A proportion of the atmospheric carbon fixed by phytoplankton is transferred to the ocean depths, where it is degraded to particulate organic carbon and dissolved organic carbon (POC and DOC), then eventually oxidized to inorganic carbon. Carbon exchanges between the ocean and atmosphere in the form of CO2. Therefore, the oxidation of organic carbon is an important process in controlling the storage of carbon in the ocean. Organic matter is oxidized by heterotrophic processes, which have, however, largely been unquestioned. Most fresh organic matter is readily oxidized in surface waters, with the proportion of refractory organic matter being increased in the deep water 1,2. We consider how the organic matter in deep water is oxidized. Does it proceed in the same manner as in shallow water? To answer that question, we revisit the existing oceanic dataset.
The relationship between silicon concentration (as silicic acid) and TIC is shown in Fig. 1a 3. TIC, dissolved inorganic nitrogen (DIN) and dissolved inorganic phosphorus (DIP) are usually strongly coupled in seawater (Fig. 1b), implying the same primary source of all three components (organic constituents of phytoplankton). The ratio carbon to nitrogen to phosphorus is a nearly constant 106:16:1 (the Redfield ratio) in dissolved nutrient pools (Fig. 1b, c). These components are remineralized on a similar timescale and are therefore subjected to the same advective flow, resulting in coherent behaviour 3. However, closer inspection of Fig. 1a reveals that TIC depends only on silicic acid concentrations in the range 50 < [Si] < 150 µmol/kg. The contours of Fig. 1a denote nitrogen concentration. The straight contour lines indicate that TIC increases with [Si], for fixed [DIN]. In other words, both TIC and Si concentrations can increase independently of [DIN] (or [DIP], not shown) with ΔSi :ΔTIC typically 1.2 ~ 1.6. This is more evident in Fig. 1c. Linear relationships between TIC and [DIP] (or [DIN], not shown), following the Redfield ratio, can be seen up to the maximum values of [DIN] or [DIP]. At high TIC values in Fig. 1c, however, additional TIC depending only on [Si] occurs at [Si] > 50 µmol/kg. This again shows the presence of a C and Si source, independent of N and P in the deep layer, where [Si] > 50 µmol/kg.
The TIC and alkalinity diagram shows TIC increase and alkalinity decrease largely with [DIP] (or [DIN], not shown) increase (Fig. 2b). This is primarily due to remineralization of planktonic organic matter, affected by carbonate formation and dissolution as indicated by the arrows in Fig. 2b. A similar feature is seen with [Si] in the region of [Si] < 50 µmol/kg, implying the supply of TIC and silicic acid due to planktonic degradation. However, alkalinity increase depends strongly on [Si] in the region of [Si] > 50 µmol/kg (Fig. 2a, inside oval). When there is a supply of carbon in the absence of P or N (Fig. 1), the slope (ΔTIC/Δalkalinity) passing through a certain [P] value (or [N] value, not shown) is ~ 1 or slightly lower (Fig. 2b). This alkalinity dependence on Si is not clear in the region of [Si] < 50 µmol/kg (Fig. 2a).
Here, we consider potential explanations for this simple but significant relationship of ΔTIC:Δalkalinity:ΔSi = 1:1:1.2 ~ 1.6.
Simple mixing of different water masses is in accord with the relationship shown in Fig. 1, if one end member has low [Si] and TIC whereas the other is characterized by high [Si] and TIC. However, [Si] in deep water is normally correlated with seawater density: isopycnic lines run almost vertically in Fig. 1a3. Therefore, the straight lines in Fig. 1a do not match isopycnic lines. The idea of mixing can thus be discounted.
This simple relationship of ΔTIC:Δalkalinity:ΔSi = 1:1:1.2 ~ 1.6 is achieved by carbonate dissolution, OC oxidation and opal dissolution at a molar ratio of 1:1:2.4 ~ 3.2. If this is a result of remineralization of settling particles, the ratio corresponds to CaCO3:OM:opal weight ratio of 1:0.28: 1.5 ~ 2, assuming that OM is void of N and P. CaCO3 dissolution and OM dissolution can be strongly coupled, as CO2 produced by OM oxidation can be consumed in CaCO3 dissolution. Also, OM is not limited to that in settling particles, but can be DOC. Therefore, OM contents in settling particles may not be restricted. The above CaCO3:opal ratio of 1:1.5 ~ 2, however, poses three problems. Firstly, dissolution of CaCO3 and opal take place independently: the former depends on carbonate saturation depth (CSD) and the latter mainly on water temperature. Close analysis reveals that the ΔTIC:Δalkalinity:ΔSi relationship at [DIP] = 2.5 µmol/kg (or [DIN] = 35 µmol/kg) includes data from depth shallower than CSD at [Si] = 50 ~ 60 µmol/kg, those from depth around CSD at [Si] = ~ 100 µmol/kg and those from depth deeper than CSD at [Si] = ~ 150 µmol/kg. This simply indicates that the three reactions (carbonate dissolution, OC oxidation and opal dissolution) cannot always be coupled with each other, but that carbonate dissolution reaction should decouple from the other two at areas of smaller [Si]. Secondly, opal export to deep water takes place in only a few regions, such as the North Pacific and the Southern Ocean4. When major dissolved Si input occurs only in North Pacific Deep Water (NPDW) and Antarctic Bottom Water (AABW), how can this input Si can be distributed to regions characterized by 50 < [Si] < 150 µmol/kg without resorting to mixing processes, which were discounted earlier? (Circulation of this input Si will be discussed later, in a different context.) Thirdly, the CaCO3:opal ratio in settling particles varies widely, by a factor up to 1004. If the Si is not sourced from NPDW or AABW, how could such a small variation in ΔTIC:Δalkalinity:ΔSi = 1:1:1.2 ~ 1.6 be attained over the wide range of 50 < [Si] < 150 µmol/kg?
A different explanation is that a ‘weathering reaction’ occurs in the ocean depths. Crustal material contains alkaline and alkaline-earth elements, dissolution of which generates not only silica, but also bicarbonate ion from oxide minerals. For example:
CaO + 2CO2 + H2O → Ca2+ + 2HCO3− (1)
Na2O + 2CO2 + H2O → 2Na+ + 2HCO3− (2)
If dust has the mineralogical composition typical of the upper crust 5, as expected, and contains 10% ‘weathering-resistant’ quartz6, the ratio of total HCO3− to Si (ΔHCO3− /ΔSi) produced by dissolution will be ~ 0.7, i.e. almost identical to the slope observed in Fig. 1a. However, CO2 should be replenished, in order to maintain ΔTIC/ΔSi ~ 0.7. In principle, this may be achieved by the parallel oxidation of organic carbon (OC). If this is indeed the case, the combined reactions yield TIC, alkalinity and Si at a ratio of 1:1:1.4. The slightly lower than unity TIC:alkalinity ratio in Fig. 2b could be the result of secondary carbonate formation/dissolution reactions. The layer of [P] around 3 µmol/kg in the Pacific Ocean corresponds to the depth where carbonate formation is supposed to occur.
It has been reported that the weathering reaction can be mediated by microorganisms in marine environments 7,8. Therefore, it is reasonable to consider that those organisms harness energy by oxidizing OC and that they obtain micronutrients (such as Fe) by dissolving silicate minerals. This mechanism can generate silicic acid and HCO3− simultaneously, thus explaining the alkalinity increase being dependent on [Si] increase (Fig. 2a). Interestingly, the vertical distribution of dissolved organic carbon (DOC) and particulate organic carbon (POC) is the mirror image of that of [Si]. Furthermore, the magnitude of [Si] increase is similar to that of DOC decrease9. Although this combination of dust weathering and OC oxidation has not been proposed previously, it does provide a potential interpretation of the relationship shown in Figs. 1 and 2.
Based on the above discussion, there are two different combinations of remineralized materials which could form the basis of a possible explanation: (1) carbonate, OC and opal; alternatively (2) dust and OC. We now further examine the possibility of novel combination 2.
The spatial variation of K/Na or Mg/Na in seawater with changes of silicic acid concentration might provide further support of the proposed weathering. If a silicic acid concentration change of 100 µmol/kg is the result of dust weathering, calculation predicts 0.02% and 0.1% increase in Mg/Na and K/Na, respectively. However, the precision of measurements is not high enough to detect such differences10.
Weathering of silicate dust in an aqueous phase produces HCO3−, with pCO2 of the water being depleted unless CO2 is replenished by the oxidation of OC. Therefore, when the silicic acid concentration increases without an increase in TIC, the aqueous phase becomes alkaline and depleted in CO2. Alternatively, if OC oxidation occurs without silicate dust weathering, pCO2 of the aqueous phase will increase and the phase becomes more acidic. A test of whether the process of silicate dust weathering to release silicic acid operates in the deep ocean is to compare the chemical composition of the seawater with the propensity of the seawater to release or absorb CO2.
We first express TIC as a function of [Si] and [DIN] by analyzing data from depths greater than 500 m to avoid the surface mixing layer, which shows a ‘noisy’ relationship between TIC, [Si] and [DIN]. The slope of ΔTIC/ΔSi tends to decrease from 0.8 to 0.4 with increasing [Si]. As the most important layer is the uppermost depth of 500 m, where [Si] is largely < 60 µmol/kg, data with [Si] < 60 µmol/kg was used. The reason for the decreasing slope with increasing [Si] will be considered in relation to upwelling water, towards the end of this paper. The function obtained is:
[TIC] = 2194 + 0.8 [Si] + 7.56([N] − 32.1). (3)
The water that satisfies this equation is assumed to be well balanced between “CO2 absorbing” dust weathering and “CO2 releasing” organic carbon remineralization. We then introduce an index BOW, denoting deviation of [TIC] from the balance associated with “ocean weathering”, to evaluate whether CO2 would be absorbed or released during water circulation.
B OW = [TIC] − 7.56([N] − 32.1) − 0.8[Si] − 2194 (4)
If the BOW-index is negative, it is likely that the pCO2 is not yet recharged from OC decay. This water is likely to absorb CO2 when it rises towards the surface. On the other hand, if the BOW-index is positive, OC degradation exceeds the weathering effect. This water should release CO2, when it flows towards the surface.
The second possibility, namely of dust and OC coupling, if true, may be seen in the present-day oceans. Surprisingly excellent matches of the BOW-index values with surface pCO2 are seen (Fig. 3a, d): the CO2-releasing east central Pacific Ocean shows positive BOW-index values. Furthermore, that locality’s Sub-tropical Mode Water forms, which are well-known to be CO2-absorbing, exhibit negative BOW-index values. In reality, the uppermost depth (500m), where BOW-index values can be calculated, is overlaid by the surface mixed layer. The tendency of water to release or absorb CO2 may be disguised (at least partially) by the presence of this layer and also by carbonate buffering of pCO2. Existing models of surface pCO2 require a supply of Fe from dust 11. Furthermore, a dependence on surface water temperature is also needed to reproduce the observed surface pCO2 11. It should be noted that neither Fe abundance nor water temperature are considered in the BOW-index we present here. If carbonate and opal dissolution + OC oxidation (combination 1) are operative, excess Si or TIC excess should not relate directly to whether CO2 is retained in the ocean water or is released to the atmosphere.
It should be noted that, as our BOW-index is defined using data for water deeper than 500 m, it is pointless to show its value for surface waters. In addition, the above discussion implies that the surface-layer water merely equilibrates with the atmosphere with respect to CO2 and is therefore of little importance in regulating atmospheric pCO2. This is consistent with the observation that surface production plays little role in carbon transport to the ocean interior 12.
If we accept the importance of airborne dust input to the sea surface (as an alternative to riverine sources), the oceanic silicon budget may be greatly affected. If the Si responsible for ΔTIC/ΔSi = 0.7 is ultimately derived from dust, this would account for about half of the silicate supplied to the oceans, on a global scale. The contribution of airborne dust is comparable to that of riverine silicic acid, thus probably the residence time of silicon may be smaller than has been estimated previously 13. We suggest that investigating for the presence of surficial bacteria on dust particles associated with weathering would be helpful here.
The proposed weathering process in the oceans implies that carbon can be retained in the oceans as inorganic carbon, coupled with the accumulation of silicic acid in a ‘normal’ deep ocean, where 50 < [Si] < 150 µmol/kg (Fig. 4a). Close comparison between the BOW-index and the observed pCO2 distribution reveals a discrepancy between them in the North Pacific: the BOW-index shows negative values (Fig. 3a), but the observed pCO2 indicates emission from the same area (Fig. 3d).
The North Pacific is an active area of upwelling and diatom production is high. We have postulated “dissolution kinetics of diatom aggregates” to explain the chemical composition of diatom frustules14,15 and to reproduce the opal distribution, which is consistent with the observed Al distribution16. Based on this kinetic explanation, frustules form larger aggregates in regions of high diatom productivity and a greater portion of opal dissolves in deeper waters. Owing to rapid remineralization, carbon tends to remain as CO2 in the surface water, whereas in upwelling areas more frustules escape dissolution. In such upwelling areas, carbon abundance is therefore decoupled from [Si], resulting in collapse of the stoichiometric relationship between TIC and Si. In the surface water of such areas, accumulated CO2 has to be removed to the atmosphere to meet a steady state, while Si is selectively exported to the deep water. In deep water of such areas, ΔTIC/ΔSi decreases to almost zero, as can be seen in Fig. 1a; this implies loss of carbon from those regions (Fig. 4b). This is able to explain the discrepancy seen in the comparison of the BOW-index and pCO2 values in the North Pacific (Fig. 3a, d). It is interesting to note that, in Fig. 1a, the Southern Ocean − another upwelling-active area − also displays a clear departure from the linear relationship. The East Central Pacific ocean is also known as an upwelling dominant area 17. Unfortunately, stations of our dataset are not exactly on the equator. The BOW index already shows the feature of CO2 releasing, which is likely to be further boosted by upwelling. This is the reason why this area is known as the most active CO2 releasing zone. The zero ΔTIC/ΔSi can also be explained by the dissolution of silicic acid from the seafloor 18. This may occur, but water showing ΔTIC/ΔSi = 0 includes intermediate water 3 and it is considered that the loss of CO2, coupled with efficient export of Si from the surface layer, is the main reason for the zero ΔTIC/ΔSi. If some opal deposition from the water column onto the seafloor occurs, this should also contribute to the collapse of the stoichiometry or deviation from the linear relationship.
The CO2 flux from upwelling areas can be estimated from the corresponding deviation (Fig. 4c), mass flow data of upwelling, and opal burial rates in those areas18. In the North Pacific, estimation of the mass transport is complicated 19. For the Southern Ocean, however, using mass transport data for upwelling 20, CO2 emission is calculated (as the product of mass transport rate and the extent of deviation from linearity) to be around 30 Tmol/year. The contribution of opal burial is not included in this calculation, but it is less than 10%. This estimate is very small compared to the rate of total carbon exchange between the ocean and atmosphere, but is comparable to the carbon removal rate attributed to terrestrial weathering 21. Nevertheless, this CO2 emission could be crucial to the oceanic carbon budget, as it is not readily ‘cancelled out’ because of originating from the deep ocean. It can therefore be a critical component for understanding carbon budget variations during the transition between glacial and interglacial stages. It is widely accepted that a proportion of the carbon sequestered by the ocean during glacial times was returned to the atmosphere during deglacial times22. Such a mechanism explains very well that CO2 is released during deglacial times, in tandem with an increase in opal burial 23.
Incidentally, it has been argued that the ‘unexploited’ high abundance of DIN in the Southern Ocean, together with Si depletion, are exported to (and affects the biological productivity of) lower latitudes by sub-Antarctic mode water (SAMW) of typical density (σθ) = 26.8 24. However, as shown in Fig. 3c, it is clear that BOW does not propagate during the transportation. The tendency of water to absorb/release CO2 should be acquired rather locally by the addition of dust or oxidation of OC.
The mechanism developed in this paper does not contradict observations of opal flux in the oceans4. Significantly, high opal export is only seen in upwelling-active areas, where the (ΔTIC/ΔSi) stoichiometric relationship has collapsed. In many regions, where the observed (ΔTIC/ΔSi) ratio of ~ 0.6 to 0.8 applies, opal export to deep water is smaller than 30 mmolSi m− 2yr− 1 (ref. 4). This is consistent with Si increasing by 10 µmol/kg or less in the deep water, during the time taken to flow from the North Atlantic to the North Pacific (~ 1000 year). However, such an increase is significantly smaller than as shown by observations.
Deep water to which silica is added in the upwelling areas (see Fig. 4) eventually flows elsewhere, transported by Pacific Deep Water (PDW) in the Pacific or by Antarctic Bottom Water (AABW) in the Atlantic. This is clearly evident in the BOW -index values at 2000 m depth (Fig. 3b), which are more uniform and closer to zero than those at 500 m depth (Fig. 3a). This indicates that the deep water is almost in balance, with respect to CO2. Nevertheless, because of the ‘excess’ Si supplied from deep water in the upwelling areas, the BOW-index values are negative in the North Pacific and in the Southern Ocean. This may appear as a depression in the slope of ΔTIC/ΔSi seen in Fig. 1a and may explain why ΔTIC/ΔSi in Fig. 1a decreases with the increase of silicic acid concentration. Selecting datasets in which [Si] < 60 µmol/kg ensures that there is balance between “CO2 absorbing” dust weathering and “CO2 releasing” organic carbon remineralization, as expressed by Eq. 3. For [Si] < 60 µmol/kg, significant deviation in the (ΔTIC/ΔSi) ratio due to influence of “upwelling active areas” is avoided.
This study is supported by Grant-in-Aid from MEXT (No. 17K05717).