Mesocosm experiments

Si:Nexport measurements

Between 2010 and 2014, we conducted five in situ mesocosm experiments to assess impacts of OA on natural plankton communities. Study sites covered a large latitudinal gradient (28 °N–79 °N) and diverse oceanic environments/ecosystems (Extended Data Fig. 1 and Extended Data Table 1). Sample collection and processing was conducted every 1 or 2 days throughout the experiments. Sinking particulate matter was obtained from sediment traps attached to the bottom of each mesocosm, thereby collecting the entire material sinking down in the enclosed water column36. Processing of sediment trap samples followed a previous protocol37. Samples for particulate matter suspended in the water column were collected with depth-integrating water samplers (HYDRO-BIOS) and filtered following standard procedures. Biogenic silica was leached from the sediment trap samples and filters by alkaline pulping (0.1 M NaOH at 85 °C). After 135 min the leaching process was terminated with 0.05 M H2SO4 and dissolved silica was measured spectrophotometrically38. Carbon and nitrogen content were determined using an elemental CN analyser (EuroEA)39.

Analysis of OA impacts

To test for a systemic influence of OA on Si:Nexport, we synthesized the datasets from the different experiments and (i) conducted a meta-analysis to quantify effect sizes, and (ii) computed probability density estimates. Because the experimental design, the range of CO2 treatments, and the time periods for our analysis of Si:Nexport varied to some extent among experiments (Extended Data Table 1), we pooled mesocosms for ambient conditions and in the ({{p}}_{{{rm{CO}}}_{2}}) range of ~700–1,000 μatm (‘OA treatment’), corresponding to end-of-century values according to RCP 6.0 and 8.5 emission scenarios15. Effect sizes were calculated as log-transformed response ratios lnRR, an approach commonly used in meta-analysis40:


where X is the arithmetic mean of Si:Nexport ratios under OA and ambient conditions (Extended Data Table 1). Effect sizes <0 denote a negative effect of OA and effect sizes >0 indicate that the effect was positive. Effects are considered statistically significant when 95% confidence intervals (calculated from pooled standard deviations) do not overlap with zero. The overall effect size across all studies was computed by weighing individual effect sizes according to their variance, following the common methodology for meta-analyses40. In addition, we computed probability densities of Si:Nexport based on kernel density estimation, which better accounts for data with skewed or multimodal distributions41. Another advantage of this approach is that it does not require the calculation of temporal means. Instead, the entire data timeseries can be incorporated into the analysis, thus retaining information about temporal variability. Confidence intervals of the density estimates were calculated with a bootstrapping approach using data resampling (1,000 permutations)41. The resulting probability density plots can be interpreted analogously to histograms. Differences among ambient and OA conditions are considered statistically significant when confidence intervals of the probability density distributions do not overlap. Numbers for suspended and sinking Si, C and N (and their respective ratios) for the analysis period are given in Extended Data Table 2.

Analysis of pH effects on Si:N in global sediment trap data

We analysed a recent compilation of global sediment trap data (674 locations collected between 1976 and 2012)35. The aim of this analysis was to assess the influence of pH on opal dissolution in the world ocean. In contrast to the mesocosm experiments, where export fluxes were measured only at one depth, the global dataset provides depth-resolved information, enabling us to examine the vertical change in the Si:N ratio of sinking particulate matter and how this correlates with pH. It has long been known that the silica content of sinking particles increases with depth, as opal dissolution is less efficient than organic matter remineralization25,42. The resulting accumulation of Si relative to N can be quantified as the change in Si:N with increasing depth, that is, the slope of the relationship of depth versus Si:N (ΔSi:N, in units of m−1). Our approach is analogous to previous studies, which used vertical profiles of Si:C as a proxy for differential dissolution/remineralization of opal and organic matter, and its regional variability in the ocean24,42. We extracted all data that (I) included simultaneous measurements of Si and N, and (II) contained vertical profiles with at least three depth levels (so that ΔSi:N [m−1] can be calculated). We then calculated linear regressions for individual Si:N profiles and subsequently extracted those for which Si:N displayed a statistically significant relationship with depth (p < 0.05). Thereby, profiles with no clear depth-related pattern of Si:N—for example, due water mass advection—were excluded. In total, 190 profiles of Si:N flux matched those criteria and were used for further analysis.

To assess the influence of pH on ΔSi:N, we applied linear regression analysis, using the ΔSi:N for each vertical flux profile and the average pH (from the GLODAP database)43 over the corresponding depth range. Because temperature is considered the primary factor driving opal dissolution16,26 and has been shown to influence the preservation of silica compared to organic matter44, we additionally conducted a multiple regression analysis for the same dataset, by including temperature as a second factor. Results confirm that temperature also has an influence on ΔSi:N [m−1], with a comparable effect size as identified for pH. However, because the role of temperature has been discussed extensively in earlier studies (see Methods section ‘The present and future role of pH for opal dissolution in the ocean’), here we focus on the previously overlooked effect of pH.

The present and future role of pH for opal dissolution in the ocean

It is noteworthy that pH effects on opal dissolution have so far been mostly neglected in oceanographic research and earlier work almost exclusively focused on temperature as the factor controlling opal dissolution16,24. Although chemical dissolution experiments have demonstrated the pH dependence of biogenic silica dissolution20,21, it has so far been considered a minor factor in oceanic silica cycling. The reason probably lies in present-day gradients of temperature and pH in the ocean, and the resulting influence on opal dissolution: A temperature gradient of ~15–20 °C (average difference between surface and deep ocean), corresponds to a three- to fourfold change in the opal dissolution rate26,44. By comparison, the effect of pH described in our study is much smaller, indicating an ~20% difference in opal dissolution for present-day pH gradients of ~0.3–0.35. Accordingly, in the present ocean, the effect of temperature is roughly 10-fold larger compared to that of pH—which is probably why the latter has been so far neglected. In addition, the present-day vertical gradients of both temperature and pH both work in the same direction, that is, towards a decrease in opal dissolution rates with depth (decreasing temperature and decreasing pH).

However, the situation changes completely under future scenarios of OA and warming. Surface ocean pH is predicted to decrease by 0.2–0.4 until the year 2100, whereas sea surface warming may reach 1–3 °C (refs. 45,46). For these changes, the effect size of both factors on opal dissolution is on a similar order of magnitude. More importantly, in contrast to present-day vertical gradients, their future changes have antagonistic effects on opal dissolution, that is, warming-driven acceleration and a pH-driven slowdown. Notably, our results suggest that the effect of decreasing pH even overcompensates for the effect of warming on a global average (Extended Data Fig. 3). This illustrates that pH becomes an increasingly relevant factor for opal dissolution and the pelagic Si cycle in the context of ongoing climate change and OA.

Possible other contributing factors to OA impacts on Si:N

Although our results provide strong evidence for a chemical effect of decreasing pH on opal dissolution, additional or alternative explanations for the observed OA impacts on Si:N in the mesocosm studies cannot be fully excluded. On the basis of previous findings, one would expect that OA impacts on Si:Nexport can be explained by responses of diatoms. From a physiological perspective, lower pH may theoretically facilitate silicification by diatoms. The solubility of Si in seawater decreases with decreasing pH, promoting precipitation and inhibiting dissolution of opal. Diatoms are known to utilize this physicochemical property to precipitate opal in a cellular compartment with low pH conditions47,48. However, experimental evidence is scarce and partly controversial, with indications for either enhanced or reduced silica production under lower pH49,50. From an ecological perspective, higher Si:Nexport may have arisen from shifts in phytoplankton community composition, with a greater proportion of particle export driven by diatoms compared to other (non-silicifying) taxa, or by more heavily silicified species within the diatom community. However, our data do not support either of these two potential explanations, as the influence of OA on Si:N is only detectable for vertical particle fluxes (collected in sediment traps), but not for freshly produced particulate matter in the water column (Fig. 1c). This suggests that OA effects on Si:N emerged primarily while the biogenic detrital particles were sinking and not due to biotic effects during their production. Another possibility is that changes in N remineralization under simulated OA additionally contributed to the increase in Si:N. However, the current consensus is that bacterial communities and organic matter remineralization are mostly resilient to OA51. Results from studies that reported effects are very variable and, in most cases, it was not possible to separate direct pH effects (for example, on bacterial activity) from indirect effects mediated through pH-driven changes in quality and/or quantity of the organic matter substrate. Thus, there are currently no indications that OA will enhance N consumption of sinking organic matter. Altogether, the various independent lines of empirical evidence (mesocosms, ocean sediment traps, chemical studies) and the consistency of their results suggest that the pH effect on Si dissolution is the most probable explanation for our findings.

Global impact assessment from Earth system model simulations

We incorporated the effects of simulated OA on Si:Nexport observed in the in situ mesocosm experiments into an Earth system model to assess the global scale impacts on nutrient availability and plankton biogeography over the coming centuries. Consequently, we applied a modified version of the University of Victoria Earth System and Climate Model (UVic ESCM), which simulates silica cycling and diatoms, as well as other functional groups of phytoplankton as described in an earlier work52. Biogenic opal is produced by diatoms, including a parameterization for iron dependency of silicification, resulting in elevated Si:N ratios of production under iron limitation52. Vertical profiles of opal fluxes and dissolved silica are instantaneously computed based on biogenic silica production in the surface ocean and dissolution throughout the water column, and silica dissolution is parameterized as an exponential, temperature-dependent rate. Simulated present-day spatial distributions of Si(OH)4 in the surface ocean agree well with observational data (Extended Data Fig. 4).

To simulate OA effects on silica dissolution and Si:Nexport, we parameterized the specific silica dissolution rate to scale with changes in pH throughout the water column relative to preindustrial conditions for each box in the three-dimensional model grid, thereby accounting for the vertical characteristics of future pH changes (see Extended Data Fig. 3). Therefore, we assumed that the pH sensitivity of biogenic silica dissolution derived from the observed OA effect on Si:Nexport (17% for ΔpH of around 0.3) is linear, corresponding to a decrease in opal dissolution rate of 57% per unit pH. This estimate agrees remarkably well with published rates from chemical dissolution experiments20,21. We note that the model also accounts for effects of warming on silica dissolution (using a temperature dependence that is similar to other global models) that work in the opposite direction as the OA effect (see Extended Data Fig. 3).

Model simulations were run for the period 1750 to 2200 using extended IPCC scenarios (RCP 8.5 and RCP 6.0) for atmospheric CO2 concentrations27. The reason why we conducted the simulations until the year 2200 (instead of 2100 as commonly done in other climate change studies) is that we expected impacts of the OA-driven slowdown of opal dissolution, such as Si trapping in the deep ocean, to emerge only on the long timescale of global circulation. However, as can be seen in Fig. 3b and Extended Data Fig. 5, OA-amplified Si trapping in the deep ocean and the resulting decline in diatoms become apparent by 2100. Thus, the reference year (2100 or 2200) only affects that magnitude of the effect; qualitatively, the results are very similar.

Generally, simulated impacts of climate change are consistent with other models, for example, reduced nutrient supply to the surface ocean and an associated decrease in phytoplankton biomass. The most important results that are relevant for the interpretation of OA effects reported here (via slower silica dissolution) are presented in Extended Data Table 3. More details on model behaviour in climate change simulations can be found elsewhere52.

We emphasize that the simulated OA effect on silica dissolution occurs on top of other climate change impacts that are already known, for example, ecosystem responses to ocean warming and lower nutrient supply. Because these impacts have been extensively discussed in previous studies, we focus here on the global-scale implications of the insights from our work: the slowdown of silica dissolution under OA as revealed by our analysis of mesocosm and ocean sediment trap data. Thus, visualization and interpretation of results from the climate change simulations mostly refer to the net effect of OA-sensitive opal dissolution (ΔOA), which is quantified as the difference between (a) the model including OA effects on opal dissolution and (b) the standard model configuration. Additional results are presented in Extended Data Fig. 5 and Extended Data Table 3.

Limitations of the global model

The model we used (UVic ESCM) is similar to other common models in terms of its ecosystem component, its skill in reproducing present-day conditions of biogeochemical quantities, and its behaviour in climate change simulations52,53. Thus, it is reasonable to assume that the simulated OA impacts on opal dissolution and Si:Nexport would yield similar results if incorporated into other global models. The driving mechanism, namely the slowdown of opal dissolution under OA, alters the vertical profile of particulate opal fluxes and regenerated Si(OH)4. Because this is a chemical effect, it should be largely insensitive to specifics of the ecosystem model structure—instead, the most important factor is probably how the spatial distribution of diatoms and opal production are reproduced by different models. Our model shows good skill in reproducing observational data of Si(OH)4 in the surface ocean (Extended Data Fig. 4), indicating that simulated spatial patterns of diatoms and opal production are reasonably realistic. However, as with most Earth system models, our simulations do not account for some potentially relevant mechanisms that may arise, for example, through ecological competition or complex food-web interactions, which may also yield potential repercussions for the global carbon cycle and will be discussed in the following.

In the model including OA-sensitive opal dissolution, a large part of the loss in diatom biomass is compensated by an increase in productivity of other phytoplankton groups. Accordingly, global primary productivity and carbon export remain largely unaffected by the OA effect on opal dissolution and Si:Nexport, despite the sharp decline in diatoms. This is largely attributable to the degree of competition in the model, which depends on the choice of zooplankton prey selectivity and grazing formulations54. By contrast, a recent modelling study focusing on functional diversity and ecological redundancy has demonstrated that changes in phytoplankton composition can entail knock-on effects on primary productivity, trophic transfer and carbon export34. Accordingly, it is possible that the OA-driven loss of diatoms (owing to slower opal dissolution) could trigger additional ecological changes, which might in turn modify carbon cycling and export fluxes. However, the low degree of ecological detail in most Earth system models (including ours) does not enable an assessment of such complex knock-on effects.

Another important factor to consider is iron. It is well known that iron limitation enhances silicification of diatoms, increase cellular Si:N by up to two- or threefold55,56,57. Thus, future changes in iron supply may cause shifts in the Si:N of particle flux from the surface ocean, which could theoretically counteract/enhance the consequences of OA-enhanced Si:Nexport to some extent. In this context, the largely iron-limited Southern Ocean is of key interest, as iron supply might increase in the future, owing to increased aerosol dust deposition and melting ice45,58. This would alleviate iron limitation and reduce the Si:N of diatoms, thereby possibly counteracting the OA-induced increase in Si:Nexport. Our model includes a parameterization for iron-limitation effects on opal production by diatoms, thereby also controlling Si:Nexport. However, as with most other Earth system models59, it does not account for the complex mechanisms that may lead to future changes in iron deposition (for example, aeolian dust deposition, input from ice melting). Simulated iron inputs to the ocean are fixed at preindustrial rates. Thus, it is not possible to directly assess how future changes in iron limitation might interact with OA-driven changes in opal dissolution. However, in this context it is important to differentiate between effects on Si:N in the surface ocean (for example, iron effects on during production) and those occurring throughout the water column (for example, due to OA effects on dissolution): changes in the total amount of produced biogenic silica would alter the magnitude of Si flux from the euphotic zone, whereas the pH effect on Si dissolution alters flux attenuation throughout the water column. The relative size of the OA effect (that is, the proportional decrease in dissolution of sinking opal) is thus independent of the magnitude of Si flux from the surface. Accordingly, it can be assumed that future changes in dust deposition and Si:N (for example, in the Southern Ocean) would be superimposed by the OA-driven decrease in opal dissolution.

Furthermore, effects of opal ballasting on sinking velocities and remineralization rates of in diatom-derived organic matter are not accounted for in the model. In theory, the OA-driven decrease in opal dissolution may increase particle sinking velocities owing to enhanced (that is, prolonged) mineral ballasting by opal60. At the same time, slower opal dissolution may enhance the protection of organic matter against remineralization to some extent61. On the basis of such considerations, it is often suggested that a decrease in diatoms (such as the OA-driven loss of diatoms reported here) may reduce the efficiency of the biological carbon pump owing to slower sinking speed and/or faster remineralization of sinking organic matter of non-diatom origin1. However, as the mechanisms on the particle scale described above are very complex and evidence on mineral ballasting in observational data is controversial, they are not included in our model (and neither in other, ecologically more complex, models)34,62. Nevertheless, these mechanisms should be kept in mind when interpreting our results, as they could potentially have repercussions for the efficiency of the biological carbon pump.

Altogether, we note that our model results should be considered as an important estimate of how OA impacts on opal dissolution and Si:Nexport may affect the marine silica cycle and the future biogeography of diatoms. As the OA-driven decrease in opal dissolution is a purely abiotic mechanism, we consider the main findings from our model simulations (decline in Si(OH)4 and diatoms) to be robust and valid on large spatiotemporal scales, and expect them to be reproducible with other Earth system models. However, the model properties outlined above result in only minor knock-on effects of OA-driven changes on primary productivity and carbon export. Whether this holds true for the real ocean is uncertain and other Earth system models, for example, those with a higher degree of ecological complexity, may yield somewhat different results. For instance, although the OA effect on Si(OH)4 and diatoms might be similar, possible responses of plankton community structure and biogeochemical processes may differ depending on the ecosystem model structure. Therefore, we hope that our study will be an incentive for the scientific community to explore OA effects on opal dissolution with different global models, and thereby assessing the potential variability of this effect among models.

Source link