Supraglacial dust and debris: geochemical compositions from glaciers in Svalbard, southern Norway, Nepal and New Zealand

Instruments Data Provenance & Structure

Tools useful in analyzing geochemical glacier composition include atmospheric residence and transport models and satellite surface composition mapping. Atmospheric 15 particulate residence and transport modeling can identify age of particulates as well as local and long-range circulation paths (e.g. Han and Zender, 2010;Huang et al., 2010). Satellite determination of supraglacial dust and debris geochemistry (e.g. Gleeson et al., 2010;Casey et al., 2012) allows for increased spatial and temporal glacial dust and debris analysis toward climate, radiative forcing and particulate provenance

Background
In the atmosphere, snow and ice crystals nucleate around particulates (except when temperatures are below −38 • C and homogeneous freezing occurs) (Zimmermann et al., 2008;Cziczo et al., 2009;Knopf et al., 2011). Whether or not a particulate is at the center of a snow or ice crystal, as precipitation falls to Earth's surface, additional 5 particulates can be assimilated. Both natural and anthropogenic particulates are transported from source areas to remote locations and incorporated in glacial geochemistry (Tatsumoto and Patterson, 1963;Barrie, 1985;Schwikowski et al., 1999;Kaspari et al., 2009). Yet, the geochemical composition of supraglacial ice and snow evolves over days, 10 seasons and years. Atmospheric deposition of particulates fluctuates with season, weather or emission event (e.g. a volcanic eruption, smelter emission, oceanic salts). Atmospheric wet and dry deposition, and local dust or rock fall, biotic processes, wind exposure, glacial slope and/or resurfacing of sediments from glacial flow influence supraglacial ice composition. Glacier motion can crush glacial sediments and bedrock 15 to rock flour, allowing for remobilization of reactive trace elements within minerals to glacial ice (Tranter, 2003). Biotic processes also influence supraglacial composition (e.g. Kohshima et al., 1992;Hodson et al., 2005). For example, biota can change inorganic element oxidation states, and thus affect solubility of inorganic elements on glacier surfaces.
lor and McLennan, 1985;Rudnick and Gao, 2003). Insoluble elements are targeted in this study to investigate particulate origins and anthropogenic vs. natural influences to supraglacial dust and debris. Soluble elements are used to study seasonal glacier surface abundances.
3 Geochemical influences to the study areas 10 The four glacier study areas located in Svalbard, southern Norway, Nepal and New Zealand contain polar and mid-latitude glaciers, Northern and Southern Hemisphere glaciers, debris covered and bare ice glaciers, as well as active volcano, marine and continental influenced glaciers. These regions provide different geographic (latitude, hemisphere, altitude), climate (precipitation, temperature, atmospheric and oceanic 15 circulation) and glaciologic (temperate, polythermal) characteristics (Fig. 2, Table 1). The Svalbard, Grønfjordbreen and Aldegondabreen small cirque glaciers are located on the west coast of Spitsbergen at 77.98 • N, 14.12 • E. This Arctic study area is influenced strongly by the North Atlantic ocean current, resulting in a relatively mild climate (mean annual temperature −6.0 • C) relative to the high northern latitude. Of the study 20 areas, these glaciers are lowest in elevation (250-500 m), receive the least amount of precipitation (less than 400 mm per year (Hagen et al., 1993)), and are closest to the ocean as well as a local emission source. Grønfjordbreen and Aldegondabreen are less than 15 km from the ocean and coal mines. field, Jostedalsbreen experiences high precipitation year-round (approximately 1200-3000 mm) (Andreassen et al., 2005) with significant snow accumulation and annual melt due to typically warm summers (mean annual temperature 6.4 • C). Although removed from direct local emission sources, long-range transport of natural and anthropogenic particulates is a primary source of Pb, Cd, Zn and Co to southern Norway 5 (Pacyna et al., 1984;Steinnes, 2001;Berg et al., 2008). Of the areas studied, the Khumbu Himalaya, Nepal glaciers are closest to the Equator and highest in elevation: Ngozumpa glacier ( (Benn and Owen, 2002). Nepalese Khumbu Himalayan glaciers are continental "summer accumulation type" and experience summer precipitation that exceeds winter precipitation. The mean annual average temperature and precipitation in the region is −2.4 • C and 465 mm, respectively (Tartari et al., 1998;Hambrey et al., 2008). Atmospheric depo-15 sition from local continental dust is a known influence on area glaciers (e.g. Risheng et al., 2003), however, the majority of glacial debris is due to frequent rock and ice avalanches from surrounding extreme terrain (Hambrey et al., 2008). This study site has the most extensive debris cover of the areas discussed, with local geology composed primarily of quartz, feldspars, carbonates and micas (Searle et al., 2003;Casey 20 et al., 2012). Approximately eight small cirque glaciers lie on the upper outer flanks and in the summit craters of the active andesitic stratovolcano Mt. Ruapehu,in New Zealand (39.27 • S, 175.56 • E) (Dibble, 1974;Chinn, 2001). The glaciers span in elevation from 2200-2797 m and are influenced by a temperate, high precipitation, maritime climate 25 (6.1 • C annual mean temperature, 1100 mm annual average) with seasonal snow cover. The warm acidic Crater Lake (typically between 15-40 • C) near the summit, provides a continual sulfur and to a lesser extent chloride and magnesium aerosol influence on glacier ice (Hurst et al., 1991;Werner et al., 2006). Mt. Ruapehu produces frequent eruptions -predominantly steam, periodically magmatic, and commonly lahars -with over 60 such documented events since 1945 (Williams, 2001;Keys, 2007;Keys and Green, 2008). The last large Mt. Ruapehu eruption occurred on 25 September 2007 (discussed in Christenson et al., 2010;Kilgour et al., 2010). From the frequent volcanic eruptions, tephra layers in glacier ice are numerous and can be up to 1/2 m thick, rang-5 ing from fine ash to ballistics, with plagioclase, pyroxene, elemental sulphur, anhydrite, pyrite, and alunite dominant mineralogy (Williams, 2001;Kilgour et al., 2010).

Sample collection
Preparation for trace element supraglacial snow and ice sample collection included the 10 following: in a clean room, soaking low density polypropylene 500 ml Nalgene bottles in an acid bath for 48 h, subsequently washing acid-soaked bottles with triple filtered deionized water, drying bottles under laminar flow, and double bagging bottles prior to transport to the field. In the field, extreme care was taken during sample collection and handling. Samples were collected with clean polyethylene gloves using standard trace and blanks were run before, every tenth sample, and at the end of measurements to monitor instrument performance. Blank sample concentration values were subtracted from all sample measurement results, and three times the standard deviation of blank concentrations was utilized for the instrumental detection limit (subsequently denoted as 3σ).

15
Mineralogy and geochemical composition of glacier surface debris collected at the Khumbu Himalaya, Nepal and Mt. Ruapehu, New Zealand study glaciers were measured as follows. Glacial debris mineralogy was determined via powder X-ray diffraction (XRD) and geochemical composition was assessed via X-ray fluorescence spectroscopy (XRF). To prepare supraglacial debris samples for XRD and XRF analysis, 20 debris was oven dried at 60 • C for 24 h. Debris was then crushed and ground to a fine powder of less than 125 µm particle size (via a vibratory disc mill). Powder XRD was conducted on two grams of homogenized debris sample powder via use of a Philips XPERT diffractometer (manufactured by PANalytical B.V., Almelo). Dominant mineral components were identified by semi-quantitative peak area (height × full width 25 at half maximum) weight factor estimates (Moore, 1997) and full pattern modeling (e.g. Chipera and Bish, 2002). Ten grams of the homogenized debris sample powder were prepared into sample tablets to be measured on a Philips PW2400 XRF. significant dust re-emission in the region, particularly during the non-monsoon season (Lee et al., 2008;Liu et al., 2011). The three other alpine glacier regions contain less debris cover and are surrounded by less arid topography and thus exhibit lower trace and rare earth elemental overall abundances. This magnitude of particulate influence at the glacier surface is quantified in Fig. 3 average magnitude abundances 1 × 10 −5 -1 × 10 −6 for Svalbard, Norway, and New Zealand compared with average magnitude abundances 1 × 10 −3 -1 × 10 −4 for Nepal (similar REE magnitude differences are displayed in Fig. 4). Trace element to UCC abundance signatures displayed additional differences. Svalbard and southern Norway snow and ice samples revealed As, Sb and Bi enrichment, 10 and slight enrichment in Cd, Pb, Zn, and Cu. Khumbu Himalaya snow and ice samples had relatively constant values for transition metals (Ti, V, Cr, Mn, Fe, Co, Ni) and were most enriched in As, Cs and Bi. Mt. Ruapehu snow and ice samples were least enriched in light transition metals (Ti, V) and heavy metals (Pb, U) and most enriched in As, Cd, Sb, Tl and Bi. The strong negative Zr anomalies plotted, particularly strongly 15 in Nepal samples, are indicative of potential incomplete nitric acid digestion of the silicates (e.g. compared to HF digestion, as discussed in Osterberg et al., 2006).
Because REE are among the least soluble and geologically altered of the trace elements (Kreutz and Sholkovitz, 2000;Marx et al., 2005;Zhang et al., 2009), REE abundances can reflect supraglacial dust and debris sources. REE abundance relative 20 to chondrite (after Boynton, 1984) is plotted in Fig. 4. Regional variations are found in both the magnitude as well as fractionation of the light (LREE: La, Ce, Pr, Nd), middle (MREE: Sm, Eu, Gd, Tb, Dy, Ho) and heavy (HREE: Er, Tm, Yb, Lu) elements. The magnitude of REE abundance was largest in the Khumbu Himalayan glacier samples, and is indicative of the high dust deposition and re-emission in the region.

Supraglacial snow, ice and debris enrichment factors relative to UCC and regional geology
Individual elemental abundances were further quantified by calculation of enrichment 5 factors (EF) (Eq. 1). EF provide a simple, robust means to evaluate the abundance of an element in a sample relative to a reference material. Enrichment factors greater than 5 indicate significant enrichment, likely due to rock or soil dust. Enrichment factors greater than 10 indicate contribution from other natural or anthropogenic sources (Dasch and Wolff, 1989;Barbante et al., 2003) (UCC, regional loess and regional de-10 bris are used as references in this study).
Where "EF x " corresponds to the enrichment factor of element "x", "X " is the concentration of the element to be measured, "Ref" is the concentration of the reference element (Ti in this study), "sample" indicates the concentration ratio of the sample, and "STD" indicates the concentration ratio of the standard material chosen.
Glacier surface sample EF relative to UCC, and relative to local geology are presented in Tables 5 and 6, respectively. Surface glacier snow, ice and debris enrichment of elements relative to UCC (reference element Ti) are presented for the following elements: Fe, Co, Ni, Cu, Zn, As, Mo, Cd, Sb, Rb, Ba, Pb, Bi and U. In the Svalbard 20 supraglacial snow and ice samples, significant crustal enrichment (EF greater than 5) was found for Ni, Cu, Zn, As, Mo, Cd, Sb, Ba, Pb, Bi and U, and high (EF greater than 10) enrichment for all previous elements except Mo, Ba and U. The southern Norway snow and ice samples displayed significant crustal enrichment in Cu, Zn, As, Mo, Cd, Sb, Pb and Bi, and high enrichment of As, Cd, Sb, Pb and Bi. The Khumbu Himalaya snow and ice samples were significantly crustal enriched in Zn, As, Mo, Cd, Sb, Pb, Bi, and U. Greater than 10 crustal enrichment was found for As, Mo, Sb, Pb, Bi and U in Khumbu Himalaya snow and ice samples. Significant crustal enrichment was found in all Mt. Ruapehu snow and ice samples. Enrichment factors greater than 10 were measured for Co, Ni, Cu, Zn, As, Mo, Cd, Sb, Pb and Bi. From the debris samples, 5 significant crustal enrichment of U was found in Ngozumpa debris. No significant enrichment was found for Khumbu glacier debris nor Ruapehu glacier debris. Local geologic compositions were taken directly from Nepal and New Zealand supraglacial debris sample XRF measurements, while Spitsbergen Quaternary Loess (averaged from Advendalen samples reported in Gallet et al., 1998) was utilized for 10 Svalbard and southern Norway EF relative to local geology calculation. Table 6 lists the EF relative to local geology calculation results from the following elements (reference element Ti): Fe, Co, Ni, Cu, Zn, Rb, Ba, Pb and U. Svalbard snow and ice displayed significant local enrichment in Ni, Cu, Zn, Ba, Pb and U, and high enrichment in Ni, Cu, Zn, and Pb. Southern Norway significant local enrichment of snow and ice samples 15 was found for Cu, Zn and Pb, high local enrichment for Cu and Pb. Khumbu Himalaya snow and ice significant local enrichment, EF greater than 10 was found for U only. Mt. Ruapehu snow and ice significant local enrichment was found for Co, Ni, Cu, Zn, Rb, Ba, Pb and U, with high local enrichment for all listed elements except U.
The crustal vs. local enrichment calculations generally agreed well for the elements 20 repeated in both calculations (e.g. Svalbard Pb at 43.0, 12.4 relative to crustal enrichment and 47.7, 13.7 relative to local geology). Although EF cannot be used to specify provenance nor anthropogenic vs. natural contributions (e.g. Correia et al., 2003), the EF greater than 10 values found for As, Cd, Pb and Bi suggest non-crustal and potentially anthropogenic deposition sources in the glacier study regions. In the Mt. Ruapehu study region, the strong metal crustal and local geology enrichment is likely predominantly due to the continual volcanic degassing, although a small anthropogenic contribution may be possible.

Element abundance ratios
Elemental abundance ratios were used to demonstrate the type and degree of various geologic source contributions to supraglacial dust and debris. Three elemental abundance ratios, Ca/S, Al/Ti and Nd/Yb, are listed in Table 7 and visualized in Fig. 5. The Ca/S ratio conveys information on relatively soluble elements Ca and S and non-5 quartz dominant mineral dust abundance in each of the regions. Snow to ice phase changes result in alterations to the Ca/S ratio in each of the study regions (with the exception of New Zealand where Ca and S were not measured in summer samples). Ca can be related to calcium carbonate CaCO 3 dust, while S is related to sulphate aerosol or volcanic influences. The Svalbard and southern Norway glacier samples Both the Al/Ti and Nd/Yb elemental ratios utilize insoluble elements to provide information on the type and provenance of mineral dust in the supraglacial snow and ice samples. The Al/Ti ratio differentiates quartz type, while Nd/Yb has been utilized to pinpoint provenance (e.g. Kreutz and Sholkovitz, 2000). The particular insolubility of REE preserves geologic differences irrespective of glacial weathering and seasonal melt 20 processes. The LREE Nd is chosen and the HREE Yb because these elements are relatively abundant and detected with greater precision than for REE potential anomaly elements. Distinct Al/Ti and Nd/Yb ranges of values can be seen from the different study area supraglacial snow and ice samples (Fig. 5). The Nd/Yb values are within the range of those reported in previous glacier snow and ice studies (e.g. Grousset 25 et al., 1992;Ikegawa et al., 1999;Svensson et al., 2000;Kreutz and Sholkovitz, 2000;Osterberg et al., 2006). Of note, such data have not been reported in Svalbard, southern Norway or northern New Zealand. Himalaya snow, demonstrating that particulate loads measured in this study are likely to have albedo reducing impacts. Further, the geochemical composition of supraglacial debris, for example granitic continental vs. basaltic tephra, determines the variation of glacier surface albedo and the absorption of solar radiation. The more silica-rich debris found in Khumbu Himalaya, generally absorbs less solar radiation than the basaltic, 15 silica-poor debris found in Mt. Ruapehu surface glacier compositions.

Modeled air mass trajectories in glacier study regions
Another means to quantify particulate transport to and deposition on glacier surfaces is use of atmospheric models. Atmospheric particulate residence times can be calculated (e.g. Han and Zender, 2010), and transport paths can be mapped (e.g. Draxler

20
and Rolph, 2012). For this study, the HYSPLIT (HYbrid Single-Particle Lagrangian Integrated Trajectory) model (Draxler and Rolph, 2012) was utilized to estimate regional atmospheric transport patterns in each of the glacier study regions during the ablation season sample collection periods (Fig. 6). In each study region, 5-day transport paths were mapped using 24 variations in path and National Center for Environmental mass transport from the northeast. Southern Norway ablation season atmospheric transport from high-altitude Atlantic westerly movement was estimated. Khumbu Himalaya, Nepal atmospheric transport was dominated by westerlies in the November portion of the ablation season. New Zealand ablation season atmospheric particulate deposition was modeled to be transported from the west, over the Tasman Sea. The 5 model scenarios in each region demonstrate likely ablation season transport paths, and the ability to conduct more sophisticated analysis of particulate residence times, transport paths and thus deposition to targeted glaciers.

Discussion
Atmospheric particulate flux and the predicted changes to particulate flux patterns 10 have significant implications for the world's glaciers and ice sheets (Petit et al., 1999;Hansen and Nazarenko, 2004;Kaspari et al., 2009). Increased atmospheric deposition of dust, in conjunction with warming climate, could lead to greater than expected snow and ice melt rates, increased snow and ice mass loss, and ultimately contribution to sea level rise. Recent studies have also highlighted changes in spatial distribution 15 of heavy supraglacial debris due to changing climate (Stokes et al., 2007;Scherler et al., 2011;Lambrecht et al., 2011). With warming climate, insulation provided by heavy supraglacial debris can lessen, resulting in downwasting, thinning or retreat of many glaciers (e.g. southern Himalayas, Caucasus), changes to mass flux patterns, increased meltwater discharge, and/or increased supraglacial melt and glacier lake 20 outburst flood potential. The geochemical composition of supraglacial dust and debris influences these supraglacial energy balance variables and thus is important to study. This study adds to the understanding of supraglacial trace and rare earth elemental compositions, by providing a synoptic data set from western Svalbard, southern Norway, Khumbu Himalaya, Nepal and northern New Zealand.

25
In situ sampling techniques and analytical geochemical methodologies continue to evolve (e.g. immediate measurement techniques detailed in Gkinis et al., 2011). Currently many trace elemental measurement studies are limited by logistic constraints, which result in small sample sizes and/or inability to conduct multi-region comparisons. Small data sets also impede use of statistical methods such as cluster analysis (e.g. positive matrix factorization, principal component analysis). The most ideal description of glacial surface chemistry involves high spatial and temporal resolu-5 tion data. Satellite surface geochemical composition mapping and atmospheric dispersion and transport modeling over glacier study regions could be promising tools for investigating changing surface glacier compositions. Atmospheric modeling on multi-seasonal or annual time scales could assist in quantifying particulate deposition, at temporal scales. Satellite remote sensing techniques for describing atmospheric and surface geochemical compositions are continually improving. Satellite observations of atmospheric particulates, including mineral dust, soot, sulfide, water vapor and other components are currently mapped via spectral, lidar and thermal remote sensing techniques (e.g. Royer et al., 2010;Liu et al., 2008;Clarisse et al., 2011). Emerging research 15 points to the ability to detect major mineral groups as well as transition metals and rare earth elements on glacier surfaces from satellite spectral remote sensing techniques (e.g. Gleeson et al., 2010;Casey et al., 2012). Yet, in situ data collection and geochemical analysis is currently the most accurate method for deriving surface glacier geochemical composition.

Conclusions
Trace and rare earth element supraglacial composition data are limited in published literature. This study provides an important synoptic data set of four diverse alpine glacier regions. The contrasting geographic, glaciologic, atmospheric, and geologic conditions of the regions maximize supraglacial composition diversity and allow for a first-order particulates. XRF geochemical bulk oxide and trace element measurements offer higher accuracies and precision, free from dissolution biases. For this reason, XRF or laser ablation ICP-MS is recommended for future analytical geochemical studies. Limitations were found in use of solution ICP-MS as demonstrated by the suspected incomplete 15 supraglacial particulate digestion of the silica-rich Khumbu Himalaya samples.