4.1. Fe Solubility of Volcanic Ash: Dissolution Rates and Sources of Soluble Fe
 The timescale at which the soluble and thus potentially bioavailable Fe that is released from volcanic ash (Figure 2a) is similar to the timescale at which ash particles sink through the photic zone of the surface ocean: few minutes for coarse (2000–500 μm), to 1–2 h for intermediate (250–150 μm), and to 1–2 days for fine (<50 μm) ash particles (based on Stokes' law estimates [Duggen et al., 2007]). Shorter residence times of about 1–2 h may arise for aggregates of ash particles formed during the humid eruption conditions (e.g., ∼1750 m/d for Pinatubo 1991 ash [Wiesner et al., 1995]). Volcanic ash is a mixture of various particles or components with <2 mm diameter that potentially may release Fe on contact with seawater and on different timescales, such as glass shards (quenched magma fragments), pyrogenic minerals (i.e., silicates and oxides), lithic particles (e.g., eroded rock material from the volcanic conduit of any origin) [Fisher and Schmincke, 1984; Óskarsson, 1981].
 The surface of the ash particles are coated by a thin layer of salts in the form of Fe sulfates and Fe halides that are formed through the interaction of ash particles with volcanic gases (S, HCl and HF) and aerosols in the eruption plume [Delmelle et al., 2007; Naughton et al., 1976; Óskarsson, 1980, 1981; Rose, 1977]. Although Fe content of these salts is still unknown, they are likely to be the most soluble components during seawater dissolution of ash particles [Duggen et al., 2007, 2010; Frogner et al., 2001; Jones and Gislason, 2008]. Volcanic glass shards on the other hand usually dominate the bulk composition and can have Fe contents ranging from <1 wt.% to well above 10 wt.% (e.g., Figure 4a; 1–5 wt.% for SZVA and 8–11 wt.% for HSVA). Fe content of pyrogenic minerals ranges from trace to major element level; from basically almost zero (e.g., plagioclase) through 10–30 wt.% FeO (e.g., clinopyroxene) to up to 50–70 wt.% (e.g., magnetite) [Nakagawa and Ohba, 2003].
Figure 4. Diagrams display the (a) seawater Fe release of volcanic ash versus the Fe content of volcanic glass shards and (b) seawater Fe release of volcanic and mineral dust versus the total Fe content. Dashed lines show the hypothetical fractional Fe solubilities (%Fes) ranging from 0.001% to 1% (%Fes = (Dissolved Fe/Total Fe)*100).
Download figure to PowerPoint
 The relation between the total iron and the dissolved iron from volcanic ash is shown in Figure 4. As illustrated in Figure 4, no correlation exists between the Fe release (or Fe solubility) and the Fe content of the volcanic glass or bulk ash samples. Ash samples with lower glass Fe content (<6 wt.%) generally release more soluble Fe than the samples with higher Fe content (>8 wt.%) (Figure 4). Although the salt coatings make up less than 1% of the mass of the bulk sample the data, Figure 4a suggests that very rapid (minute-to-hour scale) release of Fe from ash is most likely dominated by swift dissolution of the surface salts rather than the glass shards, which is in accordance with what was argued in previous studies [Duggen et al., 2007; Frogner et al., 2001; Jones and Gislason, 2008]. A recent study with Etna volcanic ash suggests that initial alteration of glass shards and mineral particles may also partly contribute to rapid Fe release [Censi et al., 2010]. On longer timescales (days through weeks to years), alteration of volcanic ash particles deposited as an ash layer at the seafloor is controlled by their bulk chemical composition and may significantly contribute to the surface ocean marine biogeochemical Fe cycle through upwelling.
 For the aerosol samples collected over the Pacific Ocean a trend of increasing Fe solubility with decreasing Fe content was reported [Zhuang et al., 1992] similar to what is found for volcanic ash (Figure 4). This trend can be linked to the low Fe contents (or low abundances) of the relatively more dissolvable Fe components. Due to their high Fe contents iron (hydro)-oxides (e.g., hematite, magnetite 60–80 wt.% Fe) are commonly assumed to be the major sources of iron into the surface ocean [Mahowald et al., 2009]. However, it has recently been found that clay minerals are much more soluble although they contain relatively less Fe (<3%–20% Fe) [Journet et al., 2008]. The lack of a correlation between total Fe and the dissolved Fe thus demonstrates that the seawater Fe solubilities of volcanic ash or mineral dust cannot be inferred from the total Fe content but has to be determined directly. Either total iron or the dissolved fraction is not constant rather changing progressively during the long-range transport in the atmosphere. Particle size distribution [Baker and Jickells, 2006], mineral composition (aeolian fractionation [Duggen et al., 2010]), and the particle-surface chemistry (chemical and photochemical atmospheric processes [Duggen et al., 2010; Jickells and Spokes, 2001; Spokes and Jickells, 1996]) may enhance the solubility and bioavailability of Fe in the oceans.
 Sample storage is another factor that possibly may affect the Fe mobilization behavior in laboratory experiments. The soluble Fe salts on volcanic ash particles are likely to be unstable and may be affected by storage duration of the sample. Based on the reanalysis of a single ash sample, Jones and Gislason  argued that aging of ash material during storage might reduce the Fe release. As inferred from our new data in Figure 5, ash samples stored for more than 10–20 years tend to release less Fe on contact with seawater than younger samples. Five ash samples from Sakura-jima volcano even display a negative correlation between their age and the amount of Fe mobilized. If considered an aging effect, the Sakura-jima ash samples point to a decrease in rapid Fe release of about 200 nmol Fe/g ash over the course of 25 years. The data therefore suggest that Fe release data inferred from volcanic ash several or more years old are generally minimum estimates and that data from younger samples is more reliable.
Figure 5. Graph showing the possible influence of storage time on the Fe mobilization behavior of volcanic ash samples. The gray field denotes the correlation between age and Fe release for ash samples from Sakura-jima volcano.
Download figure to PowerPoint
 From the larger data set available it can now be inferred (by taking into account the possible aging effect and uncertainties indicated by repeat measurements) that volcanic ash samples generally release between 35 and 340 nmol Fe/g ash, with a mean of about 200 ± 50 nmol Fe/g for SZVA and possibly around 70 nmol Fe/g for HSVA (Figures 2 and 5) during dry deposition into the surface ocean. The percental (or fractional) Fe solubilities (as commonly used for mineral dust) are calculated in order to allow comparison with the previous aerosol Fe solubility studies. The calculations are based on the Fe release data and the total Fe content of the samples as follows: %FeS = (Dissolved Fe/Total Fe)*100 (Table 1). Accordingly, the Fe solubility for volcanic ash (VA) ranges from 0.007% to 0.1% (%FeSbulk VA, using the bulk sample data) and from 0.003% to 0.2% ((%FeSVA glass, using the glass data) (Table 1). Since the composition of volcanic ash progressively approaches the composition of the glass shards contained during aeolian fractionation (see Duggen et al.  for details), the overall %FeSVA can be constrained to 0.003% to 0.2% (Table 4). As the SZVA samples stem from different volcanoes worldwide (Figure 1), the 200 ± 50 nmol Fe/g (0.01%–0.02% FeS) value is likely to be representative on a global scale (e.g., dry deposition estimate for global models), and, above all, appears to be largely independent of the bulk composition of the ash samples (Figure 4).
 For the Cape Verdian mineral dust sample, the calculated Fe solubility is 0.002% (Table 1). This is in agreement with Fe solubilities reported for other experiments performed at seawater pH (=8) ranging from 0.001% to 0.02% (Table 4) (e.g., 0.001%–0.02% [Guieu and Thomas, 1996] and <0.013% [Spokes and Jickells, 1996]). The strong dependency of Fe solubilities on the experimental setup with different starting materials and different solutions (e.g., aerosol versus soil samples, pH 4 versus pH 8 solutions) is discussed in section 4.4.
4.2. Regional Impact of Major Volcanic Eruptions on Surface Ocean Fe Concentrations
 For a case study of the regional impact a single major eruption we chose a well-constructed historical eruption of Barva volcano in the Central American subduction zone (Figure 6). The Barva eruption deposited at least 7.9 × 1016 g of ash that traveled at least 1000 km distance into the eastern equatorial Pacific Ocean [Kutterolf et al., 2008]. Sediment core data allowed the reconstruction of the distribution of the ash layer as well as the thickness that decreases from coast (up to 100 cm thick) to the remote ocean (<1 mm thick) (Figure 6). Assuming an ash density of 2400 kg/m3 at 30% porosity, ash layer thicknesses of 10 cm, 1 cm and 0.1 cm are recalculated to ash loads of ∼170 kg/m2, ∼17 kg/m2 ∼1.7 kg/m2, respectively.
Figure 6. Map showing the extent and particle load in an ash fallout area during a large-scale volcanic eruption, exemplified by the historical Barva eruption (∼322,000 years ago) in Costa Rica. Isopachs (ash thickness contours) were mapped on the basis of marine sediment core data [Kutterolf et al., 2008]. Fe concentrations (nmol/L) between the isopachs display the surface ocean Fe levels (100 m mixed layer depth) with ash loads decreasing with increasing distance from the volcano. The Fe levels were inferred from adding the calculated increase in dissolved Fe that is associated with an ash layer of a given thickness (using an average ash density of 2400 kg/m3 at 30% porosity over 1 dm2 surface area) to the background seawater Fe levels in coastal waters (typically 4 nmol/L) and the open ocean (typically 0.3 nmol/L) [Baker and Croot, 2010; De Baar and De Jong, 2001; Parekh et al., 2005]. The range of Fe concentrations originates from the variations in Fe release of volcanic ash (35–340 nmol/g ash, Figure 2). As an example, the maximum increase in Fe concentrations in the open ocean during the deposition of 1 mm ash layer was calculated as follows:[0.3 nmol/L] + [((2400 g/dm3 − (2400 g/dm3 × 0.3)) × (1 dm2 × 0.01 dm) × (340 nmol Fe/g ash)/(1000 dm × 1 dm2)] = 6 nmol.
Download figure to PowerPoint
 Volcanic ash fallout impact the surface ocean Fe concentrations depending on the initial seawater Fe concentrations prior to ash deposition, the Fe mobilization behavior of volcanic ash, ash load (ash-to-seawater ratio), the mixed layer depth, and the maximum concentration solubility of iron in seawater [Baker and Croot, 2010; Duggen et al., 2010]. In coastal waters, due to higher riverine and continental input, the surface ocean dissolved Fe levels are relatively high and range from 1 to 100 nmol/L with typical concentrations between 8 and 10 nmol/L [Baker and Croot, 2010; Boyd et al., 2007, 2000; Coale et al., 2004; De Baar and De Jong, 2001; Liu and Millero, 2002]. Due to high background levels Fe is generally not limiting phytoplankton growth in coastal waters, although exceptions such as the California upwelling region have been reported [Hutchins and Bruland, 1998]. Assuming an initial Fe concentration of about 4 nmol/L in coastal seawater, deposition of a 0.1 cm, 1 cm and 10 cm ash layer with the range in Fe mobilization (shown in Figure 2; 35–340 nmol Fe/g ash) could raise the dissolved Fe concentrations to about 5–10 nmol/L, 10–60 nmol/L, and 65–570 nmol/L, respectively (Figure 6). The inferred values are in accordance with strongly enhanced Fe levels determined in Mediterranean seawater (∼600–700 nmol/L) close to Sicily within the ash fallout area of the 2001 eruption of Etna volcano [Censi et al., 2010].
 High particle loadings in the close vicinity of a volcanic source may cause increased Fe scavenging of the particles (as seen in the mineral dust deposition [Baker and Croot, 2010; Guieu et al., 1997; Spokes and Jickells, 1996]) but the distinction of the release versus scavenging is hard to determine. The excess concentrations of dissolved Fe that are above the typical maximum Fe solubility in coastal areas (8–10 nmol/L) would most likely include a large colloidal phase, which is also potentially bioavailable and important to the overall Fe cycling. High Fe levels of several tenths to hundreds nmol/L as observed during the 2001 Etna eruption were argued to be linked to enhanced organic complexation, resulting from lysis of phytoplankton cells during a phytoplankton bloom associated with volcanic ash fallout [Censi et al., 2010]. The short residence time of Fe of about 2–3 months in the surface ocean, however, may limit the biogeochemical impact of volcanic eruptions.
 In the surface open ocean, Fe concentrations are extremely low (0.02–0.8 nmol/L) (Figure 7a) thereby limiting phytoplankton growth in HNLC areas [De Baar and De Jong, 2001; Parekh et al., 2005]. Assuming an initial Fe concentration for the upper ocean of about 0.3 nmol/L, deposition of a 0.1 cm, 1 cm and 10 cm ash layer could increase Fe levels to about 1–6 nmol/L (at 1000 km distance from the volcano), 6–57 nmol/L (up to 500 km away), and 60–570 nmol/L (up to 250 km from the volcano), respectively (Figure 6). Mesoscale Fe fertilization experiments show that an increase of Fe levels by only 2 nmol/L can stimulate massive diatom blooms in Fe-limited oceanic regions [Wells, 2003]. Therefore, even relatively low ash loads corresponding to millimeter-scale ash layers may be sufficient to cause a vigorous MPP response. Based on satellite data a recent study demonstrated a causal connection between the 2008 Kasatochi eruption in the Aleutians and a large scale (∼3500 km by 1500 km), about 2–3 months lasting phytoplankton bloom in the subarctic North Pacific [Hamme et al., 2010; Langmann et al., 2010].
Figure 7. (a) Average surface ocean iron concentrations based on the work of Parekh et al. . High-Nutrient Low-Chlorophyll (HNLC) regions are defined by comparison of the seasonally averaged surface nitrate and silica concentrations [Watson, 2001] and annual averaged chlorophyll concentrations (SeaWIFS). (b) Areas of higher versus lower likelihood of volcanic ash deposition, the extent of which are roughly estimated on the basis of the location of historically active volcanoes, low-altitude wind directions, and tephra frequencies in marine sediment drill cores in the Quarternary [Straub and Schmincke, 1998]. (c) Averaged annual mineral dust fluxes into the world ocean. Percentage inputs are as follows: North Pacific, 15%; South Pacific, 6%; Southern Ocean, 6%, North Atlantic, 43%; South Atlantic, 4%, and Indian Ocean, 25% [Jickells et al., 2005].
Download figure to PowerPoint
4.3. Flux of Volcanic Ash and Mineral Dust Into Pacific Ocean: Millennial-Scale Deposition Rates
 The flux of Fe into the Pacific Ocean can be constrained by combining Fe release with geological flux data. Although most of the explosive active volcanoes on Earth are located around the Pacific Ocean that hosts about 70% of the Fe-limited oceanic regions (Figure 7a), an estimate of the airborne volcanic ash input into the Pacific Ocean is so far not available in the literature. Below we therefore constrain the input of airborne volcanic ash into the Pacific, followed by an estimate of the volcanic ash soluble Fe flux, which is then compared with the Fe flux associated with Pacific mineral dust deposition.
 The volcanic ash flux into the Pacific Ocean can be considered constant and quasi-continuous over geological timescales, such as the past several 100 ka [Straub and Schmincke, 1998], and millenial mineral dust deposition can be considered largely constant after the last glaciation [Jickells and Spokes, 2001]. A meaningful way to compare the fluxes of volcanic ash and mineral dust, despite the differences in episodicity/seasonality of deposition, is therefore to recalculate mass and hence Fe fluxes to a postglacial millennial base. Being aware of the uncertainties and limitations of such global-scale estimates, the main goal is to constrain the order of magnitude of Fe release from volcanic ash compared to that of mineral dust, which will be useful to further improve our understanding of the potential role of volcanic ash deposition for the surface Pacific Ocean biogeochemical Fe cycle.
4.3.1. Volcanic Ash Flux Into the Pacific Ocean: Millennial-Scale Estimates
 Due to their different nature in eruption style two different approaches are advanced for estimating the fluxes of SZVA and HSVA: (1) an arc-length-based approach for subduction zone (SZ) volcanoes and (2) an apron-based approach for hot spot volcanoes (HS). In both cases we first estimate the amount of ash emitted from Pacific volcanoes and then the fraction that was deposited offshore into the Pacific Ocean. The flux estimates are briefly summarized below and outlined in more detail in Tables 2 and 3.
Table 2. Millennial Volcanic Ash Input From Subduction Zone Volcanoes Into the Pacific Ocean
|Subduction Zone Volcanic Arcs||Arc Length (km)||Volcanic Ash Emission (1015 g/ka)a||Offshore Fraction of Emitted Volcanic Ash (%)b||Offshore Deposited Volcanic Ash Into the Pacific Ocean (1015 g/ka)c|
|New Zealand–Tonga-Kermadec||2500||20.0||26.0||85 ± 5||16.0||23.4|
|Fiji Islands||340||2.7||3.5||85 ± 5||2.2||3.2|
|New Hebrides||1450||11.6||15.1||85 ± 5||9.3||13.6|
|Solomon Islands||390||3.1||4.1||85 ± 5||2.5||3.7|
|New Britian||1000||8.0||10.4||85 ± 5||6.4||9.4|
|Papau New Guinea||950||7.6||9.9||20 ± 10||0.8||3.0|
|Indonesia||4700||37.6||48.9||20 ± 10||3.8||14.7|
|Philippines||1610||12.9||16.8||60 ± 10||6.4||11.7|
|Ryuku Islands||1210||9.7||12.6||60 ± 10||4.8||8.8|
|Mariana||1500||12.0||15.6||85 ± 5||9.6||14.1|
|Izu-Bonin||1100||8.8||11.5||85 ± 5||7.0||10.3|
|Japan||1400||11.2||14.6||85 ± 5||9.0||13.1|
|Kuril Islands||1350||10.8||14.1||85 ± 5||8.6||12.6|
|Kamchatka||1000||8.0||10.4||85 ± 5||6.4||9.4|
|Aleutians||1900||15.2||19.8||85 ± 5||12.2||17.8|
|Alaska||800||6.4||8.3||40 ± 10||1.9||4.2|
|North Canadian Cascades||450||3.6||4.7||20 ± 10||0.4||1.4|
|High Cascades||1300||10.4||13.5||20 ± 10||1.0||4.1|
|Mexico||970||7.8||10.1||40 ± 10||2.3||5.0|
|Central America||1100||8.8d||11.5e||80 ± 10f||6.2||10.3|
|South America (North)||550||4.4||5.7||80 ± 10||3.1||5.2|
|South America (Central)||960||7.7||10.0||80 ± 10||5.4||9.0|
|South America (South)||1300||10.4||13.5||20 ± 10||1.0||4.1|
|Subduction zone total||29830||239||311||64 ± 8||126||212|
Table 3. Millennial Volcanic Ash Input From Hot Spot Volcanoes Into the Pacific Ocean
|Hot Spot Oceanic Islands||Volcaniclastic Production Rate (1015 g/ka)a||Offshore Deposited Volcanic Ash Into the Pacific Ocean (1015 g/ka)b|
|Easter hot spot||5.7||0.1||0.5|
|Hot spot total||104||2||9|
188.8.131.52. Subduction Zone Volcanic Ash Flux: Arc-Length-Based Approach
 Large (strato-)volcanoes, which are the sites of intermediate to major explosive volcanic eruptions, are generally found at nearly constant distances of about 60–100 km apart from each other. Therefore, the length of a subduction zone segment (arc) can be considered proportional to its potential volcanic intensity and thus emitted material flux [Sigurdsson et al., 2000]. The inferred ash flux per millennia and kilometer arc length of any active subduction zone can thus, as a first-order approximation, be applied to any other subduction zone segment in the Pacific.
 As a basis for our flux estimates we use the Central American Volcanic Arc (CAVA), which is among the most well-studied subduction zone segments in the Pacific Ring of Fire, with data for offshore deposited ash available for the past 191 ka. Volume estimates of CAVA ash deposits were obtained by fitting straight lines to data on plots of ln isopach thickness versus square root isopach area (e.g., Figure 6) [Fierstein and Nathenson, 1992; Kutterolf et al., 2008; Pyle, 1989]. It is important to note that the ash volume estimates inferred from discrete ash layers do not account for the dispersed ash, as this fraction is not visible due to mixing with nonvolcanic sediments (e.g., due to bioturbation). The missing dispersed ash fraction corresponds to ∼6% to 60% and on average for ∼30% of the total erupted mass [Peters et al., 2000; Rose and Durant, 2009; Scudder et al., 2009; Straub and Schmincke, 1998]. Hence the ash emission rate from CAVA is likely to be underestimated by about 30%.
 According to ash thickness contour (isopach) maps inferred from marine drill core data (e.g., Figure 6), about 1139 km3 of ash was emitted from the 1100 km long CAVA during the past 191 ka (Table S4) [Kutterolf et al., 2008]. By converting volume to mass, this corresponds to about 1680 Pg (Pg = petagrams = 1015 grams) of ash for the past 191 ka (using dense rock equivalent densities of 1680 kg/m3 for mafic and 1470 kg/m3 for felsic tephra [Kutterolf et al., 2008]). The ash emission rate per millennium can then be calculated to 8.8 Pg/ka and to a rate per kilometer arc of 8.0 Tg/ka/km (Tg = teragrams = 1012 grams) (Table 2). Taking into account the dispersed ash fraction (+30%), we infer an ash emission rate of between 8.0 and 10.4 Tg/ka/km for CAVA (Table 2). Assuming that the millennial ash emission rate (per kilometer arc length) is largely the same for all Pacific subduction zone segments, the total emission of all Pacific arcs can be inferred from multiplying the CAVA ash emission rate (of 8.0–10.4 Tg/ka/km) with the known lengths of individual arcs (Table 2). The millennial ash emission for Pacific SZ volcanoes is thus estimated to be on the order of 239–311 Pg/ka (Table 2).
 Only a part of the emitted ash is deposited over the ocean. In the case of the well-characterized CAVA, proximal and distal sections of ash layers are found both onshore and offshore (e.g., Figure 6) (Table S4). Based on the interception of proximal and distal facies (at ∼20–10 cm ash layer thickness [see Kutterolf et al., 2008]) it can be estimated that about 50% (220 Pg) of the proximal and about 90% (1120 Pg) of the distal ash was deposited into the Pacific Ocean over the past 191 ka (Table S4), which corresponds to about 80% ± 10% of the ash emitted from the CAVA (6.2–10.3 Pg/ka). The result is consistent with the location of Central American volcanoes within trade wind zone and with the frequency of ash layers in marine drill cores offshore Central America (Figure 7b).
 For other SZ segments along the Pacific Ring of Fire, the proportion of offshore deposited ash was estimated by taking into account: (1) the general wind directions (e.g., westerlies, trade winds), (2) the overall distance of the volcanoes from the ocean (Figure 7b), and (3) the constraints for onshore and offshore deposition of CAVA ash (Table 1). For example, in the eastern Pacific, ash from SZ volcanoes located in the trade wind area (∼30°N to 30°S) is mostly deposited offshore (e.g., 80% ± 10% for the northern section of the South American Arc; see Table 2 and Figure 7b), whereas ash from volcanoes situated in the westerlies (>30°N and >30°S) is dominantly deposited on land (e.g., only 20% ± 10% deposited offshore for the High Cascades; see Table 2 and Figure 7b). The distance of volcanoes from the ocean is generally larger in the eastern Pacific as these are mainly found at continental margins, whereas SZ volcanoes in the western Pacific often form ocean island arcs (Figure 7b). The proportion of the ash deposited offshore thus tends to be higher for island arcs volcanoes, compared to those located at continental margins (e.g., 85% ± 5% for Mariana Islands in the western Pacific; see Table 2). We also take into account the boundary of the Pacific Ocean (Figure 7b). The ash from Indonesian volcanoes located in the trade wind zone, for example, is mainly transported into the Indian Ocean rather than into the Pacific Ocean (only 20% ± 10% into the Pacific although 80% ± 10% is deposited offshore; see Table 2). Together, we infer a SZVA flux into the Pacific Ocean in the range of 126–212 Pg/ka, which takes into account uncertainties arising from the CAVA ash flux estimate, arc lengths and wind directions (Table 2).
184.108.40.206. Hot Spot Oceanic Island Volcanic Ash Flux: Apron-Based Approach
 The data basis from scientific ocean drilling is generally insufficient to construct isopach maps of offshore ash layers related to Pacific HS volcanic ocean islands (Figure 7b). In an apron-based approach, two end-member ocean island settings are distinguished: (1) caldera-forming and (2) non-caldera-forming hot spot systems (Table 3). Caldera-forming ocean islands (e.g., Hawaii) potentially create more explosive volcanic eruptions compared to non-caldera-forming ocean islands (e.g., Samoa, Marquesas) [Lipman, 2000] (Figure 7b), for which volcanic apron production rates data are provided by the literature: 230,000 km3 over the past 5.5 my for caldera-forming Hawaii and 10,000 km3 over the past 5.0 my for non-caldera-forming Samoa [Duncan and Clague, 1985; Lonsdale, 1975; Rees et al., 1993; Straub and Schmincke, 1998]. These values suggest apron production rates of 42 km3/ka for caldera-forming and 10 km3/ka for non-caldera-forming ocean islands (Table 3).
 Volcanic aprons not only consist of volcaniclastic rocks (e.g., ash, pumice, hyaloclastites) but also of hardrock formed from lava flows etc. [Rees et al., 1993; Schmincke and Sumita, 1998]. The volcaniclastic-to-hardrock ratio is estimated to be about 1:3 (e.g., Hawaii [Wolfe et al., 1994]). Applying this ratio to the apron production rates inferred above and by converting volume to mass using dense rock equivalents of 1680 kg/m3 for mafic tephra (2400 kg/m3 at 30% porosity), the volcaniclastic emission rates of hot spot-related ocean islands can be calculated to range from 5.7 Pg/ka to 23.7 Pg/ka (Table 3).
 Based on the offshore extend and the thickness of ash layers found in Deep Sea Drilling Project (DSDP) cores related to hot spot volcanism [Kelts and McKenzie, 1976; Viereck et al., 1985], the proportion of offshore ash compared to total volcaniclastics is estimated to range from 1% to 5%. The offshore ash deposition rates for individual HS oceanic islands are thus calculated to vary between 0.1 Pg/ka to 2.0 Pg/ka (Table 3). Taking into account whether a hot spot-related ocean island is caldera-forming or non-caldera-forming, the offshore ash deposition rate for Pacific oceanic islands is estimated to be between 2.0 Pg/ka and 9.0 Pg/ka (Table 3), which corresponds to <3% of the apron production rates of HS ocean islands. The HS offshore volcanic ash deposition rate is thus significantly lower than that of SZ volcanoes where in general more large-scale explosive eruptions occur (Table 2).
220.127.116.11. Overall Volcanic Ash Flux Into Pacific Ocean
 The Pacific millennial ash input from SZ and HS volcanoes through subaerial eruptions is estimated to range from 128 to 221 × 1015 g/ka (Table 4), more than 90% of which is derived from SZ volcanoes (Tables 2 and 3). Despite the uncertainties involved in such geological flux estimates (e.g., the CAVA ash flux, wind directions, arc lengths etc.) that may easily introduce a factor 2 error we argue that the ash flux estimate provided here serves well as a first-order approximation. For comparison, our millennial flux estimate for the Pacific Ocean is 15–20 times lower than the amount of material ejected from the large Pinatubo eruption (∼8.1 Pg [Wiesner et al., 1995]). Within a week, a single volcanic eruption can deposit similar amounts ash (e.g., Kasatochi 2008 eruption, at least 650 × 1012 g ash [Langmann et al., 2010]) to the yearly input of volcanic ash into Pacific Ocean (128–221 × 1012 g/yr based on short-term averaged millennial deposition rates). Therefore, the millennial-scale estimates in Tables 2 and 3 are realistic and rather conservative. The largest uncertainty of our estimate, however, is that it does not consider rare supereruptions that occur with a frequency of ≥10,000 years such as the Toba eruption in Sumatra ∼74 ka ago that produced about 2000 × 1015 g of ash, much of which was deposited into the ocean [Oppenheimer, 2002].
Table 4. Summary of the Fe Mobilization Behavior of Volcanic Ash and Mineral Dust in Different Experimental Setups With Variable pH and the Input of Soluble Fe Based on the Millennial Fluxes of Volcanic and Mineral Dust
| ||Volcanic Ash||Mineral Dust|
|nmol/g Ash||%FeSVA||nmol/g Dust||%FeSdust|
|Fe Release (nmol Fe/g or % FeS)|
|At seawater pH (8)||35–340a||0.003–0.2a||<10–125b,c||0.001–0.02b|
|In seawater without pH buffer||10–39,000d||0.001–1.8c,d||1,600–165,000c,e||0.26–26e|
|in acidic solutions (pH 1–5)||20–200,00f||0.001–22c,f||60–500,000c,g||0.01–80g|
|Millennial particle flux into the Pacific Ocean (1015 g/ka)||128–221h||39–519i|
|Soluble Fe Flux Into the Pacific Oceanj(109mol/ka)|
|At seawater pH (8)||3–75||1–65|
|In seawater without pH buffer||2–8,600||65–85,000|
|In acidic solutions (pH 1–5)||2–44,000||2–260,000|
4.3.2. Mineral Dust Input Into the Pacific Ocean: Millennial-Scale Flux
 About half of the global surface ocean dust flux is deposited into the northwest Pacific Ocean and stems from Asian deserts (Figure 7c) [Jickells et al., 2005; Jickells and Spokes, 2001]. Modeling of extrapolated aerosol dust concentrations (from island and coastal collection sites) suggests an annual dust deposition rate of 39–519 × 1012 g/y for the Pacific Ocean [Mahowald et al., 2005]. During the last glaciation the dust input was probably 2–20 times higher than during the more humid (and more vegetated) interglacial periods [Kohfeld and Harrison, 2001; Mahowald et al., 1999; Martin et al., 1990; Winckler et al., 2008]. The postglacial global pattern of dust deposition most likely did not change significantly [Jickells and Spokes, 2001]. Therefore, assuming a constant annual dust deposition rate in the Holocene, the annual dust deposition rate inferred from modeling [Mahowald et al., 2005] corresponds to a millennial dust flux rate of 39–519 × 1015 g/ka. Based on ocean sediment core data, however, the postglacial dust input is estimated to be 4–5 times lower than the modeling estimates [Rea, 1994], possibly due to a less anthropogenic input before the past 1000 years.
4.4. Significance of Volcanic Ash-Related Fe Input Into the Pacific Ocean: Biogeochemical Implications, Eruption Frequencies, and Spatial Distributions
 The millennial flux of airborne volcanic ash into the Pacific Ocean (128–221 × 1015 g/ka) is comparable to that of mineral dust (39–519 × 1015 g/ka [Mahowald et al., 2009]) or 50–115 × 1015 g/ka if corrected to sediment core observations [Rea, 1994] (see Table 4). From the similarity in both material flux and Fe solubilities in seawater at pH = 8 (Table 4) it follows that the flux of soluble Fe through dry deposition of volcanic ash into the Pacific Ocean is comparable to that of mineral dust (3–75 × 109 mol Fe/ka for volcanic ash and 1–65 × 109 mol Fe/ka for mineral dust). These estimates do not consider the effect of wet deposition (by rainwater) that, based on experimental results, would greatly enhance the Fe solubility of both volcanic ash and mineral dust as shown in Table 4 (see discussion in section 4.1) [Baker and Croot, 2010; Duggen et al., 2010; Mahowald et al., 1999]. The ratio of wet to dry deposition may, however, vary from region to region but on a global-scale dry deposition dominates. As outlined by Jickells and Spokes  for the Pacific Ocean, about 70% of the atmospheric particles are derived through dry deposition.
 Focusing on the dry deposition process, for mineral dust an overall range of 0.001%–0.02% FeS is derived based on the experiments (including this study) providing constraints for the solubility of Fe under seawater conditions (e.g., constant pH of 8) (Table 4). Previously, a value of 0.01% FeS was chosen for the dry deposition of Fe into the Saragossa Sea [Jickells, 1999], which is in accordance with the range outlined in Table 4. Results for Fe solubilities inferred from experiments at highly variable pH vary 3–4 orders of magnitude: 0.001%–22% for volcanic ash and 0.001%–80% for mineral dust (Table 4 and, for comparison with the new data, see the dashed lines in Figure 4) [Baker and Croot, 2010; Duggen et al., 2010; Frogner et al., 2001; Jones and Gislason, 2008; Mahowald et al., 2005; Schroth et al., 2009]. Table 4 emphasizes that major uncertainties in estimating Fe flux arise from the strong pH dependency of Fe solubility. Experimental studies demonstrate this with highest Fe solubilities under low pH with minimum Fe solubilities around pH 8 [Desboeufs et al., 1999; Guieu and Thomas, 1996; Spokes and Jickells, 1996]. Since results from experiments with acidic solutions are very likely to overestimate the dissolution of Fe in seawater [Baker et al., 2006], our Fe flux estimates for both volcanic ash and mineral dust were performed with data from experiments at seawater pH of 8.
 Volcanic ash and mineral dust particles can serve as cloud condensation nuclei (CCN) [Andreae and Rosenfeld, 2008; Duggen et al., 2010; Jickells and Spokes, 2001; Textor et al., 2006]. Therefore the particles that were transported long distances through the atmosphere may have been affected by interaction with low-pH cloud water prior to dry deposition (soil versus marine aerosol [Zhuang et al., 1992]). The volcanic ash samples used in this study were transported between a few and up to 130 km through the atmosphere (see Table S1), whereas the Cape Verdian loess sample was collected more than 1000 km away from its source. As long-distance atmospheric transportation may enhance Fe solubilities, Fe release data for ash particles sampled in the proximity of volcanic source craters may underestimate actual Fe solubilities in the remote ocean. It is still uncertain to what extend low-pH cloud water cycling of ash and dust particles affects the seawater solubility (hence bioavailability) of Fe. In terms of biogeochemical Fe and C cycles, however, a key parameter is the maximum concentration solubility of Fe that is controlled by various factors (e.g., nature of Fe-binding ligands) and Fe uptake mechanisms by marine photosynthetic organisms which are not completely understood [Baker and Croot, 2010].
 Due to differences in temporal and spatial deposition patterns, comparable millennial Fe fluxes into the Pacific Ocean are unlikely to have the similar marine biogeochemical impacts (Figure 7). The episodic deposition nature of mineral dust and volcanic ash (e.g., in the northern Pacific) may be less different than commonly thought. Despite its seasonality the dust input into the Pacific is not uniform throughout the year. About 30%–90% of the annual dust input is derived in the vicinity to the dust source at about 5% of the high deposition days [Mahowald et al., 2009]. Episodic dust storms occur about 20% of the days of a year, and may deposit about half of the yearly input within a 2 week period [Jickells and Spokes, 2001]. In terms of mass flux, the frequency and duration of the storm events is comparable to the moderate level of volcanic eruptions. Low-to-moderate eruptions (volcanic explosivity index of VEI <5 with ejecta volumes <1 km3 and ash plume heights of <10–25 km) are episodic but take place frequently. The relatively frequent volcanic eruptions can be grouped into (1) constant-to-daily gentle eruptions (eruption column height of hundreds of meters, e.g., Hawaiian volcanoes), (2) weekly eruptions (1–3 km ash plume height, e.g., Galeras), and (3) yearly explosive eruptions (3–5 km ash plume height, e.g., Cordón Caulle) (http://www.volcano.si.edu/).
 Each year at least 25 eruptions of VEI = 2 (<1011 grams ash), about 15 eruptions of VEI = 3 (<1014 grams ash), and about 1–4 eruptions of VEI = 4 (<1015 grams ash) occur along the Pacific Ring of Fire (http://www.volcano.si.edu/). Globally, high-magnitude eruptions with ejecta of >1015 grams (with ash plume heights of 10–25 km) occur every ≥10 years (e.g., Eyjafjallajökull 2010 eruption, VEI = 5). Supereruptions with ejecta of >1016 grams to >1019 grams (with ash plume heights of >25 km) are rare and take place every ≥10 to ≥10,000 years, depending on the magnitude (e.g., Pinatubo 1991, VEI = 6; Tambora 1815, VEI = 7; Taupo ∼26,500 years ago, VEI = 8). Fe fertilization that may arise from such large-scale eruptions to supereruptions is to date basically unknown but some studies indicate that the effect may have been significant in the Earth's history [Duggen et al., 2010, and references therein].
 In terms of biogeochemical response and C cycles, both sources can be expected to have their greatest impact on Fe-limited (HNLC) regions (Figure 7a) [De Baar and De Jong, 2001; Watson, 2001]. While mineral dust deposition is mainly restricted to the northwestern Pacific HNLC region (Figures 7a and 7c), Fe input by volcanic ash is more widespread including the north (subarctic), eastern equatorial and southwestern Pacific (Figure 7b). In the subarctic Pacific, for example, the ash fallout of the intermediate-scale August 2008 eruption of Kasatochi volcano (VEI = 4) was associated with a large-scale phytoplankton bloom in the Fe-limited North Pacific [Hamme et al., 2010; Langmann et al., 2010]. In the Fe-limited eastern equatorial Pacific mineral dust input is relatively rare so that the flux of soluble Fe from ash from Mexican, Central American, and South American volcanoes is likely to dominate the millennial atmospheric input of Fe in these regions (Figure 7). In the Pacific sector of the Southern Ocean mineral dust deposition is very limited too, whereas volcanic ash from active volcanoes in the subduction zones of New Zealand and Tonga-Kermadec can be transported into the Fe-limited Southern Ocean with the westerlies, potentially causing volcanic Fe fertilization (Figure 7b).