Melt ponds are a common feature of the Arctic sea-ice environment during summer, and they play an important role in the exchange of heat and water vapor between the ocean and the atmosphere. We report the results of a time-series study of the CO2 dynamics within melt ponds (and nearby lead) and related fluxes with the atmosphere during the summer-to-autumn transition in the central Arctic Ocean during the Multidisciplinary drifting Observatory for the Study of Arctic Climate (MOSAiC) expedition. In late summer 2020, low-salinity meltwater was distributed throughout the melt ponds, and undersaturation of pCO2 in the meltwater drove a net influx of CO2 from the atmosphere. The meltwater layer subsequently thinned due to seawater influx, and a strong gradient in salinity and low-pCO2 water was observed at the interface between meltwater and seawater at the beginning of September. Mixing between meltwater and underlying seawater drives a significant drawdown of pCO2 as a result of the non-linearities in carbonate chemistry. By the middle of September, the strong stratification within the meltwater had dissipated. Subsequent freezing then began, and cooling and wind-induced drifting of ice floes caused mixing and an influx of seawater through the bottom of the melt pond. The pCO2 in the melt pond reached 300 µatm as a result of exchanging melt pond water with the underlying seawater. However, gas exchange was impeded by the formation of impermeable freshwater ice on the surface of the melt pond, and the net flux of CO2 was nearly zero into the pond, which was no longer a sink for atmospheric CO2. Overall, the melt ponds in this Arctic sea-ice area (both melt ponds and lead water) act as moderate sinks for atmospheric CO2.
1. Introduction
The Arctic is predicted to be affected more strongly by global warming than other regions, with temperatures rising about four times faster than the global average (Ono et al., 2022; Rantanen et al., 2022). Along with rising air temperatures, the extent of Arctic sea ice has been decreasing (Polyakov et al., 2012; Stroeve et al., 2012); the lowest extent of sea ice in recorded history was reported in 2012 (National Snow and Ice Data Center, 2024). There are many impacts of the reduction of the extent of sea ice. For example, ocean circulation changes due to alteration of the density structure caused by increased meltwater flow into the oceanic surface layer.
The Arctic Ocean and its adjacent seas north of 60°N show high solubility of CO2 due to the low water temperatures in the Arctic (Bates et al., 2006), absorbing 180 ± 130 Tg C yr−1 from the atmosphere, which is 5% to 14% of the total CO2 absorption by the global ocean (Bates and Mathis, 2009; Takahashi et al., 2009; Yasunaka et al., 2018). In the region where sea ice occurs, the increased solar radiation and air temperatures during the summer promote the melting of snow and sea ice, with the meltwater entering the ocean surface layer as freshwater (Smith et al., 2023). Meltwater from sea ice and snow can enter the ocean surface layer from the edges and bottoms of sea ice, or it can flow through cracks or leads in the sea ice (Eicken, 1994; Richter-Menge et al., 2001; Eicken et al., 2002; Nomura et al., 2018a; 2023). In addition, meltwater accumulates in depressions on the sea-ice surface, forming melt ponds. The seasonal evolutionary process of melt ponds can be divided into four stages (Eicken et al., 2002). In the first stage, snow on sea ice melts and accumulates on the surface of the sea ice. The sea ice at this time has low permeability, so the meltwater does not percolate through the sea ice, but spreads horizontally and drains into cracks, leads, or seal holes and reaches the ocean surface. In the second stage, most of the snow has melted, and the melt pond area (melt pond fraction) increases and approaches a maximum. However, the size of the melt ponds will start to decrease because of horizontal transport and an increase in the number of downward meltwater runoff pathways due to increased sea-ice permeability. Melt ponds also drain due to the increased hydrostatic pressure when too much water is accumulated above freeboard, the sea-ice floating line (Polashenski et al., 2012). In the third stage, further increases in the permeability of the sea ice result in more downward drainage. In addition, meltwater is less likely to be retained on the sea-ice surface as the sea ice breaks up. In many cases, however, the fraction of melt pond area to sea-ice area may reach a maximum at this stage because the elevation of the sea-ice surface becomes lower than the freeboard as the sea ice thins, and new melt ponds are then formed from surface flooding or underlying seawater intrusion (Polashenski et al., 2012). If melt ponds persist through the summer into the winter, then they refreeze and, after snow accumulates on their surfaces, are no longer visible. This is the fourth stage of melt ponds (Eicken et al., 2002; Petrich et al., 2012; Smith et al., 2023).
Melt ponds are a common and natural feature of the Arctic sea-ice environment during summer. The presence of melt ponds accelerates sea-ice melting by reducing the albedo of the area where sea ice is present, thereby warming the upper ocean (Curry et al., 1995; Perovich et al., 2002; Eicken et al., 2004). Melt ponds are categorized as “open ponds” or “closed ponds” depending on whether they are connected to seawater or not (Eicken et al., 2004; Lee et al., 2012). Melt ponds have been found to harbor both freshwater and saltwater phytoplankton and bacteria (Melnikov et al., 2002; Gradinger et al., 2005; Nomura et al., 2011; Fernández-Méndez et al., 2014; 2015; Sørensen et al., 2017). They also temporarily trap atmospheric trace elements that have accumulated within the snow, and release them into the surface ocean after photochemical reactions (Marsay et al., 2018). A melt pond is thus a transitional environment with various physical, chemical, and biological impacts in the region of Arctic sea ice.
Studies of melt ponds have used satellites to determine the ratio of melt pond area to sea-ice surface area (Rösel and Kaleschke, 2011; Webster et al., 2022) and modeling to investigate melt pond geometry and physical processes (e.g., Taylor and Feltham, 2004; Polashenski et al., 2012). Those studies have revealed that melt ponds cover about 20−50% of the area of summer Arctic sea ice (Polashenski et al., 2012; Feng et al., 2022; Webster et al., 2022). Despite such widespread coverage, relatively few studies have been conducted to constrain the CO2 budget in melt ponds (Semiletov et al., 2004; Bates et al., 2014; Geilfus et al., 2015), and all were conducted in areas of first-year sea ice in coastal areas of Alaska and the Canadian Arctic. Semiletov et al. (2004) and Geilfus et al. (2015) found that melt ponds absorb CO2 from the atmosphere (−5.4 to −0.0 mmol m−2 day−1; negative values indicate CO2 absorption from the atmosphere) by having lower pCO2 in melt ponds on Arctic sea ice than in the atmosphere.
A long-term drifting observational study broken up into 5 consecutive legs—Multidisciplinary drifting Observatory for the Study of Arctic Climate (MOSAiC)—was conducted in the central Arctic Ocean from October 2019 to September 2020 (Nicolaus et al., 2022; Shupe et al., 2022). We collected various data over a year in the region of central Arctic sea ice, where year-round observations are difficult to make. During that study, our observations took place during Leg 5 (August to October 2020) when melt ponds had matured (late summer) to the time when the surface of the melt ponds froze in early autumn. There have been no reports of melt pond carbonate chemistry in multi-year ice areas of the central Arctic Ocean. Nor have any studies examined in detail the vertical profiles of biogeochemical components inside melt ponds, the relationship between the structure of the melt pond interior and CO2 dynamics, and the impact on the flux of CO2 between the atmosphere and sea ice. The objectives of this study were to (1) describe the relationship between the physical environment and CO2 dynamics inside the melt ponds, (2) evaluate the effect of mixing of meltwater and seawater inside the melt ponds on carbonate chemistry, and (3) describe the process of CO2 exchange between the melt ponds and the atmosphere. The resulting data will enable future predictions of the effects of further sea-ice melting and meltwater inflow on the Arctic Ocean carbonate chemistry and fluxes between the atmosphere and the Arctic Ocean.
2. Materials and methods
2.1. Observation date and site
Surveys of melt ponds on sea ice and nearby leads were carried out during MOSAiC Leg 5 (August 22 to September 24, 2020) in the multi-year pack ice region of the central Arctic Ocean on board the R/V Polarstern (Figure 1). During Leg 4, the vessel drifted to the ice edge in Fram Strait and was thereafter relocated to the central Arctic Ocean to study the onset and early freezing phase of the sea ice.
MOSAiC Leg 5 study area in the central Arctic Ocean. (a) Location of MOSAiC Leg 5 in the central Arctic Ocean. (b) Map of the sampling locations on the MOSAiC Leg 5 ice floe. This image was obtained on August 25, 2020, by a drone (operated by S Graupner). (c) Enlargement of the red box in (b) to show sampling stations (St.) in melt ponds and the lead.
MOSAiC Leg 5 study area in the central Arctic Ocean. (a) Location of MOSAiC Leg 5 in the central Arctic Ocean. (b) Map of the sampling locations on the MOSAiC Leg 5 ice floe. This image was obtained on August 25, 2020, by a drone (operated by S Graupner). (c) Enlargement of the red box in (b) to show sampling stations (St.) in melt ponds and the lead.
On the initial arrival at the Leg 5 floe, multi-year ice with a thickness of 1.6 m based on measurements at the ice-coring site (Figure 1), a survey of 12 melt ponds was undertaken, selected to cover a broad range of sizes and shape (Table 1; Figure 1). After the breakup of the floe at the beginning of the survey (August 24), several of the melt ponds were inaccessible, and two of the melt ponds had been destroyed by a crack that opened directly through them (Stations 3 and 8, which were observed as lead sites by Nomura et al., 2023). The decision was made to focus additional melt pond analyses specifically on two melt ponds: Stations 1 and 4. Station 1 was a closed pond, and Station 4 was an open pond (Lee et al., 2012). The thickness of the ice below the bottom of the melt pond was measured (0.7 m at Station 4 on September 14 and 1.3 m at Station 1 on September 18) by coring over the refrozen melt pond surface ice after vertical water sampling. In the initial survey, we also included a lead site, which we refer to as the ROV lead site, and a melt pond near the ice-coring site (Table 1; Nicolaus et al., 2022). All melt pond and lead stations are shown in Figure 1. An overview of activities for melt pond and lead observations are summarized in Table S1.
Melt pond station, type (open or closed), size, temperature, salinity, dissolved inorganic carbon (DIC), total alkalinity (TA), and pCO2 on August 22, 2020
Station (Type)a . | Pond, Size (m2)b . | Sampling Depth (m) . | Temperature (°C) . | Salinity . | DIC (μmol kg−1) . | TA (μmol kg−1) . | pCO2 (μatm) . |
---|---|---|---|---|---|---|---|
1 (closed) | 17 | 0.3c | 0.5 | 0.8 | 73 | 63 | 138 |
2 (open) | 5 | 0.3 | 0.1 | 1.1 | 102 | 79 | 296 |
3 (open) | (crack) | 0.3 | 0.1 | 1.8 | 143 | 125 | 246 |
4 (open) | 192 | 0.3d | 0.2 | 0.2 | 45 | 16 | 370 |
5 (open) | 233 | 0.3 | 0.1 | 0.2 | 44 | 15 | 364 |
6 (closed) | 14 | 0.3 | 0.1 | 0.3 | 50 | 20 | 386 |
7 (closed) | 23 | 0.3 | 0.0 | 0.3 | 56 | 28 | 363 |
8 (open) | (crack) | 0.3 | 0.0 | 0.6 | 66 | 42 | 309 |
9 (closed) | 48 | 0.3 | 0.2 | 2.3 | 179 | 159 | 277 |
10 (closed) | 8 | 0.3 | 0.1 | 0.3 | 60 | 36 | 308 |
11 (open) | 63 | 0.3 | 0.6 | 2.2 | 162 | 141 | 279 |
12 (open) | 444 | 0.3 | 0.1 | 1.4 | 124 | 97 | 362 |
Average | 0.3 | 0.7 | 1.0 | 94 | 71 | 314 | |
Standard deviation | ±0.5 | ±0.8 | ±48 | ±51 | ±74 |
Station (Type)a . | Pond, Size (m2)b . | Sampling Depth (m) . | Temperature (°C) . | Salinity . | DIC (μmol kg−1) . | TA (μmol kg−1) . | pCO2 (μatm) . |
---|---|---|---|---|---|---|---|
1 (closed) | 17 | 0.3c | 0.5 | 0.8 | 73 | 63 | 138 |
2 (open) | 5 | 0.3 | 0.1 | 1.1 | 102 | 79 | 296 |
3 (open) | (crack) | 0.3 | 0.1 | 1.8 | 143 | 125 | 246 |
4 (open) | 192 | 0.3d | 0.2 | 0.2 | 45 | 16 | 370 |
5 (open) | 233 | 0.3 | 0.1 | 0.2 | 44 | 15 | 364 |
6 (closed) | 14 | 0.3 | 0.1 | 0.3 | 50 | 20 | 386 |
7 (closed) | 23 | 0.3 | 0.0 | 0.3 | 56 | 28 | 363 |
8 (open) | (crack) | 0.3 | 0.0 | 0.6 | 66 | 42 | 309 |
9 (closed) | 48 | 0.3 | 0.2 | 2.3 | 179 | 159 | 277 |
10 (closed) | 8 | 0.3 | 0.1 | 0.3 | 60 | 36 | 308 |
11 (open) | 63 | 0.3 | 0.6 | 2.2 | 162 | 141 | 279 |
12 (open) | 444 | 0.3 | 0.1 | 1.4 | 124 | 97 | 362 |
Average | 0.3 | 0.7 | 1.0 | 94 | 71 | 314 | |
Standard deviation | ±0.5 | ±0.8 | ±48 | ±51 | ±74 |
aAn additional station was established at the ice-coring site (Figure 1) after development of an open pond of 356 m2.
bDepth of the melt pond at Station 1 was 0.3 m; at Station 4, the depth range was 0.5–1.2 m over the study period, August 23–September 14, 2020.
cMelt pond water depth was 0.3 (Table S1).
dMelt pond water depth was 0.5–1.2 over the sampling period (Table S1).
2.2. Hydrographic observations and water sampling
On August 22, the first day of observations, water was sampled from a depth of 0.3 m in 12 different melt ponds to determine the horizontal distribution of water temperature, salinity, DIC, and TA across the melt ponds. Vertical temperature and salinity profiles were then obtained at Station 4 and the ROV lead site with a RINKO profiler (model ASTD103, JFE Advantech, Japan; Nomura et al., 2022; 2023; 2024c). The uncertainty of the temperature and salinity is ±0.01°C and ±0.01, respectively. Salinity was treated as practical salinity in our study. When the surface was frozen, a 0.15 m × 0.15 m hole was cut with a hand saw 0.5 m away from the thick sea-ice (1.6 m) margin of the melt pond or lead by using a 2 m pole. The thickness of the melt pond surface ice was measured before the RINKO profiler was lowered into the hole. To avoid disrupting the configuration of the melt pond layers, the RINKO profiler was gently and slowly hand-lowered with a rope until the profiler reached the bottom of the pond. The downcast data were used because the structure of the meltwater layer was disturbed when the RINKO profiler was recovered during the upcast. The thickness of the meltwater layer in each profile was defined as the depth where the density gradient was steepest in the depth profile and exceeded 0.1 kg m–3 cm−1 (Nomura et al., 2023).
After the initial survey of melt ponds, a long-term thermistor probe (TR-5120, T&D Corp., Nagano, Japan) was installed at Station 1 for the duration of Leg 5 to monitor temperatures at water depths of 0.015 m, 0.05 m, 0.1 m, and 0.2 m. The temperature data were stored in a data logger (TR-52i, T&D Corp.; Nomura et al., 2024d). Station 1 was chosen for this installation due to the small size of the melt pond (Table 1), which allowed for the suspension of the thermistor in the center of the pond from a pole laid across the pond margins. No further sampling was undertaken at Station 1 to avoid disturbing the measurements from the probe.
Prior to collecting water samples from Station 4 and the ROV lead site to make further measurements, we checked the vertical structure and depth of the meltwater layers from the same hole used for the RINKO profiler by attaching a conductivity sensor (Cond 315i, WTW GmbH, Germany) to the lower end of a 2-m-long ruler and inserting it into the melt pond and lead waters until the salinity measured with the Cond 315i increased at the interface between meltwater and seawater (halocline) to decide the depth for water sampling. Using tubing attached alongside the sensor, water was pumped up with a peristaltic pump equipped with PTFE tubbing (L/S Pump Tubing, Masterflex, USA) at depths corresponding to meltwater (surface), halocline, and seawater (bottom), with continuous salinity measurements via the conductivity sensor. The collected seawater was subsampled for measurements of dissolved inorganic carbon (DIC), total alkalinity (TA), and the seawater oxygen isotope composition (δ18O; Nomura et al., 2024b). Immediately after subsampling for measurement of DIC and TA, a 6.0% (by wt.) mercuric chloride (HgCl2) solution (100 µL) was added to stop biological activity. All samples were stored at 4°C on the R/V Polarstern.
During the discrete water sampling at Station 4 and the ROV lead site, the CO2 concentration in the water column was measured directly on site at each depth (i.e., surface, halocline, and bottom) by passing the water from the peristaltic pump through an equilibrator Liqui-Cel® (G542, S/N: 132462, 3M Company, USA) connected to an infrared gas analyzer (LI-8100A, LI-COR Inc., USA; Nomura et al., 2024b). The analyzer was calibrated with standard gases containing 0.0, 299.3, and 501.3 ppm CO2 before the expedition. Root means square (RMS) noise at 370 ppm with 1-second signal averaging is <1 ppm (https://www.licor.com/env/products/soil-flux/LI-8100a). The sample was passed through the equilibrator for about 1 minute prior to analysis to allow CO2 to equilibrate with air. At Station 4 and the ROV lead site, vertical CO2 measurements were made every 0.05 m to obtain detailed profiles. The pCO2 (μatm) was calculated from the CO2 concentration (ppm) in the air equilibrated with water, the barometric pressure, and the saturated water vapor pressure (Weiss and Price, 1980).
2.3. CO2 flux between the atmosphere and the surface of ice and water by chambers
We measured the CO2 fluxes between the atmosphere and the surface of the melt pond or the lead, sea ice, and snow (Nomura et al., 2024a) at Station 4, the ROV lead site, and the ice-coring site. When the surface of the melt pond or lead was frozen, flux measurements were made over the frozen surface for 1−2 hours. Then, a 1 m × 1 m hole was cut with a hand saw, and chambers were installed over the water surface with buoyant material (Nomura et al., 2020; 2022).
We used LI-COR 8100–104 chambers connected to a LI-8100A soil CO2 flux system (LI-COR Inc., USA) to measure the fluxes. A chamber was connected via a closed loop to an infrared gas analyzer (LI-8100A, LI-COR Inc., USA) to measure CO2 concentrations with an air pump at a rate of 3 L minute−1 during 20-minute intervals. Power was supplied by a battery (8012−254, Optima Batteries Inc., USA).
We also used a Teflon-coated metal chamber (0.50 m in diameter and 0.30 m high with a serrated bottom edge; Nomura et al., 2010; 2012). Every 5−10 minutes during an experiment (20 minutes), about 500 mL of air was collected from the chamber using a 50 mL glass syringe with a 3-way valve and then transferred to a 3000-mL Tedlar bag (AAK 3L, GL Sciences Inc., Japan). To avoid any effect from pressure changes during the collection of air samples, a 3000 mL Tedlar bag was installed within the chamber as a pressure regulator. After collection, air samples were quickly transported in a dark container to a laboratory onboard the R/V Polarstern. The CO2 concentrations were measured with a CO2 analyzer (Picarro 2132-i) used for continuous measurements of atmospheric CO2/CH4 concentrations on board.
The CO2 fluxes (in mmol m−2 day−1; a negative value indicates CO2 being absorbed from the atmosphere) were calculated based on the changes in the CO2 concentrations within the headspace of the chambers. The detection limit of the system is about +0.1 mmol m−2 day−1 (Nomura et al., 2010; 2018b). An inter-comparison experiment between the metal chamber and the LI-COR 8100−104 chamber in the home laboratory indicated good agreement (Nomura et al., 2022). Data obtained with both methods were therefore considered comparable.
Atmospheric CO2 concentrations were measured in a Swiss container onboard the R/V Polarstern (Angot et al., 2022). Atmospheric pCO2 was calculated from the observed CO2 concentration, air temperature, and air pressure during MOSAiC Leg 5 (Schmithüsen et al., 2021).
2.4. Artificial mixing experiments in the melt pond and lead
We conducted water mixing experiments at Station 4 on September 2 and at the ROV lead site on September 5 to understand how the carbonate chemistry and CO2 fluxes in the melt pond and lead responded to changes in the marine environment caused by wind and movement of sea ice. Before the mixing experiments, we measured the salinity and CO2 concentrations at depths of 0.05 m in the melt pond and 0.1 m in the lead with an equilibrator, and concomitantly measured the sea–air fluxes using the metal chamber adapted to float on the water surface. The water in the melt pond or lead was then mixed vigorously for 30 minutes by two persons using canoe paddles. After mixing, further water column and surface flux measurements were taken.
2.5. Sample analysis
Oxygen isotope analyses were carried out with a mass spectrometer (DELTA-S Finnigan MAT, USA) at the ISOLAB Facility of the Alfred Wegener Institute in Potsdam, Germany, using the equilibration method (details in Meyer et al., 2000; 2022). The δ18O values in per mil (‰) were calculated using the 18O/16O ratio of Vienna standard mean ocean water (V-SMOW) as the standard. The standard deviation of δ18O, calculated from six subsamples of the reference water (δ18O = −0.07‰), was <0.05‰.
The DIC of seawater was determined by coulometry (Johnson et al., 1985; Johnson, 1992) using a home-made CO2 extraction system (Ono et al., 1998) and a coulometer (CM5012, UIC, Inc., Binghamton, NY, USA). The TA of seawater was determined by titration (Dickson et al., 2007) using a TA analyzer (ATT-05, Kimoto Electric Co., Ltd., Japan). Both DIC and TA measurements were calibrated with reference seawater samples (Batch AR, AU, and AV; KANSO Technos Co., Ltd., Osaka, Japan; https://www.kanso.co.jp/eng/) traceable to the Certified Reference Material distributed by Prof. A. G. Dickson (Scripps Institution of Oceanography, La Jolla, CA, USA). The standard deviations of the DIC and TA measurements, calculated from 10 subsamples taken from reference water with a DIC value of 2039.4 μmol kg−1 and a TA value of 2311.6 μmol kg−1, were ±1.4 μmol kg−1 and ±1.2 μmol kg−1, respectively. The pCO2 was calculated from the DIC and TA using CO2SYS ver. 02.05 (Orr et al., 2018). We used the carbonic acid dissociation constants (K1 and K2) of Goyet and Poisson (1989). The KHSO4 was determined as described in Dickson (1990). We estimated the error on the pCO2 to be less than 6.0% by calculating the effect of changing the DIC and TA values by their standard deviation of the DIC and TA measurements.
2.6. Fractions of snow and sea-ice meltwater and under-ice seawater
To identify the origin of the meltwater within a melt pond or lead, we estimated the fractions of snow meltwater, sea-ice meltwater, and seawater with the following equations (Østlund and Hut, 1984):
where F, S, and δ represent the fraction, salinity, and δ18O, respectively. The subscripts “snow”, “si”, “sw”, and “obs” refer to snow meltwater, sea-ice meltwater, seawater, and the observed (measured) values in water, respectively. The fractions obtained with Equations (1)−(3) are expressed as percentages in the following text. The end-member values for snow and sea ice were taken from Leg 4, as in Nomura et al. (2023), which might have contributed to the Leg 5 melt ponds (Lange et al., 2023). For seawater, values sampled on August 22 from a depth of 10 m were used. The end-member values for snow, sea ice, and seawater are shown in Table 2. According to Nomura et al. (2023), taking the standard deviations of salinity and δ18O measurements into account, the three fractions varied maximally by 1.9%.
Mean end-member values (±standard deviation) for snow meltwater, sea-ice meltwater, and seawater
Sample . | Salinity . | δ18O (‰) . | N values . |
---|---|---|---|
Snow meltwatera | 0 | −19.1 ± 9.8 | 67 |
Sea-ice meltwatera | 2.4 ± 0.3 | −0.39 ± 0.47 | 21 |
Seawaterb | 32.6 | −1.25 | 1 |
Sample . | Salinity . | δ18O (‰) . | N values . |
---|---|---|---|
Snow meltwatera | 0 | −19.1 ± 9.8 | 67 |
Sea-ice meltwatera | 2.4 ± 0.3 | −0.39 ± 0.47 | 21 |
Seawaterb | 32.6 | −1.25 | 1 |
aSamples were collected during Leg 4 (Lange et al., 2023).
bFrom 10 m depth on August 22.
3. Results
3.1. Vertical profiles of temperature, salinity, density, DIC, TA, and pCO2
At Station 4, the vertical profiles of temperature, salinity, and density were obtained from the surface of the melt ponds to a depth of about 1.2 m (Figure 2). In the top 0.6 m, the three parameters remained largely consistent around −1.5°C (temperature), 28 (salinity), and 1022 kg m−3 (density) between August 28 and September 8. In contrast, in the bottom 0.6 m depth, temperature successively decreased from a maximum of about 0.2°C at the end of August to −1.5°C in mid-September. At the same time, the salinity increased from 0.3 to 28, and the density from 1000 to 1022 kg m−3.
Temporal changes in melt pond water mass characteristics over the MOSAiC Leg 5 sampling period. Results from CTD casts at melt pond Station 4 obtained between August 23 and September 14, 2020. Panels (a), (b), and (c) depict depth profiles of water temperature (°C), salinity, and density (kg m−3), respectively, over 0.1 m depth intervals.
Temporal changes in melt pond water mass characteristics over the MOSAiC Leg 5 sampling period. Results from CTD casts at melt pond Station 4 obtained between August 23 and September 14, 2020. Panels (a), (b), and (c) depict depth profiles of water temperature (°C), salinity, and density (kg m−3), respectively, over 0.1 m depth intervals.
The vertical distributions of salinity, DIC, and TA from water samples at Station 4 were very similar on August 28, September 2, and September 14 (Figure 3). All values were lowest in the surface layer and increased with depth. Furthermore, salinity, DIC, and TA increased with time in the surface and halocline. In contrast, there was little temporal change in the bottom layer.
Temporal changes in carbonate chemistry. (a) Salinity, (b) DIC (µmol kg−1), and (c) TA (µmol kg−1) determined on melt pond water samples collected at Station 4. Panel (d) shows the vertical profile of pCO2 calculated from these results using CO2SYS. Red, green, and blue indicate the results for August 28, September 2, and September 14, 2020, respectively.
Temporal changes in carbonate chemistry. (a) Salinity, (b) DIC (µmol kg−1), and (c) TA (µmol kg−1) determined on melt pond water samples collected at Station 4. Panel (d) shows the vertical profile of pCO2 calculated from these results using CO2SYS. Red, green, and blue indicate the results for August 28, September 2, and September 14, 2020, respectively.
Surface pCO2 at Station 4 on August 28 and September 2 ranged from 364 µatm to 385 µatm and was undersaturated compared to the atmosphere (Figure 3d). Values in the bottom layer ranged from 286 µatm to 303 µatm. On August 28 and September 2, pCO2 was lower in the halocline, from 155 µatm to 166 µatm. However, on September 14, when the salinity of the melt pond water at both sampling depths was similar to that of seawater, the pCO2 values were close to 300 μatm.
At the ROV lead site, on September 4 and 5, the pCO2 was again undersaturated in the surface layer (328–364 µatm) compared to the atmosphere (Figure 4). Inside the lead, as in the melt pond, there was a pattern of meltwater layer–halocline–seawater layer, with the pCO2 clearly lowest in the halocline (79–154 µatm). From September 4 to September 5, the meltwater layer thinned, and the halocline moved closer to the lead surface. At the same time, low-pCO2 seawater moved to the surface. Below the halocline, pCO2 was vertically uniform, as was salinity.
Vertical profiles of salinity and pCO2. (a) Salinity measured at the ROV lead site using a RINKO profiler and (b) CO2 concentrations measured using an equilibrator, converted to pCO2 for the result on September 4. (c) and (b) are the results on September 5 for salinity and CO2 concentrations, respectively. Plots indicated by arrows in (b) and (d) show the pCO2 at the halocline.
Vertical profiles of salinity and pCO2. (a) Salinity measured at the ROV lead site using a RINKO profiler and (b) CO2 concentrations measured using an equilibrator, converted to pCO2 for the result on September 4. (c) and (b) are the results on September 5 for salinity and CO2 concentrations, respectively. Plots indicated by arrows in (b) and (d) show the pCO2 at the halocline.
3.2. Spatial distribution of the carbonate species in the melt pond
At the beginning of the observations on August 22, DIC and TA were analyzed at 12 melt ponds alongside water temperature and salinity (Table 1). The water temperature was 0.0 to 0.6°C and salinity was 0.2 to 2.3. The pCO2 calculated from the DIC and TA varied from 138 to 386 µatm, which was undersaturated compared to the atmosphere (400 ± 3 µatm).
3.3. Temporal variations of air and water temperatures, wind speed, and thickness of the meltwater layer
After an initial phase of air and water temperatures around 0°C from late August until September 4, the air temperatures dropped to below −10°C, and the water temperature at Station 1 decreased to about −1.5°C (Figure 5a and c). The melt water layer at Station 4 successively shrunk from 0.6 m until it completely vanished by September 13 (Figure 5d). Wind speeds varied from 1 m second−1 to 15 m second−1, averaging approximately 7 m second−1 over the observation period. Storm events with wind speeds exceeding 12 m second−1 were observed on September 7 and September 14 (Figure 5b).
Temporal changes in air temperature, wind speed, and melt pond stratification. Temporal variations of (a) 12-hour mean air temperature, (b) 3-hour mean wind speed, (c) daily mean temperature at water depths of 0.015 m, 0.05 m, 0.1 m, and 0.2 m at Station 1 measured by thermistor probe, and (d) meltwater layer thickness at Station 4 during the observation period. (a) and (b) were measured at the 25 m height mast of the R/V Polarstern (Schmithüsen et al., 2021). The dashed horizontal lines in (a) and (c) represent 0°C.
Temporal changes in air temperature, wind speed, and melt pond stratification. Temporal variations of (a) 12-hour mean air temperature, (b) 3-hour mean wind speed, (c) daily mean temperature at water depths of 0.015 m, 0.05 m, 0.1 m, and 0.2 m at Station 1 measured by thermistor probe, and (d) meltwater layer thickness at Station 4 during the observation period. (a) and (b) were measured at the 25 m height mast of the R/V Polarstern (Schmithüsen et al., 2021). The dashed horizontal lines in (a) and (c) represent 0°C.
3.4. Fractions of under-ice seawater, sea-ice meltwater, and snow in the melt pond
Based on discrete water sampling at Station 4 on September 2, salinity was low in the surface layer (0.8 at 0.05 m) and increased with depth, as observed in the RINKO salinity profiles (Figure 2b). At the bottom (0.95 m) of the melt pond the salinity (29.0) was of similar order as the salinity of the seawater measured at 10 m below the sea ice (32.6) on August 22 (Nomura et al., 2023; Table 2). The δ18O was higher in the halocline (−1.64‰ at 0.6 m) than in the surface and bottom layers (Table 3).
Values for salinity, δ18O and the fraction of snow (Fsnow), sea ice (Fsi), and seawater (Fsw) at Station 4 on September 2, 2020
Layer . | Salinity . | δ18O (‰) . | Fraction (%) . | ||
---|---|---|---|---|---|
Fsnow . | Fsi . | Fsw . | |||
Surface | 0.8 | −2.82 | 12.6 | 87.4 | 0.0 |
Halocline | 23.9 | −1.64 | 3.4 | 25.1 | 71.5 |
Bottom | 29.0 | −3.18 | 10.8 | 0.3 | 88.9 |
Layer . | Salinity . | δ18O (‰) . | Fraction (%) . | ||
---|---|---|---|---|---|
Fsnow . | Fsi . | Fsw . | |||
Surface | 0.8 | −2.82 | 12.6 | 87.4 | 0.0 |
Halocline | 23.9 | −1.64 | 3.4 | 25.1 | 71.5 |
Bottom | 29.0 | −3.18 | 10.8 | 0.3 | 88.9 |
The water sampled from the surface (0.05 m) consisted of 89% sea-ice meltwater and 11% snow meltwater (with no seawater present). As depth increased, the fraction of sea-ice meltwater decreased, and conversely, the fraction of seawater increased. In the halocline (0.06 m) and bottom (0.95 m) of the melt pond, seawater accounted for 70% and 89% of the water, respectively. The fraction of snow meltwater ranged from 4% to 11% from the surface to the bottom layer (Table 3).
3.5. CO2 flux between the atmosphere and the surface of ice and water by chambers
On both August 31 and September 7, the fluxes over the sea ice near the melt pond were slightly positive (0.4 ± 0.3 mmol m−2 day−1 on August 31, 0.6 ± 0.2 mmol m−2 day−1 on September 7). In contrast, the fluxes between the atmosphere and the melt pond surface were negative on both days (−0.7 ± 0.3 mmol m−2 day−1 on August 31, −1.0 ± 0.7 mmol m−2 day−1 on September 7). In other words, CO2 is released from the sea ice and absorbed by the melt pond surface. The CO2 flux from the frozen melt pond to the atmosphere was 0.1 ± 0.04 mmol m−2 day−1 (Table 4).
Mean CO2 flux (±standard deviation) for each surface type at the ice-coring site
Date (2020) . | Surface Condition . | CO2 Flux (mmol m−2 day−1) . | N Values . |
---|---|---|---|
Aug 31 | Sea ice (no snow) | 0.4 ± 0.3 | 3 |
Melt pond water | −0.7 ± 0.3 | 7 | |
Sep 7 | Sea ice (2 cm thick snow) | 0.6 ± 0.2 | 2 |
Melt pond water | −1.0 ± 0.7 | 3 | |
Melt pond surface ice (5.5 cm thick ice) | 0.1 ± 0.04 | 4 |
Date (2020) . | Surface Condition . | CO2 Flux (mmol m−2 day−1) . | N Values . |
---|---|---|---|
Aug 31 | Sea ice (no snow) | 0.4 ± 0.3 | 3 |
Melt pond water | −0.7 ± 0.3 | 7 | |
Sep 7 | Sea ice (2 cm thick snow) | 0.6 ± 0.2 | 2 |
Melt pond water | −1.0 ± 0.7 | 3 | |
Melt pond surface ice (5.5 cm thick ice) | 0.1 ± 0.04 | 4 |
3.6. Mixing experiments in melt pond and lead water
The mixing experiment in the melt pond showed an increase in salinity, a decrease in pCO2, and an increase in CO2 absorption (Table 5). The mixing experiment at the ROV lead site resulted in similar observations as for the melt pond.
Values for salinity, pCO2, and CO2 flux before and after the mixing experiments
Melt Pond or Lead . | Salinity . | pCO2 (μatm) . | CO2 Flux (mmol m−2 day−1) . |
---|---|---|---|
Station 4 beforea | 0.8 | 321 | −3.9 |
Station 4 aftera | 6.5 | 81 | −21.0 |
ROV lead site beforeb | 4.6 | 345 | −0.5 |
ROV lead site afterb | 20.0 | 278 | −4.9 |
Melt Pond or Lead . | Salinity . | pCO2 (μatm) . | CO2 Flux (mmol m−2 day−1) . |
---|---|---|---|
Station 4 beforea | 0.8 | 321 | −3.9 |
Station 4 aftera | 6.5 | 81 | −21.0 |
ROV lead site beforeb | 4.6 | 345 | −0.5 |
ROV lead site afterb | 20.0 | 278 | −4.9 |
aSeptember 2.
bSeptember 5.
4. Discussion
4.1. Temporal variations during melt pond evolution
Based on changes in salinity (Figure 2b) and meltwater layer thickness (Figure 5d), the melt pond at Station 4 (considered to be an open pond) was initially filled with meltwater (mainly sea-ice meltwater; see Section 3.4). Over time, seawater entered into the pond through the bottom. In the final stages, the meltwater layer thickness in the melt pond began to decrease (Figure 5d), and finally the pond was filled with seawater completely (Figures 2c and 3a). This evolution of melt pond at Station 4 suggests that it was in the transition period from the third stage to the fourth stage. In such melt ponds, the exchange of meltwater and seawater occurs through spot holes at the bottom of melt pond (i.e., cracks, seal holes) and brine channels (Polashenski et al., 2012; Marsay et al., 2018). The aerial image in Figure 1 reveals holes (dark areas) that were connected to the sea at the bottom of many melt ponds during the observation period, as indicated by the type open in Table 1. Lee et al. (2012) reported that surface water (at most within 0.1 m depth) in the open ponds had a higher salinity than closed ponds, with a value close to that of the underlying seawater, and did not mention the stage of the melt pond in their study. Judging from the observation period and the high salinity and unfrozen water in the open pond surface, it can be assumed that these melt ponds they observed were at the third stage. However, both melt ponds investigated in our study, that is, melt ponds at Stations 4 and 1, respectively considered to be open and closed ponds, had low salinities (<2.3; Table 1). As indicated in Figure 2b, the melt ponds were filled with meltwater or stratified in the early stages of observation showing no difference in surface salinity between open and closed ponds in contrast to the observations of Lee et al. (2012).
Compared to melt pond sampling (Stations 1 and 4), we found a similar progression in salinity changes at the ROV lead site and a nearby lead (Stations 3 and 8; Nomura et al., 2023) compared to the melt ponds. A low-salinity (approximately 5.0) water layer was initially present in the surface layer of the lead, with a sharp change in salinity at a depth of about 1 m, and the salinity was uniform at values of about 32 at depths below 1 m (Figure 4). The meltwater layer thickness at the ROV lead site varied from 0.44 m to 0.27 m as time progressed. Nomura et al. (2023) found that the mixing in the lead (Stations 3 and 8) could be explained by wind-induced ice floe drift. There was a strong correlation between wind speed and ice drift speed during MOSAiC Leg 5 (r2 = 0.60, p < 0.001; Nomura et al., 2023). The changes in the characteristics of the water in the leads and melt ponds were very similar, suggesting that the water in the melt ponds was mixed with under-ice water through holes at the bottom of the melt pond due to wind-induced ice floe drift. Melt ponds and leads are not independent structures; rather, they interact with each other through holes and cracks in the sea ice. In the melt pond, as in the lead, there was a meltwater layer at the surface, and after a sharp salinity change over time, it appears to be replaced by seawater (Figure 2b). Eicken et al. (2002) have reported through tracer experiments that melt ponds are connected to each other and that melt ponds interact through horizontal water exchange and other processes. Smith et al. (2023) have also reported that meltwater exchange can occur within the ice via a complex system of cracks, leads, and melt ponds. The implication is that the changes observed in melt ponds were not a local phenomenon but were likely a widespread occurrence in surrounding cracks and leads.
4.2. Effect of mixing melt pond water on carbonate chemistry
Within the melt ponds at Station 4 and the ROV lead site, the pCO2 was lowest in the halocline where meltwater and seawater mixed (Figures 3d, 4b, and 4d), and the salinity gradient was steepest in the depth profile (Figures 2b, 4a, and 4c). To investigate the cause of this phenomenon, we examined whether it could be explained by the mixing of meltwater and seawater. We first calculated the variation in pCO2 that would have resulted only from the mixing of meltwater and seawater in the halocline (pCO2_cal.), that is, excluding the effects of biological activity, gas exchange, and precipitation/dissolution of calcium carbonate. The data used in the calculations are from melt pond water on August 28 and September 2, when low-salinity meltwater was present in the surface layer of the melt pond and seawater was present in the bottom layer.
The values of DIC, TA, and salinity in the surface and bottom layers were used as end-members to examine how the relationships between DIC and salinity, TA and salinity, and pCO2 and salinity would have changed by mixing of the two end-members (Table 6). Figure 6a and b shows the DIC and TA at different salinities when surface and bottom waters were mixed on August 28 and September 2, respectively. The slope of the TA–salinity relationship is larger than that of the DIC–salinity relationship, but the intercept is smaller. As a result, the difference between DIC and TA changed sign at a salinity of 6–7; the DIC was higher than the TA in the lower salinity range (<6), and the TA was higher in the higher salinity range (>7). From these DIC and TA results, we calculated pCO2 (Figure 6c). For surface meltwater (lower salinity), the pCO2 was about 360 µatm and hence lower than the atmospheric pCO2 (400 ± 3 µatm). For the bottom seawater (higher salinity), the pCO2 was about 300 µatm. The pCO2 decreased to about 120 µatm when meltwater and seawater were mixed at the halocline, and it reached a minimum at a salinity of about 8 (Figure 6c).
End-member values for salinity, dissolved inorganic carbon (DIC), total alkalinity (TA), pCO2, and pH for the melt pond Station 4
Date (2020) . | Layer . | Salinity . | DIC (μmol kg−1) . | TA (μmol kg−1) . | pCO2 (μatm)a . | pHa . |
---|---|---|---|---|---|---|
Aug 28 | Surface | 0.2 | 47 | 17 | 385 | 6.1 |
Bottom | 30.9 | 1996 | 2114 | 288 | 8.1 | |
Sep 2 | Surface | 0.8 | 80 | 52 | 364 | 6.6 |
Bottom | 29.0 | 1948 | 2054 | 286 | 8.1 |
Date (2020) . | Layer . | Salinity . | DIC (μmol kg−1) . | TA (μmol kg−1) . | pCO2 (μatm)a . | pHa . |
---|---|---|---|---|---|---|
Aug 28 | Surface | 0.2 | 47 | 17 | 385 | 6.1 |
Bottom | 30.9 | 1996 | 2114 | 288 | 8.1 | |
Sep 2 | Surface | 0.8 | 80 | 52 | 364 | 6.6 |
Bottom | 29.0 | 1948 | 2054 | 286 | 8.1 |
aCalculated from CO2SYS ver. 02.05 (Orr et al., 2018).
Carbonate chemistry in relation to salinity. DIC (µmol kg−1) or TA (µmol kg−1) of surface and bottom water at Station 4 on August 28 (a) and September 2 of 2020 (b) against salinity, each connected by a straight line. (c) pCO2 values calculated from DIC and TA for both dates. Atmospheric pCO2 is the average of values converted from CO2 concentrations observed in the Swiss container onboard the R/V Polarstern during MOSAiC Leg 5 (August 22 to September 14, 2020).
Carbonate chemistry in relation to salinity. DIC (µmol kg−1) or TA (µmol kg−1) of surface and bottom water at Station 4 on August 28 (a) and September 2 of 2020 (b) against salinity, each connected by a straight line. (c) pCO2 values calculated from DIC and TA for both dates. Atmospheric pCO2 is the average of values converted from CO2 concentrations observed in the Swiss container onboard the R/V Polarstern during MOSAiC Leg 5 (August 22 to September 14, 2020).
Meire et al. (2015) have discussed the low pCO2 of surface waters in Godthåbsfjord, Southwest Greenland: salinity, DIC, and TA behaved conservatively during mixing, in contrast with pCO2 which exhibited a marked non-linear behavior. Indeed, in Godthåbsfjord, when two water masses (one being glacial meltwater with low salinity, DIC, and TA, and the other being high-salinity water in a fjord with high DIC and TA) in equilibrium with the atmosphere were mixed, pCO2 became undersaturated compared to the atmosphere and decreased to 200 μatm at a salinity of about 8. Horikawa et al. (2022) found in Bowdoin Fjord that the pCO2 of surface water decreased when glacial meltwater and seawater were mixed in the fjord and reached a minimum at a salinity of about 12. Although this mixing decreased the pCO2, the decrease in pCO2 was not continuous and reached a minimum around a brackish salinity of 6–12. We also observed a decrease in pCO2 at the halocline of the ROV lead site (Figure 4b and d). To examine these results in detail, we calculated the mixing ratios of melt pond and lead surface water (i.e., meltwater, with low salinity, DIC, and TA) and pond/lead bottom water (i.e., seawater, with high salinity, DIC, and TA) as end-members with their pCO2 at the time of sampling. The present melt pond study and the lead and fjord survey from Meire et al. (2015) show a similar asymmetric bell-shaped curve (non-linear behavior) that holds true across scales where meltwater mixes with seawater (Figure 7). Such a pattern has been reported repeatedly for estuaries (Mook and Koene, 1975; Salisbury et al., 2008; Cai et al., 2013; Abril et al., 2021). The pCO2 is at a minimum at a mixing ratio of around 0.3 (Figure 7). We discuss the causes of this pCO2 depression and the non-linearities in carbonate chemistry in detail below.
The role of meltwater mixing ratios for pCO2. The mixing ratio of meltwater in the surface layer and seawater in the bottom layer determined from the salinity (1 for 100% seawater), plotted on the x-axis, and the pCO2 at that time, plotted on the y-axis.
The role of meltwater mixing ratios for pCO2. The mixing ratio of meltwater in the surface layer and seawater in the bottom layer determined from the salinity (1 for 100% seawater), plotted on the x-axis, and the pCO2 at that time, plotted on the y-axis.
Figure 8a and b shows the pH (seawater scale) and the concentrations of carbonate species (CO2, , and ) as calculated by CO2SYS in the melt pond on August 28, with the mixing ratio calculated from the end-members. The concentrations of all carbonate species tended to decrease as the mixing ratio decreased from 1.0 (i.e., only seawater) to about 0.3 (i.e., less influence of seawater; Figure 8b). Once the mixing ratio became less than 0.3, that is, the influence of meltwater was high, the pH decreased more rapidly (from approximately 8 to 6) compared to higher ratios, leading to strong changes in the carbonate species. The relative abundances of carbonate species in seawater change strongly with pH (Figure 9). CO2 is dominant in the low-pH range, is dominant near neutral pH, and at higher pH, becomes dominant (Zeebe and Wolf-Gladrow, 2001). Because the pH of seawater (mixing ratio of around 1) is about 8 (Figure 8a), is the dominant carbonate species in seawater (Figure 8c). In contrast, the pH of meltwater at the surface of the melt pond (mixing ratio of around 0) was low (6.1; Figure 8a). The influence of the meltwater therefore increased near the surface, and the pH decreased rapidly. The result was a larger ratio of CO2 to in the seawater. As a result, as can be seen from Figures 6a, 6b, 8c, and 8d, the DIC in seawater decreased as the salinity (mixing ratio) decreased, but the abundance of CO2 increased as the mixing ratio decreased to less than about 0.3. The result is a pCO2 minimum at an intermediate mixing ratio, as shown in Figure 7. We applied this analysis to the results of Meire et al. (2015), using as end-members low-salinity water (salinity = 0, DIC = 81 μmol kg−1, TA = 50 μmol kg−1) and high-salinity water (salinity = 33.65, DIC = 2118 μmol kg−1, TA = 2220 μmol kg−1), with similar results as shown in Figure 7 and with the trend of pCO2 variation for different mixing ratios also similar to that in our study. Mook and Koene (1975) have reported non-linear changes of pH and carbonate species due to the mixing of freshwater and seawater in an estuary. According to their calculations, the mixing of seawater and freshwater causes a change of pH, the ratio of carbonate species changes with the change of pH, and the CO2 increases at low pH. The present calculations for the melt pond halocline are consistent with their results, which also holds true for the computations carried out by Cai et al. (2013) and Abril et al. (2021).
The role of meltwater mixing ratios for carbonate chemistry. (a) The carbonate species (, , and CO2) and pH calculated using CO2SYS based on the dissolved inorganic carbon (DIC) and total alkalinity (TA) measurements of the August 28 water sample versus mixing ratio, and (b) an enlarged view of the results for carbonate species. (c) A plot of the abundance ratio of each component when the sum of the three carbonate components in (a) is 100%, and (d) an enlarged view of the plot in (c) for carbonate species.
The role of meltwater mixing ratios for carbonate chemistry. (a) The carbonate species (, , and CO2) and pH calculated using CO2SYS based on the dissolved inorganic carbon (DIC) and total alkalinity (TA) measurements of the August 28 water sample versus mixing ratio, and (b) an enlarged view of the results for carbonate species. (c) A plot of the abundance ratio of each component when the sum of the three carbonate components in (a) is 100%, and (d) an enlarged view of the plot in (c) for carbonate species.
The role of pH for carbonate chemistry. The abundance ratio of carbonate species (, , and CO2) versus the pH calculated by CO2SYS.
The role of pH for carbonate chemistry. The abundance ratio of carbonate species (, , and CO2) versus the pH calculated by CO2SYS.
In summary, the change in the pCO2 profile within the melt pond was determined by physical mixing and non-linearities in carbonate chemistry. The change of pCO2 due to the mixing of meltwater and seawater in the melt pond decreased the pCO2 at high salinities by the dilution effect. The pCO2 reached a minimum where the proportion of meltwater was high and the pH changed significantly (Figure 8). Furthermore, the calculations indicate that at lower salinities (i.e., closer to the surface), the pCO2 approached the atmospheric pCO2.
4.3. CO2 exchange with the atmosphere in the melt ponds
The observed CO2 fluxes between the atmosphere and the surface of the melt pond (Station 4 and ice-coring site), ranged from −3.9 to −0.7 mmol m−2 day−1. By convention, negative fluxes correspond to fluxes from the atmosphere to the melt ponds. Semiletov et al. (2004) measured the CO2 flux between the atmosphere and the melt ponds near Utqiagvik (Barrow), Alaska, in June to range from −51 to −3.9 mmol m−2 day−1. Geilfus et al. (2015) also reported CO2 fluxes in first-year fast ice melt ponds in Resolute Strait during the spring and summer seasons. According to these authors, melt ponds in the early melting season are formed mainly by snow meltwater, resulting in very low pCO2 values in the melt ponds (36 µatm, Geilfus et al., 2015). As a result, the atmosphere–melt pond CO2 fluxes ranged from −5.4 to 0 mmol m−2 day−1.
According to Geilfus et al. (2015), melt ponds in the early stages of their evolution continuously absorb CO2 because their surface water becomes undersaturated compared to the atmospheric CO2 as a result of meltwater inflow. This condition is the first stage of the melt pond evolution described previously. In contrast, the melt ponds observed in this study were in the third to fourth stage. In these stages, the continuous inflow of meltwater ceases because air temperatures have begun to decrease (Eicken et al., 2004). The pCO2 in the melt pond surface layer therefore approaches equilibrium with the atmosphere (Figures 3d and 6c), and the uptake of CO2 decreases, resulting in values lower than those reported by Geilfus et al. (2015) (e.g., −0.7 ± 0.3 mmol m−2 day−1 on August 31; Table 4). The presence of CO2-near-equilibrated meltwater limited the CO2 exchange between atmosphere and surface of the melt pond (Smith et al., 2023). The near-equilibrated meltwater layer acts as a cap to the much lower pCO2 halocline, prohibiting CO2 exchange between the low pCO2 water and the atmosphere. As discussed in Section 4.2, when the melt pond was stratified, low-pCO2 water was present at the halocline (Figures 3d, 4b, and 4d), and this water could continue to absorb CO2 if it was mixed with the surface meltwater layer. This behavior was apparent in the mixing experiments carried out in our study. Mixing experiments in the melt pond surface layer (0.05 m) resulted in changes of pCO2 from 321 µatm to 81 µatm, and CO2 fluxes from −3.9 to −21.0 mmol m−2 day−1. These results indicate that significant mixing within stratified melt ponds and leads could reactivate the surface as a significant CO2 sink from the atmosphere. Natural mixing of melt pond water can occur due to wind mixing and movements of sea ice. Nomura et al. (2023) have documented two storms during the observation period (the first around September 7, and the second around September 13−14), with wind speeds exceeding 12 m second−1. The melt ponds that formed on this Arctic sea ice had the potential to absorb more CO2 from the atmosphere when they were stirred by the wind-induced cooling and the sea ice movement, bringing the low-pCO2 water to the surface layer as shown in the present mixing experiments.
Figure 10 shows the relationship between CO2 fluxes between the atmosphere and the melt pond/lead water surface and difference in pCO2 (ΔpCO2). The fluxes measured at the water surfaces of the melt ponds and the lead decreased with increasing ΔpCO2. In other words, the higher the ΔpCO2 of the melt pond/lead surface layers, the more CO2 was absorbed from the atmosphere. On the other hand, when the surface was frozen and CO2 fluxes were measured over the frozen surface, CO2 absorption from the atmosphere stopped, even if the surface layer of the melt pond was undersaturated compared to the atmosphere (Figure 10). Therefore, the CO2 flux measured over the refrozen melt pond ice surface was almost zero (0.1 ± 0.04 mmol m−2 day−1; Figure 10 and Table 4). The brine volume fraction, calculated from the temperature (−2.3°C) and bulk salinity (1.3) of the refrozen melt pond ice according to Cox and Weeks (1983), was 2.7%. A brine volume fraction of 5–7.5% or higher is typically taken to indicate that sea ice is permeable to brine (Golden et al., 1998; Pringle et al., 2009; Zhou et al., 2013). Therefore, the low brine volume fraction calculated for the refrozen melt pond ice surface indicates that no CO2 exchange could occur. Similar results (no gas exchange) were observed for the superimposed ice that formed over Antarctic sea ice and the sea ice in Saroma-ko Lagoon, Hokkaido, Japan, when surface snow was melted and the bottom of the snow refroze (Nomura et al., 2010; 2013; Delille et al., 2014). However, our result showing no CO2 exchange over the refreezing melt pond is the first observation of its kind.
Relationship between CO2 fluxes and ΔpCO2. The difference between pCO2 of surface water at Station 4, ROV lead site and ice-coring site and atmospheric pCO2 (ΔpCO2), versus CO2 fluxes measured by the chamber method between the atmosphere and melt pond/lead water (circle, square) or the atmosphere and ice (cross) at that time. Colors indicate the date (in 2020) of measurement for each sample (red: August 31, ice-coring site; green: September 2, Station 4; light blue: September 5, ROV lead site; blue: September 7, ice-coring site). Solid and dashed lines indicate the linear-fit curve for CO2 fluxes measured over the melt pond/lead water and ice interface, respectively.
Relationship between CO2 fluxes and ΔpCO2. The difference between pCO2 of surface water at Station 4, ROV lead site and ice-coring site and atmospheric pCO2 (ΔpCO2), versus CO2 fluxes measured by the chamber method between the atmosphere and melt pond/lead water (circle, square) or the atmosphere and ice (cross) at that time. Colors indicate the date (in 2020) of measurement for each sample (red: August 31, ice-coring site; green: September 2, Station 4; light blue: September 5, ROV lead site; blue: September 7, ice-coring site). Solid and dashed lines indicate the linear-fit curve for CO2 fluxes measured over the melt pond/lead water and ice interface, respectively.
According to the field observations of Webster et al. (2022), melt ponds covered approximately 32% of the MOSAiC Leg 5 sea ice floe. In addition to the melt ponds, cracks and leads were formed in the sea-ice region. The CO2 fluxes measured on the melt ponds and lead in this study indicated CO2 absorption from the atmosphere. Moreover, low pCO2 water was present in the halocline of the melt ponds and lead, and the results of mixing experiments indicated that when the water in the melt ponds and lead was mixed, pCO2 decreased sharply and CO2 absorption increased. Although our study was able to show that changes in pCO2 within a melt pond were determined by physical mixing, respiration, and photosynthesis by organisms living inside the melt pond may have altered the pCO2 to some extent as well. Diatoms and other organisms were observed floating inside the melt pond. Earlier in the season, a unique algal community (halocline flora) was identified at the interface between the meltwater and seawater layers in melt season leads, and the biomass of that algal community remained high for over 4 weeks (Smith et al., 2023). Concentrated algal aggregates were similarly observed in the meltwater layer of leads near the melt pond stations, but at variable states of productivity and decomposition (Smith et al., 2023). Semiletov et al. (2004) have suggested that photosynthesis is accelerated by an increase of photosynthetically active radiation inside a melt pond. However, we were unable to identify significant biogeochemical processes affecting the melt pond pCO2 profile in this summer–autumn time of the year. This observation suggests that there are important CO2 sinks such as melt ponds and leads in the Arctic regions of sea ice, and those sinks can play a role in the overall CO2 budget of the Arctic Ocean, which cannot be ignored when considering the global carbon budget (Steiner et al., 2013).
5. Conclusions
The melt ponds in our survey in the central Arctic Ocean have a lower pCO2 than the atmosphere. CTD observations of the melt ponds indicated that they were initially filled with meltwater, but became stratified over time as seawater intruded from below. The seawater eventually mixed with the meltwater, finally creating a uniform profile in the melt pond. The melt pond disappeared as the surface refroze.
The pCO2 of the water at the interface between the meltwater and underlying seawater is lower than the pCO2 of the meltwater (above) and seawater (below) due to the non-linearities in carbonate chemistry that is changed by the mixing of meltwater and seawater. This decrease in pCO2 has been documented in previous studies and was apparent from the direct flux measurements using the chamber method in this study. Experiments we conducted in the field showed that mixing of the meltwater and underlying seawater due to wind or sea-ice turbulence could drive a further drawdown of atmospheric pCO2. This effect has been observed in leads but also in estuarine systems where freshwater from rivers is mixed with seawater. It appears from the present work to be commonly observed at the surface of sea ice, both in melt ponds and in leads between ice floes.
According to the observations, but also theoretical carbonate system thermodynamics, we suggest that the whole Arctic sea-ice area (both melt pond and lead water) is experiencing this atmospheric pCO2 drawdown at the surface of melt ponds and leads. A minimum in pCO2 is observed below the meltwater layer, where we estimate that the mixing ratio between meltwater and seawater is 0.3. This low pCO2 mixing water is not in contact with the atmosphere. If this low-pCO2 water is brought to the surface as a result of enhanced mixing, like wind-driven mixing of ice motion, the surface layer pCO2 would decrease further. By the end of the survey, the pCO2 in the melt pond surface layer was about the same as that of the seawater under the sea ice and could potentially continue to absorb CO2 from the atmosphere, but the CO2 flux became nearly zero as the surface of the melt pond froze. Overall, melt ponds in the Arctic sea-ice area act as moderate sinks for atmospheric CO2.
Data accessibility statement
The data analyzed in this study were retrieved from links below: RINKO profiler-derived variables: https://doi.pangaea.de/10.1594/PANGAEA.966749, thermistor probe derived variables: https://doi.pangaea.de/10.1594/PANGAEA.966770, water sampler derived variables: https://doi.pangaea.de/10.1594/PANGAEA.966787, CO2 chamber-derived variables: https://doi.pangaea.de/10.1594/PANGAEA.966833, and meteorological variables: https://doi.org/10.1594/PANGAEA.935267.
Supplemental material
The supplemental files for this article can be found as follows:
Table S1. Overview of activities for melt ponds and leads during the observations. PDF
Acknowledgments
This work was conducted and data used for this report were produced as part of the international Multidisciplinary drifting Observatory for the Study of the Arctic Climate (MOSAiC) with the tag MOSAiC20192020. We thank all persons involved in the expedition of the R/V Polarstern during MOSAiC in 2019–2020 (AWI_PS122_00) as listed in Nixdorf et al. (2021; http://doi.org/10.5281/zenodo.5179739).
Funding
This study was supported by the Japan Society for the Promotion of Science (grant numbers: JP18H03745; JP18KK0292; JP17H04715; JP20H04345; 23KJ0017) and by a grant from the Joint Research Program of the Japan Arctic Research Network Center (MY, DN, MT, JI). ALW and KS were funded through the U.K. Natural Environment Research Council (NERC) (Grants Nos NE/S002596/1 and NE/S002502/1, respectively). ESD was supported by NERC through the EnvEast Doctoral Training Partnership (NE/L002582/1), as well as NERC and the Department for Business, Energy & Industrial Strategy (BEIS) through the U.K. Arctic Office. EJC was supported by the National Science Foundation (USA) NSF OPP 1821911 and NSF Graduate Research Fellowship. HM is supported through the German Federal Ministry of Education and Research (grant number 03FO869A). KMP, HA, and BB were supported by the U.S. National Science Foundation (awards OPP 1807496, 1914781, and 1807163). BD is a research associate of the FRS-FNRS and was supported by the research credit J.0051.20.
Competing interests
The authors declare that they have no conflict of interest.
Author contributions
Contributed to conception and design: MY, DN, ALW, MD, ESD, EJC, KMP, HA, BB, BD.
Contributed to acquisition of data: MY, DN, ALW, YL, MD, KS, ESD, EJC, KMP, HA, BB, HM.
Contributed to analysis and interpretation of data: MY, DN, ALW, ESD, MH, MT, BD.
Drafted and/or revised the article: MY drafted the first version of the manuscript, and all authors contributed to writing and revisions.
Approved the submitted version for publication: All authors.
References
How to cite this article: Yoshimura, M, Nomura, D, Webb, AL, Li, Y, Dall’osto, M, Schmidt, K, Droste, ES, Chamberlain, EJ, Posman, KM, Angot, H, Blomquist, B, Meyer, H, Hoppema, M, Tozawa, M, Inoue, J, Delille, B. 2025. Melt pond CO2 dynamics and fluxes with the atmosphere in the central Arctic Ocean during the summer-to-autumn transition. Elementa: Science of the Anthropocene 13(1). DOI: https://doi.org/10.1525/elementa.2024.00023
Domain Editor-in-Chief: Jody W. Deming, University of Washington, Seattle, WA, USA
Associate Editor: Kevin R. Arrigo, Department of Earth System Science, Stanford University, Stanford, CA, USA
Knowledge Domain: Ocean Science
Part of an Elementa Special Feature: The Multidisciplinary Drifting Observatory for the Study of Arctic Climate (MOSAiC)