Variations in mineralogy of dust in an ice core obtained from northwestern Greenland over the past 100 years

Our study is the first to demonstrate a high-temporal-resolution record of mineral composition in a Greenland ice core over the past 100 years. To reconstruct the past variations in the sources and transportation processes of mineral dust in 15 northwestern Greenland, we analyzed the morphology and mineralogical composition of dust in an ice core from 1915 to 2013 using Scanning Electron Microscopy (SEM) and Energy-Dispersive X-ray Spectroscopy (EDS). Analysis of the SEM-EDS reveals that the ice core dust mainly consisted of silicate minerals and the composition varied substantially on multi-decadal and inter-decadal scales, suggesting that the geological origin of the ice core minerals changed periodically during the past 100 years. The multi-decadal variation trend differed among mineral types: kaolinite generally formed in lowor middle20 latitude areas were abundant in the colder periods (1950 to 2000), whereas mica, chlorite, feldspars, mafic minerals, and quartzes formed in arid, high-latitude, and local areas were abundant in the warmer periods (1915 to 1949 and 2005 to 2013). This indicates that the multi-decadal variation of the relative abundance of the minerals can be attributed to the local temperature changes in Greenland. The trajectory analysis shows that the minerals were mainly transported from the western coast of Greenland in the two warming periods, which was likely due to an increase of dust sourced from local ice-free areas. 25 On the other hand, the abundant kaolinite was likely derived from old sediments at higher latitudes in North America, rather than from low and middle latitudes.

https://doi.org/10.5194/cp-2020-146 Preprint. Discussion started: 3 December 2020 c Author(s) 2020. CC BY 4.0 License. minerals from the northern hemisphere, such as Asia and North America (e.g., Biscaye et al., 1997;Maggi et al., 1997;Svensson et al., 2000;Donarummo et al., 2003). This trend reflects a decrease in weathering intensity with latitude. In addition to variations in kaolinite and chlorite, Type C, D, and E minerals also reflect the geological and climatic conditions of their source areas, as mentioned above. Thus, compositional variations among the five types of minerals can be used as an indicator of the source and transportation process of ice core dust in different periods. In this study, we consider each type of mineral as 145 possibly being contributed from the following sources: Type A, low-middle latitude areas (e.g. Central Africa, South America, Southeast Asia); Types B and C, high-latitude (e.g. North America, Russia, North Europe) and/or desert areas (e.g. Asia and North Africa); Type D, local areas (Greenland); Type E, desert areas (e.g. Asia and North Africa).

Backward trajectory analysis
To investigate the possible source regions of the ice core dust, the air mass transport pathways are analyzed using the Hybrid 150 Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model, which is distributed by National Oceanographic and Atmospheric Administration (NOAA, Stein et al., 2015). Points at 50 and 500 m above ground level (a.g.l.) at the SIGMA-D site are set as the initial points of an air mass for the 7-day backward trajectories. The probability distribution of the air mass at altitudes below 1,500 m a.g.l. is calculated at a 1° resolution. We assumed wet deposition for the preserved aerosol tracers Parvin et al., 2019), and thus the probability was weighted by the daily precipitation when the air mass 155 arrived at the ice-core site. We used daily precipitation from the ERA-40 and ERA-Interim reanalysis data sets, both of which are produced by the European Centre for Medium-Range Weather Forecasts (Dee et al., 2011;Uppala et al., 2005). The daily precipitation of ERA-40 ( "# ) was calibrated with that of ERA-Interim ( $%& ) via a linear regression obtained for the period of 1979-2001 ( $%& =0.47 "# , R 2 = 0.702, p < 0.001) to maintain consistency between the two precipitation datasets for the entire period . We also calculated the regional contribution from the probability distribution, for which land regions 160 were divided into the following 5 regions: the Greenland ice sheet, the Greenland coast, Canada, northern Eurasia, and China  https://doi.org/10.5194/cp-2020-146 Preprint. Discussion started: 3 December 2020 c Author(s) 2020. CC BY 4.0 License.

Snow cover fraction 165
To examine the surface condition of neighboring source for mineral dust emission, we analyzed inter-annual changes in snow cover fraction derived from multiple numerical simulations by a climate model. Various international organizations have conducted many numerical experiments simulated with global climate models to reproduce or predict climate change from past to future. The experiment products have been published by Coupled Model Intercomparison Project Phase 6 (CMIP6, Eyring et al., 2016) under the auspices of the World Climate Research Programme (WCRP). In this study, snow cover fraction 170 derived from historical simulations from 1850 to 2014 (Onuma and Kim, 2020a, b, c, d) were used to examine mineral dust sources for SIGMA-D. The dataset was produced by MIROC6, a climate model developed by a Japanese modelling community (Tatebe et al., 2019), with four reanalysis datasets such as GSWP3 (Kim, 2017), CRUJRA (Harris, 2019), Princeton (Sheffield et al., 2006), and WFDEI (Weedon et al., 2014)

Dating of the SIGMA-D ice core
Dating of the SIGMA-D ice core, which was performed by annual layer counting of δ 18 O and Na + , showing obvious seasonal variations (Fig. A1). The observed seasonality of chemical components and the water stable isotope ratio in the snowpack and 180 ice cores has previously been reported at various sites on the Greenland Ice Sheet (Whitlow et al., 1992;Legland and Mayewski, 1997;Kuramoto et al., 2011;Oyabu et al., 2016;Kurosaki et al., 2020). The winter season of the SIGMA-D ice core was defined as the depth at which δ 18 O was at its minimum value and Na + was at its maximum value, and we counted winter season to winter season as 1 year.
Other fixed dates were provided by the tritium profile and nssSO4 2− spikes. A sharp tritium peak at 11.56 m w.e. corresponds 185 to the H-bomb test in 1963 (Koide et al., 1982, Clausen andHammer, 1988), indicating an accumulation rate of 0.23 m w.e. yr −1 from 1963 to 2013. The large nssSO4 2− peak appeared at 54.17 m w.e. is assumed to correspond to the eruption of the Laki volcano in 1783. The nssSO4 2− signal of the 1783 Laki eruption has also been found in other ice cores in Greenland, Arctic Canada, and Svalbard (Clausen and Hammer, 1988;Grumet et al., 1998;Matoba et al., 2002). Similarly, we assume other nssSO4 2− spikes to be the signatures of unknown (1810), Tambora (1816), and Katmai (1912) volcanic eruptions at shallower 190 depths of 47.53, 46.03, and 23.50 m w.e., respectively. Comparing the annual layer counting and these reference horizons, we estimate that the ice core dating includes a 1-year error.
As a result of these analyses, we estimate that the upper 112.87 m (86.06 m w.e.) of the ice core is equivalent to the period from 1660 to 2013. In this study, we used ice samples to a depth of 38.60 m (22.75 m w.e.) covering 1915 to 2013 for the SEM and EDS analyses (Table 1). 195 Figure 3 shows SEM images of the mineral dust in the SIGMA-D ice core. The number size distribution of mineral dust in the SIGMA-D ice core showed that most particles had a diameter of <2 µm (Fig. 4), which is consistent with other Greenland ice core dust (e.g., Steffensen, 1997;Biscaye et al., 1997). The mean and maximum particle diameters, calculated as 5-year averaged values, ranged from 0.97 to 2.60 µm and 4.94 to 26.51 µm, respectively, with a single modal structure at the peak 200

Quantitative estimation of mineral dust
The elemental composition of individual mineral particles obtained from the EDS analysis showed that the ice core dust was composed mainly of silicate minerals in all the samples (65%-95%, Fig. 6). Based on a peak intensity ratio sorting scheme (Donarummo et al., 2003) and comparison of oxide composition and morphological information with those of standard 210 minerals, the silicates were categorized as quartz, Na/Ca-and K-feldspars, clays (kaolinite, pyrophyllite, smectite, illite, micas, and chlorite, as well as mixed layers of illite/smectite and mica/chlorite), and mafic minerals rich in magnesium and iron (Figs.   3 and 7). These minerals were also found in Greenland ice cores from glacial periods (e.g., Maggi, 1997;Svensson et al., 2000).

Source regions of SIGMA-D ice core dust
To identify the source regions of the ice core dust, we applied the HYSPLIT back trajectory model and calculated the 235 probability distributions of an air mass arriving at the SIGMA-D site from 1958 to 2014 (Figs. 2b and 8). The results show that the air mass at elevations above ground level from 0 to 1500 m came mainly from the western coast of Greenland, including the Baffin Bay, whereas a smaller part came from northern Canada (Figs. 2b and 8a). Excluding the ice sheet and ocean areas that could not be possible sources of mineral dust, the air mass is considered to have come mainly from the Greenland coast (50%-60%) and Canada (~40%), with a small contribution from northern Europe (~3%, Fig. 8b). Contributions from these 240 three possible source regions show little seasonal and inter-annual variabilities (Fig. 8b and c). https://doi.org/10.5194/cp-2020-146 Preprint. Discussion started: 3 December 2020 c Author(s) 2020. CC BY 4.0 License.

Variation in silicate mineral composition
The SEM-EDS analysis revealed that the SIGMA-D ice core dust samples collected from 1915 to 2013 mainly contained silicate minerals, which is the most abundant family of crustal minerals (Deer et al., 1993). Silicate mineral composition 245 showed variations on a multi-and inter-decadal scale, indicating that the geological origin of the minerals changed cyclically over two different time scales during the past 100 years.
Variation trends in the silicate mineral composition of the SIGMA-D ice core samples substantially differed among mineral types (Fig. 9), indicating that the minerals in the ice core were derived from multiple geological sources. The dominance of Type A minerals in the samples from 1950 to 2004 indicated that the minerals might be mainly derived from low-or middle-250 latitude areas in the periods. In contrast, the abundance of Type B, C, D, and E minerals in the samples from 1915 to 1949 and from 2005 to 2013 indicated that the minerals were likely derived from arid deserts and/or high latitude areas, including Greenland.
The morphological characteristics of the ice core dust also support the changes in the sources of silicate minerals. The size distribution of the minerals showed lower mean, maximum, and modal diameters from 1950 to 1994 compared with the other 255 periods, when the samples comprised mainly Type A minerals (Table 1, Fig. A2). The circularity also showed a similar trend, containing smaller amounts of particles with circularity values >0.80 from 1950 to 1994 (Fig. 5). The particle size depended on the mineralogy of the ice core dust. The coarser fraction (>2 µm) of the ice core dust samples contained little clay, especially for the Type A minerals (Fig. 10), but contained an abundance of Type C, D, and E minerals. This particle size dependence is https://doi.org/10.5194/cp-2020-146 Preprint. Discussion started: 3 December 2020 c Author(s) 2020. CC BY 4.0 License. consistent with other analyses of Greenland ice core dust (Biscaye, 1965;Svensson et al., 2000). Thus, the ice core dust was 260 likely derived from different geological sources in the late 1990s compared with the other periods. The low air mass contribution from low and middle latitudes indicated by the back trajectory analysis suggests that there are other possible sources for the abundance of Type A minerals found in the cooling period (1950s to 1980s). Type A minerals are typical of humid tropical climatic zones, such as those found in modern-day Africa and Southeast Asia (e.g., Bergaya et 325 al., 2006). However, some studies have shown that other possible sources of Type A minerals are ancient soils formed by chemical weathering in high-latitude areas in past warming events. Darby (1975) analyzed the clay mineral composition of marine sediments from deep-sea cores in the Arctic Ocean and revealed that there was abundant kaolinite (Type A minerals) apparently derived from shale and soils of northern Alaska and northern Canada, which were relict deposits of warmer climates in the Tertiary. These abundant kaolinites were deposited in a non-marine environment (Allen and Johns, 1960). Clay 330 mineralogy of North Sea Basin sediment cores also revealed that increased kaolinite concentrations were associated with Palaeocene-Eocene Thermal Maximum (Kemp et al., 2016). However, the mineralogical composition showed small amounts of kaolinite in the soil around the Qaanaaq Glacier in northwestern Greenland (Nagatsuka et al., 2014). Thus, the Type A minerals in the SIGMA-D ice core were likely transported from old sediments at higher latitudes in North America, rather than from low and middle latitudes and the local area. 335

Conclusions
Analysis of the SEM-EDS of individual dust morphology and mineralogy in the SIGMA-D ice core revealed that the ice core dust mainly consisted of silicate minerals, including quartz, feldspars, and mafic minerals, and clay minerals, including kaolinite, illite, smectite, micas, and chlorite. Most of the particles had a diameter of <2 µm, implying that the ice core contained mainly long-range transported wind-blown mineral dust. The silicate mineral composition varied substantially on multi-340 decadal and inter-decadal scales. The multi-decadal variation trend differed among mineral types formed in different source areas, which corresponded to that of the surface temperature change in Greenland: kaolinite generally formed in low-or middle-latitude areas were abundant in the colder periods (1950 to 2000), whereas mica, chlorite, feldspars, mafic minerals, and quartzes formed in arid, high-latitude, and local areas were abundant in the warmer periods (1915 to 1949 and 2005 to 2013). This indicates that the multi-decadal variation of the relative abundance of the minerals can be attributed to the local 345 temperature changes in Greenland. The trajectory analysis showed that the air mass arriving at the SIGMA-D site came mainly from the western coast of Greenland from 1958 to 2013 and the contribution showed little inter-annual variabilities, indicating that the variations in the geological origins of the ice core dust were unlikely due to changes in air mass transportation. The abundant minerals in the two warming periods were probably due to an increase of dust sourced from local ice-free areas because the snow/ice cover duration in the Greenland coastal region was shortened by the recent warming during the melt 350 season. On the other hand, the abundant kaolinite in the cooling period was likely derived from old sediments at higher latitudes from North America, rather than from low and middle latitudes. Although further analyses are needed to identify the cause of inter-decadal variations in ice core dust, our study is the first to demonstrate a high-temporal-resolution record of mineral composition in a Greenland ice core over the past 100 years. https://doi.org/10.5194/cp-2020-146 Preprint. Discussion started: 3 December 2020 c Author(s) 2020. CC BY 4.0 License. Figure A2. (a) Comparison of particle size distribution and log-normal fitting results (mode: mode diameter and R 2 : half peak width) of the ice core minerals among the samples. (b) Historical changes in the mode diameter and R 2 https://doi.org/10.5194/cp-2020-146 Preprint. Discussion started: 3 December 2020 c Author(s) 2020. CC BY 4.0 License.