What color should glacier algae be? An ecological role for red carbon in the cryosphere

Abstract

Red-colored secondary pigments in glacier algae play an adaptive role in melting snow and ice. We advance this hypothesis using a model of color-based absorption of irradiance, an experiment with colored particles in snow, and the natural history of glacier algae. Carotenoids and phenols—astaxanthin in snow-algae and purpurogallin in ice-algae—shield photosynthetic apparatus by absorbing overabundant visible wavelengths, then dissipating the excess radiant energy as heat. This heat melts proximal ice crystals, providing liquid-water in a 0°C environment and freeing up nutrients bound in frozen water. We show that purple-colored particles transfer 87%–89% of solar energy absorbed by black particles. However, red-colored particles transfer nearly as much (85%–87%) by absorbing peak solar wavelengths and reflecting the visible wavelengths most absorbed by nearby ice and snow crystals; this latter process may reduce potential cellular overheating when snow insulates cells. Blue and green particles transfer only 80%–82% of black particle absorption. In the experiment, red-colored particles melted 87% as much snow as black particles, while blue particles melted 77%. Green-colored snow-algae naturally occupy saturated snow where water is non-limiting; red-colored snow-algae occupy drier, water-limited snow. In addition to increasing melt, we suggest that esterified astaxanthin in snow-alga cells increases hydrophobicity to remain surficial.

adaptive coloration, glacier biology, red-snow algae, ice algae, water limitation, glacier microbiome, photo-physiology, astaxanthin, purpurogallin

INTRODUCTION

Photosynthetic microbes live on glacier surfaces in Arctic, alpine and Antarctic environments at densities sometimes visible from satellite imagery (Takeuchi et al. 2006; Ganey et al. 2017). Recent correlative (Yallop et al. 2012; Takeuchi et al. 2015; Lutz et al. 2016; Stibal et al. 2017), theoretical (Lutz et al. 2016; Musilova et al. 2016; Cook et al. 2017a,,b) and experimental studies (Musilova et al. 2016; Ganey et al. 2017) demonstrate an important role for these organisms in reducing glacier albedo. Albedo reduction in turn increases glacier melt, and so, by implication, warms climate and raises sea level. Like all near-freezing ecosystems, glaciers are liquid-water limited (Hodson et al. 2008; Anesio and Laybourn-Parry 2012; Stibal, Šabacká and Žárský 2012a), and the process of radiative forcing on their surfaces provides the melt needed by organisms living there. Resident populations of snow-algae have been shown experimentally to increase in abundance with addition of liquid-water, as well as to cause melt (Ganey et al. 2017). Thus, natural selection should favor those cells that melt the most water, conditional on the avoidance of cell overheating. Nevertheless, few studies acknowledge the possibility that physiological mechanisms for melting snow and ice are adaptive in glacier algae. Gorton, Williams and Vogelmann (2001), observing absorption in the far red (750–800 nm) for Chlamydomonas nivalis cells, inferred that absorption there likely did not drive photochemistry but instead ‘might cause warming and could thus be beneficial to an alga living in the snow’.

We hypothesize that the observed pink, orange and red colors of snow due to snow-algae and the gray and purple tint of ice due to ice-algae (Remias 2012) represent a directly functional, adaptive role for pigments in the melting of snow and ice crystals because these pigments provide liquid-water in a water-limited but well-illuminated environment. This study answers the question ‘What color should snow and ice algae be?’ by modeling the energy transfer of the spectral irradiance of the environment as a function of particle color and the spectral albedo of snow. Using the calculated radiative effects of variously colored particles and the results of a simple field demonstration with colored chalk dust in snow, we interpret cell color in the context of algal survival and reproduction in a nutrient-poor, frozen and strongly irradiated environment.

Glacier surface algae

Photosynthetic microbes occupy three distinct glacier habitats: snow (Lutz et al. 2016), bare ice (Yallop et al. 2012; Stibal et al. 2012b) and water-filled cryoconite holes (Takeuchi 2001; Hodson et al. 2010). The dominant eukaryotic algae of snow and bare ice habitats differ in life history, cell shape and pigmentation as fundamentally as bare ice differs from snow. Glacier algae preferentially settle on mineral particles (Takeuchi 2001; Remias, Lütz-Meindl and Lütz 2005; Stibal et al. 2017), and these structures may provide oases for consortiums of heterotrophic bacteria and fungi (Thomas and Duval 1995; Remias, Lütz-Meindl and Lütz 2005), as well as N-fixing cyanobacteria. Most previous studies of glacier algae indicate protection against overillumination as the primary role of their secondary pigments (reviewed in Remias 2012)—snow-algae's red-colored carotenoid, astaxanthin (C₄₀H₅₂O₄) and ice algae's brown-colored phenol, purpurogallin glucopyranoside (C₁₈H₁₈O₁₂). These extra-chloroplast pigments absorb the peak irradiance wavelengths of blue and green light, screening thylakoid membranes from light stress (Bidigare et al. 1993; Gorton, Williams and Vogelmann 2001; Remias, Lütz-Meindl and Lütz 2005; Leya et al. 2009; Remias et al. 2011; Remias 2012). What happens to the light energy these pigments absorb has yet to be directly addressed. Here, we explicitly assume that the secondary pigments concentrated in cytoplasmic lipid bodies, located between the photosynthetic plastids and the cell wall, absorb wavelengths <550 nm and conduct the absorbed energy as heat outwards where it can melt proximal snow or ice. This liquid-water increases fitness of individual algal cells. We also speculate that within snow-algae cells, the esterification of astaxanthin with fatty acids reduces overall intracellular density and allows cells to remain surficial, rather than melting into the snowpack or ice as seen with dense mineral particles and hydrophilic compounds such as brown carbon.

Snow-algae: Flagellated green algae in the Chlorophyta colonize glacier snow as an unresolved number of species, mostly among several genera in the Chlamydomonadales (Chlamydomonas, Chloromonas and Chlainomonas; Remias et al. 2016). A three-stage life cycle defined by cell mobility, activity and location in the snowpack characterizes snow-algae (Hoham and Duval 2001). Mostly visible as a red-colored but non-motile stage, snow-algal ‘hypnoblasts’ photosynthesize on the summer surface (Williams, Gorton and Vogelmann 2003; Remias, Lütz-Meindl and Lütz 2005). With no liquid-water or light available when buried beneath winter snowfall, snow-algae cease photosynthesis and overwinter in their second life stage: dormant ‘cysts’ that can also act as passive propagules. Each cyst potentially germinates in spring as multiple, green-colored ‘swarmers’ that swim through the melt-water inside the snowpack, dividing, photosynthesizing and absorbing nutrients in their third stage. All three stages, but particularly the astaxanthin-pigmented hypnoblast and cyst stages of Chlamydomonas (Remias, Lütz-Meindl and Lütz 2005) and Chlainomonas (Remias et al. 2016), are spheroid cells.

Ice-algae: Non-motile green algae in the Charophyta colonize bare-ice and firn surfaces, mostly as members of three genera in the Zygnematophyceae (Mesotaenium, Ancylonema and Cylindrocystis; Kohshima, Seko and Yoshimura 1993; Remias, Andreas and Lütz 2009; Uetake et al. 2010; Yallop et al. 2012; Stibal et al. 2017). In contrast to the snow-algae of the Chlamydomonadales, the Zygnematophyceae—as seen in Ancylonema nordenskiöldii (Remias et al. 2012a) and Mesotaenium bergrenni (Remias et al. 2012b)—display a single-stage life cycle of dark-colored, non-motile, elongated cells lacking astaxanthin (Remias, Andreas and Lütz 2009). Populations of these algae visibly tint ice surfaces gray or purple due to their abundant phenol, purpurogallin. This pigment fills vacuoles between the cell membrane and chloroplasts providing broad-spectrum absorption of photosynthetically active radiation (PAR).

The abiotic glacier environment

Because the glacier surface environment is generally near freezing through time and across space, it is water-limited, oligotrophic and overly illuminated (Hodson et al. 2008; Anesio and Laybourn-Parry 2012; Stibal, Šabacká and Žárský 2012a).

Water-limited: Even during the melt season, glacier surfaces are often water-limited for the organisms that live there (Stibal, Šabacká and Žárský 2012a) because water in its solid state is metabolically unavailable. Cold snow (e.g. below –5°C) is xeric, and snow-algae can neither colonize with flagellated swimmers nor germinate their durable cysts in the absence of liquid-water (Hoham and Duval 2001). Because of surface lapse rates, higher elevations are colder and so drier, thus establishing an upper elevation limit to snow-algae's distribution, a limit rising with warming climate (Dial et al. 2016b). Even where summer temperatures reach above freezing and melt leaves sufficient liquid-water for snow-algae's germination, cell-division, photosynthesis and colonization, the surface of a snowpack, unlike its interior, can be subject to daily cycles of drought when liquid-water freezes nightly as a crust (Hoham 1975; Remias et al. 2005). Moreover, repeated freeze-thaw metamorphism leaves snow crystals sintered and large-grained, diminishing the field capacity of the snowpack's surface to hold liquid-water as it melts (Kohshima, Seko and Yoshimura 1993). This water limitation has been revealed through experimental addition of water that can increase snow-algae abundance by nearly 50% (Ganey et al. 2017).

Oligotrophic: Like alpine lakes, glaciers and ice sheets are nutrient-poor (Hodson et al. 2008; Anesio and Laybourn-Parry 2012), often with little detectable correlation between alga abundance and nutrient concentration in snow (Lutz et al. 2016). In Alaska, experimental addition of nutrients increased snow-algae by a factor of nearly 4 (Ganey et al. 2017). Cyclonic storms may deliver marine-derived nutrients to coastal northwest American glaciers, as suggested by their rich biota of red-snowfields grazed by ice-worms (e.g. Enchytraidae: Mesenchytraeus solifugus; Goodman 1971; Dial et al. 2016a), just as sea birds are hypothesized to deliver nutrients in Antarctica (Fujii et al. 2010). Compounding the problem of inadequate mineral nutrition on glaciers is the strong likelihood that ice and snow crystals bind nutrients until melted-out, amplifying the importance of liquid-water on glaciers.

Overly illuminated: Glacier surfaces are strongly irradiated due to an absence of vegetative structure and snow's high photon fluency, a measure of light from all directions due to scattering in the snowpack, which is several times (≥2x) greater than photon irradiance (Gorton, Williams and Vogelmann 2001; Yallop et al. 2012). The spectral albedo of snow depends on snow grain size and wavelength of light (Wiscombe and Warren 1980; Fig. 1, dashed and dotted lines). Thus, the glacier surface is often oversaturated with PAR and underprotected from UV. At the large snow grain size (∼1 mm) of summer glacier surfaces occupied by snow-algae, >95% of purple through green light (350–500 nm) is reflected by snow. Snow spectral albedo, α(x), drops with wavelength (x) as a steepening curve (Fig. 1) to <30% by the near infrared (x = 1000 nm). Ignoring transmission, the complement of albedo approximates absorption as β(x) = 1 – α(x). In the visible spectrum, absorption by snow climbs steadily with increasing wavelength (Fig. 1, solid line). Snow impurities, like black carbon (Kirchstetter, Novakov and Hobbs 2004), dust (Skiles et al. 2012), glacier algae (Kohshima, Seko and Yoshimura 1993) and other organics reduce albedo, absorb visible light, and conduct the absorbed radiative energy as heat into the glacier, melting ice and snow. An impact of melt is that the presence of liquid-water enhances the rate of grain growth very efficiently, further reducing albedo.

Figure 1.

Snow grains of all sizes absorb little radiant energy at wavelengths below 500 nm. Theoretical spectral albedo, α, for clean snow by snow grain size and its complement spectral absorption, β(x) = 1 – α(x), for 1 mm grain size at wavelength x.

MATERIALS AND METHODS

Spectral irradiance (Fig. 2) of wavelength x at time t, SI(x,t), is the power density (W m⁻² nm⁻¹) of solar radiation and depends also on the atmospheric conditions, location and topography of the Earth's surface where sunlight strikes. We explore the role of colored particles in absorbing this irradiance and melting snow and ice through a simple energy transfer model and a field experiment with colored particles deposited on snow.

Figure 2.

Spectral irradiance (solid lines) and absorbed irradiance by 1 mm grain snow (dotted lines) on 14 July 2017 at 60°N, 150°W at (a) mid-day (1200 hours) and (b) early morning (0500 hours). Irradiance values are from SPCTRAL2 model (Bird and Riordan 1986) downloaded from https://www.pvlighthouse.com.au for a plane perpendicular to radiation (perpendicular plane).

A model for the role of color in melting snow and ice

Model assumptions: We assume that a particle of a given color absorbs visible wavelengths other than those wavelengths reflected as its color. The absorbed wavelength energy is then conducted as heat into the ice or snowpack. For the simple treatment here, we ignore the energy absorbed in photosynthesis. We further assume that additional energy is absorbed by the ice or snowpack due to the particle's reflected light. This reflected radiative forcing can be calculated using the product of the spectral absorption of ice and snow (β(x) as one minus albedo) with the spectral irradiance of the particle's reflected wavelengths; that is, using β(x) SI(x,t). The model views a particle of a given color as transferring light energy into a glacier through two pathways: first through conduction of light energy absorbed by the particle and second through radiative forcing of reflected light into snow and ice. Their sum yields the energy generated by a particle's color available to melt proximal frozen water. For a photosynthetic particle, the energy consumed in photosynthesis would be subtracted; here, however, we ignore this variable quantity to explore the role of color energy alone.

Southcentral Alaska's 1900 km² Harding Icefield (60°N 150°W) has been the site of two snow-algae studies (Takeuchi et al. 2006; Ganey et al. 2017) that apply band ratios of red and green light as remote sensors of alga abundance. We used modeled (SPCTRAL2, Bird and Riordan 1986; https://www.pvlighthouse.com.au) visible spectral irradiance (350–750 nm) at 1 nm intervals on 14 July 2017 at hourly intervals from dawn (0400 hours) to dusk (2000 hours) at 60°N 150°W, where mid-day spectral irradiance peaks in the energetic blue-green portion of the visible spectrum (≈500 nm; Fig. 2).

We accepted the other defaults of the SPCTRAL2 model, using both the perpendicular-to-incoming solar radiation plane and perpendicular-to-gravity (horizontal) plane for global spectral irradiance. The modeled incoming spectra assume clear skies and sum both direct and diffuse irradiance as a total (i.e. “global spectral irradiance”). Moreover, the model here applies only to the uppermost surface of the glacier; we do not attenuate irradiance according to depth or shading, but visualize each particle as a sphere on the surface. We used spectral irradiance at this single day and single place to calculate (1) the total energy in the visible spectrum (350–750 nm) and (2) the energy in every 64-nm interval of the visible spectrum between 350 and 750 nm at (3) both a plane perpendicular to solar radiation and a plane perpendicular to gravity. We chose 64 nm because we consider six colors of uniform banding across the 350–750 nm range (6 colors × 64 nm color⁻¹ = 384 nm).

Model calculations: Let SI(x,t) be the value of ‘spectral irradiance’ in W m⁻² nm⁻¹ at wavelength x and time t. Then, the area under SI(x,t) between two visible wavelengths (defined as ‘color’ λ bound by λ₁ and λ₂) gives the instantaneous power density for that color as SIA(λ,t) in W m⁻². Let BLACK(t) be the total area beneath the SI(x,t) curve in the visible interval at time t. That is, the instantaneous power density (W m⁻²) absorbed in the visible spectrum by an all-black particle is

\begin{eqnarray*} \it{BLACK} (t) &=& \mathop \int \limits_{350}^{750} SI\left( {x,t} \right)dx \\ &\approx & \mathop \sum \limits_{x\ = \ 351}^{749} SI ({x,t}) \Delta x + \ \frac{{\Delta x}}{2}\left[ {SI\left( {350,t} \right) + \ SI\left( {750,t} \right)} \right], \end{eqnarray*}

where Δx = 1 nm and the sum approximates the integral using the trapezoid rule. Similarly, SIA(λ,t), the power density reflected by a particle of color centered at λ with bandwidth of 64 nm (chosen as 64 nm color⁻¹ × 6 colors ≈ 400 nm = 750–350 nm), is

\begin{eqnarray*} \it{SIA} ({\lambda ,t}) &=& \mathop \int \limits_{\lambda - 32}^{\lambda + 32} SI({x,t})dx \\ & \approx & \mathop \sum \limits_{x\ = \ \lambda - 31}^{\lambda + 31} SI\left( {x,t} \right)\Delta x + \frac{{\Delta x}}{2}\left[ {SI\left( {\lambda - 32,t} \right) + \ SI\left( {\lambda + 32,t} \right)} \right] \end{eqnarray*}

so that IRF(λ,t) = BLACK(t) – SIA(λ,t) gives the instantaneous radiative forcing (W m⁻²) of a colored particle with bandwidth of 64 nm centered at λ at time t.

Integrating IRF(λ,t) over the day (dawn ≤ t ≤ dusk) yields solar ‘irradiation’ for that day and color λ as IRR(λ), a measure of energy density (J m⁻²):

\begin{eqnarray*} \it{IRR}\ \left( \lambda \right) &=& \mathop \int \limits_{dawn}^{dusk} \it{IRF}\left( {\lambda ,t} \right)dt \approx \mathop \sum \limits_{dawn + 1}^{dusk - 1} \it{IRF}\left( {\lambda ,t} \right){\rm{\Delta }}t + \frac{{{\rm{\Delta }}t}}{2}\nonumber\\ &&\times\,\left[ {\it{IRF}\left( {\lambda ,dawn} \right) + \it{IRF}\left( {\lambda ,dusk} \right)} \right]. \end{eqnarray*}

Integrating BLACK(t) from dawn to dusk gives the solar irradiation absorbed by an all-black particle over the day. Dividing irradiation, IRR(λ), by the latent heat of fusion for melting 0°C ice (Ω = 334 J cm⁻³), and scaling with 1 m² = 10⁴ cm² converts the units of IRR (λ) into cm of water equivalent (w.e.). This gives a measure of melt-water production due to a particle color's absorption of spectral irradiance over the course of a day and subsequent conduction into the snowpack. Because we used spectral irradiance modeled at hourly intervals (Δt = 1 h), we multiplied by 3600 s h⁻¹ to convert to joules. Heat conducted as w.e. by color λ bounded by λ₁ = λ – 32 and λ₂ = λ + 32 is thus (trapezoidal approximation not shown here but applied in Supplementary Material),

\begin{eqnarray*} \ C{\left( {\rm{\lambda }} \right)_{w.e.}} &=& \frac{{0.36}}{{\rm{\Omega }}}\ \mathop \int \limits_{{dawn}}^{{dusk}} \left(\,\,\, {\mathop \int \limits_{350}^{750} SI\left( {x,t} \right)dx - \mathop \int \limits_{\lambda - 32}^{\lambda + 32} SI\left( {x,t} \right)dx} \right)dt\nonumber\\ &&\approx\, \frac{{3600}}{{{{10}^4}{\rm{\Omega }}}}\mathop \sum \limits_{t\ = \ dawn}^{dusk} \mathop \sum \limits_{\ x\ = \ \lambda - 32}^{\lambda + 32} \left[ {\it{SIA} - SI\left( {x,t} \right)} \right]\Delta x\Delta t, \end{eqnarray*}

where Δx = 1 nm and Δt = 1 h.

To calculate the spectral irradiance energy absorbed directly by the snowpack due to a particle's reflection of color λ, multiply spectral irradiance by β(x), the spectral absorption of ice = 1 – α(x) at wavelength x (ignoring transmittance), then doubly integrate over wavelengths in color λ and all daylight time, divide by Ω and scale to arrive at w.e. due to reflection (trapezoidal approximation not shown here but applied in Supplementary Material),

\begin{equation*} {R_{w.e.}}\left( \lambda \right)\ = \frac{{0.36}}{{\rm{\Omega }}}\ \mathop \int \limits_{dawn}^{dusk} \mathop \int \limits_{\lambda - 32}^{\lambda + 32} \beta \left( x \right)SI\left( {x,t} \right)dx\,dt. \end{equation*}

We used R (vers. 3.3.1) to calculate Total_w.e(λ) = C_w.e.(λ) + R_w.e.(λ) for all integer values of visible λ (350 nm + 32 nm = 382 nm ≤ λ ≤ 718 nm = 750 nm – 32 nm) to identify the color λ that maximizes Total_w.e..

Model caveats: The absolute results modeled here depend on the chosen color bandwidths and range of visible light, here set at 350–750 nm with 64 nm bandwidths to capture six familiar colors (purple, blue, green, yellow, orange, red). The model treats the particle's optical property as a simplified, uniform spectral reflectance in the form of a step function between zero and one across the 64 nm bandwidth; such reflectance likely does not exist in any natural pigment. The model assumes the particle is spherical, but makes no special adjustments for calculations based on that geometry. As a sphere and by the law of reflection, perhaps half of the incident light will be reflected into the atmosphere, with half reflected into the nearby snowpack. However, field measurements of surface photon fluency rates are at least double those of irradiance (Gorton, Williams and Vogelmann 2001), so these fluent photons could also reflect the given color into the nearby snowpack. For the treatment here, we ignore both the half reflected back into the atmosphere and any fluent photon reflection, assuming as a first approximation that the two are equal.

Field experiment

We performed a simple field study during early April 2014 involving the application of colored chalk dust atop a 35–50 cm snowpack on an open lawn at 61°N in Anchorage, Alaska (60 m asl). After sundown on 2 April, we applied 35 ml of carpenter's marking chalk (Keson Industries Incorporated brand, ProChalk model PM8Black as 70%–85% CaCO₃ colored with 15%–30% Aeosperse5 carbon black; model 8R as 85%–90% CaCO₃ colored with 10%–15% red iron oxide; and model 8B as 85%–90% CaCO₃ colored with 10%–15% ultramarine blue; www.keson.com) to each of 36 squares (1600 cm²) using a flour sifter. This amount was likely far more than a natural density of algae. The 36 squares were distributed among four experimental blocks as three replicates per block of each color: black, red and blue (Fig. 3c). Particles melted naturally into the snow over the course of several days. Twenty-four hours after application, we estimated melt with the proxy of surface lowering, measured as the depth at the lowest point in each experimental square below the uncolored edges of the square. Five days after application (7 April 2014), we photographed one replicate of the three colors.

Figure 3.

Theory and experimental set-up for snow melt by particle color. Global (direct + diffuse) spectral irradiance on 14 July 2017 at 60°N 150°W absorbed by 1 mm grain size snowpack perpendicular to gravity (a) and perpendicular to irradiance (b). Symbol colors indicate particle color as defined by wavelengths. Circles connected by dots give energy converted to water equivalents through conduction only when a particle absorbs the complement of its color and conducts absorbed heat into the snowpack. Circles connected by solid lines give total energy absorbed by snowpack through both conduction of absorbed heat and reflection of wavelengths into snowpack. (c) Set-up (3 April 2014) for demonstration that red-colored particles melt more snow than blue-colored ones using carpenter's chalk in April snow in Alaska. (d) Experiment after several 5 days (7 April 2014) showing black particles melt to soil first, then red, with blue particles melting much less snow. Particle concentration exceeds likely snow-algae abundance in nature.

No precipitation fell during the experiment. Temperatures underwent daily freeze-thaw cycles each day but one (Merrill Field weather station, 4.1 km NW and 20 m lower; min daily temperature: –6.7° to 1.1°C; max daily: 4.4° to 8.9°C), under mostly sunny skies. We did not measure the reflectance spectra of the chalk dust, control for salts with a no-pigment chalk, nor report measurements other than the first day record of surface lowering.

To compare red to blue melt, we averaged the surface lowering for the three squares of each color in each of the four blocks, then for each block we divided mean red and blue surface lowering by mean black surface lowering and multiplied by 100, giving melt for red and blue as a percentage of black.

RESULTS

Overall, both the model (Fig. 3a and b) and the field experiment (Figs 3d and 4) suggest that red-colored particles will melt more than blue-colored particles, relative to black-colored particles. Theoretical results show that purple-colored particles melt the most, both when the particle is positioned on a horizontal surface (Fig. 3a) and when on a sloped surface that is perpendicular to irradiance (Fig. 3b). Importantly, the addition of reflected radiative forcing increases the total melt potential of red-colored particles to near that of purple-colored particles. Blue-colored particles melt the least when positioned horizontally (Fig. 3a) and green-colored particles melt the least when perpendicular to irradiance (Fig. 3b).

Figure 4.

Snow melt 24 h following chalk application. Melt due to red-colored chalk dust exceeds that of blue-colored chalk dust when both are relative to melt by black-colored chalk dust and does so near that predicted by the color melt model of Fig. 3a and b. Each solid bullet is the mean of the given color surface lowering in a block (n = 3 samples per color per block) divided by the mean surface lowering of the black samples in the same block, multiplied by 100. Dashed lines link color samples from the same experimental block (n = 4 blocks) and open circles gives median value by color. Expected lowering shows the range of lowering for red and blue particles from Fig. 3a and b, where 64 nm bandwidths (λ – 32 ≤ λ ≤ λ + 32) define colors red (λ = 718 nm) and blue (λ = 462 nm), which may or may not represent the reflectance spectra of the chalk dust.

Light absorbing color theory

Consider the visible spectrum (350–750 nm) uniformly divided into six colors, λ: purple = 398 nm, blue = 462 nm, green = 526 nm, yellow = 590 nm, orange = 654 nm and red = 718 nm, each centered in 64 nm bandwidths (λ – 32 ≤ λ ≤ λ + 32). Among these colors (Fig. 3a and b), purple- and red-colored particles melt the most snow and ice through absorbed irradiance (>85% of black melt) and blue and green melt the least (<82% black melt). The difference in melt between red- and purple-colored particles depends on the incidence angle of the incoming radiation relative to the particle (Fig. 3a and b). At low incidence angles, red-colored particles transfer 85% the energy of all-black particles into the melting snow and differ most from purple-colored (89% of all-black particle energy; Fig. 3b solid curve); at greater incidence angles, such as particles positioned horizontally (i.e. perpendicular to gravity), red- and purple-colored particles each melt ≈87% as much as an all-black particle (Fig. 3a solid curve).

Purple-colored particles do not reflect wavelengths that are absorbed by snow and ice, whereas red-colored particles do (Fig. 1). This increases the melting energy (solid curves in Fig. 3a and b) for red over its absorbed energy (dotted curves in Fig. 3a and b), but not for purple. Purple (87%–89% of black) and red-colored particles (85%–87%) potentially melt near equivalent amounts of snow or ice, but when insulated (as in snow), the temperature of a red particle will likely remain lower than a purple particle, because the red particle melts nearby snow both through conduction, which heats the particle, and by reflection, which does not.

Among particles positioned on a horizontal surface, the lowest melt is produced in a 64-nm band centered at λ = 482 nm (Fig. 3a), which we identify as blue to green using https://academo.org/demos/wavelength-to-colour-relationship/. When perpendicular to irradiance, the least energy is absorbed in the 64-nm band centered at λ = 553 nm (Fig. 3b), which we identify as green to yellow.

Field experiment

Results from the field application of black-, red- and blue-colored marking chalk to snow were consistent with theoretical calculations of the melt rate of red and blue relative to all-black particles. As expected, all-black particles melted to the soil surface first, followed by the red-colored particles; the experiment ended before the blue particles melted to the soil surface (Fig. 3d). Besides the transfer of light energy as heat conducted into the snow and causing melt, it is possible that long-wave radiation from the particles contributed to melt in the same ordering, as suggested by the sun cups produced during the experiment. Measuring surface lowering ∼24 h after the initial dusting in each block and averaging over color within blocks, then dividing the red and blue mean by black mean lowering, gave the percentage of black chalk lowering by red and blue (Fig. 4). The median lowering among the four blocks for red chalk was 87% (range 85%–90%) and for blue 77% (range 69%–82%) of the all black particles. These average values are surprisingly close to, but greater than, model predicted (Fig. 3a and b) red (85%–87%) and less than model predicted blue (80%–82%). The reflectance spectra for the chalk are unknown and unlikely represented by 62 nm wide step function as assumed by the simple model.

DISCUSSION

For glacier algae, spectral irradiance provides the energy available for both photosynthesis and melting frozen water. Because liquid-water and nutrients bound within ice crystals are more limiting than light on a glacier surface, spectral irradiance's prime function for photosynthetic microbes living there may be generating liquid-water for metabolism and nutrient mining, rather than for fixing carbon. Here, we have developed a simple model to explore the role of color in melting snow and ice and described an experimental demonstration to qualitatively test the model's predictions. In answer to ‘What color should glacier algae be?’, we reply, “Purple or red, but not green or blue”.

Black algal cells?

Very dark-colored impurities that absorb light uniformly, like black, brown and gray carbon (Kirchstetter, Novakov and Hobbs 2004), conduct the most heat to snow and ice. However, snow-algae occupy snow grain surfaces that can also be in contact with air, more so than ice-algae, because summer surface snow is less dense than surface ice (snow < 550 kg m⁻³ vs. ice ≈800 kg m⁻³) and the density difference is due to closed air spaces. If snow-algae cells were black, then they could potentially overheat when insulated within dry snow interstices, denaturing membrane and other proteins (Gates et al. 1965). In contrast, dark-colored ice-algae living on ice could cool through conduction more rapidly than dark-colored snow-algae living on surface snow, because ice's thermal conductivity is two orders of magnitude greater than air (2.2 W m⁻¹ K⁻¹ vs. 0.02 W m⁻¹ K⁻¹ at 0°C). So, in answer to the question ‘What color should ice-algae be?’, an answer based on energy transfer through radiative and conductive heat exchange might be ‘They should be dark with a large surface area:volume ratio that effectively conducts radiant energy to melt ice’.

Observations that ice-algae in the Zygnematophyceae are often cylindrical with a length up to twice their width, tinting the ice or firn gray or purplish with brown to dark-brown phenolic pigments (purpurogallin) that absorb relatively uniformly across the visible spectrum (Remias, Andreas and Lütz 2009; Remias 2012; Yallop et al. 2012) support the answer. Moreover, ice-algae cells contain pyrenoids, organelles present in aquatic algae and thought to play a role in sequestering CO₂ in low [CO₂] environments (Giordano, Beardall and Raven 2005), such as melt-water. Thus, a dark, non-spherical ice-alga cell is more likely to melt proximal ice than a lighter colored, spherical cell and, because it contacts ice or water that it melts, the cell's absorbed irradiant energy is rapidly conducted away, protecting the cell from overheating.

Red as an optimal color for snow-algae, purple for ice algae

Based on considerations of solar irradiance and particle color, we hypothesize that red is the color that snow-algae should be to melt snow without overheating and, if overheating is not a constraint, as perhaps for ice-algae, then purple melts an equivalent or greater amount of water. Thus, we advance a new function—melting ice and snow—for astaxanthin in snow-algae and for purpurogallin in ice-algae. While purple-colored snow-algae living in the coarse-grained substrate of snow may be subject to overheating in pore-spaces, purple-colored ice-algae on bare ice surfaces may be less likely to overheat. The absolute results presented here depend on the choice of both color bandwidths and range of visible light, here set as 64 nm bandwidths across 350–750 nm to capture six familiar colors (purple, blue, green, yellow, orange, red). However, as an example, using 60 nm bandwidths from 391–750 nm shifts the maximal energy from purple to red.

Previous studies (Bidigare et al. 1993; Gorton, Williams and Vogelmann 2001; Remias et al. 2005; Leya et al. 2009; Remias et al. 2011) describe the primary functional role of snow-algae's astaxanthin as protection against intense visible and UVA spectrum irradiation (but not UVB; Remias and Lutz 2007). Astaxanthin molecules (Thomas and Duval 1995; Gorton, Williams and Vogelmann 2001; Remias 2012), individual snow-alga cells (Bidigare et al. 1993; Gorton, Williams and Vogelmann 2001; Holzinger, Allen and Deheyn 2016) and their extracts (Gorton, Williams and Vogelmann 2001) show high absorbance (Thomas and Duval 1995; Gorton, Williams and Vogelmann 2001; Remias 2012), absorptance (Gorton, Williams and Vogelmann 2001) and absorption (Bidigare et al. 1993) of purple through green wavelengths (400–550 nm). Field spectra of high-density populations of snow-algae reflect far less purple through green wavelengths than low-density populations (Painter et al. 2001; Takeuchi et al. 2006; Ganey et al. 2017), as expected from laboratory measurements of absorption spectra from individual cells and their isolated pigments. These field and laboratory measurements all show relatively high reflectance and non-absorption for the red portion of the visible spectrum, consistent with the visible red of hypnoblasts.

Astaxanthin likely serves other cellular functions besides screening excess light and generating heat. For example, we speculate that the esterified form may increase cellular hydrophobicity to maintain an algal cell's position atop the snow surface. The esterified form also likely provides a storage molecule for encysted propagules (Remias et al. 2005) and acts as a cryoprotectant that displaces cell water (Bidigare et al. 1993). Boussiba (2000) suggests astaxanthin in the related Haematococcus pluvialis (Chlamydomonadales), a heat- and drought-tolerating species of ephemeral aquatic habitats (Proctor 1957), is a byproduct of a generalized defense mechanism against free-oxygen radicals generated by environmental stresses of all forms. Most studies attribute the change from green snow-alga swarmer to red hypnoblast as a response to nutrient depletion, particularly nitrogen (N), in the oligotrophic snowpack (Bidigare et al. 1993; Remias, Lütz-Meindl and Lütz 2005; Leya et al. 2009; Remias et al. 2011).

Increased astaxanthin benefits surface-dwelling hypnoblasts because surface snow is drier than lower in the snowpack, and surface hypnoblasts can produce and benefit more from liquid-water than subsurface green swarmers. The observation that green snow-algae inhabit wet snow (Fogg 1967; Mueller et al. 1998; Lutz et al. 2014) is consistent with this conclusion. Lutz et al. (2014) found that adjacent green- and red-colored snow in Svalbard differed in alga cell species, alga cell biomass, snow wetness and nutrient concentration, with the much more abundant green-colored algae inhabiting ‘naturally’ wet and nutrient-rich snow and the red-colored algae inhabiting dry, nutrient-poor snow. An exception to the dry-snow habitat of red-colored hypnoblasts is Chlainomonas, a red-snow algae that lives in wet environments (Remias et al. 2016).

Not all light energy absorbed by hypnoblasts is transferred as heat because some is captured in photosynthesis as chemical potential energy for doing cellular work, respiration, and reproduction and some fluoresces from pigment excitation. A full accounting of all absorbed energy in glacier algae has yet to be made.

Generalization

Secondary pigment coloration in photosynthetic organisms is rarely addressed in the context of heat. We have shown that red and purple can be the ‘hottest’ particle colors in oblique sunlight (Fig. 3a) and green the ‘coolest’ in direct sunlight (Fig. 3b). As argued here, a selective pressure in near-freezing environments may be for colors that generate heat, making red reflectance adaptive in freeze-thaw environments that are also well-illuminated. The hypothesis posed here may be profitably extended to deciduous low shrubs that turn red in autumn among alpine and subalpine habitats, while at lower elevations their tall shrub and tree congeners turn yellow (e.g. Betula). Moreover, while fruit and inflorescence colors in flowering plants have a well-developed natural history, the evolutionary reason that plant leaves are green remains an open question (Nishio 2000; Marosvölgyi and van Gorkom 2010). Clearly, insufficient energy receipt has not been the main selective pressure for photosynthetic organisms, and outside of near-freezing environments such as glaciers and ice sheets, energy dissipation is likely to have been a greater ecological challenge over evolutionary time. Green reflectance may help meet this challenge.

Pigments in non-photosynthetic organisms may also play a role in absorbing heat in cold environments. Among prokaryotes, red-colored cells of the bacterium Hymenobacter (Bacteroidetes: Cytophagia) have been described from various solar-illuminated, frozen environments (Klassen 2009) including soils (Oren 2006), glacier ice (Klassen and Foght 2008), and glacier snow (Fujii et al. 2010) in Antarctica; Arctic glacial till (Chang et al. 2014); and atmospheric hailstones (Alexander Michaud, personal communication), where microbes can act as an ice nucleator (Michaud et al. 2014). Absorption spectra of four carotenoids isolated from Hymenobacter show 475–500 nm peaks (Klassen 2009). On North American glaciers, the ice-worm (M. solifugus) is a black, nocturnal annelid, whose color may be an adaptation for heat absorption in a motile organism that can escape overheating by burrowing, and whose black color absorbs all wavelengths that penetrate the meter or so into the snowpack where they spend the day (Dial et al. 2016a).

CONCLUSION

We propose that one additional purpose of ‘red carbon’—astaxanthin for snow-algae and purpurogallin for ice-algae that are photosynthesizing on glacier-surfaces—is the production of locally available liquid-water in a water- and nutrient-limited, frozen environment. These pigments likely evolved to shield photosynthetic machinery from overillumination that is dissipated as heat, predisposing algae to colonize snow and ice habitats. Many evolutionary-ecology processes appear as ‘chicken-or-egg/cause-or-effect’ paradoxes, but may be better considered as compact solutions to parallel problems. Just as four walls support a roof and protect from wind, natural selection favors cellular products serving multiple purposes. In this case, the biological imperatives of algal survival and reproduction on glaciers and ice sheets make red carbon perhaps the most potent of light-absorbing impurities in the cryosphere.

SUPPLEMENTARY DATA

Supplementary data are available at FEMSEC online.

Acknowledgements

This paper was presented at the 7th International Conference on Polar and Alpine Microbiology in Nuuk, Greenland 8–12 September 2017. We acknowledge M Loso and A Burgess for critical discussions; appreciate SC Cary for pointing out that melting frees up nutrients and AB Michaud for mentioning red-colored bacteria in hailstones; thank J Zarsky for pointing out the value of walls over posts in supporting a roof; and acknowledge the questions, comments and suggestions of JM Cook, M Stibal and an anonymous reviewer.

FUNDING

This work was supported by National Aeronautics and Space Administration (NASA) Alaska Space Grants to RJD and GQG.

Conflict of interest. None declared.