Weathering controls on the Phanerozoic phosphate cycle

Although phosphate is an essential macronutrient for marine biota, critical to our under- 1 standing of marine productivity, biogeochemistry, and evolution, its long-timescale geologic 2 history is poorly constrained. We constrain weathering-derived ﬂuxes and seawater con- 3 centrations of phosphate throughout the Phanerozoic (541 Ma to present), by developing a 4 model for the coupled, long-term biogeochemical cycles of phosphate, carbon, oxygen, and 5 calcium. We ﬁnd that the relative contribution of continental and seaﬂoor weathering to the 6 total weathering rate exerts a ﬁrst-order control on ocean productivity, through a previously 7 uninvestigated mechanism. Speciﬁcally, continental weathering is a source of the limiting nu- 8 trient phosphate, but seaﬂoor weathering is not. As a result, times in Earth history in which 9 seaﬂoor weathering constitutes a large fraction of the total weathering rate (e.g., the early 10 Paleozoic and Mesozoic), are also times in which phosphate delivery to the ocean is relatively 11 low. A lower concentration of phosphate in seawater likely affected primary productivity, 12 oceanic and atmospheric oxygen concentrations, with possible implications for the evolution

of marine fauna over Earth history.
14 By limiting marine primary productivity and organic carbon (C) burial over geologic 15 timescales, the long-term availability of phosphate (P) is a key determinant of atmospheric oxygen 16 and carbon dioxide levels (pO 2 and pCO 2 , respectively) 1-4 , climate 5 , and marine faunal evolution 17 and diversity patterns 6 . On geologic timescales, the size of the marine P reservoir is set by a balance 18 between P supply from continental weathering and P loss, mostly by burial in sedimentary rocks [7][8][9] . 19 On timescales comparable to the mixing time of the ocean (thousands of years) the given long-term 20 marine P reservoir is distributed between the surface and deep ocean by a balance between P in-21 corporation into newly fixed organic matter and P release during organic matter remineralization 5 .

22
The long-term source of P into the ocean is chemical weathering of continental silicate rocks [8][9][10] . 23 Seafloor weathering, on the other hand, is considered to be a minor sink of P 11-13 . During seafloor 24 weathering, away from the spreading axis, low-temperature (generally 5-20 o C) water circulates in 25 the highly porous upper oceanic crust 14 . Though the exact mechanism is unclear, porewater and 26 bulk-rock chemical analyses suggest that P is removed from seawater during off-axis water-rock 27 interactions, likely due to adsorption onto iron-(oxy)hydroxides in sediments, secondary apatite 28 precipitation, or microbial activity in the crust 11 . 29 Both continental and seafloor weathering rates depend on temperature through the chemical 30 reaction kinetics [15][16][17][18] . Continental silicate weathering rates also depend on temperature through 31 changes in precipitation and runoff, which alter the water to rock ratio [19][20][21] . The temperature 32 dependence of both continental and seafloor weathering has led to suggestions of negative cli-33 mate feedbacks 17,[22][23][24][25][26][27] . According to these feedbacks, at a given volcanic outgassing rate, CO 2 34 will accumulate in the atmosphere to a steady-state concentration at which the global-average 35 temperature-dependent continental and seafloor weathering rates (which are equal to the rate of 36 carbonate burial) exactly match the CO 2 outgassing rate. As continental weathering is also the 37 source of the nutrients required for the production of organic matter, a similar feedback is sug-38 gested to exist between the temperature-dependent rate of continental P weathering and organic C 39 burial 23, 28 . Since continental silicate weathering is a source of both calcium (Ca) and P and seafloor 40 weathering is only a source of Ca and a minor sink of P, the relative proportion of seafloor weather-41 ing to the total weathering (seafloor + continental) has implications for the P balance. For example, 42 a gradual increase in pO 2 over 1,500-500 Ma has been suggested to reflect a gradual increase in 43 the relative proportion of continental weathering, shifting the balance between the inorganic and 44 organic C burial sinks 29 . The consequences for the Phanerozoic P cycle, to our knowledge, have 45 not been fully explored. 46 Here, to study the evolution of the Phanerozoic P cycle, we developed a model for the cou-47 pled, long-term biogeochemical cycles of P, C, O 2 , and Ca ( Fig. 1), and simulated the fluxes of 48 these elements between the ocean-atmosphere and rock reservoirs in response to major geologic 49 forcings. We account for the significant parameter and forcing uncertainties by drawing these from 50 distributions that represent the uncertainty in their values (Supplementary Information, SI). We 51 validated our model against geologic proxies for pCO 2 , the C isotopic composition (δ 13 C) of car-52 bonate minerals, and the C content of altered oceanic crust (Methods). We propose an evolutionary 53 trajectory of the P cycle, consistent with these and other observational constraints. Our results 54 suggest that the P weathering flux and marine concentration varied throughout the Phanerozoic in 55 response long-term variations in continental weatherability. 56 Continental and seafloor weathering throughout the Phanerozoic 57 Our analysis suggests that during the early Paleozoic era, continental weathering rates were low 58 and seafloor weathering rates were high, relative to their present-day values. The lower contribu-59 tion of continental weathering-derived alkalinity influx to the ocean (out of the total weathering-60 derived alkalinity influx) stems from lower continental weatherability before the evolution of land 61 plants. Land plants enhance continental weatherability in several ways, for example, by exuding 62 organic acids and metal chelators, by increasing the surface area for weathering due to their exten-63 sive root systems, and by recirculating water through evapotranspiration. The lower weatherability 64 results in a balance between the volcanic CO 2 source and the weathering CO 2 sink (comprised of 65 CaCO 3 and organic C burial) at a higher global-average temperature (Extended Data Fig. 1). The 66 higher temperature accelerates both continental and seafloor weathering. As continental weather-67 ability was significantly lower than the present, seafloor weathering becomes the dominant source 68 of alkalinity to the ocean (Fig. 2).

69
As land plants expanded, ∼400 to 350 Ma, continental weatherability increased, and a bal-70 ance between CO 2 inputs and outputs was reached at a lower surface temperature. This resulted in 71 slower temperature-dependent kinetics of both seafloor and continental weathering, though conti-72 nental weathering rates net increased due to the higher plant-related weatherability (Fig. 2). During 73 the assembly of the supercontinent Pangaea between ∼306 and 237 Ma, the average annual pre-74 cipitation over land is thought to have decreased to a mid-Triassic minimum lower than today 75 (Extended Data Fig. 2), due to the difficulty of delivering moisture to inland regions 30 . Lower 76 precipitation results in lower weathering rates at a given temperature, requiring a higher temper-77 ature for seafloor and continental weathering rates to match the volcanic outgassing rates. As in 78 the early Paleozoic, the higher temperature accelerates both continental and seafloor weathering.

79
However, due to the low continental weatherability, in this case related to lower precipitation over 80 land, seafloor weathering becomes the dominant source of alkalinity to the ocean (Fig. 2b). In 81 contrast, during the breakup of Pangaea, which started at ∼237 Ma, continental weathering rates 82 increased and seafloor weathering decreased to present-day levels (Fig. 2).

83
Our proposed geologic history of seafloor weathering rates is consistent with observed 84 changes in CO 2 content of the oceanic crust 14,26,31,32 . During seafloor weathering reactions, Ca 85 is leached from the basalts, increasing the Ca concentration in crustal pore fluids, and promoting 86 CaCO 3 precipitation in veins and void fillings 14,17,26,31 . The rate of C addition to altered oceanic 87 crust, as inferred from its measured CO 2 content 26 , is higher in the Mesozoic than over Cenozoic  As continental weathering is a source of both alkalinity and P, whereas seafloor weathering is a 95 source of alkalinity only, the variation in the relative importance of continental and seafloor weath-96 ering described above carries implications for the marine P budget. Our results (of 10 6 random 97 parameter draws; Fig. 3) indicate that the high relative importance of seafloor weathering in the 98 early Paleozoic (Fig. 3b) led to weathering-derived P fluxes and deep-ocean P concentrations that 99 are both lower-than-present by a factor of ∼3. When land plants evolved (∼400-350 Ma) and con- Ma) P delivery rates and deep-ocean concentrations increased toward their present-day values. Our 106 predicted geologic history of the P cycle (weathering and organic C burial fluxes) differs from a 107 recent prediction based on 87 Sr/ 86 Sr and δ 13 C values in marine carbonates 33 . We note, however, 108 that the aforementioned prediction accounted neither for recognized changes in continental lithol-109 ogy as drivers of marine carbonate 87 Sr/ 86 Sr records, nor for recognized deviations in the C:P of 110 buried organic matter from the "Redfield" ratio of 106:1.

111
Weathering-derived P influxes and the seawater P concentrations depend on assumptions 112 about the temperature dependence of seafloor weathering rates. Assuming that seafloor weather-113 ing rates are temperature-independent, only alkalinity influxes from continental weathering may 114 respond to changes in CO 2 outgassing, through a change in the global average surface tempera-115 ture. As continental weathering is also a source of P, in this case, higher CO 2 influxes translate 116 directly to higher P influxes. For this reason, with temperature-independent seafloor weathering 117 rates, the Paleozoic weathering-derived P influxes and deep-ocean P concentrations are higher than 118 with temperature-dependent seafloor weathering, and they closely follow the temporal evolution of 119 the CO 2 outgassing rate (dark red envelope; Fig. 3). Below, we argue that the limited observational 120 constraints on the evolution of seawater P concentrations support temperature-dependent seafloor 121 weathering.

122
To our knowledge, there is only a limited number of proxies for the marine P concentration.

123
An increase in P/Fe ratios in iron-oxide-rich marine sediments, from a pre-Cretaceous average 124 of 0.38 to a Cretaceous-Cenozoic average of 2.55, has been interpreted to reflect a change in the 125 silica cycle, rather than an increase in the seawater P concentration 4 . Specifically, a decrease in 126 the seawater silica concentration due to the expansion of siliceous phytoplankton (diatoms) was 127 inferred to have reduced the competition with P for adsorption sites on the surface of iron oxides, 128 resulting in enhanced P adsorption and higher P/Fe, without a major change in the seawater P groups that emerged later in Earth history tend to prefer progressively more P-rich conditions 38 , 141 and tend to produce biomass richer in P 39-41 . If these nutrient requirements and utilization patterns 142 are conserved group-specific traits, they suggest that P availability increased over the Phanerozoic 6 .

143
Implications for the carbon and oxygen cycles 144 Predicted marine primary productivity (Tmol C yr −1 ) tracks the dissolved surface P concentration 145 ( Fig. 4a, b), as primary productivity is calculated from the surface-ocean P concentration and 146 the elemental composition of the primary producers (here assumed to be constant at the Redfield 147 ratio; SI). Accordingly, early Paleozoic primary productivity is ∼2.5 times lower than today (Fig. 148 4b). Though the burial of marine-derived organic C is proportional to marine primary productivity, 149 early Paleozoic organic C burial is lower than today by only ∼30% (Fig. 4c). This muted response 150 to lower-than-present primary productivity is related to lower early Paleozoic pO 2 (Fig. 4e, f), 151 which led to a higher marine organic C burial efficiency 42 . Lower early Paleozoic pO 2 stems from 152 the absence of land plants and the associated terrestrial organic C burial flux (Fig. 4d), which has 153 C:P approximately an order of magnitude higher than marine organic C.

154
The δ 13 C values of marine carbonate rocks are widely used to constrain the relative propor-155 tions of organic C and CaCO 3 burial 30,43,44 . With the emergence of land plants, our model predicts 156 a ∼3-fold increase in organic C burial, driven mostly by the onset of terrestrial organic C burial.

157
Over the same period, inorganic C burial is predicted to increase by only a factor of ∼2 (Extended We identify the relative importance of continental and seafloor weathering as a key factor in the 169 marine P cycle and, through the effect of P on primary productivity, also the global C and O 2 cycles.

170
Our analysis reveals two geologic events where a change in the relative importance of continental 171 and seafloor weathering drove enhanced P delivery rates. The first event was the evolution of land plants during the Devonian, which facilitated the uptake of nutrients from rocks. The second was 173 the Mesozoic breakup of the supercontinent Pangaea, which resulted in an increase in precipita-174 tion over the continents. Both events increased continental weatherability, thereby increasing the 175 relative importance of continental out of the total weathering, and enhancing the P supply to the 176 ocean. Our biogeochemical model suggests that the enhanced P supply led to an increase in marine 177 primary productivity, with possible implications for evolutionary trajectories of the biosphere over The long-term biogeochemical cycle of P involves a source from physical and chemical weathering 182 of silicate rocks on land, mainly apatite, and oxidative weathering of organic matter in sedimentary 183 rocks. A fraction of the weathered P is buried as terrestrial organic biomass, either on land or in 184 marine environments, with much of the burial occurring in deltaic environments. However, most 185 of the P is delivered to the ocean. The forms of P that enter the ocean are particulate and dissolved, 186 organic, and inorganic. The riverine flux of particulate inorganic P into the sea is 0.23-0.65 Tmol 187 P yr −1 (refs. 9, 10, 50), the particulate organic P influx is 0.03-0.26 Tmol P yr −1 (refs. 51, 188 52), and the total dissolved P (organic and inorganic) influx is 0.01-0.06 Tmol P yr −1 (refs. 7, 189 9, 10, 53-55). Some of the P that enters the ocean, mostly the particulate P, is not bioavailable 190 and is buried as detritus or iron oxide-bound P in shelf sediments 9, 10 . However, a fraction of 191 the initially unavailable P become bioavailable due to bacterial remineralization of the organic 192 P, and/or release by reductive dissolution of the iron oxide-bound P. The flux of bioavailable P 193 from continental weathering is loosely constrained due to the large uncertainty associated with the 194 fraction of particulate P that becomes bioavailable, estimated to be 8-69% 10, 52, 56-60 . 195 Most of the bioavailable P, after being fixed in newly produced organic matter, settles to 196 the sediments 9 . In the sediments, the organic matter undergoes remineralization and dissolved 197 (bioavailable) P is released into the sediment porewater. From the sediment porewater, the dis- ratio of C to P in buried organic matter, which depends on the degree of ocean anoxia, is between 202 237 and 250 (mol C mol P −1 ). Thus, a total of 0.018-0.076 Tmol P yr −1 is ultimately buried as 203 organic P. The modern fluxes of P burial as apatite and iron oxide-sorbed P are 0.015-0.091 and 204 0.024-0.041 Tmol P yr −1 , respectively (see Supplementary Table S2).

205
In addition to the sedimentary P sinks described above, P is removed from the deep ocean by 206 adsorption onto iron-oxide particles formed by oxidation of reduced iron emitted in hydrothermal 207 plumes at mid-ocean ridges. The modern-day flux of P removed in this process is estimated at 208 0.004-0.008 Tmol P yr −1 . Furthermore, oceanic P is also removed by ridge-flank crustal reac-209 tions. It is unclear whether P is removed via adsorption onto iron (oxy)hydroxides in sediments, 210 secondary apatite precipitation, or biological processes in the crust 11 . The magnitude of this flux 211 in the present ocean is loosely constrained and is estimated at 0.007-0.028 Tmol P yr −1 (Supple-212 mentary Table S2).  Table S9).

230
To test the sensitivity of the model prognostics to the different parameters, we varied each 231 parameter (total 34) within its range, and recorded the deep-ocean P concentration at two time marine P concentration, which is the main prognostic of interest in this study, is sensitive to the 237 parameters and forcings that affect the delivery of alkalinity and CO 2 to the ocean-atmosphere.

238
The reason for this sensitivity is that the influxes of P to the ocean depend on the absolute rates 239 and relative contributions of continental and seafloor weathering (i.e., sources of alkalinity), which 240 are required to balance the outgassing source of CO 2 by a combination of carbonate and organic 241 C burial. Thus, the parameters and forcings that affect the delivery of alkalinity and CO 2 to the 242 ocean-atmosphere ultimately control the sources and major sinks of P.

244
In Extended Data Fig. 1, we compared our model output to a compilation of δ 13 C values in marine 245 carbonate rocks 44 , which contains 43,972 samples. Data before the mid-Jurassic are derived mostly 246 from shallow-water (platform), near-shore carbonates. Data from the Jurassic through the modern 247 are derived mostly from CaCO 3 secreted by planktonic foraminifera and calcareous nannoplank-248 ton, which was extracted mainly from oceanic drill-cores. Overall, the envelope covering 90% 249 of the model δ 13 C results is mostly within one standard deviation of a 10-million-year moving 250 average of δ 13 C values measured in marine carbonate rocks.

263
In Fig. 2b we compared the Ca flux from the altered oceanic crust, estimated from the bulk 264 CO 2 content of oceanic crust at different ages and an assumption of a 1:1 molar ratio of Ca:C in 265 the crustal carbonates that yielded the CO 2 , to the fluxes of Ca leaching from the oceanic crust 266 obtained in our model. The bulk CO 2 content of the altered crust was obtained from Table 1 in     Rev. Ecol. S. 13, 349-372 (1982).  Chem. Geol. 161, 181-198 (1999).   Figure 1: Model scheme. The longterm cycling of P, C, O 2 , Ca, and the sedimentary pools of organic matter (C org ) and carbonate rocks (CaCO 3 ). F wP,ocean is the continental weathering flux of P to the ocean, F bap , F bFeP , are P burial fluxes as apatite and iron oxidesorbed P, respectively, and F plume,P , F off,P are the fluxes of inorganic P that is removed by adsorption to iron oxides in hydrothermal plumes, and by offaxial seafloor reactions, respectively. F borg and F borg,land are marine and terrestrial organic matter burial fluxes, respectively. F v is the CO 2 input from metamorphic and volcanic degassing, and F wcarb and F worg are C weathering fluxes from carbonates and organic matter in sedimentary rocks, respectively. F bcarb is carbonate burial. F morg , F mcarb , F ored are fluxes associated with metamorphism of organic matterbearing and CaCO 3 bearing rocks, and oxidation of reduced gases, respectively. F wsil and F wsf are the Ca input flux from continental silicate and offaxis seafloor weathering, respectively. . Markers and error bars represent the inorganic C uptake by the oceanic crust, calculated from the CO 2 content of oceanic drill cores (see Supplementary Table S1). Assuming a 1:1 molar ratio of Ca:C, these fluxes are equivalent to the source of Ca from seafloor weathering reactions. The higher CO 2 contents of the upper crust in the Atlantic Ocean than in the Pacific Ocean, at any given age, are interpreted to reflect the incorporation of recrystallized sedimentary CaCO 3 into the slowspreading Atlantic crust 26 . Comparison between (a) the P influx to the ocean from continental weathering (Tmol P yr −1 ), and (b) the P concentration in the deep ocean (µM P) obtained in ∼10 6 default model simula tions (same as in Fig. 2, frequency in color contours), in which seafloor weathering rates de pend on temperature, and simulation results obtained with temperatureindependent seafloor weathering rates (5th to 95th percentiles delineated in dark red). (c) The effect of continental configuration, latitude and vegetation cover on surface albedo, and consequently, on average continental temperature (∆T geog ). ∆T geog is drawn from a nor mal distribution, where the mean was adopted from Goddéris et al. (2012) 62 , and the stan dard deviation was adopted from the range of climate predictions associated with the CMIP5 models 63 . (d) A timedependent forcing that accounts for the enhancement of the climate sensitivity during cold periods (f glac ). f glac is drawn from a uniform distribution between unity and two when there is evidence for longlived glaciations, and is set to unity over the rest of the Phanerozoic. (e) Weathering enhancement due to landplant evolution (f E ). (f) A timedependent forcing that represents the colonization and expansion of terrestrial biomass (f cland ). f E and f cland were drawn from uniform distributions at any given time, where the boundaries of the uniform distribution change over time to account for the evolution of land plants described in the SI.