times {[\text{P}]}_s+{\rho }_2\right)$, where ρ₁ and ρ₂ are parameters that characterize the linear geographic ${(\text{P}:\text{C})}_{\text{org}}$ dependence on surface phosphate concentrations. The intercept, ρ₂, which corresponds to the lowest P:C_org, is between 0.0048 and 0.006 (based on binned or unbinned regressions, respectively; figure S2 in ref. 26), and we adopted this range (Table 1). To normalize to present-day conditions, where at the global average surface phosphate concentration [$\text{P}{]}_s^0=0.5$ µM, P:C_org is at the Redfield ratio (P:C${\text{C}}_{org}^0=0.0094$), we set ${\rho }_1=\left(\frac{{\rho }_2-0.0094}{-0.5}\right)$. To calculate N:P_org we substitute ${(\text{N}:\text{P})}_{\text{org}}={(\text{C}:\text{P})}_{\text{org}}/{(\text{C}:\text{N})}_{\text{org}}$. As C:N_org is relatively uniform across oceanic regions (26) and among different phytoplankton taxa (23), we draw it from a distribution that represents the variability in this value across different phytoplankton taxa growing under nutrient-replete conditions (Table 1).

Parameterization 2 is based on observed correlations between surface C:N:P_org in the present ocean and sea-surface temperature (27). The geographic temperature dependence of algal elemental composition is given by the equations ${(\text{C}:\text{P})}_{\text{org}}={(\text{C}:\text{P})}_{\text{org}}^0\times e^{{\rho }_3\left(\text{T}-{\text{T}}^0\right)}$ and ${(\text{N}:\text{P})}_{\text{org}}={(\text{N}:\text{P})}_{\text{org}}^0\times e^{{\rho }_4\left(\text{T}-{\text{T}}^0\right)}$, where ${(\text{C}:\text{P})}_{\text{org}}^0,\hspace{0.17em}{(\text{N}:\text{P})}_{\text{org}}^0$ are the Redfield ratios, ρ₃ and ρ₄ are parameters that characterize the geographic temperature dependence (Table 1), and T is the global average surface temperature in Kelvins (T⁰ is the present-day temperature).

Parameterization 3 is based on a meta-analysis of laboratory experimental data, where the following relationship was reported: ${(\text{P}:\text{C})}_{\text{org}}={(\text{P}:\text{C})}_{\text{org}}^0\times {\left(\frac{{[\text{P}]}_s}{{[\text{P}]}_s^0}\right)}^{{\rho }_5}\times {\left(\frac{\text{T}}{{\text{T}}^0}\right)}^{{\rho }_6}$. Here, ρ₅ and ρ₆ are empirical parameters that represent the sensitivity of the P:C_org of different phytoplankton groups to the phosphate concentration and temperature, respectively. We used cyanobacterial ${\rho }_5=0.3$ and ${\rho }_6=-7.8$ parameters (Table 1), but note that using eukaryote parameters (${\rho }_5=0.4$ and ${\rho }_6=0$) would give similar results due to a tradeoff between phosphate and temperature acclimation (24).

Fossil Assemblage.

The fossil data in Fig. 2C represent the number of occurrences of the different phytoplankton groups in the geologic record, which were compiled from the Paleobiology database (https://paleobiodb.org) using the following parameters: time intervals = Phanerozoic, paleoenvironment = marine, order = acritarchs, prymnesiophycea, dinophycea, diatoms). Acritarchs are considered to include the earliest green algae, which use chlorophyll-b. Dinoflagellates, coccolithophores, and diatoms, i.e., “modern” red phytoplankton groups, use chlorophyll-c. Biomarkers suggest that cyanobacteria played a significant role in Paleozoic marine primary production alongside green algae (46), and we, therefore, show the elemental composition of cyanobacteria in Fig. 2 A and B in the early Paleozoic.

Data Availability

The code used in this study and the model C:Porg and N:Porg trajectories (medians and 95% confidence intervals) over Phanerozoic time are provided in Zenodo (https://zenodo.org/record/5768279#. YbttQWjMKHs; ref. 40). All study data are included in this article and/or SI Appendix.