Sediment concentrations and transport in icebergs, Scoresby Sound, East Greeland

Glaciers erode their beds by abrasion and plucking and entraining sediment on their way downwards driven by gravity. Basal ice becomes a sediment transport agent. To determine glacial transport of sediment, measurements of both the ice flux and its concentration of sediment are needed. Once glaciers reach the ocean, ice and its entrained sediment is released into the ocean by calving. Further transport takes place by icebergs. Quantification of IRD (ice-rafted debris) fluxes, which upon ultimate deposition on the ocean floor is an important climate indicator, becomes even more complicated as icebergs topple and differentially melt while being in transport. While the volume of tidewater glacier ice released by recent calving is quite well constrained by satellite measurements, there is a lack of measurements of the concentration of sediment within the moving ice. Here, we describe a method to collect samples of ice from icebergs systematically for the first time in Greenland together with a strategy to obtain representative samples. Our method is tested in Scoresby Sound, East Greenland in order to describe the transport of sediment into the fjord system related to calving from known glacial source areas. Our data clearly demonstrate that biased and sparse sampling potentially produces unrealistic values of sediment concentrations. Seventy-two samples from 24 icebergs had an average concentration of sediment of 35.5 g/l of ice with a standard deviation of 97%, between the 24 individual icebergs. The origin of the sediment is related to specific source areas. Based on the samples, we present an estimate of the annual transport of sediment out of Scoresby Sound related to calving ~100 (0.3-200) million t yr -1 . Finally, we discuss the uncertainties of our estimate.


Introduction
The Earth´s sediment cycle is driven by a variety of geomorphic processes, e.g.Walling (1983), Milliman andFarnsworth, (2011), andSyvitski et al. (2022).The relative importance of glacial erosion compared to other erosional processes, both recent and over millennial timescales is well described e.g.Hallett (1997), Antoniazza and Lane (2021), Herman et al., (2021).Whereas there are exceptions for cold-based ice sheets, and erosion rates can be variable, generally rates exceed or are on par with bedrock incision by rivers (Koppes and Montgomery, 2009;Alley et al., 2019).In Greenland, glacial erosion is the most powerful geomorphic agent.Greenland is the largest glaciated area in the Northern Hemisphere.One example to demonstrate the efficiency of glacial erosion here is the sediment flux originating from the Watson River, which drains about 12.600 km 2 of the Western sector of the Greenland Ice Sheet, which only amounts to less than 1% of the entire ice sheet area (van As et al., 2018).This well monitored proglacial river delivers 693-2353 t km -2 yr -1 of sediment derived by glacial erosion (Hasholt et al., 2018).The effects of glacial erosion are also demonstrated by the presence of hundreds of large braided river flood plains and deltas (Bendixen et al., 2017) together with large plumes of suspended sediments in fjords and coastal waters (Chu et al., 2012).Besides the output of meltwater from the rivers, Greenland is also the largest producer of icebergs by calving in the Northern Hemisphere (Enderlin et al., 2014).The average annual loss (2010 to 2017) by calving from Greenland was ~500 +-50 Gt yr -1 of water, or ~550 km 3 of ice (Mankoff et al., 2020).Field observations show that calved icebergs can carry high concentrations of sediment at their base, but quantitative data are sparse.The first observations and measurements of sediment concentration in icebergs by the authors were from an ice-dammed lake (Nordbo Lake in Johan Dahl Land in South Greenland) intended to be used as reservoir for hydropower.Measured concentrations of 15 mg l -1 up to 40 g l -1 of sediment in icebergs calved into the lake demonstrated that a substantial amount of sediment can be transported by icebergs (Hasholt and Thomsen, 1981).The impact of englacial transport was also demonstrated by the presence of IRD in bottom samples from an ice-dammed lake in East Greenland, (Hasholt, Walling and Owens, 1996).A first estimate of transport of sediment related to calving delivered to the Arctic Ocean ranged from 50 to 500 million t yr -1 (Hasholt et al. 2006;2016).The estimate was based on the volume of icebergs derived from Reeh (1985a) multiplied with iceberg concentration estimates from Hasholt and Thomsen (1981) and Ruth et al. (2003).The estimated transport involved the export of icebergs only from North-and East Greenland.Overeem et al. (2017) attempted to quantify sediment transport from Greenland with a focus on riverine suspended sediment transport assessed from remote sensing imagery.Beside calculation of the proglacial river sediment transport, based on calibrated LandSat-derived fluvial suspended sediment concentration, the paper also presented a preliminary calculation of the amount of sediment transported by calving icebergs.The combined estimate of fluvial and glacial sediment transport suggest that Greenland is the largest exporter of sediment in the Northern Hemisphere.Moreover, if the maximum estimate of the sediment transport by calved icebergs is used, this component of the sediment transport from Greenland could account for up to 17% of the total flux of sediment from the continents to the oceans worldwide.However, very few field measurements of the thickness of sediment-rich basal ice or the concentration of sediment in icebergs from Greenland have been available for these calculations.Thus, an improved knowledge of the range of concentration of sediment in icebergs and quantification of the spatial variability across different calving source areas are required.Here we aim to: 1) develop a method and strategy to obtain representative samples of calved ice.2) describe collected ice samples in front of known source areas (calving fronts).3) investigate if the sediment in the ice samples can be related to the geology of specific source areas.and 4) discuss the applied methodology in relation to the calculation of sediment transported by calved ice.

Fieldwork
An obvious reason for the lack of measured concentrations of sediment in icebergs is the difficult access to tidewater glaciers with calving fronts along the remote coasts of Greenland.Other reasons are the harsh weather conditions and the difficulty of boat travel within the melange of sea ice, which includes growlers, bergy bits and icebergs choking up the fiords near calving fronts.Moreover, the potential for sudden toppling or fracturing of icebergs makes in-situ sampling of larger icebergs unsafe.Yet, there is boat access to fiords with tidewater glaciers in their fiord heads near many Greenlandic communities and coastal research stations.In addition, potential for sampling during research cruises or other investigative efforts exists.The Danish Navy regularly patrol the waters around Greenland to control illegal access and fishing and to carry out Search and Rescue (SAR) missions.An agreement between The Danish Centre for Sea-research (DCH) and the Danish Navy provides the possibility for scientific projects to obtain logistical support and help from the Navy after an application and a positive evaluation of the quality and relevance of a given project.This agreement has allowed the author to join a mission in Scoresby Sound East Greenland in August 2018 (Fig. 1).Beside the normal tasks e.g.SAR, which have absolute priority, the main scientific objective of the expedition was to investigate the impact of seismic booming and ship traffic on narwhal behaviour.However, the author was allowed to use the Man Over Board (MOB) rubber dinghy with crew for sampling of ice at a distance of up to 4 nautical miles from the vessel HDMS I/F Lauge Koch.

Methods
Sampling strategy: Sampling sites were chosen as close to the calving front of the selected source area as possible, depending on weather and ice conditions.The iceberg to be sampled was chosen to be the one closest to sample position pre-selected on a map.The intention was to distribute samples evenly along the calving front.In order to obtain samples for describing the geology of the sediment and to get an idea of the "maximum" transport, supplementary sampling was carried out from icebergs with an observed large content of sediment, the so-called "dirty icebergs".Sampling procedure: For safety reasons only small icebergs in the WMO (World Meteorological Organization) size categories growlers (<1m freeboard, <10m waterline) and bergy bits (< 5m freeboard, 10-30m waterline) were selected for sampling.The first sample was taken at a random location where the boat bumped up to the iceberg.The following two samples were picked "systematically" within a fixed distance of 2-6 m (multiples of a used yardstick length) from the previous sample.This systematic sampling was used to avoid subjective choices of sample locations, while the triplicate samples helped to assess variability within a selected iceberg.At each sampling point, the ice was loosened by chopping with a mountaineering ice axe with a long shaft.If samples were intended for analysis of englacial sediment concentration, the surface was cleaned before chopping and if the sample was destined for analysis of trace elements e.g.iron content, a brass hammer was used instead of a steel ice pick.The shards of ice were collected in a collapsible fishing net held underneath the sampling point or picked up from the water.The pieces of ice were put into a pre-labelled plastic bag, which was carefully inspected for leakage.In case of potential leakage due to sharp shards, double bagging of the sample was employed.To avoid leakage by sharp shards a bottle or container with thicker walls was also used when available.The location of the sampling points and time of sampling was recorded with a handheld parallel multi-channel GPS or the boat GPS receiver, with a typical accuracy of 5-10 m, and date and time of sampling were noted.A description in the field of the iceberg is given according to the WMO classification.An overview photo of the iceberg and photos of each sample location were taken, if possible with a standardized colour calibration, or grey scale calibration card within the photo view.A full field description included notes on the ice matrix, i.e. blue ice/white ice, bubbly/broken/dense, visible layering or stratification, evidence for algae, red or green, and any alternative light-dark bands.Sediment sorting, angularity, grain size distribution, layering or stratification visible, and presence and abundance of pebbles or cobbles were also recorded in field notes.Environmental conditions, such as air and water temperature were measured together with wind speed and direction.If available, water depth from the boat sonar system was noted.After returning to the ship, the samples were allowed to melt, and in the case of observed leakages, new plastic bags were applied.Then all samples were weighed with a resolution of +-one g.Freeze storage was not used because of lack of available freezing capacity during all steps of the normal transport route from Greenland to the laboratory.Analysis: In the laboratory, all samples were again controlled for leakage, visually and by weighing and comparing against field weights.In a few cases, the outer bags were damaged due to heavy handling during transportation, but it was observed that no sediment had been lost from the inner bags.First, a subsample of 50-100 ml, which was used for analysis of the isotopic content, was filtered through Millipore CEM 0.45 micron filters.Then the sediment was separated from the water by filtering through Whatman GF/F mass fibre filters with a retention diameter of 0.7 microns (to avoid lengthy duration of the filtering); larger grain sizes were poured into porcelain crucibles or metal containers.Filter papers, crucibles and metal containers were dried at 65 o C and weighed with an accuracy +-0.1 mg.Loss on ignition was determined after combustion at 550 o C. Geology: Lithic fragments down to gravel size were used in a blind test.A geologist, who has carried out geological mapping of Greenland with special experience in the Scoresby Sound area was asked to characterize the fragments, and, to the extent possible, to identify the rock types and their possible source areas without being informed about the sampling location.Isotopic analysis: Measuring the stable isotope ratios of oxygen and hydrogen (δ 18 O, δD) allowed us to distinguish between thick accumulation or sea ice (with high δ 18 O) or glacial ice, iceberg originating by calving (with low δ 18 O).We also explored whether isotopic signatures provide information about the local origin of the glacial ice.If signatures are unique enough, they could possibly enable delimitation of source areas.Isotopic analysis was performed at the Niels Bohr Institute (NBI), Denmark using Cavity Ring Down Laser Spectroscopy with a Picarro L2140i analyser.Measurements were reported on the international VSMOW-SLAP (Vienna Mean Ocean Water Standard Light Antarctic Precipitation) isotope scale after calibration with local water standards.Methodological details on the instrument calibration and accuracy are described in Gkinis et al. (2021).Calculation of the sediment transport: The 2018 volume of calved ice delivered from 15 individual calving glaciers was determined by measuring the velocity of the ice from satellites and multiplying the velocity with the cross section area of flux gates based on updated bed topography from BedMachine v4.Details of the methodology and accuracy of this approach are described in Mankoff et al. (2020).The concentration of sediment from each of the 15 glaciers was determined as the average concentration of the samples collected in the present investigation that were closest to each glacier.The annual transport was determined by multiplying volume of ice from a glacier with the corresponding concentration.

Sampling and sampling conditions
Sampling was carried out from the 26 th August to 1 st September 2018, when the sea-ice had disappeared from the Scoresby Sound.The weather was fair with light winds that allowed the MOB to operate far from the fjord coast.The ship sailed north through Hall Bredning (fig.1), where several large icebergs originating from the Daugaard-Jensen Glacier were drifting southwards along the east coast of Jameson Land.Then we sailed through Ikaasakajik (Øfjord) and further south west through the sound east of Storø because the passage westwards was blocked by a melange of ice possibly originating from the Eilson Glacier.The route went south through ice from Rolige Glacier and Vestfjord Glacier, and then eastwards through Ujuaakajiip Kangertiva (Fønfjord), where the first samples (no.1-2) were taken.Samples 3-5 were taken directly south in Kangersivat (Gaasefjord).We then turned back to a whaling station closer to the fiord head of Rødefjord, and iceberg samples no.6-11 were collected nearby.Then we sailed east to the waters along the south coast of the fjord (fig. 2 and 3) where samples no.12-21 were collected.Because of the melange of ice originating from Nertiit Kangersivat (Gaasefjord) it was not possible to sail closer to the calving fronts here, the nearest samples were at the location of iceberg 17 and 18.When the whale observations along the south coast were finished, we turned northwards through Hall Bredning again and had opportunity to collect the last samples at location 22, 23 and 24 (fig.1) downstream of the Daugaard-Jensen Glacier calving front.A total of 24 icebergs were sampled.

Sediment concentration
We have assumed a density of 0.9 g cm -3 for ice and 2.65 g cm -3 for sediments in order to calculate the sediment concentration in mg l -1 for icebergs.An overview of the calculations and the results of the 72 single samples (three from each iceberg) are available in a table in the Isaaffik data portal https://isaaffik.org.. Sample weight varied from 0.3 to 1.7 kg with an average of 0.7 kg.The average concentration was 36.6 g l -1 with a maximum of 692 g l -1 and a minimum of 0.8 mg l -1 .The standard deviation was 400% of the average.The variation in sediment concentration between the single iceberg triplicate samples is several orders of magnitude.The sample with the maximum concentration was not able to float because the density was larger than one.Results from the 24 individual icebergs shown in figure 1 are calculated as the average of the three subsamples and are reported in table 1.The average of all 24 concentrations is 36.6 g l -1 ; median is 6.6 g l -1 , the standard deviation is 97% of the average.Maximum sediment concentration is 442 g l -1 and minimum 4 mg l -1 .An example of the variation within a single iceberg ( 7) is shown in fig. 4 A,  B and C. The distribution of concentrations was bimodal with a group of five below 0.5 g l -1 , and a further five evenly distributed in the interval between 0.5 and 5.5 g l -1 . Except for one concentration value there is a gap in the interval between 5.5 and 10 g l -1 .Eleven icebergs have concentrations of sediment in the interval between 10-60 g l -1 .Two outliers with respectively 120 and 442 g l -1 are found because of the sampling from selected "dirty" icebergs.Considering concentration as function of distance from nearest source area reveals a very large spread of values at distances out to 80 km.Some icebergs can have several possible source areas.A similar survey revealed a large variation out to 60 km from source.In both cases, low concentrations are found once the iceberg has travelled more than 100 km from any possible source, indicating an expected loss of sediment from source to the mouth of the extensive fiord system.

Geology
The locations of the 24 sampled icebergs are plotted on a lithological map, fig. 5 (Harrison et al., 2011, Moon et al., 2021).Two main groups of rocks were identified as shown in table 2. Basaltic rocks originate from the southern coastal area of the fjord.Samples 3-5 and 14-18 represent Kangersivat (Gaasefjord) as a source draining part of the Greenland Ice Sheet, whereas samples 12 and 13 are sourced from a local glacier (Sydbrae) draining from the basaltic area, and samples 19 -21 also represent a local glacier, the Brede glacier.Gneissic rocks are found in samples 6-8 and 9-11 representing the geology of the western part of the fjord.They originate from Rolige Brae or the Vestfjord Glacier.Samples 22, 23 and 24 also contains gneiss, and they either originate from local glaciers, such as Eielson Glacier, small glaciers calving from Renland or Milne Land or are broken off from larger icebergs originating from the Daugaard-Jensen Glacier.Overall, the results confirm that a sample of rock fragments from an iceberg often can be traced back to a distinct calving front.If the geology at the calving front is unique, the samples can be used to track the travel route of the supraglacial component of ice rafted debris.On the other hand, samples from an iceberg originating from a source area may provide useful information about the unknown geology of the source area if the geological surveying has been less intense than in the present area.

Isotopic analysis
Results from the analysis of selected samples are shown in table 3, where A, B and C refer to a subsample from a given iceberg.It appears that all sampled icebergs except one (11A) which has a very high δ 18 O are of glacial origin and do not include sea ice.This value indicates that the sample could be from sea ice or from ice deposited at a relatively warm location at a low level in a glacier.The results of the isotopic analysis show a variation within samples from the same iceberg that indicate differences in deposition temperature of up to 2 °C.This variation is to be expected as the distance between the samples are at least 2 meters.Only if the samples by chance were from the same horizontal layer they should be the values be equal within an iceberg.If the 2 meters represent vertical ice column distance, the two samples could be from layers deposited several hundred years apart.The δ 18 O and the δ Deuterium values also reveal a bimodal distribution.The δ 18 O values from the northern and western part of the fiord system, with a gneissic geology, are grouped around -32 per mil, while the samples along the south coast with basaltic rocks are grouped around -22 per mil.Similarly, for δ Deuterium, the samples from the gneiss areas group around -250, while those from basalt areas group around -180.No clear grouping is found for the D-O (Deuterium excess) values.

Transport of sediment by calving
We assume that the average concentration of samples from a source area represents the true average concentration of sediment in the ice calved from that source area.The mass of transported sediment is then calculated by multiplying the measured volume of ice calved from the source area over the year 2018 with the average sediment concentration.Here we use the measured ice flux from the 15 glaciers based on work by Mankoff et al. (2020).Our calculation of the ice transport out of Scoresby Sound is shown in Table 4, together with the coordinates of the glacial flux gates.The entire volume of calved ice is 19.5 km 3 , of which the Daugaard-Jensen Glacier alone accounts for 53%.In comparison, modelled meltwater flux from this region amounts to 10 km 3 yr -1 (as calculated from analysis of runoff predicted by RACMO, in Overeem et al., 2017).The calculation in Table 4 reveals a transport of 234 million tons with a large uncertainty.The average concentrations for the 15 source areas are determined as an average of concentrations from 3-10 icebergs.Small local glaciers on Milne land and Renland are not included because measurements of calving volume are lacking.In addition, these glaciers would have been represented only by one sample of sediment concentration (22).Only three icebergs (22, 23 and 24) are measured at locations that could represent ice from the Daugaard-Jensen and Charcot source areas.The two samples closest to the Daugaard-Jensen Glacier have concentrations of only 12 to 40 mg l -1 ; whereas the inclusion of sample 22 brings the average up to 14066 mg l -1 .This clearly demonstrates the large uncertainty resulting from the use of only a few samples of sediment concentration.The accuracy of the volume of calved ice is reported as ± 10% of the average of the total output of 500 Gt yr -1 .However, the accuracy of estimates from single glaciers can be less favourable as seen from table 4. Using the 10% on the calved volume and the lowest standard deviation on our average concentrations of 97%, then the combined uncertainty will be the square root of 10 2 +97 2 equal to ± 97.5 %.The uncertainty is dominated by the uncertainty of the concentration average.The transport calculation excludes two outliers, iceberg 3 (442486 mg l -1 ) and iceberg 9 (121256 mg l -1 ), because they were selected specifically to obtain samples of sediment.Including these two samples, the transport would have been biased towards high concentrations and our calculated transport would be larger than the "true" transport, perhaps representing an "upper limit".However, if we assume that the rest of samples with high concentration represent the right proportion of "dirty ice" in the melange of icebergs, then our estimate could be correct.Iceberg 11 is also included, although the isotopic analysis indicates that it could consist of sea-ice.This iceberg is still included because the sediment concentration was 1693 mg l -1 , which is considered unlikely for sea-ice from the open sea.However, at the beginning of the sampling, the distinction between sampling aimed at geological origin and sampling aimed at determination of sediment transport was not strict.The limited acces to the MOB resulted in sampling biased towards sampling "dirty" icebergs to get at least one sample from a source area.Here we compensate for this bias by omitting all icebergs used for geology description in Table 2.The average concentration of sediment in the remaining 12 icebergs (4973 mg l -1 ) is multiplied with the volume of calved ice, giving a transport of 97 million ton yr -1 as our best estimate.An estimate of the "minimum" transport is found by multiplying the average concentration (14.4 mg l -1 )from the 5 icebergs with the lowest concentration with the volume of calved ice, only 0.3 million tonnes, as shown in Table 4.

Drift velocity and temperature
In some cases, it was possible to calculate drift velocity and direction from the GPS observations.We recorded velocities of up to 6 km hour -1 .Most often, we found that the drift direction followed the wind direction when the wind speed was more than 5 m/sec.The maximum air temperature was 10 o C and water temperature ranged from 3 to 8 o C indicating that the melt rates at the surface can be high.

Discussion
Here we discuss strategy and limitations to our sampling approach, relate the sample descriptions to different modes of glacial sediment transport, provide discussion on provenance and possible identification of source area and put the estimated sediment transport into perspective of suspended sediment flux.It is not possible to measure the volume of an ice sample in the field and it is difficult and expensive to bring the sample to the laboratory in frozen state.Therefore, we use given densities of 2.65 and 0.9 g cm -3 for sediment and ice respectively to calculate the volume of ice.As basalt is heavier than gneiss, it could be advocated that actual densities should have been used.However, using a range of density for rocks from 1.74 to 3.3 g cm -3 would result in a change in ice volume of ~2 per mil.We don´t know the actual density of ice.Our density of ice differs slightly from the density (0.917 g cm -3 ) that has been used to calculate the measured volume of calved ice to water equivalent, a difference of 1.9% that is considered negligible compared to the other uncertainties.We collected samples with an average weight of 0.7 kg with a maximum of 1.7 and minimum of 0.3 kg.It appears that small samples can result in anomalous results, as for instance iceberg number 3, which would not be able to float, as the density of its smallest sample, is larger than that of water.This iceberg also has a very high standard deviation of 173% of average concentration, and this inhomogeneity is caused by the biased selection of this iceberg for collection of rock pebble and sediment sample to determine rock type and mineralogy.Another iceberg, number 9, also has a very large concentration, but with a standard deviation of only 78%, showing that the sediment can be more homogenously distributed within an iceberg.For safety reasons we chose only to collect samples from icebergs within the WMO size classes bergy bits and growlers.It could be argued that icebergs of this size do not represent the population of icebergs drifting around on the way from the calving front to the open sea.Hypothetically, small iceberg size could be the result of melting of a larger iceberg, with the implication that has lost its basal sediment while melting, and therefore has concentrations below average.From our field observations, we have seen very large icebergs drop off bergy bits and growlers so they are followed by a tail of these along their travel route.We have also observed large icebergs suddenly disintegrate into minor bits and pieces because of inner tensions or after capsizing.Remote-sensing based observations (e.g.Sulak et al., 2017) and theory on iceberg size distribution show that initially icebergs follow an exponential or power-law distribution (Kirkham et al., 2017) which is resulting from elastic-brittle fracture processes (Åström et al., 2020).Over periods of transport this size distribution evolves reflecting the grinding and crushing processes in an ice mélange (to a power-law distribution with a different exponent), (Åström et al., 2020).These studies support the assumption that large toppled icebergs can still generate growlers or bergy bits by grinding and crushing, even after a substantial period of melt in the mélange.Considering melt-rates, even minor icebergs have enough mass to persist all the way to the open sea.Therefore, we assume that sampling growlers and bergy bits will not result in an entirely biased estimate of the concentration of sediment in icebergs travelling from the calving front to the open sea.We collected three samples from each iceberg in order to quantify the variation of concentration within an iceberg.We find that the standard deviation can be up to 173% of the average concentration, a better measure of the average concentration could be obtained by collecting more samples.However; storing, transportation and analysis would have been too costly in this pilotstudy.A proposed solution could be to collect e.g. 10 subsamples, store them in one container and analyze only the bulk sample.To obtain samples representing the true distribution of concentrations we used random sampling to avoid subjective choices of icebergs and sampling positions as recommended by Yates (1960).We use the term "systematically random" to describe the fact that we collect samples with an equidistant distance of multiples of ruler length, claiming that our samples are true random samples because no known processes could distribute the sediment systematically over this spacing.The iceberg chosen to be sampled is found as an iceberg of the correct size closest to a GPS position pre-chosen as a sampling position on a map.We endeavored to get as close to the calving front of a source area as possible.Looking at the UAV photo fig.6, it appears that a number of "dirty" icebergs are floating around between a much larger numbers of "clean" white icebergs.By collecting our samples from randomly chosen icebergs, the "dirty" icebergs should theoretically occur in our samples in proportion to their occurrence within the whole population of icebergs.Accidentally we can either collect too many or too few "dirty" icebergs resulting in either an overor an underestimation of the average population concentration.In future UAV imagery, such as shown in fig.6, may serve to calculate the proportion of "dirty" icebergs, and allow stratified random sampling by collecting samples of a smaller number of "dirty" icebergs and a larger number of the "clean" ones according to their actual proportion.Similarly, Anderson et al. (1980) in a study from Antarctica observed icebergs from a helicopter and found 4 "dirty" icebergs out of 370.
In the Scoresby Sound catchment, the Daugaard-Jensen Glacier is the largest single producer of calved ice, responsible for more than half of the total ice flux.Because of logistical constraints, we were only able to collect two to three samples that likely represent the concentration of sediment from this source.To obtain a transport value from this source with a narrower band of uncertainty, the number of randomly chosen icebergs should ideally be taken proportional to the output of ice from the single source areas.Characteristics and origin of the sediment found in icebergs is described by Dowdeswell (1996) and Andrews (2000).They separate sediment into three types of origin: supraglacial, englacial and subglacial.The supraglacial component consists of dust and rocks fallen on the surface of the glacier, which in the ablation zone will accumulate by surface melt and form a layer of sediment that can be from a centimetre to meter thick at the terminus.Other sources of supraglacial sediment closer to the ice sheet margin, are rock material eroded from valley sides and nunataks, forming medial and side moraines and often observed as bands of sediment in icebergs.Englacial sediment partly originates from dust deposits on the surface gradually accumulating throughout the ice column and then moving with glacial flow.This englacial component is very small.Based on ice cores, Ruth et al. (2003) found concentrations of 0.05 to 8 mg kg -1 throughout the 1500 m long NGRIP ice core, originating from the ice sheet interior.Larger dust concentrations can be expected in upper ice layers formed closer to the ice sheet margin, where local outcropping rocks areas contribute to the dust deposition.The third, subglacial component originates from erosion at the glacier bed and may consist of both glacial flour from abrasion and rocks plucked from the bedrock valley bottom.Increased sediment concentrations in the basal ice are even observed in interior ice cores.Sediment entrainment is attributed to deformation at lower concentrations, regelation and freeze-on or frozen fringe processes at high concentrations (Alley, 1997;Gow and Meese, 1996;Meyer et al., 2019).Observations at exposed glacial margins indicate that a basal layer with extremely sediment concentrations can be 1 to 5 m thick.In Antarctica, icebergs have been observed with 12-15m thick sediment-laden ice (Anderson et al.1980).Samples at exposed basal ice along the margin of a tidewater glacier in Baffin Island indicated an average volume concentration of 35% (Dowdeswell, 1996).In the present study, we clearly found evidence of both supraglacial and englacial deposits within the sampled icebergs, and possibly some that represent basal ice.We observed one iceberg, see fig. 3, that most likely represented the basal layer, but unfortunately, we were not able to collect samples.Because of the basal position and the density of this layer, it will only be exposed after toppling of the icebergs during calving, or later because of iceberg capsizing due to imbalance because of underwater melting.The calving regime of the Daugaard-Jensen Glacier appears to be generating more tabular icebergs, and is perhaps less affected by toppling, and because of this, it may be difficult to obtain samples from this basal sediment-laden ice.Sampling could also be stratified according to the glaciological origin of the sample.In Overeem et al. (2017), the calculation of the transport is based on a volume of englacial ice with a volumetric concentration of sediment found from a few measurements.Such "clean" ice is described in cores from the Greenland Ice Sheet, Ruth et al. (2003).The englacial ice represents the "clean" white ice originating far from ice sheet boundaries and mainly receiving its sediment as windborne dust.More importantly, calculations account for a volume of ice from a bottom layer of ice with 20% of sediment by volume and an estimated average thickness of 3 m (Andrews et al., 1994).However, in East Greenland on its way from the drainage divide to the calving front the moving ice erodes at its boundaries with the valley bottom and mountain sides as demonstrated by the thick supraglacial debris layers originating when the ice passes around nunataks (fig.2).Closer to the calving front the glacier can receive sediment input from rock fall and sand blown in from neighboring out-wash plains.Due to the melting at the surface sediment is accumulated on the surface in the ablation zone of the glacier.This surface sediment may enter crevasses by slumping or through transport by supra-glacial meltwater.The fast flow of the ice and the formation and closing of crevasses result in a chaotic distribution of sediment in the uppermost ice column at the calving front where it is very difficult to distinguish between ice of different origin.However, at least in our study, this supraglacial component appeared to be a significant additional source term.In particular, it is difficult to obtain samples representing the bottom layer (fig.3).Growlers and bergy bits are formed at the terminus during the calving or produced from larger icebergs by calving.However, they could also be remnants of a large iceberg after transport, melting, grinding, and crushing in the melange.Complicated transport patterns of bergy bits are described by Carlson et al. (2017), in the Nuup Kangerlua (Godthaabs Fiord) where they found that bergy bits followed local gyres and were more sensitive to local wind patterns than larger icebergs.Because of their small size, growlers and bergy bits will melt fast, so that their lifetime will be short compared to that of large icebergs.The sediments found in these small icebergs will therefore be deposited close to the sampling location.Samples collected close to the calving front will represent the local transport well, but for samples collected far from a known calving front knowledge about the transport history of the iceberg is needed to interpret the sediment content.Syvitski et al. (1996) describe the complicated transport patterns in Kangerlugssuaq Fjord, East Greenland.There is an annual cycle of iceberg transport.During the winter, icebergs cannot leave the fjord because of the formation of sea-ice that has closed the fjord.The calving forms a melange of broken ice termed sikussak in front of the terminus, where the icebergs can have a residence time of up to two years.When the sea-ice gradually disappears during summer, the icebergs escape the sikussak and transit the fjord in about 68 days.In this study, the sampling was carried out in late August.At that time, sea-ice had totally disappeared, but the samples were confined to areas outside the sikussak both in Gaasefjord and in Rødefjord.Along the north coast of the Geikie Plateau we could get closer to the calving fronts.Nordvestfjord was dominated by large icebergs from the Daugaard-Jensen Glacier while rather few bergy bits and growlers occurred.Calving from a big iceberg was observed in this area but the sampled icebergs 23 and 24 were not observed as being recently broken off from a large iceberg.Tidewater glaciers further to the north of Scoresby Sound may lose most of the sediment from the bottom layer, because it is locked up in the sikussak where it is exposed to melting by the seawater, Reeh et al. (1999), Bigg (1999), Enderlin et al., 2018.The effect of this residence time in the melange and melting is that any icebergs released from the sikussak will contain less sediment to be transported long distances.Several authors have described the problems of trying to predict the occurrence and frequency of calving, and subsequent transport routes e.g.Benn et al. (2017) , Bond and Lotti (1995), Choi et al. (2018) , Death et al. (2006), Meire et al. (2017) and Todd et al. (2017).They point to the importance of basal topography, bottlenecks in fjords, trapping and bouncing into valley sides and meteorological factors.From these discussions, it is obvious that the transport of sediment by icebergs involves complicated processes that are capable within short time intervals to move the sediment both in and out of a fiord depending on freshwater outflow, tidal currents, wind direction and speed and interactions with warm ocean water.Here we simplify and assume that all calved ice from 2018 left the Scoresby Sound in 2018, and that the sediment concentrations in icebergs measured in August 2018 represents an average value for 2018.The validity of this assumption has to be tested by more detailed studies, covering iceberg transport dynamics and a longer time span.
Our other aim was to investigate whether sediment found within an iceberg could be attributed to a specific source location or in areas without geological mapping give an information about the geology underneath the ice.Preliminary analysis shows that macro particles fell into two groups; one of igneous rocks, and one of crystalline metamorphic rocks and metasediments.The igneous rock could clearly be related to the Geikie Plateau south of the fjord, where basaltic deposits are shown on geological survey maps (Fig 5).The other rock types are related to the geology in the western and north-western part of the fjord, but it was more difficult to relate these to a more specific location.The basaltic rocks from the southern part of the fjord have the potential to be used as tracers to study IRD transport within the fiord system, indicating that observed IRD was transported to the sea bottom from the Geikie Plateau.The basaltic rocks of this origin may occur in samples collected from the bottom of the Atlantic Ocean, but analyses that are more specific are needed to distinguish these basaltic rock fragments from IRD that originates from other similar source areas e.g.Iceland and the Faeroe Islands.Isotopic analysis was carried out in order to exclude samples comprising sea ice, with no or very little sediment content, from an investigation of the sediment concentration in calved ice.The isotopic analysis was able to separate the samples into uniform classes related to the assumed source areas.One sample was different from all the others.This could have originated from an iceberg that had drifted into the fiord from a source far away, it could have originated from a low altitude location of a local source, or it could have been derived from sea ice.The last possibility was ruled out by the sediment concentration, which was likely too large to be from sea ice.These results indicate that isotopic analysis is a valuable tool for selecting samples that need further analysis before being accepted as representing the population of icebergs originating from any particular fiord system.

Conclusions and perspectives
In this pilot project, we have developed a new procedure and protocol to obtain un-biased samples from icebergs in order to evaluate their role and share of the total sediment transport from Greenland to the Atlantic Ocean.We have discussed problems related to the sampling procedure and to the complicated transport routes of the icebergs from the calving front to the open sea.We demonstrate large variations in the concentrations of sediment, both within any individual iceberg and spatially between icebergs in Scoresby Sound.We present an estimate of the transport related to calving of ~100 million tons with a range from 0.3 to 200 million tons.We are fully aware that 12 samples of icebergs are not sufficient to constrain the transport of sediment reasonably accurate, but we have performed the calculation to illustrate the effects of biased sampling.The simultaneous meltwater-driven suspended sediment fluvial contribution is 21-30 million tons (from Table 1 in Overeem et al., 2017).
We have also demonstrated the potential of collecting samples from icebergs to describe the travel routes of IRD and/or to obtain new knowledge about the geology of the land underneath the calving glaciers.However, this sampling should be more focused in the selection of distance of sampling from calving front and in obtaining larger amounts of fine sediments for more sophisticated geological and mineralogical analysis, including quantification of the angularity of the coarse sediment fraction.The utility of isotopic analysis is demonstrated, but probably the interpretation of the analysis and the information that could be extracted from the analysis could be improved significantly by comparing samples collected simultaneously from the glaciers upstream of their calving fronts.It is important to state that this pilot study was of short duration and had only limited access to a major calving front (Daugaard-Jensen Glacier).Thus, this dataset has limitations when extrapolating the results in time and space.In particular, we recognize the complexity related to trace the travel route of any sampled icebergs.Larger icebergs are nowadays tracked by satellite by for example the Danish Meteorological Institute (DMI), but the smaller bergy bits and growlers are not.We envisage the potential of the use of UAV´s to fill this gap in knowledge, as also shown in Ryan et al. (2015).In relation to climate change, this component of the Global Sediment Cycle will diminish or disappear when the global amount of ice capable of calving diminish during warm periods.In cold periods, this component will increase, as the calving fronts move closer to the open sea, so that the icebergs do not lose their sediment load in fiords.This effect is clearly demonstrated by IRD deposits found far south in the Atlantic Ocean.This pilot study focus on methodology intended to be used in the study of the transport of sediment by calving in Greenland.It clearly demonstrates the importance of systematic unbiased sampling.In the future we intend to utilize the experiences from this investigation by using a clear definition of sample aim, a relevant stratification of samples to secure unbiased representative sampling and a sufficient number of samples.The results of our studies of the sediment content of icebergs calved from Greenland are published in the Isaaffik data portal (see references), where also future data will be published.

Fig. 1
Fig. 1 Map of sample locations in Scoresby Sound, East Greenland.SDFE (Agency for Data Supply and Efficiency, Denmark)

Fig. 5 A
Fig.5 A,B Isotopic analysis of selected samples

Fig. 9 A
Fig. 9 A. View of calving front at Sydbrae .Calving front 300 m high, surrounding mountains 2000m.Observe high concentration sediment stripes, medial moraines on the surface and within the calving front, photo location close to iceberg 12 in figure 1 (photo B.Hasholt).B. Satellite image of Sydbrae Glacier (USGS) illustrates how local rockwalls and nunataks source abundant supraglacial sediment.

Table 1 :
Average sediment concentration in 24 icebergs

Table 2 :
Geological evaluation of sample origin Sample 25A and 26A are collected near location 13 and 19 respectively, but the exact position was not recorded

Table 4 :
Transport of calved ice from the Scoresby Sound catchment.