Diurnal and seasonal source‐proximal dust concentrations in complex terrain, West Greenland

Diurnal and seasonal cycles of aeolian activity are well‐constrained for low latitude dust source regions and provide valuable insights into relationships between dust emissions and environmental drivers. Such cycles have received little systematic attention in high latitude dust source areas (≥50°N and ≥40°S), and understanding them will aid the modelling of atmospheric dust over different timescales. This paper examines the timing and drivers of atmospheric dust concentration close to source at ice‐free locations c. 6 km and c. 37 km from the Greenland Ice Sheet margin. Dust concentration, and associated environmental drivers including wind speed and velocity, air temperature and humidity, was measured from April–October 2018 and April–August 2019. Measured dust concentrations were a similar order of magnitude to source‐proximal values measured globally and varied from 0 to ≥1000 μg m−3. Diurnal cycles in the environmental drivers of dust emission were similar to those observed in low latitudes, but unlike in the low latitudes, there was no clear diurnal pattern of dust concentration. Only 3 out of 17 recorded dust events showed a significant positive relationship between wind speed and dust concentration, and factors such as wind direction, humidity and sediment availability are the over‐riding controls on dust activity at the event and seasonal scales. Whereas 7 dust events are attributed to down‐valley katabatic winds, 10 events were associated with a westerly sea breeze blowing up‐valley towards the ice sheet and high atmospheric moisture content. If deposited on the ice, this dust will alter ice albedo both directly and indirectly, (e.g. through the promotion of algal blooms) affecting cryospheric melt rates. Our results demonstrate an intricate relationship between first‐order controls and dust concentration that raises challenges for modelling the uplift and subsequent dust transport from high latitude sources, especially the appropriateness of assumptions based on emissions behaviour at low latitudes.

April-August 2019.Measured dust concentrations were a similar order of magnitude to source-proximal values measured globally and varied from 0 to ≥1000 μg m À3 .
Diurnal cycles in the environmental drivers of dust emission were similar to those observed in low latitudes, but unlike in the low latitudes, there was no clear diurnal pattern of dust concentration.Only 3 out of 17 recorded dust events showed a significant positive relationship between wind speed and dust concentration, and factors such as wind direction, humidity and sediment availability are the over-riding controls on dust activity at the event and seasonal scales.Whereas 7 dust events are attributed to down-valley katabatic winds, 10 events were associated with a westerly sea breeze blowing up-valley towards the ice sheet and high atmospheric moisture content.If deposited on the ice, this dust will alter ice albedo both directly and indirectly, (e.g. through the promotion of algal blooms) affecting cryospheric melt rates.
Our results demonstrate an intricate relationship between first-order controls and dust concentration that raises challenges for modelling the uplift and subsequent dust transport from high latitude sources, especially the appropriateness of assumptions based on emissions behaviour at low latitudes.

| INTRODUCTION
Broad-scale patterns of dust emissions are reasonably wellconstrained in terms of the global distribution of dust sources and the seasonality of dust emissions (e.g.Ginoux et al., 2012;Prospero et al., 2002).These patterns are better understood for low latitude (primarily sub-tropical) than high latitude dust source regions (≥50 N and ≥40 S) that until recently have received considerably less attention (Bullard et al., 2016;Meinander et al., 2022).First attempts to incorporate high latitude sources into regional models of dust emission suggest that land north of 60 N accounts for 1.7-5.3% of global dust emissions (Groot Zwaaftink et al., 2016) and is responsible for 30.7% of the dust burden in the Arctic (Shi et al., 2022).Dust originating in the Arctic and retained in the region provides an important nutrient and sediment transfer pathway linking terrestrial compartments such as soils and lakes, particularly in hydrologically disconnected areas (Anderson et al., 2017), and may also contribute nutrients to polar and high latitude oceans (Arnalds et al., 2014;Crusius et al., 2011).Mineral dust in the cryosphere is known to reduce albedo affecting snow and ice melt rates both directly through deposition and indirectly by providing nutrients such as phosphorus that support the growth of glacier ice algae (e.g.McCutcheon et al., 2021;Nagorski et al., 2019;Wientjes et al., 2011).Although high latitude-derived dust can be limited to low altitudes because of the stable and stratified atmospheric boundary layer in cold regions (Baddock et al., 2017;Ranjbar et al., 2021), it can also reach high altitudes (Dagsson-Waldhauserova et al., 2019) and provide ice-nucleating particles for the formation of Arctic mixed-phase clouds (e.g.Shi et al., 2022).Despite these critical roles, there are few contemporary near-surface measurements of airborne dust from high latitude sources with which to reduce uncertainties in modelling pan-Arctic aerosol-climate interactions and effectively tune global models of dust emission.
The impact of dust in the environment is modified by the diurnal (24-h cycle) and seasonal variation of emissions.For example, diurnal patterns of emission moderate the radiative effects of dust at the surface (Miller et al., 2004), the relative timing of dust and snow deposition in the cryosphere can influence melting and hence glacier and ice sheet hydrology, and the availability of nutrient-rich dust deposition in marine systems may depend on seasons with sufficient light for phytoplankton uptake (Bullard, 2017).In comparison with the low latitudes, diurnal and seasonal patterns of aeolian sediment transport, dust emission and associated environmental drivers have not been systematically well-quantified in the high latitudes using field data, largely due to the challenges these regions present for both access and long-term instrumentation.With few exceptions (e.g.Butwin et al., 2020;McKenna Neuman, 1990;Nickling, 1978;van Soest et al., 2022), direct measurements and observations of emission, transport or deposition of local dust in the high latitudes are often related to individual events or short monitoring periods of a few days (e.g.Dagsson-Waldhauserova et al., 2015;Fowler et al., 2018).Measurements over short sampling periods are valuable but may not be a good indicator of long-term dust activity due to the relative timing of data collection and dustiness.The analysis of secondary data, such as World Meteorological Organization (WMO) dust-related weather reporting codes, has been used to explore long term (multidecadal) seasonal and annual trends in dust events in Iceland (e.g.Dagsson-Waldhauserova, Arnalds, & Olafsson, 2014) and Greenland (Bullard & Mockford, 2018), but inconsistent and subjective classification of dust events (e.g.O'Loingsigh et al., 2010), local topography (van Soest et al., 2022) and background dust activity (Cosentino et al., 2020) can result in poor relationships between dust weather code data and field measurements.There remains a need for more direct, in situ measurements of near-surface dust activity in the high latitudes to better understand diurnal and seasonal patterns of emissions and the associated environmental drivers and impacts of the dust emission process.
Studies in low latitudes have found that aeolian activity occurs more frequently during daylight hours and is usually at a minimum in the morning whilst peaking in the afternoon.The spectrum of activity that follows this diurnal cycle includes sand transport (Gunn et al., 2021;Stout, 2010;Yang et al., 2013), the prevalence of blowing dust (Sreenivasaiah & Sur, 1937;Wiggs et al., 2022), dust-induced reduction in visibility (Basha et al., 2019;N'Tchayi Mbourou et al., 1997;Orgill & Sehmel, 1976), dust devil frequency (Lorenz et al., 2018) and atmospheric dust concentrations (Kelley et al., 2020;Stout, 2015).The diurnal pattern is attributed to incoming daytime solar radiation that produces (i) thermal instability and active atmospheric convective mixing resulting in strong horizontal wind velocities and (ii) surface drying by evaporation as temperatures increases, both of which promote aeolian activity (e.g.Ozer, 2001;Yu et al., 2021).This is followed by nocturnal cooling that promotes atmospheric stability, increased relative humidity and raised surface moisture levels all of which suppress aeolian activity.The strength of this cycle can vary with distance from the dust source region; for example, in north Africa, stations located further from source regions have a weaker daily cycle than those close to the source (N'Tchayi Mbourou et al., 1997).The diurnal dust cycle can also vary seasonally due to changes in ground surface conditions (moisture, snow, vegetation) and local wind conditions (Orgill & Sehmel, 1976).The daily cycle results in pulses of dust emission to the atmosphere, and the relative timing of these versus surface or satellite observations has implications for the estimation of total dust emission in a region (Todd et al., 2007;Wiggs et al., 2022).Some studies from the high latitudes suggest similar diurnal patterns of dust to those in the low latitudes (e.g.Cosentino et al., 2021).
For example, Nickling (1978) observed 15 dust events in May-June in the Slims River Valley (>60 N, Yukon Territory) and found minimal aeolian activity in the morning, even when wind speeds were high, due to high sediment surface moisture content.As air temperature and associated evaporation rates increased during the day and the surface dried, aeolian activity increased resulting in the occurrence of dust emissions in the afternoon.In the high latitudes, the diurnal temperature range is typically highest in the winter or spring and lowest in summer or autumn reflecting the impact of increased cloudiness in warmer months (Przybylak, 1999).In addition, the absence or reduction of incoming solar radiation during the polar night suppresses the diurnal heating cycle increasing the importance of wind associated with largescale atmospheric circulation patterns rather than locally generated winds (Gillies et al., 2013;Przybylak, 2000).In the Dry Valleys of Antarctica, this results in a clear diurnal pattern of aeolian activity in the spring and summer that is not disernible in the fall and winter when solar radiation is low or absent (Gillies et al., 2013).Spring and summer dust concentrations in Reykjavík, Iceland follow a diurnal cycle where PM 10 concentrations are lower during the night and early morning and increase in the afternoon; however, Reykjavík is several hours particle travel time downwind of dust sources suggesting a morning onset of dust emissions (Thorsteinsson et al., 2011).This is supported by Mockford et al. (2018) who measured dust emissions directly at sources in southern Iceland and found the onset of dust events in June varied from 07:30 to 14:45 (local time) with all except one event starting before midday.Butwin et al. (2020) recorded start times of dust events in the winter (October-February) in southern Iceland in both the early hours (midnight to 9 AM) and afternoon (midday-15:00 local time).Dust events of long duration can have multiple peaks during a 24-h period; for example, Dagsson-Waldhauserova et al. (2015) reported sustained high dust concentrations (≥1000 μg m À3 ) for >24 h with multiple peaks of up to 6500 μg m À3 during both day and night in March 2013 in southern Iceland (c.20 km downwind of the source).
At the seasonal scale, recent field measurements of aeolian activity in Greenland and Svalbard have identified positive relationships between mean aeolian deposition rates and mean air temperature, but the seasonal relationships between wind speed, precipitation and humidity are less clear or consistent (Rymer et al., 2022;van Soest et al., 2022).Barchyn & Hugenholtz (2012) found that the threshold wind speed for aeolian activity in Canada varied considerably more during winter than summer and they attributed this to the influence of moisture, ice and temperature on surface conditions.The relationship between aeolian activity, dust emissions and meteorological conditions is particularly variable where fine (<100 μm) sediments are supply-limited or availability-limited (Bullard, 2013).In high latitude sources, sediment supply to floodplains typically results in a spring and/or autumn increase in dust availability (Bullard et al., 2016), but there can also be a time lag between delivery of sediment to the floodplain and the occurrence of conditions suitable for deflation such that the timing of supply of sediments to the deflation area is decoupled from erosive meteorological conditions (Prospero et al., 2012;van Soest et al., 2022).In addition, local factors such as wind direction can exert a strong seasonal control on dust uplift particularly where dust sources are topographically-constrained, such as in valleys (e.g.Barchyn & Hugenholtz, 2012;Crusius et al., 2011).
Quantifying diurnal and seasonal variations in near-surface dust concentrations can help to clarify the relationship between dust emissions and environmental drivers that is essential for understanding contemporary patterns of dust activity and predicting potential future changes in the timing, magnitude and frequency of overall dust emissions.The aim of this paper is to examine the timing and potential drivers of variations in atmospheric dust concentration close to source in West Greenland.Dust derived from ice-free Greenland provides nutrients and sediments to terrestrial and cryospheric ecosystems (Anderson et al., 2017;McCutcheon et al., 2021) and is estimated to contribute approximately 1 Tg year À1 to the global dust load (Groot Zwaaftink et al., 2016).In addition, recent analyses of lake, firn and ice core records suggest dust emission from Greenland has increased over the past 20 years due to increased sediment availability (Amino et al., 2021;Kjaer et al., 2022).Within this paper, we consider first the overall annual environmental conditions during 2018 and 2019 when the data were collected, in order to situate the results in the longer-term context of a rapidly changing Arctic.We then present an analysis of diurnal patterns and trends of meteorological variables and dust concentrations, followed by an exploration of the gross relationships between them.
Finally, we consider individual dust events and associated environmental drivers that enable an analysis of the influence of geography on the activation of sources and distribution of dust in complex terrain.

| METHODS
The study area is the Kangerlussuaq region (67 00 0 N, 50 43 0 20 00 W) of West Greenland (Figure 1) for which substantial contemporary and historical ecological, geomorphological and meteorological data are available (Yde et al., 2018).This ice-free area of land is situated between the Labrador Sea and the Greenland Ice Sheet (GrIS) and is geomorphologically representative of many Arctic proglacial systems (Anderson et al., 2017).There is a continental arid climate that is changing rapidly such that since 1994 mean June air temperatures have increased by 2.2 C (from 8.3 C to 10.5 C) and mean winter precipitation has almost doubled from 21 to 40 mm (Saros et al., 2019).Daily meteorological conditions are principally governed by the stable, high-pressure air mass over the ice sheet and the relative strength of katabatic winds versus an onshore sea breeze channelled by local topography (Cappelen et al., 2001;Kopec et al., 2014).WMO dust coded weather records suggest a long-term average of c. 5 dust event days per year (Bullard & Mockford, 2018), and contemporary annual deposition of dust derived from local-regional sources is in the range 10-100 g m À2 (van Soest et al., 2022;Willemse et al., 2003).
Inter-annual dust activity is highly variable, but overall atmospheric dustiness in this part of Greenland has increased since 2000 driven by an increase in the magnitude of dust events rather than their frequency (Saros et al., 2019).Dust sources in the area include active glacial outwash plains and dunefields in Sandflugtsdalen and Ørkendalen (Bullard & Austin, 2011;Dijkmans & Törnqvist, 1991), sediments exposed at the head of Kangerlussuaq fjord and local reworking of loessic soils (Heindel et al., 2015) (Figure 1).Outwash plain sediments are typically bimodal with a coarse pebble mode and fine sand mode but can be overlain by seasonal flood deposits comprising primarily silt and clay-sized material (≤63 μm), which are the main source of local dust (Bullard & Austin, 2011;van Soest et al., 2022).The sediments are derived from glacial erosion, weathering and fluvial deposition and comprise quartz, plagioclase, K-feldspar, amphibole, pyroxine and pyrite (Eisner et al., 1995;Hasholt et al., 2018).There are small, localised anthropogenic dust sources close to the head of the fjord, but most mining activities take place >80 km west of the study area and are unlikely to affect dust concentrations at Kangerlussuaq (Søndergaard & Jørgensen, 2021).The predominant wind direction associated with WMO dust events is from the east to northeast; however, dust events associated with southwesterly winds have also been reported (Bullard & Mockford, 2018).
Two TSI DustTrak DRX Environmental Monitor 8543-M instruments were deployed from April to October 2018 and April to August 2019 (Table 1) with the inlet at a height of 2.5 m.These time periods capture spring through to early autumn, but it was not possible to measure through the polar winter due to power and instrument constraints.One DustTrak was located at an altitude of 75 m above the Sandflugtsdalen glacial outwash plain, 5.6 km from the GrIS marginhereafter referred to as the 'Ridge' site (Figure 1).This DustTrak was located on the east end of a topographic ridge, approximately 50 m below the highest point of the ridge and fully exposed to katabatic winds from the GrIS.The second was located 37.4 km west of the GrIS margin in a lake catchment characterised by loessic soilshereafter referred to as 'SS85'.This site is c. 190 m a.s.l., surrounded by hills up to 300 m a.s.l. and separated from the potential sediment sources of the fjord and outwash plains by a north-east to south-west trending topographic ridge.Previous research has been conducted at the Ridge (van Soest et al., 2022) and SS85 lake catchment (e.g.Anderson et al., 2001;Prater et al., 2021;Saros et al., 2019), and the names are retained here for consistency and to allow our data to be associated with previous work.
The DustTrak is a laser photometer and was used to measure average dust concentration for 1 min at 5-min intervals.There is some uncertainty regarding the maximum particle size over 15 μm that the DustTrak can detect (Goossens & Buck, 2012), and for this reason, only PM 10 size-constrained values were used in this analysis.The DustTrak can underestimate PM 10 by up to 20% particularly at high dust concentrations (Javed & Guo, 2021).It can also record falsely high particle concentrations at high relative humidity due to the effect of humidity on electrical components and/or the instrument falsely reading water droplets as dust particles (Jayaratne et al. 2018).To minimise this risk, both DustTraks were fitted with heated sampler inlets to remove atmospheric moisture.In our measurement record, there are 6151 observations where relative humidity is ≥85%, and T A B L E 1 Dates of instrument at each site and % successful data retrieval.%PM 10 < 1 indicates % of retrieved data where PM 10 readings were <1 μg m À3 .only 59 of these observations coincide with dust concentrations ≥10 μg m À3 .As this is <1% of all observations at this humidity, it is unlikely that these potential errors occurred and that dust concentrations recorded during high relative humidity can therefore be accepted.
At each site, wind speed and direction were measured using an RM Young 5013 sensor mounted at 2.5 m with the same sampling interval as the DustTrak.Air temperature and humidity were recorded every 10 min using Omega OMYL-RH20 self-logging instruments.All data are reported as West Greenland Standard Time (UTC-3).The use of solar power for instrumentation, interference with equipment by wild animals and minor equipment malfunction resulted in incomplete retrieval of data, that is, recording gaps of variable duration (Table 1).
Of the dust concentration measurements recorded, 47-76% had PM 10 readings of <1 μg m À3 (Table 1).In this paper, for analysis of individual variables, the maximum possible data set has been used.
Where variables are being compared, for example, the relationship between dust concentration and wind speed, only concurrent readings are used.
For the purposes of this study, background dust emissions were taken to be <10 μg m À3 , and a dust 'event' was defined as the occurrence of at least six 1-min PM 10 measurements ≥10 μg m À3 within 1 h.Where dust concentrations drop below 10 μg m À3 for longer than 1 h but a further period of at least 6 PM 10 measurements ≥10 μg m À3 within 1 h occurs again during the same day this is classed as a single event.This threshold is substantially lower than that required to trigger concerns regarding air quality (e.g.mean values over 24 h exceeding 50 μg m À3 ; de Longueville et al., 2013) and that required to reduce visibility to levels used to define a WMO dust storm (highly variable depending on particle size and distance from source but typically >400 μg m À3 ; Baddock et al., 2014;Camino et al., 2015).It is used, however, because background dust levels are very low compared with those reported elsewhere (e.g.Cosentino et al., 2020;Mockford et al., 2018;Thorsteinsson et al., 2011) and as an indicator of all dust emissions (and wind erosion activity) in the area rather than only events of prescribed magnitudes.Using the method of Wiggs Time (Fausto et al., 2021).
To identify any diurnal patterns in the meteorological drivers of dust emissions, hourly averages for wind speed, temperature, humidity and solar radiation were calculated for each site over their periods of observation.This averages approximately 300 values for each hour in each measurement period.In atmospheric and aeolian science, a 24-h cycle is generally referred to as the diurnal cycle.In ecology, the term diel is more common, and because it is often used to describe a 24-h period regardless of day or night, this term may be more appropriate for cycles in the high latitudes where at certain times of the year variations between day and night are hard to distinguish.Nevertheless, because most research into daily patterns of aeolian activity describes them as diurnal, we retain that term here and refer to morning (5 AM to 12 noon), afternoon (12 noon to 5 PM), evening (5-9 PM) and night (9 PM to 5 AM) regardless of season.

| Environmental conditions 2018-2019
Seasonal temperature anomalies (relative to the baseline average for 1981-2010) during the field sampling periods in 2018 were typically <1 C (Tedesco et al., 2018).Snow accumulation during winter 2017/2018 was above average, and snow extent was above average during April and May (Mudryk et al., 2018).Onset of the 2018 melt season-as indicated by positive degree days-was mid-May (Julian Day 132), and the area was snow-free by mid-June (Figure 2).Annual discharge from the Watson River in 2018 was well below average at 3.76 ± 0.66 km 3 (measured mean discharge 2006-2021; 6.45 km 3 ).In contrast, in spring 2019, temperatures were 2.5 C above the baseline that, combined with low prior winter snowfall, led to an early melt season leaving the terrestrial area within the region snow free by the end of April (Mudryk et al., 2019)

| Diurnal patterns of meteorological variables and dust concentration
The overall diurnal pattern for each meteorological variable is cyclical (Figure 4).Wind speeds overall are c.0.5 m s À1 higher at the Ridge than at SS85, and at both sites, the lowest wind speeds are in the morning between 03:00 and 07:00.Wind speeds increase during the day reaching a maximum in the mid-afternoon to early evening (15:00-20:00).Temperatures are lowest between midnight and 04:00 and increase to their highest values in the late morning that is sustained until early evening.Relative humidity is highest at night reaching above 60% from 22:00 to 06:00 and falling to lowest values during the day (from 11:00-18:00).Average incoming solar radiation for April to October 2018 and 2019 peaks at midday.Solar radiation varies considerably from month to month at this latitude, but separate curves for May, July and September in each year indicate the peak was always between 11 AM and 3 PM for the measurement periods used (Table 1).Our observations do not include winter when incoming solar radiation is minimal.
Average diurnal wind direction at the Ridge is from the north overnight veering to north-north easterly/easterly in the early morning to down valley (off-ice) flow during the day (Figure 5).At SS85, average diurnal wind direction is from the north to north-east during the night veering to east then east-south-east during the day.
In contrast to the meteorological variables, there is no consistent diurnal pattern to dust concentration (Figure 6).In 2018 at the Ridge, the highest average dust concentrations were recorded between midnight and 05:00.In contrast, in 2019 at the Ridge and 2018 at SS85, the lowest average dust concentrations were recorded between midnight and early morning and the highest concentrations recorded in the late morning and early afternoon.In 2019 at SS85, the highest dust concentrations occur in the early hours (01:00-06:00) and evening (18:00-22:00).

| Dust events
Seventeen individual dust events were identified across the 2018 and 2019 monitoring periods, some of which extended over two calendar days.Of these, 10 occurred in 2018 and 7 in 2019 (Table 2).Nine events were recorded only at the Ridge site, and four only at SS85.A further four events were recorded at both locations at the same time although the exact start time and duration of the dust emissions varied slightly between sites.The WMO weather-type observations at Kangerlussuaq airport (Figure 1) were examined for each event, but with the exception of Event 12, no dust weather codes were recorded coincident with dust events as defined here.For Event 12, detected only at the Ridge, the airport meteorological records show that there were strong wind gusts and an easterly wind with blowing and drifting sand observed in the vicinity.Sentinel-2 (10-m spatial resolution) imagery was examined for all dust events, but due to a combination of cloud cover and the relative timing of dust events and overpasses, none were visible using satellite remote sensing.Dust events were most frequent in May and July (Figure 8), but from all events, there is no clear diurnal pattern to the onset of dust events.Start times given in Table 2 and summarised in Figure 8 show the onset of 3-5 dust events in each 3-h time bin (from midnight).Event fluxes ranged from 0.092 g m À2 (Event 10) to 15.430 g m À2 (Event 7), and the majority were <5 g m À2 (Table 2).Excluding the exceptional event recorded at the Ridge on 28 July (Event 7), there is a significant positive relationship (P < 0.001) between event duration (in minutes) and event flux (g m À2 ) at each site (Figure 9).Four of the events listed in Table 2 are examined in detail below-these are 24-28 July 2018 (Events 6 and 7) because they featured the highest dust concentrations and 15-16 July (Event 5) and 22-24 September 2018 (Event 9) because they were detected at both measurement sites.

| Event 5: 15-16 July 2018
Event 5 was recorded at both the Ridge and SS85 nearsimultanously and lasted for over 20 h at both locations (Figure 10).At the Ridge, there was a short period of increased dust emissions for c. 3 h from mid-morning and then the main detection period at both locations was from mid-afternoon until mid-to late-afternoon the following day.Mean and maximum dust concentrations were similar at both locations but fluctuated more at the Ridge than at SS85.Dust emissions were associated with down-valley winds and the end of the event at both locations was marked by a change in wind direction to northerly at the Ridge and westerly at SS85 as well as an increase in humidity from c. 40% to >80% at both locations.(mean = 78 μg m À3 ) (Figure 11).This pulse of dust emissions was associated with south-southwesterly (up valley) low velocity winds (mean 2.6 ± 0.6 m s À1 , maximum gust 4.2 m s À1 ), a mean temperature of 3.6 ± 0.07 C and mean relative humidity of 99.9 ± 0.14%.Dust concentration then dropped to ≤10 μg m À3 until 00:50 on 28 July when a second event began as dust concentration increased from a concentration of 10 to >300 μg m À3 within 10 min.For the next 5 h, all measurements were ≥10 μg m À3 and fluctuated between 12 μg m À3 and 1000 μg m À3 with a mean of 313 μg m À3 .
The DustTrak stopped recording at 05:55 on 28 July so it is not possible to determine the full duration of the event.During this period, wind speeds were up-valley and low (mean 1.9 ± 0.4 m s À1 , maximum gust 4.1 m s À1 ); mean temperature and mean relative humidity were 3.3 ± 0.08 C and 99.0 ± 0.36%, respectively.In contrast to the Ridge, dust concentrations at SS85 were ≤2 μg m À3 throughout this 5-day period (Figure 11).The levels and timing of variations in humidity, temperature and wind speed observed at SS85 are very similar to those recorded at the Ridge, but wind direction is more variable.

F I G U R E 9
The relationship between event flux and event duration at each measurement site.Best-fit line for data from the Ridge (dotted line) excludes Event 7 that is shown separately.

| Event 9: 23-24 September 2018
Events 8 and 9 were recorded at both locations as a series of concentration peaks occurring over 1-3 h in the early morning of 22 September (Event 8) and during the day on 23 September extending through to the early hours of the following morning (Event 9; Figure 12).Wind speeds and direction were variable at both sites.At SS85, wind direction was primarily from the northern sector, but wind direction at the Ridge was variable between northerly and easterly.Minor differences between the two observation sites are apparent because diurnal cycles in the Kangerlussuaq region are modulated by distance from the ice sheet margin.Air temperature rises earlier in the day, and the highest temperatures are sustained for longer at SS85 (c.37 km from the GrIS margin) compared with the Ridge (c.6 km from the GrIS margin).Also, the influence of down-valley katabatic winds driven by thermal gradients between the ice and the land surface results in higher wind speeds with an earlier and broader peak at the Ridge compared with those recorded at SS85.This decrease in strength of katabatic winds with distance from the GrIS has been widely noted (e.g.Heinemann, 1999;Smeets et al., 2018), and the differences in the timing and profile of the daily cycles of wind strength are similar to those observed at sites 6.2 and 20 km west of the GrIS by van den Broeke et al. (1994) for July in the same region.Diurnal averages of wind direction illustrate the prevalence of daytime katabatic, easterly winds.During the night, winds at both sites have a northerly component that is attributed to the overnight weakening of the strongly directional katabatic winds and increased influence of the change in roughness length as wind flows off the ice and over vegetated tundra causing backing of easterly winds to the north-northeast (van den Broeke et al., 1994).
Relative humidity decreases during the day and is at its lowest in mid-late afternoon as also observed in low latitudes (e.g.Gunn et al., 2021;Stout, 2015).The combination of low relative humidity and higher temperatures will result in highest values of absolute humidity between 06:00 and 21:00 that accords with van den Broeke et al. 's (1994) July data from Kangerlussuaq.In this part of Greenland, Kopec et al. (2014) differentiate between days dominated by persistent easterly winds ('no reversal days') that have high overall wind speeds and low humidity all day with only a slight increase in the latter around 16:00 and 'reversal' days when winds driven by a marine-like air mass to the west result in a sea breeze effect and the occurrence of up-valley winds.The up-valley winds typically occur during the afternoon and evening and are associated with high atmospheric moisture content.Using the definition of Kopec et al. (2014), 10 of the 17 dust events listed in Table 2 occurred on 'reversal days'.Several studies have recorded the occurrence of dust on the surface of snow and ice on the GrIS and traced it mineralogically or sedimentologically to local Greenland dust sources (e.g.Bøggild et al., 2010;Kjaer et al., 2022;Wientjes et al., 2011).In the cryosphere, dust deposits can contribute to the development of cryoconite holes that reduce albedo and can increase ice melt rates (Bøggild et al., 2010).Dust is also recognised as providing important nutrients for the growth of supra-glacial ice algae on the GrIS that also accelerate surface melting (McCutcheon et al., 2021).The occurrence of up-valley dust-laden winds identified in this study provides a mechanism for the transfer of locally entrained dust towards the GrIS and its deposition on snow and ice.Calculations of event flux (Table 2) suggest that, overall, similar quantities of dust are transported up-valley (towards the ice) and down-valley.At the Ridge, the total event flux (all events) was 20.36 g m À2 for events associated with up-valley winds (reversal days) and 13.92 g m À2 for events associated with down valley winds.At SS85, total event flux was 3.99 g m À2 for up-valley winds and 5.83 g m À2 for down valley winds.
In contrast to that observed at low latitudes, there is no consistent relationship between the diurnal cycle of meteorological drivers and dust concentration at Kangerlussuaq.At SS85 in 2018 and the Ridge in 2019, dust concentrations peak from mid-morning to midday, which earlier than the meteorological drivers would suggest.At the Ridge in 2018, the highest dust concentrations are in the early hours (03:00-05:00), and at SS85 in 2019, they are in the early hours (02:00-06:00) and evening (18:00-22:00).As only 17 dust events F I G U R E 1 2 Summary meteorological and PM 10 data for Event 9 on 23-24 September 2018 for (A) Ridge and (B) SS85.Wind direction is given every 8 h.were recorded, the observed pattern is strongly influenced by the diurnal timing of those events with the highest dust concentrations (e.g.Events 6 and 7).The diurnal analysis dissociates meteorological drivers from dust concentration and does not account for the influence of additional factors not fundamentally controlled by a diurnal cycle, such as longer term sediment availability and topography.

| Meteorological conditions and dust concentration
Meteorological conditions were measured as prospective drivers of dust emissions; however for concurrent measurements at 10-min intervals, there are no significant relationships between wind speed, wind direction, air temperature or relative humidity and dust concentration (≥10 μg m À3 ) for either site.There is a complex and dynamic feedback relationship between meteorological conditions and surface conditions that can affect the threshold for aeolian particle entrainment (Barchyn & Hugenholtz, 2012).For example, humidity, which affects evaporation and sublimation fluxes (Box & Steffen, 2001;Liston & Sturm, 2004), strongly influences the wind speed required to initiate sediment transport (McKenna Neuman & Sanderson, 2008).
The effect of humidity depends on whether interparticle forces are dominated by adsorption forces or capillary forces that, in turn, is governed by particle size, density and particle-scale surface roughness (Richard-Thomas & McKenna-Neuman, 2020;Yu et al., 2017).When interparticle forces are dominated by capillary forces, such as for dense, sand-sized particles, the threshold friction wind speed typically increases with increasing relative humidity for values greater than 30-40% (Belly, 1964;McKenna Neuman & Sanderson, 2008).Gunn et al. (2021) suggested little or no sand saltation would occur when relative humidity >40% and, given saltation plays a key role in dust emissions, also suggest dust emissions would substantially reduce above the same threshold.Soil particle-size characteristics can affect the relative importance of these different forces.For example, Ravi et al. (2006) found for clay loam and sandy loam, containing a mix of clay, silt-and sand-sized particles, that threshold increases with increasing relative humidity up to 40% (below which adsorption forces dominate), decreases with increasing relative humidity between 40 and 65% and increases with increasing relative humidity >65% (above which capillary forces dominate).
Given the sediment sources in Kangerlussuaq comprise sand, silt and clay, the above suggests dust concentrations should decrease with increasing relative humidity above 60-65%, if not less, but for the events examined here, six occurred when mean relative humidity was ≥60%, and the highest concentrations of PM 10 were recorded for relative humidity >90% (Table 2).This might be explained by the fact that the relative humidity and temperature of air also affect air density that impacts aeolian particle entrainment and transport (Du et al., 2022;McKenna Neuman, 2003, 2004).In low altitude cold regions, this can result in very low threshold velocities at high relative humidity (>60%;McKenna Neuman & Sanderson, 2008).Mockford et al. (2018) proposed that for sites where there is a strong eventscale relationship between surface wind speed and dust concentration, both transport capacity and sediment supply are likely to be nonlimiting.Where the relationship is weak, factors such as relative humidity, precipitation, surface properties (such as the development of lag deposits or crusting) and wind direction in relation to source (e.g.Crusius et al., 2011) can become more important than wind speed.These factors can be elucidated by examining the evolution of dust events through time.
Of the two Greenland sites considered here, there is a significant positive relationship between wind speed and PM 10 for Event 9 (both locations), Event 14 (both locations) and Event 16 (Table 2).For Event 5, there is a significant negative relationship at the Ridge but not at SS85.For Event 5, dust concentration drops abruptly in the early evening despite an increase in wind speed.This drop is concomitant with a rapid decrease in temperature and increase in relative humidity to >80% that may have increased the required threshold velocity.The decrease in dust concentrations was also concurrent with a change in wind direction from easterly to northerly that would reduce the likelihood of particle entrainment from the approximately east-west aligned Sandflugtsdalen outwash plain source.
For Events 6 and 7, high dust concentrations were recorded at the Ridge alongside very high relative humidity (99%) and low wind speeds (<3 m s À1 ).Aeolian activity, including dust emissions, occurring during light rain and relative humidity >60% have previously been recorded in Iceland (e.g.Ashwell, 1986;Butwin et al., 2020).Dagsson-Waldhauserova, Arnalds, Olafsson, Skrabalova, et al. (2014) measured high concentrations of very fine dust (>1500 μg m À3 min À1 ) associated with low wind speeds (0-4 m s À1 ) and high humidity (77-90%) in August in Iceland.In this case, dust emission was attributed to upwards air motion associated with surface heating from solar radiation and peaked in mid-afternoon.It is unlikely that this process was operating during Events 6 and 7 presented here as they both occurred close to midnight and, although due to the time of year there was incoming solar radiation, temperatures were low.The observed wind direction indicates it is more likely that Events 6 and 7 occurred during a sea breeze-driven air flow reversal bringing humid maritime air from the south-west, inland up the fjord (Kopec et al., 2014).This wind could potentially entrain dust from sources at the head of the fjord as well as possibly Ørkendalen.Despite the high dust concentrations, these dust events were only recorded at the Ridge site and not at SS85.Local topography is likely to play an important role at both sites.At the Ridge, the instrument station is below the highest point of a topographic ridge to the west and therefore in the lee side of an obstacle for up-valley winds.Deceleration of winds on the lee side of topography and topographic steering can result in high dust concentrations (Goossens, 2006) and high dust deposition on the downwind side of obstacles (Comola et al., 2019).The instrument station at SS85 is to the west of the Kangerlussuaq fjord head sediment deposits, and there are no obvious sources to the immediate south or west from which dust could have been entrained.The wind direction driving the dust events featured in Figures 10 and 12 suggests dust observed at SS85 originated to the north or east of the site.It may also be locally derived where vegetation is disturbed enabling the erosion of loessic soils (Heindel et al., 2015).Marticorena et al., 2017;Wiggs et al., 2022), dust transport during the measurement period is dominated by a single high magnitude, low frequency event (Event 7) that accounted for 45% of total event flux.

| Variability of event, seasonal and annual dust concentration
In both 2018 and 2019, on average 1-2 dust events per month were recorded during the measurement period.The first event each year did not occur until after the start of the melt season and the region was snow-free.More dust events were recorded at the Ridge in 2018 (n = 9) compared with 2019 (n = 4), and this is reflected in the average annual dust concentration at the Ridge that was 1.81 μg m À3 in 2018 and only 0.45 μg m À3 in 2019.In contrast at SS85, there were 4 events recorded each year, and the average annual dust concentration was similar in each year (1.03 μg m À3 in 2018; 1.05 μg m À3 in 2019).Interannual variability of dustiness is widely reported and driven by the interplay of factors affecting sediment supply, sediment availability and transport capacity (Zender & Kwon, 2005) interacting with large scale periodic climate patterns such as the El Nino-Southern Oscillation (Cosentino et al., 2020;Marx et al., 2009), North Atlantic Oscillation (Stewart et al., 2023) and Interdecadal Pacific Oscillation (Rimbu et al., 2022).With only a partial 2-year record of dust concentration measurements, it is not possible to examine the impact of climate periodicity on West Greenland dust concentrations, but some interannual factors can be explored.concentration measured for this study varied each year at the Ridge but not at SS85.This may be because the Ridge is more closely coupled to the active sediment sources in the region particularly the up-valley outwash plain of Sandflugtsdalen but potentially also the sediments at the head of the fjord that lies down-valley.Higher dust concentrations measured at the Ridge in 2018 may be due to source proximity but also because low discharge through the Watson River (2018: 3.76 ± 0.66 km 3 ) would ensure the outwash plain was drier, enhancing the potential for available fine sediments to be transported by wind.In addition, very local sources of dust derived from reworking of exposed loess deposits may have contributed to the dust concentrations.In 2019, discharge from the Watson River was very high (8.33 ± 1.37 km 3 ), and although this may have delivered more sediments to the outwash plain and fjord delta, it also resulted in greater spatial and temporal extent of inundation of, and high surface moisture levels on, the outwash plain that would reduce the likelihood of desiccation and aeolian entrainment.In addition, high discharge can result in transport of meltwater suspended sediments beyond the fjord delta (Chu et al., 2012) decreasing the likelihood of entrainment from sources at the head of the fjord.There was no substantial difference in annual dust concentration or event frequency at SS85 between 2018 and 2019 possibly because it is separated from the major sources of fine sediments by a topographic ridge that may restrict atmospheric dust distribution.This may indicate that dust detected at this location is predominantly very locally derived or the result of regional (>10 km; Lawrence & Neff, 2009) dust events.Bullard and Mockford et al. (2018) used WMO dust weather codes to examine the seasonal variability in dust event days over 72 years at Kangerlussuaq.They found the number of dust event days was highest in May and June and September/October and lowest in July and August.For the current study, dust 'events' occurred most frequently in July, followed by May.No dust events were recorded in August, and we were unable to measure dust concentration during the winter months (November-March).Although the record of dust concentration we present here is one of the longest for a high latitude dust source, the two observation periods from spring to autumn remain short compared with the WMO record.Both of these data sets reveal implications for achieving a detailed understanding of seasonal dust variability based on ground-based monitoring at high latitude sources.With the use of year-round dust deposition measurements, van Soest et al. (2022) suggested there is not a good relationship between long-term dust proxy data (measured at the airport) and dust deposition in Kangerlussuaq over short (seasonal to annual) timescales for reasons that may include short measurement durations, local topography and wider issues around the proxy data set.Local topography that separates dust sources (e.g. the fjord and Sandflugtsdalen) may constrain the spatial extent of dust events; however as defined here, four events were detected near-simultaneously at both the Ridge to the east of the airport and SS85 31 km away and to the west of the airport that can therefore be considered 'regional' events (Lawrence & Neff, 2009), but none of these four were recorded as dust weather codes at the airport.Despite using high spatial resolution imagery that should enable the identification of small dust sources, none of the events identified during the measurement period in Kangerlussuaq were visible using Sentinel-2 satellite data.For 5 of the 17 events, this was due to cloud cover, and for a further five, the events did not coincide with a satellite overpass; that is, no image was available on that day.For 6 days when events occurred, the dust sources were free from cloud, and the land surface was clearly visible, but no dust plumes were visible.The only possible event detected was Event 13 where skies were clear and small dust plumes might be visible, but further work would be required to confirm this.Satellite remote sensing has proven a very useful tool for identifying regional to global scale patterns of dust emissions in the low latitudes (e.g.Ginoux et al., 2012;Prospero et al., 2002), but it is recognised that the contribution of small dust sources to global dust loading is considerably underestimated by this method (Urban et al., 2018).Local high latitude dust events have been identified and characterised using remote sensing (e.g.Ranjbar et al., 2021), but the key challenges of remote sensing at high latitudes remain and include those associated with identifying small sources and plumes that do not rise to high altitudes as well as high levels of cloud cover and the relative timing of dust emissions and satellite overpasses (Bullard et al., 2016).

| CONCLUSIONS
Our results present the first systematic multi-season surface measurements of source-proximal dust concentrations in West Greenland.
These data provide a quantification of dust activity in a source region that is geomorphologically representative of the type of fjord or glacial valley that characterises many contemporary dust sources in high latitudes.Atmospheric dust concentration in the Kangerlussuaq region of West Greenland is typically low ≤10 μg m À3 , but measurements of dust concentrations from April to October demonstrate that increased dust concentrations can occur in associated with dust events that are detectable at both the local (<10 km) and regional (10-100 km) scale.
The largest events reached maximum PM 10 concentrations of several hundred μg m À3 , and two such events occurred during the measurement period.Both of these events were driven by up-valley winds that transported dust towards the GrIS.If deposited in the cryosphere, this dust would have the potential to reduce ice albedo and increase snow and ice melt rates.The total of all derived event fluxes suggests similar quantities of dust were transported towards and away from the ice during the measurement period.The diurnal cycles of wind speed, temperature and humidity are similar to those observed at low latitudes, but unlike at low latitudes, there is no clear diurnal cycle of dust concentration.This lack of a clearly discernible diurnal dust pattern may be due to the low number of observed events or a decoupling from the diurnal variation of meteorological drivers caused by the influence of additional controls on emissions, such as event-based wind direction or surface erodibility state, through constraints imposed by sediment availability.
The complexity of the role played by meteorological drivers in the regional dust system is further indicated by the intricate overall relationship between dust concentration and any of wind speed, air temperature or relative humidity for the spring to autumn period.For specific events in some instances, there is a relationship between dust concentration and wind speed, but this is not universal.Such intricacy concerning the first-order controls on dust emissions raises challenges for the modelling of uplift and subsequent dust transport from such high latitude sources, especially the appropriateness of assumptions based on emissions behaviour in low latitude sources.It also highlights the benefits of in situ measurements for understanding the interplay of those drivers particular to high latitude dust dynamics, for instance in the case here where aeolian activity resulting in the highest dust concentrations occurred in conditions of very high humidity.

F
I G U R E 1 (a) True colour (RGB) Sentinel-2B image from 11 June 2019 (ESA Copernicus https://scihub.copernicus.eu/dhus/#/home);(b) elevation map of study area derived from the AsterGDEM; (c) measurement site on the Ridge in April 2018 looking east towards the Greenland Ice Sheet; (d) measurement site at SS85 in August 2018.[Color figure can be viewed at wileyonlinelibrary.com] et al. (2022), for each dust event, an indicative event flux was derived from the co-located measurements of dust concentration and wind speed at 2.5-m height.Additional variables can affect dust emissions and consequently atmospheric dust concentration.These include sediment supply, snow cover, melt rates and incoming solar radiation.The main sediment supply for dust emissions in the Kangerlussuaq area is from meltwater suspended sediments.Published meltwater discharge data for the Watson River (van As, 2022) were used as an indication of potential sediment supply to the floodplain dust sources as there is a positive relationship between the concentration of suspended sediments ≤2-mm diameter and discharge for this river(Hasholt et al., 2018).Hourly air temperature data from Kangerlussuaq airport WMO station (67.01 N 50.71W) were used to calculate positive degree days-the sum of air temperatures above the melting point during a period-to indicate the timing of the onset of the melt season for 2018 and 2019 where pdd M ¼ 1 The positive degree-day sum for each day is pdd M ( C d), and T m is the temperature ( C) measured M times in the day (d À1 ) (Braithwaite & Hughes, 2022).Daily time series estimates of overall regional snow cover (spatially averaged for the domain 66.5-67 N, 50.25-51.25W) and snow cover at a point specifically representing the Sandflugtsdalen outwash plain were obtained from the snow depth parameter of the ERA5 reanalysis data (as liquid water equivalent, m).Incoming shortwave radiation measurements were obtained from the KAN_B PROMICE automatic weather station (67.1252N, 50.1832W; Figure 1) and converted to West Greenland Standard

F I G U R E 2
Snow depth for Sandflugtsdalen and regionally for 2018 and 2019.Positive degree days and the occurrence of dust events are indicated for each year.X indicates the start and end of the field measurement records in each year for this study.[Color figure can be viewed at wileyonlinelibrary.com]F I G U R E 3 Wind roses for all observations where dust concentration ≥10 μg m À3 at (a) the Ridge and (b) SS85.[Color figure can be viewed at wileyonlinelibrary.com]3.3| Relationship between dust concentration and meteorological variablesThe relationships between dust concentrations greater than background levels (≥10 μg m À3 ) and wind speed, wind direction, temperature and humidity are shown in Figure7.The highest dust concentrations were recorded between 24 and 28 July 2018 and are differentiated from other data in Figure7and discussed further below.For the Ridge, dust concentrations ≥10 μg m À3 are associated with wind speeds ranging from 0.25 to 13.4 m s À1 , but there is no significant statistical relationship between the two variables using the whole data set.For individual dust events (see Section 3.4), six had a statistically significant positive relationship between wind speed and PM 10 (at 95%), and one had a significant negative relationship (Table2).The very high dust concentrations recorded between 24 and 28 July 2018 are associated with winds 1-4 m s À1 .At SS85, most dust measurements are associated with winds <6 m s À1 .Dust concentrations ≥10 μg m À3 recorded at the Ridge are primarily associated with winds blowing from the north to east quadrant or from south to southwest.At SS85, dust concentrations ≥10 μg m À3 are associated with a wide range of wind directions.There is no statistically significant relationship between dust concentration and air temperature.Air temperature associated with dust concentrations ≥10 μg m À3 at the Ridge range from À4.5 C to +20 C and at SS85 from À3.8 C to +24.1 C. Air temperature during the high F I G U R E 4 Diurnal distributions of (a) wind speed, (b) temperature and (c) relative humidity for maximum length of record at each site for measurement periods indicated in Table 1.(d) Incoming shortwave radiation for KAN_B automatic weather station averaged for May, July and September 2018 and 2019 and averaged for April-October 2018 and 2019.[Color figure can be viewed at wileyonlinelibrary.com]F I G U R E 5 Diurnal distribution of wind direction averaged for 2018 and 2019 at (a) the Ridge and (b) SS85.F I G U R E 6 Diurnal distributions of dust concentrations >0 μg m À3 by site and year.Note different scale for dust concentrations at the Ridge in 2018.F I G U R E 7 Relationship between dust concentration (≥10 μg m À3 ) and (a) mean wind speed, (b) wind direction, (c) temperature and (d) relative humidity.[Color figure can be viewed at wileyonlinelibrary.com]T A B L E 2 Dust events (for definition see main text) recorded in the Kangerlussuaq region from 19/04/18 to 7/10/18 and 13/04/19 to 22/08/19.from 24 to 28 July 2018 was 3.1-3.8C. At the Ridge, other than 24-28 July 2018, dust concentrations ≥10 μg m À3 occur when relative humidity is 25-67% and at SS85 when it is 22-87%.During the 24-28 July 2018 event, relative humidity exceeded 98%.These observations occurred on days with showers or light rain.
Events in bold occurred on 'reversal days' sensu Kopec et al. (2014).Spearman's rank test correlations in bold are statistically significant at 95% level.a Variable winds.b DustTrak stopped recording at 05:50.3.4.2| Events 6 and 7: 24-28 July 2018 Following 8 days during which all recorded dust concentrations at the Ridge site were ≤2 μg m À3 , Events 6 and 7 represented the highest dust concentrations in the record by an order of magnitude.Between 23:00 on 24 July and 05:00 on 25 July, a series of ≥10 μg m À3 dust concentrations were recorded, the highest at 396 μg m À3

F
I G U R E 8 (a) Frequency of dust events recorded by month for the 2018 and 2019 monitoring periods; monthly mean dust concentration (both years) is also indicated; (b) diurnal variation in onset time of all dust events (see text for details).

|
Diurnal patterns of meteorological drivers and dust concentration Our observations from spring to autumn for 2 years in West Greenland show similar diurnal cycles in the environmental drivers of dust emission to those recorded in low latitudes (e.g.Stout, 2015; Kelley et al., 2020).Measurements were only made between April and F I G U R E 1 0 Summary meteorological and PM 10 data for Event 5 on 15-16 July 2018 for (a) Ridge and (b) SS85.Wind direction is given every 8 h.F I G U R E 1 1 Summary meteorological and PM 10 data for Events 6 and 7 recorded on 24 and 28 July 2018, at (a) Ridge and (b) SS85.Wind direction is given every 8 h.Note logarithmic scale for dust concentration in panel A. October and do not include the polar night where incoming solar radiation is minimal.Although the quantity of incoming solar radiation varies through the measurement period, it exhibits a clear diurnal cycle and the timing of the maximum is consistent.
It is difficult to compare measurements of dust concentration across different global locations because factors such as instrumentation, horizontal and vertical proximity to the emitting surface, sampling interval and the definition of 'dust event' will affect results, and all vary from study to study.Nevertheless, it is clear the dust concentrations recorded in West Greenland are of a similar order of magnitude, but at the lower end of the range, to source-proximal dust concentrations reported from elsewhere.The maximum dust concentration recorded over the 17 events reported here was ≥1000 μg m À3 , but the majority were <50 μg m À3 .In southern Iceland, Mockford et al. (2018) recorded nine events in 6 weeks with maximum (10 s) PM 10 concentrations of 869-8860 μg m À3 , and Dagsson-Waldhauserova, Arnalds, Olafsson, Skrabalova, et al. (2014) recorded a maximum concentration of 1757 μg m À3 .At lower latitudes, Stout (2015) recorded a peak hourly dust concentration of 13 984 μg m À3 (West Texas), Kimura & Shinoda (2010) recorded a maximum dust concentration of 5400 μg m À3 (1 min average; Mongolia) and Baddock et al. (2014) a maximum dust concentration of 260 μg m À3 (SE Australia).The average dust concentration for each event recorded for this study ranged from 12 to 292 μg m À3 , which is similar to that recorded by Csavina et al. (2014) in the SW USA (20-45 μg m À3 ) but lower than Mockford et al. (2018) where event averaged concentrations in Iceland ranged from 269 to 2570 μg m À3 .As observed at low latitudes (e.g.

Van
Soest et al. (2022) measured dust deposition in the Kangerlussuaq area from spring 2017 to spring 2019.They suggested that high meltwater discharge from the Watson River in 2016 (8.2 ± 1.3 km 3 ) provided large quantities of fine sediments that were entrained and deposited in the following winter and spring 2017.Reduced supply of fine sediments due to low meltwater discharge in 2017 (4.3 km 3 ) and 2018 (3.76 km 3 ) was suggested to contribute to low dust deposition in 2018 and spring 2019.Average annual dust AUTHOR CONTRIBUTIONSConceptualization and methodology: JEB, CP, MCB, NJA; funding acquisition: JEB, NJA; data collection: JEB, CP, MCB, NJA; data analysis: JEB, MCB; data curation: JEB, CP; writing initial draft: JEB; writing, reviewing and editing: JEB, MCB, CP, NJA.
. Annual discharge from the Watson À3).By comparison, mean dust concentration was very similar for both years at SS85 at 1.03 μg m À3 (range 0-44 μg m À3 ) in 2018 and 1.05 μg m À3 (range 0-44 μg m À3 ) in 2019.Mean wind speed (all winds) at the Ridge in 2018 was 2.86 ± 2.07 m s À1 (mean maximum wind speed 4.13 ± 2.99 m s À1 ), slightly lower than that recorded in 2019 that was 3.38 ± 1.96 m s À1 (mean maximum wind speed 4.7 ± 2.69 m s À1 ).Mean wind speed at SS85 was more consistent across the 2 years at 2.38 ± 2.17 m s À1 (mean maximum 3.33 ± 2.68 m s À1 ) in 2018 and 2.64 ± 1.79 m s À1 (mean maximum wind speed 3.77 ± 2.17 m s À1 ) in 2019.Figure3shows wind roses for each site for observations where dust concentrations are ≥10 μg m À3 .Wind direction at the Ridge is dominated by northeasterly winds that are strongly controlled by valley topography.Wind direction at SS85 is more variable.