Saharan dust and giant quartz particle transport towards Iceland

Mineral dust emissions from Saharan sources have an impact on the atmospheric environment and sedimentary units in distant regions. Here, we present the first systematic observations of long-range Saharan dust transport towards Iceland. Fifteen Saharan dust episodes were identified to have occurred between 2008 and 2020 based on aerosol optical depth data, backward trajectories and numerical models. Icelandic samples from the local dust sources were compared with deposited dust from two severe Saharan dust events in terms of their granulometric and mineralogical characteristics. The episodes were associated with enhanced meridional atmospheric flow patterns driven by unusual meandering jets. Strong winds were able to carry large Saharan quartz particles (> 100 µm) towards Iceland. Our results confirm the atmospheric pathways of Saharan dust towards the Arctic, and identify new northward meridional long-ranged transport of giant dust particles from the Sahara, including the first evidence of their deposition in Iceland as previously predicted by models.


Results
Fifteen (#1-15) Saharan dust events were identified in Iceland between 1 January 2008 and 29 February 2020. A systematic combined analysis of Moderate Resolution Imaging Spectroradiometer (MODIS) aerosol optical depth (AOD) data and Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) data resulted in the identification of nine of these fifteen dust events. These nine episodes matched distinct AOD peaks (Fig. 1). The remaining six episodes were determined based on the numerical simulations of the NASA's Modern-Era Retrospective analysis for Research and Applications, Version 2 (MERRA-2) and Barcelona Supercomputing Center (BSC), and verified using satellite images and backward trajectories of the HYSPLIT model.
All 15 dust episodes were independently examined by analyses of Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) vertical aerosol subtype profiles. The presence of dust over Iceland was confirmed by the lidar data for all 15 episodes. Among the events, two rare wet depositional episodes were detected; in these cases, the sampled Saharan dust materials were analysed using automated static image analysis.

Meteorology of the Saharan dust events #1-15.
Based on geopotential height maps and wind vectors at 700 hPa, two different synoptic meteorological situations were distinguished (Fig. 2). The Omega blocklike pattern determined the Saharan dust transport of nine episodes. The high-pressure centre of the blocking system was located over Western-South-western Europe. The steepest pressure gradients and connected strongest flows were situated at the southwest side of the anticyclonic system.
Synoptic patterns of the remaining six events were associated with a low-pressure system situated over (or close) to the Iberian Peninsula. The highest wind speeds were observed at the foreside of the stationary cyclone. Central and Eastern European blocking systems were also responsible for the development of the extended meridional atmospheric 'conveyor belt' .
Dust transport pathways were mapped using back trajectories calculated at different heights from Iceland (Fig. 3). The two main types of synoptic patterns defined two specific transport routes: one directly over the Atlantic Ocean and another over Europe (Spain, France, and the British Isles). Typical trajectory lengths were ~ 4500 to 5000 km, with possible source areas of the dust material located in Northwest Africa. Daily surface observations of the United States Naval Research Laboratory (Monterey aerosol page: https:// www. nrlmry. navy. mil/ aeros ol/# aeros olobs ervat ions) allowed a more detailed source appointment. The most important dust sources can be regarded as NW African dust hot spots 68 , like (1) the ancient lakebed deposits of the Taoudenni Basin, (2) large alluvial fans and extensive wadi-system on the western and north-western slopes of the Hoggar Mountains (Ahaggar Plateau and foothills), (3) intramountain sediments of the Atlas Mountains, and (4) coastal sebkhas of Western Africa. Figs. 1,  2, 3). The general meteorological background and related transport routes, as well as the possible source areas of the two depositional events, in Iceland were very similar. The first deposition event (SDE #1) occurred during the last days of March 2014, when a low-pressure system governed the atmospheric patterns of the eastern basin of the North Atlantic. The cyclonal flow of the system reached Northwest Africa, where the steep pressure gradient between the mid-latitude low and Saharan highs triggered sand and dust storms. A high amount of lofted mineral dust was transported northwards along the eastern flank of the stationary atmospheric trough system blocked by the Central European ridge. Dust material reached the atmosphere of France on 28 March 2014 and consequently Great Britain and Ireland on 29 March 2014. Intense wet deposition of dusty material and 'blood rain' events were reported for the affected areas 58,69 . Atmospheric dust was deposited in Iceland on 3 April 2014, as shown by wet deposition in the Icelandic domain in the dust numerical simulations of the NMMB/BSC (Nonhydrostatic Multiscale Model/Barcelona Supercomputing Center) (Fig. 4a). Precipitation did not affect all parts of Iceland, and samples collected in Keldnaholt were dry. Particulate matter (PM 10  www.nature.com/scientificreports/ of North Africa, and the cyclonal flow lofted enormous mass of desert dust into the atmosphere. In the following few days, the atmosphere from Africa through Crete over Europe to Iceland was loaded with an enormous amount of Saharan dust, reaching a PM 10 concentration of > 150 µg m −370 . The blocking mechanism of an Eastern European anticyclone was responsible for the stationary behaviour of the atmospheric system. Saharan dust    The forecasts of the BSC for dust load and wet deposition clearly showed the north(west)ward Saharan dust transport governed by the previously presented meteorological flow patterns (Fig. 4b). Mineral dust material from the foreside of the Atlas Mountains reached high latitudes across Western and North-western Europe over a 4-day period (the exact amount of deposited dust can only be roughly estimated from the models). Particulate matter (PM 10

Granulometry of SDE samples compared to local Icelandic dust source samples.
A detailed laboratory analysis of the samples was necessary to distinguish between local and remotely transported particulate matter and to exclude possible local dust addition and contamination of the deposited dust material. An automated static optical image analysis of several thousand individual particles was performed to obtain a robust granulometric characterisation based on the distribution curves of grain size and shape (Fig. 5). The grain size distribution curves of deposited material during the SDEs in Iceland showed clear unimodal distributions, which fell predominantly into the coarse silt (20-62.5 µm) and (very) fine sand (62.5-(125)-250 µm) particle size fractions. The modal grain sizes, median diameters, and mean diameters D [3,4] were 62.9 µm and 81.2 µm, 65.7 µm and 87.0 µm, and 70.6 µm and 89.3 µm for SDE #1 and SDE #2, respectively.
The circle-equivalent (CE) diameters of bulk samples from Icelandic sources were diverse. The grain sizes of samples from Landeyjarsandur and Mýrdalssandur fell into the coarse silt/fine sand fractions, whereas the particles of other samples were much smaller, especially the samples of very fine-grained deposits from Dyngjusandur and Maelifellssandur.
According to previous studies (e.g. Varga et al. 71 ), the grain shape parameters of automated static optical image analysis are size-dependent variables (direct comparison of shape parameters of particles with significantly different sizes do not provide realistic results). Thus, particles in the size ranges of the SDE samples (i.e. 20-250 µm) were selected to create adequate conditions for a representative particle shape comparison. The SDE samples exhibited a more convex and circular particle shape compared to the local Icelandic samples. The general patterns of the aspect ratio distribution curves were similar, but the solidity curves clearly indicated major differences among the different sample groups. The SDE samples and Landeyjarsandur (a sandy beach that is not directly affected by glacial processes) sample presented similar curves.
Cluster analyses of different size and shape parameters were also performed ( Fig. 6). Based on the grain size distribution and solidity shape parameter, the SDE samples were clustered close to Landeyjarsandur, whereas other Icelandic dust hotspots consisted of significantly finer grains. The analyses of other major shape parameters (aspect ratio, convexity, and circularity) resulted in various trends for different materials, with the solidity parameters showing a similar clustering for the same particle size-based groups of material. In general, the SDE samples were closely clustered, whereas the Icelandic samples were not; however, there were similarities between Raman spectroscopy also confirmed the differences between the materials from SDEs and local sources. Due to the high percentage of particles with nonapplicable Raman signals, robust quantitative analyses could not be made. Our general findings, however, clearly showed that Icelandic surface samples obviously reflected the local basaltic environment, and that no quartz particles were found in these samples. Quartz particles, including particles larger than 100 µm, were identified in large numbers in both SDE samples.

Discussion
Enhanced atmospheric meridionality and climate change. The surface temperature anomaly during the two SDEs compared to the climatological mean for the period from 1981 to 2010 (by applying the reanalysis data of the National Center for Environmental Protection (NCEP) and National Center for Atmospheric Research (NCAR) 72 ) was 3.1 °C. Enhanced meridionality of the upper-level atmospheric flow was the main synoptic driver of the warm incursions and the identified Saharan dust events. The importance of a more meandering polar jet and associated meridional flow patterns during the formation of the North African cyclone and poleward dust transport was discussed by Francis et al. (2018). According to the authors' findings, the intense warming of Arctic regions reduces the temperature difference between the high-and mid-/low-latitude areas, thus leading to increasing planetary wave amplitudes and more meridional flow patterns at high altitudes 73 .
The southward propagation of the upper-level atmospheric trough and the orographic blocking of the Atlas Mountains play vital roles in the formation of severe surface wind storms and dust entrainment in the northwest regions of the Sahara 74 . The area of these Northwest African dust hotspots was identified as the main source of the discussed Icelandic dust events; hence, the Icelandic samples were collected approximately 4000-4500 km from their potential Saharan sources.
Long-range meridional transport of giant Saharan dust. In the recent years several studies were published on long-ranged giant dust transport. It has been shown that coarse particles with a diameter of > 30 µm and giant particles with a diameter of > 100 µm are capable of long-range transport within the Saharan air layer [75][76][77][78][79] . Previous studies focused on the zonal (east-to-west) transport of giant Saharan dust, whereas meridional (south-to-north) transport towards higher latitudes has only been mentioned in some recent papers 62,80 . In particular, aerosol studies on the long-range transport of coarse and giant particles to the Arctic are generally lacking 81 . Here, we provide the first evidence of giant Saharan particles deposited in Iceland. Our results confirm the existence of established atmospheric pathways of Saharan dust towards the Arctic, with Saharan dust contributing ~ 32% of the total atmospheric dust load in the Arctic 59 . Several studies have reported that dust deposition was poorly simulated by climate models and other numerical simulations 82,83 , primarily due to the false parametrization of the grain size of dust particles. There is a general bias in the size distribution of atmospheric dust, which is skewed towards finer size fractions 79,82,84,85 . Underestimation of the transport and deposition of these coarse mineral particles leads to further biases in the derived climatic effects (and other environmental effects of) aeolian dust (e.g. radiative forcing, which possibly causes net atmospheric warming, and a role in biogeochemical cycles and soil formation). So, the presented size data also has a climatic relevance as the interpretation questions of coarse-sized windblown dust are standing in the focal point of recent studies. Climatic impacts of mineral dust are also sensitive to particle size and shape. Radiative effects of coarse mineral grains in the shortwave spectrum is determined by the lower single scattering albedo compared to finer particles, causing more absorption of solar radiation and so, heating of the atmosphere 86,87 . Ryder et al. 87 reported an increased atmospheric heating by up to a factor of three when as a result of calculated single scattering albedo dropping from 0.92 to 0.8 when larger particles were taken into account. Kalashnikova and Sokolik 88 showed that irregular shape character of windblown mineral particles also have an effect on radiative properties via larger surface-to-volume ratio relative to that of an equal volume sphere. According to the calculations of Kok et al. 86 , average global dust extinction efficiency results are underestimated by 20-60% for larger than 1 micron particles.  (2) collected mineral material of the SDEs showed different characteristics. These differences in the two sediment groups were also confirmed by cluster analysis. Clustering of the shape parameters of the SDE and local surface samples showed an outlying nature of the particles from the washout samples. The washout samples mainly contained coarse silt and fine sand-sized mineral dust material, which could be clearly distinguished by their different size and shape parameters in comparison to the samples from local source areas [89][90][91] . The more circular and convex shape along with higher solidity values indicated a more mature nature of the SDE particles, which was significantly different to the nature of the fresh volcanoclastic deposits of Iceland. We infer that these granulometric differences could relate to (1) the relatively young geological evolution history of local samples, and (2) the initial shape properties and long-range transport processes of Saharan mineral dust particles. Similar shape features of far-travelled Saharan dust particles have been reported in other studies 62,80,92 that applied the same approach and used the same Malvern Morphologi G3-ID device as used in the present study. We are not aware of any other automated static image analysis data of windblown North African mineral dust material (shape data made by other methodological approaches cannot be quantitatively compared).
The well-rounded (but not spherical) shape of coarse silt and fine sand particles (caused by rapid abrasion and edge-rounding) is a known property of windblown deposits; however, sharp edges of finer silt-sized fractions (typical for fine-grained aeolian silts) were also visible on the captured images of individual particles of the SDE samples. Contrary to these properties, the Icelandic samples were much more irregular in shape and dominated by sharp features, even in the case of larger particles [89][90][91] .
Mineralogical comparisons of Icelandic and Saharan dust have shown that the latter is mainly comprised of quartz and feldspar, while the former primarily consists of volcanic glass 34 . Mineralogical fractionation is not observed in Icelandic dust due to the lack of larger mineral grains such as quartz. It is clear from the detailed laboratory analyses that the local addition of coarse particulate matter to the sampled deposited particles can be excluded.

Conclusion
In this paper, we presented the first systematic observation, synoptic background, and long-range transport routes of Saharan dust storms towards Iceland. Fifteen Saharan dust events were identified during a 12 year period (1.25 events annually) by applying MODIS satellite products of the NASA Terra Satellite, backward trajectories of the HYSPLIT model, and the numerical model forecasts of NASA's MERRA-2 and Barcelona Supercomputing Center. Samples of Saharan mineral dust that was deposited in Iceland during two outstanding wet deposition episodes were analysed in detail by using automated static image analysis of several tens of thousands of individual particles. There were clear granulometric differences between the SDE samples and the Icelandic dust source samples, indicating the external nature of the sampled grains and we were able to exclude possible contamination of local dust material. In addition, giant quartz particles (> 100 µm) were found among the deposited material, which confirmed the possible long-range meridional transport of these exceptionally large particles towards the Arctic region. Previous studies on giant particle transport from North Africa were focused on zonal (east to west) dust transport routes. Saharan dust episodes were associated with the enhanced meridionality of high-altitude winds. The meandering flow patterns of jets could be closely related to the increased warming of the Arctic and decreasing temperature difference between high-and low-latitude regions.

Methods
Study site for local samples and transported Saharan dust samples. The surface sedimentary units of six major Icelandic dust hotspots were sampled to obtain the granulometric (grain size and shape) properties of local source materials (Fig. 7).
The Hagavatn dust hotspot is located in the north-eastern part of the Lambahraun plain, and is bordered by moraines. Glacial surges and meltwater release in the outwash plain are responsible for the accumulation of finegrained deposits in the lake, where the fluctuating water level often leaves loose unconsolidated sediments on the surface. Landeyjarsandur is a floodplain area of the Markarfljót glacial river and sandy beach area. The glaciofluvial plain of Maelifellssandur is situated to the north of the Mýrdalsjökull glacier, and receives a huge amount of unconsolidated silty material every summer during melting episodes of nearby glacial areas. Mýrdalssandur is located to the southeast of the Mýrdalsjökull glacier, and is also a glacio-fluvial lowland plain under the effect of fluctuating glacial waters and jökulhlaups. Dust is generally transported towards the ocean from this source area. Dyngjusandur is situated in the rain shadow area of the Vatnajökull glacier, and is bordered by Askja (an active volcano) to the north, by the major glacial drainage system of Jökulsá á Fjöllum to the east, and by Vaðalda (a shield volcano). A significant part of the region is covered by an active and continuous sand sheet covering an area of approximately 170 km 2 (large parts were covered and blocked by Holuhraun lava during 2014-2015) with a thickness of up to 10 m. Fine-grained material originates from the summer floods associated with meltwater from the Vatnajökull glacier, although autumn dust events also occur in some years. Sandkluftavatn is a small lake bed approximately 50 km northeast of Reykjavik.
Samples of Saharan mineral dust that was deposited during two identified SDEs in Iceland were collected at the Agricultural University of Iceland in Keldnaholt, which is in the vicinity of Reykjavik, on 3 April 2014 and 25 April 2019 (a detailed description of each episode is provided in the Results). Outlying AOD values indicate an elevated dust concentration; therefore, backward trajectories at multiple altitudes were calculated for all AOD peaks to confirm the possible Saharan origin. The dust activity of each assumed source area was also confirmed using the visibility-reducing surface weather reports of the Naval Research Laboratory (https:// www. nrlmry. navy. mil/ aeros ol/# aeros olobs ervat ions) at the end-points of the trajectories.

Identification of Saharan dust in
Aerosol v4.10 subtype vertical profiles obtained from CALIPSO (https:// www-calip so. larc. nasa. gov/) were applied to provide another independent confirmation of the atmospheric presence of mineral dust over Iceland.
Dust load and wet deposition data of DREAM8 and NMMB/BSC dust numerical simulations were used to identify Saharan dust over Iceland and designate days of possible dust samplings 2,95 . The online multi-scale atmospheric dust model of the BSC provides daily dust load and deposition forecasts, which can also be used for the study area, even though the region was not part of the modelled domain before 2015. The movement of dust-loaded air masses towards Iceland could be clearly seen from the numerical simulations.
Synoptic meteorology and HYSPLIT backward trajectories. Mean geopotential height and wind vector maps at 700 hPa were compiled to define the governing synoptic meteorological patterns by using the daily mean composite application of the Earth System Research Laboratory at the United States National Oceanic and Atmospheric Association (NOAA) (http:// www. esrl. noaa. gov/ psd/). To determine the dust transport pathways, 120-144 h backward trajectory analyses from the sample collection location were applied using the HYSPLIT model 96,97 . The meteorological input data for the synoptic calculations and trajectory model were gathered from the Reanalysis Project dataset 72 of the NCEP/NCAR. Automated static image analysis. Direct optical image-based granulometric measurements of mineral particles provide a unique opportunity to determine the size and shape parameters of dusty material. The automated microscopic approach allows the characterisation of hundreds of individual mineral particles, thus providing the opportunity for a robust statistical analysis.
In this study, a Malvern Morphologi G3-ID advanced automated static image analyser was used. Approximately 7 mm 3 of mineral material per sample was dispersed onto a flat glass slide with an instantaneous (10 ms) pulse of compressed air (4 bar) and a settling time of 60 s. Particles were scanned using a 20× magnification lens (0.025 µm 2 per pixel resolution) attached to a Nikon eclipse microscope with z-stacking enabled (two focus layers were added above and below the initial focal plane, equivalent to a ± 13.5 μm z-height focus range).
Particle size and shape parameters of ~ 80,000 to 160,000 individual particles per sample were automatically recorded by the device software. Circle-equivalent (CE) diameter, aspect ratio, circularity, convexity, and solidity granulometric parameters were determined in this study. The CE diameter is calculated as the diameter of a circle with the same area as the projected two-dimensional particle image. The aspect ratio is the ratio of the particle width/particle length. Circularity is a proportional relationship between the circumference of a circle equal to www.nature.com/scientificreports/ the object's area and perimeter. Convexity is the ratio of the convex hull to a particle's perimeter, while solidity is the ratio of the area of the convex hull to the particle area. Aggregated and stacked particles were filtered by their certain parameters; particles with a circularity of < 0.6 were excluded from the granulometric characterisation. A cluster analysis was applied to determine the similarities of size and shape patterns in samples from different sources via a quantitative approach. Hierarchical cluster trees were created using the Euclidean distance pairs of the selected parameters. A built-in Kaiser Rxn1 Raman spectroscope was used for the chemical analysis of particles. The spectra acquired using a 785 nm (< 500 mW) laser over 5 s were correlated to library spectra (BioRad-KnowItAll Informatics System 2017, Raman ID Expert) to identify quartz particles among the targeted sedimentary grains. The employed automated static image analysis with Raman spectroscopy has been previously successfully applied to identify Saharan quartz particles in the unique basaltic-calcarenite sequences of Fuerteventura in the Canary Islands 92,98 .

Data availability
All data needed to evaluate the conclusions in the paper are present in the paper.