The Spatial Variation of Dust Particulate Matter Concentrations during Two Icelandic Dust Storms in 2015

Particulate matter mass concentrations and size fractions of PM1, PM2.5, PM4, PM10, and PM15 measured in transversal horizontal profile of two dust storms in southwestern Iceland are presented. Images from a camera network were used to estimate the visibility and spatial extent of measured dust events. Numerical simulations were used to calculate the total dust flux from the sources as 180,000 and 280,000 tons for each storm. The mean PM15 concentrations inside of the dust plumes varied from 10 to 1600 μg ̈ m ́3 (PM10 = 7 to 583 μg ̈ m ́3). The mean PM1 concentrations were 97–241 μg ̈ m ́3 with a maximum of 261 μg ̈ m ́3 for the first storm. The PM1/PM2.5 ratios of >0.9 and PM1/PM10 ratios of 0.34–0.63 show that suspension of volcanic materials in Iceland causes air pollution with extremely high PM1 concentrations, similar to polluted urban areas in Europe or Asia. Icelandic volcanic dust consists of a higher proportion of submicron particles compared to crustal dust. Both dust storms occurred in relatively densely inhabited areas of Iceland. First results on size partitioning of Icelandic dust presented here should challenge health authorities to enhance research in relation to dust and shows the need for public dust warning systems.


Introduction
Air pollution from natural sources accounts for a significant part of the total particulate matter pollution on Earth.Deserts, stratovolcanoes, and arable land areas contribute to global air pollution in addition to emissions from industrialized and densely inhibited regions.Desert dust has a remarkable influence on Earth's ecosystems as well as human health.Several studies have shown that suspended desert dust can increase mortality hundreds of kilometers downwind from dust sources [1][2][3][4][5][6].Mortality was found to increase by 2%-12% with every 10 µg¨m ´3 increase in particulate matter (PM 10 ) concentration.Some of these studies reported increased mortality for the PM 2.5 dust particulate matter.A decrease of mixing layer height was associated with an increase of daily mortality while the effect of mixing layer thinning on particle toxicity was exacerbated when Saharan dust outbreaks occurred [7].Positive associations between mass concentrations of larger sizes of particles, such as PM 15 , were observed for cardiopulmonary and ischemic heart disease causes of death during the long-term studies on air pollution and mortality of the American Cancer Society [8].The studies on levels and speciation of PM 1 in Europe are, however, scarce [9][10][11].The PM 1 fraction has considerable importance in relation to health because of high potential for entering the lungs [12].Moreover, submicron particles are more likely to travel further distances during the long range transport than larger particles [13].
In spite of a cold and moist climate, Iceland has been identified as the most active and largest Arctic and European desert [14].Other cold climate and high latitudes regions with considerable dust inputs are Alaska, Greenland, Svalbard, Antarctica, and South America [15][16][17][18][19][20].One of the most extreme wind erosion events on Earth was measured in Iceland in 2010 [21].Annual dust day frequency in Iceland is comparable to the major desert areas of the world [22,23].Emissions from local dust sources, enhanced by strong winds, affect regional air quality in Iceland, such as in the capital Reykjavik [24].Particulate Matter (PM 10 ) concentrations during dust events in Reykjavik often exceed the health limit of 50 µg¨m ´3 over 24 h [25,26], while PM 10 concentrations measured during dust events in the vicinity of dust sources (<30 km) exceed the health limit in order of 10-100 times [26][27][28].This shows that atmospheric aerosols, mainly dust, can markedly impair air quality in non-polluted Arctic/sub-Arctic regions.
Icelandic dust differs from dust originating from most continental deserts.It is volcanogenic in origin, of basaltic composition, with lower SiO 2 proportions (<50%) and higher Al 2 O 3 , Fe 2 O 3 , and CaO contents than crustal dust [23,[27][28][29].This volcanic dust contains about 80% volcanic glass with numerous large gas bubbles and massive shards.It is extremely angular with sharp-tipped shards and often with curved and concave shard-faces.Fine pipe-vesicular structures of glass, as known from asbestos, can be also found.All these factors suggest that volcanic dust can be easily suspended and have highly negative effects on human health as concluded by Carlsen et al. [30].
Studies that provide PM mass concentration measurements during dust storms in Iceland are few and limited in scope, excluding research related to volcanic eruptions [23,24,27,28,31].None of these studies dealt with the size partitioning of the PM components.Here we presented a study on PM source characteristics of volcanic dust during two dust events from different dust sources in Iceland.The emphasis was given to the fine dust fraction of PM 1 .An effort has been made to measure the transverse horizontal profile of dust storms and estimate the spatial extent of such storms in the terms of PM concentrations, dust load computation, and visibility information obtained from cameras.

Instrumentation and Measurement Setup
Two dust storms were measured in southwestern Iceland in the summer of 2015.Measurements of both storms began after the dust plume was visible from Reykjavík with transverse horizontal profile measurements through the dust plumes.The source area of the first dust event on 15 June 2015 was Landeyjasandur (Storm 1, Figure 1A), about 100 km from Reykjavík, while the second dust storm on 4 August 2015 originated from the Hagavatn dust source (Storm 2, Figure 1B), about 85 km distance from Reykjavik.
A mobile instrument, aerosol monitor DustTrak DRX 8533EP, was used to measure particulate matter (PM) mass concentrations at several places within the dust plume.The DustTrak instrument provides measurements of mass concentration from 0.001 to 150 mg¨m ´3 with the mass fraction concentrations corresponding to PM 1 , PM 2.5 , PM 4 , PM 10 , and the total PM 15 .Five minute measurements were made at each stop in the dust plume, consisting of 60 five-second sampling periods.The measurement time was short to allow for travel through the dust storm with measurements at as many places as possible.All measurements were calculated as 5-min averages.The TSI DustTraks have given similar results in parallel measurements of PM mass concentration as Beta attenuation instruments (Thermo ESM Andersen FH 62 I-R) [32].This instrument has been used in Iceland by the Environmental Agency of Iceland (EAI) since 1996 and we have found good relation between PM 10 concentrations obtained by DustTrak DRX and the Thermo ESM.The error range for absolute values for other PM size ranges could be ˘10% [33].We emphasize that the DustTrak measurements in both storms are complemented by an independent measurement by the Thermo ESM instrument at Similar agreement was found for Storm 2. These numbers indicate that large scale errors sometimes reported for measurements with the TSI DustTraks do not apply here [32,33].
measurements in both storms are complemented by an independent measurement by the Thermo ESM instrument at the EAI in Reykjavik (PM10 30-min maxima of about 300 µ g•m −3 for Storm 1 and about 200 µg•m −3 for Storm 2).Measurements for Storm 1 with DustTrak at Keldnaholt (Reykjavik) and Thermo ESM at Grensásvegur (Reykjavik), approx.5 km apart are very similar (PM10 30-min maxima of about 224 µg•m −3 vs. 280 µg•m −3 at the same time, with the latter being closer to the center of the plume).Similar agreement was found for Storm 2. These numbers indicate that large scale errors sometimes reported for measurements with the TSI DustTraks do not apply here [32,33].
In this study, we used a unique dataset from a network of active cameras operated by the Icelandic Road and Coastal Administration to find the exact extent and area of the dust plume [34].A total of 25 cameras were used to estimate the visibility changes during these dust storms.Figure 2 presents the time laps images from the web cameras with corresponding visibility estimations.The houses on the photos are situated at about 1 km from the camera.The mountains in the background are about 3.5 km from the camera.Subsequently, maps of the dust storms were made in ArcMap 10.1.In this study, we used a unique dataset from a network of active cameras operated by the Icelandic Road and Coastal Administration to find the exact extent and area of the dust plume [34].A total of 25 cameras were used to estimate the visibility changes during these dust storms.Figure 2 presents the time laps images from the web cameras with corresponding visibility estimations.The houses on the photos are situated at about 1 km from the camera.The mountains in the background are about 3.5 km from the camera.Subsequently, maps of the dust storms were made in ArcMap 10.1.These numbers indicate that large scale errors sometimes reported for measurements with the TSI DustTraks do not apply here [32,33].
In this study, we used a unique dataset from a network of active cameras operated by the Icelandic Road and Coastal Administration to find the exact extent and area of the dust plume [34].A total of 25 cameras were used to estimate the visibility changes during these dust storms.Figure 2 presents the time laps images from the web cameras with corresponding visibility estimations.The houses on the photos are situated at about 1 km from the camera.The mountains in the background are about 3.5 km from the camera.Subsequently, maps of the dust storms were made in ArcMap 10.1.

Meteorological Conditions and Transport of Dust
Figure 2 shows the observed 10 min average wind speed at the weather stations Landeyjahöfn (Storm 1), which is close to the dust source, and Gullfoss (Storm 2) which is at a 15 km distance from the source, but does not capture the catabatic wind effects at the source.The wind speed was gradually increasing from about 6 to 16 m¨s ´1 during the Landaeyjasandur dust event (Storm 1).The Hagavatn dust event (Storm 2) occurred with wind speeds from about 4 to 14 m¨s ´1 measured at Gullfoss.A numerical simulation of surface winds during the two events is shown in Figure 3.The simulation was carried out with the numerical model HIRLAM with a horizontal resolution of 5 km.The simulation was initialized at 00UTC the same day, using initial and boundary conditions from the operational suite at the European Centre for Medium range Weather Forecasts.The true winds in the dust source area on 4 August were probably a few m¨s ´1 stronger than observed at Gullfoss.There are substantial horizontal gradients in the wind field close to the dust sources in both cases.In Storm 1 on 15 June, the mountains in S-Iceland generated a corner wind of 6-8 m¨s ´1 greater speed than in the incoming flow.This corner wind extended over a large area over the sea and the SW-coast of Iceland.The dust was lifted where this windstorm blew over the coastal areas.Once in suspension, some of the dust entered a wake area over land with weaker winds.In Storm 2 on 4 August, dust was also generated by locally enhanced winds.There was local acceleration in flow which ran along a major mountain range and had a downslope component.The local acceleration and the winds in general were underestimated by a few m¨s ´1 in the simulation (Figure 3), but the flow pattern and the wind directions are realistic.The orographic flow perturbation is of a smaller scale in Storm 2 on 4 August than in Storm 1 on 15 June, and consequently is harder to reproduce in a numerical simulation.

Meteorological Conditions and Transport of Dust
Figure 2 shows the observed 10 min average wind speed at the weather stations Landeyjahöfn (Storm 1), which is close to the dust source, and Gullfoss (Storm 2) which is at a 15 km distance from the source, but does not capture the catabatic wind effects at the source.The wind speed was gradually increasing from about 6 to 16 m•s −1 during the Landaeyjasandur dust event (Storm 1).The Hagavatn dust event (Storm 2) occurred with wind speeds from about 4 to 14 m•s −1 measured at Gullfoss.A numerical simulation of surface winds during the two events is shown in Figure 3.The simulation was carried out with the numerical model HIRLAM with a horizontal resolution of 5 km.The simulation was initialized at 00UTC the same day, using initial and boundary conditions from the operational suite at the European Centre for Medium range Weather Forecasts.The true winds in the dust source area on 4 August were probably a few m•s −1 stronger than observed at Gullfoss.There are substantial horizontal gradients in the wind field close to the dust sources in both cases.In Storm 1 on 15 June, the mountains in S-Iceland generated a corner wind of 6-8 m•s −1 greater speed than in the incoming flow.This corner wind extended over a large area over the sea and the SW-coast of Iceland.The dust was lifted where this windstorm blew over the coastal areas.Once in suspension, some of the dust entered a wake area over land with weaker winds.In Storm 2 on 4 August, dust was also generated by locally enhanced winds.There was local acceleration in flow which ran along a major mountain range and had a downslope component.The local acceleration and the winds in general were underestimated by a few m•s −1 in the simulation (Figure 3), but the flow pattern and the wind directions are realistic.The orographic flow perturbation is of a smaller scale in Storm 2 on 4 August than in Storm 1 on 15 June, and consequently is harder to reproduce in a numerical simulation.

Dust Concentrations and Visibility
The spatial extent and PM concentrations measured during the two dust storms are shown in Figure 1 and Table 1.The mean PM 15 concentrations inside the dust plume 1 varied from 162 µg¨m ´3 to 1260 µg¨m ´3 (PM 10 = 158-583 µg¨m ´3), and from 10 µg¨m ´3 to 1600 µg¨m ´3 (PM 10 = 7-486 µg¨m ´3) inside dust plume 2, respectively.These numbers represent rather low concentrations for an ongoing dust storm compared to the long term PM measurements provided by the Environmental Agency of Iceland (EAI) [23,[26][27][28].This was partly caused by moderate winds not exceeding 16 m¨s ´1 (Figures 2  and 3).The mean PM 1 concentrations were, however, 97-241 µg¨m ´3 during Storm 1 and reached up to 164 µg¨m ´3 during Storm 2. Such high fine dust concentrations have been reported during massive dust storms from lacustrine sediment areas in Iran and during African dust episodes in Barcelona (hourly means 60-70 µg¨m ´3) [10,35].The PM 1 maximum of 261 µg¨m ´3 measured during moderate Icelandic dust storms is comparable to the maximum of 495 µg¨m ´3 reported from Iran during a massive dust storm when PM 10 concentrations exceeded 5000 µg¨m ´3.Relatively high PM 1 annual means are regularly measured over Greece during African dust outbreaks [36].The distance of the measurements from the dust sources was <100 km.The source material contains extremely fine particles.Storm 1 originated from the Landeyjarsandur dust hot spot which mostly consists of fine volcanic material from active volcanic systems such as Eyjafjallajökull and the Katla systems [14].Figure 1B shows the values for Storm 2 which originated from the Hagavatn glacial floodplain.The Hagavatn dust materials are more crystalline in nature compared to most other Icelandic dust sources [29].The higher PM 1 concentrations during the Storm 1 than Storm 2 can be related to this difference in crystallinity and also to the early suspension in June rather than in August when submicron particles had been already removed.
The camera network from the Icelandic Road and Coastal Administration (IRCA) was used to determine the dispersal area of these two dust storms.Figure 4 depicts a time lapse series of photos from one of these web cameras which is located in Sandskeið near Reykjavík (Point 1 at the Figures 1B and 5B).The visibility was reduced down to 1 km during Storm 2. The camera is located about 3.5 km from the mountain and about 1 km from the house shown in the middle of the photos.The corresponding PM 10 concentrations to these visibility reductions can be obtained from DustTrak measurements close to this location for the left image only.The closest instrument (EAI) is located about 20 km downwind from this camera.The left picture shows the visibility was >3.5 km corresponding to the PM 10 of 50 µg¨m ´3 by the EAI and 71 µg¨m ´3 measured using the DustTrak instrument at the site.The middle photo shows the visibility was about 3.5 km, corresponding to the PM 10 of 70 µg¨m ´3, while the photo on the right shows visibility was about 1 km, corresponding to PM 10 of 100 µg¨m ´3.The PM 10 concentrations for the middle and right picture at the location of the camera can, therefore, only be retrieved using the visibility-dust formula given by Dagsson-Waldhauserova et al. [23].This formula is based on the long-term observations of PM 10 and visibility in Iceland.Applying the formula to the visibility estimations from the camera, the PM 10 concentrations are calculated as <190 µg¨m ´3 for the left photo, 370 µg¨m ´3 for the middle photo, and >780 µg¨m ´3 for the photo on the right.
Impaired visibility was observed at all spots where PM measurements were conducted (Figure 5).These images together with the IRCA camera network allowed us to estimate the spatial extent of the dust plumes as depicted in Figure 1.The total land area affected by Storm 1 was about 2450 km 2 but 4220 km 2 for Storm 2. Numerical simulations and operational radiosoundings at Keflavik revealed the thickness of the well-mixed atmospheric boundary-layer as about 1 km in Strom 1 on 15 June with mean winds of about 15 m¨s ´1.The event lasted for about 8 h, giving a total dust flux from source of about 180,000 tonnes.In Storm 2 on 4 August, the boundary-layer thickness was about 1.3 km and the mean winds were 12 m¨s ´1.This event lasted for about 12 h and the total flux from the source is estimated to be about 280,000 tonnes.Both events can thus be characterized as medium-sized (e.g., [22,37]).
Atmosphere 2016, 7, 77 6 of 12 The distance of the measurements from the dust sources was <100 km.The source material contains extremely fine particles.Storm 1 originated from the Landeyjarsandur dust hot spot which mostly consists of fine volcanic material from active volcanic systems such as Eyjafjallajökull and the Katla systems [14].Figure 1B shows the values for Storm 2 which originated from the Hagavatn glacial floodplain.The Hagavatn dust materials are more crystalline in nature compared to most other Icelandic dust sources [29].The higher PM1 concentrations during the Storm 1 than Storm 2 can be related to this difference in crystallinity and also to the early suspension in June rather than in August when submicron particles had been already removed.
The camera network from the Icelandic Road and Coastal Administration (IRCA) was used to determine the dispersal area of these two dust storms.Figure 4 depicts a time lapse series of photos from one of these web cameras which is located in Sandskeið near Reykjavík (Point 1 at the Figures 1B and 5B).The visibility was reduced down to 1 km during Storm 2. The camera is located about 3.5 km from the mountain and about 1 km from the house shown in the middle of the photos.The corresponding PM10 concentrations to these visibility reductions can be obtained from DustTrak measurements close to this location for the left image only.The closest instrument (EAI) is located about 20 km downwind from this camera.The left picture shows the visibility was >3.5 km corresponding to the PM10 of 50 µ g•m −3 by the EAI and 71 µ g•m −3 measured using the DustTrak instrument at the site.The middle photo shows the visibility was about 3.5 km, corresponding to the PM10 of 70 µ g•m −3 , while the photo on the right shows visibility was about 1 km, corresponding to PM10 of 100 µ g•m −3 .The PM10 concentrations for the middle and right picture at the location of the camera can, therefore, only be retrieved using the visibility-dust formula given by Dagsson-Waldhauserova et al. [23].This formula is based on the long-term observations of PM10 and visibility in Iceland.Applying the formula to the visibility estimations from the camera, the PM10 concentrations are calculated as <190 µ g•m −3 for the left photo, 370 µ g•m −3 for the middle photo, and >780 µ g•m −3 for the photo on the right.Impaired visibility was observed at all spots where PM measurements were conducted (Figure 5).These images together with the IRCA camera network allowed us to estimate the spatial extent of the dust plumes as depicted in Figure 1.The total land area affected by Storm 1 was about 2450 km 2 but 4220 km 2 for Storm 2. Numerical simulations and operational radiosoundings at Keflavik revealed the thickness of the well-mixed atmospheric boundary-layer as about 1 km in Strom 1 on 15 June with mean winds of about 15 m•s −1 .The event lasted for about 8 h, giving a total dust flux from source of about 180,000 tonnes.In Storm 2 on 4 August, the boundary-layer thickness was about 1.3  Figure 6 captured how Storm 1 passed relatively densely populated areas such as the town of Hveragerði, population of 2300, and the neighboring town of Selfoss, the eighth largest community in Iceland, with about 6500 inhabitants.The highest PM10 concentrations of >500 µ g•m −3 and PM1 > 200 µ g•m −3 were measured in this area.The margin of the dust plume is very visible on Figure 6A and it can clearly be seen how the visibility changed due to dust in the dust plume.Reduced visibility due to dust in Mosfellsbae r (near Reykjavik) is shown on picture 6B.Here the concentrations exceeded 400 µ g•m −3 for PM10 and 100 µ g•m −3 for PM1.The long-term frequency of dust events in Iceland reports about one dust day annually for the capital of Reykjavik [23].This number of dust storms is highly underestimated judging from our own observations as well as the measurements provided by the EAI.The meteorological stations at the towns of Hveragerði and Selfoss report 3.7 to 6.8 dust days a year.6A and it can clearly be seen how the visibility changed due to dust in the dust plume.Reduced visibility due to dust in Mosfellsbaer (near Reykjavik) is shown on picture 6B.Here the concentrations exceeded 400 µg¨m ´3 for PM 10 and 100 µg¨m ´3 for PM 1 .The long-term frequency of dust events in Iceland reports about one dust day annually for the capital of Reykjavik [23].This number of dust storms is highly underestimated judging from our own observations as well as the measurements provided by the EAI.The meteorological stations at the towns of Hveragerði and Selfoss report 3.7 to 6.8 dust days a year.

Size Partitioning of the PM Components of Icelandic Dust
Mineral dust outbreaks increase both fine and coarse PM concentrations [3,6,10,13,24,38,39]. Table 1 shows the mean PM1-15 concentrations at different locations (Figure 1) inside the dust plume.Although the PM10 concentrations are moderate (<600 µ g•m −3 ) for an ongoing dust storm in Iceland, the mean levels of PM1 are considerably high, such as >97 µ g•m −3 for Storm 1.The proportions of PM1/PM10 are significantly higher for Iceland than for any other dust events we have found in the literature.The PM1/PM10 ratio ranged from 0.41 to 0.63 for Storm 1, and 0.34 to 0.57 for Storm 2, respectively (Table 1).Perez et al. [9] reported the PM1/PM10 ratios were relatively stable during African dust outbreaks, where the mean annual average decreased from 0.48 to 0.42 for dust days.Arizona dust outbreaks also had even more stable PM1/PM10 ratios of between 0.17 and 0.22 (0.18 on average) [38].Claiborn et al. [39] reported no increase in PM1 during dust storms within the USA.In Iceland, the more severe the dust event was the lower PM1/PM10 ratio was observed.The same trend was reported from Iran where the PM1/PM10 ratio decreased from 0.4 to 0.05 during the dust storms with the mean ratio of 0.14.Generally, the PM1/PM10 ratio <0.4 is attributed to the summer season with high dust suspension as summarized from 22 studies on size-segregated particulate matter ratios [40].These comparisons showed that Icelandic volcanic dust is extremely fine compared to the crustal dust.Such high proportions (>60% of PM1 in PM10) as obtained during Storm 1 have been reported for urban air pollution, but not for natural dust.The PM1 proportion of 57%-60% in PM10 was found, for example, at four sites in Austria [41] while the PM1/PM10 ratio between 0.45 and 0.74 was found at a polluted urban site of Taipei in Taiwan [42].
Table 1 shows that the PM1/PM2.5 ratio ranged from 0.89 to 0.94, thereby confirming that most of the fine dust particles were of submicron size.This is contrary to what has been reported on such ratios during dust events elsewhere.The PM1/PM2.5 ratio was 0.49 during dust events in the USA [39] while the PM1/PM2.5 was ranging from 0.05 to 0.8, with a corresponding mean value 0.55, during dust storms in Iran [10].Values such as 90% of PM1/PM2.5 were reported from urban sites or cities such Graz in Austria [41].High PM1/PM2.5 and PM1/PM10 ratios imply that PM2.5 or PM10 mainly consist of submicron particles that have a greater health impact than larger particles [6,10,12].Detailed mineralogical and geochemical analyses of Icelandic dust revealed fine pipe-vesicular structures of volcanic glasses, as known from asbestos and high content of heavy metals [27,28].Such structured submicron particles can likely have even more destructive effects on human and animal health, as reported by Carlsen et al. [30].Figures 1 and 6 illustrate that the dust plumes with high PM concentrations are passing the most densely inhibited areas in Iceland.The frequency of such events is up to 135 dust days annually in Iceland with many crossing populated areas [23].Currently, no warnings for the general public are issued.

Size Partitioning of the PM Components of Icelandic Dust
Mineral dust outbreaks increase both fine and coarse PM concentrations [3,6,10,13,24,38,39]. Table 1 shows the mean PM 1-15 concentrations at different locations (Figure 1) inside the dust plume.Although the PM 10 concentrations are moderate (<600 µg¨m ´3) for an ongoing dust storm in Iceland, the mean levels of PM 1 are considerably high, such as >97 µg¨m ´3 for Storm 1.The proportions of PM 1 /PM 10 are significantly higher for Iceland than for any other dust events we have found in the literature.The PM 1 /PM 10 ratio ranged from 0.41 to 0.63 for Storm 1, and 0.34 to 0.57 for Storm 2, respectively (Table 1).Perez et al. [9] reported the PM 1 /PM 10 ratios were relatively stable during African dust outbreaks, where the mean annual average decreased from 0.48 to 0.42 for dust days.Arizona dust outbreaks also had even more stable PM 1 /PM 10 ratios of between 0.17 and 0.22 (0.18 on average) [38].Claiborn et al. [39] reported no increase in PM 1 during dust storms within the USA.In Iceland, the more severe the dust event was the lower PM 1 /PM 10 ratio was observed.The same trend was reported from Iran where the PM 1 /PM 10 ratio decreased from 0.4 to 0.05 during the dust storms with the mean ratio of 0.14.Generally, the PM 1 /PM 10 ratio <0.4 is attributed to the summer season with high dust suspension as summarized from 22 studies on size-segregated particulate matter ratios [40].These comparisons showed that Icelandic volcanic dust is extremely fine compared to the crustal dust.Such high proportions (>60% of PM 1 in PM 10 ) as obtained during Storm 1 have been reported for urban air pollution, but not for natural dust.The PM 1 proportion of 57%-60% in PM 10 was found, for example, at four sites in Austria [41] while the PM 1 /PM 10 ratio between 0.45 and 0.74 was found at a polluted urban site of Taipei in Taiwan [42].
Table 1 shows that the PM 1 /PM 2.5 ratio ranged from 0.89 to 0.94, thereby confirming that most of the fine dust particles were of submicron size.This is contrary to what has been reported on such ratios during dust events elsewhere.The PM 1 /PM 2.5 ratio was 0.49 during dust events in the USA [39] while the PM 1 /PM 2.5 was ranging from 0.05 to 0.8, with a corresponding mean value 0.55, during dust storms in Iran [10].Values such as 90% of PM 1 /PM 2.5 were reported from urban sites or cities such Graz in Austria [41].High PM 1 /PM 2.5 and PM 1 /PM 10 ratios imply that PM 2.5 or PM 10 mainly consist of submicron particles that have a greater health impact than larger particles [6,10,12].Detailed mineralogical and geochemical analyses of Icelandic dust revealed fine pipe-vesicular structures of volcanic glasses, as known from asbestos and high content of heavy metals [27,28].Such structured submicron particles can likely have even more destructive effects on human and animal health, as reported by Carlsen et al. [30].Figures 1 and 6 illustrate that the dust plumes with high PM concentrations are passing the most densely inhibited areas in Iceland.The frequency of such events is up to 135 dust days annually in Iceland with many crossing populated areas [23].Currently, no warnings for the general public are issued.About 90% of the PM 2.5 particles were attributed to submicron particle fractions, and the ratios of PM 2.5 /PM 10 were similar to the PM 1 /PM 10 (Table 1).The mean PM 2.5 /PM 10 ratio was 0.61 for Storm 1 and 0.44 for Storm 2. This is similar to what was found during dust outbreaks in Spain [9], but considerably higher than the 0.3 ratio reported from the USA [39].The PM 2.5 /PM 10 ratios in Iran ranged from 0.1 to 0.5 with the mean of 0.23 [10].The lower values were related to the high dust season.The high proportion of fine materials in Iceland is attributed to the small grain size at the dust sources, which is a result of glacial action producing fine-grained materials.These materials are further sorted by glacio-fluvial processes at the glacial margins and in glacial rivers [14].This unceasing glacial and glacio-fluvial action ensures continuous re-supply of the fine grained materials, in contrast to larger aeolian bodies on the continents.
The high ratios of PM 10 /PM 15 shown in Table 1 suggest a low proportion of particles >10 µm.However, <50% proportion of PM 10 on PM 15 in many cases suggest the presence of larger particles.We did not find any relation between the PM 10 concentration and PM 10 /PM 15 ratio.There is, however, limited information on this size range of PM in the literature [8].
Our results show that Icelandic dust contains fine submicron particles, as was reported by Dagsson-Waldhauserova et al. [27].Suspended dust measured at the Maelifellssandur glacial floodplain resulted in the high proportion of close-to-ultrafine particles, such as 0.3-0.37 µm.Generation of such fine particles is associated with mechanical processes of glaciers and fluvial processes [27].However, even the finest lacustrine sediments produced from the most active dust hot spots of the world, such as the Bodele Depression in the Sahara or the Sistan Basin in Iran do not consist of such high amounts of submicron particles as observed in Icelandic dust [3,10,13,43,44].

Conclusions
The study of two dust events shows the usefulness of combining photos obtained by surveillance cameras and portable dust measurement instruments for identifying the extent, magnitude, and grain size characteristics of single dust storms in Iceland.It shows that common dust storms are of several hundred thousand tons of magnitude from relatively well-defined main dust sources.The in situ measurements of two moderate Icelandic dust storms in 2015 show that aeolian dust can be very fine.The study highlights that suspended volcanic dust in Iceland can have extremely high PM 1 concentrations that are comparable to urban pollution in Europe or Asia.The PM 1 /PM 2.5 ratios are generally low during dust storms outside of Iceland, much lower than >0.9 and PM 1 /PM 10 ratios of 0.34-0.63found in our study.The extremely high proportions of submicron particles are predicted to travel long distances.Moreover, such submicron particles pose considerable health risks because of their high potential for entering the lungs.Icelandic volcanic glass often has fine pipe-vesicular structures, known from asbestos, and has a high content of heavy metals.The two dust events with high PM concentrations reported here passed the most densely inhabited areas of Iceland and influenced an area of 2450 km 2 during Storm 1 and 4220 km 2 in Storm 2. The mean frequency of dust events in Iceland is about 135 dust days annually, however, health risk warnings for the general public are not being issued.The data provided stresses the need for such a warning system and is an important step towards its development.In light of the small size of the dust reported here, in addition to the high frequency of the dust events, it is vital to step up integrated dust and health research in Iceland.Furthermore, dust has influence on weather and climate in general, and the fine fraction of the Icelandic dust has bearings for weather forecast and climate change predictions.
Reykjavik (PM 10 30-min maxima of about 300 µg¨m ´3 for Storm 1 and about 200 µg¨m ´3 for Storm 2).Measurements for Storm 1 with DustTrak at Keldnaholt (Reykjavik) and Thermo ESM at Grensásvegur (Reykjavik), approx.5 km apart are very similar (PM 10 30-min maxima of about 224 µg¨m ´3 vs. 280 µg¨m ´3 at the same time, with the latter being closer to the center of the plume).

Figure 1 .
Figure 1.Spatial location of the dust storms.Left figure (A, Storm 1) shows the dust storm from Landeyjasandur on 15 June 2015.The measurement locations are marked on the map with the numbers corresponding to the information on the PM15 and PM1 concentrations in the table.Right figure (B, Storm 2) shows the dust storm from the Hagavatn on 4 August 2015.The lines on the figures depict the dispersal area of the dust plumes estimated from the images of the Icelandic Road and Coastal Administration web camera network [34].

Figure 2 .
Figure 2. Observations of mean 10 min wind speed from 01-24 UTC at Landeyjahöfn (at the dust source for Storm 1) on 15 June 2015 and at Gullfoss (about 15 km SE of the dust source for Storm 2) on 4 August 2015.

Figure 1 .
Figure 1.Spatial location of the dust storms.Left figure (A, Storm 1) shows the dust storm from Landeyjasandur on 15 June 2015.The measurement locations are marked on the map with the numbers corresponding to the information on the PM 15 and PM 1 concentrations in the table.Right figure (B, Storm 2) shows the dust storm from the Hagavatn on 4 August 2015.The lines on the figures depict the dispersal area of the dust plumes estimated from the images of the Icelandic Road and Coastal Administration web camera network [34].
storms are complemented by an independent measurement by the Thermo ESM instrument at the EAI in Reykjavik (PM10 30-min maxima of about 300 µ g•m −3 for Storm 1 and about 200 µg•m −3 for Storm 2).Measurements for Storm 1 with DustTrak at Keldnaholt (Reykjavik) and Thermo ESM at Grensásvegur (Reykjavik), approx.5 km apart are very similar (PM10 30-min maxima of about 224 µg•m −3 vs. 280 µg•m −3 at the same time, with the latter being closer to the center of the plume).Similar agreement was found for Storm 2.

Figure 1 .
Figure 1.Spatial location of the dust storms.Left figure (A, Storm 1) shows the dust storm from Landeyjasandur on 15 June 2015.The measurement locations are marked on the map with the numbers corresponding to the information on the PM15 and PM1 concentrations in the table.Right figure (B, Storm 2) shows the dust storm from the Hagavatn on 4 August 2015.The lines on the figures depict the dispersal area of the dust plumes estimated from the images of the Icelandic Road and Coastal Administration web camera network [34].

Figure 2 .
Figure 2. Observations of mean 10 min wind speed from 01-24 UTC at Landeyjahöfn (at the dust source for Storm 1) on 15 June 2015 and at Gullfoss (about 15 km SE of the dust source for Storm 2) on 4 August 2015.

Figure 2 .
Figure 2. Observations of mean 10 min wind speed from 01-24 UTC at Landeyjahöfn (at the dust source for Storm 1) on 15 June 2015 and at Gullfoss (about 15 km SE of the dust source for Storm 2) on 4 August 2015.

Figure 3 .
Figure 3. Numerical simulations of 10 m winds (m•s −1 ) over Iceland at 15 UTC on (A) Storm 1 on 15 June 2015 and (B) Storm 2 on 4 August 2015.The locations of the dust sources are indicated with a circle.
).The mean PM1 concentrations were, however, 97-241 µg•m −3 during Storm 1 and reached up to 164 µg•m −3 during Storm 2. Such high fine dust concentrations have been reported during massive dust storms from lacustrine sediment areas in Iran and during African dust episodes in Barcelona (hourly means 60-70 µg•m −3 ) [10,35].The PM1 maximum of 261 µg•m −3 measured during moderate Icelandic dust storms is comparable to the maximum of 495 µg•m −3 reported from Iran during a massive dust storm when PM10 concentrations exceeded 5000 µg•m −3 .Relatively high PM1 annual means are regularly measured over Greece during African dust outbreaks [36].

Figure 3 .
Figure 3. Numerical simulations of 10 m winds (m¨s ´1) over Iceland at 15 UTC on (A) Storm 1 on 15 June 2015 and (B) Storm 2 on 4 August 2015.The locations of the dust sources are indicated with a circle.

Figure 4 .
Figure 4. Changes in visibility during Storm 2 on 4 August 2015 when dust was blowing from Hagavatn.Left photo (A) shows visibility >3.5 km, corresponding to PM10 of 71 µ g•m −3 , measured by a DustTrak instrument at the site.The middle photo (B) shows visibility of about 3.5 km corresponding to an estimated PM10 of 370 µ g•m −3 , while the photo on the right (C) shows about 1 km visibility, corresponding to an estimated PM10 of >780 µ g•m −3 based on the formula from Dagsson-Waldhauserova et al. [23].The images are from a web camera from the Icelandic Road and Coastal Administration.

Figure 4 .
Figure 4. Changes in visibility during Storm 2 on 4 August 2015 when dust was blowing from Hagavatn.Left photo (A) shows visibility >3.5 km, corresponding to PM 10 of 71 µg¨m ´3, measured by a DustTrak instrument at the site.The middle photo (B) shows visibility of about 3.5 km corresponding to an estimated PM 10 of 370 µg¨m ´3, while the photo on the right (C) shows about 1 km visibility, corresponding to an estimated PM 10 of >780 µg¨m ´3 based on the formula from Dagsson-Waldhauserova et al. [23].The images are from a web camera from the Icelandic Road and Coastal Administration.

Figure 5 .
Figure 5.The measurement spots (place number: PM15 concentration/PM1 concentration, µ g•m −3 ) including a photo from every measurement spot.It can be seen that visibility was more reduced with higher dust concentrations.(A) Storm 1 from the Landeyjasandur on 15 June 2015 and (B) from Storm 2 on 4 August 2015.

Figure 5 .
Figure 5.The measurement spots (place number: PM 15 concentration/PM 1 concentration, µg¨m ´3) including a photo from every measurement spot.It can be seen that visibility was more reduced with higher dust concentrations.(A) Storm 1 from the Landeyjasandur on 15 June 2015 and (B) from Storm 2 on 4 August 2015.

Figure 6
Figure 6 captured how Storm 1 passed relatively densely populated areas such as the town of Hveragerði, population of 2300, and the neighboring town of Selfoss, the eighth largest community in Iceland, with about 6500 inhabitants.The highest PM 10 concentrations of >500 µg¨m ´3 and PM 1 > 200 µg¨m ´3 were measured in this area.The margin of the dust plume is very visible on Figure6Aand it can clearly be seen how the visibility changed due to dust in the dust plume.Reduced visibility due to dust in Mosfellsbaer (near Reykjavik) is shown on picture 6B.Here the concentrations exceeded 400 µg¨m ´3 for PM 10 and 100 µg¨m ´3 for PM 1 .The long-term frequency of dust events in Iceland reports about one dust day annually for the capital of Reykjavik[23].This number of dust storms is highly underestimated judging from our own observations as well as the measurements provided by

Figure 6 .
Figure 6.(A) The dust front of Storm 1 approaching to the town Hveragerði, population of 2300, which is near the town Selfoss, the eighth largest community in Iceland, with about 6500 inhabitants.The highest PM concentrations were measured in this area.The margin of the dust plume is very clear.Reduced visibility due to dust in Mosfellsbae r (near Reykjavik), overlooking river Leirá, is shown on the right picture (B).Here the concentrations exceeded 400 µ g•m −3 for PM10 and 100 µ g•m −3 for PM1.

Figure 6 .
Figure 6.(A) The dust front of Storm 1 approaching to the town Hveragerði, population of 2300, which is near the town Selfoss, the eighth largest community in Iceland, with about 6500 inhabitants.The highest PM concentrations were measured in this area.The margin of the dust plume is very clear.Reduced visibility due to dust in Mosfellsbaer (near Reykjavik), overlooking river Leirá, is shown on the right picture (B).Here the concentrations exceeded 400 µg¨m ´3 for PM 10 and 100 µg¨m ´3 for PM 1 .

Table 1 .
Particulate matter concentrations PM 1-15 (µg¨m ´3) for both storms.Ratios between different PM values are given.