Over the past two decades, with recognition that the ocean’s sea-ice cover is neither insensitive to climate change nor a barrier to light and matter, research in sea-ice biogeochemistry has accelerated significantly, bringing together a multi-disciplinary community from a variety of fields. This disciplinary diversity has contributed a wide range of methodological techniques and approaches to sea-ice studies, complicating comparisons of the results and the development of conceptual and numerical models to describe the important biogeochemical processes occurring in sea ice. Almost all chemical elements, compounds, and biogeochemical processes relevant to Earth system science are measured in sea ice, with published methods available for determining biomass, pigments, net community production, primary production, bacterial activity, macronutrients, numerous natural and anthropogenic organic compounds, trace elements, reactive and inert gases, sulfur species, the carbon dioxide system parameters, stable isotopes, and water-ice-atmosphere fluxes of gases, liquids, and solids. For most of these measurements, multiple sampling and processing techniques are available, but to date there has been little intercomparison or intercalibration between methods. In addition, researchers collect different types of ancillary data and document their samples differently, further confounding comparisons between studies. These problems are compounded by the heterogeneity of sea ice, in which even adjacent cores can have dramatically different biogeochemical compositions. We recommend that, in future investigations, researchers design their programs based on nested sampling patterns, collect a core suite of ancillary measurements, and employ a standard approach for sample identification and documentation. In addition, intercalibration exercises are most critically needed for measurements of biomass, primary production, nutrients, dissolved and particulate organic matter (including exopolymers), the CO2 system, air-ice gas fluxes, and aerosol production. We also encourage the development of in situ probes robust enough for long-term deployment in sea ice, particularly for biological parameters, the CO2 system, and other gases.
1. The rise of sea-ice biogeochemical studies
Sea ice covers up to 8% of the Earth’s ocean surface (Steele et al., 2001), and despite global warming trends, both polar oceans are still mainly covered by sea ice in winter (Comiso, 2010) and likely will continue to be for the foreseeable future. The changes in sea-ice extent and physical structure associated with a warming climate (Perovich and Richter-Menge, 2009; Massom and Stammerjohn, 2010) are causing dramatic shifts in sea-ice ecosystems and the interactions between sea ice and both the atmosphere and the underlying waters. Long assumed to be a passive barrier to both light and matter, sea ice was relatively neglected in biogeochemical studies until the early 1990s. Since then, intensive research in the Arctic and Southern Oceans, as well as in sub-polar seas, has shown that, in reality, sea ice is an active player in biogeochemical processes, making significant contributions to regional and possibly global cycles of many elements (e.g., Arrigo et al., 2010; Deming, 2010; Thomas et al., 2010; Loose et al., 2011; Rysgaard et al., 2011; Vancoppenolle et al., 2013). Future changes in the sea-ice environment will be accompanied by changes to these biogeochemical cycles, generating an urgent need to better understand the chemical-physical-biological function of the ocean-ice-atmosphere system.
Although this review is focused on recent methodological developments in sea-ice biogeochemistry, the first formal biological studies of sea ice date back to the mid-19th century (Horner, 1985), and chemical studies extend back to the early 20th century (e.g., Ringer, 1928; Wiese, 1930). Post-war studies of sea-ice chemistry were mainly motivated by efforts to understand sea-ice structural properties in support of cold-war military operations and potential industrial development in the polar regions (e.g., Nelson and Thompson, 1954; Assur, 1958; Bennington, 1963; Tsurikov, 1965), but that work also provided extremely useful information on the geochemistry of ice brines. Incremental work continued at a slow pace for several decades, but with recent increases in access to polar regions and technological developments, as well as with the growing urgency in climate-change research (e.g., Post et al., 2013) and a need to improve representation of sea-ice processes in numerical models at all spatial and temporal scales, the study of sea-ice biogeochemistry has expanded rapidly. Scientists have come to this field from a variety of disciplines, including glaciology, oceanography, sedimentology, and even tundra ecology; as we attempt to understand the complex biogeochemical processes occurring in sea ice, many creative modifications have been applied to methods not originally designed for sea-ice applications.
Sea ice presents a particularly challenging environment for biogeochemical studies (see Petrich and Eicken, 2010, for a comprehensive description of sea-ice types, characteristics, and life-cycles). Perhaps most significantly, the sea-ice environment is cold. Standard seawater with a salinity of 35 g kg-1 freezes at -1.9 °C (Petrich and Eicken, 2010). Sea-ice temperatures below -30 °C have been measured (e.g., Miller et al., 2011b), while the air temperatures above the ice, where people and instruments must operate, can drop to -60 °C or even lower. Standard oceanographic equipment, however, has been built to operate at temperatures only as low as 0 °C, while most chemical analyses and instrumentation are designed to operate between 20 and 25 °C. If it is not cold, i.e., if the temperatures are around 0 °C, as occurs during the spring and autumn transition seasons, the sea ice is often very thin or deteriorating, making sampling dangerous. Additional difficulties arise from the heterogeneity of sea ice, which is a complex mixture of pure ice, solid salts and particulate organic matter, liquid brines, and gas bubbles. Even the boundary conditions are variable, with snow, frost flowers, and melt ponds at the top and skeletal ice, platelet ice, and algal mats at the bottom. In many ways, soils or sediments are a better conceptual model than seawater for describing sea-ice spatial variability, for multi-phase theories are required to describe sea-ice physical-chemical properties (i.e., a “mushy layer”; Vancoppenolle et al., 2010; Hunke et al., 2011). Finally, the high and variable salinities of sea-ice brines compromise chemical analyses by complicating calibrations and corroding delicate instrument components.
Comprehensive guides for biological and physical methods in sea-ice research have already been published (Horner et al., 1992; Eicken et al., 2009; Michel and Niemi, 2009); we will not attempt to reiterate them here. Rather, we are accepting the challenge issued by Eicken et al. (2009) in their preface, wherein they hoped their book would “spark broader collaboration among sea-ice researchers to document and refine the best-practice approaches to sea-ice field studies.” We discuss not only some remaining ambiguities in determining biomass, nutrient concentrations, and the rates of biological processes in sea ice, but also the challenges of quantifying gas concentrations and fluxes, aerosol emissions, trace metals and their chemical speciation, the complex inorganic carbon and organic sulfur systems, and the sticky problem of organic matter in sea ice, including the difficulties in defining and distinguishing between the dissolved, colloidal, and particulate fractions. We also make concrete recommendations for what ancillary physical data should be collected in conjunction with biogeochemical measurements, to allow effective interpretation of the resulting data sets, as well as for the most critical and potentially useful directions for future methodological developments. We do not intend this paper to be a stand-alone methods manual. Rather, we direct readers to the sources in which individual methods are described.
2. General considerations
Each element, compound, or process has a unique set of sampling and analytical requirements for accurate, precise, and useful results. The bulk of this paper addresses the specialized procedures for sampling and analyzing a wide range of parameters in sea ice. However, there are several generalities in sampling ice and its components that are worth discussing before tackling the specifics.
2.1. Patchiness and scaling
Sea-ice physical, chemical, and biological properties are highly variable, both temporally and spatially. During most seasons, sea ice is characterized by strong vertical gradients in temperature, brine salinity, habitable pore space, and permeability (Eicken, 1992; Petrich and Eicken, 2010; Vancoppenolle et al., 2013). Particularly during spring and, possibly, autumn, those gradients can change rapidly on daily or even hourly timescales (e.g., Mundy et al., 2005; Nomura et al., 2010a). Horizontal variability is also extremely high; for example, biomass can vary by an order of magnitude on the sub-meter scale (Spindler and Dieckmann, 1986; Steffens et al., 2006). Horizontal patchiness of ice algae has been mainly attributed to the spatial variability in physical sea-ice properties (Eicken et al., 1991) and light exposure (Raymond et al., 2009), which are affected by ice formation processes, parent seawater salinity, and meteorological events, among other factors (Gosselin et al., 1986; Rysgaard et al., 2001; Granskog et al., 2005a; Fritsen et al., 2011). This sea-ice heterogeneity has consequences for sampling design: not only do the goals of the project dictate the type of ice that should be sampled (i.e., first-year, multi-year, smooth, ridged, young, melting, etc.), but the most appropriate sampling scheme and the minimum number of samples required will depend on the representativeness of any single sample.
The scales of horizontal spatial variability in biological parameters in different sea-ice regimes have been investigated using transects and nested equilateral triangle sampling patterns combined with parametric and non-parametric statistical analysis techniques (e.g., Gosselin et al., 1986; Swadling et al., 1997; Granskog et al., 2005a; Steffens et al., 2006; Søgaard et al., 2010, 2013). We recommend that, to the extent possible and relevant to the specific study, researchers design their sampling using a nested approach that facilitates extrapolation of detailed information to larger scales by distinguishing hierarchical layers of detail (Figure 1). In a nested sampling regime, the primary scale defines the study area, in all its variation, and the secondary scale serves to determine the representativeness of each site within the study area. The tertiary scale (i.e., the number of individual sampling sites), along with the number of replicates (the quaternary scale), defines the accuracy of a parameter.
Designing a nested sampling program begins with a visual survey of the sampling area to establish the various spatial scales of the variability and determine how many sites and samples are required at each hierarchical level. There is no simple, universal algorithm that can be used; every sampling site must be assessed in relation to the goals and resources of the project. For example, Sturm (2009) provides recommendations for sampling densities required to minimize errors in mean snow depth (see his Figure 3.1.13), but the ease of snow-depth measurements allows for a large number of data points that are not always practical for studies of biogeochemical properties.
In addition, the practical and safety requirements associated with ice coring unavoidably produce a bias towards stable ice floes with low levels of deformation. The true scale of the horizontal variability can be determined more accurately by combining methods that provide information on varying scales (e.g., ice coring surveys combined with optical investigations from underwater platforms; Williams et al., 2013). State-of-the-art methods to determine floe-scale sea-ice physical properties, such as sea-ice surface elevation, snow thickness, freeboard, and sea-ice draft, need to be linked with new methods in sea-ice ecology (e.g., determination of ice algal biomass from transmitted under-ice irradiance spectra; Mundy et al., 2007) to evaluate coupled physical-biological sea-ice processes on relevant scales. Developing ice buoy networks and autonomous underwater vehicle technology, combined with improved physical (e.g., CTDs, upward looking sonars), chemical (e.g., oxygen and pH sensors), and bio-optical (e.g., hyperspectral radiometers, fluorometers) sensors, is an active field of research and promises a step-change in our understanding of horizontal patchiness and physical-chemical controls over biological properties in sea ice.
2.2 Sampling techniques and considerations
In general, because sea ice is highly heterogeneous (section 2.1), collecting different samples for analyses of different parameters makes it difficult to confidently link the biogeochemistry of those parameters. Therefore, to the extent possible, each individual sample should be analyzed for as many parameters as is feasible. However, this ideal goal is severely constrained by realities of required sample volumes for analyses, incompatible sample processing requirements, and vulnerabilities to different contamination sources. Here we summarize the basic techniques to collect samples from sea-ice environments, noting that the best approach may vary depending on the project goals and the sea-ice conditions. With this in mind, we have also included what we feel are important considerations to take into account when sampling each particular medium.
2.2.1 Bulk ice
A standard approach to processing sea-ice samples is to collect a core and divide it into sections using a clean stainless steel saw, depending on the scientific question and the demands of analytical sample volume and of collaborative cooperation between researchers (Figure 2). The core sections are then melted and analyzed as any other aquatic samples. The thickness of the core sections may vary, according to the needs of the project, but parameters to be compared should be analyzed on sections of comparable thickness. Cores are usually sectioned from the bottom, to limit brine loss from more permeable parts of the ice (the vertical extent of which depends on ice temperature, salinity, and texture), although the individual sections should be identified according to their depth from the top of the ice (see section 2.3). Core extraction invariably results in at least some brine loss (e.g., Notz et al., 2005), contributing to losses of both dissolved and particulate matter, and therefore, cores must be sampled and processed quickly, preferably sectioned into melt containers in the field. Divers have collected sea ice from below in attempts to more quantitatively recover the components of bottom ice (e.g., Welch et al., 1988; Horner et al., 1992; McMinn and Hegseth, 2007). In addition, although analyses of melted ice cores are most common, many interesting analytes and processes (such as microorganism abundances and species compositions, metabolic rates, gas partitioning, salt precipitation and dissolution, etc.) are strongly affected by the drastic changes in temperature and salinity that result when sea ice melts (see section 3.2, below); more complex sampling and analysis methods are often required to avoid melting samples.
The early stages of ice formation play a fundamental role in partitioning material and organisms between the atmosphere, ice, and underlying water (e.g., Giannelli et al., 2001; Notz and Worster, 2009; Müller et al., 2013), but sampling young ice that cannot bear a load (i.e., frazil, grease, nilas, or pancake ice) requires special safety considerations. Thin, broken ice (brash ice and small pancakes) can be sampled from a dinghy (e.g., Grossmann and Dieckmann, 1994) or with a bucket or basket lowered from the side of the ship (e.g., Gradinger and Ikävalko, 1998). Intact, young ice sheets can be cored or cut directly by researchers from a ship’s basket (Figure 3a) or using a flat-bottomed boat (Figure 3b). Young, poorly consolidated ice samples contain high quantities of interstitial water and brines that are easily lost, which makes it important to record whether young ice samples were drained to separate the pore water or if special care was taken to retain the brines (Grossmann and Dieckmann, 1994). Cottier et al. (1999) and Smedsrud and Skogseth (2006) described specialized tools for collecting frazil and young sea ice. High-salinity brine waters surrounding frazil and brash ice as it is forming can also be collected using open-mouth jars covered with mesh (Miller et al., 2011a). Similarly, Kristiansen et al. (1998) sampled the infiltration community at the snow-ice interface of flooded ice floes by immediately filtering the slush, at low temperature, through a 200 µm net to remove ice crystals.
Thick, ridged sea ice is even more severely undersampled than young, thin ice, due to a number of significant difficulties with travelling over and through ridged ice, accessing ridge keels and sails, and collecting representative samples from such a varied environment (see section 2.1). However, ridged ice constitutes a large fraction of the total area and volume of sea ice and, according to the limited data available, represents an important biomass pool (Gradinger et al., 2010; Meiners et al., 2012). The methods used for sampling thick, deformed ice have been largely ad hoc; standardized approaches are needed.
The fractionation between the solid (ice and particulates) and liquid (brine) phases of the ice is often important to understanding sea-ice biogeochemical cycles, but effectively collecting and accurately analyzing representative brine samples from mature sea ice has been an exceptional challenge. Traditionally, sea-ice brines have been collected by drilling sackholes (partial core holes) to a desired depth within the ice and then allowing brines from the surrounding ice to drain into the hole (Figure 4a). The most obvious problem with this approach is that, because of the three-dimensional structure and variable connectivity of the brine network, the sackhole brine integrates the geochemical properties of numerous individual brine channels from an undefined volume of the sea ice surrounding the hole. Therefore, sackhole brines provide data on the vertical and horizontal macro-scale (over several tens of centimeters, depending on the depth of the sackhole); our ability to obtain data on the brines at the micro-scale level (a few centimeters or less) in sea ice is still limited for the majority of solutes of biogeochemical interest. Also, if the ice is warm and highly permeable, the collected brines can be contaminated with upward seeping seawater or draining snowmelt and meltpond water; the significance of this problem can be assessed by comparing the measured brine salinity to that estimated from the in situ temperature profile. Conversely, the colder the ice, the less brine it contains, the less connected the brine network is, and the longer it takes for a sufficient volume of brine to accumulate in the sackhole. At very cold temperatures, it may not be possible to collect enough brine within a suitable timeframe. Furthermore, the longer it takes for the brines to accumulate, the greater the risk that the sample will be compromised by interaction with the air, which is likely at a different temperature (often much lower) than the ice interior, causing the brines to freeze further within the sackhole. Potential brine-air exchange is a particular problem for analyses of insoluble gases (section 4.4). To minimize such air-brine interactions, the sackhole can be capped with plugs made from thick insulated material. When sampling sackholes, any snow needs to be removed from the sampling site before coring, and care is required to prevent ice core shavings from entering the hole and contaminating the brines. The accumulated brine can be extracted from the sackhole with a large pipette (a “turkey baster”; Figure 4b), with tubing attached to a syringe or a peristaltic pump, or simply by dipping a bottle into the brine.
Other methods used to collect sea-ice brines include gravity-draining full cores into containers (Figure 4c; Nomura et al., 2009) or crushing and/or centrifuging ice samples (e.g., Grossmann and Dieckmann, 1994; McMinn et al., 2009; Munro et al., 2010). However, both of those approaches generally deliver relatively small volumes that are suitable for only a limited number of analyses and may not be statistically representative (see section 2.1). Also, the pressure generated by centrifugation can melt some of the ice, diluting the extracted brines (Papadimitriou et al., 2004), apparently by as much as 15%.
Particulates (including organisms) and organic matter appear to be under-represented in brines collected in sackholes (e.g., Weissenberger, 1992; Sime-Ngando et al., 1997; Lannuzel et al., 2008) by up to 98% (Becquevort et al., 2009). Likely explanations include preferential adsorption onto the ice walls, “filtration” by the brine channel network, and impeded transport by sticky, gelatinous exopolymeric substances (EPS) or aggregates of ice algae (Meiners et al., 2004; Krembs et al., 2011).
Concentrations of dissolved materials in brines can also be estimated from calculated brine salinity (based on in situ temperature; Cox and Weeks, 1983; Petrich and Eicken, 2010) and measured bulk ice concentrations (Figure 5), using the equation
where C indicates analyte concentration, S is salinity, and the subscripts B and I represent brine and bulk ice (e.g., Dieckmann et al., 1991a; Norman et al., 2011). This calculation assumes that none of the analyte is in solid or gaseous form and that, therefore, the concentration of the analyte in the brine is directly proportional to salinity, an assumption that is probably only valid for highly soluble substances. Not only do supersaturated brines precipitate salts (even the highly soluble ions Cl- and Na+ precipitate from sea-ice brine solutions to a significant extent at temperatures below -24 °C; Assur, 1958) and release gas bubbles, but both dissolved and particulate organic matter can adsorb onto the surface of the brine channels. In addition, high concentrations of organic matter, particularly EPS (section 4.2.3), can impact sea-ice microstructure (Krembs et al., 2011), with hypothesized (Ewert and Deming, 2013) but still unknown implications for the sea-ice equations of state (Cox and Weeks, 1983).
The third phase in sea ice, gas bubbles, are particularly important in carbon, oxygen, and sulfur cycling, but gas inclusions are even more challenging to recover and analyze than brines. The methods that have been developed to tackle this problem are summarized in section 4.4.
2.2.4 Snow and frost flowers
On top of sea ice, the snow cover thermally insulates the ice, limits the penetration of visible and UV radiation, and exchanges brines and gases with the underlying ice (e.g., Kelley and Gosink, 1985; Massom et al., 2001; Zemmelink et al., 2008; Ewert et al., 2013). The snow cover is generally sampled in layers by excavating snow pits, as described by Sturm (2009). Established sampling methods for other environments can usually be adapted to snow over sea ice. For example, clean methods used for sampling trace elements in snow over land have been successfully implemented for sampling of halogens and trace elements in snow over sea ice (Simpson et al., 2005, Poulain et al., 2007).
Frost flowers are usually collected by simply scraping or scooping them into sample containers (Figure 6; Obbard et al., 2009; Bowman and Deming, 2010; Miller et al., 2011a; Aslam et al., 2012; Douglas et al., 2012; Bowman et al., 2013; Fransson et al., 2013; Granfors et al., 2013a), although Alvarez-Aviles et al. (2008) used tweezers. Brine skims on the top of the ice (often associated with frost flowers) can be collected with a scooped spatula (Bowman and Deming, 2010; Roscoe et al., 2011) or an eyedropper (Alvarez-Aviles et al., 2008). However, exclusive sampling of frost flowers separately from brine skims remains a challenge, nor do current approaches likely adequately capture volatile components. Effective study of these sea-ice micro-environments can benefit from application of non-invasive techniques, such as infrared imaging (Barber et al., 2014), and from development of sensitive but robust microsensors for in situ analyses.
Sea-ice scientists still use non-standardized, ad hoc systems for identifying samples, but if we are to establish comprehensive, accessible, and useful sea-ice biogeochemical databases, then instituting basic standards to classify samples will be necessary. These metadata requirements apply not only to observations at the time of sampling, but also to sample processing and preparing data sets for archiving. In an effort to begin to meet the need for standardized record-keeping, we encourage researchers to always identify their samples with the following information:
latitude and longitude;
date and time (preferably UTC, but if local time is used, clearly identify the time zone and whether it is under Summer, or Daylight Savings, Time);
weather conditions, including air temperature, cloud fraction, wind speed and direction, and contact information for complete meteorological data from a nearby ship or station;
water depth, particularly in coastal waters;
ice description, including approximate age (i.e., whether it is multi-year or first-year ice, and for young ice, estimates of time since initial ice formation), thickness, freeboard, and texture (according to Eicken et al., 2009);
if applicable, depth in the ice core, measured from the top, in cm; and
if applicable, estimates of snow or melt-pond coverage and measurements of their depths on the surface of the ice.
For each core, a standard data sheet should be prepared, giving the important metadata, information on how the core was processed, and what subsamples were taken for what analyses (Figure 7); eventually these sheets should be also populated with the analysis results for archiving. As it is usually best to collect several ice samples or cores, the distance between samples or the approximate total area from which the samples originated must also be specified in the metadata. If possible, replicates should be collected according to the guidelines for nested sampling design in section 2.1.
To facilitate data archiving and retrieval, we suggest that each sample be identified (at least in data files, if not also on the samples themselves) by an expedition code, followed by the date in YYYYMMDD format (with a lower-case letter for each sampling ‘event’ on a given day, associated with a unique time and location), followed by an identifier of the sample type (i.e., “ice” for sea-ice cores, “br” for sackhole brines, “ff” for frost flowers, “sn” for snow, “gap” for gap layers, and “pond” for melt ponds), with each replicate core or sample receiving a different sequential number. Individual core sections should then be identified by the depth from the air-ice or snow-ice interface. For example, a specific core section from an Antarctic sea-ice camp in the year 2015 might be identified as BS2015-20150409b-ice-03-20-30, where BS2015 would be the expedition code, followed by the date (April 9th) and “b” indicating that it was the second location sampled that day, the 3rd replicate core from that location, and the 20–30 cm section down in that core.
2.4. Ancillary measurements
Any biogeochemical study of sea ice should include a number of ancillary measurements (Table 1). Beyond the needs for working up and interpreting our own results, when archived data are used by later researchers investigating questions we have not yet conceived, the ancillary data may prove critical.
Most importantly, the physical properties of the ice should be described in as much detail as possible (see Eicken et al., 2009, for standard methods). At a minimum, the in situ ice temperature and bulk salinity should be determined and reported, preferably in 5-cm vertical increments throughout the depth of the ice. Indeed, bulk salinity is sufficiently important (and simple enough to measure) that it should be determined on all sections analyzed for any biogeochemical parameter.
We also encourage sea-ice investigators to report brine salinity, which can be calculated from temperature (assuming thermodynamic equilibrium; Petrich and Eicken, 2010), and the brine volume as a fraction of the bulk sea-ice volume, which provides insight into the connectivity of the brine channel network (Golden et al., 1998). Estimating the brine volume fraction (Cox and Weeks, 1983, for T < -2°C; Leppäranta and Manninen, 1988, for T > -2°C) requires that in addition to temperature and bulk salinity, the ice density be measured (Eicken, 2009) or estimated based on an assumed air volume content. It is also helpful to describe the ice texture throughout the full profile (Eicken, 2009), at least based on a direct visual inspection of the thicknesses and sequence of granular and columnar layers (which can be recorded with photographs). A polarized light source, if available, can provide more detailed information (Figure 8). The δ18O ratio of the sea ice (section 4.8) helps identify its meteoric versus marine origins (e.g., Macdonald et al., 1999; Tison et al., 2008). Finally, macronutrient (nitrate, nitrite, phosphate, silicic acid, ammonium; section 4.1) and chlorophyll a (Chl a; section 3.3.2) measurements, in combination, allow us to assess the overall status of the biological community in the ice (i.e., bloom versus non-bloom states; degree of nutrient limitation, etc.).
At a minimum, the presence of a snow cover and its thickness need to be reported. If resources allow, the snow cover should be fully characterized, as described by Sturm (2009), including basic stratigraphic analysis, with measurements of thickness, temperature, density, hardness, and grain properties (shape and size) for each layer. Snow samples from different strata should be also collected and analyzed for δ18O (for insight into brine exchange with the underlying sea ice), as well as for the same analyses as those planned for the ice samples. More specialized information on atmospheric forcing, both light and heat, is also valuable, and standard methods are given by Perovich (2009).
3. Biological factors
Sea ice contains an abundant and taxonomically diverse community, including Bacteria and Archaea (hereafter referred to jointly as bacteria), autotrophic algae, hetero- and mixo-trophic protists, fungi, and metazoans. Algal cell concentrations in bulk sea ice range over six orders of magnitude (from 101 to 106 cells/mL; Arrigo et al., 2010), whereas those of bacteria range over four orders of magnitude (from 103 to 107 cells/mL; Deming, 2010) and those of viruses over three orders of magnitude (from 105 to 108 viral particles/mL; Deming, 2010). In order to understand and budget the exchanges of matter and energy within this rich sea-ice microbial community and between it and the surrounding environment, including the atmosphere, the water column, and sediments (i.e., the paleorecord), it is essential to have good measurements of biomass, its chemical composition (C, N, P, Si, etc.), taxonomic composition, and metabolic functions and transformation rates. Such measurements are challenging in any environment, and particularly so in sea ice.
Sampling considerations for determining the biological properties of sea ice depend on the parameter, organism of interest, and type of ice sampled. When sampling for microorganisms, especially bacteria, application of sterile aseptic technique, to the extent possible in the field, is needed; e.g., the use of ethanol-sterilized tools for coring, cutting or otherwise collecting the samples, and sterile receptacles (e.g., Bowman et al., 2012). Tools must be ethanol-cleaned between samples to prevent cross-contamination. Samples should be kept as close to the in situ temperature as possible, to prevent thermal or osmotic shock to the microorganisms, and light exposure needs to be minimized during transport and storage to limit growth, light shock, and photochemical degradation.
When the planned analysis requires full, intact ice cores, the cores should be placed into sterilized, black plastic sleeves, to retain brines and protect organisms from light stress, immediately after extraction from the core barrel (e.g., Maas et al., 2012). Typically, however, the core is divided into sections to determine detailed vertical profiles (section 2.2.1). When cutting ice cores into sections, contamination may be prevented by placing them on autoclaved foil (Bowman et al., 2012) or, as when sampling for organic compounds (section 4.2.1), removing the outer layers of the cores (Song et al., 2011; Fripiat et al., 2014a). If the bottom ice community is of interest, the core must be handled gently to preserve the lower skeletal layer. Dieckmann et al. (1992) developed an apparatus for sampling unconsolidated platelet layers that can be deployed through a hole in the ice only 5 cm in diameter and was successfully deployed by Arrigo et al. (1995) and Robinson et al. (1998) to study the biogeochemistry and photophysiology of the platelet ice community. When large sample volumes are required, sections from different cores can be pooled during melting, although information on the horizontal heterogeneity (section 2.1) is lost.
3.2. Sample processing
Ideally, we would study the sea-ice biological community in situ in order to fully understand the relationship between the organisms and their environment (Junge et al., 2001; Krembs et al., 2002). However, most standard methods for evaluating biomass and activity are unable to accommodate the bulk ice matrix; investigators usually melt ice samples before examining the biological community. Unfortunately, large temperature changes or osmotic stress, as the salinity drops during ice melt, can cause sympagic cells to burst, a significant concern in biological sea-ice studies. In addition, even if cells remain intact, photosynthetic stress has been observed in sea-ice algae subjected to dramatic salinity decreases (Ralph et al., 2007). Therefore, samples should generally be processed in the cold (i.e., at or only slightly above the freezing point of the final melt solution) to limit temperature changes (Deming, 2010; Mikkelsen and Witkowski, 2010), but limiting osmotic shock is more difficult. A variety of methods are used to melt sea ice, with a range of potential impacts on the sea-ice community. Available protocols include simply melting the ice (direct melt), melting in filtered or artificial seawater (seawater melt), and melting in concentrated brine to give a final salinity similar to the in situ brine salinity (brine, or isohaline, melt). Use of direct and seawater melts in samples collected for analyses of nutrients, organics, extracellular polysaccharides, and sulfur species is discussed in sections 4.1, 4.2.1, 4.2.3, and 4.5, respectively.
Although some studies have observed no difference between direct and seawater melts for Chl a measurements (Dieckmann et al., 1998; Kaartokallio, 2004) or diatom counts (Mikkelsen and Witkowski, 2010), or for the culturable fraction of sea-ice bacteria (Helmke and Weyland, 1995), direct melting can cause loss of anywhere from 13 to 97% of eukaryotic cells, with ciliates and flagellates most susceptible to bursting (Garrison and Buck, 1986; Mikkelsen and Witkowski, 2010). However, Mikkelsen and Witkowski (2010) found that slow, direct melting under refrigerated conditions appeared to be suitable for most of the eukaryotic cells, with no significant differences from seawater-buffered melts, except for one flagellate group. Winter ice, which can contain sharp vertical gradients in brine salinity, varying from salinities similar to seawater (approximately 35 g kg-1) at the bottom to over 200 g kg-1 near the upper surface, requires particular care in melting. In a comparison of direct and brine melts for different sections of the winter ice column, Ewert et al. (2013) found that direct melts resulted in a significant loss of up to 55% of bacterial cells in the upper ice column but no difference in the lower sections, indicating that in situ brine salinity is a key factor in selecting the appropriate melting method. An important caveat to melting under refrigerated or buffered conditions is that it can take several days (Mikkelsen and Witkowski, 2010), during which ongoing biological processes can modify the sample and its community structure in ways that are still difficult to assess.
As a general recommendation, the choice of melt protocol should be based on the target measurement and the expected change in salinity resulting from the melting process. Although isohaline melts often appear to be the best approach to melting sea-ice samples for analyses of biological parameters, isohaline melting at low temperature is time consuming, which may introduce artifacts into the analyses (even while protecting against others). In addition, adding seawater or brines (either natural or artificial) can dilute or contaminate the sample, both in terms of organisms and analytes. Therefore, usually the concentrations of relevant analytes (such as macronutrients) need to be quantified in those seawater or brine solutions, which in turn must be pretreated to remove or inactivate contaminants (e.g., by filtration or UV oxidation). Finally, whereas the loss of a specific group of organisms can unacceptably bias the community composition, it may not necessarily impact total biomass determinations (POC, PON, bSiO2, Chl a) to a significant extent, particularly if the community is dominated by diatoms.
3.3. Biomass and community structure
3.3.1. Particulate organic matter and biominerals in sea ice
Particulate organic matter, which includes both living and non-living material, has been measured in sea-ice melts at concentrations up to 103 µmol C L-1 (e.g., Gradinger, 1999; Kennedy et al., 2002). Such maximum concentrations are up to two orders of magnitude higher than in the surface open ocean (Martiny et al., 2013), confirming that sea ice is an important pool of biogenic organic matter in the polar oceans.
The differentiation between the particulate and dissolved phases in sea-ice melts and brines is arbitrary and based on the filters used, a problem in all aquatic biogeochemical studies (see Hilmer and Bate, 1989; Knefelkamp et al., 2007, and Wang et al., 2007, for reviews of the best practices for filtration in aquatic science). At the molecular level, the distinction between dissolved, colloidal, and particulate material is physically ambiguous and variable, particularly at high concentrations, such as in sea-ice brines. Also, the filtration process, itself, can cause particulate organic matter to break apart, while dissolved macromolecules can adsorb onto the filters (e.g., Wangersky, 1993). This latter effect can be a particular problem for sea-ice samples with high concentrations of EPS that might clog the filters and reduce their effective pore size. Most but not all bacteria are captured by 0.2 µm pore-size filters; capturing viruses requires filters of even smaller pore size (0.02 µm; e.g.,Wells and Deming, 2006). In practice, many researchers rely on the convenience of glass and quartz fiber filters (e.g., GF/F, 0.7 µm nominal pore-size, pre-combusted at high temperatures to remove organic contamination), although they, of course, are not suitable for analyses of biogenic silica (see below).
After melting (section 3.2) and filtration, the C and N contents of the particulate organic matter (particulate organic carbon, POC, and nitrogen, PON) in sea-ice samples are generally analyzed by combustion to CO2 and N2, the standard method used in seawater (e.g., Ehrhardt and Koeve, 1999). Although studies of Chl a in sea ice have found no difference between cores melted with or without filtered seawater (section 3.2), to our knowledge, similar confirmation that melt procedures do not impact POC and PON analyses has not yet been published.
Measurements of cellular abundance in sea ice can be reported in different units for different purposes (Horner et al., 1992). For bulk macroscale analyses important in biogeochemical modelling, depth-integrated abundances (i.e., cells m-2) can be useful, as can further converting cellular abundance to biomass (typically in mmol C m-2), either by direct measurement of elemental composition or by using a published conversion factor (e.g., Miller et al., 2011b). For insight into the ecology of the organisms, results are sometimes scaled to the in situ brine volume (e.g., cells mL-1 brine; Junge et al., 2004; Wells and Deming, 2006; Collins et al., 2008).
Diatoms are the only ecologically significant group in sea ice producing biogenic silica (bSiO2), which is filtered from ice melts using polycarbonate membranes (Fripiat et al., 2007). Analysis generally follows a double/single wet-alkaline digestion method (Ragueneau et al., 2005) to also assess lithogenic contamination, which could be an issue in landfast sea ice. Biogenic calcium carbonate (CaCO3), in the form of foraminifera, has been observed and quantified by visual counting in Southern Ocean sea ice (Spindler and Dieckmann, 1986; Dieckmann et al., 1991b; Eicken et al., 1991; Thomas et al., 1998). The analysis of abiotic CaCO3 minerals is discussed in section 4.6.5.
3.3.2. Ice algal pigments and absorption spectra
Algal pigments provide both quantitative and qualitative information on the composition of the sea-ice community over a variety of temporal and spatial scales; Chl a, a ubiquitous pigment in algae and phytoplankton, is the most commonly used proxy of viable algal biomass in sea ice (e.g., Dieckmann et al., 1998; Meiners et al., 2012). Pigments are collected by filtering melted ice core sections in the dark using either GF/F or polycarbonate membrane filters; in general, GF/F filters capture nearly all of the algae and are compatible with standard pigment extraction methods (Mantoura et al., 1997; Roy et al., 2011). Investigators usually melt the ice in filtered seawater to prevent cell rupture (e.g., Garrison and Buck, 1986; Becquevort et al., 2009), although some have found no significant difference in Chl a concentrations between samples melted with and without filtered seawater (section 3.2). Exposure to light can affect the algal pigment composition; therefore, melting should always take place in the dark.
Chlorophyll a is usually determined by fluorometric detection following an acetone/methanol extraction (e.g., Arar and Collins, 1997; Gosselin et al., 1997). Many investigators apply a constant ratio to convert Chl a to carbon biomass, but in sea ice (as well as in other environments), this ratio is highly variable among photosynthetic organisms, as well as with changes in light, temperature, and nutrient concentrations (e.g., Arrigo et al., 2010). Therefore, although POC is a more expensive analysis than Chl a, sea-ice investigations benefit from parallel measurements of POC and Chl a.
High Performance Liquid Chromatography (HPLC) allows simultaneous analyses of a number of other pigments, in addition to Chl a. Standard methods are presented by Bidigare et al. (2005) and Roy et al. (2011). The large variations in light conditions throughout the sea-ice column and horizontal heterogeneity (section 2.1) strongly affect cellular pigment contents, so that special care is required in implementing the complex algorithms used to process HPLC data (Wright and Jeffrey, 2006; Latasa, 2007; Alou-Font et al., 2013). In particular, different layers of an ice core may have to be treated as separate community groups.
3.3.3. Bacteria and viruses
Bacteria and viruses contribute directly to the particulate nutrient and carbon pools in sea ice, participate in the still poorly understood sympagic food web, and mediate changes in the chemical composition of the ice brines. Knowing the abundance of bacteria and viruses is the first step in assimilating these microorganisms into conceptual and mathematical models of biogeochemical cycling in sea ice. Additional information on microbial metabolic rates (section 3.4.3) and diversity (sections 3.3.2 and 3.3.4) further facilitate the conceptual integration of biological and chemical perspectives on sea-ice processes.
Epifluorescence microscopy of melted ice samples (subject to the limitations and ambiguities discussed in section 3.2) is the principal technique by which bacterial and viral abundances are determined in sea ice. The methods have not changed substantially since they were first introduced in the late 1970s (e.g., Hobbie et al., 1977; Porter and Feig, 1980; Noble and Fuhrman, 1998; Kaartokallio et al., 2008). Challenges faced by users of this method in sea-ice samples include the difficulty of microscopic observations aboard moving vessels, high background fluorescence (attributed to non-specific staining or staining of nucleic acids in EPS), and extremely low or high cellular or viral abundances in some melted ice samples. Flow cytometry, with its high sample throughput, is commonly used for measurements of bacterial and viral abundance in marine systems and can also be useful in sea-ice biology (Riedel et al., 2007a; Kaartokallio et al., 2013; Piwosz et al., 2013), although high EPS concentrations could interfere with the analysis in some sea-ice samples.
3.3.4. Genetic community assessments
Since 2001, investigators have extracted environmental DNA from sea ice for various sequencing or prokaryotic community fingerprinting applications, enabling determination of prokaryotic community composition, structure, and metabolic potential within sea-ice samples (Table 2). In a few cases, RNA has been extracted with DNA to determine the structure and function of the active prokaryotic community.
aPC = phenol chloroform; DNeasy and RNeasy available from Qiagen.
bFYI = first year sea ice, MYI = multiyear sea ice, YI = young sea ice, FF = frost flowers.
Samples are often size-fractionated with filters of different pore sizes to separate major components of the community before nucleic acid extraction. To facilitate intercomparison with the GOS (Global Ocean Sampling Expedition) dataset, we recommend standard pore sizes of 0.1, 0.8, and 3.0 µm (Rusch et al., 2007), although there is little evidence that these cutoffs correspond to natural ecological boundaries in sea ice. For accurate DNA extraction from cells, free from background contamination, it is essential that cells not be lysed until after capture on the filter, making the sample melting conditions critical (section 3.2). Low biomass samples, such as winter sea ice and some multi-year sea ice, require large melt volumes (greater than 1 L), while higher biomass samples can have an overabundance of eukaryotic DNA, along with interfering compounds such as EPS and non-specific humics. Because these compounds are chemically similar to nucleic acids, they often co-extract and interfere with downstream applications, such as the polymerase chain reaction (e.g., Tebbe and Vahen, 1993). An ideal nucleic acid extraction protocol for sea ice would produce a high yield of the target molecule, reducing the bias toward or against any member of the microbial community and minimizing the co-extraction of EPS, humics, and other interfering compounds.
The available extraction methods fall into three broad categories: phenol chloroform (PC), kit-based, and electrophoretic methods. Although electrophoretic extraction is a promising new technology that may overcome some of the challenges regarding biomass and interfering compounds (e.g., So et al., 2010), to date, only PC and kit-based methods have been applied to sea-ice samples. Cowie (2011) evaluated PC and kit-based methods in Antarctic sea ice and found that PC (using the methods of Moeseneder et al., 2001) is suitable for sea-ice samples. The same study also found that the RNA extraction kit RNeasy (Qiagen) in combination with bead beating was the most effective RNA extraction method, when samples were preserved with RNAlater (Qiagen). Although this and other studies show that nucleic acids can be extracted from sea ice, we encourage more comprehensive intercomparisons. In the meantime, investigators should carefully test their methods on the specific ice types they are studying.
3.4 Metabolic processes
Numerous studies have attempted to adapt methods for determining community metabolic rates in aquatic systems to the sea-ice environment (Table 3). None of these methods has been entirely satisfactory. Intercalibration experiments and further method development are high priorities in sea-ice biogeochemistry.
aPhysical biases include bubble formation, sea ice-atmosphere exchange, and solubility changes.
bRate depends on the incubation time, with shorter incubation times more closely approximating gross primary production (e.g., Laws et al., 2002).
cSatoh and Watanabe (1988) incubated algae scraped off the bottom of ice cores, but the method should be applicable to any depth in the ice core.
3.4.1. Primary production and elemental uptake rates
The term “primary production” largely refers to organic matter synthesis by photosynthetic organisms, harvesting light to convert inorganic to organic carbon. The conversion of inorganic to organic carbon by chemosynthetic microorganisms in sea ice, in particular by nitrifying bacteria (which are also chemosynthetic), has been implicated in studies using stable isotopes (e.g., Fripiat et al., 2014a) and DNA sequencing (Barber et al., 2014), but chemosynthesis is generally considered a minor contribution to overall primary production in sea ice. Several methods exist to directly estimate photosynthesis-based primary production (gross and net) in aquatic systems (e.g., Falkowski and Raven, 2007); each method has its own assumptions, ambiguities, and biases, which have been extensively discussed in the oceanographic literature (e.g., Bender et al., 1987; Laws et al., 2002). However, the complexity of the sea-ice/brine matrix presents particular problems in quantifying metabolic rates.
Incubations for determining primary production in sea ice, usually based on the incorporation of a tracer into particulate organic matter over a known amount of time, can be conducted either in vitro, in refrigerated incubators with spectral filters to mimic natural light conditions, or in situ, by embedding inoculated samples back into the sea-ice environment (e.g., Horner and Schrader, 1982; Mock and Gradinger, 1999; McMinn and Hegseth, 2007; Gradinger, 2009). In vitro incubations remove the in situ variability, allowing easier comparison between different experiments, and are still the best method to determine maximum photosynthetic rates and efficiencies as functions of light intensity (e.g., Burkholder and Mandelli, 1965; Arrigo and Sullivan, 1992). By combining these parameters with measurements of Chl a and light in the field, primary production can be derived (Arrigo et al., 2010). On the other hand, in situ incubations more directly assess primary production under the specific conditions to which the natural community is exposed.
For an incubation to be informative, the sample needs to represent the in situ community and environment as well as possible. Although brines collected from sackholes or by centrifugation are convenient to handle and provide a solution that represents the environment experienced by sea-ice algae, the biomass collected with such brines is not representative of the sea-ice community (section 2.2.2). On the other hand, while bulk sea-ice melts seem to provide representative biomass of some taxonomic groups, the dramatic changes in temperature and salinity associated with melting, even if the ice samples are melted in seawater, destroy the natural habitat, as well as often rupturing cells (particularly the flagellated taxa; section 3.2). In addition, the melting process at low temperature can take some time (often more than a day), further alienating the community from in situ conditions. The cost-benefit balance of the length of time required for melting (longest for isohaline melts if also isothermal; Junge et al., 2004) versus the modification of the samples (greatest for direct melts) is unknown.
Some investigators have incubated whole ice sections in closed containers filled with seawater bearing the relevant isotopic tracer either in vitro (e.g., Grossi et al., 1987) or in situ by replacing the inoculated ice samples, sealed in transparent containers, into the core holes from which the ice had been extracted (Figure 9; Mock and Gradinger, 1999; Mock, 2002). These methods using unmelted ice sections have the advantage of maintaining a representative sample, but questions remain as to whether the tracer is adequately distributed within the brine network. If the tracer preferentially remains in the surrounding seawater or brines, primary production could be severely underestimated. Crushing the ice (Rysgaard et al., 2007) should improve homogenization.
Changes in O2 concentration can be used to estimate net community production (NCP), defined as primary production minus respiration. This method has been applied to sea-ice communities using in situ oxygen optodes (Mock et al., 2003) and microelectrodes (McMinn and Ashworth, 1998; McMinn et al., 2000; 2007; McMinn and Hegseth, 2007). Determining NCP solely from O2 dynamics in sea ice is complicated by exchange with the atmosphere arising from solubility changes, as temperature and salinity vary with ice formation and melt (Glud et al., 2002). In seawater, Ar measurements are used to correct for such physical contributions to the O2 concentration changes (e.g., Craig and Hayward, 1987); Zhou et al. (2014b) have used O2/Ar ratio measurements to estimate NCP in winter and spring sea ice.
With recent technological advances, underwater eddy covariance (EC) has potential for investigating sea-ice primary production (Long et al., 2012). The technique (which is analogous to the atmospheric technique described in section 5.1.3) requires the under-ice deployment of a three-dimensional current velocimeter in conjunction with a fast-response oxygen electrode (Figure 10). The instruments can be deployed through a hole in the ice and are usually positioned within 1 m of the ice-seawater interface. By correlating oxygen concentrations with vertical current velocity, oxygen fluxes representative of conditions in an area upstream of the instruments (typically on the order of 100 m2) can be calculated. The technique is thus a true in situ measurement, it does not disturb the ice under investigation, and it avoids some of the small-scale patchiness issues associated with other techniques. However, at best, EC only provides a measurement of net productivity (as respiration cannot be separated from photosynthesis). At worst, EC may be strongly influenced by abiotic processes, such as gas rejection (during ice growth) and dilution (during melt), which need to be accounted for in certain cases. The technique is also only useful for measuring production in the bottom-most layers of the ice (i.e., those that are freely exchanging with the underlying seawater).
Incubations with other isotopic tracers, such as 15N (Harrison et al., 1990; Kristiansen et al., 1992; 1998), can be used to assess the nature of primary production (e.g., new versus regenerated; Dugdale and Goering, 1967) and concomitant biogeochemical dynamics (such as N-uptake and nitrification). Because diatoms are important members of sea-ice communities, the information on Si uptake and bSiO2 dissolution that could be generated by adapting methods for incubations with 30Si and 32Si (Fripiat et al., 2009) to sea-ice samples would be particularly valuable. Such incubations depend on many of the same assumptions as carbon-based primary production incubations and, therefore, suffer from the same challenges and limitations. In addition, because these macronutrients occur in sea-ice brines at much smaller concentrations than inorganic carbon, labelled N and Si substrates are added to the incubations in ‘trace’ quantities (generally, 10% of the ambient level is recommended; Dudgale and Goering, 1967).
3.4.2. Variable fluorescence methods to determine ice algal photosynthetic parameters
Over the last 20 years, chlorophyll a variable fluorescence methods have proven to be useful tools for understanding photophysiological properties of marine algae. The application of pulse amplitude modulated (PAM) fluorometry to study sea-ice algae was pioneered by Kühl et al. (2001) in the Arctic and McMinn et al. (2003) in Antarctica. This approach measures the energy conversion efficiency of photosystem II (the quantum efficiency) to derive maximum relative electron transport rates and estimate photosynthetic efficiency and the photoadaptive index of the algae (Ralph and Gademann, 2005). In general, because PAM fluorometers do not provide a direct measurement of carbon fixation, the overall value in PAM fluorescence techniques lies more in their ability to measure ice algal photophysiological responses to varying physicochemical conditions, particularly on small scales relevant to the sea-ice skeletal layer and brine channels (e.g., Hawes et al., 2012).
The most widely used PAM methods in sea-ice research involve ice shavings, brines, and melted ice samples that are analyzed in vitro (McMinn et al., 2007; Ralph et al., 2007; Manes and Gradinger, 2009; Meiners et al., 2009; Hawes et al., 2012; Granfors et al., 2013a); these methods are thus biased by sea-ice sampling procedures (sections 2.2 and 3.2). To bypass sample extraction artifacts, divers have successfully deployed non-invasive PAM fluorometers under sea ice to measure the spatial variability of bottom ice algal biomass and algal photosynthetic parameters in Greenland fast ice (Kühl et al., 2001). Detailed studies utilizing PAM techniques have also confirmed vertical variability in ice algal distributions and photosynthetic properties (Manes and Gradinger, 2009; Hawes et al., 2012).
A second type of variable fluorescence instruments, so-called Fast Repetition Rate (FRR) fluorometers, are increasingly used in phytoplankton research but have not yet been employed widely in sea-ice research. In contrast to PAM fluorometers, FRR fluorometers provide measurements of quantum efficiency, absorption cross section, and turnover times of photosystem II and can, therefore, be used to estimate primary production (Robinson et al., 1998; Suggett et al., 2003).
3.4.3. Bacterial production
Marine bacterial biomass production is usually estimated from incubations with radioisotope-labelled precursors of DNA or proteins (e.g., Ducklow, 2000). As with primary production measurements (section 3.4.1), incubations of melted ice samples are of limited utility, and methods that attempt to maintain the in situ conditions during the incubations are preferred. Tritiated-thymidine incorporation (TTI) into DNA (e.g., Smith and Clement, 1990; Deming, 2010) and 3H-leucine (LEU) incorporation into protein (e.g., Kaartokallio, 2004; Paterson and Laybourn-Parry, 2012) are the most common techniques applied to sea-ice samples. While TTI and LEU incorporation often co-vary in the marine environment, suggesting that both methods address bacterial production-related processes, the two methods measure distinctly separate physiological processes; a range of ratios of LEU incorporation to TTI has been reported, with very high ratios in some sea-ice habitats (Mock et al., 1997; Kaartokallio et al., 2008; 2013). High ratios have been attributed to unbalanced growth (e.g., investment in cell growth, measured by LEU incorporation, versus cell division, measured by TTI), but also to incorporation of the radiolabeled LEU tracer into other, non-protein components of sea-ice organisms. Bacterial activity in sea ice has also been estimated from transformations of 15N-labelled nitrogen substrates (Rysgaard and Glud, 2004) and from uptake and reduction of the dye 5-cyano-2,3-ditolyl tetrazolium chloride (CTC; Junge et al., 2004; Meiners et al., 2008).
Estimating bacterial biomass production from the uptake of labelled substrates requires tracer-to-carbon conversion factors, but such conversion factors are not easy to measure routinely; conversion factor determination usually involves incubating natural samples for several days. Therefore, literature values from open-water studies are often applied to sea-ice data, although the accuracy of the conversion factor is strongly affected by community composition and physiology. In order to accurately estimate sea-ice bacterial carbon production, conversion factors for both the LEU incorporation and TTI methods need to be determined for a variety of sea-ice communities in different habitats and regions. In addition, bacterial processes can be particularly sensitive to temperature (e.g., Rivkin and Legendre, 2001); the impact of incubation temperature on measured metabolic rates specifically for sympagic organisms warrants further investigation.
4. Chemical components
Methods for analyzing the major seawater ions in sea ice, its overlying snow, and frost flowers are well established and uncontroversial (e.g., Nelson and Thompson, 1954; Domine et al., 2004; Granskog et al., 2004; Douglas et al., 2012). Modern methods almost universally utilize ion chromatography. However, the analyses of nearly all other chemical parameters in sea ice are still subject to debate. Our focus here is on those methods in which we have less confidence.
Our lack of confidence in most chemical analyses in sea ice stems from difficulties in sampling, handling, and processing the samples, and rarely from the actual analyses of the aqueous samples. An important exception is analyses of undiluted brine samples, in which the high salinities often reduce precision and complicate calibrations.
4.1. Inorganic macronutrients
Fundamental information on the nutritional state of the sympagic biological community can be derived from the distributions of the inorganic macronutrients (phosphate, PO43-; silicic acid, Si(OH)4; and the nitrogen species: nitrate, NO3-; nitrite, NO2-; and ammonium, NH4+) in sea ice (e.g., Dieckmann et al., 1992; Kaartokallio, 2001). Variations in their concentrations throughout the ice column and with time indicate biological activity and exchange with the underlying water. Most sea-ice nutrient analyses are based on spectrophotometric analysis, using a continuous flow analyzer (e.g., Hydes et al., 2010). A large range of nutrient concentrations occurs in sea ice (Thomas et al., 2010), from depleted to replete conditions, and sampling requires the usual precautions against contamination. To the extent possible, equipment preparation and analyses should follow standard repeat hydrography protocols (e.g., Granskog et al., 2005b; Hydes et al., 2010).
Nutrients are often measured in melted bulk sea-ice samples, but brines for nutrient analyses have also been collected both by centrifuging ice samples (e.g., McMinn et al., 2009; Munro et al., 2010) and from sackholes (e.g., Gleitz et al., 1995; Papadimitriou et al., 2007). The costs versus benefits of filtering ice melts before nutrient analysis are unclear. Particulate (intracellular) nutrient concentrations can be quite high in sea ice, because of high biomass accumulation, particularly near the ice-water interface (e.g., Arrigo et al., 2010). On the other hand, particle-bound nutrients can be released to solution during melting (section 3.2), although Thomas et al. (1998) found no evidence during their spring-time study that cell lysis significantly impacted dissolved nutrient measurements of bulk ice samples melted without adding seawater (direct melt). Large cells generally would have higher intracellular nutrient contents to release upon lysis than small cells, but large diatoms, which often dominate sea-ice communities, appear to have reduced susceptibility to lysis during melting (section 3.2; Mikkelsen and Witkowski, 2010), minimizing the effect of cell lysis on nutrient concentration estimates for diatom-rich ice. Additional intercalibration exercises are needed to more thoroughly assess the effects of core melting protocols and filtration on measured nutrient concentrations with different microbial communities and in different seasons.
Ideally, nutrient samples, particularly those for ammonium, should be analyzed immediately after sampling to avoid artifacts associated with biological growth or decay during sample storage (Holmes et al., 1999; Hydes et al., 2010). When analysis cannot be completed within hours, the samples generally should be stored frozen (< -20°C) in the dark. However, silicic acid in seawater polymerizes when it freezes and only very slowly redissolves when the sample is thawed for analysis, possibly resulting in underestimation of the original dissolved silicic acid concentration (e.g., Hydes et al., 2010). The impacts of this phenomenon on estimates of in situ concentrations of biologically available silicic acid in sea-ice brines is unknown. Therefore, filtration to remove Si-bearing diatoms and storage at room temperature in the dark are more suitable procedures for silicic acid analyses of sea-ice samples (Fripiat et al., 2014b). Alternatively, filtered samples for nutrient analyses can be poisoned with agents such as mercuric chloride and stored refrigerated but unfrozen until analysis (e.g., Kattner, 1999).
4.2. Organic compounds
Sea ice acquires organic matter from seawater during ice formation and through in situ biological production (Thomas et al., 1998; 2001a; Giannelli et al., 2001; Granskog et al., 2004; Riedel et al., 2007a; Stedmon et al., 2007; 2011; Müller et al., 2013). Extremely high concentrations of dissolved organic carbon, several thousand µmol L-1 (up to two orders of magnitude higher than seawater values; Hansell et al., 2009), have been reported in sea-ice samples (e.g., Thomas et al., 2001a; Junge et al., 2004; Riedel et al., 2008). The resulting sea-ice organic pool is a complex mixture of living and non-living particulate material and dissolved compounds. Operational definitions of dissolved, colloidal, and particulate carbon are particularly tenuous in sea ice, as the temperature and salinity changes associated with melting can cause phase changes in the organic matter. In addition, we do not know how well sampled brines represent the organic content of brine pockets and channels within the undisturbed ice; even if not rigorously particulate, the colloidal and dissolved organic matter may still be “sticky” and adhere to brine channel walls.
In most cases, concentrations should be measured and reported in units of moles of carbon (or nitrogen or phosphorous) and not grams of bulk organic matter. This approach allows the data to be used in biogeochemical cycling studies, including incorporation into numerical models, without unverified assumptions about carbon content and the C:N:P ratios of the organic matter.
Most studies of organic biogeochemistry in sea ice to date have focused on the bulk parameters of total, dissolved, and particulate organic matter. Methods for determining the particulate fractions are detailed in section 3.3.1. Here we address sample collection for the total (unfiltered) and dissolved (filtered) aqueous fractions. Photochemically active colored dissolved organic matter (CDOM) is discussed in section 4.7; halocarbons are discussed in section 5.2.
4.2.1. Sample handling
Organic matter concentrations in sea ice vary over an order of magnitude, from low values similar to those observed in the deep ocean to high values in spring brines and gap layers (e.g., Thomas et al., 1995; Song et al., 2011). When dissolved organic matter (DOM) is present in high concentrations in sea-ice samples, some of the stringent protocols required to successfully sample seawater for DOM are eased. In low-DOM sea-ice samples, however, care is required to avoid not only environmental contamination during sampling and processing, but also cross-contamination between samples during both processing and analyses. Therefore, the exact requirements for preventing contamination of sea-ice DOM samples have not yet been established and probably depend on the samples; that is, upper levels of multiyear ice probably need to be handled more rigorously than gap-layer slush samples. To eliminate potential contamination during core extraction and manipulation, the outer layers of cores can be removed before subsampling for organic matter (Granskog et al., 2004; Song et al., 2011; Fripiat et al., 2014a). In addition, although glass equipment that has been combusted at high temperatures is preferred, plastics are being used more often and may be acceptable under some conditions (e.g., Thomas et al., 1998; Miller et al., 2011b), particularly if contact times are kept short. Rigorous experiments are still required to confirm the conditions under which plastics can be used.
The question of whether or not to filter samples for organic analysis can only be answered within the context of each specific study. Particularly in concentrated brines, the physical and behavioral distinction between “dissolved” and “particulate” organic matter may have little relationship with the operational definition based on the type of filter used. Filtration also introduces artifacts, as dissolved organic matter can stick to filters and turbulence or pressure gradients associated with the filtration process can break apart particles (section 3.3.1).
Furthermore, the salinity changes associated with melting ice can cause coagulation and disaggregation of organic matter, as well as cell lysis (section 3.2). On the other hand, while melting the ice in artificial or filtered seawater will minimize artifacts associated with salinity changes, it is almost impossible to generate truly “organic-free” seawater and thus avoid contributing to the analysis in ways that are difficult to quantify.
4.2.2. Total organic carbon and nitrogen (TOC and TON)
The generic, undifferentiated pool of organic matter has been analyzed in melted bulk ice (e.g., Thomas et al., 2001a; Cozzi, 2008; Dumont et al., 2009) and in brines collected from sackholes (e.g., Papadimitriou et al., 2007; Meiners et al., 2009). In parallel to seawater methods, the samples are often passed through precombusted glass fiber filters, in which case the resulting analytical results are termed “dissolved” organic carbon or nitrogen (DOC, DON), although this pool also includes some bacteria and viruses (Deming, 2010).
High temperature catalytic oxidation is generally used for TOC and DOC analyses (Qian and Mopper, 1996; Spyres et al., 2000), while DON is often inferred from the difference between total dissolved nitrogen (TDN; Bronk et al., 2000) and the inorganic nitrogen species (NO3-, NO2-, NH4+; section 4.1). Sea-ice studies have generally measured TDN using UV oxidation (Thomas et al., 2001a; Papadimitriou et al., 2007; Cozzi, 2008), although other methods may also be suitable (e.g., chemical oxidation with persulfate; Bronk et al., 2000; Fripiat et al., 2014a). Less common are studies of urea in sea ice (Harrison et al., 1990; Kristiansen et al., 1998; Conover et al., 1999; Garrison et al., 2003; Papadimitriou et al., 2009), generally measured using the urease or diacetyl monoxime methods (e.g., Price and Harrison, 1987).
4.2.3. Exopolymeric substances (EPS)
Over the past decade, extracellular or exopolymeric substances (EPS) have been recognized as extremely important components of sea ice. This high-C, low-N material, similar to the transparent exopolymeric particles (TEP) found in seawater, may act as a cryoprotectant for sympagic biota and has a measurable effect on the microstructure and salinity of sea ice (Krembs et al., 2011). Also present in surface sea-ice environments, such as frost flowers and snow, EPS has additional implications for air-ice interactions (Bowman and Deming, 2010; Ewert et al., 2013).
Ice core samples for EPS analysis have been melted (section 3.2) directly (e.g., Krembs et al., 2002; Juhl et al., 2011), in seawater (e.g., Meiners et al., 2003; Riedel et al., 2006), and in concentrated brines (e.g., Collins et al., 2008; Ewert et al., 2013). Direct melts are convenient, because lower salinities simplify further chemical analyses, but additional studies are required to confirm whether the melting protocol has an effect on measured EPS content or on the physical and chemical properties of this complex material.
After melting, EPS is separated into dissolved (dEPS) and particulate (pEPS) fractions by filtration through different types of filters (Table 4). Additional size fractionation can be achieved with sequential precipitation of dEPS fractions of varying solubility across an ethanol gradient (Underwood et al. 2010; 2013; Aslam et al., 2012). Three methods are commonly used to quantify particulate and dissolved EPS in sea-ice research (Table 4): the standard colorimetric Alcian blue method developed for TEP analysis of seawater (Passow, 2002), the colorimetric TPTZ (2,4,6-tripyridyl-s-triazine) method of Myklestad et al. (1997), and the phenol-sulfuric acid assay (PSA) of Dubois et al. (1956), modified for small sample volumes. In a direct comparison, van der Merwe et al. (2009) found that results from analyses of Antarctic sea ice using the Alcian blue and PSA methods agreed at high but not at low EPS concentrations near the detection limits. More such intercomparisons are needed to confirm the validity and comparability of results from the Alcian blue, TPTZ, and PSA methods under varying conditions. Some investigators have quantified specific components of EPS using a carbazole assay for uronic acids (Bitter and Muir, 1962) and hydrolysis followed by gas chromatography for the neutral monosaccharide composition (Table 4). Methods need to be developed to more thoroughly characterize EPS in sea ice, including: polymer composition, structure and molecular size in different EPS types; interactions between EPS and other components of the sea-ice biogeochemical system (salts, trace metals, other forms of organic matter); and the potential for EPS to interfere with other chemical analyses.
Filters stained with Alcian blue can also be analyzed microscopically to determine particle abundance, size distribution, and the number of EPS-associated bacteria (Meiners et al., 2003; 2004; 2008). Krembs et al. (2002; 2011) also used in situ visualization of EPS distribution within brine pockets and channels to directly examine the association between EPS and ice biota.
4.2.4. Specific organic compounds
Very few studies have attempted to measure specific organic classes or compounds, either natural or anthropogenic, in sea ice. The steps required to limit contamination, gas exchange, or brine loss strongly depend on the nature of the compounds of interest: their concentration ranges, sources, volatility, and particle affinity.
Herborg et al. (2001) and Dumont et al. (2009) distinguished between the mono- and polycarbohydrate fractions of the DOC pool in sea ice. Belt et al. (2013) measured the lipid paleo-biomarker IP25, sterols, and fatty acids in filtered bulk sea-ice melts. Mycosporine-like amino acids (MAAs), which may serve as photoprotectants under some circumstances, have been determined in both Arctic and Antarctic sea ice (Ryan et al., 2002; Uusikivi et al., 2010; Mundy et al., 2011). Stedmon et al. (2007; 2011) and Granskog et al. (2015) used fluorescence to quantify humics and “amino acid-like” organic matter in sea-ice melts and brines. A number of organic anthropogenic contaminants have been measured in sea-ice melts (Rahm et al., 1995; Pućko et al., 2010a; b), in brines (Pućko et al., 2010b), and in snow (Garbarino et al., 2002), frost flowers, and brine skims (Douglas et al., 2012) over sea ice.
4.3. Trace metals
The seasonal ice cover represents a key reservoir, storing and transporting potentially bio-active trace metals and thus likely playing major roles in not only trace metal cycles but also the carbon cycle in polar and subpolar oceans (e.g., Lannuzel et al., 2010). However, trace metal research in sea ice is subject to the same draconian restrictions required in seawater to avoid contamination (Figure 11; e.g., Bruland and Rue, 2001; Cutter et al., 2010). In general, all equipment, preparations, and analyses should follow standard GEOTRACES protocols (Cutter et al., 2010). Several studies have successfully developed sampling and measurement techniques for trace metals in the cryospheric environment and have reported data for numerous elements in sea ice, snow, and brines in the Arctic and Southern Oceans (Table 5).
aMethods not specified
Samples for trace metal analyses need to be recovered from a dedicated sampling site, upwind from all other operations, and specific sampling procedures must be performed against the wind. Clean-room garments and plastic gloves should be worn over cold-weather clothing. While trace metal-clean sampling is relatively straightforward for snow above sea ice (e.g., Edwards and Sedwick, 2001; Lannuzel et al., 2006), ice core sampling carries a greater risk of contamination. Ideally, a titanium or an electropolished stainless steel corer (Lannuzel et al., 2006) should be used, although standard corers can also be used, if the outer layer of the core can be removed without contamination or substantial brine loss (Figure 11b; Hölemann et al., 1999; Granskog and Kaartokallio, 2004; Granskog et al., 2004; Grotti et al., 2005; Aguilar-Islas et al., 2008). The ice should be cored by hand, although electric auger motors have also been used with the generators downwind (e.g., de Jong et al., 2013). A number of investigators have collected brine samples from sackholes for trace metal analyses (Lannuzel et al., 2006; 2007; van der Merwe et al., 2009; 2011a; Chaulk et al., 2011; Cossa et al., 2011).
Changes to the in situ chemical speciation and fractionation between oxidation states and particulate/colloidal/soluble phases during sample collection and processing are exceptionally problematic in sea-ice studies of trace metals. In particular, iron speciation is very poorly understood in sea ice, and the definitions of what is actually measured are highly operational (e.g., Bruland and Rue, 2001). To date, trace metal speciation measurements in sea ice and brines have been limited mainly to separations between operationally-defined particulate and dissolved fractions, separated by filtration and varying dissolution procedures (e.g., Grotti et al., 2005; Lannuzel et al., 2006; 2007; 2011; van der Merwe et al., 2011a; Hendry et al., 2010a). Boye et al. (2001) analyzed iron-organic complexation in one sea-ice sample and confirmed that a large fraction of the iron in sea ice can be complexed by organic matter, indicating that organic complexation may be as important in sea ice as in seawater.
Mercury is another special case, involving not only a highly contamination-prone metal in solution, but also gaseous and organic phases. Most studies of Hg in the marine cryosphere have been focused above the ice, analyzing snow, frost flower, or surface brine skim samples (Garbarino et al., 2002; Douglas et al., 2005; 2008; Poulain et al., 2007; Sherman et al., 2012). The analyses utilize either standard cold vapor atomic fluorescence spectroscopy (EPA, 2002) or atomic absorption spectroscopy methods. Chaulk et al. (2011) and Cossa et al. (2011) measured total Hg and dissolved Hg chemical speciation, respectively, within sea ice. Sherman et al. (2012) also used stable mercury isotopes (Δ199Hg, analyzed by inductively coupled plasma mass spectrometry) to investigate air-ice mercury fluxes.
With the realization that sea ice is porous comes an understanding that it could serve as a source or sink of climatically active gases. Most of the gases measured in sea ice, to date, have been found at relatively high concentrations; in general, the precision of the analyses has been a greater challenge than detection limits. The most obvious difficulty in sampling sea ice for gases is the potential for exchange with the air. In most cases, gases are lost from the ice to the atmosphere, but ice samples can also be contaminated by contact with air, particularly by pollutant volatile organics and hydrocarbons. In addition, gases occur in sea ice both as solutes and as bubbles; some methods will extract both fractions, while others collect only the fraction dissolved in the brines. In this section, we discuss general concerns with analyzing gases in sea ice and summarize the specific methods used to date (Table 6). The details of measuring dimethylsulfide and carbon dioxide are discussed in respective sections 4.5 and 4.6, elemental mercury is covered in section 4.3, halocarbons are discussed in section 5.2, and studies of other volatile organic compounds are included in section 4.2.4.
aGC: Gas Chromatography with suitable detectors
bNDIR: Non-dispersive infrared spectroscopy
cMethods not specified
dDetermined total DMS+DMSP
eProton-transfer-reaction mass spectrometry
The oldest methods for extracting gases from sea ice are wet extractions, involving sequential melting and refreezing of ice samples (Matsuo and Miyake, 1966). For insoluble gases at high concentrations in the ice or for which analyses with low detection limits exist, the refreezing step is not required. With knowledge of the gas partitioning coefficient between the aqueous melt and air, the initial melt can simply be equilibrated with a volume of ambient air, which is then analyzed. This approach has been used successfully for analyses of total gas content (Tison et al., 2002), CO (Xie and Gosselin, 2005; Song et al., 2011), N2O (Kelley and Gosink, 1979; Randall et al., 2012), methane (Zhou et al., 2014a), and organohalides (section 5.2).
Dry-extraction (or dry crushing), involving crushing an ice sample with steel balls in a vacuum chamber, has proven effective for determining O2, N2 (Tison et al., 2002), and Ar (Zhou et al., 2013) in sea ice. The size of the crushed ice sample mainly depends on the concentration of the target gas in the ice and the sensitivity of the gas chromatography detector. For trace gases (such as DMS, section 4.5), the gas may need to be preconcentrated before injection into the gas chromatograph. Crushing the ice has one intrinsic problem: contamination by methane released by the metal-metal friction between the stainless steel balls and the container during the crushing process (Higaki et al., 2006). Therefore, other extraction methods are used for analyses of carbon-containing species.
Sea-ice brines for gas analyses are usually collected from sackholes, although Nomura et al. (2009) used full-core gravity drainage (Figure 4c) to collect brines for O2 analysis. In general, extracting brines for gas analyses is only satisfactory for relatively soluble gases, such as O2 and CO2, and even so, only if temperatures are high enough for the brines to accumulate quickly and the sackhole is capped (Papadimitriou et al., 2007). In contrast, the bulk of the insoluble gases in sea ice is probably located in bubbles within brine pockets and channels; when only the brines are analyzed, such gas bubbles within the ice are lost. Zhou et al. (2014a) found that brine CH4 concentrations deduced from measurements in bulk ice can be up to 10 times higher than the concentrations directly measured in brine samples, a difference almost certainly due to exchange with the atmosphere during brine percolation.
In situ probes hold great potential for determining gas concentrations in sea ice (McMinn et al., 2009). Probes based on photochemical detection (i.e., optodes) are particularly promising for sea-ice applications; oxygen has been measured successfully in sea ice using commercially available optodes (Figure 12a; Mock et al., 2002; Rysgaard et al., 2008), although reaction times are slow and calibrations are potentially complicated by both high salinities and organic matter concentrations. Another potential drawback of microprobe measurements is uncertainty in the specific microenvironment sampled. For example, an oxygen optode might only sense the liquid phase, missing the (often dominant) gas phase. In addition, any in situ probe will change the thermodynamic environment of the ice to at least some extent; particularly under sunny conditions, objects frozen into the ice absorb heat, causing localized excess melt. Gas-permeable silicone chambers developed by the soil science community have been deployed for measuring in situ CO2 mole fractions in sea ice (Figure 12b; section 4.6.4; Miller et al., 2011a;b) and theoretically could be used for analyses of other gases, including O2, CH4, and N2O (e.g., Holter, 1990; Kammann et al., 2001). Electrochemical probes are generally unsuitable for sea-ice applications, because their internal electrolyte solutions freeze, but have potential for further development. The primary drawbacks to in situ sensors, in general, are that the study site must be occupied for an extended period and, unless the sensors can be deployed at freeze-up, their installation requires disturbing the ice cover.
4.5. Sulfur species
Dimethylsulfide (DMS) is one of the most abundant volatile sulfur compounds in the ocean and accounts for more than half of the global biogenic sulfur flux to the atmosphere (e.g., Liss et al., 1997). Sea ice usually contains larger amounts of DMS and its precursor dimethylsulfonioproprionate (DMSP) than does the under-ice water (up to two orders of magnitude higher), although the amounts in sea ice are highly variable (e.g., Trevena and Jones, 2006). Therefore, seasonal ice melting introduces elevated DMS concentrations to surface waters from the release of sea-ice DMS and DMSP (e.g., Tison et al., 2010). The sulfur compounds studied in sea ice to date are DMS, DMSP (which occurs in both particulate and dissolved fractions), and dimethylsulfoxide (DMSO). Other sulfur species important in the sulfur cycle that have not yet been investigated in sea ice include sulfur dioxide (SO2), hydrogen sulfide (H2S), carbonyl sulfide (COS), and carbon disulfide (CS2).
Total and dissolved DMSP have been measured in ice brines recovered from sackholes (Trevena and Jones, 2006; Asher et al., 2011), although it is not clear how well dissolved DMSP measurements from brines distinguish the in situ partitioning between dissolved and particulate fractions within the brine network of undisturbed ice (sections 2.2.2, 4.2.1). Total DMSP is often analyzed in bulk ice melts (Levasseur et al., 1994; Curran and Jones, 2000; Trevena et al., 2000; Trevena and Jones, 2006), whether obtained by melting in filtered seawater (Levasseur et al., 1994), in acidified filtered seawater (Trevena et al., 2000; Trevena and Jones, 2006), or in concentrated brine (Stefels et al., 2012). However, in a direct comparison of DMSP analyses from twin cores, one melted in concentrated filtered seawater and one dry-crushed (section 4.4), Stefels et al. (2012) found that large amounts of DMSP can be converted to DMS during the melting process.
Because of its very low solubility, accurately measuring DMS concentrations in ice samples has been a challenge. Although brines have been analyzed for DMS (Table 6), it is generally considered too insoluble to be recovered confidently from brines collected from sackholes. Trevena and Jones (2006) melted ice samples directly into acid within purge chambers. Small ice samples have also been directly crushed to recover and analyze DMS (Tison et al., 2010; Stefels et al., 2012). Fluxes of DMS out of sea ice have been estimated using both chamber (section 5.1.1; Nomura et al., 2012) and micrometeorlogical (section 5.1.4; Zemmelink et al., 2008) techniques. The analyses generally use gas chromatography with either flame photometric or mass spectrometric detection.
Analyses of DMSO in sea ice are still scarce and complicated by the need to eliminate interference from DMSP (Brabant et al., 2011). Lee et al. (2001) measured DMSO associated with the particulate (algal) fraction in sea ice, Brabant et al. (2011) determined total DMSO in bulk ice, and Asher et al. (2011) measured dissolved DMSO in sea-ice brines. All of these studies analyzed DMSO as DMS, after purging and reduction.
The rates at which sulfur species are formed and degraded is one of the largest uncertainties in the global sulfur cycle (e.g., Ayers and Cainey, 2007); in only one study have natural rates of interconversion between the various sulfur compounds in sea ice brines been determined directly (using short-term incubations with 2H- and 13C-labelled DMS, DMSP, and DMSO tracers; Asher et al., 2011). Stefels et al. (2012) described a method for spiking samples with deuterated DMS and DMSP before melting, in order to document and then correct for DMSP degradation. Although used to examine sample storage and processing artifacts (Stefels et al., 2012), this tracer method also has potential for investigating conversion rates by the in situ community. As for primary production and bacterial production (sections 3.4.1 and 3.4.3), a satisfactory method for determining sulfur cycling rates in undisturbed ice environments has not been reported, making continued method development a high priority. In general, we encourage immediate processing and analysis of sea-ice samples for sulfur cycle studies to limit interconversion between sulfur species and loss of insoluble DMS.
4.6. The carbon dioxide system
The inorganic carbon system in sea ice is controlled by complex biogeochemical processes that transform carbon between phases (gas bubbles, brine, and particles) and between inorganic and organic forms, while also exchanging with the atmosphere and the underlying water (e.g., Thomas et al., 2010; Loose et al., 2011). A number of methods have been used to collect and process samples for determining CO2 system parameters in sea ice (Table 7), but sample storage and analyses have generally followed standard protocols for seawater (Dickson et al., 2007); methodological intercalibrations are desperately needed. For example, in the one study that determined pCO2 within sea ice by multiple methods (calculated from TIC and alkalinity in brines, section 4.6.1; calculated from TIC and alkalinity in bulk ice melts, section 4.6.3; and with in situ peepers, section 4.6.4), poor agreement was found between them (Miller et al., 2011a).
Interpretations of inorganic carbon data from sea ice have also been heavily influenced by our understanding of the seawater inorganic carbon system (e.g., Zeebe and Wolf-Gladrow, 2001). This reliance on the seawater model is perilous, because the seawater methods have been optimized for seawater and thus calibrated for very narrow concentration and salinity ranges; both sea-ice brines and bulk melt samples almost always fall outside those ranges. In addition, pH is defined analytically on a number of different scales (e.g., Dickson, 1993). The standard scale used for seawater (the total hydrogen ion scale, pHT), and standard buffers certified on that scale, cannot be applied rigorously to sea-ice brines (e.g., Miller et al., 2011a). Finally, the conditional stability constants used to convert between CO2 partial pressure (pCO2), dissolved inorganic carbon (DIC), total alkalinity (AT), and pH in seawater are only rigorously valid for temperatures above 0 °C and salinities between 5 and 50 g kg-1. Studies in spring ice (Delille et al., 2007) indicated that seawater thermodynamic relationships may be acceptable in warm, low-salinity sea ice, but in sea-ice brines at even moderate brine salinities of 80 g kg-1, Brown et al. (2014) found that measured and calculated values of the CO2 system parameters can differ by as much as 40%. On the other hand, because the CO2 system parameters are much more variable in sea ice than in seawater, sea-ice measurements demand less precision than those in seawater.
As indicated by Tables 6 and 7, most carbonate system parameters have been measured in both bulk ice and brines. Attempts to use those data to understand sea-ice biogeochemistry always involves uncomfortable assumptions about the validity of the sample handling and the in situbehavior of the CO2 system. For example, whereas brine samples are compromised by gas exchange during sampling, interpreting measurements in bulk sea ice requires assumptions about the presence of gaseous and solid inorganic carbon. Therefore, the best way to sample the ice and the best parameters to measure depend on both the conditions and specific questions targeted by the study.
Nonetheless, we can make some recommendations. In general, samples for CO2 system analyses in sea ice should be collected upwind from any ship, camp, or generator to avoid contamination by CO2 or soot from fossil fuel combustion. Hand-coring is preferable, although electric auger motors can be used, as long as the generator is located a substantial distance downwind. Some investigators have filtered their samples for DIC analyses (Papadimitriou et al., 2004; 2007; 2012) using specialized techniques to avoid significant gas exchange (McCorkle et al., 1985), but vacuum filtration is not recommended. Because the concentration of particulate inorganic carbon (PIC) can be high in sea ice (Dieckmann et al., 2008; 2010; Rysgaard et al., 2013), the results from DIC analyses of unfiltered sea-ice samples are properly termed total inorganic carbon (TIC).
Samples for pCO2, TIC, and pH are sensitive to gas exchange and need to be isolated from the atmosphere during sampling. Therefore, analyses of these parameters in brines, which are difficult to sample without exposing them to the air, can be problematic (see also sections 2.2.2 and 4.4).
Electrochemical pH measurements are particularly challenging in ice brines, because the high sample salinities result in large liquid junction potentials and severely slow electrode response times. In addition, stable, certified standard buffers are not available for brine solutions, compromising electrode calibration. Although the first sea-ice brine pH measurements were made electrochemically (Gleitz et al., 1995), spectrophotometric measurements are becoming more common (Miller et al., 2011a; Hare et al., 2013). Wren and Donaldson (2012) have developed a spectrophotometric method for analyzing pH in surface brine films that may have potential for in situ applications. Particular care is needed in spectrophotometric analyses to use optical and thermodynamic parameters for the dyes that have been defined for appropriate temperature and salinity ranges (e.g., Millero et al., 2009).
4.6.2. Gas bubbles
The standard method of crushing ice under vacuum to retrieve gases trapped in bubbles within an ice sample, as developed for glacial ice, may give artificially high pCO2 values when applied to sea ice (e.g., Tison et al., 2002). The vacuum likely disrupts the CO2 system equilibria within the brines, causing CO2 outgassing from the brine solution and possibly also precipitating CaCO3 (Geilfus et al., 2012b). Therefore, Geilfus et al. (2012b) developed a method to accurately measure CO2 in gas bubbles and brines in sea ice by equilibration with a headspace of known volume and CO2 mole fraction (dry head-space equilibration). The headspace must be as small as possible to assure that the CO2(g) released from the ice dominates the CO2 signal, with only a small contribution from the standard headspace gas. Gosink (1978) described an in situ head-space equilibration technique that involved sealing sampling cuvettes to the ice surface.
4.6.3. Bulk ice melts
Total alkalinity of bulk ice melts is a relatively uncomplicated analysis that has been performed for decades (e.g., Lyakhin, 1970; Anderson and Jones, 1985; Nedashkovsky et al., 2009). In a standard potentiometric titration (i.e., Dickson et al., 2007), the measured AT will include not only that which was in the brines in situ, but also a contribution from any particulate inorganic carbon (e.g., CaCO3(s)) that dissolves as the ice melts or when the sample is acidified during the titration.
On the other hand, because TIC is impacted by gas exchange, its analysis in bulk ice is more complicated; a method for confidently melting ice samples without allowing interaction with ambient CO2 has not yet been devised. In the field, ice cores need to be retrieved, sectioned, and isolated from the atmosphere as quickly as possible. The most common approach to melting ice samples for TIC analysis is to use gas-impermeable bags (made from fluoropolymers such as ALTEF® or Kynar®; Rysgaard et al., 2009; Fransson et al., 2011; 2013; Miller et al., 2011a). After sealing the sample in the bag, the headspace should be removed using a hand pump, to assure that the sample is not exposed to an excessive vacuum. As long as the container in which the sample is melted is sealed and the headspace (after melting) is less than 2% of the total volume, the TIC concentration in the solution should be correct to within 0.01% (Dickson et al., 2007); in fact, because the pCO2 of ice melts is generally low, the melt solution should also absorb essentially all of the gaseous CO2 initially present as bubbles trapped in the ice. Alternatively, the ice sample can be melted without a headspace in water of known TIC concentration (Rysgaard et al., 2007; 2009).
Unless AT and TIC samples can be analyzed within 1–2 hours of collection, they generally should be poisoned with small quantities of HgCl2 (Dickson et al., 2007). Although straightforward and relatively safe for samples that are initially aqueous, like seawater and brines, poisoning is more complicated for sea-ice melts, which cannot be bottled for long-term storage until after melting is complete. Some researchers have added the HgCl2 directly to the bag with the melting sample (Rysgaard et al., 2009; Fransson et al., 2013). However, because ice melts are poorly buffered, the Hg(OH)2 complexes formed from the added mercury may impact the carbonate system chemistry (Fransson et al., 2013). In addition, it is difficult to completely contain the mercury when working with gas-impermeable bags (during cleaning between samples, but also because the bags can fail at low temperatures, developing small holes and leaking), creating an exposure risk for all personnel using the core-processing laboratory. Therefore, HgCl2 is often added to samples only after they are transferred from the bags into bottles for long-term storage. We still lack a satisfactory method for safely preserving sea-ice samples for carbonate system analyses during melting.
Unlike TIC and AT (in units of mol kg-1), which are total quantities unaffected by the temperature and salinity changes associated with melting, pH and pCO2 are potentials; their values measured in sea-ice melts cannot be directly converted to the original conditions in the solid ice/brine matrix. Therefore, pCO2 and pH measurements in ice melts do not provide information about the initial, in situ conditions of the ice, although the measurements can be used to calculate other CO2 system parameters or to derive the theoretical characteristics of the melt that will influence the surface waters in summer (Nedashkovsky and Shvetsova, 2010; Fransson et al., 2011; Geilfus et al., 2013). Samples for pCO2 and pH analyses have the same issues with potential degradation during melt as TIC and AT samples: standard seawater protocols (e.g., Dickson et al., 2007) indicate that pCO2 samples that cannot be analyzed within a couple of hours should be poisoned with HgCl2; and, presently, pH samples cannot be stored for they change significantly within hours of collection (i.e., within the time it takes for an ice sample to melt).
4.6.4. In situ sensors
Ideally, in situ sensors would provide the most meaningful pCO2 and pH measurements in sea ice, particularly if the sensors could be deployed at freeze-up, so that installation would not disrupt an established ice cover. To date, available pH microelectrodes are still unsuitable for deployment in sea ice, because their electrolyte solutions freeze at low temperatures, in addition to the calibration and response-time issues discussed in section 4.6.1. Although in situ silicone gas exchange chambers (“peepers”) have been used for pCO2 measurements (section 4.4; Miller et al., 2011a;b), gas diffusion rates in silicone decrease dramatically with temperature, and peepers require long equilibration times, limiting deployments to extended occupations of a single site. In addition, peepers have not yet been fully tested or calibrated under controlled conditions. As noted in section 4.4, in situ CO2 or pH probes will modify their local thermodynamic environment within the ice to at least some extent.
4.6.5. Particulate inorganic carbon (PIC)
A number of carbonate salts are known to precipitate from brines in sea ice, including calcium carbonate (generally in the form of ikaite, CaCO3·6H2O; e.g., Dieckmann et al., 2008) and magnesium- and mixed magnesium-calcium-carbonates (e.g., Assur, 1958). Precipitation of any of these minerals likely has a strong influence on pCO2 and the entire CO2 system within the ice. Ikaite has been identified in both Antarctic and Arctic sea ice (Delille, 2006; Dieckmann et al., 2008; 2010; Rysgaard et al. 2012; 2013; Geilfus et al., 2013; Nomura et al., 2013a), but solid salts are not recovered from all samples (e.g., Nomura et al., 2013a). The lack of observable PIC in some samples is likely due to natural variability in sea ice (possibly related to phosphate concentrations, thermal history of the ice, and/or sea-ice permeability; Nomura et al., 2013a; Rysgaard et al., 2013; Papadimitriou et al., 2013; 2014) and not methodological differences, as the presence and absence of precipitates are observed by the same groups. Recent laboratory studies have defined the thermodynamics and kinetics of ikaite precipitation in abiotic ice brines (Papadimitriou et al., 2013; 2014), but the conditions controlling the formation and preservation of solid CaCO3 in natural sea ice are still largely unknown.
Because carbonate salts are water soluble to varying extents at temperatures above freezing, the ice samples must be melted at a low temperature and processed as soon as melting is completed. Although the melted samples can be filtered to collect the solid precipitate (Dieckmann et al., 2010; Fischer et al., 2012; Søgaard et al., 2013), the quantity of particulate organic matter present in the sea ice can interfere with subsequent visualization and analysis of the inorganic salts, so the precipitate is often collected from sea-ice melt samples using a pipette (Dieckmann et al., 2008; Miller et al., 2011b; Geilfus et al., 2013; Nomura et al., 2013a). Rysgaard et al. (2013) have developed a microscopic method to visually identify and quantify ikaite crystals from the ice as it melts. Even when dry, ikaite is unstable at temperatures above 4 °C; if confirmation of the specific calcium carbonate mineralogy (i.e., ikaite versus calcite, aragonite, or vaterite) is required, the sample must be kept below 4 °C throughout sample recovery, melting, storage, transport, and analysis. Analysis is usually by x-ray diffraction spectrometry, but facilities able to keep a sample cold throughout the analysis are rare. If the specific mineralogy of the salt is not required, the precipitate sample can be stored indefinitely at room temperature and analyzed on any x-ray diffraction instrument (Dieckmann et al., 2008; Miller et al., 2011b) or with a standard calcium assay (Fischer et al., 2012).
The question of whether carbonate minerals that precipitate within sea ice are mobile with the brines has not been completely resolved. Despite observations that particulates are under-represented in percolating sea-ice brines (section 2.2.2), solid CaCO3 has been recovered from centrifuged brines (Tison et al., 2002), and circumstantial evidence from the Sea of Okhotsk (Lyakhin, 1970) and the Beaufort Sea (Fransson et al. 2013) has indicated that abiotic CaCO3 precipitates from sea ice may be released to the water column.
4.7. Photochemistry: CDOM, hydrogen peroxide, and ozone
Photochemical processes are likely to be very important in many sea-ice biogeochemical cycles, but while numerous studies have examined the transmission of electromagnetic radiation through sea ice (e.g., Perovich, 2009), there has been little research on photochemistry within the ice (Belzile et al., 2000). Colored dissolved organic matter (CDOM, usually measured spectrophotometrically and reported as absorption coefficients, in units of m-1) likely represents the most photochemically active fraction of the non-living sea-ice components and has been measured in sea ice by a number of investigators (Belzile et al., 2000; Scully and Miller, 2000; Granskog et al., 2005b; 2015; Uusikivi et al., 2010; Norman et al., 2011). Fluorescence has also been used to measure and differentiate the components of CDOM in natural and laboratory sea ice and in frost flowers (Stedmon et al., 2007; 2011; Müller et al., 2013; Granskog et al., 2015). Although generally assumed to be less prone to contamination than bulk DOC, samples of CDOM, particularly those measured by fluorescence, can easily be contaminated. In addition, the effects of melting protocols on the absorbance of organic matter from sea ice has not been investigated explicitly. Hydrogen peroxide and other photochemically produced oxidizers, such as ozone, are likely important players in any photochemical reactions occurring in sea ice (Klánová et al., 2003; King et al., 2005), but no one has reported direct sea-ice measurements of these compounds.
4.8. Stable isotopes: 18O, 2H, 13C, 15N, 30Si
The stable oxygen isotope ratio (18O relative to 16O) in the water molecules of a sea-ice sample is controlled by many of the processes that influence sea-ice biogeochemistry, including freezing, melting, flooding, and snowfall, making δ18O a powerful tool for sea-ice studies (e.g., Eicken, 1998; Granskog et al., 2003; Tison et al., 2008; Nomura et al., 2009; 2011). Particularly in the Arctic Ocean, where δ18O measurements can also help distinguish between riverine versus sea-ice melt sources of freshwater in the surface ocean (e.g., Macdonald et al., 1989), δ18O is often considered a mandatory parameter, along with salinity.
Sampling δ18O is simple and inexpensive; the main concern is evaporation during sample storage. Although evaporation during melting could also be an issue, if the container is open or the headspace is large, evaporation is no more of a problem for δ18O than for salinity. Mass spectrometric analyses of δ18O require as little as a few milliliters of sample; both bulk sea-ice melts and brines have been analyzed for δ18O (e.g., Zhou et al., 2013). Glass containers with caps forming a tight seal are preferred, particularly if the samples are likely to be stored for more than a few months before analysis (e.g., McLaughlin et al., 2012). Parafilm can also be wrapped around the outside of the cap, to further protect against leakage (Miller et al., 2011b).
Sea-ice measurements of the stable isotopes of other elements are summarized in Table 8. Deuterium fractionation in the water molecules in sea ice is greater than that of 18O, providing additional information on brine convection and ice melt. Sampling and handling of 2H samples for mass spectrometric analysis should follow those for 18O samples (Zhou et al., 2013). Stable isotope ratios of carbon (13C relative to 12C), nitrogen (15N relative to 14N), and silicon (30Si relative to 28Si) are proving to be useful tools in investigations of sea-ice biogeochemical cycles, including interpretations of sedimentary records in the polar oceans. In sea ice, stable isotope measurements can potentially help distinguish between the origins (i.e., land, seawater, or sea ice) of the particulate organic matter, as well as between biogeochemical cycling pathways. Samples for stable isotope analyses are typically collected from bulk sea ice and sackhole brines following the methodologies for POC, PON, bSiO2 (section 3.3.1), NO3-, Si(OH)4 (section 4.1), TDN (section 4.2.2), and TIC (section 4.6); the analyses have used standard mass-spectrometric techniques (e.g., McCorkle et al., 1985; Kennedy and Robertson, 1995; Sigman et al., 2001; Cardinal et al., 2003).
aδ18O analyses are routine and, therefore, not included.
bTDN = NO3- +NO2- + NH4+ + DON and DON = dissolved organic N.
5. Ice-atmosphere and ice-ocean fluxes
Because sea ice is porous, it exchanges material with both the overlying atmosphere and the underlying water. Brine rejection from sea ice has long been recognized as a primary driver of oceanic deepwater formation and global circulation (e.g., Chu and Gascard, 1991), with likely implications for the biogeochemical cycles of many elements. Likewise, the atmosphere, both in the boundary layer directly above the sea ice and at higher altitudes (e.g., Begoin et al., 2010), is strongly influenced by sea-ice biogeochemistry.
Qualitative information about the direction of fluxes can be derived from measurements of concentration gradients between the ice and the atmosphere or the water: the larger the gradient, the larger the flux might be. However, confirming and quantifying those presumed fluxes requires more sophisticated methods. In particular, when material fluxes above sea ice are measured directly, the fluxes estimated from the measured ice-air gradients can be wrong both in magnitude and direction, because of reactions occurring at the interface, in the surface brines, frost flowers, and snow cover, that produce or consume gases and aerosols.
5.1. Air-ice gas fluxes
A number of methods have been developed to estimate gas fluxes above sea ice, but no systematic intercomparisons between various gas flux techniques over ice surfaces have been published. In particular, micrometeorlogical (sections 5.1.2–5.1.6) and chamber (section 5.1.1) methods measure fluxes on very different temporal and spatial scales, which has confounded efforts to compare the resulting flux estimates. Comparisons between eddy covariance and chamber methods have been carried out over terrestrial surfaces (Wang et al., 2013; Riederer et al., 2014), but carefully designed intercalibration experiments over sea ice are still needed to resolve remaining important questions about how each type of flux measurement performs in the sea-ice environment.
5.1.1. Flux chambers
Enclosure methods, which were widely used in early biosphere-atmosphere exchange studies (e.g., Mosier, 1989), are still common for some applications, including sea-ice biogeochemistry (Figure 13; Table 9). The method is based on the rate of increase (or decrease) of the trace gas concentration with time within a chamber placed directly on the ice (or snow) surface providing the flux (McMinn et al., 2009).
changes in the radiation balance (both short and long wave);
changes in the temperatures of the air and the surface (i.e., the ice or snow);
changes in turbulence, wind speed, and the vertical density profile;
a pressure gradient between inside and outside the chamber; and
a surface-atmosphere gas concentration gradient, including changes in the gas concentration inside the chamber resulting from the flux.
To overcome these issues, flux measurements should be completed quickly, before changing conditions introduce artifacts. Although the optimum length of time for a measurement is dependent on the situation (e.g., weather, gas concentration gradient), 20 to 30 minutes is generally recommended. However, over snow, even a perfect chamber deployment may underestimate the flux, because in an undisturbed system, wind-driven pressure pumping within a snow cover can enhance the transport beyond molecular diffusion by up to 40% (Bowling and Massman, 2011).
Two fundamentally different types of chamber systems are available (Mosier, 1989; Luo and Zhou, 2006): in closed systems, the change in the gas concentration inside the chamber is measured directly, usually by connecting the interior to an on-line gas analyzer in a closed loop; in open systems, ambient air is pumped through the chamber, and the flux is calculated from the air flow rate and the difference in gas concentrations between the inlet and outlet. Changes in the trace gas concentration within the enclosure are a larger concern with a closed system, but open chamber methods are more susceptible to artifacts arising from interior-to-exterior pressure gradients. To date, only closed-chamber systems have been used to measure gas fluxes above sea ice; because of the difficulty in controlling the pressure differential in open systems, we recommend using closed systems, particularly if they incorporate dampening to equalize the pressure (as described by Xu et al., 2006). Current generation closed-chamber systems (e.g., the Li-COR 8100-104) include such dampening mechanisms, making them probably the best chambers for measuring gas fluxes over sea ice (Fischer, 2013).
The greatest advantages and disadvantages to using enclosure methods are both due to spatial variability. Chamber enclosures only integrate the signal from the area they cover (generally, a few hundred cm2); if the exchange is governed by factors that vary on larger horizontal scales (i.e., the thickness and wetness of the snow cover, melt ponds, leads, under-ice hydrology, etc.), a prohibitive number of individual chamber measurements over a large area may be required to estimate the flux accurately (section 2.1). On the other hand, the method is ideal for studying specific, small-scale processes influencing variations in the flux (i.e., brine channel distributions, ice algae respiration, etc.), and enclosure methods are the only technique available to determine fluxes on the same scale as most sea-ice biogeochemical measurements. In contrast, the micrometeorlogical techniques (sections 5.1.2–5.1.6) cover areas several orders of magnitude larger than chambers, integrating fluxes from different ice types and any open water in the footprint; micrometeorlogical results can, therefore, be difficult to interpret over heterogeneous surfaces.
5.1.2. Micrometeorological methods: general
Micrometeorological techniques, such as eddy covariance (EC), can be used to measure fluxes of gases, as well as of momentum, sensible heat, and latent heat (e.g., Vihma et al., 2009) above sea ice. Unlike enclosure techniques, micrometeorlogical methods do not modify the observed environment, and they integrate processes occurring over relatively large spatial areas (up to several hundred m2). This integration can be a problem if the surface is heterogeneous (i.e., sea ice, leads, or other open water features may contribute to the observed fluxes), but typically the issue can be addressed by calculating the “footprint” of the flux measurement (e.g., Vesala et al., 2008). The EC method has been used widely to examine CO2 fluxes over all types of surfaces, but its application to many other chemical compounds has been hampered by a lack of fast-response sensors and by small signal levels. These limitations have led to development of a variety of alternative methods, including eddy accumulation (Businger and Oncley, 1990) and gradient techniques (e.g., Businger et al., 1971). However, EC is, by definition, a direct flux measurement method (e.g., Swinbank, 1951), while the others are based on several assumptions that break down over heterogeneous surfaces and in stably stratified atmospheres, such as often occur over sea ice.
The basic framework for measuring and interpreting micrometeorlogical flux data, including those from EC systems, is based on simultaneous measurements of the gas concentration and vertical, turbulent motion in the atmosphere (Figure 14). Ideal conditions for micrometeorlogical flux measurements include a horizontal and homogeneous surface, no source or sink in the atmosphere that can alter the concentration above the surface, and consistent atmospheric conditions (i.e., air temperature, wind velocity, and mean gas concentration).
5.1.3. Eddy covariance
A general description of EC methods is given by Lee et al. (2004), but briefly, both the three-dimensional motion field and the analyte must be measured at the same place and at the same high frequency. Sonic anemometers are used for the motion field, while the analyte can either be measured in situ by an instrument exposed to the atmosphere (an “open path” system) or by transporting air in a constant stream to an analyzer (a “closed path” system). Non-dispersive infrared (NDIR) analyzers are generally used to measure CO2 fluxes in EC systems (e.g., Baldocchi, 2003), but other types of detectors are also available for CO2 and CH4 (Crosson, 2008; Detto et al., 2011) and for volatile organic compounds (VOCs; Müller et al., 2010). Fluxes are derived from the covariance between the vertical velocity and the concentration of the species of interest (Figure 14); the EC approach is, of course, only valid when the fluctuations in the concentration are caused by the vertical turbulence, not by sources or sinks within the boundary layer. In order to resolve the turbulence and measure a sufficient fraction of the vertical transport, the sampling frequency must be higher than 10 Hz.
Over sea ice, the EC method has primarily been used to measure CO2 exchanges (Table 9) through deployments either on ships or directly on the ice (Figure 15). All published studies, with the possible exception of Semiletov et al. (2004) who did not specify their instrumentation, have used the LI-7500 open-path CO2/H2O analyzer, manufactured by LI-COR Biosciences. This instrument has been used widely at lower latitudes, but it suffers from some biases under cold conditions and in the marine environment. In cold weather, the instrument may significantly heat the column of air it is sampling, which lowers the measured gas concentration and biases flux measurements towards CO2 uptake (Burba et al., 2008). In addition, salt deposited on the lens of the instrument interferes with the infrared absorption measurement, resulting in flux overestimations (Prytherch et al., 2010). In most cases, researchers have taken steps to minimize or correct for these biases, but results obtained using these open-path sensors should still be interpreted with a healthy dose of skepticism.
Closed-path EC systems potentially avoid many of these problems associated with the open-path sensors, because the temperature of the air sample can be controlled and instrument lenses can be protected by filters. However, the closed-path systems also suffer from their own shortcomings. Most importantly, closed-path systems inevitably attenuate gas concentration fluctuations and thus degrade the flux signal (Leuning and King, 1992; Lee et al., 2004), making it difficult to confidently identify gas exchanges between sea ice and the atmosphere, which are often much smaller than those observed over terrestrial or open-water surfaces. This difficulty may be why no sea ice-air CO2 flux measurements using closed-path EC systems have yet been reported, despite several attempted deployments. However, Muller et al. (2012) have reported ozone fluxes over sea ice measured using a closed-path EC system.
Eddy covariance measurements can be subject to significant biases and random errors (Businger, 1986; Finkelstein and Sims, 2001), but there are no straightforward ways to calibrate or validate EC flux measurements in the field. These errors are exacerbated over sea ice, where the observed fluxes are often small. Particularly when temperatures and, therefore, sea-ice permeability are low, the fluxes are likely to be close to eddy covariance detection limits (estimated to be 1 µg C m-2 s-1 for CO2; Wang et al., 2013), and the uncertainty in the measured flux can be over 200% in some cases (Sørensen et al., 2014). However, evaluating the frequency spectra of the sampled EC data can eliminate some sources of error (e.g., advection, noise), and coupling fluxes estimated from EC and by spectral techniques gives more robust results (Kaimal et al., 1972; Sørensen and Larsen, 2010; Norman et al., 2012; Sørensen et al., 2014). Flux estimations based on the spectral methods use the same instrumental configuration as EC, but corrections for atmospheric stability are also required.
5.1.4. Eddy accumulation
For many trace gases, fast-response sensors, as required for EC flux measurements, are not available; relaxed eddy accumulation (REA) provides an alternate method for estimating the fluxes of these gases. Measurements by REA rely on conditional sampling (Hicks and McMillen, 1984; Businger and Oncley, 1990) of the gas into separate reservoirs depending on whether the bulk air movement is upward or downward. The “relaxation” refers to the fact that samples are taken with a constant flow rate and are not weighted according to the vertical wind speed; the data consequently lack information on the vertical wind speed. To date, only DMS fluxes have been measured over sea ice by REA (Zemmelink et al., 2008), but the method has potential for application to fluxes of other gases.
5.1.5. Gradient techniques
Gradient techniques provide another alternative for measuring fluxes of trace species for which no fast response sensor is available. Basically, the gradient technique requires measurements of gas concentrations from at least two levels (one typically within 1 m of the air-ice surface, the other 1–10 m from the surface), along with some estimate of how rapidly the gas is transported between those levels. This transport rate estimate is generally based on observations or parameterizations of atmospheric turbulence, which are usually derived from wind velocity and temperature measurements (preferably made at the same two levels). The method is indirect, can require a number of empirical functions to account for thermal stratification of the atmosphere, and is based on the assumption that turbulent transfer is analogous to molecular diffusion. The most common gradient technique is the aerodynamic method, which is based on the momentum flux equation and the wind speed-gradient relationship (Businger et al., 1971; Businger, 1986; Baldocchi et al., 1988; Sørensen et al., 2005). To date, the gradient technique has only been applied to measure nitric acid (Beine et al., 2003) and ozone (Bocquet et al., 2011) fluxes over terrestrial snow, but the method has potential for applications over sea ice.
5.1.6. Best practices for micrometeorlogical techniques
Several textbooks and review articles (we recommend Lee et al., 2004) have dealt extensively with issues of best practice for micrometeorlogical techniques, particularly with respect to eddy covariance. The interested reader should consult those texts, but we address several points here that are unique to the sea-ice environment.
The first important consideration is the installation of the meteorological tower, especially if the measurements are to last through a portion of the melt season. A melting sea-ice surface is inherently unstable; towers can tilt and even topple in such conditions. We recommend one of two approaches to combat such tower instability. First, the tower can be mounted on a qamutiq (sled), which can be periodically moved or repositioned to keep the tower level (Figure 15). Alternatively, the tower should be frozen into the ice by drilling holes (to a depth of about 50 cm), in which the ends of the tower posts are anchored with frozen fresh water. This system can be improved by also passing the tower posts through a plywood base on the ice surface, as this will shade the posts and prevent localized melt. The tower can also be directly fixed to a sheet of plywood (preferably with rigid insulation underneath), but freezing the tower legs into the ice is preferred.
The orientation of the tower is also important. Even a low-profile lattice-style tower can create significant snow drifts; it is best to place the meteorological equipment on the side of the tower that faces the prevailing wind direction, so that it will measure fluxes predominately over an undisturbed, upwind surface. Most eddy covariance installations over sea ice are by necessity close to the ground (less than 5 m above the surface), leading us to recommend sampling at a rate of 20 Hz to capture the small-scale eddies that exist near the surface. Finally, it is important to choose instruments that have low-temperature ratings and are well sealed against blowing snow. As discussed previously, particular attention should be paid to the choice of infrared gas analyzer to ensure that it operates properly in cold environments.
5.2. Aerosols, frost flowers, and saline snow
The formation of atmospheric particulate aerosols is another important interface flux in the sea-ice system. The polar regions frequently receive aerosols from lower latitudes, but wintertime sea-ice formation is also associated with vertical particle fluxes. Whereas under open-water conditions, the sea-to-air transport of both solid and liquid aerosols is controlled by breaking waves, in polar regions brine-wetted saline snow and frost flowers formed on new sea ice provide additional sources of atmospheric aerosols. In particular, frost flowers may be broken up by blowing wind, and the submicron fraction may have lifetimes up to a week. Both frost flowers and saline snow contain NaCl, other ocean salts from sea-ice brines, and organic components, including both organisms and their products (Alvarez-Aviles et al., 2008; Bowman and Deming, 2010; Douglas et al., 2012; Ewert et al., 2013). Particles from frost flowers and saline snow can be lofted into the atmosphere and transported long distances by wind, eventually deposited in new places under dry or wet conditions. Therefore, the size and composition of such particles should be characterized in studies of atmospheric aerosols in polar regions (Yang et al., 2008; Obbard et al., 2009; Roscoe et al., 2011).
Several methods exist to measure aerosols: in-line with mass spectrometers and ion chromatographs, as well as off-line by gas chromatography-mass spectrometry and Fourier transform infrared spectroscopy (e.g., Russell, 2014). Each of these recommended methods characterizes different, complementary chemical qualities. However, comparability between methods requires well-characterized inlets and careful techniques for minimizing artifacts during collection and storage. New advances in quantifying aerosol fluxes require simultaneous analyses of sea-ice components, with careful attention to the local meteorology, so that the measured aerosols can be linked to their upwind source regions.
Of particular interest is the role of halogens in atmospheric chemistry of the polar regions (Simpson et al., 2007; Abbatt et al., 2012), including the function of Br in tropospheric ozone and mercury depletion and the possible importance of iodine in new particle formation. High concentrations of BrO observed in satellite datasets during the polar spring, in particular, suggest that halogens originate at the sea-ice surface; the available studies of halocarbons in sea ice, to date, have confirmed variable and often high concentrations of biogenic halocarbons in ice, brines, and overlying snow (Sturges et al., 1997; Simpson et al., 2005; Atkinson et al., 2012; Mattson et al., 2012; Granfors et al., 2013a;b). Inorganic halides are easily sampled and analyzed (after melting and filtering) by, for example, ion chromatography for bromide, cyclic voltammetry for iodide, and spectrophotometry for iodate (e.g., Atkinson et al., 2012). On the other hand, sampling sea ice for halocarbons, which are volatile, should follow methods to limit gas exchange (section 4.4), including melting samples in gas-impermeable bags with minimal headspace (as for inorganic carbon species, section 4.6.3). Halocarbon samples are usually analyzed by gas chromatography, requiring aqueous (i.e., melted, see section 3.2) or gas samples. Shaw et al. (2011) measured iodated organics in artificial sea ice from laboratory tank studies by melting ice samples directly in sealed syringes to limit gas exchange. New methods to confidently analyze the halogen chemistry of frost flowers are particularly needed.
5.3. Ice-water fluxes
Most efforts to quantify biogeochemical exchanges at the ice-water interface have used mass balance and budgeting tactics. This approach has been most successful in identifying water-to-ice fluxes of nutrients (Cota et al., 1987; 1990; Rahm et al., 1995; Nishi and Tabeta, 2008) and of trace metals (Granskog and Kaartokallio, 2004; Lannuzel et al., 2010; 2011; van der Merwe et al., 2011a). Recent models of ice-brine dynamics have confirmed that seawater pumping into the brine network of growing sea ice could cycle enough surface seawater through the lower parts of the ice cover to account for observations of the nutrient and iron distributions in the ice (Vancoppenolle et al., 2010). On the other hand, efforts to estimate inorganic carbon fluxes from sea ice into the underlying surface waters have been confounded by the need to identify small changes in large, variable concentrations (e.g., Miller et al., 2011b; Fransson et al., 2013).
Attempts to use the gradient method to derive sea ice-water chemical fluxes have been limited by the difficulty of collecting high-vertical resolution water samples from under the undisturbed ice sheet. Although Dieckmann et al. (1992) developed a promising device for detailed under-ice sampling, it has not been utilized widely, and efforts by divers to collect biogeochemical samples from under the ice have not produced consistent results. In situ microsensors, such as those deployed for measuring oxygen by Rysgaard et al. (2001) and McMinn et al. (2000), show some promise for measuring gradients near the ocean-ice interface, but deployment of such instruments (usually by divers) remains difficult.
Eddy covariance shows promise for measuring ice-water, as well as ice-air, fluxes. Although to date only oxygen fluxes have been determined (section 3.4.1; Long et al., 2012), the method should be applicable to any analyte for which an in situ sensor is available with a response time less than 1 second. Sensors for measuring fluxes of nitrate (Johnson et al., 2011) and hydrogen sulfide (McGinnis et al., 2011) have been employed in benthic studies and could be adapted to the sea-ice environment.
Sinking particle fluxes from the sea ice can be measured using particle interceptor traps tethered below the ice (Michel et al., 1996; Fortier et al., 2002; Juul-Pedersen et al., 2008; Nishi and Tabeta, 2008); time series from such “sediment” traps have proven to be a valuable tool in estimating carbon budgets of sea-ice primary production, export, and transfers to pelagic and benthic grazers (e.g., Michel et al., 2002; Renaud et al., 2007). The traps are typically deployed at shallow depths specifically to capture particles exported from the ice (as opposed to produced within the water column) and to avoid excessive drag on the trap line. However, the traps can also be deployed at deeper depths, thereby providing insights into sea ice-pelagic-benthic coupling (Fortier et al., 2002). The extensive literature pertaining to sediment trap methodology does not deal specifically with under-ice deployments, although standard protocols for particle trap deployments should be followed as much as possible (i.e., Gardner, 2000). Particular challenges associated with under-ice deployments include over- and under-trapping if the traps are deployed under moving ice floes or under fast ice in high current regimes.
6. The future
Biogeochemists working in the sea-ice environment have made tremendous strides over recent years in learning how to measure and then understand the biogeochemical processes occurring in sea ice. Nonetheless, we still have much work ahead of us to resolve the uncertainties in the measurements we are making and identify which of the methods we are using are most suitable under various conditions (Table 10).
aThe order in which items are listed does not necessarily imply priority.
Dedicated, collaborative, field and laboratory studies are required to compare and intercalibrate a number of methods, most notably those for primary production and the CO2 system, but also for nutrients, biomass, cell abundance, and size fractionation of organic matter. Time on icebreakers and in ice camps is expensive; it is often difficult to rationalize, within interdisciplinary process studies, the kind of redundant sampling and analyses that are required for rigorous intercalibrations. Therefore, to resolve methodological discrepancies, we will need expeditions and experiments that are focused first on methods. This kind of approach is difficult for many sea-ice scientists, as well as our funding agencies, who are motivated to use every opportunity in remote polar environments to address questions on how sea ice impacts the functioning of our planet. However, in the end, the confidence in our measurements provided by focused methodological studies will make these efforts worthwhile. In the meantime, to the extent possible, we encourage interdisciplinary process studies to include methodological intercalibrations.
For the most part, large icebreakers are not essential (and may be overkill) for the intercalibration and method validation experiments required (Table 10). Rather, a number of coastal laboratory facilities with ready access to fast ice, such as those in Barrow (Alaska, USA), Svalbard (Norway), McMurdo Sound (Antarctica), Saromo-ko (Hokkaido, Japan), and Tvärminne (Hanko, Finland), could be very useful sites for this work. Laboratory studies, including those in large-scale ice-tank facilities, will also be a critical component of our efforts to understand our methods and their limitations. The spatial and temporal variability in natural sea ice often prevents us from accurately constraining our methods; in controlled laboratory tanks can we determine the true methodological precision and accuracy of chemical measurements in sea ice. On the other hand, to date, laboratory tanks have not been successfully inoculated with a natural sea-ice community, which, when coupled with universal challenges to laboratory-based biological experiments (i.e., maintaining nutrient supply, etc.), severely limits the utility of laboratory experiments for studying the complexity of biological processes in sea ice.
The field of sea-ice biogeochemistry is also ripe for new technological developments; we have only begun by trying to adapt methods we know from other disciplines. As the field matures, we hope that entirely new approaches and methods will be developed. These efforts could be facilitated by recruiting more students with degrees in analytical chemistry, biotechnology, bioengineering, and electronics, as well as through new collaborations. In particular, in situ probes are critically needed to answer many of our questions effectively. Electrochemical and optical technologies show the most promise; throughout this paper, we have highlighted a number of biogeochemical parameters that we believe are particularly well-suited for sensor development, but there will certainly also be others. Robust ice buoys on which such chemical sensors could be deployed received a substantial boost from the 2007–08 International Polar Year (e.g., Knepp et al., 2010) but require additional development to recover data dependably throughout the entire cycle of freeze-up and melt. Recent advances in under-water vehicle technology, including successful under-ice deployments of instrumented remotely operated vehicles (ROVs) and autonomous underwater vehicles (AUVs) also promise new tools to study sea-ice physical-ecological-biogeochemical interactions on scales from meters to kilometers (Wadhams and Doble, 2008; Williams et al., 2013).
Finally, a key challenge for the future is integrating interdisciplinary measurements on different scales to tell a more complete story of sea-ice biogeochemistry. For example, measurements of biogenic gas fluxes paired with analyses of bacterial gene expression and in situ nutrient concentrations are far more valuable than any of these observations, alone.
In closing, we have described many of the problems with existing methods for studying sea-ice biogeochemistry, while also noting the successes, which have been significant. This important and exciting research field is now beginning to mature. We hope that the insights presented here will provide inspiration for new scientists to boldly tackle some of these challenging methodological problems, because we must solve these problems, if we are to understand how sea ice impacts the Earth’s biogeochemical cycles.
Data accessibility statement
This manuscript does not present original data.
© 2015 Miller et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
This manuscript is a product of SCOR working group 140 on Biogeochemical Exchange Processes at Sea-Ice Interfaces (BEPSII); we thank BEPSII chairs Jacqueline Stefels and Nadja Steiner and SCOR executive director Ed Urban for their practical and moral support of this endeavour. This manuscript was first conceived at an EU COST Action 735 workshop held in Amsterdam in April 2011; in addition to COST 735, we thank the other participants of the “methods” break-out group at that meeting, namely Gerhard Dieckmann, Christoph Garbe, and Claire Hughes. Our editors, Steve Ackley and Jody Deming, and our reviewers, Mats Granskog and two anonymous reviewers, provided invaluable advice that not only identified and helped fill in some gaps, but also suggested additional ways to make what is by nature a rather dry subject (methods) at least a bit more interesting and accessible. We also thank the librarians at the Institute of Ocean Sciences for their unflagging efforts to track down the more obscure references we required. Finally, and most importantly, we thank everyone who has braved the unknown and made the new measurements that have helped build sea-ice biogeochemistry into the robust and exciting field it has become.