Aerosols in atmospheric chemistry and biogeochemical cycles of nutrients

Atmospheric aerosols have complex and variable compositions and properties. While scientific interest is centered on the health and climatic effects of atmospheric aerosols, insufficient attention is given to their involvement in multiphase chemistry that alters their contribution as carriers of nutrients in ecosystems. However, there is experimental proof that the nutrient equilibria of both land and marine ecosystems have been disturbed during the Anthropocene period. This review study first summarizes our current understanding of aerosol chemical processing in the atmosphere as relevant to biogeochemical cycles. Then it binds together results of recent modeling studies based on laboratory and field experiments, focusing on the organic and dust components of aerosols that account for multiphase chemistry, aerosol ageing in the atmosphere, nutrient (N, P, Fe) emissions, atmospheric transport, transformation and deposition. The human-driven contribution to atmospheric deposition of these nutrients, derived by global simulations using past and future anthropogenic emissions of pollutants, is put into perspective with regard to potential changes in nutrient limitations and biodiversity. Atmospheric deposition of nutrients has been suggested to result in human-induced ecosystem limitations with regard to specific nutrients. Such modifications favor the development of certain species against others and affect the overall functioning of ecosystems. Organic forms of nutrients are found to contribute to the atmospheric deposition of the nutrients N, P and Fe by 20%–40%, 35%–45% and 7%–18%, respectively. These have the potential to be key components of the biogeochemical cycles since there is initial proof of their bioavailability to ecosystems. Bioaerosols have been found to make a significant contribution to atmospheric sources of N and P, indicating potentially significant interactions between terrestrial and marine ecosystems. These results deserve further experimental and modeling studies to reduce uncertainties and understand the feedbacks induced by atmospheric deposition of nutrients to ecosystems.


Background
Atmospheric aerosols are liquid or solid particles suspended in the atmosphere; they range from about 3 nm to 100 m in diameter and they have different shapes, chemical compositions and hygroscopic and optical properties Pandis 2006, Pöschl andShiraiwa 2015). The properties of aerosols depend on their source, transport and transformation in the atmosphere, and they are modified by the so-called atmospheric ageing that includes chemical and physical processing in the atmosphere (Kanakidou et al 2005, Boucher 2015. Aerosols affect human health by inhalation and subsequent production of reactive oxidative stress Shiraiwa 2015, Lelieveld et al 2015), and long-term exposure can lead to lung cancer and cardiopulmonary disease (Pope III et al 2002, Pope III andDockery 2013). It is currently estimated that each year around 4.5 million premature deaths occur due to air pollution, with the most affected countries being China, India and Russia; most of these effects are linked to atmospheric aerosols (Lelieveld and Pöschl 2017). Aerosols also affect climate directly since they interact with radiation by scattering or absorbing light (Boucher 2015, Carslaw et al 2010, although quantification of the impact of absorbing aerosols such as black and brown carbon remains challenging (Lack et al 2012, Zhang et al 2017. The indirect effect on climate is associated with so-called aerosol-cloud interactions that alter the atmospheric water cycle (Ramanathan et al 2001, Stevens and Feingold 2009, Christensen et al 2016. Less attention has been given to the effect of aerosols on climate associated with the modification of ecosystem functioning (e.g. Arneth et al 2010), especially via atmospheric deposition of nutrients carried by aerosols (e.g. Mahowald et al 2008, Aumont and Bopp 2006, Aumont et al 2008 and the carbon sequestration sink (e.g Guieu et al 2014a, 2014b, Pitta et al 2017, Christodoulaki et al 2016, Duce et al 2008. Atmospheric O 2 and biomass production, both required to sustain life on Earth, are controlled by the photosynthetic activity of terrestrial and marine ecosystems (Pepper et al 2015), which in turn is limited (Moore et al 2013) by the availability of trace elements needed for synthesis of amino acids, nucleic acids and proteins, the building blocks of life. Aerosols carry such trace elements and provide them to ecosystems via atmospheric deposition (Mahowald et al 2017). During atmospheric transport, the bioavailability of this atmospheric pool of trace elements in aerosols increases (Baker et al 2006b) and the amounts deposited affect sequestration of atmospheric carbon (Jickells et al 2017) into terrestrial and marine environments, and the levels of O 2 in subsurface oceans (Ito et al 2016)-with profound implications for climate, fisheries, marine life and biodiversity .
In the atmosphere, aerosols, aerosol-associated water and cloud water also serve as media for multiphase chemistry, the main difference among them being the percentage water content and thus the level of dissolution of various compounds in water, which strongly affects pH levels. Cloud water chemistry has attracted attention for more than two decades (e.g. Erel et al 1993, Herrmann et al 2000, Warneck 2005 and references therein), in particular with regard to the oxidation of organics in the atmosphere. Recently, multiphase chemistry (in aerosol, cloud and fog waters) has been recognized as an important driver of atmospheric oxidant levels, with redox reactions and organic chemistry being involved (e.g. Lim et al 2010, Myriokefalitakis et al 2011, Mao et al 2013, and it affects the composition and properties of aerosols (e.g. Rudich 2003, Carlton et al 2007, Ervens and Volkamer 2010, Carlton and Turpin 2013, Ervens 2015, Myriokefalitakis et al 2015, 2016. Multiphase chemistry also contributes to the fixation of nutrients like N onto aerosols, and to the conversion of insoluble forms of trace elements like Fe and P into soluble, and thus readily bioavailable, material for uptake by ecosystems. These changes affect the nutrients carried by aerosols to ecosystems and remain understudied. The present review aims to fill this gap by integrating knowledge about the impact of multiphase chemistry on atmospheric water acidity, composition and the solubility of nutrients that determines their bioavailability, focusing on the biogeochemical cycles of N, P and Fe, including their organic fractions and impact on marine ecosystems (figure 1). This study builds upon earlier review papers on organic aerosol formation, properties and impacts on the climate system (Kanakidou et al 2005, Hallquist et al 2009, on fog and cloud chemistry, hereafter both referred to by the term cloud chemistry (Ervens 2015), and aerosol water (McNeill 2015), on multiphase chemistry at the biosphere-atmosphere interface (Pöschl and Shiraiwa 2015), and on aerosol deposition and its impact on ecosystems (Mahowald et al 2017). It integrates recent global modeling studies of the atmospheric deposition of nutrients (N, P, Fe) that, based on laboratory and field experiments, account for the changes in biogeochemical cycles induced by chemical formation and processing of aerosols in the atmosphere. It also summarizes the findings on the impact of atmospheric deposition of these nutrients on ecosystems. Focus is put on the organic fraction of nutrients and the multiphase chemistry-driven solubilization of trace elements, like Fe and P, to their readily bioavailable form, on dust aerosols. Simulations using past and future anthropogenic emissions of pollutants show the importance of human-induced emissions on nutrient atmospheric deposition and biogeochemical cycles. Uncertainties and implications are also presented.

Nutrients in the atmosphere
Natural and anthropogenic aerosol sources are very diverse (figure 1). Natural sources include emissions from soils, particularly in arid areas, the sea surface, volcanoes, terrestrial vegetation and wild fires; anthropogenic sources are associated with various combustion processes, including energy production, the residential and transport sectors, solvent use, agriculture and industrial activities (Lamarque et al 2010, Carslaw et al 2010, Kanakidou et al 2012, Boucher 2015, Fröhlich-Nowoisky et al 2016.
Humans have significantly modified the composition of atmospheric aerosols due to direct emissions of various aerosol components (carbonaceous aerosols that include black, brown and organic carbon) or their precursor molecules, like sulfur dioxide, nitrogen oxides and volatile organic compounds (e.g. Galloway et al 2004, Lamarque et al 2010, Kanakidou et al 2005, Fuzzi et al 2006, Goldstein and Galbally 2007, Hallquist et al 2009. However, both natural and anthropogenic sources also emit into the atmosphere various other trace Figure 1. Terrestrial vegetation and marine phytoplankton are responsible for producing the O 2 in the atmosphere of our planet and the majority of the removal of CO 2 from the atmosphere. Photosynthesis requires nutrients that can be supplied to ecosystems by atmospheric deposition of trace elements, carried by aerosols, which can be considered as a major source of nutrients for ecosystems. Multiphase chemistry in the atmosphere converts insoluble trace elements to soluble (bioavailable) nutrients. This chemical process has been neglected so far in most global modeling of biogeochemical cycles. The feedback of aerosols on climate via atmospheric deposition and the impact of the biogeochemical cycling of nutrients have been neglected in climate modeling until now. elements attached to the aerosol phase, such as P (Anderson et al 2010, Paytan and McLaughlin 2007) and metals, including Fe, Cu, Cd, Co, Ni, Hg, Zn, Pb and As, among many others (Paytan andMcLaughlin 2007, Rauch andPacyna 2009). Metals can also be present in the aerosol phase as free ions or complexes in inorganic or organic forms. N, P and many metals (e.g. Fe, Cu, Ni, Co, Cd, Zn) are essential micronutrients for marine phytoplankton, but at high levels they can be toxic

Organics as carriers of nutrients
Organics are key players in biosphere-atmosphereclimate interactions (e.g. Kulmala et al 2004) and their role as carriers of nutrients has only recently been recognized (Anderson et al 2010, Cornell 2011, Kanakidou et al 2012. They have a significant human-driven component due to primary anthropogenic emissions (Lamarque et al 2010), (Hoesly et al 2018) or the interaction of natural emissions with pollution (Kanakidou et al 2000, 2012Spracklen et al 2011. This organic pool in the atmosphere has a variable content of C, H, O, N and other elements depending on the origin and the history of organics in the atmosphere, especially between anthropogenic (fossil) and biogenic sources. In the early 1950s van Krevelen introduced a method for studying the H:C content of organic compounds as a function of their O:C content (van Krevelen 1950) in order to track the origin and maturity of kerogen and petroleum. With the increasing power of analytical techniques (Aiken et al 2008) these plots, now known as van Krevelen diagrams, started to be used to understand the origin and properties of organic aerosols in the atmosphere in order to simplify the representation of a very complex pool of compounds (Heald et al 2010). Recently, van Krevelen diagrams have been also extended to include elements such as N (Altieri et al 2012). Element ratios have also been used to understand the organic N and organic P atmospheric cycles based on the relatively better understood organic C cycle (Kanakidou et al 2012, 2016.
During the last decade attention has been paid to biological material that is widely found as part of the organic aerosols directly emitted by vegetation (Jaenicke 2005), as well as to the amounts of organics present in soils and potentially carried by soil dust aerosols (e.g. Schlesinger 1977, Kanakidou et al 2012 andreferences therein), further discussed in section 3.1. Remarkably, organic aerosols are found to be more abundant at several locations in the atmosphere (Jimenez et al 2009) compared with the sulfate aerosols (Tsigaridis and Kanakidou 2007) that have been mitigated the past decades (Lamarque et al 2010, Hoesly et al 2018. Following this reduction in sulfate, sulfate-associated organics in the aerosol phase have decreased, as well as aerosol hydroscopicity and thus the aerosol-associated water over polluted regions in the USA (Attwood et al 2014, Boulanger 2015, with potential implications for multiphase chemistry and aerosol acidity.

Primary aerosol sources
Aerosols are present in the atmosphere in different sizes, ranging from submicron to supermicron aerosols, with diameters smaller and larger than 1 m, respectively. Until recently, most of the modeling studies of aerosols have considered the PM 10 aerosol fraction, i.e. those with diameters smaller than 10 m (Textor et al 2006), and have neglected the larger supercoarse mode. Therefore, in the following discussion we neglect the supercoarse aerosols, except when specifically mentioned.

Dust
Dust and sea spray are the largest components of atmospheric aerosols, with mass emissions to the atmosphere two to three orders of magnitude higher (of one to several thousand Tg yr −1 ) than other aerosol components (Boucher 2015). Large uncertainties exist in their fluxes, some of which is due to the contribution of the different size distributions of aerosols taken into consideration in individual studies. This is also reflected in model estimates of dust fluxes, where differences of almost an order of magnitude in dust emissions are calculated between models with a wide variety of assumptions on dust size distribution (Textor et al 2006). Indeed, dust particles have diameters spanning from 100 nm to a few tens of micrometers. Escribano et al (2017), in a regional inverse modeling study, estimated that mineral dust emissions over northern Africa and the Arabian Peninsula ranged between 2550 and 4210 Tg yr −1 , accounting for supercoarse particles with diameters smaller than 30 m; this estimate is reduced by a factor of four to five when only diameters smaller than 6 m are considered. The fine dust particles can be transported for long distances after being lifted from the surface, while the larger dust particles are quickly deposited by gravitational settling, therefore affecting only local ecosystems (Boucher 2015, Tipping et al 2014. Globally, dust is a major source of trace elements contained in minerals (Claquin et al 1999, Nickovic et al 2012, Usher et al 2003, in particular Fe (Jickells et al 2005, Mahowald et  and even for specific minerals it varies significantly, for instance between 0.24% and 0.7% for kaolinite and between 55% and 77.4% for hematite (Nickovic et al 2013, Journet et al 2008. Note that these values are associated with large uncertainty. Estimated ranges for total Fe and total P dust emissions, neglecting supercoarse particles, range from about 35 to as high as 115 Tg Fe yr −1 (Myriokefalitakis et al 2015, Ito andKok 2017) and from 1.1 to 3.8 Tg P yr −1 , Brahney et al 2015. Observations have shown that freshly emitted dust aerosols contain significant amounts of insoluble trace elements, in particular Fe and P, but only a very small fraction of these are in a readily bioavailable soluble form (Tagliabue et al 2017).
The soluble Fe fraction is <0.5% of total Fe mineral dust emissions (Fung et al 2000, Hand et al 2004, Sedwick et al 2007 Atmospheric chemical processing, transport and removal drastically alter Fe solubility in airborne dust (Baker and Jickells 2006 Sholkovitz et al (2012)) while solubilities in the range of 2.3%-67% have been observed in atmospheric processed aerosols, based on observations during oceanographic cruises (e.g. Baker et al 2006b, Furutani et al 2010), and high solubilities of 82 ± 18% have been measured in precipitation in the Southern Ocean (Heimburger et al 2013). Theodosi et al (2010) observations of Fe solubility in rainwater in the eastern Mediterranean for over 2 years have shown a pH-driven Fe solubility, with the lowest solubility of 0.5% associated with Saharan dust episodes (rainwater pH = 8) and the highest (27.5%) with polluted air masses. Overall, the observed atmospheric Fe solubility for various types of aerosols spans over more than three orders of magnitude from 0.01% to 80% (Sholkovitz et al 2012).

Sea spray
Sea spray aerosol formation is mainly driven by wind, and occurs by the ejection of surface seawater to the atmosphere in the form of droplets during the bursting of bubbles in seawater. Evaporation of part of the water from the ejected droplets leads to the formation of aerosols spanning sizes from fine to supercoarse (Lewis and Schwartz 2004). Therefore, sea spray aerosol composition reflects that of surface seawater and, in particular, of the surface microlayer, which is enriched in several substances, primarily organics, compared with the underlying seawater (e.g. Duce and Quinn 1972, Aller et al 2005, Ebling and Landing 2017. Sea spray aerosols contain salts, organic material (fatty acids, monosaccharides, polysaccharides and siliceous material) and trace elements in both organic and inorganic forms, including N and P (Cochran et al 2017).
It has been demonstrated that some elements like P , but also marine bacteria and viruses (Aller et al 2005), can be enriched in sea spray aerosols during natural bubbling of the surface seawater (outgazing) induced by the overlying wind. Sea spray and ocean bacteria and viruses can penetrate inland and affect coastal ecosystems (Aller et al 2005, von Glasow 2008, von Glasow et al 2013. Gong et al (2002) estimated residence times in the marine boundary layer for supercoarse, coarse and fine sea spray aerosols of about 30 min, of the order of few hours, and of days to weeks, respectively, which allows small sea spray particles to be transported for large distances in the atmospheric circulation.

Bioaerosols
Bioaerosols include airborne particles of biogenic origin (e.g. bacteria, viruses, fungi, fungal spores) or various fragments or products released from living organisms (e.g. pollen) (Ariya et al 2009). Their sizes range from 10 nm to roughly 100 m depending on their origin. Until recently attention has been given to bioaerosols because of their impact on human health (causing or enhancing diseases and allergies), on the spread of organisms, some of which might be pathogenic for humans, allowing genetic exchange between habitats and geographic shifts of biomes, and on climate (acting as ice-nucleating agents). These impacts are summarized in the comprehensive review by Fröhlich-Nowoisky et al (2016) and will not be further discussed here, where we focus on cloud chemistry which has received less attention to date. Observa  (Jaenicke 2005). Spanning almost two orders of magnitude, this uncertainty requires reduction. Bioaerosols carry several trace elements that are present in biological material, such as N and P, in bioavailable forms, making a significant contribution to their atmospheric biogeochemical cycles (Kanakidou et al 2012, 2016. Specifically, they contribute about 9 Tg N yr −1 (range 0.6-18.6 Tg N yr −1 ), or about 8% of the total N emissions (Kanakidou et al 2012), and about 0.156 Tg P yr −1 , or about 50% of the primary bioavailable P emissions and about 10% of the total P emissions . However, large uncertainties are associated with these estimates of the amounts bioaerosols emitted (Burrows et al 2009, Myriokefalitakis et al 2017 as well as their P and N content relative to C. The P and N content of bioaerosols has been used to derive their emissions, solubility and atmospheric ageing that converts insoluble to soluble material (Kanakidou et al 2012). More precisely, their N:C and P:C elemental ratios have been observed to range by over one and two orders of magnitude, respectively (Kanakidou et al 2012 and references therein). In addition, bacteria are considered as rather insoluble aerosols, in contrast to pollen grains and fungal spores that contain many highly polar sugars and sugar alcohols (Ariya et al 2009). It is worth mentioning the observed coexistence of bioaerosols and dust aerosols during dust events, as indicated by measurements of airborne microorganisms in the eastern Mediterranean and the west coast of the Atlantic Ocean at levels two to three times higher under dust conditions than during clear atmospheric conditions (Griffin et al 2001(Griffin et al , 2007. These measurements are supported by the observations of highly water-soluble organic nitrogen amino acids in the dust outflows of the Sahara and Gobi deserts (Mace et al 2003, Wang et al 2013.

Volcanoes
Volcanoes are important contributors to both volcanic ash and sulfur dioxide emissions, with passive degassing contributing about 20% on top of total anthropogenic SO 2 emissions to the present day (Klimont et al 2013, Boucher 2015, Carn et al 2017. Although the volcanic contribution to the S budget is more important than to the other co-emitted materials (water, halogens, CO 2 ), the joint contribution of all emitted species can have a different impact in the atmosphere than the changes induced by SO 2 due to changes in atmospheric chemistry (LeGrande et al 2016). Rauch and Pacyna (2009) estimated that volcanic emissions are a greater contributor to the emissions of some metals (Fe, Ni, Cr and Al) than natural biomass combustion sources of these metals. On the basis of volcanic plume observations, Mahowald et al (2008) estimated the volcanic global source of P to be about 0.006 Tg P yr −1 . Although, on a global scale, volcanic ash is a small source of P, regional impacts on ocean nutrient distributions and marine productivity have been detected that are the result of concurrent injections of volcanic S, N, Fe and P (Uematsu et al 2004, Henson et al 2013, Olgun et al 2013. An upper limit to the P and Fe emissions from volcanic eruptions of 0.218 Tg P yr −1 and 0.305 Tg Fe yr −1 , respectively, can be derived by using the volcanic ash content in Fe, P and NH x measured by Duggen et al (2007) and the NH x estimate of emissions from volcanic eruptions. The NH x flux can be derived from Uematsu et al (2004) concurrent observations of NH x and SO 2 during volcanic eruptions and the estimate of global SO 2 emissions from volcanoes (Boucher 2015).
In addition to volcanic eruptions (Benitez-Nelson et al 2003), meteors (Johnson 2001) can also deliver Fe to the atmosphere in the form of iron oxides, which are also potentially available to marine biota, but these emissions make a negligible contribution to the global Fe budget and are not considered in global models.

Biomass burning and fossil fuel combustion sources
In terms of mass, dust and sea-salt aerosols are by far (two orders of magnitude) the largest aerosol emissions on the global scale when considering the PM 10 aerosol fraction (Boucher 2015). However, combustion processes emit more soluble, and thus more bioavailable, aerosols (Ito 2013(Ito , 2015a. Recently the contribution of coarse aerosols to biomass burning emissions has been pointed out (Luo et al 2008, Wang et al 2014. Open fires can be both natural and anthropogenic in origin (Bowman et al 2009, Keywood et al 2011. They emit large amounts of aerosols, mainly black and organic carbon. Combustion processes also emit organic and inorganic trace gases that can form secondary aerosols (organic, sulfate and nitrate), as well as trace elements like Fe and P that are contained in the burnt biomass. According to Rauch and Pacyna (2009), the largest anthropogenic emissions of Al, Fe and Pb to the atmosphere are from fossil fuel combustion, while those of Ag, Cr, Cu, Ni and Zn are from industrial metal production. Gerst and Graedel (2008) estimated that, assuming a similar per capita use of metals as currently, in the future the amount of global in-use metal stocks required could be as high as three to nine times those existing at present. Therefore, anthropogenic inputs of these metals to the environment are expected to increase in the future.
While there is consensus that primary P sources from combustion processes make a significant contribution to global P fluxes in the atmosphere, in particular when the soluble fraction of P is concerned  The reason for the divergence in the estimates is the consideration of different forms of P (i.e. residual or P-containing ash, gaseous or particulate P produced during combustion processes) and different size distributions of the emitted P-containing particulate matter, in particular consideration of significant combustion emissions (Wang et al 2014) or natural emissions (Tipping et al 2014, Brahney et al 2015 in the supercoarse (>10 m) mode.
Biomass burning (Guieu et al 2005) and anthropogenic combustion processes, such as oil and coal fly ash (Luo et al 2008, Sedwick et al 2007, emit aerosols that contribute to atmospheric concentrations of Fe close to source regions but also over the open ocean (Sholkovitz et al 2012). Fe contained in oil fly ash has been observed to be associated with ferric sulfate salts and is thus very soluble (77%-81%) (Schroth et al 2009). Lower solubility has been measured for coal fly ash (20%-25%; Chen et al 2012) and even lower solubilities for biomass burning (∼18%) (Bowie et al 2009). Total Fe emissions from combustion sources (biomass burning, fossil fuel, domestic and agricultural waste burning) have been estimated by Luo et al (2008) based on observed correlations between Fe and black carbon concentrations at 1.07 Tg Fe yr −1 from biomass burning and 0.66 Tg Fe yr −1 from industrial combustion. Ito (2013) has estimated the emissions of Fe from shipping to be 100 times lower, equal to about 0.016 Tg Fe yr −1 for the year 2001. Wang et al (2015b) have evaluated total Fe emissions from fossil fuel and biomass combustion as being almost five times higher, at 5.3 Tg yr −1 (90% confidence of 2.3-12.1 Tg yr −1 ; averaged for the period 1960-2007). The largest estimates account for 72% emissions from combustion in the supercoarse mode, sizes at which aerosols, as already discussed, are expected to have only local impact since they are removed quickly by gravitational settling.

Agriculture
Agricultural activities that use fertilizers and insecticides are known to emit aerosols and precursor gases. In particular, ammonia, nitrogen oxides and nitrous oxide are emitted from fertilized soils and livestock into the atmosphere, and ammonia and nitrogen oxides can produce aerosols (Pozzer et al 2017, Lelieveld et al 2015, Bauer et al 2016, which in turn, after atmospheric transport and via transformation, can return to the surface by deposition and fertilize ecosystems (Duce et al 2008). In addition the use of insecticides (e.g. organophosphates; FAO 2001 www.fao.org/ faostat/) is a source of organic P particles entering the atmosphere (Kanakidou et al 2012).
Based on the above discussion, the primary emissions of N, P and Fe per source category are summarized in table 1.

Secondary sources-multiphase chemistry
Chemical transformation of sulfur dioxide, nitrogen oxides, ammonia and volatile organic compounds in the atmosphere leads to the formation of aerosols. This can occur by the formation of gas phase oxidation products like sulfuric acid, nitric acid and multifunctional organic compounds. Such compounds can nucleate (H 2 SO 4 /H 2 O, H 2 SO 4 /H 2 O/NH 3 and extremely lowvolatility highly oxidized organics) to produce new particles (Zhang et al 2012, Kulmala et al 2013, Riccobono et al 2014, Dunne et al 2016, Kirkby et al 2016 in the atmosphere or partition onto existing ones (Seinfeld and Pandis 2006). They thus change the mass, size, composition and physicochemical and optical properties of the aerosol. Chemical ageing of aerosols through multiphase reactions also leads to particle growth, by modifying aerosols in the aqueous phase of the atmosphere (e.g. Pöschl and Shiraiwa 2015).
Di-nitrogen (N 2 ) is abundant in the atmosphere and is converted to reactive nitrogen by bacteria (NH 4 + , NO 3 − ), lightning (NO) or anthropogenic activities (mainly as NO, NO 2 , HNO 3 , NH 3 , N 2 O); this is then assimilated by living organisms. Reactive nitrogen is also an important driver of atmospheric chemistry, since the availability of NO x (NO + NO 2 ) controls tropospheric ozone chemistry. In the atmosphere NO x is chemically converted to organic nitrogen compounds and HNO 3 ; together with (primarily) sulfuric acid this contributes to atmospheric acidity, which controls the partitioning of HNO 3 /NO 3 − to the particulate phase. Ammonia has important emissions to the atmosphere, similar to or even higher than NO x , depending on the region and year, and is the main neutralizing gas for these acidic compounds (Seinfeld and Pandis 2006). The availability of ammonia, together with environmental factors like temperature and relative humidity, controls the amount of ammonium and nitrate aerosols.
Multiphase chemistry is also well known to make a significant contribution to atmospheric sulfate production by oxidation of the dissolved sulfur dioxide to SO 4 2− via reactions mainly with H 2 O 2 and O 3 . Transition metals play an important role in this conversion when photochemical processes are slow (at mid to high latitudes during winter) (Seinfeld and Pandis 2006, Alexander et al 2009, Harris et al 2013. Indeed, several metals have been measured in the atmospheric aqueous phase (aerosol water, fog, cloud and rain), among which Fe and Cu are the most important chemically when taking into account the measured concentrations and reactivity. There, the presence of the transition metal ions of Fe, Cu and Mn affects HO x (OH, HO 2 , H 2 O 2 ) concentrations and recycling, and their involvement in 'Fenton' reactions with H 2 O 2 is an important source of radicals in the aqueous phase (Sedlak et al 1997).
A large variety of organic volatile compounds (VOCs) of anthropogenic and biogenic origin have the potential to form secondary organic aerosol (SOA) (Kanakidou et al 2005, Hallquist et al 2009. Biogenic VOCs, including aldehydes, ketones, alcohols and terpenoides, are emitted to the atmosphere mainly by vegetation; the largest single compound emitted is isoprene (Guenther et al 2012). Gas phase oxidation of organic compounds with several carbon atoms in their molecules and subsequent partitioning of their products to the aerosol phase has been found to contribute to the formation of SOA (Kanakidou et al 2005, Hallquist et al 2009, Shrivastava et al 2017, while multiphase organic photochemistry is also able to produce SOA (Blando and Turpin 2000) from the oxidation of smaller molecules such as acetylene (Volkamer et al 2009, Lim et al 2010. Chamber and field observations show that acidity of aerosols increases SOA formation and that sulfate esters are formed during the oxidation of isoprene in the presence of SO 4 2− . The formation of SOA from isoprene and monoterpenes is also observed to be mediated by anthropogenic SO 2 and NO x emissions through the presence of SO 4 2− in aerosols and NO 3 radical chemistry (Xu et al 2015b). Therefore, SO 4 2− is involved in the transfer of biogenic volatile organic compounds as gas-phase reaction products to the condensed phase and has been suggested to be important in the formation of humic-like substances (HULIS) in ambient aerosols (Graber andRudich 2006, Surratt et al 2007).
Multiphase chemical reactions produce oxygenated organics and thus increase the O:C content of aerosols, reconciling model results with observations. For this reason, the O:C ratio has been suggested as a marker for aqueous multiphase chemistry (Waxman et al 2013). Glyoxal could make a significant contribution to the mass of ambient organic aerosol (OA) after hydration through multiphase chemistry that favors oxidation to form organic ligands like oxalic acid (Lim et al 2010, Ervens and Volkamer 2010, Myriokefalitakis et al 2011 and oligomerization reactions even in the absence of acidity (Wang et al 2010, Zhang et al 2012. Di-acids are important components of water-soluble organic C. They thus affect the water associated with aerosols and the aerosol pH. They also act as bidentate (chelating) ligands forming complexes with metals like Fe and Cu, capturing the potential toxic free ions into less harmful forms for ecosystems, facilitating metal solubilization and providing a metal buffering capacity that maintains the activities of free metal ions at a natural level against relatively small perturbations of metals in the marine environment (Hirose andSugimura 1985, Hirose 2006).
Atmospheric reactions of inorganic nitrogen (nitrogen oxides and ammonia) with organic compounds in the gas and the aqueous/aerosol phases can lead to the formation of organic nitrogen compounds and subsequent transfer of inorganic nitrogen to the aerosol phase (Atkinson and Arey 2003, Ervens and Volkamer 2010). Recent laboratory investigations have shown that photosensitizers can initiate aerosol growth in the presence of gaseous volatile organic compounds, in particular the imidazole derivatives, like imidazole-2 carboxaldehyde, that are formed by reaction of glyoxal with ammonium cations (Rossignol et al 2014). Dark reactions including acid-catalyzed aerosol chemistry of hydronium or ammonium ions, which protonate epoxide and carbonyl groups thus enabling attack by nucleophiles such as water, SO 4 2− , NH 3 and NO 3 − , may be responsible for the ring opening mechanism of epoxides, organosulfate formation by epoxides and carbonyls and C-N-containing lightabsorbing organic species as part of the 'brown carbon' pool (Surratt et al 2010, McNeill 2015, Laskin et al 2015. Night-time chemistry driven by NO 3 radicals also produces organic nitrogen in the aerosol phase that was found to be 5%-12% of OAs during the summer months (Xu et al 2015a). Kiendler-Scharr et al (2016) found that 34%-44% of nitrate in submicron aerosols in Europe is organic nitrogen and a significant fraction of it can be attributed to gas-phase NO 3 radical night-time chemistry. The secondarily formed organic N contributes to the atmospheric deposition of organic N, which in turn makes up around 20%-25% of the global total of soluble N deposition (Cornell 2011, Kanakidou et al 2012, Jickells et al 2013. Overall, the SOA source estimates span more than an order of magnitude from 12 to about 500 Tg yr −1 ; these estimates were made using different assumptions of SOA volatility and SOA fragmentation that also affects volatility (Tsigaridis et al 2014, Shrivastava et al 2017. SOA formation in both cloud and aerosol water has been estimated to range between 13 and 47 Tg yr −1 , and between a small loss (−0.4 Tg yr −1 ) and a significant source (13 Tg yr −1 ), respectively (Myriokefalitakis et al 2011. A significant amount of organic N is attached to these aerosols. According to global model estimates, this amount lies between about 3 and 18 Tg N yr −1 and contains both oxidized and reduced organic N . The secondary production of organic N can be of similar magnitude to primary emissions of organic N from specific sources, as has been estimated for biomass burning (Ito et al 2015).

Aerosol processing: solubilization of nutrients and the role of acidity
Desert dust contains a large variety of minerals and metals, most of them in insoluble form (Claquin et al 1999, Nickovic et al 2012. However, to be bioavailable to ecosystems elements have to be in soluble form (or in nanoparticles of diameter smaller than 0.2 m). In addition, atmospheric aerosol observations show the presence of noticeable amounts of soluble elements such as Fe (Sholkovitz et al 2012) and P (Mackey and Paytan 2009, Baker et al 2006a. The mechanisms that convert insoluble material to the soluble material that is readily bioavailable to ecosystems remain under investigation.
Uptake of Fe, the fourth most abundant element in the Earth's crust, by living organisms is indeed limited because it is mainly present in the form of insoluble iron oxides. However, various cellular mechanisms have evolved to capture Fe from the environment in biologically useful forms. For instance, microbes excrete siderophores to capture Fe as six-coordinate complexes with Fe(III), or they use other mechanisms to reduce iron from the insoluble ferric form (Fe . Both biotic and abiotic processes proceed initially by the production of inorganic or organic acids, with a simultaneous decrease of pH, and of the ligands needed to react with minerals and release nutrients to the solution.
Indeed, solubilization of the Fe and P present in atmospheric aerosols is found to be affected by atmospheric acidity (Shi et al 2011a(Shi et al , 2012 and to increase with the acidity of the aerosol and cloud water (Theodosi et al 2010). Furthermore, analysis of observations using aerosol equilibrium simulations has supported the importance of the aerosol acidity for the conversion of Fe(III) to the more soluble Fe(II) and the role of anthropogenic sulfate in modulating acidity in the Pacific (Meskhidze et al 2003). Fe requires much more acidic conditions than P in order to be solubilized (van Cappellen and Berner 1991, Guidry and Mackenzie 2003, Myriokefalitakis et al 2015, 2016, Stockdale et al 2016. Three different chemical mechanisms for Fe solubilization have been identified so far: proton-driven, ligand-driven (Paris et al 2011, Paris andDesboeufs 2013) and photo-induced reductive Fe solubilization (Chen and Grassian 2013). Reductive dissolution involves conversion of Fe(III) to Fe(II) by electron transfer (Larsen andPostma 2001, Fu et al 2010) and is significantly accelerated in the presence of Fe(II) or Fe(II) ligand complexes (Litter et al 1994). Al-Abadleh (2015) reviewed the bulk and surface chemistry of Fe in the presence of HULIS and reported direct photoreductive formation of Fe(II) or organic Fe complexes in photoreductive experiments, with the highest rates being observed at low pH values. Shi et al (2012) made a comprehensive laboratory investigation of the factors that affect Fe solubilization from dust aerosols. Their results point to acid processing rather than cloud processing as a prime mechanism that increases the solubility of dust Fe during transport in the atmosphere. Furthermore, laboratory studies have shown that Fe solubilization from minerals occurs on different timescales-from hours to weeks depending on the particle size and mineralogical composition (Shi et al 2011b). For instance, in acidic conditions octahedral cations were found to be released more readily than tetrahedral cations and the preference for metal release depends on the metal-O bond; in particular the number of protons needed to release the metal depends on the metal valence (Bibi et al 2011). Thus phyllosilicates are solubilized much faster than aluminosilicates (Ito and Shi 2016a). Laboratory experiments have also shown that the formation of mononuclear bidentate complexes of oxalic acid with Fe surfaces enhances dissolution of Fe in dust and combustion aerosols more than sulfuric acid when both acids are present in the atmosphere (Chen and Grassian 2013). The breaking of Fe-O bonds at the mineral surface controls the rate of dissolution of Fe, which increases as pH decreases. Chen and Grassian (2013) also found that when oxalate concentrations are low the dissolution of Fe is suppressed because dissolved Fe competes with surface Fe for complexation.
Experiments on laboratory dust solubilization have shown that solubilization of P and Ca 2+ , i.e. release from dust, depends on the amount of H + in the dust aerosol water, i.e. the aqueous layer surrounding the dust particle (Stockdale et al 2016). At low H + concentrations the release of P is controlled by water volume alone, while above 10 −4 mol. g −1 an approximately linear relationship exists between P solubilization and H + concentration until the acid reactive phosphate pool of the dust aerosol is exhausted. Ca 2+ behaves similarly, but with lower threshold values for H + than P. Indeed, H + can react with PO 4 3− and the OH or F groups of apatite minerals at the crystal surface, thus weakening the Ca 2+ bonds and mobilizing PO 4 3− from the crystal surface (Christoffersen and Christoffersen 1981). Such a mechanism of desert dust solubilization under polluted conditions has been suggested to explain the observed PO 4 3− levels over the eastern Mediterranean, a typical region where Saharan dust can interact with polluted air masses from Europe and the Middle East (Nenes et al 2011).
Natural aerosols, especially the larger sized ones, are enriched in cations like Ca 2+ , Mg 2+ and Al 3+ (dust), Na + , K + and Mg 2+ (sea spray), and K + (biomass burning). These aerosols show more basic characteristics than anthropogenic aerosols, which are smaller in size, richer in acids and thus have a more acidic behavior than natural aerosols (Bougiatioti et al 2016). This reduced acidity of coarse aerosols significantly suppresses dissolution of Fe and P, while it favors the partitioning of nitrate to the aerosol phase (Weber et al 2016, Karydis et al 2016. It also partially explains the observed inverse relationship between Fe solubility and particle size (Baker and Jickells 2006, Ito andFeng 2010).
Dust aerosols coated by acidic compounds (e.g. sulfates and nitrates), which form by heterogeneous uptake of SO 2 and HNO 3 during atmospheric transport (Bauer et al 2007), show increased Fe solubility compared with fresh dust particles, in particular in fine mode (Meskhidze et al 2005). Such particles can be transported and deposited in remote regions, for example over the Atlantic and Pacific oceans (Baker et al 2006c, Sedwick et al 2007. In line with these findings, a positive correlation between Fe solubility and sulfur emissions has been observed for acidic atmospheric samples collected at urban sites (Oakes et al 2012), supporting substantial atmospheric processing or anthropogenic combustion sources of soluble Fe.
Atmospheric acidity also affects nutrient deposition via its control on the partitioning of HNO 3 /NO 3 − to the aerosol phase (Weber et al 2016). Indeed, in order to be captured into the aerosol phase, NO 3 − requires an aerosol pH higher than 2 to start transfer to the aerosol phase; at a pH of about 4 or higher it is mostly present in the aerosol phase, due to the semivolatile character of NH 4 NO 3 . Such pH constraints can be fulfilled by coarse dust aerosols, where nonvolatile cations like Ca 2+ keep pH levels relatively high, or on biomass burning aerosols, where K + cations play a similar role. On the other hand, anthropogenic or acid-coated fine aerosols suppress HNO 3 /NO 3 − partitioning to the aerosol phase. Because of the different removal rates of HNO 3 and NO 3 − through deposition, and thus removal lifetimes, aerosol acidity, via its impact on HNO 3 /NO 3 − partitioning, will also affect the distribution of HNO 3 /NO 3 − deposition.

Nutrient biogeochemical cycles (including organic fractions)
Atmospheric aerosols, either directly emitted or chemically formed in the atmosphere, undergo atmospheric transport and physical and chemical transformations before being deposited to the surface where they release nutrients. Atmospheric aerosols are therefore a significant component of nutrient biogeochemical cycles, and the spatiotemporal variability of the atmospheric deposition fluxes and their intensity depend on atmospheric transport and ageing. In the following we will focus on the N, P and Fe cycles, as these elements are essential nutrients for ecosystems and are known to limit primary productivity depending on their availability and location. N, P and Fe are integral parts of living organisms. N is an essential component of amino acids, which are the structural basis of proteins, and is also contained in several other biological molecules such as nucleotides and nucleic acids (deoxyribonucleic acid or DNA and ribonucleic acid or RNA), making it necessary for life (Galloway et al 2008). In addition, N-containing molecules carry electrons and hydrogen, both of which are necessary for organismal functions such as respiration and photosynthesis. Indeed, N is a component of chlorophyll, where N atoms bind Mg 2+ at the center of the chlorophyll structure. P is also essential for the formation of DNA and RNA; it is required for bone formation and for adenosine triphosphate (ATP), which stores energy in cells and supplies it to cellular processes, and for hemoglobin, which transports oxygen in blood (Smil 2000). The availability of P strongly influences the fixation of inorganic C into cellular biomass by photosynthetic organisms. P supply is a major factor governing the productivity of ecosystems, especially during long-term ecosystem development (Benitez-Nelson 2000). Fe is an essential element for almost all living organisms as it participates in a wide variety of metabolic processes, including oxygen transport, DNA synthesis and electron transport (Abbaspour et al 2014). The importance of Fe comes from its involvement in redox (reduction/oxidation) reactions and its ability to reversibly change its oxidation state by electron transfer, which varies depending on the ligand with which Fe is complexed (Guerinot 1994). Several enzymes involved in photosynthesis, respiration and N fixation (Twining et al 2004) require Fe, which makes it an important micronutrient for marine organisms; shortage of Fe can limit primary productivity over large regions of the ocean, the socalled high-nutrient/low-chlorophyll (HNLC) regions (Martin et al 1991, Jickells et   They also evaluated that about 40% of the total N deposition is in the form of atmospheric aerosols. Since the start of industrialization, human activities have dramatically increased the amounts of N emitted to the atmosphere and deposited to the surface, modifying the N biogeochemical cycle (Galloway et al 2008). Using the anthropogenic and biomass-burning Atmospheric Chemistry and Climate Model Intercomparison Project (ACCMIP) historical emissions and the Representative Concentration Pathway (RCP6.0 and RCP8.5) future emissions scenarios in a global model, and accounting for the organic N as explained above, human-driven N deposition has been shown to have globally increased by about 80 Tg N yr −1 since 1850. This change, from about 50 Tg N yr −1 in 1850 to about 130 Tg N yr −1 in 2005, is more pronounced over the land (almost a three-fold increase), where most of the anthropogenic sources are found, compared with over the ocean (a two-fold increase) .
A similar global increase (by a factor of 1.9) has been estimated by Brahney et al (2015). Projections for the year 2050 show a further increase in N deposition by less than 5 Tg N yr −1 (to about 135 Tg N yr −1 ) due to the implementation of air pollution abatement strategies leading to a reduction in NO x and SO 2 emissions . These deposition estimates also include the emission and re-deposition of marine organic N and ammonium, which will not lead to a net increase in marine productivity (Jickells et al 2017). The contribution of soluble organic N to total soluble N has been reduced from 40% in the past to 20% currently, and is projected to remain at roughly this level in the future (figure 3). Large uncertainties are associated with these results because they are based on an incomplete understanding of the N content of organics and neglect future changes in biogenic emissions driven by climate. The future projections also indicate significant changes in the regional distribution of N deposition and its chemical composition, with reduced compounds gaining in importance relative to oxidized ones (figure 4), but small changes in the global total N flux (Lamarque et al 2013. A qualitative change in the chemical composition of the deposited N is also found in particular with regard to NO 3 − deposition, which has increased since 1850 and is projected to be reduced in the future due to reductions in SO 2 and NO x emissions to improve air quality. This is not the case for NH x , which is projected to continue to increase due to lack of a legislative requirement to reduce its emissions. Total organic N deposition is projected to remain almost unchanged in the future. This difference in the composition of N deposition is expected to affect ecosystem growth, since ammonium is taken up more quickly by vegetation than nitrate, while little is known about how or how fast organic N is assimilated by organisms. The form of N that is used by an organism should reflect that organism's enzymatic systems for absorbing and incorporating N (Rice 1992). Plants that take up NO 3 − will need to reduce it to NH 4 + before it can be used for synthesis of amino acids and other nitrogenous organic compounds: this transformation, however, requires energy. Nitrate reductase is involved in the first step of the reduction of NO 3 − into a form that can be used by plants, and is the enzyme that controls the ability of plants to use NO 3 − ; therefore those species that use mostly NO 3 − as their N source should have relatively high levels of nitrate reductase activity (NRA) (Rice 1992). For algae, Sigee (2005) reported that because uptake of ammonia does not require energy, and there is no induction period for enzymes required for assimilation when both nitrate and ammonia are available, ammonia is the preferred choice. The energy required for nitrate reduction is derived from photosynthesis. Organic compounds can be only used via specialized enzymatic processes (Gattuso and Hansson 2012).
In the ocean, organic N is converted to dissolved organic N by heterotrophic bacteria, and is then remineralized to NH 4 + by several potential processes such as enzymatic breakdown and degradation of biomass (Sigee 2005). Therefore, the model-simulated changes in the composition of the deposited N  might favor the development of species that are more able to assimilate the available forms of deposited N that have increased.
Large continental areas of the world are subject to N deposition fluxes that exceed 10 kg N ha −1 yr −1 , i.e. that are more than an order of magnitude greater than natural fluxes and exceed the critical loads that have detrimental impacts on ecosystems (Bobbink andHettelingh 2011, Galloway et al 2008). N availability limits primary productivity over large areas of the global ocean (Wang et al 2015a, Duce et al 2008. Thus, in low-N marine ecosystems atmospheric deposition of N enhances primary productivity (e.g. Zamora et al 2010). Duce et al (2008) estimated that the humaninduced increase in deposition of atmospheric N to the oceans may account globally for up to 3% of the annual new oceanic primary productivity, in parallel with export into suboxic or anoxic zones that will increase marine emissions of N 2 O to the atmosphere (Suntharalingam et al 2012). Higher contributions of atmospheric N deposition to marine productivity have been estimated for semi-enclosed marine ecosystems such as the Mediterranean Sea, for which atmospheric deposition of N (and P) may account for up to 35%-60% of new production (Christodoulaki et al 2016). Recently, Jickells et al (2017) used these updated estimates of N influx external to the ocean to re-evaluate the global impact of anthropogenic atmospheric deposition of N on C sequestration by marine ecosystems. Their new estimate of annual marine primary productivity is much lower than earlier estimates, and this effect is almost compensated by N 2 O emissions. Interestingly, potential ecological shifts through changes in nutrient regimes have been also reported, driven by atmospheric deposition of nutrients (e.g. Krishnamurthy et al 2009, Chien et al 2017, Wang et al 2015a. These results show the complexity of the effects of atmospheric deposition that involve several feedback mechanisms and need to be evaluated synergistically.

Phosphorus
Unlike atmospheric reactive N, P (and Fe) is present in the atmosphere mainly in the aerosol phase, with the exception of a few phosphines which are rather scarce in the environment. P has a variety of natural sources including dust, volcanoes, sea spray, freshwater wetlands (that emit phosphine) (Glindemann et al 2005, Han et al 2011 and biomass burning. It is also produced by a number of human activities, including deforestation, P mining and agricultural practices (fertilizers and insecticides). The human-driven emissions increase the solubility of atmospheric P and cause the accumulation of P in soils and aquatic environments (Mackey and Paytan 2009). The first estimates of the total global deposition fluxes of P based on extrapolations from scarce observations varied between 3 and 4 Tg P yr −1 Duce 1979, Smil 2000). Of this amount, only a fraction is soluble or in the nanoparticle size range that is readily available to ecosystems.
The first global simulations of the atmospheric P cycle were focused on dust sources, considering a fixed 10% solubility of P on dust aerosols, 100% solubility of P in volcanic emissions and 50% from other sources . Recent studies have included dust, combustion and biological aerosols in the supercoarse mode as P sources (Tipping et al 2014, Wang et al 2014, Brahney et al 2015, thus accounting for the total P in the atmosphere. Brahney et al (2015) argue that aquatic organisms have evolved numerous ways to rapidly acquire P from organic molecules, therefore in the long-term all P would potentially be available.
Recently, Myriokefalitakis et al (2016) provided the first global modeling study of the P atmospheric cycle with an explicit parameterization of P solubilization on mineral dust under acidic conditions in the atmosphere, occurring both in the aerosol and in cloud water. They showed that neglecting the dissolution of P degrades the comparison of model results with observations. They also found that the chemical processing of dust contributes about one-third (0.14 Tg P yr −1 ) of the global atmospheric source of soluble P; this has increased over the past 150 years by about 70% due to changes in anthropogenic and biomass burning emissions that have affected atmospheric acidity. Future reductions of about 30% were projected for 2100 due to the implementation of air quality regulations that also affect atmospheric acidity. The present-day global annual soluble P deposition flux of 0.45 Tg P yr −1 was calculated to have a strong spatial and temporal variability, with about 40% of this amount being deposited over the oceans (figure 2). This soluble P deposition flux, which accounts for soluble organic P and phosphate, is almost double the estimate of phosphate emissions by Mahowald et al (2008), in which neither organic P nor the chemical processing of dust aerosol were considered. Accounting for changes in both the anthropogenic emissions of P and the solubilization of P due to chemical processing, Myriokefalitakis et al (2016) estimated that about 20% of the increase in the soluble P flux was due to changes in anthropogenic emissions since 1850, and projected a reduction of ∼15% by 2100 due to changes in anthropogenic emissions. They also evaluated that the contribution of bioaerosols to the total deposition of bioavailable P to the marine environment can seasonally exceed 50% regionally, indicating the existence of potentially important interactions between terrestrial and marine ecosystems.
The computed contribution of soluble organic P to the bioavailable dissolved deposition (DP) varies between 37% and 40%, with the lowest contribution occurring at present. As discussed in section 2, the contribution of biological particles to bioavailable P is also considerable, as pointed out by Myriokefalitakis et al (2016).
The computed atmospheric deposition of readily bioavailable P over the global ocean varies between 0.096 Tg P yr −1 (Mahowald et al 2008) and 0.17 Tg P yr −1 . While riverine inputs mainly affect coastal regions, atmospheric deposition is a source of nutrients for the open sea (e.g. Okin et al 2011).
While N and Fe are limiting nutrients for large fractions of the ocean (Moore et al 2013), P is known to co-limit primary productivity over certain regions like the Mediterranean (Krom et al 1991), where the P limitation has been attributed to the anomalously high N:P ratio in the atmospheric deposition fluxes . The disproportionate increase of atmospheric deposition of anthropogenic N over P in the eastern Mediterranean can explain such behavior (Christodoulaki et al 2016) and the atmospheric fluxes of N and P may have induced a total increase in phytoplankton biomass of 16% over the past 150 years. Furthermore, Richon et al (2017) simulations for the entire Mediterranean Sea have shown that atmospheric deposition of N and dust P increases primary productivity over vast regions on average by 30%-50% and that the fertilization effects are transmitted along the biological chain. They also calculated that the dust input of P increases primary productivity by up to 30% when surface stratification occurs, results that are in agreement with those of Christodoulaki et al (2013) for the eastern Mediterranean.
Similarly to the Mediterranean case, the increased anthropogenic N flux to the north Pacific Ocean and the observed resulting increase in surface nitrate levels provide an excess N input compared with P and this is expected to potentially induce a P limitation in that area (Kim et al 2014). Brahney et al (2015), analyzing data from 700 upland lakes globally, found a P limitation induced by the much faster atmospheric flux of anthropogenic N. They also suggested that despite the smaller changes in the flux of P compared with that of N, P may have a greater impact on primary productivity, because, on average, based on the Redfield ratio, one unit of increased P deposition has 16 times the influence of one unit of N deposition (Brahney et al 2015). Human-induced P limitation has also been recorded in the forests of south China, where both N and P deposition have increased but the increase in anthropogenic N deposition is faster than that of P deposition.

Iron
Fe is present in the aerosol and water phases of the atmosphere. Mineral dust is the dominant source of Fe, contributing ∼95% of the Fe atmospheric load on a global scale while the remaining ∼5% is contributed by volcanoes, biomass burning and anthropogenic combustion processes . However, Fe from non-dust sources is more soluble than mineral dust on aerosols (Luo et al 2008, Ito 2013, Ito 2015a. Global emissions estimates of total Fe vary by a factor of three, from 37-115 Tg Fe yr −1 , Ito and Kok 2017, Ito and Shi 2016b, Ito and Xu 2014, Johnson and Meskhidze 2013, Luo et al 2008, Myriokefalitakis et al 2015, Wang et al 2015b, with a small fraction (1-2 Tg-Fe yr −1 ) emitted as soluble Fe (Mahowald et al 2017). Representation of Fe in global models has continuously improved over the years by accounting for the variable mineralogy of dust emissions and different Fe contents in the various minerals (Nickovic et al 2012, Journet et al 2014. Combustion emissions have also been considered in more detail (Luo et al 2008, Ito 2013, Wang et al 2015b, but this has increased the uncertainty range for these emissions estimates, which vary between ∼2 and 5.3 Tg Fe yr −1 . The uncertainty in the Fe emission estimates has been evaluated to be a factor of four to five for both dust and combustion sources (Luo et al 2008).
Solubilization through atmospheric processing has recently been parameterized in models with everincreasing detail: Luo et al (2008) performed a sensitivity study; Johnson and Meskhidze (2013) considered only dust solubilization and scaled oxalate to SO 4 2− ; Myriokefalitakis et al (2015) explicitly calculated oxalate through the multiphase chemistry of carbonyls (Myriokefalitakis et al 2011), protonpromoted Fe solubilization for three different types of Fe minerals and combustion aerosols in both aerosol water and cloud droplets and ligand-promoted Fe solubilization in cloud droplets only; Ito (2015b) accounted for proton-promoted solubilization considering one type of Fe mineral, and oxalate dissolution processes accounted for the photo-reductive dissolution process but for both dust and combustion aerosols. This latter mechanism has been found to have only a minor impact on Fe solubilization. Myriokefalitakis et al (2015) calculated a solubilization flux of 0.175 Tg Fe yr −1 that is half of the total primary soluble Fe emissions from both mineral and combustion sources in their model. About 22% of this solubilization flux is attributed to organic ligands.
Overall, modeled estimates of total Fe deposition fluxes to the ocean range between 7 and 11 Tg Fe yr −1 (Ito and Kok 2017, Ito and Shi 2016b, Ito and Xu 2014, Johnson and Meskhidze 2013, Luo et al 2008, Myriokefalitakis et al 2015, Wang et al 2015b and show significant spatial and temporal variability. Soluble Fe deposition fluxes to the ocean have been reported to range between 0.11 and 0.34 Tg Fe yr −1 (Johnson and Meskhidze 2013, Myriokefalitakis et al 2015, Ito 2015a, Ito and Shi 2016a. The uncertainty in the simulated deposition fluxes of soluble Fe to the ocean has been evaluated to range between −50% and +150%; this is associated with the uncertainty in Fe-containing minerals commonly found on dust (Johnson and Meskhidze 2013).
Humans have significantly modified the Fe atmospheric cycle (Tagliabue et al 2017) and atmospheric deposition by increasing direct emissions of Fe from industrial (e.g. smelting) and combustion processes , such as combustion of fossil fuels and coal, and the acidity of atmospheric aerosols, which affects the conversion of mineral Fe to more soluble and thus bioavailable forms. Myriokefalitakis et al (2015), based on historical and projected emission scenarios (as for P), calculated that atmospheric deposition fluxes of soluble Fe have more than doubled since 1850 due to increases in the primary non-dust emissions and in dust-Fe solubilization. They also projected that the deposition flux of soluble Fe will decrease by about 25% due to reductions in primary anthropogenic and biomass burning emissions as well as in the dust-Fe solubilization flux. Similarly, Ito and Xu (2014) estimated that the improvement in air quality projected for the future will lead to a decrease of the deposition of filterable Fe (soluble and particulate of diameter <0.02 m) from Fe-containing mineral dust to the eastern North Pacific due to lower acidification in Asian dust, which is mainly associated with the reduction of NO x emissions. Ito and Shi (2016a), accounting for dust and combustion emission of Fe, estimated that the soluble Fe flux to the ocean has increased from 0.05-0.07 Tg Fe yr −1 in the pre-industrial period to 0.11-0.12 Tg Fe yr −1 at present due to air pollution. These numbers do not account for the oxalate-promoted solubilization of Fe on dust aerosols. Myriokefalitakis et al (2015) have estimated that Fe-oxalate complexes, produced by ligand-promoted dust-Fe solubilization in cloud water, made a global contribution of between 7% and 18% to soluble Fe deposition, with the smallest contribution calculated for the present and the highest when considering anthropogenic emissions in 1850 (figure 3). This fraction has to be considered as a lower limit estimate since ligand-promoted Fe solubilization in aerosol water was not considered in their study.
Note, however, that these studies do not take into account changes in dust emissions for past and projected estimations. Dust emissions have been estimated to have increased by about 40% over the last century (Mahowald et al 2017). Overall, the changes due to atmospheric pollution as well as the projected air quality improvements are expected to affect the oceanic productivity, in particular over the high-nutrient/ low-chlorophyll oceanic regions.
The external supply of Fe in the open ocean can affect a number of biogeochemical processes, as nicely reviewed by Tagliabue et al (2017). Fe deposition can significantly stimulate N fixation and biological C export on a global scale. Increased Fe deposition has also been proposed to lead to increased regional respiration and productivity and expansion of minimum oxygen zones (Ito et al 2016), thus affecting life in the ocean. Fe supply also has an impact on ecosystem biodiversity. For instance, a high Fe supply favors algae with red-line plastids (diatoms and cryptomonads) to have a lower Fe:P content than algae with green-line plastids, such as chlorophytes and prasinophytes (Twining and Baines 2013). Furthermore, phytoplankton in the Fe-rich tropical Atlantic exhibit Fe:P stoichiometries more than five times higher than in the Southern Ocean under low-Fe conditions (Twining and Baines 2013). In the Fe-poor Southern Ocean, ferric reductases enable the use of the more dominant organically complexed Fe pool by phytoplankton, and this strategy appears to be common in organisms adapted to a 'bioavailability envelope' (Tagliabue et al 2017).

Importance of humans in changing atmospheric acidity and nutrient biogeochemical cycles
As discussed earlier, anthropogenic emissions of atmospheric acid precursors (NO x and SO 2 ) have been increasing over the past 150 years and only started to decline in the last 30 years due to measures taken to mitigate air pollution (Lamarque et al 2010, van Vuuren et al 2011. In parallel, emissions of ammonia, the main alkaline gas that neutralizes these acids in the atmosphere, have also changed, although at different rates than NO x emissions (Lamarque et al 2010). This is because of the absence of direct regulations on NH 3 emissions which have been indirectly affected, at least in Europe, by changes in agricultural practices since 1990 (EEA 2010). These emission changes have also modified the acidity of the various atmospheric compartments (gases, aerosols, clouds, rain), which responded differently to these changes because different solution densities and different fractions of pollutant emissions are involved (Driscoll et al 2016, Pascaud et al 2016, Watmough et al 2016, Weber et al 2016. This led to changes in all three nutrients discussed here, namely N, P, and Fe. Because of the different mechanisms for the dependence of atmospheric deposition of the soluble fraction of N, P and Fe on atmospheric acidity, it is expected that the relative abundances of the soluble fraction of these nutrients in the deposition flux has also been modified since 1850 and will change in the future. Indeed, based on recent results obtained with the same global model for N , P  and Fe (Myriokefalitakis et al 2015) (total soluble fluxes shown in figure 2 for present-day simulations), the changes in the N:P:Fe stoichiometry of the deposition fluxes can be derived. These changes reflect not only chemistry-driven modifications of the deposition fluxes but also those due to the primary anthropogenic emissions of these nutrients. N deposition, in particular nitrate deposition, is also affected by aerosol acidity since the partitioning of HNO 3 /NO 3 − to the aerosol phase is prohibited under very acidic conditions (Weber et al 2016). Because of the high deposition rates of HNO 3 when compared with fine aerosols (Brook et al 1997), Kajino et al (2008) concluded that atmospheric deposition of total NO 3 − (sum of HNO 3 gas and NO 3 − particulates) can increase, modulated by the presence of SO 4 2− , even when NO x emissions remain constant or slightly decrease. This behavior is confirmed by the analysis of global simulations of the atmospheric N cycle  which show that at present a larger fraction of total NO 3 − is in the gas phase (as HNO 3 ) than has been in the 'past' and will be in the 'future'. As a consequence, the lifetime of the total NO 3 − due to atmospheric deposition is calculated to have decreased by 35% from the 'past' to the present day and to increase by about 20% in the 'future'. These changes in partitioning between gas and aerosol phases are modulated by changes in acidity, with key roles played by SO 2 and NH 3 emissions, and lead to changes in the geographical pattern of emissions.
Global model simulations enable a distinction between the two factors (chemical processing and primary emissions) affecting the soluble fractions of Fe and of P in atmospheric deposition. In particular, the proton-and ligand-driven solubilization of dust-Fe (Myriokefalitakis et al 2015) and the acidity-driven solubilization of dust-P  have been considered. The deposited soluble fluxes of Fe (and P) carried by dust alone, calculated as the sum of the initially emitted soluble dust-nutrients and the solubilized amounts during chemical processing (ageing) in the atmosphere, were found to be higher at present by 12% and 24% for Fe and P, respectively, compared with 'past' depositions (anthropogenic and biomass burning emissions of the year 1850) (Lamarque et al 2010). An almost equivalent decrease was estimated for when these emissions reach 2100 levels ('future'), as in the RCP6.0 scenario (van Vuuren et al 2011). These decreases are driven by much larger changes in the solubilized amounts during atmospheric ageing and are entirely due to the modification of atmospheric acidity and the ligand content of aerosol and cloud water induced by the changes in anthropogenic and biomass burning emissions.
As already mentioned, Fe and P show a different dependence on atmospheric acidity-Fe requires a more acidic environment to solubilize than Pso the molar ratio of the atmospheric deposition of soluble Fe to that of soluble P associated with dust aerosols is expected to be affected by the changes in anthropogenic and biomass burning emissions. The results of Myriokefalitakis et al (2016) indicate significant increases from the 'past' to the present in the ratio of soluble Fe to soluble inorganic phosphate (DFe/DIP) when accounting for both primary and secondary sources (figure 5), in particular over tropical oceans in the outflow of desert dust emissions mixed with the atmosphere from Asian and Central American pollution and tropical biomass burning. In the 'future' precursor emissions should continue to be limited by legislation so the solubilization of dust-Fe will be almost back to 'past' levels, with only small differences that are mainly due to biomass burning emissions. On the other hand, over North America, where air quality has improved since the 'past' due to the limitation of extensive coal combustion and will continue to improve due to the applied legislation, decreases in DFe/DIP are calculated from the 'past' to present and to the 'future'. However, these results have to be interpreted with caution and used only as a qualitative demonstration of the impact of acidity on the deposition of nutrients carried by dust, since they are associated with significant uncertainties in the aerosol acidity calculations, in particular due to the mixing state description of aerosols. In addition, as discussed earlier, changes in nutrients can favor certain species over others and so the overall functioning of ecosystems, while others can adapt to the new conditions (Guerinot 1994). The evaluation of such impacts requires the use of coupled land/atmosphere or land/air-sea exchange modeling and the consideration of nutrient ratios that deviate from Redfield stoichiometry.

Summary and recommendations
This review, based on a compilation of current experimental and modeling studies, has documented the importance of atmospheric processing via multiphase chemistry for the formation of nutrients or their transformation to forms that are readily bioavailable to ecosystems. This process needs to be further understood and considered when modeling atmospheric deposition of nutrients.
It has also shown that there is a clear consensus on the importance of the organic fraction of nutrients as an integral part of the atmospheric cycles of the main nutrients considered here (N, P, Fe), and that atmospheric aerosols are the key carriers of these components. Interestingly, bioaerosols have been found to make a significant contribution to atmospheric sources of N and P (Kanakidou et al 2012, indicating potentially significant interactions between terrestrial and marine ecosystems. However, further experimental and modeling studies are needed to provide accurate quantification of organic nutrient sources and impacts on ecosystems. In particular, measurements in remote locations are logistically difficult but invaluable for the documentation and understanding of atmospheric processing of aerosols. There is also the need for better characterization of organic nutrients present in atmospheric aerosols and their association with organic matter present in the ocean and the ecosystem response to the corresponding atmospheric inputs. Little is also known about the bioavailability of the organic material that penetrates to ecosystems via atmospheric deposition; it is expected that when there is a shortage of inorganic nutrients organisms will search for other forms of nutrients (organics) even if they might be less easy to assimilate. It is also expected that those which are able to adapt to the environmental changes will be favored over others in such a case. It has been suggested that changes in the relative composition of atmospherically deposited nutrients are driving human-induced limitations of the ecosystems to specific nutrients. These modifications are also favoring the development of certain species over others and impacting on the overall functioning of ecosystems (Guerinot 1994).
Furthermore, we have focused here on the abiotic solubilization of nutrients. However, it is worth mentioning the dual role of microbes in the atmosphere, which contribute to bioaerosols as carriers of nutrients (mainly N and P) but also as mediators increasing or reducing the bioavailability of nutrients in the environment (Mackey and Paytan 2009). The presence of microbes in soils has been used to explain the occurrence of bioavailable Fe and P, and soil microbes could therefore also have a potential for use in sustainable agricultural development (Alori et al 2017). However, to our knowledge, no work has been done on the presence of such microbes in atmospheric aerosols and their potential contribution to further solubilization of nutrients present in aerosols during atmospheric transport. Instead, the focus has been on the solubilization of trace elements under acidic conditions driven by abiotic chemical reactions forming acids and ligands in atmospheric aerosols.
Finally, it has been seen that although Redfield stoichiometry is commonly used as a first step in identifying the limiting nutrient for ecosystem development, the functioning of several species deviates from Redfield stoichiometry, in particular those that are able to adapt to nutrient stress conditions. Therefore there is a close link between community structure and the average relative proportions of nutrients in suspended and sinking particles in the ocean that requires a deeper understanding (Twining et al 2004). These impacts, including the contribution of organic nutrients to atmospheric deposition, need to be understood and taken into account in Earth system modeling with a new 'non-Redfield' approach to modeling biogeochemistry. and the European Union's Seventh Framework Programme (FP7/2007(FP7/ -2013