Atmospheric Inputs as a Source of New Nutrients
How important is the atmosphere as a source of nutrients? The answer to this question depends on the large-scale variation in annual inputs, the magnitude of that input relative to other sources of nutrients, and the short-term temporal and spatial variability in inputs. Large-Scale Distribution of Atmospheric Inputs Global estimates of the annual input for at- mospherically derived mineral matter to the ocean in 10 ° X 10 ° (latitude X longitude) areas have been made from measurements of mineral dust in air and rain and from consideration of the wet and dry removal mechanisms for the dust (Duce et al., 1991). Numerous measurements from is- land sites and ships now enable estimation of the geographic and temporal variability of the at- mospheric concentrations and fluxes of mineral matter to the ocean. The atmospheric input of iron to the world ocean (Fig. 1) is derived from these mineral dust measurements. Atmospheric iron input varies spatially by over four orders of magnitude (Fig. 1). The highest inputs occur in the western North Pacific, the Indian, and the Equatorial Atlantic Oceans, with lesser deposition in the regions around Australia and North and South America. The lowest inputs of atmospheric iron occur in the south central Pacific and South- ern Ocean. The importance of the above spatial differences in atmospheric inputs is dependent on the degree that nutrients limit biological production and the magnitude of alternate sources. In general, bio- logical production in surface waters of the ocean is limited by light and the supply of nutrients, including nitrogen and phosphorus (Dugdale,1967; Eppley, 1981), and possibly metals such as iron (Martin and Fitzwater, 1988; Sunda et al., 1991). For oceanic systems at considerable dis- tance from land or shallow benthic regeneration sites, there are three possible sources of these crit- ical elements (in addition to nitrogen fixation). First, regeneration by zooplankton and bacteria supplies a significant fraction of the nutrients re- quired to maintain primary production in the eu- photic zone. Some of these nutrients, however, are removed from the euphotic zone with the gravitational flux of biogenic particles. In order to sustain primary production there must be a supply of”new” nutrients to replace those lost by the particle flux (Dugdale and Goering, 1967). Second, vertical mixing and upwelling processes in the pycnocline can provide an upward flux of new nutrients to the euphotic zone. Finally, at- mospheric deposition provides another source of new nutrients (Menzel and Spaeth, 1962; Duce, 1986). The role of atmospheric inputs as an annual source of new nitrogen, phosphorus, and iron to the euphotic zone has been recently evaluated for the Sargasso Sea and North Pacific Gyre (Duce, 1986). Sources of new nutrients consid- ered were deep waters transported up into the euphotic zone by diffusive and advective pro- cesses, atmospheric inputs, and, in the case of nitrogen, in situ fixation by marine organisms. The results for iron and nitrogen, modified by more recent data on dissolved iron concentra- tions in surface seawater, are shown in Table I. Despite the uncertainty in these calculations (see Table 1 legend), the atmosphere is clearly the dominant source of new iron in the North Pacific Gyre and Sargasso Sea. The atmosphere is a po- tentially significant, but not a dominant, source of new nitrogen on yearly time scales in these areas. In contrast, the atmosphere is not a sig- nificant source of phosphorus for either region (Duce, 1986). These results indicate that annual variation in atmospheric inputs should have dra- matically different impacts on production pro- cesses in these regions depending on whether iron, nitrogen, or phosphorus is most limiting. The average percentages given in Table 1 may underestimate the importance of atmospheric supply of nitrogen on smaller temporal and spatial scales. The calculations used to obtain the esti- mates in Table 1 assume that the euphotic zone
Fig. 2: Variability of atmospheric iron in dust at Midway Island in the North Pacific gyre. Iron concentrations were estimated by multiplying measured aluminum concentrations by the iron to aluminum ratio found in dust. The aluminum data are from Uematsu et al. (1985) and Arimoto (personal com- munication). These data were usually collected at weekly intervals, thus al- lowing calculation of weekly average aluminum concentrations per cubic meter of air. During May-June 1986, data were collected each day, thus allowing an estimate of variability on a daily time scale (insert). Weekly sampling was nearly continuous from 1981 to the spring of 1984; after that,data were collected during periods of maximum dust levels.
is continuously well mixed, and thus all inputs are equally available to all parts of the system. In reality, the euphotic zone in most subtropical oceanic gyres (where nitrogen is usually limiting) is composed of a thin surface layer that mixes on diel frequencies and a deeper layer that is mixed only episodically (Jenkins and Goldman, 1985). In such regions much of the nutrient input from below the pycnocline will be consumed in the deep euphotic layer (Jenkins and Goldman, 1985) and most of the atmospheric input will be consumed in the thin surface layer. This will make the sur- face-layer euphotic zone far more dependent on atmospheric inputs. Even if the low estimates in Table 1 are the most accurate (as argued by Knap et al., 1986), atmospheric inputs may still be a significant or even dominant source of new nitro- gen on smaller temporal and spatial scales.
Variability in Atmospheric Input Events
The magnitude and frequency of atmospheric input events depend on the temporal and spatial characteristics of the processes controlling dust injection into the atmosphere, long-range dust transport over the ocean, and deposition of dust into the ocean. The primary source of atmospheric iron is mineral dust generated in arid desert, semi- desert, and loess regions of central and eastern Asia, Africa, Australia, India, and the Arabian peninsula. For example, in eastern Asia local weather events, such as frontal passages or strong thunderstorms, generate large dust clouds that are carried into the upper levels of the troposphere where the dust is transported out over the North Pacific by strong westerly winds. The dust settles gradually into the lower atmosphere as it moves over the ocean, where some 80% is removed by rain (Uematsu et al., 1985) and the remainder by dry deposition. Most dust storms in Asia occur in the spring just after the winter snows have melted and when little or no vegetation is present. Fallout from such spring dust events can be observed at island sites in the central Pacific 5-10 days after an event in China. The spring maximum in the dust transport can be seen clearly over a several-year period at Midway Island in the central North Pacific (Fig. 2). Over much of the North Pacific most of the mineral dust (and thus iron) input to the ocean typically takes place during 3-5 events in the spring, each of which lasts 1-3 days. Actual inputs to any given location in the ocean may be even more variable than predicted by the large-scale gradients (Fig. l) or atmospheric dust levels (Fig. 2). This is because wet deposition requires the co- occurrence of rainfall and of dust or fixed nitrogen in the atmosphere. Because rainfall events vary from small intense storms of a few square kilo- meters to large frontal systems that stretch for thousands of kilometers, inputs vary on the same spatial scales. Thus, atmospheric inputs are both episodic and spatially patchy. Phytoplankton responses to input events de- pend not only on the size of the event but also on the pre-event nutrient concentration, the extent of vertical mixing during the input event, the chemical fate of the added nutrients, and any in- duced changes in bioavailability of other limiting or toxic elements. In an effort to evaluate the im- portance of event size and of physical mixing, we have used data sets from the Pacific and Atlantic to calculate expected changes in sea-water con- centrations of iron and nitrogen following several types of atmospheric events. In the upper half of Table 2, we have used the time-series data in Fig. 2 to estimate the change in the iron concentration in surface sea water near Midway Island after low- dust, high-dust, and average storm input events during periods of shallow and deep physical mix- ing. In the lower half of Table 2, we have used the direct daily measurements of iron and nitrogen deposition to calculate expected concentrations of each of these elements in the Sargasso Sea near Bermuda after a similar range of events and mix- ing conditions. Both sets of calculations indicate that detectable changes will be dominated by a few relatively infrequent large input events that occur during periods of weak to moderate mixing. This conclusion is strengthened by the remarkable similarity of the range for estimated increases in surface nitrogen concentrations at Bermuda (Ta- ble 2) as compared with increases measured by Menzel and Spaeth (1962) in the same area after storms. These large atmospheric input events can cause several order-of-magnitude increases in both iron and nitrogen concentrations. As a result, in situ biological responses to atmospheric inputs may reflect changes in the bioavailability of iron and/or nitrogen.
Chemical and Biological Fate of Atmospheric Inputs
What is the fate of the nutrients in the atmo- pheric material once it enters the ocean? The an-wer to this question is simple for nitrogen because most of the input is in the form of nitrate or am- monium that can be readily used by phytoplank- on. This question is more difficult to answer for ron present in mineral particles because only truly issolved forms of iron appear to be taken up and sed by phytoplankton (Wells et al., 1983; Rich nd Morel, 1990; Morel et al., 1991, this issue). t is also made difficult by critical gaps in our un- erstanding of the underlying chemistry. erosol Dissolution The surface-layer retention and potential vailability of iron from atmospheric particles is artly controlled by the rate of aerosol dissolution n sea water. The magnitude of this dissolution as been estimated conservatively at 10% for mineral iron (Moore et al., 1984; Duce, 1986), ut as much as 40-60% of the atmospheric iron can apparently dissolve in a few minutes when the total iron concentration in sea water is very low (i.e., less than a few nmol kg -~) (Zhuang et al., 1990). These higher estimates were obtained by transferring samples of mineral aerosols col- lected over the mid-North Pacific Ocean into stirred aliquots of aged filtered open-ocean sea- water, then determining changes in the concen- tration of iron in the filtrate that passed through 0.45-um pore size filters. Some caution must be exercised in assuming that all this iron is bio- available, because the iron present in the filtrate may include dissolved iron plus submicrometer colloidal iron oxides. This rapid conversion to dissolved and colloidal iron is extremely impor- tant, because it indicates that a major portion of the aerosol iron may be retained in the euphotic zone rather than settling into deeper waters• Zhuang et al. (1990) hypothesized that the rapid dissolution of a significant fraction of the aerosol iron depends on chemical alteration of the mineral particles during atmospheric transport from their source regions. Andreae (1986) found that individual aerosol particles over the ocean are often a mixture of mineral, sea salt, and acidic sulfate aerosols. It is likely that these mixed aero- sols form as a result of coalescence within clouds, with subsequent evaporation of the cloud droplets• The result is acidic hygroscopic aerosol particles in which partial dissolution occurs during atmo- spheric transport (Winchester and Wang, 1990). Significant concentrations of Fe(II), formed by photoreduction of Fe(III), have been observed in atmospheric samples in urban areas (Behra and Sigg, 1990) and over the remote North Pacific (Zhuang, 1991). Thus, aerosols can be rich in la- bile forms of Fe(III) and/or Fe(II) when the aerosol particles enter the ocean (Duce and Tindale, 1991).
Fate of Atmospheric Iron in Surface Waters
Once the atmospheric iron enters surface sea- water, the chemical form of the dissolved iron may be altered, thus changing its solubility, retention in the euphotic zone, and bioavailability. The chemical speciation of iron in seawater is an im- portant unresolved question in marine chemistry. Some of the most important forms and processes are summarized in the conceptual model shown in Fig. 3. Iron can exist in seawater in two oxi- dation states [Fe(II) and Fe(III)], as various dis- solved chemical species [e.g., Fe +3, Fe(OH) +2, Fe(OH)~-, Fe(OH)~, Fe +2, FeOH +, Fe(OH~, FeCI ÷, FeSO~, and FeCO~, in multiple colloidal forms with different reactivities, and in a variety of particulate forms ranging from inorganic iron colloids to iron in or adsorbed onto bacteria, plankton, detritus, and mineral particles. In ox- ygenated seawater at pH near 8, the thermody- namically stable oxidation state of iron is Fe(III) (Kester, 1986). Aerosol Fe(II) inputs should be oxidized rapidly (within a few minutes) to Fe(III), a form that has extremely low solubility in sea- water (Ksp = l0 -3s) and precipitates as amor- phous colloidal hydrous oxides (Byrne and Kester, 1976). Hydrous iron oxides have surface-active properties that allow the adsorption of organic substances and other trace-metal ions from so- lution. Hydrous iron oxides may coagulate or they may be adsorbed or incorporated into larger par- ticles, resulting in size distributions that range from colloidal particles that will pass through a 0.4-tzm filter up to sizes that can settle from the system. These chemical processes tend to limit the concentrations of dissolved iron in oceanic waters to <1 nmol kg -t. Although bioavailable iron may represent a small fraction of the total iron in seawater, a va- riety of microbial and chemical processes may re- sult in exchange between bioavailable and non- bioavailable forms (Fig. 3). Some possible mech- anisms for increased dissolved iron availability in seawater include the following: 1) photochemical production of dissolved Fe(II) (Hong and Kester, 1986; O’Sullivan et aL, 1991); 2) the reductive photodissolution of colloidal iron oxides followed by immediate oxidation and reprecipitation of more soluble, faster dissolving, amorphous iron oxides on the colloidal surface (Waite and Morel, 1984; Morel et aL, 1991, this issue); 3) thermal dissolution of colloidal iron oxides (Wells et aL, 1983); 4) microbial reduction (Jones et aL, 1982) or remineralization of iron oxides; and 5) com- plexation by dissolved organic ligands (Kester, 1986). Atmospheric inputs of iron can be expected to interact with the chemical pathways discussed above (solubilization, redox cycling, colloid for- mation) in several ways that depend significantly on the size of the input and its initial dilution by physical processes. From a chemical perspective, small inputs under all mixing conditions and in- termediate inputs during strong mixing periods will produce in situ changes in dissolved iron that are analytically difficult to detect (Table 2). In contrast, large inputs during physically quiescent periods will not only be detectable but may result in precipitation of colloidal Fe(III). Such a process would put an upper limit on the potential in- creases in bioavailable iron and may coprecipitate other trace metals from the system. Thus, the bio- availability of other trace metals may depend not only on their own input rates and redox chemistry but also on that of iron. Another important effect of these chemical processes will be to increase the duration but reduce the size of the immediate in- crease in bioavailable iron. The magnitude of these buffering effects is poorly known but almost cer- tainly important. The chemical and biological processes con- trolling iron remineralization may influence strongly bioavailability and the temporal scales of biological responses. The concentration of bio- available iron is also influenced by the rate and extent that iron taken up by organisms is recycled when those organisms are eaten or die. Such re- cycling has long been recognized to play a dom- inant role in controlling primary production in nitrogen limited oligotrophic gyres of the ocean (Eppley et al., 1976). As a result of both iron redox chemistry and recycling, it seems likely that a sig- nificant fraction of the aerosol iron entering the dissolved pool during an input event will be re- partitioned into nonbioavailable forms that are slowly rereleased after the event. This release will tend to lengthen the time scale of any biological response and to provide a source of iron to support primary-production processes during periods be- tween events. Fig. 3: Potential reservoirs, pathways, and processes controlling iron in the surface ocean.
Ecosystem Responses to Episodic Nutrient Inputs
What will be the effect of an input event on phytoplankton and zooplankton productivity, biomass, and composition? It is tempting to sug- gest that the answer would be an increase in plankton production and biomass (assuming the event increased the bioavailability of a limiting nutrient). However, there is growing evidence that the biomass response also depends on the impor- tance of grazing and predation by higher trophic levels, the temporal and spatial extent of the input event, and the magnitude of physical mixing and shearing processes during and after the event. Each of these factors will be considered in sequence below.
Regulation by Resources and Higher Trophic Levels
The three major types of phytoplankton re- sponse are summarized in Fig. 4. If iron or nitro- gen were not limiting the growth of any phyto- plankton species in a given system, no changes should be observed in primary production, bio- mass, species composition, or sedimentation rates (left column, Fig. 4). This, of course, assumes that the input event is not associated with changes in other factors that control plankton dynamics. However, if iron or nitrogen does limit phyto- plankton growth rates, two very different re- sponses might occur, depending on the impor- tance of grazing processes in controlling phyto- plankton biomass and species composition. If grazing by micro- and macrozooplankton is not important, then addition of the limiting nutrient should increase primary production and phyto- plankton biomass (center column, Fig. 4). In- creased vertical sedimentation fluxes to deep wa- ters might occur as the added nutrient is depleted, but the magnitude and timing of such changes would depend on the species of phytoplankton that bloom. Recent experiments indicate differ- ences in the degree of nutrient limitation between phytoplankton species may be large enough to in- duce shifts in composition (Martin and Gordon, Fig. 4: Control diagram of potential phytoplankton responses to an atmo- spheric-nutrient input event to the surface ocean. These diagrams contain three elements: (1) an initial perturbation at the top of the diagram (atmo- spheric iron or nitrogen input for phytoplankton or food resource changes for zooplankton); 2) a series of sequential “if statements'” to define critical al- ternatives (contained in ellipses in the figures); and 3) primary organism or system responses. These primary responses are divided further into those that can be measured directly in the field (rectangles with rounded corners) and those that are di~cult or impossible to measure accurately in the ocean (rectangles with square corners). We assume in this figure that bioavailability of ambient trace metals (other than iron) is such that atmospheric inputs will not elicit a toxic biological response, lf phytoplankton growth is suppressed by ambient toxic trace metals, then removal of these metals by co-precipitation with iron may lead to biological responses similar to those of an iron-limited population. *The term “‘non-bioavailable” is particularly relevant to iron (Fig. 3), but may also apply to low levels of organic nitrogen found in at- mospheric inputs (Knap et aL, 1986). Fig. 5. Control diagram of potential zooplankton responses to changes in food resources caused by an atmospheric-nutrient input event to the surface ocean in a region where either iron or nitrogen are limiting. Control diagram elements are as defined in Figure 4 caption. The term zooplankton refers to both micro- and macrozooplankton although specific pathways may vary in importance for these two groups. 1988; Morel et al., 1991, this issue). However, if grazing processes are sufficient to balance exactly increases in the growth rates of all phytoplankton species (as argued by Miller et al., 1991 this issue), no changes would be observed in phytoplankton abundance or composition, despite increases in growth rates and physiological condition of in- dividual species (right column, Fig. 4). Toggweiler (1990) has used a similar biological model to sug- gest that a combination of grazing and high nu- trient inputs from upwelling may explain the ex- cess nutrients in the equatorial Pacific. Although a strong case can be made for such a grazing- dominated response in the subarctic Pacific (Miller et al., 199 I, this issue), the phytoplankton blooms that frequently follow nutrient input events in temperate and subtropical waters indi- cate that there are many cases where increases in grazing losses are insufficient to balance growth increases for all species (Donaghay, 1988). As a result, grazer control and iron limitation must be viewed as viable alternative explanations for excess nutrients in the tropical Pacific until experiments are conducted to test both hypotheses rigorously. The magnitude of the zooplankton response will depend on the extent and nature of the changes in phytoplankton, the degree to which zooplankton are food limited, the rate at which zooplankton can respond to changing resources, and the degree to which zooplankton biomass and species composition are regulated by predation from higher trophic levels (Fig. 5). If the zoo- plankton are not food limited, then zooplankton feeding, physiological condition, growth rates, and reproduction should remain unchanged and should result in unaltered species composition, age-class structure, and biomass (left column, Fig. 5). However, if zooplankton are limited by the quantity, quality, or spatial distribution of food, then changes in these parameters induced by at- mospheric inputs should alter feeding, improve physiological condition, and increase reproduc- tion and/or growth rates. The extent to which these changes are reflected by changes in species composition, age-class structure, and biomass will depend on the temporal scale of the phytoplank- ton response, the rate zooplankton can respond to changes in food (Donaghay, 1988), and the im- portance of predation in controlling these vari- ables (center versus right column, Fig. 5). As with phytoplankton, higher trophic level effects may totally or partially mask the effects of improved food.
Patch Induction and Biological-Physical Interactions
Spatial aggregations or patches of phytoplank- ton and zooplankton have long been recognized as a major characteristic of oceanic ecosystems (Hardy and Gunther, 1935; Haury et aL, 1978). One explanation for these patches is that they re- flect a response ofphytoplankton growth rate and biomass to spatial and temporal heterogeneity of new nutrient inputs from below the pycnocline into surface waters. For example, phytoplankton patches can be interpreted to represent the product of recent episodic inputs, whereas zooplankton patches can be interpreted as older events that have stimulated phytoplankton, and then in situ zooplankton productivity. Episodic atmospheric inputs provide an important alternative source of new nutrients to stimulate such patchiness. The temporal and spatial scale of the input event can be expected to influence strongly both the magnitude and mechanisms of zooplankton patch induction. Increases in zooplankton bio- mass from in situ growth should occur first in the rapidly growing microzooplankton, then in the slower growing macrozooplankton. Zooplankton patches also can be formed by immigration (Fig. 5). Interactions of vertical migration and vertical current shear in the ocean have been hypothesised as an important mechanism for the formation and dispersal of zooplankton patches (Hardy and Gunther, 1935; Haury et al., 1978; Steele, 1978). Immigration could be an important process if vertically migrating macrozooplankton respond to a phytoplankton patch by not migrating down to levels where currents can carry them away from the patch. Such immigration should be particu- larly important for increases of relatively slow growing macrozooplankton in submesoscale in- puts. Blooms in such patches should be too brief for multiple generations of macrozooplankton, et small enough to allow aggregation of zooplankton from larger areas. The temporal and spatial scale of input-in- duced patches can depend on the physical regime during and after the input event. Vigorous mixing of the surface layer during the input event will increase initial dispersal of the iron and nitrogen both horizontally and vertically. This increased dispersal will reduce the magnitude of any bio- logical response but increase its aerial extent. The intensity of surface-layer mixing and current shear after an input event will control the rate at which the resulting chemical and biological patch re- mains intact and spreads horizontally. As long as mixing is isotropic, the patch will remain coherent and spread horizontally at a decreasing rate with increasing size as defined by the ocean diffusion diagrams of Okubo (1971). As a result, larger patches will tend to persist longer, thus allowing more time for biotic responses to lead to detectable biomass differences (Haury et al.. 1978). However, if mixing is anisotropic (i.e., associated with strong vertical or horizontal shear), even large patches will tend to fragment into filaments and smaller patches (Okubo, 1971; Garrett, 1983). Work with patches of conservative tracers has shown that al- though coherent patches can persist for > 10 days in isotropically mixed systems (Watson et al., 1991), shear-induced filamenting in anisotropic systems can make patch tracking and in situ de- tection of chemical and biological effects ex- tremely difficult (Garrett, 1983; Watson et al.. 1987). These results indicate that persistence of the nutrient patch and therefore, the temporal and spatial extent of biotic responses will depend on the match of the time scales for biological and chemical responses to the time scales of physical processes controlling the coherence of the patch.
Atmospheric inputs of iron and nitrogen occur on a diversity of flux, time, and space scales. This diversity of scales leads to a diversity of potential chemical and biological responses to individual input events. Our preliminary analysis suggests that this range of responses extends from unde- tectable changes in chemical concentrations (and presumably biological responses) to relatively rare large events with easily detectable chemical sig- natures. The temporal persistence of such signa- tures depends on the chemical and biological fate of the nutrients from the atmosphere and on the magnitude of physical dispersion. Similarly, the in situ biological response depends not only on the magnitude of the input event and the degree of limitation of primary production by the nu- trient but also on the degree to which higher trophic levels limit the expression of changes in physiological condition as changes in biomass and composition. Regrettably, our understanding of these control mechanisms is extremely limited in many of the oceanic systems potentially affected by atmospheric inputs (for an exception, see Miller et al., 1991, this issue). Hence, future interdisci- plinary studies designed to test hypothesized ef- fects of atmospheric inputs in these systems rep- resent both a challenge and an opportunity to ex- pand our understanding of the interactions of physical, chemical, and biological processes con- trolling open-ocean ecosystems. An important component of such future stud- ies will be efforts to examine directly episodic at- mospheric-nutrient input events, either by fol- lowing small (1-20 km) natural patches created by atmospheric input events or by creating such patches by artificial addition of the appropriate nutrients. A critical challenge of both approaches will be to identify control areas and separate the patchiness induced by the input event from the inherently patchy character of the oceanic envi- ronment. One promising approach to this prob- lem is to use natural (Donaghay et al., 1987) or artificial (Watson et aL, 1991 ) conservative tracers of the input event to follow the patch in time and space. Such tracer techniques also are essential to separating in situ biological and chemical re- sponses from the effects of physical mixing and shearing. The success of such efforts will depend heavily on selecting sites where the combination of strong nutrient limitation and reduced physical dispersion/sheafing maximize the magnitude and persistence of chemical and biological responses to the input event. The success of such experi- ments also will depend critically on measuring chemical and biological factors and processes (see Figs. 3-5) that provide unambiguous tests of the