Author Correction: Cryoconite as a temporary sink for anthropogenic species stored in glaciers

A correction to this article has been published and is linked from the HTML and PDF versions of this paper. The error has not been fixed in the paper.

radioactivity, elemental geochemistry and carbonaceous content, determined respectively through γspectroscopy, instrumental neutron activation analysis (INAA) and thermo-optical analysis.
The Alps are set in the middle of Europe. Being surrounded by some of the most industrialized and inhabited regions of the planet. Their position is extremely sensitive to atmospheric pollution 12,18 , including radioactive fallout associated to punctual accidents and to global contamination [19][20][21] . Such features make Alpine glaciers an ideal terrain to better comprehend the role played by glaciers and cryoconite in determining the fate of many natural and anthropic atmospheric species in glacial environments. Environmental radioactivity. The analysis of γ-spectra revelaed the presence of both natural and artificial radionuclides. Average activities concerning both the categories are shown in Fig. 2a. Natural species belong to 238 U and 232 Th decay chains or are primordial radioactive nuclei ( 40 K). Using data concerning total U and Th concentrations, determined through INAA, it was also possible to calculate the activity of primordial 238 U and 232 Th, partially reconstructing their decay chains (two examples displayed in Fig. 3).

Results and Discussion
The activity values of natural species are in perfect agreement with common environmental ones, with the exception of 210 Pb. Its activity reaches an average value of 2.8 kBq kg −1 , with a peak of 4.2 kBq kg −1 , well above what reported in environmental matrices 22 . The biogeochemical cycle of 210 Pb is complex, being related to the diffusion in the atmosphere of the radioactive gas 222 Rn. A scheme of its cycle is given in Fig. 3. Both the nuclides are intermediate products along the 238 U decay chain. The peculiarity of 222 Rn is its possibility to escape minerals and rocks diffusing into the atmosphere, where it rapidly decays to 210 Pb 22 . Its presence in the environment is related both to local production due to intrinsic uranium decay (supported 210 Pb) or to the deposition from the atmosphere in association to precipitations (unsupported or excess 210 Pb). Comparing the activity of upper and lower chains, it is possible to distinguish the two fractions 23 . In cryoconite the unsupported fraction is on average 40 ± 11 times higher than the supported one (standard deviation). Given its origin, 210 Pb is mainly found in superficial environments. Indeed also in moraine sediments, exposed to the atmosphere, an unsupported contribute is present, but in this case the ratio between the two fractions is 3.3 (Fig. 2). If the only scavenging effect of precipitation explains the presence of unsupported 210 Pb in the moraine sediments, this is not the case for cryoconite, where the accumulation of unsupported 210 Pb is too high. High activities of unsupported 210 Pb are typical in matrices whose composition is strongly influenced by deposition from the atmosphere, as lichen, moss or peat, but even in these cases they don't exceed few hundred 22 Bq kg −1 , an order of magnitude lower with respect to what observed in cryoconite. 232 Th decay chain, where no long-lived mobile radionuclides are encountered, reveals a substantial equilibrium (Figs 2 and 3). Also artificial radionuclides were detected: 137 Cs, 241 Am, 207 Bi (see Figs 2a and 4). 137 Cs is globally found in the environment as a product of atmospheric nuclear tests and nuclear accidents, it is a high-yield fission product from 235 U. Its chemical properties (solubility and volatility) and relatively long half-life (30.1 yr) made this carbonaceous content (panel c). In the first two panels average results about cryoconite and moraine sediments are presented (dark and light colors respectively). Not all nuclide labels are reported for U and Th chains, see Fig. 3 for the complete list. In panel c data related to single sample are shown, lines refer to average and standard deviation values. Asterisks highlight to values below detection limit. radionuclide extremely mobile and able to disperse worldwide 24,25 . In Europe a relevant role in determining its diffusion was played by the Chernobyl accident in 1986 26 .
The observation of 241 Am and 207 Bi is by far rarer. 241 Am (T 1/2 432 yr) is produced through multiple neutron capture and decay from 238 U. It was released in the environment through nuclear accidents, authorized discharges and nuclear weapon tests 27 . As a consequence of the concurrent decay of its parent nuclide ( 241 Pu, T 1/2 14.3 yr), its activity in the environment is not decreasing, but it is expected to peak by the end of 21 st century 28 . The occurrence of 207 Bi (T 1/2 31.6 yr) is even more rare and poorly investigated. No more than 30 papers report about its occurrence 29,30 . Its nucleosynthesis pathway is debated, but many evidences point to a role played by specific thermonuclear tests 31 , as the explosion of the Tzar thermonuclear device in 1961, occurred in Novaja Zemlya. This event is the prime suspected to explain the presence of 207 Bi in Europe 32 .
Average (and standard deviation) activities of 137 Cs, 241 Am and 207 Bi in cryoconite found on the Morteratsch glacier are respectively: 2.7(3.8) kBq kg −1 , 30(35) Bq kg −1 , 12(6) Bq kg −1 . The high deviations reflect the significative differences found among the samples. As in the case of 210 Pb, also for these artificial nuclides the observed activities are remarkable. One sample in particular (CR3) showed maximum activities for all the 3 considered nuclides: 13.6 ± 0.7 kBq kg −1 for 137 Cs, 120 ± 9 Bq kg −1 for 241 Am and 25 ±  As shown in Fig. 2a, activities of cryoconite are always well above what observed in moraine sediments, where it was not possible to detect 207 Bi. The enrichment is particularly high for 137 Cs. In sample CR3 137 Cs concentration is more than 2500 times higher with respect to the moraine material, while on average the ratio between activities of cryoconite and moraine sediments is 250, with a strong variability (standard deviation 720). Considering environmental samples, only other data relative to cryoconite collected on Austrian (Oher Dachstein massif 16 ) and Italian (Val d' Aosta region 36 ) glaciers are comparable to our results. In Fig. 4 it can be seen that the activities of the three radionuclides are strictly related to each other. The coefficients of determination (R 2 ) of linear regression calculated for each pair of artificial nuclides ranges from 0.78 for Am-Bi, to 0.94 for Cs-Am. What distinguishes the different Alpine contexts is the slope of the regression curve. While Italian and Swiss samples share similar coefficients, close to global fallout reference 29,37 , Austrian samples present higher relative 137 Cs activities. This is probably related to the differential dispersion of radio-cesium after the Chernobyl accident. Due to geographical and meteorological issues Austrian Alps were more affected by Cs deposition with respect to Western and Central ones 38 . Taking into account the nuclides related to global fallout (the pair Am-Bi, Fig. 4c), a substantial homogeneity is observed. This reveals that cryoconite is capable to record events associated both to regional and global atmospheric phenomena.  Elements presenting an enrichment in cryoconite with respect to moraine sediments. In panel a the average concentration of cationic elements in cryoconite, normalized with respect to moraine sediment composition, is shown. Elements are ordered following their affinity for clay minerals 42,43 . In panel b full data about the anthropic elemental group, an additional crustal element (Ce, light blue bars) is included for comparison. For each element data are presented as absolute concentration (bars, lower part) and as enrichment factor with respect to UCC (black curve, upper parts). Moraine sediment data are also presented (grey bars). Asterisks are referred to values below the detection limit.
Elements were divided as follows (the element order of each group depends on the average concentration found in the samples): crustal elements, sub-divided into major (Si, Fe, Ca, Na, K) and trace ones (V, Rb, La, Ce, Sc, Cs); heavy elements (Hf, Th, U W, Ta and Au); anthropic ones (Zn, As, Se, Sb, Hg). Crustal elements are mostly associated to Earth crust, their biogeochemical cycle is only marginally affected by human activities. The heavy element group is quite heterogeneous. Hf, Th, U and Ta are considered geochemically incompatible 39 , their origin is mainly crustal and they are associated to heavy and resistant minerals. W, as also Ta, is a refractory element, characterized by extremely high melting point and resistance. Its behavior in the environment and the impact of human activities on its geochemistry are not yet well constrained, as in the case of gold. The last group, referred as "anthropic" elements, includes elements whose biogeochemical cycle is strongly influenced by human activities and anthropogenic emissions 40, 41 . Crustal elements. For most of these elements moraine sediments show higher concentration than cryoconite (Fig. 2b). This could be related to the presence of organic matter in cryoconite which dilutes the mineral fraction. Na and Ca present the higher degree of depletion, their concentration in cryoconite is halved with respect to moraine sediments. Being both quite soluble, such a significative depletion can be related to water running on glacier surface. The flowing of liquid water washes out soluble species from cryoconite, as Na and Ca. In accordance to their solubility, a depletion would be also expected for other alkali elements (as K, Rb and Cs), but this is not the case, on the contrary they are enriched in cryoconite. The process responsible for accumulation of radioactive cesium probably accounts also for the concentration of chemically similar species, as natural Cs and other alkali elements. Comparing the composition of cryoconite and moraine sediments it appears that the affinity degree for clay is an important factor in governing the accumulation of ionic soluble species in cryoconite (see Fig. 5a). Elements with high affinity for clay (Cs and Rb 42,43 ) are adsorbed from meltwater and concentrated in cryoconite. On the opposite, solubilization prevails on absorption for those elements with lower affinity, determining a depletion.
Heavy elements. They are more abundant in cryoconite than in the moraine sediments (respectively +5, +108, +96, +30, +33 and +550% for Hf, Th, U, W, Ta, Au). This is an additional evidence about the effect of water flowing on the surface of the glacier during summer. Meltwater does not only remove soluble species from cryoconite, but also the lighter mineral fractions, which are easily mobilized. The removal of lighter minerals determines a relative enrichment of heavy minerals, where many trace elements are found at high concentration 44 . A similar process was noted to occur also in the atmosphere, where elements as Ta, Zr and Hf (associated to heavy minerals) are naturally found in excess in background aerosol (see Vlastelic et al. 45 and references therein). Through different processes (chemical and physical) liquid water removes soluble, mobile and light elements both in the atmosphere and on the ground. A hypothetical influence of anthropic activities has to be ruled out, since the human impact on the mobilization of such elements is not relevant 40 . The only exception concerns gold. The environmental behavior of this noble and rare metal is definitely poorly investigated, despite it is estimated that the anthropogenic contribution to its biogeochemical cycle represents more than 90% of total 40 . Gold concentration in cryoconite is on average 6.5 times higher than in moraine sediments and 8.5 times higher than average upper continental concentration (UCC 46 ), pointing to a non crustal contribution.

Anthropic elements.
Similarly to what observed for gold, also the elements grouped as anthropic show anomalous concentration in cryoconite. As shown in Fig. 5b, absolute concentrations and crustal enrichment factors highlight that the presence of these elements in cryoconite cannot be solely attributed to its mineral fraction. On the opposite in the moraine sediments they present a concentration perfectly comparable to UCC. The presence of these elements in cryoconite cannot be explained neither in the light of bare atmospheric deposition. Moraine sediments, being continuously exposed to the atmosphere, can be intended as a monitor for local atmospheric deposition. It doesn't present any significant enrichment, as also noted for radionuclides. The elements found in excess in cryoconite (Zn, As, Sb, Se and Hg) are quite mobile and volatile and are largely emitted in the atmosphere as a consequence of mining, metal production, fossil fuel combustion and other activities 41 . Volatility and mobility are indeed important features in determining the atmospheric behavior of an element. Not by chance Zn, As, Sb, Se and Hg, among the elements whose biogeochemical cycle is more affected by anthropogenic disturbances 40 , share relatively low melting point and high volatility. Several observations demonstrated that Alpine ice and snow are effectively contaminated with these and other chemically related metals and metalloids, as Zn, Cd, Bi, Pb 12 . But concentrations found in cryoconite are 5 orders of magnitude higher than the ones observed in ice and snow.
Carbonaceous Content. Data about organic and elemental carbon (OC and EC) in cryoconite are found in Fig. 2c. Organic carbon content ranges from 3.5 to 5.7% m/m. From OC, total OM can be inferred, assuming that 50% of organic matter is composed by organic carbon 47 . Average OM in cryoconite samples is 9.5% of dry mass (first datum ever relatively to the Alps), with a standard deviation of 1.6%, in perfect agreement with other continental glaciers 3 . Concerning elemental carbon no previous data were available for cryoconite. Elemental carbon, strictly related to black carbon 48 , is of extreme importance in relation to the optical properties of glaciers, since it strongly adsorbs visible light, altering the radiative budget of ice and snow surfaces where it is found 49 . Average concentration of EC is 0.49 (0.25) % of dry mass (standard deviation). Two samples, CR10 and CR11, present peculiar features, with low OC content (below 4%) and relatively higher concentration of EC (close to 1%). The low content of OC may be related to a recent formation of these cryoconite holes, where the microbiological community is still at its early stages. Elemental and black carbon were already found in ice and snow samples from the Alps, as a consequence of combustion processes and atmospheric transport to high altitude regions 50 , but typical concentrations were in the range of 10 0 -10 2 ng g −1 , 8-9 orders of magnitude lower than what observed in cryoconite 51 . Finding a so relevant amount of EC supports the hypotheses proposed by Cook et al. 3 about the entrainment of EC in cryoconite, but it also suggests that it could represent an important source of carbon for cryoconite microbial communities. In the Morteratsch pro-glacial plain, BC represents an essential source of carbon for the developing soils 52 . The origin of such a substantial amount was not constrained, but cryoconite could be involved. As long as it remains on the glacier, cryoconite accumulates OC and EC, but when, due to glacier retreat, it is finally released in the environment, it represents an important source of carbon for the pro-glacial environment, where few other sources are available.

Multivariate analysis.
To better comprehend the relationships between the different analyzed species, principal component analysis (PCA) was applied. Results concerning the first two components are presented in Fig. 6. The first one (PC1, explained variance 32%) is mainly related to natural radioactivity (K and Th-U decay chains, including 210 Pb) and secondarily to OC. Interestingly 210 Pb and OC seem tightly correlated (Pearson correlation coefficient, r = 0.79), confirming that organic matter plays a relevant role in adsorbing lead 53 . PC1 efficiently separates moraine sediments from cryoconite samples, but it only distinguishes them as a whole, no further details are noticed. Conversely the second component (PC2, explained variance 26%) presents a finer resolving power (Fig. 6). Indeed not only moraine sediments are at one of the two extremes of the sample distribution, but cryoconite samples are also displayed along a preferential direction. The original variables with the highest loadings in PC2 are artificial radionuclides, anthropic elements, Au and EC, i.e. all the species whose main origin is associated to anthropic activities and which are enriched in cryoconite. Such a picture suggests that this component can be interpreted as a "pollution" index.
On the accumulation capability of cryoconite. Our results show that cryoconite can accumulate radioactive isotopes, specific elements and elemental carbon with unprecedented efficiency. The comparison between cryoconite and moraine sediments ruled out the possibility to associate such strong enrichments to anomalous atmospheric depositional fluxes or to composition anomalies related to the local mineral material. Atmospheric deposition certainly plays an important role in influencing the composition of cryoconite, since species as unsupported 210 Pb, whose origin is essentially atmospheric 22 , are found in abundance (Figs 2 and 3). But from previous observations it is deduced that atmospheric deposition alone is not sufficient to explain such an accumulation. Alternative and complementary processes need to be taken into account. The only reasonable source of impurities other than direct atmospheric deposition, is meltwater. A glacier can be seen as a temporary repository of atmospheric deposited matter, having snow, ice and their impurities (including pollutants), a substantial atmospheric origin. Ablation mobilizes and releases into the environment what is preserved in the glacier through meltwater. Taking 210 Pb as an example, it can be assumed that the local depositional flux ranges from 50 to 150 Bq m −2 yr −122 , with an annual deposition on the glacier of 1-2 GBq (surface of Morteratsch glacier from Zekollari et al. 54 ). During the melting season a fraction of this amount, including the deposition from previous years, is mobilized and transported downward along the surface of the glacier, where cryoconite is found. The latter is the only available substrate on the glacier which natural and anthropic substances can be adsorbed on. It can be hypothesized that a part of transported 210 Pb is retained in cryoconite, determining the enrichments we observed. This conceptual scheme not only fit for lead, but for all the impurities transported with meltwater and presenting affinity for cryoconite.
The differential enrichment of cationic species (Fig. 5a) and the relevant accumulation of natural and radioactive Cs, the element with the highest affinity for clay 55 , are indicative of the important role played by clay minerals, a major constituent of cryoconite 56,57 . Conversely the high enrichments of elements as Pb, Sb, Se and Hg, can be related to the organic component which is present in cryoconite, since all of them present affinity for it 53,[58][59][60] . PCA (Fig. 6) and Pearson coefficients (see SI) partly confirmed this scenario since OC and elements as Hg, As and 210 Pb are positively correlated to it (r = 0.44; 0.82; 0.79 respectively). The correlation with other anthropic elements is lower (r = 0.33; 0.01; 0.36 for Zn, Se, Sb), but the variability of OC content in cryoconite samples is low (see Fig. 3c). What is important is the presence of OM, not its quantity, which is a quite stable parameter.
The contemporary presence of fractions with specific affinities for different substances and the seasonal availability of large quantities of atmospheric related impurities are the main factors which account for the high accumulation capability of cryoconite. Such features make it a temporary sink for many inorganic and organic substances stored in glaciers. They could be used for long-term atmospheric monitoring purposes, also in relation to exotic species, since they proved to be able to preserve extremely rare radionuclides as 207 Bi, released in the environment only punctually 29,30 . Models attempting to reproduce the secondary mobilization of pollutants into the environment 61 , should take into account cryoconite, which, in the light of our result, delay and concentrate the final release from the glacier. Heterogeneity, a question of age? What remains unclear is the high variability observed among different cryoconite holes. In Figs 4 and 5 it can be seen that the concentration of artificial radionuclides and anthropic elements is quite heterogeneous. Looking at the data it can also be appreciated that despite being extremely variable, a close correlation exists between different species. A striking case concerns artificial radionuclides (see Fig. 4). PCA shows that most of the anthropic species are positively correlated among each other. Looking at the samples (Fig. 6b), PC2, interpreted as a "pollution index", is capable to separate them in function of their pollution content. Moraine sediments present the lowest value, while sample CR3 (where maximum activities of artificial radionuclides were found) the highest one. The only reasonable factor which could explain the different degree of contamination is the age of cryoconite. Considering the selection of sampling sites, the influence of surficial hydrology should not be relevant. Also the OM content doesn't seem important, since its variability is low (Fig. 2c). The older is the cryoconite hole and the higher is the accumulation of pollutants. Up to now no successful attempts to determine the formation age of cryoconite are known. Probably a maximum possible age, depending on several factors, exists in relation to each glacier 11 . On the Alps this age cannot exceed few decades. This limit is posed by glacier dynamics, since the movement of ice and its melting prevent the preservation of older glacier surfaces where cryoconite can accumulate. Few years, at most few decades, could be sufficient to induce a significative heterogeneity, as observed on the Morteratsch glacier.
Determining the age of cryoconite remains a challenging target. It could shed important lights on the dynamics which involve ablation, the release of legacy pollutants and their accumulation in cryoconite, but also on the cycling of relevant atmospheric species and on their environmental fate, without forgetting the important role of microbiological communities.

Methods
Samples and location. Cryoconite samples were collected during summer 2015 and 2016 from the Vadret da Morteratsch, a north faced glacier located in the Bernina Massif (Swiss Alps). Cryoconite was sampled from the terminal part of the glacier, located between 2100 and 2400 m.a.s.l. and subjected to intense seasonal melting (Fig. 1). Each sample represents a distinct cryoconite hole. Samples were gathered from the central region of the ice body, where the surface of the glacier presented the simplest possible hydrological features. Areas close to moulins and main bedières were avoided. An additional sample of fine glacial diamicton was collected from a lateral moraine lying few meters from the glacier. Samples were kept frozen until they were processed for successive measurements. After melting, samples were dried at 70° for 12 hours and successively sieved to remove the coarser fraction (>1 mm). Three aliquots were obtained: one for γ-spectroscopy (≈1 g), one for Instrumental Neutron Activation Analysis (INAA, ≈0.2 g), one for carbon analysis (few grams).
Gamma-spectroscopy and radioactivity. Gamma-radioactivity of cryoconite samples was measured with a high purity germanium (HpGe) well detector. The instrument is dedicated to low background measurements [62][63][64] . A full description of the instrument and of its customization to reduce the radioactive background is presented here 64 . Given the low amount of available material and the low concentration of some radionuclides, the application of low background techniques was mandatory not only because of the low amount of available material, but also to achieve an appropriate signal to background ratio in a reasonable time 64 . Samples were counted for 5-7 days. Before counting, they were sealed in closed vials so as to allow the secular equilibrium between 222 Rn and its progenies to be reached 23 . To determine radioactive activity, absolute efficiencies were calculated at each γ-line of interest through Monte Carlo simulations 63 , using a GEANT4 based code 65 , which reproduce the instrumental apparatus, a virtual sample, the decay of the selected radionuclides and the radiation detection process by the crystal. According to this protocol differential self-adsorption phenomena and coincidence summing events are intrinsically taken into account. Further details about the procedure and the processing of γ-spectra are found here 63,64 , including information about calibration and quality assurance. Emissions of interest were selected so as to minimize interferences. Only for 224 Ra (241.0 keV) it was necessary to apply a correction to eliminate an interference from 214 Pb. All data are corrected for decay to January 2017; full details are found in the Supporting Information.
Neutron Activation. The elemental composition was determined through Instrumental Neutron Activation Analysis (INAA). Neutron irradiation was performed at LENA laboratories 66 (University of Pavia). To detect both short-lived and long-lived radionuclides two irradiations were planned, following the scheme adopted by Baccolo et al. 64 . The acquisition of γ spectra was carried out at different times both at LENA and at the Radioactivity Scientific RePORtS | 7: 9623 | DOI:10.1038/s41598-017-10220-5 laboratory of Milano-Bicocca university. Calculation of elemental concentrations was done in accordance to a relative method, as fully described here 67 . Twenty elements were quantified, details are presented in the SI. All nuclear data necessary to analyze and process the results presented in this work were taken from Chu et al. 68 .
Organic and elemental carbon. For Elemental Carbon (EC) and Organic Carbon (OC) determination, a Thermo-Optical Sunset EC/OC analyzer (Sunset Lab Inc.) was used, following the NIOSH 5040 protocol 69 . Cryoconite powder samples were suspended on quartz fiber filters (Pall, 2500QAO-UP, 47 mm diameter) pre-fired at 700 °C for 1 h in order to remove any possible OC contamination 70 . All the filters were weighed before and after cryoconite deposition in an air-conditioned room (T = 20 ± 1 °C; rel. um. = 50 ± 5%), after 48 h conditioning. The gravimetric determination of cryoconite mass was performed using an analytical microbalance (precision: 1 μg) operated inside the conditioned room; electrostatic effects were avoided by the use of a deionizing gun. Combining the information about the mass of cryoconite deposited on the filters and about the EC/OC surface concentration, m/m concentration was calculated.