We investigated amorphous Si (ASi) in a boreal wetland in northern Sweden. We found enormous stocks of ASi in the upper soil layers (up to 11% of dry weight), in the form of diatom frustules and plant ASi. A consistent exponential decrease in ASi concentrations was observed with increasing depth in the soil profile. An inverse modeling approach shows that vegetation takes up a substantial part of weathered dissolved Si (DSi). Concurrent analysis of N and C indicates a faster turnover in and a higher leakage from the ASi pool. The magnitude of the biological buffering we observed is unprecedented and supports the emerging paradigm of the importance of biological uptake of DSi governing the export of DSi from terrestrial ecosystems. Our results complicate current models of silicate transport, highlighting the necessity to incorporate ecosystem biological buffering in our concept of Si biogeochemistry.
 All dissolved silica (DSi) transported from the continents toward the coastal zone eventually originates from the weathering of silicate minerals. Uptake of soil CO2 during this Si weathering is the most important control mechanism on atmospheric CO2 concentrations on timescales longer than 105 years [Berner et al., 1983]. The links between the global C and Si cycle are even tighter. Export of terrestrial DSi is important for DSi availability in the ocean. Diatoms, which have an obligate requirement for Si, form the most important permanent ocean carbon sink [Dugdale et al., 1995]. About 3% of diatom amorphous Si (ASi) production in the oceans (6 Tmole yr−1) is buried permanently on the seafloor and continuous replenishment of DSi by riverine input (6 Tmole yr−1) is necessary to maintain a balance in the oceanic biogeochemical Si cycle [Tréguer et al., 1995], and maintain the biological C and Si pump.
 In state-of-the-art continental watershed models describing DSi fluxes, a direct link between silicate weathering and riverine DSi fluxes and concentrations is assumed [Millot et al., 2003; Cugier et al., 2005]. Thus, chemical weathering occurring on geological timescales (>104 years), is considered the controlling factor governing the export of DSi occurring at biological timescales (<102 years). However, recent studies have shown that the uptake of weathered DSi by plants in terrestrial ecosystems and the subsequent storage as amorphous silica (ASi) in soils and vegetation, buffers the DSi export from the continents [Conley, 2002; Derry et al., 2005; Farmer et al., 2005; Street-Perrott and Barker, 2008]. The magnitude of the buffering is as of yet uncharted scientific territory. A correct quantification is necessary to better constrain the relation between riverine DSi fluxes and chemical Si weathering and quantify the carbon sink associated with silicate weathering, both on shorter and longer timescales.
 Weathering products exported from a basaltic Hawaiian watershed show that vegetation takes up between 60 and 90% of all DSi released during chemical weathering, with only 10–40% of DSi in rivers directly originating from bedrock weathering [Derry et al., 2005]. The remaining part of the riverine Si has been incorporated at least once into vegetation providing a significant biological buffer. Other results have shown the existence of intense biological cycling of Si in forests [Gérard et al., 2008; Fraysse et al., 2009], with significantly increased DSi fluxes observed after experimental forest harvesting [Conley et al., 2008] as a result of ASi dissolution. Yet, apart from these pioneer results, the impact of ecosystems and biotic uptake of soil DSi on eventual export toward rivers is essentially unquantified. This limited knowledge of the biological control by ecosystems is a critical knowledge gap in the rapidly evolving concept of the terrestrial Si cycle and the direct link to carbon cycling and ocean productivity [Street-Perrott and Barker, 2008]. Wetlands attract special attention in this context. Located at the interface between aquatic and terrestrial environments, they are rich in ASi accumulating vegetation, sponges and diatoms; wetlands have been hypothesized to be key ecosystems in the emerging paradigm of biological control on DSi fluxes [Struyf and Conley, 2009]. Yet, apart from freshwater tidal marshes [Struyf et al., 2006], which have a very specific hydrology and biogeochemistry, there are little previous studies directly addressing storage and recycling of ASi in wetlands [Struyf et al., 2009].
 More specifically, boreal wetlands are well known biogeochemical hot spots, controlling both the amount transported and the speciation of carbon, nitrogen and phosphorus, upon their transfer from terrestrial to aquatic ecosystems [Updegraff et al., 1995; Bridgham et al., 1998]. Vegetation is often composed of wet sedges, plants with a high capacity to store ASi. Furthermore, diatoms are abundant in boreal wetland soils: their abundance is tightly linked to changes in hydrology and vegetation. Diatom ASi storage could constitute a currently unaccounted for sink of ASi in boreal wetlands.
 We investigated the stocks of ASi in a boreal wetland in northern Sweden. We hypothesize that boreal wetlands accumulate a significant amount of ASi. Using inverse modeling and recycling rates estimated from the depth gradient, we will quantify the importance of the wetlands in controlling watershed-scale Si fluxes. We hypothesize that ASi storage in wetlands exert essential control on the flux of Si through boreal watersheds.
2. Materials and Methods
2.1. Study Area and Sampling
 The Muddus National Park is unique through the protection of a large area (49 340 ha) of undisturbed Swedish subarctic taiga. The typical landscape is mosaic with virgin pine forest and Norway spruce, mosaic bogs and fens. The terrain is mostly flat with only a few peaks. Average temperature was close to 0°C, annual rainfall was 613 mm and annual runoff was 373 mm for the period 1983 to 2006. The Muddus catchment (450 km2) is typical for forested catchments in this region, where spring flood comprises at least 50% of the yearly runoff: high concentrations of dissolved elements occur during low discharge and low concentrations during high discharge, but vice versa for DOC and H+ [Smedberg et al., 2006].
 Thirteen soil cores were taken at three wetland sites within Muddus National Park on 10 September 2007 (Figure 1). All sites were characterized by dominant sedge and Spaghnum vegetation. Cores were between 30 and 80 cm long and were subsampled into 5 cm intervals. We opted for 5 cm sampling intervals rather than analyzing only one core in more detail, as to better cover potential variability of ASi concentrations within the chosen sampling sites. Samples were stored at 4°C until freeze-dried ca. 1 month after sampling. The selected sampling sites were chosen to be representative for wetlands in the National Park. We expected no seasonality in the ASi concentrations in the cores: yearly fluxes of DSi from the ecosystems were hypothesized to be low compared to the standing ASi stock (which was confirmed by our results, see further).
 The freeze-dried samples were sieved through a 250 μm screen to separate fine peat from macroscopic roots and litter. The macrolitter was subsequently ground to a 200–1000 μm size. Both fractions were analyzed for ASi content by alkaline extraction of 30 mg of soil material in 1% Na2CO3 solution during 3 h and subsequent colorimetric analysis of the DSi concentration in the extract [Conley and Schelske, 2001] (measuring error < 1 μM). ASi can be composed of biological remains (diatoms, sponges, plant phytoliths) and mineralogenic ASi derived from DSi weathered from minerals and biological remains. Roots and litter were only analyzed if they comprised more than 10% of the dry sample weight. No simultaneous dissolution of DSi was observed from minerals in the ASi analysis during sequential extractions of randomly selected samples. The fine peat fraction was analyzed for total N and C content using a Carlo Erba NC2500. About 1–2 mg of the sample was put in small tin capsules and combusted at 1000°C in a He stream flow.
2.3. Carbon Dating and Microscopy
 Plant remains collected from the 30–35 cm depth from two cores at each site were C-14 dated using AMS (accelerator mass spectrometry) at Lund University C-14 laboratory [Hellborg and Skog, 2008]. Radiocarbon dates were calibrated using OxCal 3.10. For the same cores, ca. 1 g of peat from the surface layer and from the 30–35 cm layer were ashed at 550°C during 3 h, and subsequently used to microscopically examine ASi composition.
2.4. PHREEQCI Inverse Modeling
 An inverse modeling approach was used to explain the sources and sinks for Si and other major elements (i.e., base cations, Al, Cl−, SO42−) and to link the formation of alkalinity to different bedrock and soil mineral fractions. The stoichiometry of the final solution is deduced from a given set of minerals in combination with the mineral reactivity. Inverse modeling has been extensively used in other studies [e.g., Garrels and Christ, 1965]. In later years, computer programs developed by USGS have adopted this approach, e.g., Netpath [Plummer et al., 1994] and PHREEQCI [Parkhurst and Appelo, 1999]. In the current calculations PHREEQCI was used (the abbreviation emphasizes the features that the program stands for: PH (pH), RE (redox), EQ (equilibrium), C (the programming language) and I (interactive)). The calculations assume steady state and are based on data for the period 1983 to 2006 for both the precipitation (Pålkem station, about 100 km NE from Muddus, data from www.ivl.se) and the monitoring station at the Muddus catchment outlet (data from www.ma.slu.se). The entire Muddus catchment area was treated as one unit where the deposition was the input solution to the catchment, i.e., PHREEQCI was used to calculate mineral reactions from the deposition to the monitoring station at the Muddus outlet in river Luleälven. Winter data was used in the calculations (base flow runoff data, from October to April): here control by mineral weathering (lack of surficial runoff) will be highest. The turn over time for water is on average approximately 2 months using groundwater storage volumes from modeled hydrology by Smedberg et al. . The primary mineral composition of the soil used for the inverse modeling consisted of plagioclase, biotite, dolomite and halite (see Table 1 for composition of each mineral). The Ca/Na ratio is used to separate Ca for silicates and Ca coming from carbonates [Mortatti and Probst, 2003]. In this study the plagioclase Ca/Na ratio was assumed to be 1: a nearby river (Råne River), which has no known carbonate sources for Ca, has this ratio (data from www.ma.slu.se). The error in the modeled data was less than 5% for all elements except Al. This error is not an error for the whole model as such, but indicates how well the model fits observed data (Table 2). The error for Al was much higher, but considering the very low concentrations of both modeled and measured Al this is not critical for the model outcome.
Table 1. Chemical Composition of Minerals and Other Substances Used in the Inverse Modeling
Table 2. Summary of Deposition and Runoff Chemistry and Weathering of Minerals and Release of Elements in the Inverse Model With Al(OH)3 as Preferred Al Sinka
Here a, amorphous; g, gas. Adjusted deposition is corrected for evapotranspiration. All values are in mmol L−1.
3.1. Amorphous Si
 ASi concentrations in the fine peat fraction (<250 μm) comprised up to 11 wt % of dry matter (Figure 2a). ASi concentrations in the macrolitter and root fraction (>250 μm) were considerably lower (Figure 2b). In both fractions, ASi concentrations were higher in the upper soil layer. A consistent exponential decrease with depth was observed in total ASi storage (g ASi m−2), either from the surface or 5–10 cm depth (Figure 3). Lower ASi storage in the surface layers was attributed to a lower bulk density of the top peat layer, which consisted mostly of unconsolidated macrolitter (Figure 5). Light microscopy of ashed material indicates the presence of both diatoms and phytoliths (ASi remains from plants) and little or no mineralogenic matter. Site 1, which has the highest ASi concentrations, has more abundant diatoms compared to the other sites, which have a larger contribution of plant material (Figure 4). Intact diatom frustules are found in the surface layers with highest ASi content, whereas in deeper layers the ASi is primarily composed of siliceous debris, indicating gradual ASi dissolution (Figure 4). Similar patterns were observed in the other cores, but diatoms were less abundant, consistent with lower ASi contents at the respective sites (not shown).
3.2. ASi Dissolution Rate
 The sharp exponential decrease in ASi concentrations and total storage with increasing depth was assumed to be caused by gradual dissolution of the ASi. Assuming a constant rate of peat deposition over the upper 35 cm depth section at all sites, we estimated the annual dissolution rate of ASi. The age of the 30–35 cm depth layer ranges from 715 to 1045 years BP at sites 1 and 2 and between 3450 and 4650 years BP at site 3. At site 3, ASi stocks drop by a factor of 16 or more, at site 1 and 2 by a factor of 4, consistent with the different accumulation rates suggested by the peat age at 30–35 cm. The resulting estimated ASi recycling rate (from exponential decrease in average ASi storage per layer, Figure 3) ranges between 0.2 and 1.4 g DSi m−2 yr−1. This simple calculation does not account for compaction in the deeper part of the 30–35 cm section. However, any compaction in the deeper section, would result in a shortening of the age span covered by the upper sections, where decrease in ASi concentrations is steepest, resulting in a faster turnover of the ASi. The increase in nitrogen and carbon total weight per peat volume (see further) in the deeper part of the upper 35 cm section of the cores, indicates that compaction is actually occurring, as does the increasing bulk density with depth (Figure 5). The estimated ASi recycling rates are therefore conservative lower estimates from the limited dating we applied compared to the potentially occurring ASi recycling rates. Given we dated only one depth layer, we prefer the conservative estimate.
Table 2 summarizes deposition and runoff chemistry and weathering of minerals and release of elements in the inverse model with Al(OH)3 as Al sink. We did not use kaolinite as Al sink, as reactions kinetics for kaolinite formation in colder/temperate regions impede its formation [Gustafsson et al., 1999]. Still, a recent study on Iceland assumed kaolinite as the formed clay mineral [Gislasson et al., 2009]. In a study from Svartberget [Lundström et al., 2000], a comparable catchment in the same climate region (approximately 300 km S), it was found that Al found in the soil was either bound to organic material (high and low molecular weight acids) or in hydroxide form. The abundance of imogolite and allophane was too low [Karltun et al., 2000] for their formation to be considered an important removal mechanism for Al. In the case of the Al(OH)3 model, there is an uptake of Si into ASi, corresponding to 37% of the whole Si budget (Table 2). Assuming, although unlikely, imogolite and allophane formation (Table 3) as dominant Al sink, results in a net release of about 3% ASi. In the case of kaolinite formation this amounts to a DSi release from ASi of approximately 30% of the whole DSi budget (Table 3).
Table 3. Contribution of ASi to Base Flow Boreal Wetland Si Fluxes in the PHREEQCI Inverse Modeling Assuming Different Pathways for Al Removala
Sink for Al
Percent From BSi
A negative number indicates a sink.
3.4. Nitrogen and Carbon
 In contrast to ASi, carbon concentrations (wt %) were consistently higher with depth at all sites ranging from 30 to 40% in the upper soil layers to about 53% in the deeper soil layers (Figure 5). Total carbon content also increased from 10 to 30 mg cm−3 in the upper soil layers, to 60 to 80 mg cm−3 in the deeper soil layers.
 At sites 1 and 2, nitrogen concentrations (wt %) increased from 1.5 to 2% in the upper soil layers, to about 3% in the deeper soil layers. At site 3, highest N concentrations were observed in the upper soil layers (3.5 wt %), and a consistent decrease was observed in the deeper soil layers up to 80 cm (Figure 5). In the 20 to 50 cm depth layer, N concentrations were consistent over all sites (around 3%). Sites 1 and 2 were not measured in the deeper soil layer. The nitrogen content increased consistently over all sites up to a depth of 30 cm, from less than 2 mg cm−3 in the upper soil layers, to 3–6 mg cm−3 at 30 cm depth.
 For the fine peat fraction, which comprised over 80% of the biomass for the majority (>85%) of samples, we observed the opposite pattern for N and C content (increase in total storage with depth) and for ASi content (strong decrease in total storage with depth), resulting in a strong increase in C/Si and N/Si ratios (by weight) with depth. In the surface layers C/Si and N/Si are around 8 and 0.4 respectively, while at 50 cm depth C/Si and N/Si range between 3 and 19 (N/Si) and 50 and 400 (C/Si). In the deeper soil layers at site 3, ratios increase even further to 45 (N/Si) and 1100 (C/Si).
4.1. Biological Concentration of Si in Boreal Wetlands
 The high concentrations of ASi we observed in the top 10 cm of peat at all three sites, in combination with the steeply decreasing depth gradient of the ASi, are characteristic for the efficient biological uplift and fixation of weathered DSi and for subsequent intense recycling of this biologically fixed Si [Jobbagy and Jackson, 2004]. Sedges are well known active Si accumulators, using energy to take up DSi from soil water. The concentrations higher than 5% of ASi in dry weight are unprecedented in wetland research [Struyf and Conley, 2009]: in other wetland soils concentrations often range between 0.2 and 2% [e.g., Struyf et al., 2006; Struyf and Conley, 2009]. Our findings confirm the abundance of diatoms in boreal wetland soils, as previously observed in the Stordalen wetland in north Sweden.
 We hypothesize that bedrock plagioclase weathering in the Muddus catchment is increased by the secretion of organic stimulants from vegetated topsoil layers. Vegetation can increase the release of DSi from silicate minerals in a number of ways, e.g., the secretion of organic acids and chelates [Cochran and Berner, 1996; Moulton and Berner, 1998]. The (partial) uptake of this DSi in the vegetation forms the prime driver for the accumulation of the ASi pool in upper wetland soil layers: sedges actively transform DSi into ASi. The substantial cover of forest (52% [Smedberg et al., 2006]) in the Muddus catchment likely adds to the enhanced DSi weathering in the catchment. As the wetlands are situated most closely to the riverine continuum, and wetlands are mostly groundwater fed, DSi weathered in the catchment is available for uptake by the sedge vegetation in the wetlands.
 The abundance of ASi rich litter in the surface layers enriches surface water with DSi: ASi rich litter in wetlands rapidly releases DSi, increasing DSi concentrations in pore water [Struyf et al., 2007]. We hypothesize that this DSi rich environment is ideal for the development of diatoms. The abundance of diatoms is thus the by-product of the active plant uplift of soil Si. Diatom growth is likely limited to the few upper cm where light can penetrate through the loose macrolitter, allowing growth of both benthic and epiphytic species.
 In this hypothesis, the uptake of plagioclase weathered DSi in the biological Si buffer retards DSi export from the Muddus catchment: this is also seen in the most likely PHREEQCI model for the catchment (with Al-hydroxides and biological adsorption accounting for Al sinks), in which uptake of DSi into ASi net accounts for about 37% of weathering produced DSi at base flow.
 Extrapolation of our results indicates the current storage of ASi in the wetlands of Muddus National Park (which comprise 39% of the total surface area with a further 52% forest and 6% grassland) is 21,000–143,000 tons of ASi in the top 30 cm. This is 2–3 orders of magnitude greater than annual average DSi fluxes (400 tons DSi) from the entire catchment [Smedberg et al., 2006]. Although there was a large variability in total Si storage among the sites we sampled, all results indicate that once Si is incorporated into the biological buffer, it can remain stored and recycled within the wetland ASi buffer for several centuries before its release.
4.2. Role of Vegetation and the Amorphous Si Stock in Total Si Fluxes
 It might be expected that the uptake of weathered DSi in this huge biological buffer actually decreases net Si export from the Muddus catchment. Yet, comparison of Muddus with a nearby, similar sized catchment with no wetland cover, shows yearly export fluxes of DSi from the Muddus watershed (0.9 g Si m−2 yearly) [Smedberg et al., 2006] are 300% higher compared to its neighbor without a mosaique of wetland and forest cover. The neighbor watershed, Vuojatätno, had an annual average DSi export of only 0.2 g m−2, comparable to bedrock weathering rates (= base flow groundwater fluxes) in other boreal taiga regions (0.1–0.2 g DSi m−2) [Zakharova et al., 2007], although specific water yield from the latter watershed was higher. Moreover, a compilation of data from boreal watersheds indicated yearly average concentrations of DSi in rivers were higher if more wetlands were present [Struyf and Conley, 2009].
 The export of DSi from the Muddus watershed observed by Smedberg et al.  corresponds reasonably well with our estimated ASi recycling rates from the observed ASi depth profiles. Recycling rate estimates were between 1 and 1.4 g m−2 yr−1 for the upper 40 cm at site 1, and between 0.2 and 0.3 g m−2 yr−1 for the same layers at sites 2 and 3. These are conservative estimates: they do not account for any compaction of the soil in deeper layers: the N and C concentration gradients, as well as total N and C storage indicate compaction occurs in the deeper layers. The steepest depth gradient in ASi, observed in the upper soil layers, therefore spans a shorter time interval than estimated by our crude linear age model, meaning our age model will likely underestimate the yearly recycling rate, as it overestimates residence time of ASi in the most active dissolution layers. More detailed dating profiles in future studies will allow a better quantification of residence time of ASi within the buffer, as well as more detailed recycling rates.
 The net modeled uptake of base flow DSi into vegetation in our preferred PHREEQCI model, as well as the fairly good correspondence between ASi recycling rates and net export DSi fluxes from the catchment, all indicate an intense biological cycling of Si in the boreal wetlands of the Muddus catchment, that impacts strongly on net watershed Si fluxes. The results indicate that a large part of the biotically increased weathering fluxes (organic chelates, acids, ligands) end up being stored in a huge, but intensely cycled biological buffer. It could well be that the Muddus wetlands are currently in a near steady state, where uptake in the biological buffer is compensated for by the about equal release of DSi from ASi dissolution. A large part of the export of DSi from the ASi buffer might actually occur during snowmelt and summer precipitation events, as surficial soil water rapidly runs off during these periods. Although concentrations of DSi in river water are up to 6 times diluted during these periods [Smedberg et al., 2006], lowest concentrations in the Muddus catchment are always higher than highest observed DSi concentrations in the Vuojat catchment (without wetlands and forests) [Smedberg et al., 2006]. Precipitation related event discharge and relatively young soil water (residence time shorter than 3 months) discharge combined constitute about 80% of water fluxes through the Muddus watershed [Smedberg et al., 2006]. TOC export is also tightly linked to recycling from surface organic soil layers. The significant positive relationship observed between yearly TOC (mainly exported in undeep soil water fluxes) and DSi export from boreal Swedish watersheds seems to confirm the large biological influence on boreal DSi fluxes [Humborg et al., 2004].
4.3. A Leaky Reservoir
 Taking other Al sinks into account (though less likely), such as kaolinite or imogolite and allophane formation, the ASi buffer would be a net release of DSi, adding to net plagioclase weathering fluxes. This would suggest that the Muddus wetlands are currently not in steady state, and are actually loosing ASi from a previously accumulated stock. The dependence of our inverse modeling approach on the chosen Al sink implies it is difficult to precisely establish the impact of the sedge and diatom ASi buffer on export DSi fluxes. Still, our results do imply the existence of an intensely cycled, enormous soil ASi pool in boreal wetlands, not or incompletely considered in previous studies of boreal Si fluxes [Millot et al., 2003; Zakharova et al., 2007], but exerting a substantial influence on boreal Si fluxes nevertheless. Zakharova et al.  did acknowledge the potential influence of biotic cycling, but did not account for diatoms or soil stocks of ASi, only for recycling of annually produced biomass and litterfall, which will cause a strong underestimate of the biological buffer on watershed-scale fluxes.
 The intense recycling of ASi from the surficial peat layers is supported by the large increase in C/Si and N/Si ratios with depth. Nitrogen and phosphorus are in short supply in the Muddus watershed (yearly average inorganic N concentration in runoff is 3.4 μmol L−1, the dissolved inorganic P is 0.08 μmol L−1) [Humborg et al., 2004], while yearly average DSi concentration is 78 μmol L−1. Compared to N and P, Si is available to plants in unlimited supply through the groundwater. Vegetation in (sub) arctic bogs and fens is adapted to low nutrient concentrations, by the efficient mechanisms of resorption and reallocation and direct uptake of soil organic N and P [Jonasson and Shaver, 1999]. Continuous groundwater input probably reduces the need for efficient resorption of dissolving ASi. Low temperatures and pH also decrease biological decomposition processes, catalyzing efficient retention of C. In contrast to C, release of DSi from decomposing vegetation is primarily a physical rather than a biological process [Struyf et al., 2007]. The strongly declining Si/C and Si/N ratio with depth suggests the enormous ASi pool is much more leaky than the stocks of C and N.
 Known Si accumulators present in the peatlands of Muddus National Park, such as sedges (Eriophorum sp. and Carex sp.), are present in abundance throughout the Arctic. In the next century, climate change is predicted to significantly warm subarctic and arctic environments, and intensification of precipitation is expected. This may increase dissolution and export of DSi from the large ASi pool, and might shift the apparent steady state of the ASi pool in a release state. How this could affect Arctic Ocean silica cycling is as of yet unknown: a large inflow of DSi due to climate change could induce dramatic changes in phytoplankton composition and food webs downstream. On a longer timescale, DSi export as a result of physical weathering has doubled during glacial periods [Froelich et al., 1992]. The role of biology in this process is unknown, but our results suggest peatland ASi pools could significantly add to increased DSi export. Future research should also test the effect of pH, which is known to have an important influence on ASi recycling [Loucaides et al., 2008]. The uptake of DSi into the ASi buffer is not a simple time lag between weathering and eventual release into rivers. In the end, it is the complex interaction between vegetation, bedrock weathering, hydrology, diatom abundance, pH and climate which will control not only uptake of DSi into the ASi stock, but also its recycling and export. The sheer size of the ASi stock indicates that changes in recycling and export rates of DSi from the buffer could strongly change watershed DSi yield.
 Our results complicate geochemical models of Si weathering, which currently do not account for any biological buffering of DSi weathering, and link it only to bedrock lithology. Incorporation of biological buffering, as shown here, in future models of silica weathering is necessary. The enormous biologically active pool supports the emerging paradigm of the importance of biological uptake of DSi and recycling of ASi on global scales, and its feedback to ocean productivity and the carbon cycle.
 Eric Struyf and Daniel Conley acknowledge EU Marie Curie Actions (SWAMP MEIF-CT-2006-040534, COMPACT MEXC-CT-2006-042718) for funding. Eric Struyf acknowledges FWO (Flemish Research Foundation) for funding his postdoc grant. We would further like to acknowledge the Crafoord Foundation (Sweden) for funding our research. This study was also supported by the Swedish Research Council (VR; 70384101) and the Swedish Research Council for Environment, Agricultural Sciences, and Spatial Planning (FORMAS; 214-2008-202).