A novel probe to sample dissolved and particulate matter in sea ice at high vertical resolution

Sea ice contains dissolved and particulate matter scavenged and concentrated from the ocean and atmosphere. Matter is mostly partitioned between the solid-ice fraction and brine, though there can also be a considerable gas fraction (Salganik et al., 2024). Brine is a salty liquid that forms when seawater freezes, as the dissolved salt is expelled from the forming matrix of ice crystals. Considering that the growing interface of ice crystals in composed of ice lamellae with brine layers in between them, the scale of this brine formation is of the order of one millimetre (Weeks and Ackley, 1982; Wongpan et al., n.d.). Openings can form in sea ice when cold brine cracks the ice or when very salty brine melts ice that is warm enough, creating a network of channels within sea ice with channel diameters ranging from sub-millimetre to centimetres and network extents reaching up to metres (Bennington, 1967; Lake and Lewis, 1970). Within this brine network, matter can be transported due to buoyancy, diffusion and turbulence and be transformed by dissolution, precipitation, primary production and heterotrophy (Meiners and Michel, 2017, and citations therein). These processes, which begin occurring at the finest scale of the brine network (order of one millimetre), are closely tied to sea-ice biogeochemistry and control, for example, of the supply of nutrients available for ice-algal production, which is important for carbon uptake and food supply for marine ecosystems (Delille et al., 2014; Swadling et al., 2023). Thus, fine-scale in situ measurements with associated process studies are crucial to understanding the role of sea ice in the Earth system.

Sea-ice texture, which is observed by cutting cores vertically so that no part of the ice column is missed, can show horizontal layers around 0.01 m thick with texture and thus properties different from their surroundings (Eicken and Lange, 1989; Golden et al., 1998; Wongpan et al., 2018; Tison et al., 2020; Corkill et al., 2023). This heterogeneity is difficult to capture with ice coring where vertical resolution ranges from 0.05 m ice sections to whole cores (>1 m). Observations at low vertical resolution are an issue for sea-ice modelling, where the fine-scale processes mentioned above are important and layers of 0.03 m and thinner are considered (Tedesco et al., 2010; Duarte et al., 2015; Vancoppenolle and Tedesco, 2017). Gap layers within ice are also challenging to sample; they cannot be collected by corer and are only seen as breaks in texture cores but are thought to be important for sea-ice algae and ice mass balance (Ackley et al., 2008).

Variability over fine vertical scales (millimetres to centimetres) has also been noted for continental ice cores, and continuous melting techniques have been in development for glacial ice applications since at least the 1990s (Sigg et al., 1994). The principle is loading a stick of ice cut from a core onto a melt head, with galleries for collecting sample. Considerable brine drainage/movement would occur before a sea-ice core could be processed using a system such as this one, making it not very suitable for sea ice. Melting devices have also been prototyped for penetrating extraterrestrial ice layers, such as the polar areas of Mars and Jupiter’s moon Europa, though their primary goal is to reach underlying water rather than sample ice and they are designed accordingly (Kaufmann et al., 2009a; Kaufmann et al., 2009b). More recently, a sea-ice endoscopic platform has been under development (Babin et al., 2019; Perron et al., 2021; Larouche et al., 2023), though the endoscope focuses mainly on small-scale non-destructive optical measurements, rather than larger volumes required, for example, for particle filtrations. Inspired by these previous works, we developed a system to bore into sea ice and collect 0.01 m scale meltwater samples with a focus on measuring biogeochemical parameters without contamination, including trace metals. Here we report initial results from deployment of our melt probe on both natural and laboratory-grown sea ice. Our research questions were: (1) from a technical perspective, can we collect melted sea-ice samples at 0.01 m vertical resolution?; and (2) are concentrations of dissolved and particulate matter in sea ice retrieved by melt probe higher than by ice-core sampling; that is, can we collect brine that is often lost by ice-core extraction and processing?

The melt head of the probe was a bespoke 63.5 mm diameter cylinder with a hemispherical tip made from electropolished 316 stainless steel (SS; Figure 1; Ezy-Fit Engineering, Australia). Electropolished SS has previously been proven as a material of choice for trace-metal ice coring due to being contamination-free (Lannuzel et al., 2006). The melt head had a central well for the heater element and opposing off-centre ports for a Pt100 Resistance Temperature Detector (RTD; 1/4” diameter 316 SS sheath with Teflon cable; Technitemp, Australia) and a sample inlet (5 mm diameter). The heater element was a bespoke 200 W, 10 mm in diameter and 2 m long with only the bottom 40 mm (inside well) heated, and powered by a 240 VAC supply (Helios, Australia). The heater element was retained within the melt head using a 316 SS compression fitting with tetrafluoroethylene ferrules to prevent from crushing the heater element (Swagelok, USA). The overall length of the heated solid section of the melt head was 53.5 mm, with the highest point of the sample inlet located 10 mm above the tip of the hemisphere (Figure 1). For ingress monitoring, 10 mm increments were marked on the heater element with a very shallow cut made by a tubing cutter. The bottom of the RTD port was 2 mm from the melt head–ice interface, and the RTD was connected to a proportional-integral-derivative controller (ATR244-12ABC, Pixsys, USA) which output a 420 mA signal to a solid-state relay (RM1E23AA25, Carlo Gavazzi, Switzerland) to switch the heater. The heater element was made using 10 mm 316 SS tube which was filled with a compacted refractory powder (mineral-insulated). Power was supplied through a 30 mA residual current circuit breaker with overcurrent protection. The melt probe currently requires a 100–240V generator or ship/shore power.

A peristaltic pump using 7/16” outside diameter (OD), 1/4” inside diameter (ID) C-Flex silicone tubing and an Easy-Load II head (Masterflex, Germany) was connected to the sample port in the melt head with 2 m of 1/4” OD, 5/32” ID perfluoroalkoxy tubing (approximately 25 mL internal volume and therefore lag before sample reaches the container; Savillex, USA). The pump speed was approximately 270 mL min⁻¹ and the ingress rate was 4.0 ± 1.3 mm min⁻¹ with the melt head operating at 15.1 ± 1.5°C. The pump did not have flow-rate feedback to control the pump speed; the speed was adjusted manually to achieve minimal flow while keeping the pump tubing primed with liquid. The ingress by the melt head was gravity-driven so the rate was governed by melting speed. In the field but not laboratory, a second peristaltic pump drawing from a well at the top of the melt head and discharging into a core hole a small distance away (to avoid flooding the snow on site) extracted brine entering the borehole via lateral percolation. In the −15°C constant-temperature laboratory, sample tubing was heated using a 10 W m⁻¹ (at 5°C) self-regulating heating cable (XL-Trace LSZH pre-terminated; nVent Raychem, USA) and insulation to prevent freezing in the sample tubing. Care was taken to leave as little sample tube unheated/uninsulated as possible to maintain pumping.

An experiment was undertaken at the Hokkaido University Institute of Low Temperature Science using two ice tanks, one for the melt probe and one for ice coring, to test for carryover after the melt head passes through a layer of some concentrated solute (visualised in Section 3.2). Sea ice was grown from seawater collected from the Sea of Okhotsk in a −15°C constant-temperature laboratory. After the sea ice had grown to a few centimetres thick, a 5 mm layer of freshwater near the freezing point was poured on top to seal the surface. After that layer had frozen, a 5 mm layer of red food colouring (85% Dextrin, 15% Edible Red No. 102, an aromatic hydrocarbon) in freshwater near the freezing point was added. The final thickness was 0.05 m, with 0.01 m of fresh ice including the red colour underlain by 0.04 m of sea ice. The fresh sealing layer, final salinity of 1.6, was added to prevent the dyed freshwater from percolating through the existing sea-ice layer. Ice temperature was monitored using Testo 110 probe thermometers (accuracy: ± 0.2°C; Testo, Germany) at the top and bottom of the ice. Samples for dye concentration and salinity were collected at 0.01 m vertical resolution, using press-seal bags for melted samples. A jab saw was used to cut ice from the second tank. The ice core was cut into 0.01 m sections using a bandsaw. Critically, this high-resolution sectioning in a subzero constant-temperature laboratory using a large vertical bandsaw is not feasible in the field, especially for longer cores where the extended processing time will lead to considerable brine drainage/movement. The concentration of red dye was determined using a spectrophotometer (Hitachi, U-1900) using absorbance at 506 nm, which was linearly related to dye concentration (n = 4, R² = 1.00, root mean square error = 0.02%). Salinities were determined using a WTW Cond 3110 conductivity meter (accuracy: ± 0.1; Xylem Analytics, Germany).

Saroma-ko Lagoon (Figure 2a and b) connects to the Sea of Okhotsk and is covered by sea ice from January to April (e.g., Nomura et al., 2024). The sea ice grows to about 0.4 m thick and resembles fast ice formed in polar regions. Our observations contribute to longer-term efforts that have been conducted at this sampling site since 2019 (Station Nosaka, or St. N; 44.12°N, 143.96°E; Nomura et al., 2020). This work, conducted from February 28 to March 2, 2023, was part of the Sea Ice Physico-Chemistry and Ecosystems 2023 (SLOPE2023) campaign (Nomura et al., 2024). In the field, the team followed trace-metal-clean sea-ice sampling procedures (Miller et al., 2015), that is, by wearing coveralls (Tyvek, USA) and polyethylene gloves over warm clothing (Figure 2d). Care was taken to work downwind of and not to walk on the trace-metal-clean site (Figure 2c and d).

Three field experiments were conducted, each lasting for approximately 3 hours including melt-probe deployment and ice-core processing:

1. a high-resolution experiment (0.01 m) to collect samples for determining sea-ice bulk salinity of snow-free sea ice;

2. a medium-resolution experiment (0.07 m) to determine several sea-ice parameters, that is, salinity, stable oxygen isotope ratio, dissolved and particulate trace metals, dissolved organic carbon (DOC), chlorophyll a (chl a), pulse-amplitude-modulation fluorometry parameters and macro-nutrients of snow-free sea ice; and

3. a deployment on top of the snowpack and 0.1 m into the sea ice.

Prior to the snow-free experiments (1 and 2), snow was removed from the sampling quadrant using an acid-cleaned plastic shovel. Melt-probe deployments were compared to ice cores collected using a bespoke grade 2 titanium 0.09 m ID ice corer (Commonwealth Scientific and Industrial Research Organisation, CSIRO, Australia) cut into approximately 0.05 m (depending on natural breaks in the ice) sections using a grade 2 titanium hacksaw (CSIRO, Australia). Temperature measurements were made on ice cores with the same Testo 110 probe thermometers used for laboratory experiments.

The ice corer and hacksaw were cleaned thoroughly with a strong alkaline detergent (2% decon 90) and copious ultrapure water. The ice corer was used to collect several ice cores, rinsing with ultrapure water in between, before collecting the trace metal core. Similarly, the hacksaw was used to cut ice grown from ultrapure water as well as the ice cores collected prior to the trace-metal core, with ultrapure rinses in between, for cleaning. Ultrapure-water rinses might not be required in areas with low enough trace-metal concentrations such as parts of Antarctica. However, rinsing was employed here as Saroma-ko is a shallow lake and as field conditions were warm enough for the ultrapure water not to freeze in the container or onto the equipment. All plasticware used for trace-metal samples was cleaned following GEOTRACES protocols (Cutter et al., 2017). Samples were collected in press-seal bags for the 0.01 m experiment and opaque white acid-cleaned buckets for the 0.07 m experiment, as per Figure 1b for the melt probe and with each bucket housing a section for the ice cores. Brine samples were also collected from the top of the flooded borehole by pipette during the 0.07 m experiment (Figure 3), after each 0.07 m of ingress. Sample bags/buckets were transferred back to the field laboratory and melted in the dark at approximately 5°C before being analysed/subsampled. Downstream of the buckets, organic carbon samples were processed using metalware and glassware that had been combusted at 450°C for 8 hours (to burn any organic carbon present) and rinsed with ultrapure water between samples.

Samples were homogenised by gently swirling buckets before subsampling. Trace-metal samples were collected first in a bubble setup for trace-metal-clean work in the field laboratory. A sample volume of 45 mL was transferred to a vacuum filtration system (1 L jar with transfer closure and filter holder assembly, Savillex, USA; Air Admiral, Cole-Parmer, USA) loaded with a 47 mm diameter, 0.2 µm pore size polycarbonate filter (Isopore, Merck Millipore, Germany). The filter was collected for particulate trace metals; the filtrate was collected for dissolved trace metals and acidified to pH <1.8 using hydrochloric acid (TAMAPURE-AA-100, Tama Chemicals, Japan).

As quickly as possible and with minimal light introduction, pulse-amplitude-modulation fluorometry (PAM; WATER-PAM with cuvette, WALZ, Germany) was used to measure dark-adapted maximum quantum yield (Fluorescence_variable/Fluorescence_maximum; Fv/Fm) in triplicate, as a measure of photosystem II health. The dark acclimation period was approximately 24 hours, which is very long, but was necessary to allow complete melt of the ice-core sections. Ice algae may also have been subjected to hypoosmotic stress as filtered seawater was not added during melting to replicate in situ salinity (Campbell et al., 2019). This decision was in favour of replicating melt-probe sampling which did not include any salinity adjustment, though an adjustment could be implemented in the future.

Organic carbon was collected using a 100 mL glass syringe (Eterna Matic 210-100, Sanitex, Switzerland) filled to 40 mL, coupled to an SS filter holder (XX3002500, Merck Millipore, Germany) loaded with a combusted 25 mm quartz micro-fibre filter (T293, Sartorius, Germany). The filtrate was collected for DOC in a precleaned certified 60 mL borosilicate glass vial (Environmental Sampling Supply, USA) and stored at −20°C. The 100 mL glass syringe was also used to collect 20 mL samples for stable oxygen isotope ratios (δ¹⁸O).

Macro-nutrient (nitrate, nitrite, phosphate, ammonium, silicic acid) samples were filtered through a 0.22 µm pore size Durapore polyvinylidene fluoride membrane filter (Millex-GV Filter Unit, Merck Millipore, Germany), placed in the 10 mL polyethylene screw-cap vials (Eiken Chel Co. Ltd., Japan) and stored at −20°C. Samples for chl a measurements were immediately filtered onto 25 mm diameter Whatman (UK) GF/F filters. Pigments on the filters were extracted with dimethylformamide (Suzuki and Ishimaru, 1990) for 24 hours. Salinity for the 0.07 m experiment was measured last to avoid sample contamination.

Trace-metal samples were transported back to the Institute for Marine and Antarctic Studies, Hobart, Australia, for preparation and analyses. Dissolved trace-metal samples were analysed using NASS-7 as a reference material (following dilution 10 times with 10% v:v distilled nitric acid). Particulate trace-metal analysis employed BCR-414 (plankton), Arizona Test Dust and MESS-3 (marine sediment) as reference materials. All samples were oxidised (1 mL 30% w:w hydrogen peroxide and 1 mL distilled nitric acid at 100°C for 30 minutes) and then digested in closed 15 mL Teflon vials with a mixture of trace-metal-clean acids (0.25 mL hydrofluoric, 0.25 mL hydrochloric and 0.75 mL nitric at 95°C for 12 hours; Lannuzel et al., 2014). The strong acid mixture was then dry-evaporated, samples were resuspended in 10 mL, and then diluted 50 times using 10% v:v distilled nitric acid (with indium added as an internal standard). Dissolved and particulate samples were analysed for trace elements using a Sector Field Inductively Coupled Plasma Mass Spectrometer (SF-ICP-MS; Element 2, Thermo Fisher Scientific, Germany) offering multiple spectral resolution settings, at the University of Tasmania’s Central Science Laboratory. Iron (Fe), which was at risk of contamination by the melt probe, ice corer and hacksaw, was quantified alongside manganese (Mn) which should not have been at risk of contamination. Titanium (Ti) was also measured to assess the level of Ti contamination contributed by the ice corer and hacksaw. Fe, Mn and Ti were all analysed using SF-ICP-MS in medium-resolution mode.

In addition to Mn, strontium (Sr) was also measured as a contamination-free element for the data analysis. Sr is required by marine phytoplankton in low amounts comparable to Fe and Mn (extended Redfield ratio C₁₀₆:N₁₆:P₁:Si₁₅:Sr_0.005:Fe_0.0075:Mn_0.0038; Redfield, 1934; Brzezinski, 1985; Ho et al., 2003). Thoracosphaera heimii (dinoflagellate with calcified cell wall during vegetative life phase; Tangen et al., 1982) and two coccolithophores, Emiliania huxleyi and Gephyrocapsa oceanica, were excluded from the extended Redfield ratio due to their having a high requirement for Sr but not being previously documented in Saroma-ko Lagoon or the Sea of Okhotsk (WoRMS Editorial Board, 2024). Sr measurements were undertaken via SF-ICP-MS using low spectral resolution mode.

DOC was analysed using combustion catalytic oxidation and nondispersive infrared (TOC-V CSH, Shimadzu, Japan) with four-point linear calibrations (R² = 1.00, root mean square error = 0.007–0.008 mg L⁻¹) and using the Deep Seawater Reference material (Batch 19 Lot # 03-19). The δ¹⁸O was determined with an Elementar isoprime precisION isotope ratio mass spectrometer with the iso DUAL INLET system (standard deviation <0.025‰, n = 4) and referencing Vienna Standard Mean Ocean Water 2.

The concentrations of macro-nutrients were determined in accordance with the Joint Global Ocean Flux Study (JGOFS) spectrophotometric method (Intergovernmental Oceanographic Commission, 1994) using an autoanalyzer (QuAAtro 2-HR system; BL-tec, Japan). The analyzer was calibrated with reference materials for nutrient analysis (CRMs, Lot: CE, CL, CO, CG; KANSO Technos Co., Ltd. Japan). The concentrations of chl a were determined with a pre-calibrated (n = 5, R² = 1.00, root mean square error = 1.8 µg L⁻¹) fluorometer (Model-10AU, Turner Designs, Inc., USA) in accordance with the method described by Welschmeyer (1994).

Salinities were measured directly in sample bags/buckets, wiping the probe clean and dry between samples. Brine salinities and ice porosities (combined gas and brine volume fraction) were calculated from interpolated ice-core temperature and salinity profiles following Vancoppenolle et al. (2019).

The melt-probe setup performed operationally well in both the laboratory and field. Some noteworthy points are that the melt-probe temperature setpoint was 40°C but the 200 W heater element was not powerful enough to reach the setpoint with the melt head within the ice in the −15°C constant temperature room or in the flooded borehole in the field. The melt head could reach 40°C in the constant temperature room if not inside the ice. The waste pump could not keep up with liquid entering the borehole under the field conditions. Irrespective of these caveats, the melt probe penetrated the ice layer with a stable positive melt-head temperature and collected samples for every 0.01 m ingress; the borehole being full of liquid may have been beneficial for maintaining hydrostatic balance during the melt-probe field deployment.

For the carryover test with the melt probe, 96% of the dye signal was gone within 0.01 m and it had vanished within 0.02 m of the concentrated layer (Figure 4e). The sample volume increased below 0.03 m, which may have diluted the lower samples. The dye signal mostly dissipated almost immediately for the ice core, but there was some small signal all the way to the bottom of the underlying ice (between 0.6% and 1.1%). Melt-probe salinities were mostly closer to ice-core salinities than to brine salinities for the dye test (Figure 4d). Ice temperatures were as cold as −4.8°C at the top to as warm as −2.6°C at the bottom in the ice tanks (Figure 4c). These temperature bounds were determined from temperature measurements made at the top and bottom of the ice at the start and end of the melt-probe deployment. In the ice tanks, below the fresh layer (Figure 4c), the sea ice was permeable with the porosity exceeding 0.05 (Golden et al., 1998).

Salinities were higher for melt-probe samples than for ice-core samples in both the laboratory and field experiments. For field experiments, melt-probe salinities were close to calculated brine salinities, except at the ice surface where samples included some positive freeboard ice and melt-probe salinities were relatively low (Figures 3d and 4a). There were salinity peaks around 0.15 m and 0.3 m depths in melt-probe data from the 0.01 m experiment (Figure 5b). These peaks coincided with a generator failure and the melt head deviating slightly from vertical, both of which caused ingress to stall while pumping continued. This situation likely led to the sample including a high proportion of saltier brine that had percolated laterally into the borehole above; thus, interpolated lines smoothing these peaks are also shown in Figure 5b. The sea ice in the field was warm (above −2.4°C) with an isothermal profile. Correspondingly, the full ice column was permeable with porosity exceeding 0.1. Salinities measured in brine at the top of the flooded borehole for the 0.07 m experiment ranged from 27.2 to 34.6.

Acid cleaning was kept to a minimum during field processing for safety, and a small amount of Fe carryover was measured in procedural blanks. The carryover was negligible compared to sample values and as demonstrated by the efficiency of the ultrapure water rinses (84 ± 3% wash out after 9–82 nM dissolved Fe samples, and 99 ± 1% wash out after 303–497 nM particulate Fe samples).

Trace-metal recoveries for (10× diluted) NASS-7 were 94.4% and 97.0% for Fe (n = 2) and 108.7% and 108.7% for Mn (n = 2); Sr and Ti were not certified. Particulate trace-metal recoveries were between 90.7% and 109.2% (Table 1). Particulate Sr was on average 10 times more concentrated than particulate Mn, making it likely a more stable reference element. Particulate trace metal:Sr ratios were also generally lower than crustal and biological values (except ice-core Fe:Sr), supporting relatively high Sr concentrations (Table 2).

Chl a profiles were L-shaped and remarkably similar for the ice core and melt probe (Figure 6e). Patterns in photosystem II health as measured by PAM were also similar, although the melt-probe Fv/Fm values were higher. The DOC profiles were vaguely C-shaped and the melt-probe values were considerably higher than the ice-core values.

Stable oxygen isotope ratios showed similarly shaped profiles for the melt probe and ice core, with the melt-probe offset low (Figure 6c). For the snow experiment, even with the pump speed increased, no sample flowed while the melt head traversed the snowpack, suggesting that the melted snow was likely wicking into the surrounding snowpack. In support of the wicking argument, δ¹⁸O from 0–0.05 and 0.05–0.1 m (−6.7‰ and −6.3‰, respectively) in the sea ice was not lower than the 0–0.07 m sample from the previous day’s experiment (−6.7‰), as would be expected if snowmelt (−18.1‰ to −12.6‰) had percolated downwards and mixed with the brine/sea-ice melt.

Here we assess the different aspects of this new sea-ice sampling system as tested in this study. Strengths and weaknesses are identified, with possible explanations offered for anomalies in the datasets. Avenues for improving melt-probe performance in future studies are also highlighted.

The carryover test showed that the melt probe was able to sample most of a layer of concentrated solute while traversing the layer itself, measuring 96% of the solute from a 0.01 m layer within 0.02 m. This finding suggests that we could collect sea-ice samples at a vertical resolution of 0.02 m but not the 0.01 m from our first research question on the high-resolution capability of the melt probe. Some carryover may be due to non-laminar flow inducing turbulence and mixing in the sample tubing, rather than solute sinking into underlying ice which might be made more permeable due to warming by the melt head. Also, the sample inlet in the melt head is 0.01 m higher than the tip of the hemisphere which might have transferred some solute to underlying layers. Ingress tests with the melt probe going just beyond the solute layer before coring the underlying ice could help to resolve this uncertainty in future.

Melt-probe-derived “bulk ice” salinities from the field were close to calculated brine salinities (Figures 5b and 6b). Given the hemispherical melt-head design, fresh melt could have escaped and floated to the top of the flooded borehole causing the melt probe to sample mostly brine rather than bulk ice. Thus, in response to our second research question of whether melt-probe samples showed higher concentrations than ice-core samples, the answer was yes for salinity and to a greater extent than we anticipated, with the brine fraction not only being present but also apparently dominating. The expected volume of pure ice melt from the 63.5 mm diameter melt-head penetrating 0.07 m of ice with a solid fraction of 0.85 (mean from 0.07 m experiment) is 173 mL, ideally collected along with 33 mL of brine (206 mL total). After fine-tuning the pump speed, samples were at most approximately 100 mL over the expected amount, which should not have been enough dilution by neighbouring brine or seawater to completely mask the ice-melt signal without significant loss of melt to floating. However, salinity measured at the top of the borehole did not undergo appreciable freshening throughout the 0.07 m experiment. The missing fresh-ice melt is difficult to reconcile. Some possible explanations are that a fresh layer at the top of the borehole was made very thin by percolating into neighbouring ice and was missed when collecting samples by pipette, or that the melt was displaced into surrounding ice at the melt head–ice interface.

The δ¹⁸O profiles from the melt probe and ice core were similar in shape (Figure 6c), which supports the suggestion that our samples were collected from the melt head–ice interface rather than being a mixture leaked in from the borehole walls or seawater from below. The melt-probe values being on average 4.5‰ lower than the ice-core values could be explained by melt-probe samples being mostly brine that has been subjected to Rayleigh fractionation during sea-ice freezing (the heavier ¹⁸O is preferentially entrapped in sea ice, enriching ¹⁶O and lowering δ¹⁸O in brine) and previous snowmelt entrainment (O Crabeck and M Thomas, personal communication, 08/06/2023; Eicken, 1998; Toyota et al., 2013). The range of −6.7‰ to −5.6‰ from the melt probe is also in good agreement with a previous study that drained brine from ice cores at Saroma-ko Lagoon and reported values in the range of −7.5‰ to −5‰ (Nomura et al., 2009). Due to the positive freeboard of 0.025 m, the brine contribution and therefore salinity being lower in the top ice due to brine drainage seem sensible (Timco and Frederking, 1996). We do see low salinity; however, we would also expect δ¹⁸O to be higher than the underlying ice, which is not evident, perhaps due to an ex-snow fraction (snow ice or melt) lowering the δ¹⁸O.

Melt-probe salinities were higher than underlying seawater (Figures 5b and 6b) which supports our samples not being mostly seawater that has percolated up into the borehole. The borehole flooding due to lateral percolation of brines may have helped prevent upward percolation of seawater. With little-to-no mixing with sample at the melt head–ice interface, and no refreezing to trap the melt probe (both of which are foci of future designs), the flooded borehole could be considered useful for maintaining hydrostatic balance. In colder conditions such as for the ice-tank dye test, and in ice old enough for the permeability of the upper ice column to have been reduced by desalination (Notz and Worster, 2009), upward percolation of seawater may become an issue in the bottom few tens of centimetres of ice which are permeable. Perhaps flooding the borehole with an artificial brine or filling it with an inflatable plug could help to reduce seawater infiltration.

The geometry of the melt head is also an important consideration for future developments. The hemispherical melt head employed here was selected to promote downward boring and even heat distribution from the central cartridge heater. A flatter melt head–ice interface would be more optimal for collecting sample from a horizontal plane in the sea ice. However, given that our system collected mostly brine that could percolate in from ice directly beside the sample inlet rather than the mixture of melt from the entire melt head–ice interface, the resolution should be better than the height of the melt head; we suggest 0.02 m based on the dye test (96% recovery). A hemispherical melt head appears useful for brine sampling, though collecting higher resolution δ¹⁸O measurements and testing different melt-head geometries should be prioritised in future studies to corroborate this usefulness.

The melt-probe collected higher concentrations of Sr and Mn particles than the ice core (Table 2), providing further support for the melt probe to be able to sample higher concentrations of particles as per our second research question. Given that melt-probe samples were likely mostly brine, this collection of higher concentrations suggests that particle loads were higher in the brine than in the solid-ice fraction which dominates the ice-core samples. This finding is interesting considering that sack-hole brines (where a partial core is extracted, and brines are allowed to fill the hole) usually bear fewer particles than bulk sea ice because particles tend to remain within the ice matrix (Becquevort et al., 2009; Lannuzel et al., 2016; Duprat et al., 2019). The melt probe seems to resolve this undersampling issue by flushing the particles out of the ice-brine system. Previous work has also highlighted increased particle release from the solid-ice fraction of decayed sea ice like we sampled (see large voids in ice textures in Figure 5c, as well as high porosities in Figures 5a and 6a; van der Merwe et al., 2011). Interestingly, dissolved-to-particulate ratios for Sr and Mn were quite different between the melt probe and ice core, with the melt probe being about twice the ice-core concentration for Sr and 10 times higher for Mn. One potential explanation could be that size fractions differed between the two metals and a particular size fraction was collected more efficiently, presumably smaller Sr particles. This possibility could be assessed by size-fractionating samples in the future (e.g., Lannuzel et al., 2014).

Higher concentrations of Fe and Ti particles were measured for the ice core, but we suspect that these were due to contamination by the Ti ice corer which was deployed for the first time and, even though cleaned thoroughly, could have been subject to mechanical abrasion by the ice. According to Mohs scale hardness, Ti (Mohs hardness of 6) is much harder than ice and salt crystals like halite (approximately 2–2.5), though sufficient friction can cause softer ice to abrade harder metal and friction can be pronounced during an initial “run-in” period (Abdelnour et al., 2006). Additionally, diatom frustules and sand from sediment may be present in Saroma-ko sea ice (Takata et al., 2016; Nomura et al., 2024) and can be as hard as or harder than Ti (Mohs hardness as high as 6–7; Durham, 1973). Abrading the corer would release any remnant Fe impregnated in the Ti from Fe-rich tools used during manufacturing. Ice-core Fe:Sr and Ti:Sr (considering Sr as an optimal contamination-free reference element) were also high and unstable compared to melt-probe values, supporting contamination by the ice corer but not by the melt probe (Table 2). The melt-probe sampling showed relatively low concentrations of Fe and Ti, as well as L-shaped profiles for pFe and dFe (Figure 6d), which are commonly observed in sea ice (Lannuzel et al., 2016).

For macro-nutrients, melt probe concentrations were all considerably higher than ice-core concentrations in alignment with other evidence for the melt probe collecting higher concentrations of dissolved matter, with the exception of phosphate. Phosphate adsorption onto particles, such as sediment, has been shown to increase about 1.3 times for each 10°C temperature increase (Zhang and Huang, 2011). However, the same study showed that doubling salinity should decrease phosphate adsorption by a factor of 1.5, and this effect should have dominated. Another possible explanation is that phosphate in the surrounding ice from which the brine was being drawn by the melt probe was more strongly tied to microbial biofilm than the other macro-nutrients (Roukaerts et al., 2021). The silicon pump tubing could have been a possible source of contamination for the melt probe, but other macro-nutrients (except phosphate) showed similar levels of elevation which, again, are likely due to the melt probe collecting primarily brine.

Interestingly, even with the melt probe generally measuring higher particle and macro-nutrient concentrations, the melt-probe and ice-core chl a vertical profiles were remarkably similar, in contrast to the Sr and Mn particles which were enriched for the melt probe. The only explanation we have for these contrasting results is that perhaps the chl a concentrations were very similar in the brine and the solid-ice fraction and anything trapped inside its polycrystalline lattice. As for the different stresses that sampling by melt probe and ice core might present to algae, photosystem II health was higher for the melt probe. This better health could be due to the algal community having a greater thermal than osmotic tolerance (ice-core sections were not melted into filtered seawater and so were relatively fresh). Melt-probe samples were also subjected to a dark acclimation period much longer than necessary in order to match the treatment of ice-core samples; melt-probe PAM samples could have been measured immediately in the field. The ability to measure parameters immediately or even in line with the melt probe would be a major advantage of the melt-probe method, especially for time-sensitive parameters like PAM, as well as dissolved chemical species like Fe (II) that exist in sea ice but are quickly lost to oxidation after melting samples (Nomura et al., 2022).

This newly developed melt probe appeared to sample brine effectively with respect to the suite of biogeochemical parameters measured, including dissolved and particulate fractions. While the initial aim was to sample bulk ice, collecting brine profiles is also important for sea-ice biogeochemical studies and modifications could be made to the current melt-head design to target brine or bulk ice preferentially based on the findings from this study. Anomalies yet to be resolved include the missing fresh melt signal, low salinities without high δ¹⁸O in top ice, different dissolved-to-particulate ratios between metals, and low phosphate recoveries. Suggested modifications to address these anomalies include implementing an inverted cup-shaped head to catch fresh melt (improving bulk-ice sampling efficiency) and remove the risk of the tip of the hemisphere transferring material downward, a sample intake more evenly distributed across the melt head (i.e., more numerous smaller holes, like a showerhead), a wiper seal for isolating the melt head–ice interface from the borehole and improved priming and flow control for the pump.

Regarding flow control, a flow meter measuring sample flow rate and a rotary encoder measuring the ingress rate could be interfaced with the pump speed control to collect the expected volume of melt for bulk ice measurements. A paddle-wheel flow meter with only compatible materials in contact with the sample, for example, fluoropolymers for trace-metal work, would be suitable due to being robust for transport over/to the ice and not requiring specification of liquid properties that change with, for example, the concentration of salt and organic material. Alternatives to diesel-generator power should also be considered to make a melt probe system more portable. The melt probe described here (without considering peripherals like pumps) requires approximately 1 A at 200 W which is feasible for Lithium-ion batteries, but their transport is regulated as dangerous goods. Current portable methanol fuel cells which are a more environmentally friendly option are typically 12–24 V and would struggle to meet the power requirements.

Additional comparison with trace-metal corers is required to confirm the trace-metal cleanliness of the melt probe, but initial results are promising. Results from 0.01 m resolution field and lab experiments are also promising, with a little carryover but still improving on the current sample resolution from ice cores (96% recovery in 0.02 m). Future developments should also include speeding up ingress (melt-probe deployment is currently slower than coring), which Babin et al. (2019) have explored using ultrasonic vibration, and adapting the system to sample microplastics (it can be plastic-free from the sample inlet all the way to filtration) and gases (e.g., by integrating with a chamber; Nomura et al., 2010). Additional work could allow the probe to be deployed over the side of a ship to access floes that would otherwise require personnel on the ice. The effect of increasing the melt-head temperature on, for example, biology, gas evolution and particle dissolution should be carefully considered.

Overall, this proof-of-concept study presents an alternative method of sampling sea ice that, unlike traditional ice-coring methods, has the ability to support new research on fine-scale structures, with applications for sampling dissolved and particulate sea-ice constituents that include trace metals and macro-nutrients. With further development, important problems involving microplastics and gases could also be studied at higher spatial resolution.

Data can be accessed in the Supplemental materials.

The supplemental files for this article can be found as follows:

Supplemental Data (.xlsx)

This work was made possible by the Tsuneichi Fujii scholarship as well as tremendous support from Hokkaido University. Authors would like to thank Jun Nishioka, Naoya Kanna, Youhei Yamashita, Aiko Murayama, Deng Huailin, Manami Tozawa and Thomas Rodemann for helping with trace metal and carbon work; and Shigeru Aoki and Megumi Kitagawa for helping with stable oxygen isotope work. Authors would also like to thank those who provided insightful discussions which helped to guide the current and future versions of the melt probe, including Samantha Twiname, Alex Fraser, Rolf Gradinger, Mariko Honda, Anna Kelly and Pam Quayle.

The Tsuneichi Fujii scholarship and Hokkaido University sponsored this study. This project contributes to the Australian Antarctic Program Partnership (project ID ASCI000002) and an Australian Research Council’s Future Special Research Initiative grant (SR200100008), the SCOR Working Group 152–Measuring Essential Climate Variables in Sea Ice (ECV-Ice), and Biogeochemical Exchange Processes at Sea Ice Interfaces (BEPSII). This work was also funded by an Australian Research Council (ARC) Future Fellowship FT190100688, Japan Society for Promotion of Science (20H04345) and Arctic Challenge for Sustainability II (ArCS II) project. Access to SF-ICP-MS was supported by an ARC LIEF grant (LE0989539).

The authors have declared that no competing interests exist.

Contributed to conception and design: MC, KMM, PW, TC, DL.

Contributed to acquisition of data: MC, TT, DN, RA, NS, MY, ATT.

Contributed to analysis and interpretation of data: MC, TT, DN, KMM, PW, RA, NS, MY, ATT, TC, DL.

Drafted and/or revised the article: MC, TT, DN, KMM, PW, RA, NS, MY, ATT, TC, DL.

Approved the submitted version for publication: MC, TT, DN, KMM, PW, RA, NS, MY, ATT, TC, DL.

How to cite this article: Corkill, M, Toyota, T, Nomura, D, Meiners, KM, Wongpan, P, Akino, R, Samori, N, Yoshimura, M, Townsend, AT, Corkill, T, Lannuzel, D. 2025. A novel probe to sample dissolved and particulate matter in sea ice at high vertical resolution. Elementa: Science of the Anthropocene 13(1). DOI: https://doi.org/10.1525/elementa.2024.00053

Domain Editor-in-Chief: Jody W. Deming, University of Washington, Seattle, WA, USA

Guest Editor: Stefanie Arndt, Alfred Wegener Institute Helmholtz Centre for Polar and Marine Research, Bremerhaven, Germany

Knowledge Domain: Ocean Science

Part of an Elementa Special Feature: Understanding the Trajectory and Implication of a Changing Southern Ocean: The Need for an Integrated Observing System