Introduction
The surface of the Greenland Ice Sheet is darkening (He and others, Reference He2013; Tedesco and others, Reference Tedesco2016; Tedstone and others, Reference Tedstone2017), which accelerates its surface melt during the summer. The surface darkening is attributed to changes in the physical properties of snow and ice during the melt season, as well as the presence of light-absorbing particulates (LAPs; Dumont and others, Reference Dumont2014, Tedesco and others, Reference Tedesco2016, Tedstone and others, Reference Tedstone2020). The physical changes in ice structure and variability in the concentration of LAPs in bare ice areas are not well represented in regional climate models (RCMs; Tedesco and others Reference Tedesco2016), leading to underestimations of predicted melt rates, particularly in the southwestern margin of the ice sheet (Alexander and others, Reference Alexander2014; Tedesco and others, Reference Tedesco2016; Antwerpen and others, Reference Antwerpen, Tedesco, Fettweis, Alexander and van de Berg2022). In this region, part of the discrepancy between observed and modelled albedo is explained by the presence of biological LAPs, in particular blooms of pigmented microalgae, which darken large areas of the ice surface (Wang and others, Reference Wang, Tedesco, Alexander, Xu and Fettweis2020; Tedstone and others, Reference Tedstone2020; Cook and others, Reference Cook2020; Williamson and others Reference Williamson2021). The prolonged exposure of bare ice areas, due to the accelerated snowline retreat induced by climate warming (Ryan and others, Reference Ryan2018), is thought to extend the growth season of these microalgae. This could lead to larger, darker and more prolonged algal blooms in future (Benning and others, Reference Benning, Anesio, Lutz and Tranter2014). At the same time, changes in precipitation and intensified run-off may alter the physical and chemical conditions on the ice surface, with unknown consequences for algal bloom development and, thus, biological albedo reduction.
The main algal species darkening the ice surface are Ancylonema alaskana and A. nordenskiöldii (Lutz and others Reference Lutz, McCutcheon, McQuaid and Benning2018), hereafter referred to as glacier ice algae (Fig. 1a). The strong light absorption and albedo-reducing effect of glacier ice algae is due to their intracellular accumulation of the pigment purpurogallin carboxylic acid-6-O-β-D-glucopyranoside (Remias and others, Reference Remias, Holzinger, Aigner and Lu2012a, Reference Remias2012b; Williamson and others, Reference Williamson2020; Halbach and others, Reference Halbach2022; Fig. 1a, b). The optical properties of glacier ice algae have recently been measured (Chevrollier and others, Reference Chevrollier2022; Fig. 1d) and incorporated into the radiative transfer model, BioSNICAR (Flanner and others Reference Flanner, Zender, Randerson and Rasch2007; Cook and others, Reference Cook2020), along with an adding-doubling solver (Briegleb and Light, Reference Briegleb and Light2007) that was shown to accurately predict glacier ice albedo (Whicker and others, Reference Whicker2022). BioSNICAR predicts algal-driven albedo reduction from the surface concentration of algal cells, illumination conditions, as well as vertical profiles of ice density and bubble/pore size (https://github.com/jmcook1186/biosnicar-py). Chevrollier and others (Reference Chevrollier2022) showed this model could be used to accurately recreate the spectral reflectance of ice surfaces populated by algal blooms (Fig. 1e), suggesting that it can be used to study the algal impact on bare ice albedo, including an integration into remote sensing and predictive modelling systems. In particular, Chevrollier and others's (Reference Chevrollier2022) model validations imply that synthetic albedo datasets covering wide ranges of possible ice column, impurity and illumination conditions can be used to derive algal detection algorithms. So far, these algorithms had been derived from relatively small empirical datasets because radiative transfer models did not have access to reliable algal optical properties or validation data.
Fig. 1. Pigments, light absorption and albedo reduction by glacier ice algae. (a) Microscope picture of Ancylonema alaskana, scale bar = 5 μm (b) average pigment composition of glacier ice algae, (c) picture of the weathering crust surface, (d) average absorption cross-section of glacier ice algae and (e) measured and modelled spectra of an ice surface colonised by glacier ice algae. Figures adapted from Halbach and others (Reference Halbach2022) and Chevrollier and others (Reference Chevrollier2022).
BioSNICAR could eventually be refactored for coupling to RCMs to forecast algal-driven melt, but this will require models for algal biomass accumulation and surface crust development to be built. However, the controls on surface algal biomass distribution and weathering crust state as well as the feedback between them are not yet sufficiently well understood (Fig. 2). Here we outline the knowledge and process-level understanding needed for these modelling developments, using a combination of field measurements, laboratory experiments, remote sensing and modelling.
Fig. 2. Schematic overview of the ice-algal system and key feedbacks with environmental variables and surface albedo. Note that not all interactions are included, for example among environmental variables. The incoming shortwave radiation available for the algae will also indirectly depend on the presence and properties of a snow cover.
Controls on algal growth and mortality
The availability of liquid water, light and nutrients could limit glacier ice algal growth (Fig. 2). As CO2 is constantly supplied from the atmosphere to the ice surface, where it easily dissolves, it is unlikely to be a limiting factor at the ice surface that the algae inhibit. To date, algal growth has been modelled as a function of the cumulative growth period where environmental conditions allow for algal growth (Williamson and others, Reference Williamson2018, Reference Williamson2020; Onuma and others, Reference Onuma2022). The thresholds of these environmental conditions were defined by Williamson and others (Reference Williamson2020) as a snow-free ice surface (snow cover <2 cm), sufficient solar irradiance to drive photochemistry (shortwave radiation >10 W m−2) and the availability of liquid water (air temperature >0.5°C). In contrast, Onuma and others (Reference Onuma2022) used an ice temperature >0°C as a threshold for liquid water availability and did not define a solar irradiance threshold. However, there remains a lack of data to support the environmental conditions and values used as thresholds.
For example, the environmental conditions under which liquid water limits algal growth remain unknown. It has been estimated that glacier ice algae direct 48 to 65% of the incident irradiance to ice melting (Williamson and others, Reference Williamson2020), and thereby create their own liquid micro-environment, which would need to be considered for assessing a potential lack of liquid water for the algae (Fig. 2). Light availability may restrict photochemistry and subsequently growth during the beginning of the melt season when glacier ice algae are buried under a thick snow cover and, therefore, growth onset also depends on the timing of winter snowpack retreat (Williamson and others, Reference Williamson2018). Yet, algal activity measurements from below the snow cover are currently not available. Light availability is unlikely to be growth-limiting later in the season due to typically high incident solar irradiance and long day length duration, increasing with decreasing longitude. On the other hand, high incident irradiances in the middle of the season may potentially suppress glacier ice algal productivity, as demonstrated by in situ incubations that show the suppression of photochemistry at 100% ambient light during a mid-ablation season (Williamson and others, Reference Williamson2020). To define the thresholds for algal light requirements, incubation experiments or field measurements of glacier ice algal activity pre-and during snowline retreat, together with the light transmission through the snow are required. Sampling glacier ice algal populations from beneath the snow or the analysis of regression fits of their biomass progression throughout the season could additionally provide estimates of their initial population size, which is required to model algal bloom progression throughout the season (Williamson and others, Reference Williamson2020; Onuma and others, Reference Onuma2022).
So far, nutrient availability has not been incorporated into glacier ice algal growth models for the Greenland Ice Sheet (Fig. 2). McCutcheon and others (Reference McCutcheon2021) found that glacier ice algal growth is limited by phosphorous after long incubations (120 h) and suggested that the delayed response could be due to an internal phosphorus storage by the cells. No indication of nutrient limitation was found for short incubation times, suggesting that in situ nutrients can support algal nutrient requirements (Halbach, Reference Halbach2022). Moreover, the low intracellular carbon:nitrate:phosphorous ratios of algal dominated particulate organic matter (Williamson and others 2021) and single glacier ice algal cells (Halbach, Reference Halbach2022) suggest that they are well adapted to their oligotrophic environment. To fully comprehend the role of nutrients in glacier ice algal growth and their potential relevance for models, future studies should quantify algal nutrient uptake and storage (e.g. by using stable isotope tracers), as well as the competition and recycling of nutrients by other organisms of the community (e.g. by following examples from other aquatic systems in Klawonn and others, Reference Klawonn2019 or Adam and others, Reference Adam2016) (Fig. 2).
Controls on cell mortality are also currently understudied (Fig. 2). It has been shown that Chytridiomycota, a group of parasitic fungi, are widespread on the ice sheet and can infect algal cells (Perini and others, Reference Perini2019, Reference Perini2022; Fiolka and others, Reference Fiołka2021). Recently, it has been found that the prevalence of infection with chytrids among glacier ice algal populations from Alaska is 3.9 ± 5.6% on the bare ice (Kobayashi and others, Reference Kobayashi, Takeuchi and Kagami2023). However, the relative proportions of uninfected and infected cells on the Greenland Ice Sheet and a potential variability of infection prevalence throughout the season remain unknown. Ratios of viable, dead, infected or uninfected cells could be estimated by microscopy techniques, multi-stain, flow-cytometry and cell-sorting (Davey and Kell, Reference Davey and Kell1996; Klawonn and others, Reference Klawonn, Dunker, Kagami, Grossart and Van den Wyngaert2021a) as well as algal activity assessments (e.g. by Secondary-Ion-Mass-Spectrometry or BONCAT; Hatzenpichler and others, Reference Hatzenpichler2016; Gao and others, Reference Gao, Hutchins and Beardall2021; Klawonn and others, Reference Klawonn2021b, Grujcic and others, Reference Grujcic, Taylor and Foster2022).
Experiments under controlled conditions are needed to quantify growth and mortality specifically, isolated from other physical and hydrological redistribution processes (Fig. 2). Available glacier ice algal population doubling times using this approach amounted to 3.75 ± 0.36 days (net primary productivity measurements in the field Williamson and others (Reference Williamson2018)). It should be noted that any incubation of glacier ice algae in a liquid medium within bottles changes the growth conditions relative to in situ conditions (growth substrate, nutrient conditions, irradiance, temperature). Thus, in situ incubations in the field directly on the ice or short incubation times should be generally preferred to avoid or minimise extensive bottle effects. Laboratory cultures of glacier ice algae may allow the design and implementation of future experiments to study the role of microbial interactions on cell mortality and nutrient recycling (Fig. 2).
Physical and hydrological controls on the spatio-temporal variability of algal blooms
In addition to growth and mortality, it is hypothesised that glacier ice algal biomass distribution is also affected by hydrological processes, which can transport algal cells within supraglacial meltwater flow of the near-surface ‘weathering crust’ (see Cooper and others, Reference Cooper2018; Stevens and others, Reference Stevens2018) and supraglacial channels. Subsequently, such hydrological controls are likely to have important implications for albedo reduction (Fig. 2) (Stibal and others, Reference Stibal2017; Christner and others, Reference Christner2018; Tedstone and others, Reference Tedstone2020; Irvine-Fynn and others, Reference Irvine-Fynn2021; Stevens and others, Reference Stevens2022). However, the interplay between the retention of algae and their hydrological transport from the ice surface presents a key research avenue, with rates and controls upon biomass transport remaining poorly characterised.
Recent work has revealed the simultaneous advection of microbes and accumulation of microbial biomass throughout the melt season on the western Greenland Ice Sheet, with estimated bare-ice advection rates of microbes (<20 μm) for a typical melt season of 5.73 × 1012 cells km−2 d−1 (Irvine-Fynn and others, Reference Irvine-Fynn2021). However, the controls upon microbial abundance and, hence, transport mechanics of weathering crust meltwaters remain poorly defined, with no apparent link between depth-integrated hydraulic conductivity and microbial abundance (Stevens and others, Reference Stevens2022). Henceforth, more sophisticated approaches are required to fully understand and quantify the transport and retention dynamics of glacier ice algae (and by extension, all glacier microbes) from the ice surface, through the weathering crust and into the channelised glacial hydrological system. Ultimately, addressing this knowledge gap will help to elucidate the role of the supraglacial hydrological system in the spatio-temporal variability of algal blooms and their associated albedo reduction.
It has been suggested that the retention and transport of microbes from the surface to and within the saturated weathering crust system is size selective (Irvine-Fynn and others, Reference Irvine-Fynn2021; Stevens and others, Reference Stevens2022), with hypothesised controls including mechanical filtration by the weathering crust matrix (Mader and others, Reference Mader, Pettitt, Wadham, Wolff and Parkes2006), extra-cellular polymeric substances (e.g. Langford and others, Reference Langford, Hodson, Banwart and Bøggild2010; Holland and others, Reference Holland2019) or so-called hydrological ‘flushing’ (Adrien, Reference Adrien2004) with large melt events and/or weathering crust removal events. Future investigations should aim to ascertain the role of more precise hydrological metrics in partnership with algal concentrations and size distributions, specifically examining the following variables in depth-integrated and/or depth-specific manner: effective porosity of the weathering crust, direct weathering crust meltwater velocities, water delivery and drainage (from surface melt, internal melt and rainfall) and water table height. In addition, the application of tracer studies and/or high resolution 4D models at the micro-catchment scale, using the variables outlined above will help to ascertain algal mobility and near-surface residence periods beyond the point scale. Moreover, these variables are hypothesised to be highly temporally variable on a sub-hourly scale (Stevens, Reference Stevens2018), in response to diurnal and synoptic scale meteorological cycles, and should be examined under the widest possible range of antecedent and ongoing meteorological conditions. Physical, topographic and meteorological data can be gathered at different times and spatial scales from field studies, digital elevation models, automatic weather station networks (e.g. PROMICE: Fausto and others, Reference Fausto2021) or outputs from RCMs (e.g. MAR, Fettweis and others, Reference Fettweis2017; RACMO, Noël and others, Reference Noël2018). Links between these environmental data and changes in algal biomass distribution should be explored to unravel potential correlations between them.
Algal biomass can be assessed by field sampling or remote detection. Field samples can be analysed to retrieve detailed information about cell concentrations and community composition but can only represent very small areas within the very heterogeneous ice surface, which limits upscaling. Furthermore, it is common to retrieve surface ice using the adze of an ice axe or a trowel which, due to the very heterogeneous structure of the upper surface, can lead to uncertainty in the actual dimensions of the sampled area. It is also challenging to measure the specific surface area and effective grain radius in the field to configure a radiative transfer model to accurately represent a give sample area. However, new techniques for resolving the near-surface ice structure are emerging (e.g. Allgaier and others, Reference Allgaier2022).
Remote sensing enables us to study ice albedo and algal biomass over large spatial scales but changes in algal biomass can only be inferred indirectly using algorithms. Ground truthing these algorithms can be challenging due to the scale mismatch between satellite pixel size and field samples. Using hyperspectral images mounted on low-flying unmanned aerial vehicles might help to bridge that gap because the ground resolution of the resulting imagery is fine enough to match precisely to field sampling areas and can also be coarsened to match the footprint of satellite sensors. Several algorithms have been used to detect algal biomass. The simplest is based on Chlorophyll-a absorption (e.g. Painter and others, Reference Painter2001; Ganey and others, Reference Ganey, Loso, Burgess and Dial2017), however, they rely on a high spectral resolution to identify ‘biologically unique’ reflectance patterns. There have also been several papers that used classification algorithms such as K-nearest neighbours (Ryan and others, Reference Ryan2018), random forest decision trees (Cook and others, Reference Cook2020) and optimal estimation (Bohn and others, Reference Bohn2022). Further, algal growth has also been estimated using the ratio of reflectance at two red wavelengths as a proxy for algal concentration at the scale of the entire Greenland Ice Sheet ablation zone (Wang and others, Reference Wang, Tedesco, Xu and Alexander2018, Reference Wang, Tedesco, Alexander, Xu and Fettweis2020) and for Alpine glaciers (Di Mauro and others, Reference Di Mauro2020). To date, the main limitation of all these approaches has been the small training dataset sizes, but with empirical algal optical properties now incorporated into radiative transfer models (Chevrollier and others, Reference Chevrollier2022) there is potential to train these algorithms on very large synthetic datasets that cover the complete range of weathering crust conditions and biomass concentrations. By informing algal detection algorithm development and validation, improvements to both radiative transfer models and field sampling techniques will lead to more accurate quantification of glacier algal spatio-temporal dynamics and albedo impacts at the ice-sheet scale.
Indirect darkening effect of algal blooms
The state of the weathering crust is further an important control of surface albedo (Tedstone and others, Reference Tedstone2020; Fig. 2). Glacier ice algae can indirectly impact surface albedo by modifying the weathering crust surface (stimulating ice melting), in addition to their direct impact on albedo through shortwave radiation absorption (Cook and others, Reference Cook2017; Williamson and others, Reference Williamson2019). Thereby, glacier ice algae influence weathering crust decay by enhancing surface lowering relative to subsurface melting (Fig. 2), which creates a denser, less porous crust (Schuster, Reference Schuster2001; Cook and others, Reference Cook2020; Tedstone and others, Reference Tedstone2020), reducing the albedo. The indirect effect of algal blooms on weathering crust decay may be one of the reasons why algal abundance alone does not explain the discrepancy between MAR and MODIS albedo for the southwestern margin of the ice sheet in the study by Wang and others (Reference Wang, Tedesco, Alexander, Xu and Fettweis2020). However, our understanding of the indirect albedo-reducing effect of algal blooms through weathering crust densification is currently limited by a lack of empirical data and validated weathering crust development models.
Simultaneous monitoring of algal abundance and weathering crust physical properties through field samplings, experiments or remote sensing could help to elucidate these feedbacks. These physical properties include for example the density of near-surface ice (defined as ice to a 2 m depth), which describes the weathering crust growth and decay (Schuster, Reference Schuster2001). The development of algorithms for predicting weathering crust physical properties from remote sensing data has so far been restricted. This is due to a ‘many-to-one’ problem, where the same spectral albedo can be generated with different combinations of ice density and effective bubble sizes, and it is impossible to assess the correct combination. For snow, effective grain sizes can be inferred from spectral measurements (Nolin and Dozier, Reference Nolin and Dozier2000), but it is not clear that this method transfers to glacier ice. To address this, model inversions could be performed on many field spectra that have associated density profile measurements to establish a relationship between ice density and effective bubble size. This could enable remote quantification of weathering crust physical properties simultaneously with detecting algal cells.
To validate and improve weathering crust development models (Schuster, Reference Schuster2001; Woods and Hewitt, Reference Woods and Hewitt2022), empirical datasets are needed combining simultaneous measurements of weathering crust physical properties, meteorological parameters and surface albedo through time (Fig. 2). Eventually, weathering crust models could be coupled to BioSNICAR or other radiative transfer models to predict the indirect darkening role of algal blooms, as other studies have done for LAPs in snow (Tuzet and others, Reference Tuzet2017; Huang and others, Reference Huang, Qian, He, Bair and Rittger2022).
Conclusions
The presented comprehension of research priorities demonstrates the need for an interdisciplinary approach combining biological, hydrological and geophysical tools to understand the supraglacial algal system and its implications for albedo reduction. Key research questions that should be addressed in future studies are:
(i) under which conditions does liquid water limit glacier ice algal growth at the micro-scale?
(ii) what is the viable and dead fraction of glacier ice algal populations throughout the season and how do environmental factors and microbial interactions (e.g. parasitic infections by fungi) impact those?
(iii) what controls the rate of advection of algal cells from the ice surface to and through the hydrological system in the near-surface weathering crust?
(iv) what is the impact of algal blooms on the state of the weathering crust?
Addressing these questions about the biological darkening of the Greenland Ice Sheet requires the combination of mechanistic and descriptive study approaches. The mechanistic approaches are essential to isolate the controls on algal growth and mortality and may include laboratory and/or field experiments (e.g. bottle incubations or in situ experiments as done on large scales by Ganey and others, Reference Ganey, Loso, Burgess and Dial2017). These need to be complemented with observations in the field to validate the interactions under ‘true’ in situ conditions during complex feedbacks with the weathering crust and other environmental processes of the algal habitat (Fig. 2). In addition, net algal biomass assessments by remote sensing must be combined with environmental data to identify the interaction between the algal-driven darkening and environmental variables at a larger scale. The assessments of the supraglacial algal system from a micro-to-macro-scale will improve its parametrisation within models and enable a better understanding and prediction of algal bloom development and associated albedo decline over space and time.
Acknowledgements
The presented work is part of the project DeepPurple which has received funding from the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation programme (Grant Agreement No. 856416). Alexandre Anesio and Martin Hansen received support from the Aarhus University Research Foundation (grant numbers AUFF-T-2017-FLS-7-4 and AUFF-2018). Liane G Benning acknowledges the support of the Helmholtz Recruiting Initiative (grant no. I-044-16-01).
Author contributions
LH and LC wrote together a major part of the manuscript. JC, IS, AA, LB and MT contributed to the discussion and helped to draft the final version of this manuscript.
Open access
