Clearly layered mats of dense microalgae communities, hereafter referred to as melt pond algal mats, were collected from sea ice melt ponds at N 82° 24.9 and E 30° 26.2 during a cruise in the Barents Sea in July 2004 (Figures 1A, B) as part of a large Arctic ecosystem research project (CABANERA, Wassman et al., 2008). The mats were initially between 3 and 10 cm wide but broke up in smaller pieces (1-3 cm across) during sampling even at gentle handling. The mats were 3.5 to 4 mm thick with distinct and coherent layers stacked in visually recognized brown colored layers (Figure 1C). The sea ice melt ponds from where the mats were sampled were open at the bottom, i.e. there was a free flow of water between the melt pond and the ocean below. The melt pond was located near the center of the ice floe characterized as pack ice due to the occurrence of ridges and hummocks, and ice thickness varied from 1 to 1.5 m. During the available time on the ice we collected 10 to 15 algal mats from five melt ponds on the same ice flow, that were subsequently brought back to the onboard lab. The mats was gently sampled in cleaned polyethylene containers with in situ water from the melt pond, having a salinity of 34.0 and a temperature of ~0°C. The sampling location is further described in Wassmann et al. (2008).

Photosynthesis and respiration rates were measured after installing intact pieces of the algal mat in a microcosm that mimicked the in situ conditions within half an hour after sampling (Figure 1D). Three mats were sequentially installed and measured in the microcosm. The microcosmos was established inside a cold room (2.0°C) on board the research vessel and the microcosms temperature was secured at 0°C using an additional cooling device, in a setup equivalent to what was used by Hancke and Glud (2004). The algal mat was carefully pinned to a white styrofoam plate submerged in the microcosms that mimicked the light reflection properties of sea ice and kept the mat in a fixed position during measurements. Flushing gently with an air pump ensured both a stable diffusive boundary layer (DBL) of 300 to 500 μm over the mat and keeping the oxygen (O₂) concentration of the water at atmospheric saturation level. The mat surface was exposed to a photosynthetically active radiation (PAR) of 400 µmol photons m^–2 s^–1 during photosynthesis measurements by a halogen lamp with an optical fiber (Schott KL 1500). The irradiance was chosen to represent average light conditions in the melt ponds based on measured ambient light, which diel variations between 200 and 1200 µmol photons m^–2 s^–1 and an average of ~400 µmol photons m^–2 s^–1 at sampling time in July. Incident PAR was measured with a LiCor cosine-corrected quantum sensor and logger (LiCor LI-190SA, LiCor LI-1000, LiCor, US.).

Oxygen micro profiles were measured using electrochemical O₂ microsensors with a guard cathode (Revsbech, 1989) provided by RN Glud’s laboratory (University of Southern Denmark, DK). Gross and net photosynthesis, and dark community respiration were estimated from profiles measured using electrodes with tip diameters <15 μm, stirring sensitivity <1% and a 90% response time <0.5 s. Electrodes were calibrated by a 2-point calibration performed in both anoxic and air-saturated samples at ambient temperature. See Glud et al. (2000) and Hancke and Glud (2004) for additional details of the microsensor setup.

Gross photosynthesis was measured by the light-dark shift method (Revsbech and Jorgensen, 1983; Glud et al., 1992), by estimating the gross O₂ production from the initial concentration decline after a sudden eclipse of light (triplicates, with 3 min intervals). Total gross photosynthesis of the mat was calculated from integrating stepwise measurements down through the mat. Gross O₂ production was eventually converted to gross C fixation by multiplying with the molecular weight for C (12), the photosynthetic coefficient (1.4), and assuming 24h of daylight.

Rates of community net photosynthesis and dark respiration were calculated from the derived slope coefficient of the oxygen concentration profiles across the diffusive boundary layer and the top layer of the algal mat during light and dark, respectively (Jørgensen and Revsbech, 1985, Figure 2). Net photosynthesis equaled the flux of O₂ out of the illuminated mat, while respiration equaled influx during darkness. The community net photosynthesis and respiration rates were calculated from the sum of the upward and downward flux rates, derived from the linear concentration gradient during steady-state conditions, using Fick’s first law of diffusion (Jørgensen and Revsbech, 1985). Rates were corrected for the molecular diffusion coefficient according to Broecker and Peng (1974) and for temperature and salinity (Li and Gregory, 1974). Oxygen profiles varied little laterally across several investigated mats pieces and all derived rates were obtained from triplicated profiling. Flux rate calculations and the underlying principles are further described in Jørgensen and Revsbech (1985); Kühl et al. (1996); Glud et al. (2002) and Hancke and Glud (2004).

Gross ¹⁴C fixation was measured using the benchmark ¹⁴C method (Stemann-Nielsen, 1952). Subsamples of the three mats were incubated for four hours in pre-filtered (GF/F microfiber filters, Whatman, US) in situ water in light (400 μmol photons m^–2 s^–1) and in dark polycarbonate bottles positioned in the microcosms, with a final concentration of 0.0125 μCi mL^–1. Samples were subsequently filtered and immediately frozen onboard. Within 2 months filters were thawed, fumed with HCl acid for 8 hours, and 10 mL of Ultima GoldTM XP (Packard) were added before counting on a liquid scintillation analyzer with quench correction (PerkinElmer Tri-Carb 2900TR). The dark bottle values were subtracted from the light bottle values. The method is described in detail in Hodal and Kristiansen (2008), and the gross carbon fixation rate was corrected for total inorganic carbon based on Kivimae (2007).

Subsamples were taken by cutting cross sections from top to bottom of the mat and thus collecting pieces with known surface area and ensuring that the complete mat-structure was represented, and subsequently filtered on GF/F filters. Chlorophyll a (Chl a) was calculated from three subsamples extracted in MeOH (6h, 5°C, darkness) and measured absorbance at 665 nm, subtracting the absorbance at 750 nm, and using an extinction coefficient of 74.5 L g^–1 cm^–1 (Mackinney, 1941). The procedure is further described in Hancke et al. (2008). Likewise was particulate organic carbon (POC) analyzed following standard procedures after treatment of the samples with hydrochloric acid to remove the inorganic fraction using a Carlo Erba Elemental Analyzer (Model Na; Carlo Erba, Italy). Particulate organic nitrogen (PON) was analyzed according to Grasshoff et al. (1999) using a Scalar autoanalyzer (Scan Plus System, Netherlands). Mat wet and dry weights (dried in oven for >12h at 80°C) were determined to provide uniform measures and convert from areal to carbon units for gross photosynthesis measures.

Pigment samples were stored at -80°C until analyzed using a Hewlett-Packard HPLC 1100 Series system, equipped with a quaternary pump system and diode array detector. Mat subsamples were extracted in MeOH overnight at –20°C, and pigments were separated in Waters Symmetry C8 column (150 × 4.6 mm, 3.5 μm particle size) according to Zapata et al. (2000) and modified by Rodriguez et al. (2006). The extract was re-filtered (Millipore 0.2 μm) to remove debris. Chlorophylls and carotenoids were quantified according to their absorbance (350-750 nm). Identification of pigments and specific extinction coefficients for quantification followed Jeffrey et al. (1999). Light microscope analysis was used to identify dominant algal groups and species along with characterizing the community of grazers in the mat, before and after addition of Lugol (Tomas, 1997).

Steady state O₂ concentration profiles through the mat showed distinct gradients from the water above to the center of the algal mat, which was almost mirrored from the mat center and downward to below the mat. In the light (Figure 2A), O₂ concentration increased with depth until the center of the mat which indicated a photosynthetically active algal community with a high O₂ production, that by far exceeded the respiratory O₂ demand. At 400 µmol photons m^–2 s^–1, steady state O₂ profiles showed concentrations of >800 nmol O₂ cm^-3 (equivalent to μmol O₂ L^–1) in the mat center, a more than doubled O₂ concentration relative to the ambient level. In the dark (Figure 2B), the O₂ concentration decreased from atmospheric saturation in the above water (363 nmol O₂ cm^-3) to <40 nmol O₂ cm^-3 at the center of the mat, about 1.5 to 2 mm into the mat. The steep decrease in the O₂ concentration reflected a pronounced O₂ consumption in the mat during darkness.

The mat net community production of O₂ in light was 0.077 ± 0.008 nmol O₂ cm^–2 s^–1 (n=3), equivalent to 798 ± 83.0 mg C m^–2 d^–1 (assuming a molar ratio for O₂:C of 1:1 and a 24h day period equivalent to midnight sun, Figure 3). Algal mat community O₂ consumption in the dark was 0.067 ± 0.0054 nmol O₂ cm^–2 s^–1 (n=3), equivalent to 695 ± 55.6 mg C m^–2 d^–1. The photosynthetic gross O₂ production was 0.101 ± 0.016 nmol O₂ cm^–2 s^–1 (n=3), equal to 1467 ± 246 mg C m^–2 d^–1 (24h daylight, Figure 3). This measure was derived from the light-dark shift method and integrated over the center 1 mm thick photosynthetic zone in the mat (Figure 2A). The measure is independent of the net photosynthetic rate (Figure 3). Gross photosynthesis was in addition measured as the rate of gross ¹⁴C fixation in intact subsamples of the mat of a known area (~1.5 cm²) and weight. Gross ¹⁴C fixation equaled 2627 ± 305 mg C m^–2 d^–1 (n= 3). The ¹⁴C method showed gross carbon fixation rates more than twice as high as the measured gross O₂ production, which demonstrates a highly productive photosynthetic community. The method is, however, sensitive to the precision of the determined area, in contrast to the microsensor results.

The Chl a concentration in the mat was 46.9 ± 6.7 mg Chl a m^–2, the POC content 3412 ± 686 mg C m^–2, and the PON content 406 ± 92.8 mg N m^–2. This corresponded to ratios of C:Chl a = 75.1 ± 25.7, and of C:N = 9.8 ± 0.5 (Table 1). HPLC analyses showed a fucoxanthin content of 60% and a diadinoxanthin content (typical sunscreen pigment) of 30% of the Chl a concentration. The ratio of total photoprotective versus photosynthetic pigments were 0.18 ± 0.004 (n=3).

Light microscope analyses showed a dominance of the following algal genera Nitzschia spp., Navicula spp., Amphiprora spp., Entomoneis spp., and Pleurosigma spp. Also, a high representation of grazers including ciliate species were observed in the mats (Figure 1E), corresponding to previous observations of melt pond algal mats (Sørensen et al., 2017). Species composition of the grazer community was not further investigated.

The sampled ice algal mats were evidently different in physical structure and form than what has been reported previously of ice algal communities and accumulations beneath sea ice, including Melosira aggregates (Boetius et al., 2013) and floating aggregates (Fernández-Méndez et al., 2014). In their physical structure, they also differed from free-floating spherical algal accumulations observed in leads and below sea ice (Assmy et al., 2013). The species composition were, on the other hand, similar to previously observed floating algae aggregates and typical ice algae communities (Hegseth, 1992; Assmy et al., 2013).This indicate, that the melt pond algal mats may have a common point of origin with the spherical algal aggregates reported by Assmy et al. (2013). The observed melt pond algal mats were elongated, 3-5 mm thick with alternating light to dark brown layers (Figure 1C), with clear vertical layers in resemblance to other microbial and benthic mats (Franks and Stolz, 2009; Glud et al., 2009), which occur in a variety of environments as tidal flats (Barranguet et al., 1998), sublittoral soft-bottom sediments in the photic zone (Glud et al., 2002; Hancke and Glud, 2004; Woelfel et al., 2010), and in polar lakes and rivers (Quesada et al., 2008).

We speculate that the reported melt pond mats have formed at the bottom of melt ponds as a result of inflow of water and algae into open melt ponds (Fernández-Méndez et al., 2014) where the change from low light below the sea ice (<20 µmol photons m^–2 s^–1, Lund-Hansen et al., 2015) to high light in the pond (200 to 1200 µmol photons m^–2 s^–1) sparked the development of these dense microbial mat communities. This pair with the observations of a microalgal community that appear high-light acclimated, from the presence of the photoprotective pigment diadinoxanthin and the ratio of photoprotective versus photosynthetic pigments (Joy-Warren et al., 2019; Lund-Hansen et al., 2020). While Lee et al. (2011) reports about a new ice algal habitat formed by holes in the progressing thinning Arctic sea ice, they describe the formation and aggregation of ice algae to form long strands extending into the water below. This contrasts with our observations of distinct algal mats that have physical characteristics as benthic microphyte mats. Moreover, Lee et al. reports a dominance of Melosira arctica (>95%) with Nitzschia and Navicula species contributing the remaining. To our knowledge there is no previous reports on ice algal mats as we describe them here.

We do not know, over how much time the melt pond algal mats developed. The melt ponds were open upon sampling (free connection to the under-ice water), which we speculate is a precursor for the mat formation, and enables a continuous supply of nutrient rich water from below, as closed melt ponds typically hold low nutrient concentrations (Lee et al., 2011; Sørensen et al., 2017). The open melt ponds are regarded as the last development stage of the sea ice before freezing over in early autumn (Lee et al., 2011). This leaves a proposed window for the development of these melt pond algal mats of about one month around July-August before freezing and snowfall in late August/early September.

The Chl a concentration in the melt pond algal mats (46.9 ± 6.7 mg m^–2) was higher than most reports of ice algal communities in the Arctic when scaled up to a common areal unit, including aggregated ice algae (Assmy et al., 2013; Fernández-Méndez et al., 2014; Glud et al., 2014), and under-ice algal communities (Gosselin et al., 1997, Table 1). Even when compared to microphytobenthic mats in arctic and temperate environments with dense and photosynthetic active algal communities (Barranguet et al., 1998; Glud et al., 2002; Hancke and Glud, 2004; Woelfel et al., 2010). Nevertheless, higher concentrations of Chl a have been reported for ice algae in both Arctic and Antarctica than we observed here (Vincent et al., 1993; Arrigo, 2017). See Leu et al. (2015) for a pan-Arctic review of ice algal abundances.

In general, melt ponds of the Arctic Ocean are considered as low productive however empirical evidence is sparse and scattered (Lee et al., 2015; Sørensen et al., 2017). The rapidly warming Arctic favors first year ice over multiyear ice, which again is leading to an increase in melt pond formation and coverage (Polashenski et al., 2012). Warming has also been suggested to decrease ice algal primary production throughout the Arctic due to a shorter growth season (Leu et al., 2015), at the same time the contribution to primary production and carbon turnover by melt pond algal might increase due to the increase in areal coverage.

We here report on highly productive melt pond algal mat with a gross carbon production (2627 mg C m^–2 d^–1) that largely exceeded the dark consumption (695 mg C m^–2 d^–1), and thus resulted in a high net community production. The net community production was approximate two times higher when calculated from gross ¹⁴C fixation (minus the dark C consumption, as conventionally done) than when estimated independently from the O₂ steady state profiles (Figure 3). This was however anticipated, as ¹⁴C-samples were incubated in suspension which decreased the degree of self-shading between algae and thus can lead to overestimation of the gross carbon fixation (Stemann-Nielsen, 1952). Contrasting, the O₂ electrode method likely underestimates true gross O₂ production as it only includes photosynthetic production from the core of the mat (green bars in Figure 2) excluding peripherical algae photosynthesis (Glud et al., 1999). Nevertheless, the ice algal mat community demonstrated the highest net production rate of Arctic ice algal communities reported (Table 1). In fact, the present melt pond algal mat rates were approximately 3 times higher compared to measured rates from benthic algal microbial mats on temperate tidal flats.

Largely, ice algae are reported to sustain a low primary production in the Arctic, but large variations have been observed between regions, ice types, and habitats (Hegseth, 1992; Gosselin et al., 1997; Assmy et al., 2013; Glud et al., 2014; Arrigo, 2017; Lund-Hansen et al., 2018; Campbell et al., 2022). Low productivity of ice algal and aggregate communities is often associated with snow and ice cover and consequently often is light-limited (Woelfel et al., 2010; Leu et al., 2015; Hancke et al., 2018; Lund-Hansen et al., 2020b; Lund-Hansen et al., 2020a). In contrast, temperate benthic algal mats and the here described melt pond mats are both exposed to high light. In this case, it is reasonable to assume, that high primary production rates were supported by high light conditions prevailing at the bottom of melt ponds in combination with access to nutrient rich water from below the ice. This compares well with high rates of biological activity reported for floating ice-algal aggregates when scaled to the individual aggregate size published by Assmy et al. (2013), leading the authors to suggest that algae aggregates may provide a concentrated food source for ice-associated fauna during the oligotrophic Arctic summer months.

The fraction of melt ponds covering the sea ice in the Arctic Ocean in July is reported to be as high as 50 to 60% (Lee et al., 2020) but carbon fixation often low, estimated to <1% of the carbon production (Lee et al., 2012), with rates around 1.0 mg C m^–2 d^–1 (Sørensen et al., 2017). The net rate of 798 mg C m^–2 d^–1 measured in the algal mat demonstrates that melt pond algal mats may represent a significant carbon source for grazing and benthic organisms, as also suggested by Boetius et al. (2013), but how prevalent these mats are is still unanswered. Considering the research focus on ice algae, melt ponds and aggregates during the last two decades it is remarkable that mats similar to what we report here, have not been published even after we made our observations in 2004, which might imply that the mats have a low prevalence. Thus, their overall contribution to the Arctic Ocean carbon turnover is uncertain. However, our results suggest that melt pond algal mats might contribute to the Arctic Ocean carbon and energy flow at least on local and possible on regional scales. In addition, one might speculate that summer ‘blooms’ of algal mats in melt ponds could add an unpreceded supply of carbon to the Arctic food web late in the season. Additional measures of primary productivity and carbon turnover in melt pond algal mats alongside with quantification of their abundance and distribution are needed. Not least to understand the impact of increased warming on Arctic carbon pathways. Possibly, modern remote sensing techniques such as flying drones with high resolution imaging sensors (typical >1000 times better resolution than satellites) might provide tools to disclose the abundance and distribution of ice algal mats in melt ponds in the future.