4.1 Alaska

Dust storms in southern Alaska were first reported at least a century ago [Tarr and Martin, 1913]. Most of our understanding of contemporary Alaskan dust has been gained from analysis of dust deposits, in the form of modern loess in central Alaska, or from satellite remote sensing. However, it was only in 2011 that the first publication described widespread dust storms occurring, often simultaneously, at many different locations spanning much of the Gulf of Alaska coastline [Crusius et al., 2011] (Figure 6). Dust storms occur in these locations in response to a fairly predictable set of phenomena described in more detail in Crusius et al. [2011]. The many glaciers that cover much of this very mountainous coastline melt during the summer, driving a maximum in river discharge at that time (Figure 7). This river discharge contains considerable quantities of fine glacial flour, part of the bedrock eroded as glaciers advance over the landscape. In the autumn, after most of the melting ceases, but before significant snowfall, extensive river floodplains are exposed that are covered in recently deposited glacial flour. High‐pressure systems over central Alaska, coupled with autumn cooling events, can drive strong northerly winds that can be channeled down the steep mountain valleys (katabatic winds), resuspending the fine glacial flour as dust and transporting it far over the ocean. The most prominent valley where this occurs is that of the Copper River, the single largest source of freshwater to the Gulf of Alaska. However, similar phenomena have been observed to occur at many sites along the roughly 1000 km of glacierized coastline. From 2011 to 2014, dust observations were carried out on Middleton Island (J. Crusius, personal communication, 2015), near the continental shelf break in the pathway of the most prominent dust plume [see Crusius et al., 2011] (Figure 8). Those observations confirm the events as summarized above, in addition to revealing strong interannual variability.

MODIS satellite images show that dust transported offshore over the Gulf of Alaska extends at least a few hundred kilometers over the ocean. Alaskan dust has been demonstrated as containing a high proportion of iron [Schroth et al., 2009], and Crusius et al. [2011] estimated that the quantity of bioavailable iron transported to this iron‐limited region of the north Pacific Ocean via dust is comparable to that transported offshore in coastal eddies [e.g., Xiu et al., 2011]. As yet there is insufficient evidence that these autumn dust events lead directly to phytoplankton blooms, most likely because light levels are low when they occur. However, the dust nonetheless contributes significantly to the available inventory of bioavailable iron in the Gulf of Alaska, and it is possible that it could remain in solution long enough to fuel phytoplankton blooms in the spring when bloom conditions improve. Dust is also deposited on land, where observations from loess deposits demonstrate that similar dust deposition has been occurring for millennia [Muhs et al., 2013, 2016]. Dust storms originating from the glacial river valleys of the Knik, Matanuska, and Susita Rivers in Alaska can affect air quality in the state capital Anchorage. The U.S. Environmental Protection Agency standard for air quality is 150 µg m³ of particles <10 µm diameter (PM₁₀) averaged over 24 h, and this standard was exceeded due to natural dust storms in 2001, 2003, 2007, 2009, and 2010 with 24 h average PM₁₀ concentrations >500 µg m⁻³ during the most severe events [Department of Environmental Conservation, 2012].

4.2 Canada

Contemporary dust events are relatively rare in Canada compared to other high‐latitude countries, with the majority of documented observations pertaining to events occurring in late winter through spring in the Prairie Provinces, particularly within a seasonally cold, dryland area referred to as Palliser's triangle in southern Saskatchewan and Alberta. In selected periglacial settings within Arctic Canada such as on glaciofluvial outwash and dry lake beds, small‐scale dust emission events are also known to occur. However, many reports relating to these events are either anecdotal or buried in the grey literature (e.g., aviation and air quality records). To date, the most well‐known and rigorous investigation of dust emission in a western Arctic setting within Canada was carried out by Nickling [1978] who directly measured the vertical flux of dust produced from outwash sediments within the Slims River valley in the Yukon Territories under strong off‐glacier winds. This seminal work provides the only detailed set of measurements of dust emission carried out in a proglacial setting within Canada. Occasional, serendipitous sightings and observations of dust plumes have been made by other researchers working in remote locations while studying other phenomena. For example, Church [1972] made some “indicative” measurements of dust storms on the Lewis River sandur in July 1965 during his study of Arctic fluvial processes. He noted that “considerable quantities of dust were found at 100 cm above the ground” (p. 64) and recorded dust covered snowpacks in the spring. The remote sensing record of dust in Canada is very limited, although occasional events have been detected.

Lacustrine, glaciofluvial, and aeolian (loess) deposits are extensive and well documented in maps of the surficial geology of Canada [Fulton, 1989], and it would be expected that dust may be emitted from these surface types and transported tens of kilometers up to as much as 500 km from their local source. Relatively warm, dry (Chinook) föhn winds that descend from the Rocky Mountains in late winter and early spring are strongly associated with dust emission from unprotected, tilled soils in western Canada. Cold, dense katabatic winds draining from upland ice caps also play a role in proglacial systems throughout northern Canada, particularly in the high and eastern Arctic. Topographic funneling through deep glacial valleys within fjord settings leads to substantial acceleration of such winds, which may exceed 28 m s⁻¹ (100 km hr⁻¹) in extreme situations. In terms of seasonality, dust emission from undisturbed sources is relatively uncommon in summer, except under conditions of severe drought, as the surfaces are usually either well protected by a vegetation cover (e.g., moss, lichen, and macrophytic plants) or submerged in the case of the nival melt affecting outwash plains. However, aeolian transport is reported to increase again through the fall months as water levels drop, the vegetation cover becomes dormant or dies off with heavy frosts, and wind speeds increase with large‐scale meteorological shifts associated with the change of season and the positions of the jet stream and the Icelandic and Aleutian low‐pressure systems. In the Canadian Arctic, aeolian transport is believed to peak at this time of year but may continue into winter in isolated areas where a snowpack is not well developed or absent altogether. Where the snowpack is present and is relatively dry and granular, blowing snow may expose large areas of bare unprotected surface to subsequent sublimation and entrainment, as well as abrasion under particle impact [Edlund and Woo, 1992]. Extensive niveo‐aeolian deposits observed in winter throughout Canada provide some of the best evidence we have of the coincident transport of mineral and snow particles within a given boundary layer flow [e.g., Lamoureux and Gilbert, 2004].

The surface characteristics and conditions described above suggest that dust storms should be frequently observed in Canada, but this is not the case (Figure 3). A very low population density, limited number of observing meteorological stations, and problems with the use of remote sensing data (section 5) may partially explain this. Another contributing factor is improvements in soil conservation practices which have reduced, if not completely eliminated, the occurrence of wind erosion events on tilled soils. The presence of vegetation, a snowpack, and cementing or crusting (e.g., by pore ice, salt, clay, or organic material) do suppress a lot of dust emission, but anthropogenic activities associated with agriculture, transport, construction, and mining can disrupt these suppressants and lead to localized dust storms.

Most dust is deposited locally (<30 km) on glacier surfaces, higher elevation valley slopes, upland permanent snowpacks, and on ice‐covered lakes, rivers, and fjords (ultimately ending up in lacustrine and marine sediments). Long‐range transport from sources in the Canadian Arctic has not been documented. Where there is significant accumulation of wind‐blown silt, the albedo of the surface is altered affecting the energy balance and melt rate of either the snow or ice pack. For example, an experiment by Edlund and Woo [1992] on Ellesmere Island showed that snowmelt could be accelerated by a week by a dust layer with a concentration of 1230 g m⁻². The finest particles (e.g., PM_2.5) sampled from snow and ice packs in the Canadian Arctic are associated with long‐range transport on a global scale, much of it believed to arrive from Asia in late winter through spring. In comparison, mineral dust particles deposited during the autumn season generally contain coarser particles, so that local sources are believed to play a more significant role at this time of year [Zdanowicz et al., 2000].

4.3 Greenland

Early expeditions to Greenland reported strong winds and dust storms in ice‐free areas prompting Hobbs [1931, 1942] to conclude that wind was likely to be more important than meltwater for the transportation of sediment in proglacial regions. One of the earliest comparisons of the geomorphological effectiveness of the wind between the Arctic and Tropical latitudes was also based on work in west Greenland. In this, Fristrup [1953] suggested that geomorphologically active dust storms are likely to be more frequent and longer lasting in the Arctic than in the Tropics with the effects enhanced by the abrasion of snow and ice. More recently, Dijkmans and Tornqvist [1991] described dust clouds on ice and vegetation‐free sandur plains near Kangerlussuaq, SW Greenland, reaching over 100 m high during winds of 14–18 m s⁻¹ which are not uncommon (Figure 9). To date, all records of dust storms in Greenland have been based on field observations from expeditions or dedicated field campaigns [e.g., Bullard and Austin, 2011]. Greenland dust storms remain undetected by remote sensing although deposition of locally derived dust on to the ice sheet has been observed [Wientjes et al., 2011].

Potential dust sources in Greenland are confined to ice‐free areas of the land mass. Although this accounts for ~19% (~400,000 km²) of the land mass, only a small fraction, primarily vegetation‐free active glacial outwash plains, is actually dust sources. Bullard and Austin [2011] found predominantly bimodal sands and gravels at the surface of the outwash plains with lag deposits that limit aeolian entrainment but identified a clear link between meltwater floods that deposit thick layers of fine sediment (<100 µm diameter) across the floodplain and seasonal dust storms. Figure 10 shows the seasonal variability of days when dust was recorded at Kangerlussuaq meteorological station (1942–2015) based on dust codes [O'Loingsigh et al., 2010]. For dust code 06 (widespread dust in suspension, away from station) and dust code 07 (dust or sand raised by wind at or near the station at time of observation) there is a peak in dust days in both spring and autumn. Events coded 09 (dust or sand storm within sight at time of observation or preceding hour) occur most frequently in the autumn. Sediment supply to the floodplains is closely coupled to the delivery of material during spring and fall flood events; however, the record of dust days suggests that all three types of dust event can occur year round. Hobbs [1931] suggested that strong winds may cause a high frequency of dust storms in winter in Greenland; however, the monthly variability in average maximum wind gusts suggests no strong seasonal variation for the Kangerlussuaq area (Figure 10).

In addition to the floodplains, loess deposits 0.5–1 m thick and covering 70 × 10⁶ km² around the western margin of the Greenland Ice Sheet actively accumulate aeolian sediments. Despite a vegetation cover >70% in some areas, these aeolian deposits are characterized by deflation hollows caused by local wind erosion [Dijkmans and Tornqvist, 1991]. These deflation hollows are more common closest to the ice sheet and thought to be caused primarily by katabatic winds [Heindel et al., 2015]. Ongoing wind erosion of fine soils (~40 µm) in the hollows may also contribute to local atmospheric dust.

The majority of dust generated in Greenland is likely to be deposited locally or regionally. The modern loess deposits near Kangerlussuaq (Søndre Strømfjord) extend about 80 km west of the ice margin. Some locally derived dust is thought to be transported iceward and may be deposited on the ice causing surface darkening as it changes the ice albedo [Wientjes et al., 2011]. Recently, there has been increased recognition of the possible impacts of dust inputs on the ecology of lakes in Arctic and alpine regions [Mladenov et al., 2011, 2012]. Anderson et al. [2016] suggest that aeolian deposition adds 50 mg of carbon m⁻² yr⁻¹ to soils and lakes near Kangerlussuaq. In southwest Greenland there are ~20,000 lakes but the effects of dust input to lacustrine sediments and processes in this region have not been explored in detail.

4.4 Iceland

The best studied high‐latitude dust area is Iceland. More than 15 research papers have been written focusing on the sources and impacts of wind erosion in the country (Figure 11). This is due to the prevalence of aeolian activity driven by climate, volcanic activity and glacial sediment supply, and also the impact of humans. Iceland was settled around AD 874, and land‐use practices and the introduction of grazing animals has caused severe vegetation depletion and soil erosion [Dugmore et al., 2009; Gísladóttir et al., 2011]. Evidence of widespread and sustained wind erosion can be found in historical records, such as sagas, annals, old farm surveys and place names, as well as from expeditions [Arnalds et al., 2001a].

Sandy deserts with active aeolian processes cover >20,000 km² (~20%) of Iceland and their distribution is closely associated with the island's ice caps, as well as with volcanic systems and eroded soils [Arnalds et al., 2001b; Arnalds, 2010]. Unlike many other high‐latitude dust source areas, Iceland has a good and relatively dense network of weather stations. Meteorological records (1949–2011) suggest that an average of 34 dust days per year occur in Iceland, but that the frequency increases substantially when dust haze and events caused by the resuspension of volcanic materials are taken into account [Dagsson‐Waldhauserova et al., 2014]. The most active source regions for dust storms are the sandur areas on the southern coast (Figure 12) and the area northeast of the largest ice cap Vatnajökull. The seasonal temporal pattern is driven primarily by sediment supply and winds in southern Iceland and snow cover in the north. Glacial meltwater on the southern coast is distributed across broad glacial outwash plains where peak discharge is typically in spring and highest suspended sediment loads are in April with a second weaker peak in September [Gíslason et al., 1997; Old et al., 2005]. This fine suspended sediment load is deposited as extensive silt drapes across the sandur from which it can easily be entrained by strong winds. Meltwater sediment supply can dramatically increase during catastrophic flood events (jökulhlaups) triggered by volcanic or glacial activity, which have been linked to increases in dust storm activity [Prospero et al., 2012]. Dust storms generally occur when weather is, or has been, dry and wind speed ranges from 5 to 10 m s⁻¹ depending on surface roughness (Figure 13) [Arnalds et al., 2001b; Gísladóttir et al., 2005]. Extraglacial snow cover is low in southern Iceland and not thought to be a major control on dust emissions. By comparison, snow cover is higher and lasts longer in the northeast of the country, increasing the required threshold wind velocity for dust entrainment and reducing the frequency of dust storms [Dagsson‐Waldhauserova et al., 2013]. The differences in the seasonal distribution of dust days in the south and northeast of Iceland are clearly seen in Figure 14. Although dust storms occur all year round in southern Iceland, dust storm frequency is highest during the spring, with a second, lower peak in the fall. In northeast Iceland, dust storms are prevalent in the summer months.

Icelandic dust storms are not only linked to the glacial system but can also be caused by the resuspension of widespread volcanic ash deposits associated with the active volcanic zone. The eruption at Eyjafjallajökull, 14 April to 20 May 2010, produced abundant particulate matter due to its explosive eruption style. Even after the volcanic activity ceased, high particulate matter (PM) concentrations were still measured on several occasions, due to resuspended ash (Figure 2). After the eruption ceased, values as high as 8000 µg m⁻³ (10 min average), and 900 µg m⁻³ (24 h average), were measured because of resuspension of freshly deposited fine ash. In Reykjavík, 125 km WNW of the volcano, the PM₁₀ concentration reached over 2000 µg m⁻³ (10 min) during a resuspended ash storm on 4 June 2010 [Thorsteinsson et al., 2012].

The impacts of dust storms in Iceland can be identified both at the dust source and off‐site. Local redistribution of soils and sediment results in coarser soil textures [Ólafsdóttir and Guðmundsson, 2002; Jackson et al., 2005; Dugmore et al., 2009; Gísladóttir et al., 2010] as soils lose their fine material and light particles such as soil organic carbon. Some material is transported offshore [Óskarsson et al., 2004; Gísladóttir et al., 2010], and some is deposited locally in soils or into sediment traps such as lakes [Gathorn‐Hardy et al., 2009; Geirsdóttir et al., 2009]. Dust deposition rates in Iceland range from 13–26 g m⁻² yr⁻¹ in the northwest to >250 g m⁻² yr⁻¹ along the south coast, through the central volcanic belt and in the northeast [Arnalds, 2010]. Dust can be transported over the ocean to the northeast, but most is transported south where it may contribute bioavailable iron to the Atlantic [Arnalds et al., 2014] or west‐north‐west toward the capital Reykjavík. Dust storms account for around one third of the air quality exceedance events (PM₁₀ concentration >50 µg m⁻³) in Reykjavík each year [Thorsteinsson et al., 2011] (Figure 15). They can occur at any time of year and may occur in combination with snow storms during the winter [Dagsson‐Waldhauserova et al., 2015; Liu et al., 2014].

4.5 Antarctica

Antarctica is predominantly ice covered with only 2% comprising ice‐free land. Most of the focus on dust in Antarctica has centered around ice core records, which clearly indicate that the continent receives dust from two key global dust sources. One of these sources is Australia which is the largest contributor of dust under the current climate [Li et al., 2008]. The second major dust source affecting Antarctica is another high‐latitude region, southern South America. Dust from this source, which includes Patagonia and Tierra del Fuego, dominates records in the Antarctic Peninsula and West Antarctica and accounts for up to 85% of dust in the Lake Vostok core [Li et al., 2008]. Although southern South America is not currently the dominant source, it is thought to have been more important than Australia during the Last Glacial Maximum [Albani et al., 2012]. Southern South American dust sources are discussed in section 4.7. Analysis of dust in coastal ice cores, e.g., Taylor Dome, suggests that present‐day Holocene dust originates from the ice‐free areas of Antarctica [Delmonte et al., 2013].

The ice‐free areas of Antarctica have recently been recognized as important local dust sources. Of these, the McMurdo Dry Valleys are the largest ice‐free area (4800 km²) [Doran et al., 2002] and the most notable dust source. The remoteness and harsh conditions of Antarctica mean that there have been few field studies of aeolian processes, and many of these have focused on dunes and granular ripples rather than dust storms [Lindsay, 1973; Selby et al., 1974; Speirs et al., 2008; Gillies et al., 2012]. Proglacial sandur and deflation lag surfaces characterize the lower parts of the dry valleys with aeolian sand dunes restricted to relatively confined areas. Some parts of the dry valleys may also have been occupied by proglacial lakes [Hendy et al., 2000], and moraines formed by previous ice advances are also present and contain fine sediments. Selby et al. [1974] noted that although the aeolian deposits of the Dry Valleys comprise coarse sands and gravels, finer sediments are present in the glacial outwash and could be blown iceward onto the nearby glaciers. Lancaster [2002] found that silt and clay deposition dominated on nearby glaciers and ice‐covered lakes (up to 95% of aeolian sediment flux) compared with the valley floors which had more aeolian sand deposition and received higher total aeolian fluxes.

Winds in the Dry Valleys are primarily topographically driven katabatic and föhn winds. Easterly onshore (up valley) winds are very persistent, but less frequent down valley winds are stronger [Clow et al., 1988; Doran et al., 2002]. The magnitude of the difference is seasonally variable as the strength of the up‐valley winds weakens during the winter [Lindsay, 1973]. Short duration field studies (a few years) have found aeolian activity in the Dry Valleys to be annually very variable making it difficult to draw reliable conclusions about typical annual rates of dust transport or deposition [Lancaster, 2002; Malin, 1992; Gillies et al., 2013]. Snowpits offer a greater timespan of contemporary dust deposition and using a 35 year chronology (1965–2000) Ayling and McGowan [2006] suggested that aeolian transport is greatest in the winter when westerly föhn winds dominate. Atkins and Dunbar [2009] and Chewings et al. [2014] measured deposition of aeolian sediments on sea ice in McMurdo Sound (Figure 4) and found that substantial quantities of material are transported offshore each year (0.2–55 g m⁻²yr⁻¹). These sediments are deposited as a plume containing increasingly fine particles with distance downwind. Sand‐sized aeolian sediments are typically transported <5 km offshore, but silts and clays were found >100 km from source and when the sea ice melts these contribute to both ocean floor sedimentation and marine nutrients. The source of these aeolian plumes is the McMurdo Ice Shelf supraglacial debris bands rather than the McMurdo Dry Valleys. Chewings et al. [2014] suggest that the lack of input from the Dry Valleys is due to the well‐developed deflation lag that protects the surface sediments in the valleys and topographic barriers between the Valleys and the ice shelf.

Although total local dust emission from Antarctica may currently be low compared with some other high‐latitude dust sources, it is predicted to increase in future as the sediments have the potential to produce dust under slightly warmer and windier conditions [Bhattachan et al., 2015; Hedding et al., 2015]. In addition, any shift from marine‐terminating to land‐terminating glaciers will increase the area of sediment‐rich proglacial floodplains [Bullard, 2013].

4.6 New Zealand

Studies of contemporary dust processes in New Zealand are relatively few but have nevertheless provided a good overall understanding of both the primary sources of deflated sediment and the mechanisms driving its emission. On the country's South Island, the floodplains of extensive braided river systems are well established as a supply of fine sediment that is prone to deflation [McGowan, 1997; Eden and Hammond, 2003]. Exposed bars and inactive channels in these alluvial systems commonly act as emissive surfaces, and several field studies have quantified rates of dust deposition downwind of braided floodplains [e.g., Cox et al., 1973 cited by Marx and McGowan, 2005; McGowan et al., 1996; Eger et al., 2012]. In the semiarid rain shadow region east of the Southern Alps, degraded tussock grasslands are also recognized as a prominent source area for dust, with land erodibility within the intermontane basins of such high‐country rangelands (e.g., Mackenzie Basin) having been exacerbated by anthropogenic activity [McGowan and Ledgard, 2005]. Vegetation removal associated with agriculture following the arrival of European pastoral practices and rabbit infestations led to peak levels of land degradation occurring in the 1940–1950s [McGowan and Ledgard, 2005]. In the rangelands, wind erosion has left remnant pedestals, perhaps partly related to surface‐disrupting frost action [Basher and Webb, 1997; McGowan and Ledgard, 2005].

While the rain shadow effect of the Southern Alps range promotes dust emission in drier conditions on its leeside, the uplift of dust from alluvial floodplains and its subsequent local deposition have also been reported in superhumid areas on the South Island west coast, where the process is linked to active loess formation [Eden and Hammond, 2003; Marx and McGowan, 2005; Eger et al., 2012]. Approximately 19% of the South Island has been assessed as affected by wind erosion to some degree [Salter, 1984]. For instance, wind erosion of cultivated arable lands on the Canterbury Plain is well recognized, and rates of bulk soil loss have been estimated; however, specific fluxes of dust‐sized material in suspension are not well quantified for these agricultural systems [Basher and Painter, 1997].

An important consideration in understanding New Zealand's dust sources is its location within a major transport pathway of desert dust originating from Australia [e.g., McGowan et al., 2001; Eden and Hammond, 2003]. For suspended sediment collected in New Zealand, approaches that can determine the provenance of the dust have proved especially useful in determining the component of dust from local versus long distance sources [Marx et al., 2014]. The geochemical modeling of collected aerosol, referenced to a database of trace elements, has been used to estimate the proportion of sampled dust coming from New Zealand or remote Australian sources. Using this approach, dust deposition episodes in which high proportions of New Zealand material were detected have been tied to braided river systems as probable sources [Lavin et al., 2012; Marx et al., 2014].

Episodes of blowing dust from New Zealand source surfaces are commonly associated with particular meteorological conditions and are typically the product of infrequent high‐magnitude storms. In the austral winter, entrainment of dust on the South Island is typically associated with a strong anticyclonic presence in the northern Tasman Sea and the generation of strong south westerly winds [Marx and McGowan, 2005; Marx et al., 2014]. In spring and autumn, westerlies of a circumpolar nature are strongest over southern New Zealand and are capable of producing localized dust storms [McGowan et al., 1996]. Topographically enhanced gradient winds also play a significant role in generating dust activity in New Zealand [McGowan et al., 1996]. To the lee of the Southern Alps divide, föhn nor'westers occur in intermontane basin areas, with these winds having significant dust‐lifting potential due to their combined low humidity and warm, strongly gusting flows that often exceed 30 m s⁻¹ [McGowan et al., 1996]. With respect to the downwind pathways for dust entrained in New Zealand, very little research has been undertaken. The first paper to focus on this was published in 2015 by Neff and Bertler who used forward trajectory modeling to examine Southern Hemisphere dust trajectories from Australia, New Zealand, Patagonia, and southern Africa (Figure 16). They found that unquantified dust emissions from New Zealand would be transported southward with an efficiency comparable to that from Patagonia and that these high latitude sources were far more likely to impact the Southern Ocean and Antarctica than lower latitude dust sources in Australia and southern Africa (Figure 17). Neff and Bertler [2015] also suggested that even low dust emissions from New Zealand (~10 Tg yr⁻¹) could contribute as much as 8.5% total deposition of dust over the Southern Ocean and 13.7% over Antarctica.

4.7 Patagonia, Southern South America

Patagonia, including Tierra del Fuego, is a large territory of more than 900,000 km², located between 39° and 55°S in southern South America. The topography is dominated to the west and south by the Andean cordillera and by dissected plateaux and low plains to the east. The regional climate of the area is strongly affected by the southern westerly winds (SWWs) which, coupled with the Andean cordillera, produce a strong, east‐west gradient with annual precipitation of ~4000–7000 mm in the west to ~200 mm yr⁻¹ in the east concentrated in the austral autumn and winter seasons [Garreaud, 2007]. To the east of the Andes, scrub‐grassland vegetation closely reflects the semiarid conditions. The sparse vegetation cover and strong surface winds provide appropriate conditions for dust emissions. In general, dust activity is highest during the drier summer months, but winter and autumn dust events have also been reported [Gaiero et al., 2003].

The Patagonian region has been identified as a dominant source of dust in the Southern Hemisphere during glacial periods [Basile et al., 1997; Delmonte et al., 2008]. Successive maritime ice sheets, fed by the SWWs and high precipitation have, in response to climate forcing, waxed and waned along the southern Andes. Each glaciation has left a legacy of glacial landforms, ice‐scoured troughs, and extensive sandur plains extending across the Patagonian steppe [Clapperton, 1993]. During the LGM mineral dust concentrations 20–50 times higher than the present have been identified in Antarctic ice cores and radiogenic isotopic signatures of the dust indicate Patagonian sources [Lambert et al., 2008]. Atmospheric transport models coupled to global circulation models have been unable to explain the pattern of dust peaks in the ice cores [Mahowald et al., 1999]. As a possible explanation, the Antarctic dust signal may have been mediated by glacial geomorphological mechanisms leading to a “switching off” of dust emissions as the Patagonian ice sheets waned at the termination of the last glaciation [Sugden et al., 2009].

In contrast to the high dust flux of the glacial period the volumes of Holocene Patagonian dust emissions are smaller but probably continue to be a significant input into the HNLC southern ocean ecosystems and the Antarctic ice sheets (Figure 16) [Erickson et al., 2003; Wolff et al., 2006; Gassó and Stein, 2007]. The present Patagonian dust sources are from reworked loess, alluvial fans, large dessicated lake beds (Figure 18), and particles produced from explosive volcanism. Emissions and persistent major dust sources have been attributed to sustained meteorological and climatological conditions [Gaiero et al., 2013]. It is also likely that sustained over grazing of the scrub grassland since European settlement of the region has increased dust emissions [McConnell et al., 2007; Gaitan et al., 2009].

High‐latitude Patagonian sources south of ~39°S are less well understood than those farther north on the continent due to the region being cloudier and the dust events being more sporadic [Gassó et al., 2010]. This has made it difficult to identify dust source areas and to track dust events except by proxy through the analysis of dust deposition in Antarctica [Bory et al., 2010]. The monitoring of air quality and sky turbidity at meteorological stations along the Argentine south Atlantic coast provides the longest records of surface dust measurements [Mahowald et al., 2006; Gassó et al., 2010]. There is a strong relationship between the austral summer drying and increased intensity of the SWWs, probably exacerbated by strong katabatic winds across the Patagonian steppe, leading to increased dust activity. Estimates that ~30 × 10⁶ t yr⁻¹ of Patagonian dust are supplied to the South Atlantic shelf [Gaiero et al., 2003] are considered by Simonella et al. [2015] to be too low.

Recent improvements in the detection capabilities of satellites have significantly advanced our understanding of the nature and extent of point sources of dust from Fuego‐Patagonia and the distance the dust may travel—as much as ~1800 km from source [Gassó and Stein, 2007; Gassó et al., 2010]. Combined observation‐modeling studies have improved our understanding of the transport pathways, seasonality, and efficiency of the different potential dust sources [Genthon, 1992; Li et al., 2008, 2010a, 2010b; Mahowald et al., 1999, 2006; Albani et al., 2012]. Satellite data have also verified Eulerian model outputs of aerosol distribution and identified dust sources from topographic depressions [Li et al., 2008; Johnson et al., 2010]. However, as with other high‐latitude dust sources, satellite observations and modeling of Patagonian dust activity face a number of challenges (section 5.2).

4.8 Other High‐Latitude Dust Sources

In addition to the seven main regional high‐latitude dust sources considered above, there are other less well studied or smaller dust sources primarily in the Northern Hemisphere ≥50°N. Dörnbrack et al. [2010] describe a dust storm on the Norwegian archipelago of Svalbard (78°N) that occurred in May 2004. The dust sources comprised both natural sediments deflated from dry riverbeds and anthropogenic sources at active coal mines. Dust storms are not frequent on the islands but primarily occur during the autumn when river levels are low exposing dry, snow‐free sediments; the May event followed dry, sunny conditions throughout April and May. The geochemistry of aerosols in Svalbard suggests not only local dust sources but also that dust is transported to the islands from Greenland and Iceland [Moroni et al., 2015].

Numerous locations of potential dust sources exist in central Asia and Siberia [Engelstaedter et al., 2003]. This mostly arid continental polar, or humid, cool climate is characterized by tundra or boreal forest. Nevertheless, locally sourced dust storms have been reported especially in those areas affected by desertification and land degradation. The frequency of dust storms in southern Siberia varies both spatially and temporally. For example, in 1988 there were 2 dust storms recorded in the Nazarovskaya area (forest steppe) but 17 on the Koibalskaya steppe (~200 km south). High annual frequencies of storms are associated with droughts (e.g., early 1920s, 1950s, 1970s, and early 1980s) and exacerbated by agricultural activities. Bazhenova and Tyumentseva [2015] suggested that the average rate of deflation in subarid southern Siberia is 0.1–2.5 mm yr⁻¹ with the wind‐blown material dominated by silts and clays. The rate of deflation and accumulation of sediments, however, are strongly linked to local topography and vegetation cover. Farther east, dust storms caused by the resuspension of volcanic ash have been observed on the Kamchatka peninsula [Gimsey et al., 1998].

5.1 Characterizing High‐Latitude Dust Sources

Table 2 summarizes the key characteristics of the main regional high‐latitude dust sources examined here. It highlights the patchiness of our current understanding of dust emissions in these areas and also some common elements. Similar to subtropical dust sources, high‐latitude dust sources contain fine‐grained, wind erodible sediment in lows such as riverbeds, valley floors, and glacial outwash plains. Other common sources in high latitudes are reworked loess deposits which are less significant in the subtropics due to the global spatial distribution of loess [Pécsi, 1990; Muhs, 2013].

There appears to be no typical annual rainfall associated with high‐latitude dust sources. Some, such as the Dry Valleys of Antarctica, receive very little precipitation; Keys [1980] suggested 100 mm rainfall equivalent per year, but a more recent study by Fountain et al. [2009] measured spatially very variable precipitation of 3–50 mm rainfall equivalent per year (1995–2006). Other high‐latitude dust regions such as New Zealand (~600 mm yr⁻¹) and southern Iceland (1500–2500 mm yr⁻¹) receive high quantities of precipitation. Rather than total precipitation, it is more likely that the balance between precipitation and evaporation, combined with the distribution of snow cover and wind speed, are more important characteristics of high‐latitude dust sources.

The overall relationship among sediment supply, sediment availability, and transport capacity controls the magnitude, frequency, and intensity of dust storms [Bullard, 2013]. Distinct seasonalities to dust storm frequency in subtropical dust sources are primarily controlled by seasonal variations in wind regimes (both wind speed and direction). The controls on dust storm seasonality in high latitudes are arguably more complex due to the very close coupling between sediment supply and dust emissions with often a limited time period when sediments are available and wind regimes conducive to entrainment from dust sources. In many high‐latitude dust areas, air temperature affects glacial meltwater discharge and consequently suspended sediment yield. Interannual variation in meltwater suspended sediment yield is very variable at high latitudes [e.g., Hodgkins et al., 2003] and has a nonlinear relationship with discharge. For the same discharge, suspended sediment yield can be higher during the spring and lower in the autumn due to seasonal exhaustion of sediment supplies [McDonald and Lamoureux, 2009]. Discharge is closely linked to the magnitude of seasonal snowmelt even in catchments dominated by ice sheets [Chu et al., 2009; Lawler et al., 2003]. As a consequence of the way in which these different variables combine, seasonality of dust emissions can vary from source to source (Table 2). In some areas, early spring season meltwater containing high concentrations of suspended sediment can replenish floodplains. Given a lack of, or limited snow cover, these sediments can be deflated to produce spring dust storms as in Iceland, New Zealand, and Alaska. Where snow cover persists for longer and/or a second pulse of sediments is delivered to the floodplain at the end of the summer, the main dust storm season (e.g., Alaska) or a second weaker dust season (e.g., New Zealand and Canada) occurs during the autumn months. In Alaska, snow accumulation in winter and temperatures in the summer are good predictors of dust activity in the autumn. Strong winds may serve to increase dust storm magnitude and/or frequency during winter months in Canada and Greenland.

Several factors limit field studies of dust emission in high‐latitude regions. These factors include remoteness, very low temperatures, snow and ice cover, and lack of daylight during winter months and mean that there have been very few year‐round field studies of dust emissions in these regions. As with many subtropical dust source areas, the low densities of population (typically <0.5 km² outside cities) and of meteorological stations mean that many small‐scale ephemeral events go unobserved. In some cases this may be due to the vastness of the uninhabited areas of the northern Arctic and Antarctica; in other cases it may be that dust was detected but not attributed to local sources due to lack of awareness of dust originating from high latitudes. For example, while reexamining data collected in North Canada during the 1970s and 1980s [Barrie, 1986; Barrie and Barrie, 1990] for a study on intercontinental aerosol transport, Fan [2013] noted an unexplained increase in dust observations at Alert, Canada (82°30′05″N, 62°20′20″W) during the fall months. It is likely that the fall peak at Alert and at other locations in Alaska was caused by local dust entrainment [Polissar et al., 1998; Stone et al., 2014].

5.2 Challenges of Satellite Remote Sensing of High‐Latitude Dust

A sparse population and low density of meteorological stations are also characteristics of many subtropical dust source areas, but here this limitation has largely been overcome through the use of satellite remote sensing data. These have provided a rich source of information on the spatial distribution and temporal cycles of dust storms at regional to near‐global [Prospero et al., 2002; Washington et al., 2003; Ginoux et al., 2012] and subcatchment scales [e.g., Bullard et al., 2008; Lee et al., 2009]. Although there are limitations with some of these satellite‐derived data sets, such as difficulty detecting dust plumes over some surface types [Baddock et al., 2009], they have provided genuine insights into subtropical dust storm dynamics and dust pathways.

Satellite detection of dust is a useful tool for assessing the extent of a dust event and source location, but its application is particularly challenging in the high latitudes. First, there are long periods of darkness above the polar circles (66°33′N/S) in late fall and winter that limit the use of passive remote sensing, which is reliant on reflected sunlight, as a year‐round tool. Second, cloud cover is far greater at high latitudes than in the subtropics, especially during the nonwinter months, such that the presence of cloud can obscure dust plumes and confound the performance of dust retrieval algorithms. For example, in the Northern Hemisphere July cloud cover at high latitudes is 70–80% compared with 10–20% in the subtropics; in the Southern Hemisphere cloud cover is 70–80% during the summer (January) compared with 30–40% January cloud cover in the subtropics [NASA, 2014]. In the subtropics dust activity is known to be marginally underrepresented in satellite retrievals when events occur during the night or under cloudy conditions [Prospero et al., 2002; Ginoux et al., 2012], but the combined effect of long dark winters and frequent cloud cover during the summer magnifies these problems in high latitudes. Even in high‐latitude areas that receive several hours of daylight during the winter months, for example, Patagonia, the low‐pressure systems that trigger the intense winds associated with dust uplift in this region also bring abundant cloudiness, making consistent remote sensing of dust activity difficult.

Active remote sensing approaches using depolarization such as Cloud‐Aerosol Lidar with Orthogonal Polarization (CALIOP) [Winker et al., 2003] or Cloud and Aerosol Transport System (CATS) [McGill et al., 2015] are highly effective as they not only identify whether sensed aerosols are likely to be dust or not but also detect dust as a function of height during daytime and nighttime (Figure 19). However, the period between revisits of these sensors over the same dust‐bearing area is lengthy, and given that the chance of cloud obstruction is very high for each overpass, effective capture of high‐latitude dust events is serendipitous. This means that characteristics like dust plume height are only infrequently available. Multiangle Imaging Spectroradiometer (MISR) [Kahn and Gaitley, 2015], Advanced Along Track Scanning Radiometer [Curier et al., 2009], Polarization and Directionality of the Earth's Reflectances (POLDER) [Hasekamp et al., 2011], or ultraviolet range observations from Ozone Mapping Instrument (OMI) [Levelt et al., 2006] have been proven useful at identifying dust at low latitudes but have some deficiencies that become apparent at high latitudes. For example, sensors such as OMI and POLDER have large ground pixel resolution and suffer from pervasive cloud contamination; MISR and the lidars CATS and CALIOP have a narrow instrument swath which means effective revisits to the same area are infrequent.

Geostationary satellites can overcome some of the difficulties associated with a low number of repeat observations. The GOES satellite series has good coverage of high latitudes to approximately 60°N and S—these satellites are stationed over the equator at 36,000 km distance while polar satellites fly at 700 km. GOES is very effective for monitoring dust activity in Patagonia, and there is good coverage of the Gulf of Alaska, Southern Greenland, and Iceland particularly in the summer months. However, although the spatial coverage is adequate, GOES sensors are not very good at quantitative retrievals of dust due to the lack of adequate wavelengths for performing such a task. New generations of satellites should offer additional capabilities for studying high‐latitude dust. For example, the recently deployed satellites Himawari‐8 and the Second Generation Meteosat have sensors with high spatial resolution, the appropriate channels for dust detection (especially over the ocean) and coverage of different high‐latitude regions, potentially overcoming some of the limitations noted for the polar sensors.

Clear skies are not the only prerequisite for dust detection. It is also important to establish whether observations made are of mineral dust or other aerosols. Fortunately, most high‐latitude dust is produced in clean environments where the dominant aerosols are background continental, marine, and occasionally smoke aerosols. This is not the case in many low‐latitude dust source regions such as central and SE Asia or the Sahel in Africa. Strong absorption of blue light means that true color satellite imagery can be effective in identifying dust, especially over dark oceans (e.g., Figure 12). Over high albedo surfaces such as snow and ice, however, visual identification of suspended dust that relies on visible wavelengths is less effective. This condition is analogous to some degree with the difficulties of identifying dust over bright desert surfaces [Baddock et al., 2009]. Some methods for enhancing the appearance of dust in satellite imagery take advantage of the thermal contrast between a warmer land surface and the cooler temperature of dust plumes that are in suspension. In high latitudes with generally colder land surfaces, there is less of a contrast between the surface and elevated dust, however, and certain enhancement approaches found to be effective in warm desert environments are less useful over colder lands [Miller, 2003].

Postprocessed satellite data products are now available globally in many forms, and for aerosol studies, estimates of the total concentration of dust in the atmospheric column as expressed by the aerosol optical depth (AOD) are useful for quantifying the distribution of dust. Climatologies of the spatial and temporal variability of dust based on AOD have been successfully developed for subtropical desert regions and have elucidated many large‐scale controls on dust emission [e.g., Ginoux et al., 2012] and transport pathways [Kaufman et al., 2005]. Unfortunately, the application of products like AOD is less straightforward in high latitudes. In many instances, where dust may well be apparent in the visible images captured by a satellite, retrievals of AOD might be lacking for the same scene. Many of the conditions required by automatic detection algorithms to successfully identify aerosols and return accurate AOD (e.g., degree of cloudiness in a pixel, viewing angles, homogenous dark surface, moderate to high aerosol concentrations, and number of views per day) are not fulfilled poleward of 45–50° latitude, which reduces the extent and frequency of AOD retrieval in those regions.

There are also a number of artifacts in high‐latitude remote sensing of dust that arise from the most commonly used satellites and sensors aligned on polar‐orbiting paths [e.g., MODIS sensor on board the Terra and Aqua spacecraft). In low latitudes, the issue of satellite overpass times leading to failures in the detection of dust events is well known [Schepanski et al., 2009]. The inclination of the Terra and Aqua satellites (one is ascending and the other is descending) results in frequent but grouped overpass timings at high latitudes. This can mean that there is an increased chance of missing dust events in these environments. The Terra and Aqua imagery used by Thorsteinsson et al. [2011] over Iceland was taken at 1330 and 1350 UTC respectively, whereas in equatorial regions the separation of the two MODIS overflights is approximately 3 h. Such short observation intervals at high latitudes also mean that the timespan over which the evolution of dust plumes can be monitored using differently timed images is reduced. In summer seasons, however, the convergence of polar‐orbiting satellite tracks at high latitudes leads to the potential for additional satellite coverage and imagery for a given dust event day. For example, north of 58°, almost any given sector is observed twice by MODIS bringing the number of daytime observations to four. This increases the chance of observing a clear‐sky pixel or permitting the study of the evolution of a dust cloud. Satellite data algorithms that have been developed for aerosol and dust detection are typically optimized for performance in the global dust belt associated with low latitudes. Due to the satellite geometry of high latitudes, sensor viewing angles and off‐nadir pixel resolution can be increased, especially at the extremity of scenes, and this has effects on dust retrievals and spatial resolution. For assessments of high‐latitude dust based on AOD, contributions from maritime aerosols (e.g., sea salt) can be confounding, as values around 0.06 can be attributed to nonaeolian dust [Kaufman et al., 2005]. While less of an issue for source detection and AOD quantification over inland, moderate to high‐intensity dust sources at low latitudes, proportionally, this error is more significant in less intense, commonly coastal, high‐latitude source areas.

These issues associated with high‐latitude remote sensing mean that there has been an absence of systematic high‐latitude dust studies or remote sensing‐based climatologies of dust. As noted by Tomasi et al. [2015], until more sophisticated aerosol detection algorithms are implemented and new instruments are deployed with features that overcome the problems mentioned above, systematic remote sensing of aerosols in the polar regions in general and in particular aerosol identification (in this case dust) is not currently possible. While not a dedicated high‐latitude remote sensing dust monitoring study, Kaufman et al. [2005] is one of the few studies that presents multiyear AOD from MODIS as far north as between 50° and 60°, revealing peaks in this value during April for the region 10°–20°W. This is an ocean region known to be under a transport pathway for dust sourced from southern Iceland, and the peak agrees with known timing of dust activity (section 4.4).

5.3 Quantifying High‐Latitude Dust Emissions

The entrainment, transport, deposition, and impacts of dust in the earth‐atmosphere‐ocean system depend on surface fluxes, i.e., the exchange of energy, momentum, and matter among these components. These include both material fluxes (e.g., moisture and aerosols) and turbulent fluxes (e.g., wind stresses and heat fluxes). The frequency, accuracy, and spatial resolution of measurements of these fluxes have to be appropriate to the spatial and temporal scale of the component of the dust cycle under investigation [Shao et al., 2011].

As discussed in section 3, a number of environmental drivers of dust entrainment processes at high latitudes do not occur, or occur to a far lesser extent, in the subtropics. Consequently, an understanding of physical processes operating at low latitudes may not fully explain those observed at high latitudes. It is therefore important to undertake measurements of high‐latitude dust entrainment processes and fluxes in the field as well as in appropriately calibrated wind tunnels [McKenna Neuman, 2004]. Unlike subtropical dust source regions, those in high latitudes often have very low winter temperatures, the presence of ice and deep, often drifting snow, and extremely high wind speeds. In order to capture year‐round in situ data quantifying meteorological conditions and sediment fluxes, it is necessary to use instruments that will function during long periods of low temperatures and polar darkness, can withstand icing, and are reliable in locations often very far from support services [Palecki and Groisman, 2011; Bourassa et al., 2013]. The harsh conditions during high‐latitude winters can also pose serious challenges to researchers [Kadir et al., 2013] making it preferable to use self‐logging instruments that can be installed at site and downloaded remotely or manually after extended intervals rather than requiring daily servicing.

Most instruments used in dust research are designed for, and have only been tested in, temperate and subtropical conditions. For example, there are numerous different types of dust trap used to sample wind‐blown particles; these can be categorized as either active or passive samplers. An active sampler is furnished with a pumping device to maintain a constant flow of air through the intake. This means that it requires a reliable and continuous source of power. Passive samplers depend on the local wind to drive air through the device. As passive samplers are cheap and require no power supply, they are widely used. One of the most common samplers for field measurements is the BSNE (Big Spring Number Eight) trap [Fryrear, 1986]. Wind tunnel experiments indicate that the BSNE has a sediment trapping efficiency of 40% [Goossens and Offer, 2000]. While the trapping efficiency varies with particle size [Shao et al., 1993], a genuine advantage over other samplers is that the efficiency of the BSNE is invariant with wind speed as tested at speeds from 1 to 5 m s⁻¹ [Goossens and Offer, 2000]. There has been no study into how the efficiency of this trap (or other dust samplers) may vary at higher wind speeds such as those more typical of many high latitudes. Nonetheless, the BSNE has been used for field studies at high latitudes in Iceland [Arnalds et al., 2012] and in Greenland by Bullard and Austin [2011] who added a rain/snow hood to the original design [Shao et al., 1993]. The disadvantage of the BSNE and other passive traps is that the temporal resolution of the data usually depends on manual emptying of the trap which does not make it appropriate for long‐term studies where researchers are not on site. To address this, in Antarctica, Gillies et al. [2012, 2013] successfully deployed for 1 year a modified BSNE‐style trap with the capability to open and close the sampling orifice and log the elapsed time, thus allowing a time‐resolved flux to be calculated.

Active samplers have been used for collecting aerosols, including dust, at some remote high‐latitude sites. Most of these deployments have been associated with short field campaigns and so provide discontinuous data [e.g., Dibb, 2007]. In order to obtain sufficient sample mass for detailed analysis, sample integration times are typically as long as 3–4 days [Colin et al., 1997; Hagler et al., 2007] which makes it challenging to differentiate the precise timing of dust events. More recently, VanCuren et al. [2012] developed a bespoke active sampler that they successfully used to capture a continuous record of aerosols at Summit, Greenland (72°N, 38°W). The sampler was configured to run unattended for 48 weeks continuously collecting particles in 8 size bins (from 0.09 µm to 10 µm aerodynamic diameter). Due to mechanical difficulties, only 170 days of data were captured but included both summer and winter seasons. This instrument demonstrates potential for future year‐round high temporal resolution sampling at high latitudes [VanCuren et al., 2012].

Regional‐scale estimates of fluxes are often inferred using measurements from instrument networks, for example, from meteorological stations (as in Figure 3) or ground‐based aerosol sensor networks such as Aerosol Robotic Network (AERONET) and SKYNET [Tomasi et al., 2015]. Such instruments must be accurate and reliable to ensure quality and completeness of data and must also be robust enough to understand the rigors of harsh winters at high latitudes. At high latitudes, although the instruments themselves can often operate in extreme conditions (e.g.,−49°C recorded in Barrow, Alaska, in 2006), they may require custom engineering to mount and power the instruments and despite combined power systems (solar, wind, and cold weather‐tolerant batteries) there can still be considerable challenges of ensuring data continuity during long, cold, dark winters [Palecki and Groisman, 2011]. Other challenges of autonomous instrument networks include drifting snow that can make year‐round observations of near‐surface measurements difficult. Improvements in telemetry do now permit remote downloading of data from isolated stations but this can be expensive.

Notwithstanding these challenges, some regional networks of stations do exist and provide data that can be used to quantify high‐latitude dust to some extent. One example is the world meteorological station network that records visibility, as utilized to produce Figure 3. Meteorological stations often record dust codes that indicate the timing and severity of dust events which have been combined with visibility information to estimate dust storm frequency and intensity at low latitudes [e.g., Middleton, 1984; McTainsh et al., 1990; Lim and Chun, 2006]. They have been used here to produce Figures 10 and 14. There are a number of limitations to the use of dust codes [O'Loingsigh et al., 2010], but the widespread distribution, and often long records, held by meteorological stations may outweigh these. Meteorological data are available from local stations and from online databases such as that hosted by the National Oceanic and Atmospheric Administration (https//www.ncdc.noaa.gov). Other online data sets are available via EBAS (http://ebas.nilu.no) and the NOAA Earth System Research Laboratory (www.esrl.noaa/gov) both of which include aerosol data. The challenges with obtaining regional estimates of dust emissions from these data sets include reconciling the differences between different instruments and observers and obtaining sufficiently detailed information to be able to differentiate locally derived mineral dust from other aerosols such as far‐traveled dust, sea‐salt, anthropogenic aerosols (e.g., black carbon) and Arctic haze.

5.4 High‐Latitude Dust in Global Dust Models

Most of our large‐scale understanding of the dust cycle comes from global dust models that include dust emission, transport, and deposition schemes based on parameterization of processes observed in field studies and tested against natural dust archives such as marine and ice cores and terrestrial sedimentary records [e.g., Albani et al., 2015]. The basic structure of global dust models is via a series of horizontal cells at resolutions of ~0.5° to 5° and 20–60 vertical layers with a dust emission scheme that simulates dust aerosols in various size ranges [Huneeus et al., 2011]. Global dust models are designed to quantify global, regional, and seasonal variations in dust loads. The estimated total contemporary global dust load is typically 1000–2000 Tg yr⁻¹. Variations in this value are often attributed to how sources and sediments are parameterized [e.g., Uno et al., 2006; Yin et al., 2007] and assumptions about the dust emission and deposition schemes [Textor et al., 2007; Huneeus et al., 2011]. The importance of surface conditions for model performance is demonstrated by the improvements in earlier modeling outputs that resulted from the inclusion of preferential source areas in descriptions of the land surface [Ginoux et al., 2001; Zender et al., 2003].

Until recently, global dust models did not include glacigenic dust sources, including those at high latitudes, even for simulating aerosol concentrations during dustier climate periods such as the LGM. During the LGM ice covered around 25% of Earth's land area and potential dust source areas were considerably more extensive in both the Northern Hemisphere, where they were generally associated with maximum ice limits [Mahowald et al., 1999], and the Southern Hemisphere where glacigenic dust sources were more prevalent in South America and greater aridity promoted dust emissions in Australia and southern Africa [Hall, 2004]. Without explicitly taking into account glacigenic sources, Mahowald et al. [2006] suggest that the total global dust source area at the LGM was up to 35% larger than at present (based on vegetation cover and CO₂ fertilization effects) with total dust loadings up to 60 Tg m⁻². When selected glacigenic dust sources in Europe, Siberia, North America, and the Pampas region of South America are factored in, modeled global dust loadings for the LGM are predicted to have been 70–80 Tg m⁻² demonstrating their importance to the overall budget.

Under contemporary conditions, models suggest that the dust emission rates from South America are 35–45 Tg yr⁻¹ [Zender et al., 2003; Tanaka and Chiba, 2006], but this estimate does not differentiate high latitude (i.e., southern South America ≥40°S) from more northern sources. The compilation in Table 2 suggests that high‐latitude dust emissions from Patagonia are ~30 Tg yr⁻¹. Icelandic dust emissions are estimated at 35 ± 5 Tg yr⁻¹, with a minimum annual emission rate from Alaska of >0.06 × 10⁶ Tg yr⁻¹. There are no estimates of dust emissions for the other high‐latitude dust sources, but conservatively, we suggest that in total, globally, they might contribute at least 80–100 Tg yr⁻¹ to the global dust cycle. This estimate is considerably higher than most model estimates of emission rates from North America (excluding Alaska and Canada; 2–53 Tg yr⁻¹), a similar order of magnitude to that from southern Africa (63 Tg yr⁻¹) and within the range suggested for Australia (37–148 Tg yr⁻¹) [Bullard, 2011; Shao et al., 2011]. Calculations of global dust emissions are very variable, but estimates of the amount entering the atmosphere each year converge around 2000 Tg [Bullard, 2013] which suggests that high‐latitude sources contribute approximately 5% of the global dust budget.

There are a number of challenges to incorporating contemporary high‐latitude dust emissions into global dust models. One of the most important is the scale of the models as their typical degree‐scale grid resolution renders them too crude for smaller‐scale enquiries. Unlike subtropical dust sources which can be extensive, many high‐latitude dust sources are relatively small and discrete and therefore below the spatial scale that can realistically be parameterized within the models. Equally important is the challenge of modeling wind fields at appropriate scales. Field studies suggest that katabatic winds are crucial for aeolian entrainment in some high‐latitude regions. Katabatic winds are rarely incorporated into global models because the grid size prevents the simulation of such small‐scale phenomena [Moore et al., 2015; Oltmanns et al., 2014] although they have been included with some success in regional climate models [Jourdain and Gallée, 2011]. Wind gustiness is also excluded from most models despite its importance for dust emissions [Engelstaedter and Washington, 2007].

More broadly, the feedback role of high‐latitude dust in global climate models is poorly understood. Although modern high‐latitude dust activity can be intense, it is typically limited to short timespans (a few days‐weeks per season) and does not appear to travel long distances at high concentrations. These two features suggest that the direct (i.e., blocking sunlight) radiative forcing of modern high‐latitude dust is a minor component in the annual global radiation budget, particularly when compared to its effects during glacial periods when global atmospheric dust loadings were 2 to 20 times current amounts [Lambert et al., 2008; Albani et al., 2015]. Dust in the atmosphere also has potential indirect effects on radiative forcing [Lohmann and Feichter, 2005], but the specific contribution of dust from high‐latitude sources is unknown. It is known that dust provides nuclei for ice and droplet formation in clouds [Atkinson et al., 2013] and that natural mineral aerosol is the important source of ice nuclei for colder temperatures (compared to biological sources which are more important at warmer temperatures) [Bromwich et al., 2012]. High‐latitude dust, particularly in the Northern Hemisphere, tends to travel at or below cloud level and may impact cloud properties, particularly in mixed‐phase clouds [Lohmann and Diehl, 2006]. Investigations of the effect of mineral dust aerosols on clouds in polar regions have tended to focus on dust sources from Asia and the Sahara in the Arctic [Fan, 2013; Breider et al., 2014] and from Australia and South America in Antarctica [Bromwich et al., 2012], and the impact of dust sources from within the high latitudes has yet to be assessed. The effect of nutrient deposition on marine biota and subsequent climate biofeedback [Mahowald et al., 2011] also has yet to be evaluated for high‐latitude dust sources. Simulations of the radiative forcing of dust and black carbon in seasonal snow in North China suggest that the magnitude of radiative warming in the snowpack is comparable to the magnitude of surface radiative cooling due to black carbon and dust in the atmosphere [Zhao et al., 2014].

5.5 Future Prospects and Opportunities for High‐Latitude Dust Research

An important conclusion to be drawn from this study is that understanding high‐latitude dust in the Earth system, even at relatively small scales, requires a multidisciplinary approach combining expertise in geomorphology, glaciology, meteorology, oceanography, sedimentology, atmospheric sciences, and other specializations. It also requires the use of a range of research tools including fieldwork, experimentation, observational networks, remote sensing, and modeling.

A distinctive feature of high‐latitude dust sources is that they are located in paraglacial environments. Paraglacial environments respond strongly to large‐scale external climate forcing and are sensitive to climate changes associated with glacial‐interglacial fluctuations and are also expected to be sensitive to future enhanced global climate change [Mercier, 2008; Knight and Harrison, 2009]. Many glaciers have been retreating over recent decades; for example, during the twentieth century all glaciers in Iceland have retreated [Björnsson and Pálsson, 2008] and high‐latitude glaciers and ice sheets are expected to retreat further during the 21st century [Radić et al., 2013; Clarke et al., 2015]. Glacier retreat is anticipated to expose large source areas of dust associated with glaciofluvial sediments. In some regions sediment supply, coupled to ice sheets and glaciers, to high‐latitude dust sources via meltwater systems and outwash plains may increase over the short term (decades) [Jansson et al., 2005; Bliss et al., 2014] but the overall relaxation time in different landscapes will vary. Countering this, factors that decrease sediment availability such as vegetation growth or colonization [e.g., Klaar et al., 2015] or the development of proglacial lakes may cause a rapid decrease in sediment availability to the aeolian system [Bullard, 2013].

These changes and the preceding evaluation of existing research suggest that a wide range of research questions still need to be tackled to fully understand the drivers, roles, and importance of high‐latitude dust in the Earth system. At the process scale, very little work has been undertaken to determine how dust entrainment and emission at high latitudes differs from that in the subtropics. While research on the behavior of sand‐sized material in cold climates has been conducted, reviewed, and is ongoing [McKenna Neuman, 1993, 2003], insights gained from this research have been applied to, but not tested on, fine particles. In the context of understanding the origins of loess, some experiments have examined the production of fine particles by aeolian abrasion of glacial sediments [Smith et al., 1991]; however, the impact of air temperature on the rate of abrasion has not been explored. Both aeolian abrasion [Smith et al., 1991; Bullard et al., 2004] and frost weathering [Wright et al., 1998] are known to contribute independently to the production of fine particles (silts and clays). Moreover, aeolian abrasion may occur more rapidly in cold than in temperate or warm environments due to increased particle brittleness caused by weakening of microfractures [Moss and Green, 1975].

As highlighted in section 5.3, field measurements of high‐latitude dust can be logistically challenging. To date, most research has been event based or focused on short, summer seasons. A key goal for the subdiscipline must be to obtain detailed year‐round measurements of dust flux and deposition concomitant with high‐resolution meteorological data. An integrated dust observation network across the high latitudes would provide valuable data to assist in quantifying the magnitude and frequency of dust events. Technological advances are improving automated instruments that can be used to quantify dust, and the ability to download data from remote areas via satellite will both help to overcome temporal sampling and access issues. Such a network of stations would be very valuable, but field‐ or plot‐scale research (10⁻² to 10² m) is also required to evaluate the impacts of spatial variability in sediment supply, redistribution of sediments. and snow by winds, and the role of surface and near‐surface microclimate on dust emissions. One possible way to further increase the density of dust observations could be through the establishment of community dust monitoring and reporting networks [Leys et al., 2008]. Although the population of many high‐latitude dust source regions is sparse, numbers are growing. For example, the population of Alaska has increased by 17% and of Iceland by 15% since 2000 with most population growth concentrated on larger settlements. Both the greater Anchorage area, Alaska, [Department of Labor and Workforce Development, 2014] and Reykjavík, Iceland, [Landshagir, 2015] have predicted population increases over the next few decades, and both areas are vulnerable to dust storms that decrease air quality. Consequently, more research into the effects of local dust storms on human health in cold environments may be welcomed.

For each high‐latitude dust region, with the exception of Iceland, large gaps still exist in our understanding of some of the basic characteristics of the dust sources (Table 2). These include a quantification of the areal extent of dust sources, comparable information concerning the wind thresholds required to entrain dust, and, importantly, estimates of total annual dust emissions from each region. Our first estimate (section 5.4) is that under contemporary environmental conditions 80–100 Tg yr⁻¹ of dust is contributed to the global dust cycle from high‐latitude sources; this represents up to 5% of the global dust budget. Considerably more field data and modeling efforts are required to test the robustness of this value.