Field environment

East Antarctica’s largest BIA is situated on the Amery Ice Shelf. Adjacent to the ice shelf lies a near-continuous blue-ice zone, which we accessed from Davis Base (Fig. 1a). Two areas of blue ice were investigated by helicopter during a reconnaissance trip: first along the flank of the westward-flowing Sørsdal Glacier, and second along the ice-sheet margin further north. The blue ice in both areas is continuous and maintained by strong katabatic airflows. However, neither debris nor subsurface melting was found at the latter site during February 2008. Instead, melting was associated with deposits of debris particles along the flank of Sørsdal Glacier in the Flanders Moraine area (Fig. 1b). Here it also likely that the advection of warm air from the Vestfold Hills increases melting, because runoff troughs and channels are a conspicuous feature (although no flowing water was observed in them during this study: Fig. 1b). The distribution, biogeochemistry and biological production of cryoconite hole habitats in the Flanders Moraine area of the Vestfold Hills were examined over a 2 week period during February 2008.

Cryoconite mapping

A combination of rapid transect survey and high-resolution mapping was employed on the ice surface in the Flanders Moraine area. For the former, the percentage of the ice surface covered by cryoconite was deduced at 30 m intervals along a 390 m transect perpendicular to the ice margin. At each point the maximum and minimum diameters of all cryoconite holes were measured in a 12.6 m² area (defined by circle of 2 m radius about a randomly selected point). At the down-glacier end of the transect, a digital elevation model (DEM) of the ice surface was then produced after intensively surveying a 20 m × 20 m area (hereafter the ‘Cryoplot’) using a Leica Total Station. All cryoconite holes were surveyed within this area in the same manner as the transect. In addition, the depths of all cryoconite within half of this area were deduced by drilling with a Kovacs ice auger. Time constraints did not allow us to drill to all the holes because field operations from Davis Base were being drawn to a close.

The light environment of the Cryoplot was established using two Skye Instruments pyranometers: one ∼0.5 m above the ice surface and one installed on the surface of a ∼84 cm deep cryoconite debris layer, initially submerged beneath 20 cm water, an air gap and then a surface ice lid ∼5 cm thick. Both instruments measured incident radiation (400–1200 nm) in W m⁻², which was also used to estimate photosynthetically active radiation (PAR: 400–700 nm) for these locations using an empirical relationship established with a LiCor LI-192SA PAR sensor. In addition, the PAR sensor was used to derive six light profiles at 5 cm intervals after drilling a 7 cm diameter hole down 150 cm into the ice with a Kovacs auger. Profiles were measured during peak daily irradiance at solar noon. In order to consider the effects of light reflection down the open hole, the profiles were measured with and without an opaque cap over the hole. Three holes were drilled on exposed ice ridge tops and three in shallow troughs where a thin snow layer had accumulated (e.g. Fig. 1b).

Air temperature, wind speed and wind direction data were supplied courtesy of the Australian Antarctic Division for two locations: Davis Base (18 m a.s.l.) and Whoop Whoop Airfield (550 m a.s.l.) on the ice-sheet plateau east of the base. An air temperature record was then constructed for the study site (150 m a.s.l.) by interpolation between these two locations.

Biogeochemistry

Sampling was conducted within the Cryoplot during its survey (12 cryoconite water samples and 10 glacier ice samples), and at every cryoconite hole before biological production measurements were taken (see below). The subsequent analyses were used to assess the dissolved oxygen (DO₂), total dissolved inorganic carbon (TDIC) and major-ion (Ca²⁺, Mg²⁺, Na⁺, K⁺, F⁻, Cl⁻, NO₃ ⁻ and SO₄ ²⁻) concentrations. The pH measurements were discarded owing to poor performance of the electrode in the cold. TDIC was analysed using the headspace method after acidification of 15 mL samples with 5% HCl immediately upon return from the field and as described in full by Reference HodsonHodson and others (2010b). Otherwise, all samples were filtered through a 0.2 μm membrane and frozen prior to analysis in the United Kingdom using Dionex DX90 ion chromatography for ions and a Thermalox TC/TN analyser for dissolved organic carbon (DOC). Precision errors were 10% or less for DOC (calibration range 1–10 mg L⁻¹), 5% or less for DO₂ and the ion determinations (detection limit 1 μg L⁻¹ and calibration range 0–2 mg L⁻¹) and 3% or less for TDIC. DOC assays were only conducted on the 12 Cryoplot samples.

In situ measurements of primary production and respiration

Batch incubations were undertaken at 12 sites following the precise methodology developed by Reference HodsonHodson and others (2010a). This is based on the long-established use of light and dark flask incubations for assessing net ecosystem production (NEP) and respiration (R) respectively. Primary production (PP) can then be estimated by difference (between batch averages) since:

The terms NEP and R were estimated using the TDIC change (ΔTDIC) of the light and dark incubations after correction for the production of TDIC from carbonate weathering. Therefore:

where [ΔCa] represents the production of TDIC by carbonate weathering (1 µM (12 µg) of TDIC per µM (41 µg) of Ca²⁺) during the incubations.

At each site, six large UV-transparent Whirl-Pak^® bags were carefully filled with ∼20g of cryoconite debris and 350 mL of cryoconite hole water. Three of them were foil-wrapped for dark incubation, and all of them were placed back in a cryoconite hole for 1–3 days incubation. Then 15 mL gas-tight Eppendorf tubes were filled with incubation solution and analysed immediately after return to the laboratory for TDIC using the headspace method. The DO was also measured three times in the field using a luminescence meter (Hach LDO) and further 15 mL samples taken for major-ion determination in the United Kingdom. Changes in all chemical conditions were deduced by subtraction of the initial incubation conditions and then normalized for the dry mass of cryoconite used after oven-drying the debris at 40°C. Combustion at 450°C for 4 hours was then used to assess the total organic carbon content through the loss on ignition (LOI) method. Additionally, the chlorophyll-a concentration of the debris was established spectrophotometrically after methanol extraction in the dark at −20°C (Reference Talling, Vollenweider, Talling and WestlakeTalling, 1969).

Laboratory determinations of primary and bacterial production

Incubations using the above methods were also repeated in an environment chamber at 1°C using cryoconite waters and samples collected from the field and incubation waters with lower DO₂ levels than those encountered in the field. The purpose of this was to remove the problem of high ambient DO₂ levels caused by photosynthesis and freeze-concentration (see below) and to examine rates of PP under controlled (artificial) light conditions. Large samples of cryoconite hole waters were allowed to equilibrate to the laboratory atmosphere at 18°C (causing the surplus DO₂ to degas) and were then sealed, chilled to 1°C and added to the incubations. This process had no significant effect upon the TDIC of the initial solution used to commence the incubations (which was between 0.7 and 1.1 mg C L⁻¹ and therefore in the same range as the in situ incubations). However, the DO₂ was ∼40% lower than in the field (average 10 mg L⁻¹) and far more stable during the initial handling of the solutions and debris. Four light incubations were therefore conducted at 100 µmol m⁻² s⁻¹ PAR and a further two at 75 µmol m⁻² s⁻¹ for a duration of 5 and 3 days respectively. Foil-wrapped dark incubations were also conducted at the same time.

Finally, rates of gross photosynthesis and light curve characteristics were also deduced using short-term laboratory incubations of six fresh ¹⁴C-HCO₃ ⁻ -labelled cryoconite water and debris samples at 21 irradiance levels between 0 and 485 µmol m⁻² s⁻¹ and following the exact methods reported by Reference Westwood, Griffiths, Meiners and WilliamsWestwood and others (2010). Briefly, 7 mL working solutions of NaH¹⁴CO₃ with an activity of 39.183 × 10³ Bq mL⁻¹ (1.1 µC mL⁻¹) were inoculated with 2 mL of sample in a transparent glass scintillation vial and incubated for 1 hour at 1°C. This was done immediately after returning from the field using samples that were less than 3 hours old. Light curve parameters were defined statistically after estimating the photosynthesis using the protocols described by Reference Platt, Gallegos and HarrisonPlatt and others (1980).

Bacterial production was determined for debris from 12 cryoconite holes using the micro-centrifuge method of Reference Kirchman and PaulKirchman (2001). These measurements were conducted immediately on return to the laboratory. [¹⁴C] leucine (specific activity 86 Ci mmol⁻¹) was added to triplicate 2.0 mL samples to a final concentration of 35 nM and incubated at 1°C for 90 min. One control sample was killed with 90 µL of 100% trichchloroacetic acid (TCA) and labelled leucine added as above. Incubations were terminated by the addition of 90 mL of 100% TCA. Tubes were centrifuged at 16000 × g for 10 min and the supernatant aspirated. Then 1.7 mL of ice-cold 5% TCA was added to each tube and the centrifugation procedure repeated. The TCA was aspirated and 1.7 mL of ice-cold ethanol added and a further centrifugation step performed. The final supernatant was aspirated and 1 mL of scintillation cocktail added prior to counting in a Beckman LS6500 scintillation counter. For the calculation of bacterial production we used an intracellular dilution factor of 2, leucine/protein ratio of 0.073 and a carbon/protein ratio of 0.86 (Reference Smith and AzamSmith and Azam, 1992).

Microbial enumeration

Accurate enumeration of eukarya and bacteria associated with cryoconite is notoriously difficult because of the problems associated with masking by debris. Therefore indicative numbers of these microorganisms were sought using a combination of flow cytometry and epifluorescence microscopy. Phytoplankton and cyanobacteria were therefore enumerated in four of the water samples and eight of the debris suspensions immediately after sampling for the bacterial production measurements. They were screened through a 50 µm mesh immediately after sampling and analysed using a FACScan flow cytometer (Becton Dickinson) with a 488 nm argon laser. Samples were chilled in an ice-filled beaker and analysed for 5 min at a flow rate of 60 µL min⁻¹. Nanophytoplankton and cyanobacteria were discriminated using their red and orange autofluorescence respectively, using bivariate scatter plots of red versus orange fluorescence (FL3 and FL2 respectively). Samples were weighed with ±0.0001 g accuracy before and after each run to determine the volume analysed. Concentrations of microbes were then calculated using event counts from bivariate scatter plots and the volume analysed. Ten microlitres of PeakFlow Green 2.5 µm beads (molecular probes) were added to all samples as an internal standard.

Enumeration of bacteria in eight samples (four debris and four water) was undertaken using 2 mL samples of debris suspension fixed in buffered glutaradehyde (final concentration 2%), stained with SYBR Gold, filtered on to 0.2 μm black polycarbonate filter and counted using an epifluorescence microscope. SYBR Gold stained the bacteria very strongly in the sediment suspensions, making them clearly visible. However, as experienced elsewhere (e.g. Reference Langford, Hodson, Banwart and BøggildLangford and others, 2010), masking by the debris means that the bacterial counts are for indicative purposes only.

Light and climate conditions during the study

Solar radiation received at the surface of the debris in the 84 cm deep instrumented ridge-top cryoconite hole (Fig. 2b) represented up to 14% (average 9.16%) of surface irradiance throughout the study. The pyranometer data suggest that average PAR receipt at the ice surface was 409 μmol m⁻² s⁻¹ (maximum 1860 μmol m⁻² s⁻¹), while in the cryoconite hole it was 37 µmol m⁻² s⁻¹ (maximum 302 µmol m⁻² s⁻¹).

PAR profiles into the ridges and valleys showed quasi-exponential extinction curves at all sites (Fig. 6). However, the proportion of surface PAR irradiance penetrating the ice was greatly reduced beneath the valleys. The average percentage of PAR incident at the modal depth values for cryoconite beneath ridges and valleys (i.e. at 65 cm and 35 cm respectively, according to Fig. 5b) was 11.5% and 9.07% respectively. It should be noted that reflection down the hole walls was apparent in the profiles when the measurements were repeated with the opaque cap, especially in the upper 80 cm of the profile (data not shown). The implications of this are explored in the discussion.

Fig. 6. Photosynthetically active radiation profiles with depth for three ridge-top and four trough profiles

Biogeochemical conditions in cryoconite holes

Table 1 shows the biogeochemical conditions within the cryoconite holes sampled immediately after their ice lids were broken. Water temperatures were almost always ∼0.1°C, so the high DO₂ conditions reveal saturation in almost all holes. In fact, just one (H10) was under-saturated with respect to atmospheric equilibrium. This was the largest cryoconite hole sampled during the study, which was also notable for a higher TDIC concentration (1.5 mg C L⁻¹) relative to average concentrations (0.84 mg C L⁻¹). However, TDIC concentrations were exceeded by DOC concentrations in the hole waters (average 1.40 mg C L⁻¹) and were far lower than the TDIC levels found in glacier ice (average 1.41 mg C L⁻¹). The cryoconite waters are therefore notable for their low ionic strength and otherwise appear dominated by Ca ²⁺, SO₄ ²⁻, Na⁺ and Cl⁻ (Table 1). When corrected for marine salts using standard marine ratios (Reference DreverDrever, 1997) only the residual quantities of Ca²⁺, Mg²⁺ and SO₄ ²⁻ were significant (data not shown). The non-marine Mg²⁺ showed strong covariance with non-marine SO₄ ²⁻ but not with non-marine Ca²⁺, suggesting that only the latter was produced by carbonate weathering. An additional source of solute is therefore required to explain these ions, and the most likely explanation is an efflorescent salt.

The chemistry of ten glacier ice samples collected while drilling at the Cryoplot is also presented in Table 1. Their similar composition to the hole waters is immediately apparent and reveals the importance of ice wall melting in the supply of water and solutes to the cryoconite habitats. In most cases, average solute concentrations of the holes are between one and two times that of the ice (data not shown). However, the average hole water concentrations are depleted in NH₄ ⁺, while the SO₄ ²⁻ concentrations are far more enriched than the other solutes. This further supports the assertion that geochemical processes other than freeze concentration occur in these holes.

Microbiology and in situ biological production

Table 2 shows that microalgae, cyanobacteria and other bacteria were present in the Vestfold Hills cryoconite holes according to the flow cytometry and epifluorescence microscopy. The total numbers of bacteria far exceeded the number of microalgae and cyanobacteria, especially in the case of the debris. In the water, far fewer types of all cells were encountered. For example, no cyanobacteria were detected in the water samples. Flow cytometry has yet to become a standard method for the study of cryoconite microbial communities. However, estimates of bacterial biomass from epifluorescence microscopy are more common, so our results are compared with a worldwide selection of data in the discussion below.

Table 3 shows the estimates of NEP, R and PP derived from the field incubations. Values normalized for the dry mass of debris range across an order of magnitude in the case of respiration (0.46–4.5 µg C(g cryoconite)⁻¹ d⁻¹) and slightly less in the case of photosynthesis (0.61–4.82 μg C (g cryoconite)⁻¹ d⁻¹). A strong, positive correlation (r² = 0.96, p<0.05) between PP and R was also found with a slope that was not significantly different from 1 (Fig. 7). Other statistically significant chemical changes noted in the incubations included NH₄ ⁺ production by dark respiration that was correlated to both respiration (r² = 0.71) and LOI (r² = 0.83). This pattern was also observed during dark incubations in the laboratory. Finally, the DO₂ changes were not significantly correlated with TDIC changes in either the light or the dark incubations, which was most likely due to the supersaturation of O₂ in the initial solutions.

Fig. 7. The relationship between photosynthesis (PP) and respiration (R) established using the values for in situ and laboratory incubations in Table 3. Values are normalized for the dry mass of cryoconite.

Laboratory determinations of biological production

Rates of photosynthesis (PP) conducted under 100 and 75 µmol m⁻² s⁻¹ PAR ranged from 4.31 to 9.17 µg C (g cryoconite)⁻¹ d⁻¹ and were therefore significantly greater than the field incubations described above (Table 3). Figure 7 also shows that the relationship between PPand R was identical to that observed in the field, in spite of their longer incubations.

Table 4 shows sediment biogeochemical conditions associated with the 12 holes from which sediment samples were abstracted for estimation of bacterial production in the laboratory. Bacterial production (BP) was far lower (average 5.73 ± 6.00 ng C (g cryoconite)⁻¹ d⁻¹) than community respiration rates shown in Table 3. Further, no statistically significant relationships (Pearson’s correlation coefficient, significance level 5%) between BP and the organic carbon content (average 1.48 ± 0.41%) were evident. The chlorophyll-a content of the debris showed considerable variability (1.70 ± 1.06 µg chl ag⁻¹) (Table 3).

The light curves produced for the six suspensions of cryoconite debris are described in Table 5. They show a significant range of maximum PP (0.014–1.78 μg C (g cryoconite)⁻¹ h⁻¹) occurring at low light intensities (5.72– 30.5 µmol m⁻² s⁻¹). The light curves therefore reveal adaptation to low light levels relative to the PAR measured at the surface of the ice at solar noon. The coefficient of determination also varies significantly, showing that at least one of the light curves was poorly defined (hole 5). In this case, photo-inhibition was not apparent but there was still a clear maximum rate of photosynthesis. Application of the light curves in Table 5 to the estimated in situ PAR levels in the instrumented cryoconite hole produced an average PP of 1.49 μg C(g cryoconite)⁻¹ d⁻¹. This compares favourably with the average PP values from the in situ incubations (Table 2) of 2.29 μg C (g cryoconite)⁻¹ d⁻¹ . Therefore our short-term incubations have probably captured realistic photosynthetic rates and confirm that the cryoconite communities are low-light-adapted.

Microbial community composition in the cryoconite holes

Reference Cameron, Hodson and OsbornCameron and others (2012) used molecular methods to examine the microbial communities in nine Vestfold cryoconite holes sampled randomly from the Cryoplot as part of a comparative study involving Arctic and maritime Antarctic cryoconite. They used terminal restriction fragment length polymorphism (T-RFLP) to create signature ‘fingerprints’ of each community, facilitating a comparison of the abundance and diversity of dominant community members from hole to hole. Bacterial communities were found to be at least 50–60% similar to each other, with the exception of two communities, which had at least 30% or less similarity to other Vestfold bacterial community members. All of the dominant eukaryotic communities were found to be over 50% similar to each other. When further analysis of one of these communities (A-V4) was performed using 16S (bacterial) and 18S (eukaryotic) rRNA sequencing techniques, a diverse community was identified that was dominated by bacterial Betaproteobacteria, Gammaproteobacteria, Bacteriodetes and Cyanobacteria, and by eukaryotic Cercozoa, Alveolata and Tubulinea (Reference Cameron, Hodson and OsbornCameron and others, 2012). Of the autotrophic organisms in the A-V4 community, bacterial Cyanobacteria and eukaryotic Chloroplastida were most important, and they constituted 15% and 9% of the total abundance found in the analysis.

Many of the 16S rRNA A-V4 clones were highly similar to GenBank clone submissions originating from a wide range of Antarctic or cryospheric ecosystems. These included soils (e.g. EF219607, Reference Yergeau, Newsham, Pearce and KowalchukYergeau and others, 2007; FJ380174, Reference Aislabie, Jordan, Ayton, Klassen, Barker and TurnerAislabie and others, 2009), glacial ice (e.g. AF479342, Reference ChristnerChristner, 2002), sea ice (e.g. AF468440, Reference Brinkmeyer, Knittel, Jürgens, Weyland, Amann and HelmkeBrinkmeyer and others, 2003) and even penguin faeces (e.g. AY218729, Zhang and others, unpublished). There was no numerical sway towards one particular habitat being more represented in this analysis, suggesting that these communities may have been seeded from a mixture of surrounding ecosystems. Additionally, clones were often related to GenBank submissions isolated from terrestrial, freshwater and marine ecosystems that are not necessarily associated with the cryosphere (data not shown). In fact the majority of the 18S rRNA clones were most closely related to GenBank clones associated with temperate soils (e.g. EF025015, Reference LesaulnierLesaulnier and others, 2008; AM114807, Reference Moon-van der Staay, Tzeneva, Van der Staay, de Vos, Smidt and HacksteinMoon-van der Staay and others, 2006).

Light propagation to the autotrophs

Wind polishing produces optical properties in the blue ice that allow the deep penetration of light and thus subsurface melting. The optical properties of the ice and the reflectance (albedo) of the debris beneath its surface therefore become the crucial factors that enable biological production within the ecosystem (with the additional benefit of providing protection from potentially harmful UV light). Although not measured in the present study, the albedo of cryoconite is reasonably well documented (see Reference HodsonHodson and others, 2008 and references therein) and typically lies between 0.10 and 0.4. However, the optical properties of blue ice are not well known, so PAR attenuation coefficients (K) were calculated by statistically fitting the following model to the profiles in Figure 6:

where Z is the depth below the ice surface (m) and PAR₀ is the PAR just beneath the surface. We used PAR profiles derived with an opaque cap over the hole, because it was apparent that up to 18% of the light in the upper 80 cm of the hole was being reflected down toward the sensor by the ice walls. Further, only the upper 60 cm of the profiles were used due to the nonlinearity of the log-transformed data at lower depths. This is a common problem owing to PAR readings being a spectral integral between 400 and 700 m and the fact that different wavelengths of light within this bandwidth are attenuated at different rates (hence the apparent blue colour of the ice). Table 6 shows the separate K estimates for the snow-free ridge top and snow-covered (by ∼5 cm) valley profiles, along with published values for snow, water, lake and sea ice. The ridge-top (snow-free) profiles indicated K values of ∼2 m⁻¹, while the trough profiles produced marginally greater values (2.6 m⁻¹). Therefore, PAR propagation through glacier ice is far superior to dry snow (published range 14–20 m⁻¹). The blue-ice values also lie within the range (1–4 m⁻¹) described for Scandinavian lake ice covers by Reference Lei, Leppäranta, Erm, Jaatinen and PärnLei and others (2011). However, the K values reported in polar lakes are far lower, for example 0.034 m⁻¹ beneath lake ice cover of the McMurdo Dry Valleys (Reference Vincent, Rae, Laurion, Howard-Williams and PriscuVincent and others, 1998).

Very low PAR attenuation coefficients can also be expected in the water layer above the cryoconite debris, because low levels of DOC were encountered here (Table 1) and no water colour was visible (in contrast to tea-coloured water described in certain holes by Reference TranterTranter and others (2004)). This most likely reflects either the effective photolytic destruction of any coloured DOC evolving within the holes, or, more likely, the short duration of time since the holes were last flushed by new meltwater (in January). The new ice lids that were then observed on all the holes were notable in that they were level with the ice surface on the ridge tops, yet lower in the case of the trough holes (Fig. 1b inset). The latter resembled half-filled pipes, with the upper part of the hole acting as an effective snow trap. Therefore, PAR penetration beneath these holes was probably far lower than our PAR profiles suggest, because the snow thickness was sometimes more than double that observed in the troughs where the boreholes were made. The shallower depths of cryoconite holes shown in the troughs (Fig. 5b) were therefore largely controlled by the effects of wind-blown snow accumulation and perhaps some shading by the valley sides.

In spite of the above complexity associated with PAR downwelling into Cryoplot, a key finding in this study is that the levels of irradiance reaching the cryoconite debris were broadly uniform, regardless of their depth (i.e. 9.0–11.5% according to the midday boreholes and the continuous measurements shown in Fig. 2b). Such near-uniform light conditions suggest that cryoconite depths were at equilibrium with respect to the ablation rate at the ice surface (discussed below) and the solar heating of the debris within the hole. Several authors have described this equilibration process (e.g. Reference GribbonGribbon, 1979), and experiments on Arctic glaciers and Antarctic lake ice (Reference HodsonHodson and others, 2010a; Reference Jepsen, Adams and PriscuJepsen and others, 2010) show that the rates of debris melting into the ice surface allow it to reach its equilibrium depth within 1 or 2 days. This is sufficiently rapid to mean that short-term (e.g. diel) variations in solar radiation such as those shown in Figure 2b will force the most important changes in the rate of photosynthesis upon the debris surface.

The observation that ice lids only lie flush with the ice surface upon the ridge holes suggests they were full of meltwater and began to refreeze before meltwater drainage was possible. By contrast, partial drainage of holes in the troughs prior to refreezing seems to have occurred. Two things might explain this: first, the hydraulic efficiency of the near-surface ice beneath the troughs might have been greater; and second, the thin snow cover in the troughs might have delayed refreezing there, allowing more time for partial drainage. We believe thermal-induced cracking is one reason why drainage might be more efficient in the troughs. Cracking was audible every day, resulting in a popping sound whose tone varied in accordance with the width, depth and water volume of the hole being cracked open. Since holes in the troughs were nearest to the ice surface, cracks were more likely to propagate to the water layer in the hole and initiate drainage. It is therefore possible that cracking enhances the ventilation and interconnectivity of the trough holes compared with their ridge-top counterparts. The ecological implications of these differences should therefore be explored in future.

Biological production and nutrient balance in the Vestfold Hills cryoconite ecosystems

Figure 8 shows average rates of NEP, PP and R according to the in situ incubations conducted during the present study and also conducted by the authors using the same equipment in southwest Greenland and central Svalbard (Reference HodsonHodson and others, 2010a, Reference Hodsonb). The rates are clearly low in the Vestfold Hills cryoconite compared with the Arctic. Low rates of PP (0.4–1.4 µg C (g cryoconite)⁻¹ d⁻¹) have also been found with defrosted samples taken from Canada Glacier, McMurdo Dry Valleys, Antarctica, and incubated in the laboratory (Reference Foreman, Sattler, Mikucki, Porazinska and PriscuForeman and others, 2007). However, when the biological production measurements are normalized for total organic matter (rather than total debris mass), the rates of biological production in the various sites become rather similar (Fig. 8b). A shortage of organic debris supply might therefore limit biological production in the Vestfold Hills. In our study, the average chlorophyll-a specific rate of production (or assimilation number) was 0.06 ± 0.02 µg C (µg chl a)⁻¹ h⁻¹ for the in situ incubations and 0.32 ± 0.13 mg C (μg chl a)⁻¹ h⁻¹ for laboratory incubations. The rates are therefore typical of the lower values reported for Antarctic lakes by Reference Henshaw and Laybourn-ParryHenshaw and Laybourn-Parry (2002), and also the ultra-oligotrophic Crooked Lake in the vicinity of Flanders Moraine (Reference Bayliss, Ellis-Evans and Laybourn-ParryBayliss and others, 1997). As far as we are aware, these are the first estimates of the assimilation number for any cryoconite ecosystem. Photosynthetic efficiencies were low compared with polar lakes (i.e. 0.007 ± 0.003 and 0.004 ± 0.001 µg C (µg chl a)⁻¹ h⁻¹ µmol m⁻² s⁻¹ for the in situ and laboratory incubations respectively. Canadian high-Arctic lakes had photosynthetic efficiencies ranging between 0.53 and 2.30 g C (g chl a) ⁻¹ h⁻¹ µmol m⁻² s⁻¹ (Reference Markager, Vincent and TangMarkager and others, 1999), while in Antarctica the photosynthetic efficiencies ranged between 0.002 and 0.39 µg C (µg chl a)⁻¹ h⁻¹ µmol m⁻² s⁻¹ at Beaver Lake (Reference Laybourn-Parry, Madan, Marshall, Marchant and WrightLaybourn-Parry and others, 2006) and between 0.02 and 5.19 µg C (µg chl a)⁻¹ h⁻¹ µmol m⁻² s⁻¹ at Crooked Lake (Reference Henshaw and Laybourn-ParryHenshaw and Laybourn-Parry, 2002). One reason the estimates in Vestfold Hills cryoconite holes are low is that they are being compared with planktonic and not benthic systems, so shading effects are likely (see, e.g., Reference TellingTelling and others, 2012). A further reason is the entombment beneath the ice lid, which will limit the supply of TDIC and nutrients to periods of sporadic surface/near-surface runoff and melting at the hole floor during periods of surface ablation. Representative values for supply by the hole floor melting can be calculated easily because melting at the base of the cryoconite hole appears to be in equilibrium with the surface ablation rate (the so-called ‘equilibrium depth scenario’). A further assumption that the holes are hydrologically isolated is also necessary, which we think is reasonable on the ridges, where the hole depths appear to be safe from crack-induced drainage and were full prior to formation of the ice lid. This being the case, fluxes of nutrients into the holes approximate to the product of the daily surface ablation rate and the average nutrient content of glacier ice (Table 1). Average surface ablation was calculated to be 5.03 mm d⁻¹ using ice surface measurements at the stake supporting the pyranometer in the Cryoplot. Table 7 shows the resulting nutrient flux calculations for dissolved inorganic N (i.e. NH₄ ⁺ + NO₃ or ‘DIN’) and TDIC. Unfortunately, both PO₄ ³⁻ and DOC in the glacier ice were below detection limits, so we limit our analyses here to just carbon and nitrogen. What is immediately clear is that the observed rates of PP and R impose demands upon DIC and DIN resupply that cannot be met by ice melting at the base of the hole. For example, just 1.39 mg C m⁻² d⁻¹ TDIC was typically liberated from ice melt, while 31.7 mg C m⁻² d⁻¹ was fixed by primary production in the incubations. The implication therefore is that respired carbon is crucial for photosynthesis during entombment. Note that the glacier melt resupply rate, while low, is of the same order of magnitude as NEP (average 4.0 mg C m⁻² d⁻¹). Tight coupling between the heterotrophic and autotrophic communities can therefore be expected on account of this TDIC limitation. The same can be deduced when the nitrogen balance is considered, because NH₄ ⁺ production was a clear feature of the dark incubations, yet it was not observed in the light incubations. This suggests that NH₄ ⁺ was being effectively recycled by primary production during photosynthesis. Table 7 suggests that a likely demand of DIN imposed by the observed PP (according to Redfield stoichiometry) was 5.61 mg N m⁻² d⁻¹, which is almost satisfied by the observed liberation of NH₄ ⁺ in the dark incubations (3.33 mg N m⁻² d⁻¹). Therefore, like the entombed holes of McMurdo Dry Valleys glaciers (Reference TranterTranter and others, 2004), recycling of organic nutrients seems likely to meet the demands imposed by biological production.

Fig. 8. Average rates of respiration (R), net ecosystem production (NEP) and photosynthesis (PP) normalized for the total dry mass (a) and total mass of organic carbon (b). Data are shown for the present study (EAIS) and two other sites in the Arctic for comparison: Kangerlussuaq in Greenland (GIS), and Longyearbreen in Svalbard (LYR) (after Reference HodsonHodson and others 2010a, Reference Hodsonb respectively). Error bars are standard deviation of the multiple incubations conducted in each study

Another feature of the respiration measurements is that the average rates of bacterial production according to Table 4 were ∼300 times lower (average 5.73 ng C g⁻¹ d⁻¹) than average community respiration (1870 ng C g⁻¹ d⁻¹; Table 3), and also lower than most rates deduced on Northern Hemisphere glaciers using an identical methodology to the present study (3.1–950 ng C (g cryoconite)⁻¹ d⁻¹; Reference AnesioAnesio and others, 2010). It is therefore difficult to argue that bacteria are a significant source of the respired carbon, even if the bacterial growth efficiency is low. For example, just 30% of the average community respiration in Table 3 could be explained by average rates of bacterial production if their growth efficiency was only 1%. Inorganic carbon production by carbonate weathering has already been accounted for in the carbon budget, and organic carbon photolysis cannot influence the dark respiration values. Therefore, unless experimental artefacts greatly influence the comparison of our ex situ bacterial production measurements and our in situ community respiration measurements, respiration by the autotrophic community and/or other heterotrophic organisms must be invoked. Reference Cameron, Hodson and OsbornCameron and others’ (2012) study of the 16S rRNA gene clone libraries revealed several protists and other potential heterotrophs that could have been responsible for the unexplained community respiration. However, their abundance and activity were not established, so it remains unclear where the respired carbon originated.