Glacier algae accelerate melt rates on the south-western Greenland Ice Sheet

Abstract

Melting of the Greenland Ice Sheet (GrIS) is the largest single contributor to eustatic sea level and is amplified by the growth of pigmented algae on the ice surface, which increases solar radiation absorption. This biological albedo-reducing effect and its impact upon sea level rise has not previously been quantified. Here, we combine field spectroscopy with a radiative-transfer model, supervised classification of unmanned aerial vehicle (UAV) and satellite remote-sensing data, and runoff modelling to calculate biologically driven ice surface ablation. We demonstrate that algal growth led to an additional 4.4–6.0 Gt of runoff from bare ice in the south-western sector of the GrIS in summer 2017, representing 10 %–13 % of the total. In localized patches with high biomass accumulation, algae accelerated melting by up to 26.15±3.77 % (standard error, SE). The year 2017 was a high-albedo year, so we also extended our analysis to the particularly low-albedo 2016 melt season. The runoff from the south-western bare-ice zone attributed to algae was much higher in 2016 at 8.8–12.2 Gt, although the proportion of the total runoff contributed by algae was similar at 9 %–13 %. Across a 10 000 km2 area around our field site, algae covered similar proportions of the exposed bare ice zone in both years (57.99 % in 2016 and 58.89 % in 2017), but more of the algal ice was classed as “high biomass” in 2016 (8.35 %) than 2017 (2.54 %). This interannual comparison demonstrates a positive feedback where more widespread, higher-biomass algal blooms are expected to form in high-melt years where the winter snowpack retreats further and earlier, providing a larger area for bloom development and also enhancing the provision of nutrients and liquid water liberated from melting ice. Our analysis confirms the importance of this biological albedo feedback and that its omission from predictive models leads to the systematic underestimation of Greenland's future sea level contribution, especially because both the bare-ice zones available for algal colonization and the length of the biological growth season are set to expand in the future.

Full text

The Cryosphere, 14, 309–330, 2020
https://doi.org/10.5194/tc-14-309-2020
© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.
Glacier algae accelerate melt rates on the south-western
Greenland Ice Sheet
Joseph M. Cook1,2, Andrew J. Tedstone3, Christopher Williamson4, Jenine McCutcheon5, Andrew J. Hodson6,7,
Archana Dayal1,6, McKenzie Skiles8, Stefan Hofer3, Robert Bryant1, Owen McAree9, Andrew McGonigle1,10,
Jonathan Ryan12, Alexandre M. Anesio13, Tristram D. L. Irvine-Fynn11, Alun Hubbard14, Edward Hanna15 ,
Mark Flanner16, Sathish Mayanna17 , Liane G. Benning5,17,18, Dirk van As19 , Marian Yallop4, James B. McQuaid5,
Thomas Gribbin3, and Martyn Tranter3
1Department of Geography, University of Sheffield, Winter Street, Sheffield, South Yorkshire, S10 2TN, UK
2Institute of Biological, Environmental and Rural Sciences, Aberystwyth University, Aberystwyth, SY23 3DA, UK
3Bristol Glaciology Centre, School of Geographical Sciences, University of Bristol, Berkely Square, Bristol, BS8 1RL, UK
4School of Biological Sciences, University of Bristol, Tyndall Ave, Bristol, BS8 1TQ, UK
5School of Earth and Environment, University of Leeds, Leeds, LS2 9JT, UK
6Department of Geology, University Centre in Svalbard, Longyearbyen, 9171, Norway
7Department of Environmental Sciences, Western Norway University of Applied Sciences, 6856 Sogndal, Norway
8Department of Geography, University of Utah, Central Campus Dr, Salt Lake City, Utah, USA
9Faculty of Science, Liverpool John Moores University, James Parsons Building, Byrom Street, Liverpool, L3 3AF, UK
10School of Geosciences, University of Sydney, Sydney, NSW 2006, Australia
11Department of Geography and Earth Science, Aberystwyth University, Wales, SY23 3DB, UK
12Institute at Brown for Environment and Society, Brown University, Providence, Rhode Island, USA
13Department of Environmental Science, Aarhus University, 4000 Roskilde, Denmark
14Centre for Gas Hydrate, Environment and Climate, University of Tromsø, 9010 Tromsø, Norway
15School of Geography and Lincoln Centre for Water and Planetary Health, University of Lincoln, Think Tank, Ruston Way,
Lincoln, LN6 7DW, UK
16Climate and Space Sciences and Engineering, University of Michigan, 2455 Hayward St. Ann Arbor, Michigan, USA
17German Research Centre for Geosciences, GFZ, Potsdam, Germany
18Department of Earth Sciences, University of Berlin, Berlin, Germany
19Geological Survey of Denmark and Greenland, Copenhagen, Denmark
Correspondence: Joseph M. Cook (joc102@aber.ac.uk)
Received: 18 March 2019 – Discussion started: 3 April 2019
Revised: 16 December 2019 – Accepted: 18 December 2019 – Published: 29 January 2020
Abstract. Melting of the Greenland Ice Sheet (GrIS) is the
largest single contributor to eustatic sea level and is ampli-
fied by the growth of pigmented algae on the ice surface,
which increases solar radiation absorption. This biological
albedo-reducing effect and its impact upon sea level rise has
not previously been quantified. Here, we combine field spec-
troscopy with a radiative-transfer model, supervised clas-
sification of unmanned aerial vehicle (UAV) and satellite
remote-sensing data, and runoff modelling to calculate bio-
logically driven ice surface ablation. We demonstrate that al-
gal growth led to an additional 4.4–6.0 Gt of runoff from bare
ice in the south-western sector of the GrIS in summer 2017,
representing 10 %–13 % of the total. In localized patches
with high biomass accumulation, algae accelerated melting
by up to 26.15 ±3.77 % (standard error, SE). The year 2017
was a high-albedo year, so we also extended our analysis
to the particularly low-albedo 2016 melt season. The runoff
from the south-western bare-ice zone attributed to algae was
much higher in 2016 at 8.8–12.2 Gt, although the propor-
tion of the total runoff contributed by algae was similar at
Published by Copernicus Publications on behalf of the European Geosciences Union.

310 J. M. Cook et al.: Glacier algae accelerate melt rates
9 %–13 %. Across a 10 000 km2area around our field site, al-
gae covered similar proportions of the exposed bare ice zone
in both years (57.99 % in 2016 and 58.89 % in 2017), but
more of the algal ice was classed as “high biomass” in 2016
(8.35 %) than 2017 (2.54 %). This interannual comparison
demonstrates a positive feedback where more widespread,
higher-biomass algal blooms are expected to form in high-
melt years where the winter snowpack retreats further and
earlier, providing a larger area for bloom development and
also enhancing the provision of nutrients and liquid water
liberated from melting ice. Our analysis confirms the impor-
tance of this biological albedo feedback and that its omission
from predictive models leads to the systematic underestima-
tion of Greenland’s future sea level contribution, especially
because both the bare-ice zones available for algal coloniza-
tion and the length of the biological growth season are set to
expand in the future.
1 Introduction
Mass loss from the Greenland Ice Sheet (GrIS) has increased
over the past 2 decades (Shepherd et al., 2012; Hanna et al.,
2013) and is the largest single contributor to cryospheric sea
level rise, adding 37 % or 0.69 mm yr−1between 2012 and
2016 (Bamber et al., 2018). This is due to enhanced surface
melting (Ngheim et al., 2012) that exceeds calving losses at
the ice sheet’s marine-terminating margins (Enderlin et al.,
2014; van den Broeke et al., 2016). Surface melting is con-
trolled by net solar radiation, which in turn depends upon the
albedo of the ice surface, making albedo a critical factor for
modulating ice sheet mass loss (Box et al., 2012; Ryan et
al., 2018a). The largest shift in albedo occurs when the win-
ter snow retreats to expose bare glacier ice. However, there
are several linked mechanisms that then change the albedo of
the exposed ice and determine its rate of melting, including
meltwater accumulation, ice surface weathering and the ac-
cumulation of light-absorbing particles (LAPs), such as soot
(Flanner et al., 2007) and mineral dust (Skiles et al., 2017).
Photosynthetic algae also reduce the albedo of the GrIS (Ue-
take et al., 2010; Yallop et al., 2012; Stibal et al., 2017; Ryan
et al., 2017, 2018b). Despite being identified in the late 1800s
(Nordenskiöld, 1875), their effects have not yet been quanti-
fied, mapped or incorporated into any predictive surface mass
balance models (Langen et al., 2017; Noël et al., 2016; Fet-
tweis et al., 2017). Hence, biological growth may play an im-
portant yet underappreciated role in the melting of the Green-
land Ice Sheet and its contributions to sea level rise (Benning
et al., 2014).
The snow-free surface of the GrIS has a conspicuous dark
stripe along its western margin that expands and contracts
seasonally, covering 4 %–10 % of the ablating bare-ice area
(Shimada et al., 2016). The extent and darkness of this “dark
zone” may be biologically and/or geologically controlled
(Wientjes et al., 2011, 2016; Tedstone et al., 2017; Stibal et
al., 2017). There is a growing literature demonstrating the
albedo-reducing role played by a community of algae that
grow on glacier ice on the eastern (Lutz et al., 2014) and
western (Uetake et al., 2010; Yallop et al., 2012; Stibal et al.,
2017; Tedstone et al., 2017; Williamson et al., 2018) GrIS.
The algal community on the GrIS is dominated by Mesotae-
nium berggrenii and Ancylonema nordenskioldii (Yallop et
al., 2012; Stibal et al., 2017; Williamson et al., 2018, 2019;
Lutz et al., 2018), which are collectively known as “glacier
algae” to distinguish them from snow algae and sea ice al-
gae. The presence of these glacier algae reduces the albedo
of the ice surface, mostly due to a purple purpurogallin-like
pigment (Williamson et al., 2018; Stibal et al., 2017; Remias
et al., 2012).
An equivalent albedo reduction due to algae has also
been studied on snow. Worldwide, snow algal communities
are dominated by unicellular Chlamydomonaceae, the most
abundant of which belong to the collective taxon Chlamy-
domonas nivalis (Leya et al., 2004). These algae have been
shown to be associated with low-albedo snow in eastern
Greenland (Lutz et al., 2014) and to be responsible for 17 %
of snowmelt in Alaska (Ganey et al., 2017). However, for
glacier algae, quantification of the biological albedo reduc-
tion, radiative forcing and melt acceleration has remained
elusive due to the difficulty of separating biological from
non-biological albedo-reducing processes and a lack of di-
agnostic biosignatures for remote sensing. For snow, remote
detection has been achieved by measuring the “uniquely bio-
logical” chlorophyll absorption feature at 680 nm (Painter et
al., 2001), a broader carotenoid absorption feature (Takeuchi
et al., 2006), a normalized difference spectral index (Ganey
et al., 2017) and a spectral unmixing model (Huovinen et
al., 2018). However, these signature spectra can be ambigu-
ous for glacier algae due to the presence of the phenolic pig-
ment with a broad range of absorption across the UV and VIS
wavelengths that obscures features associated with other pig-
ments in raw reflectance spectra and is further complicated
by the highly variable optics of the underlying ice and mix-
ing of algae with other impurities.
The dark zone is of the order 105km2in extent and is un-
dergoing long-term expansion (Shimada et al., 2016; Ted-
stone et al., 2017). Quantifying the impact of algal coloniza-
tion on the dark zone is therefore paramount. Upscaling of
unmanned aerial vehicle (UAV) observations made in a small
sector of the dark zone to satellite data has demonstrated that
“distributed impurities” including algae exert a primary con-
trol on the surface albedo, but isolating the biological effect
and upscaling to the regional scale has been prevented by
a lack of spectral resolution and ground validation (Ryan et
al., 2018a). Recently, Wang et al. (2018) applied the vege-
tation red edge (difference in reflectance between 673 and
709 nm) to map glacier algae over the south-western GrIS
using Sentinel-3 OLCI data at 300 m ground resolution. Nei-
The Cryosphere, 14, 309–330, 2020 www.the-cryosphere.net/14/309/2020/

J. M. Cook et al.: Glacier algae accelerate melt rates 311
ther of these previous studies quantified the effect of glacier
algae effect on albedo or melt at the regional scale.
Here, we directly address these issues, resolving a major
knowledge gap limiting our ability to forecast ice sheet melt
rates into the future. First, we use spectroscopy to quantify
the effect of glacier algae on albedo and radiative forcing in
ice. We then use a new radiative-transfer model to isolate
the effects of individual light-absorbing particles on the ice
surface for the first time, enabling a comparison between lo-
cal mineral dust and algae and providing the first candidate
albedo parameterization that could enable glacier algae to be
incorporated into mass balance models. To determine spa-
tial coverage, we apply a supervised classification algorithm
(random forest) to map glacier algae in multispectral UAV
and satellite data. Runoff modelling informed by our empir-
ical measurements and remote-sensing observations enables
us to estimate the biological contribution to GrIS runoff for
the first time.
2 Field sites and methods
2.1 Overview
In this study we present a suite of empirical, theoretical and
remote-sensing data to quantify and map algal contributions
to melting on the south-western GrIS. At our field site we
paired spectral reflectance and albedo measurements with re-
moval of surface ice samples for biological and mineralogi-
cal analyses in order to quantify the relationship between cell
abundance and broadband and spectral albedo. The imagi-
nary part of the refractive index of the local mineral dusts
and the purpurogallin-type phenolic compound that domi-
nates absorption in the local glacier algae were measured in
the laboratory and incorporated into a new radiative-transfer
model. The albedo effects of each impurity were thus ex-
amined in isolation and compared. At the same time, we
also undertook a sensitivity study with other bulk dust op-
tical properties from previous literature to further test the
potential role of mineral dusts in darkening the ice surface.
Furthermore, by combining albedo measurements with in-
coming irradiance spectra and measurements of local melt
rates, we estimated the radiative forcing and the proportion
of melting that could be attributed to algae in areas of high
and low algal biomass (Hbio and Lbio). At our field site we
made UAV flights with a multispectral camera in order to
map algal coverage at high spatial resolution. We achieved
this by training a random-forest (RF) algorithm on our field
spectroscopy data to classify the ice surface into discrete cat-
egories including Hbio and Lbio. This enabled estimates of
algal coverage in a 200 ×200 m area at our field site. We
then retrained our classifier for Sentinel-2 satellite imagery
and used this to upscale further within the south-western re-
gion of the GrIS (to a 100 ×100 km Sentinel-2 tile covering
our field site and UAV image area). With these estimates of
algal coverage from our remote-sensing imagery and calcu-
lations of the proportion of melting attributed to algae from
our field data, we were able to estimate runoff attributed to
algae using the runoff model by van As et al. (2017). The
details of each stage of our methodology are provided in the
following Sect. 2.2–2.10.
2.2 Field site
Experiments were carried out at the Black and Bloom Project
field site (67.04◦N, 49.07◦W; Fig. 1), near the Institute for
Marine and Atmospheric research Utrecht (IMAU) Auto-
matic Weather Station “S6” on the south-western Green-
land Ice Sheet between 10 and 22 July 2017. We estab-
lished a 200 ×200 m area for UAV mapping (centred on
67.07789444◦N, 49.350000◦W) where only essential ac-
cess was allowed (e.g. for placing ground control points,
GCPs, for geo-rectifying our UAV images) and sample re-
moval was prohibited. We also delineated an additional ad-
jacent 20 ×200 m area that we referred to as the “sampling
strip” in which we made spectral reflectance and albedo mea-
surements paired with removal of samples for biological and
mineralogical analyses, as detailed in the following sections.
The sampling strip was subdivided into smaller subregions
that were then systematically visited each day of our field
season. This was necessary because ice surface samples were
destructively removed for analysis and this method ensured
that each area visited had not been disturbed by our pres-
ence on previous days. Some ancillary directional reflectance
measurements were also made at the same field site between
15 and 25 July 2016 and appended to our training dataset for
supervised classification (Sect. 2.8).
2.3 Field spectroscopy
At each site in our sampling strip, albedo was measured us-
ing an ASD (Analytical Spectral Devices, Colorado) Field-
Spec Pro 3 spectroradiometer with an ASD cosine collector.
The cosine collector was mounted horizontally on a 1.5 m
crossbar levelled on a tripod with a height between 30 and
50 cm above the ice surface. The cosine collector was po-
sitioned over a sample surface, connected to the spectrora-
diometer using an ASD fibre optic. Following this, the spec-
troradiometer was controlled remotely from a laptop, mean-
ing the operators could move away from the instrument to
avoid shading it. Two upwards- and two downwards-looking
measurements were made in close succession (∼2 min) to
account for any change in atmospheric conditions, although
the measurements presented were all made during constant
conditions of clear skies at solar noon ±2 h. Each retrieval
was the average of >20 replicates.
Immediately after making the albedo measurements, the
cosine collector was replaced with a 10◦collimating lens,
enabling a nadir view hemispheric conical reflectance fac-
tor (HCRF) measurement to be obtained. For HCRF mea-
www.the-cryosphere.net/14/309/2020/ The Cryosphere, 14, 309–330, 2020

312 J. M. Cook et al.: Glacier algae accelerate melt rates
Figure 1. (a) Map of Greenland showing the bounding box of the Sentinel-2 tile (22WEV) containing our field site (red box) and the
latitudinal extent of our runoff modelling (red line). The basemap was created using the MATLAB Arctic Mapping Toolbox (Greene et al.,
2017). The Sentinel-2 tile outlined by the red box is shown in detail in (b) with the field site marked with a yellow dot. Sentinel-2 bands 2,
3 and 4 were combined into a true-colour composite using GDAL (geographical data abstraction library, 2019), by the lead author. Panel (c)
shows the area immediately surrounding the field camp as captured by a DJI Phantom Pro by the lead author.
surements the upwards-looking measurements were replaced
with HCRF measurements of a flat Spectralon®panel with
the spectroradiometer in reflectance mode. This protocol was
followed for every sample surface, with both albedo and di-
rectional reflectance measurements taking less than 5 min.
We closely followed the methodology described by Cook et
al. (2017b). Albedo is the most appropriate measurement for
determining the surface energy balance, while the HCRF is
closer to the measurements made by aerial remote sensing
and less sensitive to stray light reflecting from surfaces other
than the homogeneous patch directly beneath the sensor. We
therefore used the albedo for energy balance calculations and
the HCRF for remote-sensing applications in this study.
2.4 Biological measurements
Immediately following the albedo and HCRF measurements,
ice from within the viewing area of the spectrometer was re-
moved using a sterile blade and scooped into sterile Whirl-
Pak bags, melted in the dark and immediately fixed with
3 % glutaraldehyde. The samples were then returned to
the University of Bristol and University of Sheffield where
microscopic analyses were undertaken. Samples were vor-
texed thoroughly before 20 µL was pipetted into a Fuchs–
Rosenthal haemocytometer. The haemocytometer was di-
vided into 4 ×4 image areas. These were used to count a
minimum of 300 cells to ensure adequate representation of
species diversity (where possible, as some low-abundance
samples had as few as one cell per haemocytometer). The
volume of each image area was used to calculate cells per
millilitre. Biovolume was determined by measuring the long
and short axes of at least 10 cells from each species in each
sample using the measure tool in the GNU Image Manipu-
lation Program (GIMP). The morphology of the cells in the
images was used to separate them into two species: Mesotae-
nium berggrenii and Ancylonema nordenskioldii. These di-
mensions were then used to calculate the mean volume of
each species in each sample, assuming the cells to be circu-
lar cylinders (following Hillebrand, 1999, and Williamson et
al., 2018). The average volume was multiplied by the num-
ber of cells for each species and then summed to provide the
total biovolume for each sample.
2.5 Mineral and algal optical properties and
radiative-transfer modelling
A new radiative-transfer package, BioSNICAR_GO, was de-
veloped for this study and was used to predict the albedo
of snow and ice surfaces with algae and mineral dusts. We
made a series of major updates and adaptations to the BioS-
NICAR model presented by Cook et al. (2017b). The pack-
age is divided into a bio-optical scheme wherein the optical
properties of light-absorbing impurities and ice crystals can
be calculated using Mie scattering (for small spherical parti-
cles such as black carbon or snow) or geometrical optics (for
large and/or aspherical particles such as glacier algae, larger
mineral dust particles, and large ice crystals) and a two-
stream radiative-transfer model based on SNICAR (Flanner
et al., 2007), which incorporates the equations of Toon et
The Cryosphere, 14, 309–330, 2020 www.the-cryosphere.net/14/309/2020/

J. M. Cook et al.: Glacier algae accelerate melt rates 313
al. (1989). A schematic of the model structure is provided
in the Supplement (Sect. S1).
To incorporate glacier algae into BioSNICAR_GO, ge-
ometrical optics were employed to determine the single-
scattering optical properties of the glacier algae, since they
are large (∼103µm3, making Mie calculations impractically
computationally expensive) and best approximated as circu-
lar cylinders (Hillebrand, 1999; Lee and Pilon, 2003). Our
approach is adapted from the geometric optics parameteriza-
tion of van Diedenhoven (2014). The inputs to the geomet-
ric optics calculations are the cell dimensions and the com-
plex refractive index. The imaginary part of the refractive in-
dex was calculated using a mixing model based upon Cook
et al. (2017b), where the absolute mass of each pigment in
the algal cells was measured in field samples. The absorp-
tion spectra for the algal pigments is provided in Fig. 2a.
We updated the mixing model by Cook et al. (2017b) to ap-
ply a volume-weighted average of the imaginary part of the
refractive index of water and the algal pigments so that the
simulated cell looks like water at wavelengths where pig-
ments are non-absorbing. We consider this to be more phys-
ically realistic than having cells that are completely non-
absorbing at wavelengths >0.75 µm, especially since a wa-
ter fraction (Xw) is used in the calculations to represent the
non-pigmented cellular components of the total cell volume.
This approach also prevents the refractive index from be-
coming infinite when the water fraction is zero, removing
the constraint 0 < Xw<1 from the bio-optical scheme in
the original BioSNICAR model. Based upon experimental
evidence in Dauchet et al. (2015) for the model species C.
reinhardii, the real part of the refractive index has been up-
dated from 1.5 (in Cook et al., 2017b) to 1.4. The absorp-
tion coefficients from which the imaginary refractive index
is calculated are from Dauchet et al. (2015), apart from the
purpurogallin-type phenol, whose optical properties were de-
termined empirically (Fig. 2a). The calculated optical prop-
erties were added to the lookup library for BioSNICAR-GO
for a range of cell dimensions. For the simulations presented
in this study, we included two classes of glacier algae rep-
resenting Mesotaenium berggrenii and Ancylonema norden-
skioldii with length and diameter and also the relative abun-
dance of each species matching the means measured in our
microscopy described in Sect. 2.4. In simulations (Sect. S2)
we found that ice albedo was relatively insensitive to the di-
mensions of the cells within a realistic range of lengths and
diameters. This low sensitivity to cell length and diameter is
likely because all of the cells considered here are large from
a radiative-transfer perspective.
For mineral dusts, we took measured values for surface
dust composition and particle size distribution (PSD) ob-
tained at our field site from McCutcheon et al. (2020; here-
after, referred to as McC). We then used complex refrac-
tive indices for the appropriate minerals obtained from the
existing literature, mixed them using the Maxwell Garnett
dielectric mixing approximation according to the measured
mass fractions (after converting to volumetric fractions us-
ing the mineral densities), generated the single-scattering
optical properties using a Mie scattering code and applied
a weighted average using the PSD to obtain the bulk opti-
cal properties for the dust. Since the mineralogy of the dust
varied between sites we generated three dust “scenarios”.
In the low-absorption scenario (LO-DUST) all the minerals
were set to the minimum volume-fraction measured across
all of McC’s samples except for quartz, which comprised the
remainder. In the high-absorption scenario (HI-DUST) all
the minerals were set to their maximum measured volume-
fraction, apart from quartz, which comprised the remainder.
Finally, in the mean scenario (MN-DUST) all the minerals
were present with their volume fractions equal to the mean
across all of the field samples. The mineralogy of each of
these scenarios is described in Table 1. Refractive indices
were not available for all of the individual minerals present
in McC’s analysis, so we represented the feldspar minerals
using the refractive index for andesite (Pollack et al., 1973)
and all pyroxenes with the refractive index for enstatite (Jäger
et al., 2003), and, in the absence of a refractive index for
amphibole phases, we used the refractive index for the simi-
larly green mineral olivine (OCDB, 2002). Refractive indices
for all other minerals were available (Rothman et al., 1998;
Roush et al., 1991; Pollack et al., 1973; Egan and Hilgeman,
1983; Nitsche and Fritz, 2004).
The ice optical properties in BioSNICAR-GO were also
calculated using a parameterization of geometric optics
adapted from van Diedenhoven et al. (2014). A geometri-
cal optics approach to generating ice optical properties was
chosen because it enables arbitrarily large ice grains with
a hexagonal columnar shape to be simulated, in order to
better estimate the albedo of glacier ice where grains are
large and aspherical. While the real ice surface is composed
of irregularly shaped and sized grains, this approach en-
abled us to simulate our field spectra much more accurately
and circumvented the requirements that individual grains be
small and spherical in the case of the Lorenz–Mie approach.
The optical properties of the ice grains were modelled us-
ing refractive indices from Warren and Brandt (2008). The
radiative-transfer model is a two-stream model described in
full in Cook et al. (2017b) and Flanner et al. (2007). For the
radiative-transfer modelling presented in this study, the fol-
lowing model parameters were used: diffuse illumination; ice
crystal side-length and diameter per vertical layer =3, 4, 5,
8, and 10 mm; layer thicknesses =0.1, 1, 1, 1 and 1 cm; un-
derlying surface albedo =0.15; and layer densities =500,
500, 600, 600 and 600 kg m−3. These ice physical properties
were chosen to reduce the absolute error between the simu-
lated albedo for ice without any impurities (“clean ice”) and
our mean field-measured clean-ice spectrum.
To realistically simulate measured dust and algal mass
loadings on the ice surface, we took measured values for
Hbio field samples. For mineral dusts we took the mean and
maximum mineral mass mixing ratios from McC. They mea-
www.the-cryosphere.net/14/309/2020/ The Cryosphere, 14, 309–330, 2020

314 J. M. Cook et al.: Glacier algae accelerate melt rates
Figure 2. (a) Mass absorption coefficients of the major algal pigments including the purpurogallin-type phenol. (b) Measured spectral albedos
for each surface type (Hbio is heavy biomass loading, Lbio is light biomass loading, CI is clean ice and SN is snow). (c) Plot showing the
natural logarithm of cell abundance against broadband albedo. (d) Microscope image showing examples of both algal species and mineral
fragments from a melted Hbio sample.
Table 1. Composition of each mineral dust “scenario” in percent of total by volume.
Fraction of total (% by volume)
Scenario Quartz Andesite Olivine Enstatite Kaolinite Illite Muscovite
HI-DUST 3.42 67.12 10.53 8.42 3.36 1.70 5.46
LO-DUST 45.39 50.67 3.31 0.64 0 0 0
MN-DUST 24.19 61.03 6.95 3.90 1.37 0.19 2.37
sured 394 ±194 µgLAP mLice −1, of which ∼95 % was inor-
ganic, giving mean and maximum mineral dust loadings of
373 and 567 µgLAP mLice−1. Assuming 1 mL of ice to weigh
0.917 g, this gives mean and maximum mass mixing ratios
of 342 and 519 µgdust gice−1. For glacier algae we calculated
mass mixing ratios by taking the mean cell volume across all
cells in our microscope images, converting to per-cell mass
using a constant cell density (0.87 g cm−3; Hu, 2014) and
multiplying by our mean and maximum Hbio cell abundance.
This gave mean and maximum mass mixing ratios of 349
and 646 µgalgae gice −1. We also varied the mass mixing ratios
over a range of hypothetical values to study the sensitivity
of ice surface albedo to dust and glacier algae. Glacier algae
and each of the mineral dusts (LO-DUST, HI-DUST, MN-
The Cryosphere, 14, 309–330, 2020 www.the-cryosphere.net/14/309/2020/

J. M. Cook et al.: Glacier algae accelerate melt rates 315
DUST) were added individually to the upper 0.1 cm layer in
mixing ratios of 10, 100, 500 and 1000 µgLAP gice−1, plus the
mean and maximum measured mass mixing ratios for dust
and algae, to quantify their effects on the surface albedo. We
also ran a sensitivity study where we repeated the simulations
with two other dust types, sourced from previous literature,
with contrasting mineralogies to our field site.
2.6 Empirical measurement of mineral dust reflectance
For two samples of local mineral dusts obtained from Hbio
sites, we chemically removed the organic matter and mea-
sured the PSD using scanning electron microscopy (full de-
tails in Sect. S3). The chemical cleaning method avoided the
artificial “reddening” of the mineral dust sample associated
with removing organic matter by ignition. We then arranged
the mineral dust samples into an optically thick layer on a
microscope slide and pressed them tightly against the open
aperture of a Thorlabs IS200-4 integrating sphere to mea-
sure their reflectance. The other apertures were covered with
SM05CP2C caps and the sample reflectance was measured
using the same ASD Field Spec Pro 3 as was used for field
measurements.
2.7 Radiative forcing and biological melt acceleration
The biological radiative forcing was calculated by first differ-
encing the albedo for algal surfaces and the albedo for clean
ice surfaces measured at our field site. This gives the differ-
ence in albedo between the clean and algal ice surfaces, αdiff.
The product of each αdiff and the incoming irradiance, I∗,
provided the instantaneous power density (PDalg) absorbed
by the algae. We assume that photosynthetic processes uti-
lize 5 % of this absorbed energy – at the upper end of a re-
alistic range for photosynthetic microalgae (Blankenship et
al., 2011; Masojídek et al., 2013). The remainder of PDalg
is conducted into the surrounding ice, giving the instanta-
neous radiative forcing due to algae (IRFalg). Since these
cells are coloured by the purple purpurogallin pigment, we
assume the reflective radiative forcing to be negligible, as
demonstrated by Dial et al. (2018). IRFalg was calculated
at hourly intervals using incoming irradiance simulated for
our field site using the PVSystems solar irradiance program
(https://pvlighthouse.com.au, last access: July 2019) at 1 nm
spectral resolution, following Dial et al. (2018). The radiative
forcing was assumed to be constant between each 1 h time
step, meaning the radiative forcing over 1h (HRFalg) could
be calculated by multiplying IRFalg by 3600 s h−1, assuming
that instantaneous radiative forcing is equal to radiative forc-
ing per second. Daily radiative forcing due to algae (RFalg)
was then calculated as the sum of HRFalg between 00:00 and
23:00 UTC.
To calculate the algal contribution to melting (Malg),
IRFalg was multiplied by 104to convert the radiative forcing
from units of W m−2to W cm−2and then divided by the la-
tent heat of fusion for melting ice (334 J g−1) and integrated
over the entire day as described above. This provided a value
for the amount of melting caused by the presence of algae
per day assuming the cold content of the ice to be depleted.
We calculated uncertainty by running these calculations for
every possible combination of our measured algal and clean
ice spectra and calculating the mean, standard error, and stan-
dard deviation of the pooled results.
We corroborated these estimates using a point surface en-
ergy balance model (Brock and Arnold, 2000; Tedstone,
2019). This model predicts melting in millimetres of wa-
ter equivalent given local meteorological data and informa-
tion about the ice surface albedo and roughness. We ran this
model with the albedo set equal to the broadband albedo for
each clean ice (CI), heavy biomass (Hbio) and light biomass
(Lbio) spectrum in our field measurements. The hourly mete-
orological data for 21 July 2017 used to force the model were
from a Delta-T GP1 automatic weather station positioned at
our field site. The difference in predicted melt between the
algal surfaces and the clean ice surfaces provided the melt
attributed to the presence of algae. As for the radiative forc-
ing calculations, the uncertainty was calculated by running
the energy balance model for every possible combination of
algal and clean ice spectra and calculating the mean, standard
error, and standard deviation of the pooled results.
2.8 UAV and Sentinel-2 remote sensing
Having quantified algal melt acceleration in localized
patches using the methods described in Sect. 2.2–2.6, we
then used a multispectral camera mounted to a UAV to quan-
tify algal coverage across a 200 ×200 m area at our field site.
This sample area was kept pristine throughout the study pe-
riod to minimize artefacts of our presence appearing in the
UAV imagery. Inside the sampling area we placed fifteen
10 ×10 cm ground control points (GCPs), whose precise lo-
cation was measured using a Trimble differential GPS. At
these markers we also made ground spectral measurements
using an ASD-Field Spec Pro 3 immediately after each flight.
The UAV itself was a Steadidrone Mavrik-M quadcopter,
onto which we integrated a MicaSense Red-Edge multispec-
tral camera. The camera is sensitive in five discrete bands,
with centre wavelengths of 475, 560, 668, 717, and 840 nm
and bandwidths of 20, 20, 10, 10, and 40 nm, respectively.
The horizontal field of view was 47.2◦and the focal length
5.4 mm. The camera was remotely triggered through the au-
topilot, which was programmed along with the flight coordi-
nates in the open-source software Mission Planner (Osborne,
2019). Images were acquired at approximately 2 cm ground
resolution with 60 % overlap and 40 % sidelap. The flights
were less than 20 min long and at an altitude of 30 m above
the ice surface.
We applied radiometric calibration and geometric dis-
tortion correction procedures to acquired imagery follow-
ing MicaSense procedures (Micasense, 2019). We then con-
www.the-cryosphere.net/14/309/2020/ The Cryosphere, 14, 309–330, 2020

316 J. M. Cook et al.: Glacier algae accelerate melt rates
verted from radiance to reflectance using time-dependent re-
gression between images of the MicaSense Calibrated Re-
flectance Panel acquired before and after each flight (i.e.
a regression line was computed between the reflectance of
the white reference panel at the start and end of the flight
and used to quantify the change in irradiance during the
flight). Finally, the individual reflectance-corrected images
were mosaicked using AgiSoft PhotoScan following pro-
cedures developed by the United States Geological Survey
(USGS, 2017), yielding a multi-spectral ortho-mosaic with
5 cm ground resolution, georectified to our GCPs. There was
generally close agreement between the ground, UAV and
satellite-derived albedo, although there are some differences
that we believe to be the result of different radiometric cal-
ibration techniques for satellite, UAV and ground measure-
ments, and the differing degrees of spatial integration have
been examined in detail in Tedstone et al. (2019).
To upscale further, we used multispectral data from
the Copernicus Sentinel-2 satellite. We selected the
100 ×100 km tile covering our field site (T22WEV) on the
closest cloud-free day to our UAV flight on 21 July. The
L1C product was downloaded from SentinelHub (Sinergise,
Slovenia). The L1C product was processed to L2A using the
European Space Agency (ESA) Sen2Cor processor, includ-
ing atmospheric correction and reprojection to 20 m resolu-
tion.
2.9 Supervised classification algorithms and albedo
mapping
To map and quantify spatial coverage of algae over the
ice sheet surface we employed a supervised classification
scheme. A random forest (RF) classifier was trained on the
field spectra collected on the ice surface (see Sect. 2.3) and
then applied to multispectral images gathered by the UAV
and Sentinel-2. We also included spectra obtained at the same
field site in July 2016 to our training set, giving a total of 231
labelled spectra. A schematic of the classification workflow
is provided in Sect. S4. Our HCRF measurements were first
reduced to reflectance values at five key wavelengths coinci-
dent with the centre wavelengths measured by the MicaSense
Red-Edge camera mounted to the UAV (blue: 0.475; green:
0.560; red: 0.668; red edge: 0.717; NIR: 0.840 µm), yield-
ing reflectance at each wavelength as a feature vector for
the classifier (in this case the spectral response function of
the camera was not accounted for). The classification labels
were the surface type as determined by visual inspection: SN
(snow), CI (clean ice), CC (cryoconite), WAT (water), Lbio
(low-biomass algae) and Hbio (high-biomass algae). For the
algal surface classes our visual assessment was corroborated
by microscopy, as described in Sect. 2.2. This dataset was
then shuffled and split into a training set (80 %) and a test
set (20 %). The training set was used to train three individual
supervised classification algorithms: Naive Bayes, k-nearest
neighbours (KNN) and support vector machine (SVM). For
the SVM, the parameters C and gamma were tuned using
grid search cross validation. Two ensemble classifiers were
also trained: a voting classifier that combined the predictions
of each of the three individual classifiers and a RF algorithm.
The performance of each classifier was measured using pre-
cision, accuracy, recall, and F1 score and also by plotting the
confusion matrix and normalized confusion matrix for each
classifier. In all cases the RF outperformed the other classi-
fiers according to all available metrics (Sect. S5). The perfor-
mance of the RF classifier was finally measured on the test
set, demonstrating the algorithm’s ability to generalize to un-
seen data outside of the training set. Overfitting is not usually
associated with the RF classifier, and the strong performance
on both our training and test sets confirms that the model gen-
eralizes well. For these reasons, we used the RF algorithm to
classify our multispectral UAV and Sentinel-2 images. Train-
ing the classifier using data from field spectroscopy ensures
the quality of each labelled data point in the training set, since
our sampling areas were homogeneous and surface samples
were analysed in the laboratory, circumventing issues of spa-
tial heterogeneity and uncertainty in labelling that could lead
to ambiguity for direct labelling of aerial images. Compar-
isons between the directional reflectance spectra gathered us-
ing the ASD field spectrometer and those measured using the
UAV and Sentinel-2 are provided in Fig. 3. Simultaneously
with the surface classification, we calculated the albedo in
each UAV pixel using the narrowband to broadband conver-
sion of Knap et al. (1999) applied to the reflectance at each
of the five bands.
This protocol was repeated for Sentinel-2 imagery. Addi-
tional bands are available for use as feature vectors in the
case of Sentinel-2. Directional reflectance data gathered us-
ing the ASD field spectrometer were reduced to only those
nine wavelengths coincident with the centre wavelengths
measured by Sentinel-2 at 20 m ground resolution (0.480,
0.560, 0.665, 0.705, 0.740, 0.788, 0.865, 1.610, 2.190 µm).
Training on reduced hyperspectral data has several advan-
tages over training directly on aerial multispectral data. First,
the method is sensor agnostic because the classifier can be
retrained with a different selection of wavelengths for other
upscaling platforms, enhancing the reusability of the field
measurements. Second, we have confidence in our labels be-
cause each sample has been analysed in a laboratory to con-
firm its composition, reducing label ambiguity. Finally, the
limited field of view of the field spectrometer reduces errors
arising from mixing of spectra from heterogeneous ice sur-
faces. Sentinel-2 imagery was masked using the MeASUREs
Greenland Ice Mapping Project ice mask (Howat, 2017) to
eliminate non-ice areas. Pixels with more than 30 % prob-
ability of being obscured by cloud were masked using the
Sentinel-2 L2A cloud product generated by the Sen2Cor pro-
cessor. For the calculation of albedo in each pixel, the addi-
tional bands available in the Sentinel-2 images enabled the
application of Liang et al.’s (2001) narrowband to broadband
conversion.
The Cryosphere, 14, 309–330, 2020 www.the-cryosphere.net/14/309/2020/

J. M. Cook et al.: Glacier algae accelerate melt rates 317
Figure 3. Inter-sensor comparisons. (a–e) Each UAV band reflectance plotted against ASD reflectance in uncorrected (blue) and corrected
(red) form. The correction was applied to account for a systematic offset shown in the header for each plot. (f) Mean reflectance and ±1
standard deviation error bars at each spectral band for each surface class for the ASD field spectrometer and the UAV-mounted multispectral
camera. (g) Mean reflectance and ±1 standard deviation error bars at each spectral band for each surface class for the ASD field spectrometer
and Sentinel-2.
2.10 Comparing 2016 and 2017
In 2017, the GrIS dark zone had a relatively small spatial
extent, high albedo and short duration in comparison to the
other years in the MODIS record, particularly since 2007,
whereas the dark zone was especially dark, widespread and
prolonged in 2016 (Fig. 4; Tedstone et al., 2017). We there-
fore conducted a comparison between the algal coverage on
the same dates in 2016 and 2017. First, we examined varia-
tions in the extent and duration of the dark zone, along with
www.the-cryosphere.net/14/309/2020/ The Cryosphere, 14, 309–330, 2020

318 J. M. Cook et al.: Glacier algae accelerate melt rates
snow depths and snow clearing dates for the south-western
ablation zone using MODIS, extending the time series of
Tedstone et al. (2017). Bare ice was mapped by applying a
threshold reflectance value (R < 0.60 at 0.841–0.871 µm) to
the MOD09GA Daily Land Surface Reflectance Collection
6 product. Within the bare-ice area, dark ice was mapped us-
ing a lower reflectance threshold (R < 0.45 at 0.62–0.67 µm).
The area of interest was the “common area” defined by Ted-
stone et al. (2017) bounded within the latitudinal range 65–
70◦N and is equal to that used by Wang et al. (2018). To
measure the annual dark-ice extent (in km2) we counted the
pixels that were dark for at least 5 d each year. The an-
nual duration was defined at each pixel as the percentage
of daily cloud-free observations made in each JJA (June–
July–August) period that were classified as dark. The timing
of bare ice appearance was calculated from MODIS using
a rolling window approach on each pixel (see Tedstone et
al., 2017). The mean snow depths were extracted from out-
puts from the regional climate model MAR v3.8 (Fettweis et
al., 2017) run at 7.5 km resolution forced by ECMWF ERA-
Interim reanalysis data (Dee et al., 2011). These data enabled
a comparison of the extent and timing of dark ice in 2016 and
2017.
To examine algal coverage in each year we identified the
Sentinel-2 tile covering our field site (22WEV) on the closest
cloud-free date to the UAV flight day (21 July) in each year.
These were 26 July 2017 and 25 July 2016. Since we were in-
terested in the bare-ice zone, snow-covered pixels were omit-
ted from the calculations.
2.11 Runoff modelling
Runoff at the regional scale was calculated using van As
et al.’s (2017) surface mass balance (SMB) model, forced
with local automatic weather station and MODIS albedo ob-
servations (van As et al., 2012, 2017). The model interpo-
lates meteorological and radiative measurements from three
PROMICE automatic weather stations on the K-transect
(KAN-L, KAN-M and KAN-U) and bins them into 100 m
elevation bands (0 to 2000 m a.s.l.). Surface albedo is from
MODIS Terra MOD10A1 albedo and is averaged into the
same 100 m elevation bins. For every 1 h time step, the model
iteratively solves the surface energy balance for the surface
temperature. If energy components cannot be balanced due
to the 0 ◦C surface temperature limit, a surplus energy sink
for melting of snow or ice is included. If surface temperature
is greater than the melting point, the surplus energy is used
for melting of snow or ice. When calculating turbulent heat
fluxes, aerodynamic surface roughness for momentum was
set to 0.02 and 1 mm for snow and ice, respectively (after van
As et al., 2005, 2012; Smeets and Van den Broeke, 2008). We
extrapolate modelled runoff across the south-western GrIS
(65–70◦N) by deriving the areas of each elevation bin using
the Greenland Ice Mapping Project (GIMP) digital elevation
model (DEM; Howat et al., 2014). Total summer runoff from
Figure 4. (a, b) Dark-ice duration on the south-western GrIS in
summers 2016 and 2017, expressed as a percentage of the total daily
cloud-free observations made during June–July–August (JJA). Each
year is labelled with dark-ice extent. In each year, pixels that are
dark for fewer than 5 d are not shown. (c, d) Average snow depth
modelled by MAR (blue) and cumulative dark-ice extent observed
by MODIS (red) (Tedstone et al., 2017) during April to August.
Vertical bars (grey) denote median date of snow clearing derived
from MODIS. Horizontal bars denote the interquartile range of the
day of year of bare ice appearance. Tick marks denote the start of
each month.
bare ice was calculated by summing runoff in elevation bins
that had mean daily albedo of less than 0.60. Total summer
runoff from dark ice only was calculated in the same way
but using a 0.39 threshold. The study by van As et al. (2017)
compared the performance of the model with independent
observations and found errors to be negligible in the bare-ice
zone.
To determine the algal contribution to runoff, we used
Eq. (1):
Ralg =Rtot ×((MHbio ×CHbio )+(MLbio ×CLbio )), (1)
where Ralg is the runoff due to algae, Rtot is the total runoff
from the bare-ice zone calculated using our runoff model,
MHbio and MLbio are the mean percentage of total melt at-
tributed to algae in Hbio and Lbio areas as calculated by our
energy balance modelling described in Sect. 2.6, and CHbio
The Cryosphere, 14, 309–330, 2020 www.the-cryosphere.net/14/309/2020/

J. M. Cook et al.: Glacier algae accelerate melt rates 319
and CLbio are the proportion of Ctot comprised of Hbio and
Lbio areas in our UAV or Sentinel-2 images. As discussed
later, the Sentinel-2 algal coverage estimate is conservative
because it often fails to resolve Hbio surfaces and therefore
provides a lower bound on the runoff attributed to algae.
An upper bound was therefore also calculated by assuming
the spatial coverage derived from our UAV remote sensing
– which can accurately distinguish Lbio and Hbio – surfaces
is representative of the south-western dark zone. We were
thereby able to estimate upper and lower limits for the runoff
attributed to algal growth on the south-western ablation zone.
3 Results and discussion
3.1 Algae reduce ice albedo
The ice surfaces we studied were divided into four classes de-
pending upon the algal abundance measured in the melted ice
samples: high algal abundance (Hbio), low algal abundance
(Lbio), clean ice (CI) and snow (SN). The algal abundance
(cells mL−1) in each class was as follows: Hbio =2.9×104±
2.01 ×104;Lbio =4.73 ×103±2.57 ×103; CI =625 ±381;
and SN =0±0 (1 SD). These cell abundances were signifi-
cantly different between the classes (one-way ANOVA: F=
10.21; p=3×10−5), which Bonferroni-corrected ttests in-
dicated to be due to variance between all four groups. The
dominant species of algae were Mesotaenium berggrenii
and Ancylonema nordenskioldii (Fig. 2d), confirming ob-
servations made by Stibal et al. (2017) and Williamson et
al. (2018) in the same region. Their long, thin and approxi-
mately cylindrical morphology has been shown to be near-
optimal for light absorption (Kirk, 1976). The albedo of
the ice surface also varied significantly between the sur-
face classes (one-way ANOVA for broadband albedo: F=
7.9; p=2.8×10−4), again with Bonferroni-corrected ttests
showing variance between all four groups (Sect. S6a, b).
Greater algal abundance was associated with lower albedo,
with the albedo reduction concentrated in the visible wave-
lengths (Fig. 2b) where both solar energy receipt and al-
gal absorption peak (Cook et al., 2017b; Williamson et
al., 2018), diminishing towards longer near infra-red (NIR:
>0.70 µm) wavelengths where ice absorption, represented
by the effective grain size, is most likely to cause albedo dif-
ferences (Warren, 1982). A strong inverse correlation (Pear-
son’s R=0.75, p=2.74×10−9) was observed between the
natural logarithm of algal cell abundance (cells mL−1) in the
surface ice samples and broadband albedo (Fig. 2c). The
linear regression coefficient of determination between the
albedo and the natural logarithm of cell abundance was 0.57.
It is unsurprising that the cell abundance does not account for
all variation in albedo because there are also albedo-reducing
effects related to the physical structure of the ice and pres-
ence of melt water (as demonstrated for snow by, for ex-
ample, Warren, 1982). An inverse relationship was also ob-
served between broadband albedo and biovolume (calculated
as the sum of the products of the mean measured cell volumes
and the cell counts for each algal species), but the coefficient
of determination was lower (r2=0.42). This may well be the
result of larger cells having a smaller effect on albedo than
more numerous, smaller cells for a given total volume. The
relationship between absorption and scattering coefficients
and cell size may also not be straightforward for algal cells
due to an increasingly important contribution to the cell op-
tical properties from internal heterogeneity, organelles, cell
walls and the pigment packaging effect in larger cells (Morel
and Bricaud, 1981; Haardt and Maske, 1987).
The albedo of Hbio and Lbio surfaces is depressed in the
visible wavelengths (0.40–0.70 µm, Fig. 2b), creating a red-
edge spectrum commonly used in other environments as a
marker for photosynthetic pigments (Seager et al., 2005) and
for mapping algae over the GrIS by Wang et al. (2018).
Chlorophyll ahas a specific absorption feature at 0.68 µm
which is hard to discern in the raw spectra but clear in the
derivative spectra (Fig. 5a) for Hbio and Lbio but not CI and
SN. This feature has previously been described as “uniquely
biological” (Painter et al., 2001) and supports the hypoth-
esis that the albedo reduction observed in these samples is
primarily due to algae. Our measurements therefore strongly
indicate a biological role in reducing the albedo of the GrIS
surface; however, to test that the lower broadband and spec-
tral albedo observed on algal surfaces is primarily due to the
presence of algal cells, it was also necessary to compare the
albedo-reducing effects of the algae to that of local mineral
dust.
3.2 Algae have greater impact on albedo than mineral
dust
Radiative-transfer simulations demonstrated that at measured
mass mixing ratios mineral dusts only have a very small
(<0.003) albedo-reducing effect at our field site on the
south-western GrIS, whereas glacier algae reduce the ice
albedo by up to 0.06, not accounting for indirect albedo-
reducing feedbacks. The effect of adding the mean measured
mass mixing ratio of MN-DUST to the clean ice was a very
small albedo reduction of 0.002 (Table 2; Fig. 5b). In con-
trast, adding the mean measured mass mixing ratio of glacier
algae reduced the albedo by 0.03, preferentially in the short
visible wavelengths in a similar way to our field-measured
reflectance spectra (Table 2; Fig. 5b). This effect was greater
when the mass mixing ratio was increased to the maximum
measured values (646 µgalgae gice −1and 519 µgdust gice−1)
which caused an albedo reduction of 0.06 for glacier algae
and 0.003 for MN-DUST. Changing the proportions of the
minerals in our simulated local dusts had a very small effect
on the albedo reduction. At the mean measured mass mix-
ing ratio, HI-DUST reduced the albedo by just 0.0023, while
LO-DUST reduced the albedo by 0.0016. Even with a mass
mixing ratio of 1000 µgdust gice−1, the albedo reduction due
www.the-cryosphere.net/14/309/2020/ The Cryosphere, 14, 309–330, 2020

320 J. M. Cook et al.: Glacier algae accelerate melt rates
Figure 5. (a) The first and second derivative spectra for each surface class. (b) BioSNICAR-GO modelled spectral albedo for clean ice (blue)
and ice with each of the simulated local dusts and algae in their measured mass mixing ratios in the upper 1 mm. (c) Reflectance for an
optically thick layer of two samples of the local mineral dust.
Table 2. Albedo change relative to clean ice caused by the addition of each LAP to the upper 1 mm of ice in a range of mass mixing ratios
from 10 to 1000 µgLAP gice−1.
Hypothetical mass mixing ratios (µgLAP gice−1) Measured mass mixing ratios (µgLAP gice−1)
10 100 500 800 1000 342 349 519 646
Glacier algae −0.0010 −0.0110 −0.0460 −0.0670 −0.0800 −0.030 −0.040 −0.0487 −0.056
HI-DUST −0.0001 −0.0006 −0.0030 −0.0048 −0.0060 −0.0021 −0.0023 −0.0033 −0.0039
LO-DUST −0.0001 −0.0004 −0.0021 −0.0034 −0.0042 −0.0015 −0.0016 −0.0023 −0.0028
MN-DUST <0.0001 <0.0001 −0.0020 −0.0043 −0.005 −0.001 −0.002 −0.0029 −0.0035
to local mineral dusts was only 0.006, 0.004 and 0.005 for
the HI-DUST, LO-DUST and MN-DUST, compared to 0.08
for glacier algae.
Across all our simulations, the broadband albedo-reducing
power of glacier algae exceeded that of the local min-
eral dusts, often by several orders of magnitude. At field-
measured mass mixing ratios for heavily laden Hbio surfaces,
mineral dusts cannot account for the broadband albedo re-
duction observed in the field. This is consistent with the lo-
cal mineralogy being dominated by weakly absorbing min-
erals with small grain sizes, as measured in our field sample
(Figs. 5c, 6; Table 2). In Sect. S7 we demonstrate that these
conclusions are robust to different dust types, including those
with typically Saharan optical properties and dusts with vary-
The Cryosphere, 14, 309–330, 2020 www.the-cryosphere.net/14/309/2020/

J. M. Cook et al.: Glacier algae accelerate melt rates 321
Figure 6. Particle size diameter for our local mineral dust sample (panel (b) shows magnification of 0–4µm range).
ing hematite concentrations. The radiative-transfer simula-
tions do not account for feedbacks related to grain size and
shape, near-surface meltwater accumulation, and the pres-
ence of other light absorbing particles, such as humic sub-
stances, that might modify the spectral reflectance and ex-
acerbate the biological albedo reduction. Furthermore, the
albedo-lowering effects of both the glacier algae and min-
eral dusts is reduced by the low albedo of the underlying ice.
In simulations using smaller diameter, higher-albedo snow
grains (whose optical properties were estimated using Mie
theory) the albedo reduction caused by 1000 µgdust gice of
MN-DUST increased to 0.009, 0.010 and 0.012 for grains
of diameter 1500, 1000 and 500 µm, respectively.
The small direct albedo-reducing effect from local miner-
als on the ice surface is seemingly in contrast to some pre-
vious studies, such as Wientjes et al. (2010, 2011) and Bøg-
gild (2010); however, we highlight that neither of the Wien-
tjes et al. (2010, 2011) studies directly measured the surface
albedo or any optical properties of the mineral dusts retrieved
from their GrIS sampling sites and only inferred mineralogi-
cal darkening from low spectral resolution MODIS data and
the presence of a “wavy pattern” observed across the dark
zone. We argue that while this may be indicative of geolog-
ical outcropping onto the ablation zone, it does not neces-
sarily follow that these minerals are responsible for surface
darkening. In support of this, Wientjes et al. (2011) found
strongly scattering and weakly absorbing quartz to be the
dominant mineral in surface ice and speculated that biota
may be having a darkening effect. Bøggild et al. (2010) found
mineral dust to be an albedo reducer in Crown Prince Chris-
tian Land (80◦N, 24◦W), but this area is geologically and
climatologically distinct from our field site and their tran-
sect only spanned ∼8 km from the ice sheet margin, being
an area prone to local dust deposition. Overall, our study is
consistent with previous studies that have identified that the
local bare-ice mineral dust is poor in hematite and rich in
weakly absorbing quartz and feldspar minerals (e.g. Tedesco
et al., 2013). Tedesco et al. (2013) reported their dusts be-
ing redder than algae. However, their minerals were sourced
from cryoconite, not the ice surface, where glacier algae are
scarce and the biota is dominated by a rich consortium of
other microbes that lack the characteristic pigmentation of
glacier algae. Furthermore, Tedesco et al. (2013) reported an
average of only 0.3 % goethite in their Greenland cryoconite
samples. This may have been present as hematite prior to
their sample processing, which involved heating the samples
to 500–1000 ◦C. This heating treatment likely oxidized Fe-
bearing mineral phases, thereby artificially introducing the
observed reddening.
While these radiative-transfer simulations indicate that
mineral dust is unlikely to be directly causing the albedo
decline on the GrIS, they may still influence the ice albedo
indirectly by acting as substrates for the formation of low-
albedo microbial mineral aggregates known as cryoconite
granules, which are often found in quasi-cylindrical melt
holes or scattered over ice surfaces (Wharton et al., 1985;
Cook et al., 2015a) or by providing a nutrient source stim-
ulating algal growth (Stibal et al., 2017). This is especially
true because there is evidence in the previous literature that
the dust present on the GrIS bare-ice surface is likely derived
www.the-cryosphere.net/14/309/2020/ The Cryosphere, 14, 309–330, 2020

322 J. M. Cook et al.: Glacier algae accelerate melt rates
from a local source with no contribution from Asian dusts
or volcanic ash (Wientjes et al., 2011) and that red miner-
als such as hematite, goethite and ilmenite are present only
in very low concentrations (Wientjes et al., 2011; Tedesco et
al., 2013; Sanna and Romeo, 2018) that would have a negli-
gible effect on the ice albedo.
Therefore, we have demonstrated using empirical mea-
surements and radiative-transfer modelling that glacier algae
are potent albedo reducers on the south-western Greenland
Ice Sheet and mineral dusts are not. These findings are con-
sistent with several previous studies (Stibal et al., 2017; Yal-
lop et al., 2012) that found mineral dust to be insignificant
for explaining albedo variations in the same region.
3.3 Indirect effects of algae
Algae predominantly reduce the ice albedo in the visible
wavelengths (0.40–0.70 µm), whereas variations in the NIR
result mainly from changes to ice grain radii and the pres-
ence of liquid water (Warren, 1982; Green et al., 2002). Vari-
ations in the NIR albedo between the surface classes there-
fore suggest that the lower albedo of algal surfaces is not ex-
plained entirely by enhanced absorption due to algae but also
by the smoother, wetter ice surface with fewer opportuni-
ties for high-angle scattering of photons (Jonsell et al., 2003)
compared to the well-drained and porous CI surfaces. The
spatial and temporal development of the weathering crust
is therefore an important control on ice albedo (Muller and
Keeler, 1969; Jonsell et al., 2003). Algal growth is stimu-
lated by melt, which can be enhanced by algal growth (Yal-
lop et al., 2012; Ganey et al., 2017; Stibal et al., 2017; Cook
et al., 2017a, b; Dial et al., 2018) – an example of a bio-
cryomorphic process where biota alter the physical, chemical
and hydrological conditions of the ice surface with beneficial
consequences to the biota (Cook et al., 2015b).
3.4 Algae enhance radiative forcing and melt
Having determined that glacier algae reduce the ice surface
albedo, we took an empirical approach to quantifying their
impact upon energy balance following Ganey et al. (2017),
which includes both direct albedo effects (enhanced absorp-
tion of shortwave solar radiation by the algal cells) and the
indirect effects explained above. Integrated over the entire
day, this indicated a daily mean biological radiative forcing
of 116 and 65 W m−2for Hbio and Lbio surfaces, respec-
tively, similar to RFs for Alaskan snow algae calculated by
Ganey et al. (2017). We used the biological radiative forc-
ing integrated over the entire day and the latent heat of fu-
sion for ice (334 J cm−3) to estimate 1.35 ±0.01 (standard
error, SE) cm w.e. of melting due to algae in Hbio areas on
21 July. For Lbio sites, biological melting on 21 July 2017
was 1.01 ±0.01 (SE) cm w.e.
We corroborated this estimate using a point surface energy
balance model (Brock and Arnold et al., 2000). The melt at-
tributed to the presence of algae predicted by the energy bal-
ance modelling method was similar to that predicted using
the radiative forcing method, with 1.37 ±0.48 (SE) cm w.e.
attributed to Hbio and 0.95 ±0.41 (SE) cm w.e. attributed to
Lbio. Expressing the melt attributed to algae as a proportion
of the total melting in the algal sites gives 26.15 ±3.77 %
(SE) of the local melting attributed to algae in the Hbio sur-
faces and 21.62 ±5.07 % (SE) for Lbio surfaces.
3.5 Algae are widespread across the south-western
ablation zone
Our analyses demonstrate that algae have a dramatic dark-
ening effect on the ice surface, leading to increased melt-
ing. However, the importance of this effect depends upon the
spatial extent of the algal blooms over thousands of kilome-
tres. To determine spatial coverage at our field site we clas-
sified multispectral images acquired from a UAV flown over
a 200 ×200 m area. The classified UAV image indicated that
78.5 % of the area was covered by algal blooms, of which
61.1 % was Lbio and 17.4 % was Hbio (Table 3; Fig. 7). The
high ground resolution of the imagery enabled a qualitative
assessment of the algorithm performance by visual compar-
ison between the classifier and the raw imagery (following
Ryan et al., 2018a). The algorithm produced qualitatively re-
alistic bloom shapes, correctly placed water in channels and
individual cryoconite holes in their correct positions. The
confusion matrix indicates that occasional misclassifications
are generally between water and cryoconite (Sect. S8). This
is unsurprising since both cryoconite and water have rela-
tively flat spectral shapes with few spectral features and cry-
oconite is often found beneath pools of surface water. We
also point out that our cryoconite spectral reflectance mea-
surements were made with cryoconite filling the entire field
of view of the spectrometer and thus best represent large
cryoconite holes or dispersed cryoconite rather than surfaces
peppered with many small holes. There was also some am-
biguity between thin, wet snow and bare glacier ice, as these
surfaces are spectrally similar. Nevertheless, these misclassi-
fications affect a small area of the pixel and do not affect our
estimate of algal bloom coverage.
We also classified Sentinel-2 satellite data (Fig. 7). The
confusion matrices (Sect. S8) indicate similar misclassifi-
cation types and frequencies to the UAV model. The pre-
dicted algal coverage was 58.87 %. Hbio surfaces were much
less common than Lbio (Hbio =2.53 %; Lbio =56.54 %; Ta-
ble 3). The spatial coverage by algae was different in the Sen-
tinel and UAV datasets especially for Hbio, likely because
(a) the Sentinel-2 imagery includes ice that is outside of the
dark zone, raising the overall reflectivity, and (b) even in the
UAV image, which was retrieved from within the dark zone,
Hbio surfaces comprise just 17 % of the ice surface and have a
patchy distribution. The lowest-albedo surfaces – cryoconite
and water – cover a small fraction (<3 %) of the total area
in both UAV and Sentinel-2 images (Table 4), although we
The Cryosphere, 14, 309–330, 2020 www.the-cryosphere.net/14/309/2020/

J. M. Cook et al.: Glacier algae accelerate melt rates 323
Table 3. Summary of the albedo for each surface class as predicted from our classified UAV image, the Sentinel-2 image for 2016 (S2 2016)
and 2017 (S2 2017), and as measured using field spectroscopy (ASD Field Spec) at our field site in 2017 (we do not have cosine collector
albedo measurements for water or cryoconite surfaces). The reported values are the mean, the standard deviation in brackets and the number
of observations.
WAT CC CI Lbio Hbio SN
UAV 0.31 (0.017)
n=160 448
0.09 (0.031)
n=154 070
0.53 (0.026)
n=2 735 603
0.44 (0.055)
n=12 098 635
0.25 (0.039)
n=3 447 152
0.74 (0.025)
n=63 647
S2 2016 0.08 (0.044)
n=52 060
0.13 (0.035)
n=272 419
0.46 (0.042)
n=6 771 763
0.32 (0.046)
n=8 388 680
0.23 (0.028)
n=1 410 095
0.60 (0.05)
n=9924
S2 2017 0.08 (0.039)
n=174 791
0.11 (0.034)
n=258 520
0.46 (0.075)
n=5 947 314
0.31 (0.042)
n=8 740 186
0.22 (0.026)
n=2 270 206
0.76 (0.058)
n=16 333 853
ASD field spec n/a n/a 0.50 (0.02)
n=22
0.36 (0.07)
n=28
0.24 (0.03)
n=22
0.56 (0.10)
n=5
Figure 7. (a) Classified map of the area shown in (c) for 2016. (b) Broadband albedo map of the area shown in (c) for 2016. (c) RGB “true
colour” image showing the Sentinel-2 tile covering our field site in the Kangerlussuaq area. (d) Classified map of the area shown in (c) for
2017. (e) Broadband albedo map of the area shown in (c) for 2017. (f) Classified map of a 200 ×200 m area at the field site marked in (c)
imaged using a UAV-mounted multispectral camera. (g) Broadband albedo map of a 200×200 m area at the field site marked in (c) imaged
using a UAV-mounted multispectral camera. Panels (a–e) all have pixel resolution of 20 m and use the base geographic coordinate reference
system WGS84 and the Universal Tranvserse Mercator Projection (UTM) 22N. The scale bar beneath (b) is common to (a, b, d, e), and the
scale bar beneath (f) is common to (f) and (g).
www.the-cryosphere.net/14/309/2020/ The Cryosphere, 14, 309–330, 2020

324 J. M. Cook et al.: Glacier algae accelerate melt rates
Table 4. Percentage of each image covered by each surface type as predicted by our trained RF algorithm. Snow was removed from the
calculation in the Sentinel-2 images to enable quantification of surface coverage in the bare-ice zone, i.e. below the snow line, only.
UAV Image Sentinel-2 (2016) Sentinel-2 (2017)
Total image area (km2) 0.04 10 000 10 000
Total algae (%) 78.5 57.99 58.87
Hbio (%) 17.4 8.35 2.54
Lbio (%) 61.08 49.65 56.33
Cryoconite (%) 0.82 1.61 1.67
Clean ice (%) 13.81 40.08 38.34
Water (%) 0.78 0.31 1.13
Snow (%) 6.09 n/a n/a
note that many individual cryoconite holes will not be de-
tected as they are smaller than the spatial resolution of ei-
ther Sentinel-2 or the UAV. The spatial coverage reported
here from our multispectral UAV imagery is consistent with
a k-nearest neighbours classification scheme applied to RGB
(red, green, blue) imagery from a fixed wing UAV flight over
the Kangerlussuaq region by Ryan et al. (2018a). They found
up to 85 % of the ice surface to be composed of “ice contain-
ing uniformly distributed impurities” in the same region of
the dark zone in July 2014, which our observations confirm
were dominated by algae. They also found <2 % of the ice
surface to be cryoconite-covered and that water coverage was
<5 % (except for a supraglacial lake in their imaged area).
This analysis demonstrates that algae are a major component
of the ice surface. The larger spatial coverage of algae ob-
served in UAV images compared to Sentinel-2 images likely
results from spatial integration occurring at the coarser spa-
tial resolution associated with Sentinel-2 data, where pixels
are likely to be classified as CI unless the majority of the
pixel is covered by algae. Smaller Hbio patches are rarely
detected, presumably because they are unlikely to cover the
majority of a 20 m pixel. The higher detection limit for al-
gae with decreasing ground resolution makes our estimate of
spatial coverage from Sentinel-2 conservative. We highlight
that this will have a much larger effect on studies aiming to
quantify cell abundance using Sentinel-3 where the ground
resolution is 300 m.
3.6 Algae reduce the ice albedo across the
south-western ablation zone
There was a significant difference between the albedos of
each surface class in all four datasets, consistent with the
findings from our ground spectroscopy (Table 4). The albedo
of each surface class is approximately consistent between
the datasets, despite the variation in spatial coverage, giv-
ing confidence in the accuracy of our remote-sensing albedo
retrievals and the classification algorithm. In the expansive
areas where algae are present (Fig. 7), the ice albedo is
on average 0.13 lower for Lbio and 0.25 lower for Hbio
compared to clean ice (Table 4). This, combined with our
ground-based spectroscopy, radiative forcing calculations,
and radiative-transfer and energy balance modelling, pro-
vides robust evidence in support of algae having a significant
melt-accelerating effect on the GrIS. We cannot yet explic-
itly separate mineral and biological effects, but our theoreti-
cal and empirical analyses indicate that (a) local mineral dust
cannot explain the observed albedo reduction, (b) low-albedo
areas had significantly elevated algal cell numbers relative to
clean ice, (c) uniquely biological features were detectable in
the spectra and derivative spectra for the lower-albedo sites,
and (d) radiative-transfer models incorporating algal cells
with realistic pigment profiles demonstrate the mechanism of
albedo reduction. These observations confirm that supervised
classification of Hbio and Lbio surfaces is indeed detecting
surfaces with high algal loading and can be used to estimate
algal bloom extent. Again, we point out that this estimate
is conservative because there is certain to be glacier algae
present in low numbers in some of the areas that are classified
as clean and Hbio patches are often smaller than the ground
resolution of Sentinel-2, raising their detection limit (Ted-
stone et al., 2019). Furthermore, these calculations consider
the total albedo-reducing effect, inclusive of ice structure and
meltwater feedbacks, not only the direct light-absorbing ef-
fects of the algal biomass.
3.7 Algae cause enhanced GrIS runoff
We ran a SMB model forced with local automatic weather
station and MODIS albedo observations (van As et al., 2012)
to estimate 45.5 Gt runoff from all bare ice and 33.8 Gt from
dark ice in 2017. We used the mean spatial coverage deter-
mined using our remote sensing in each year and our radia-
tive forcing calculations that attributed 21.62 ±5.07 (SE) %
of melting to algae in Lbio sites and 26.15 ±3.77 (SE) % in
Hbio sites to generate estimates for the GrIS runoff caused by
algal growth. We have provided upper and lower estimates
based on our two remote-sensing datasets because, while our
UAV is able to accurately map Hbio and Lbio surfaces, we
cannot be certain that the spatial coverage derived from the
200 ×200 m area is representative of the south-western dark
zone. At the same time, our Sentinel-2 remote sensing under-
The Cryosphere, 14, 309–330, 2020 www.the-cryosphere.net/14/309/2020/

J. M. Cook et al.: Glacier algae accelerate melt rates 325
estimates algal coverage because it includes ice outside of the
dark zone and Hbio patches are often too small to be resolved
at 20 m pixel resolution (Tedstone et al., 2019). Therefore,
we used the spatial coverage determined by our Sentinel-2
classification as a lower bound and spatial coverage deter-
mined by our UAV classification as an upper bound on our
estimate of total runoff attributed to the presence of algae.
We found that in 2017 between 4.4 and 6.0 Gt of ice loss
could be attributed to the growth of algae, representing 10 %–
13 % of the total runoff from the south-western GrIS, with the
lower estimate generated using algal coverage from Sentinel-
2 and the upper estimate generated using spatial coverage at
our field site from our UAV. When the calculations were re-
stricted to the dark zone only (i.e. excluding areas in the ab-
lation zone not classified as “dark”), algal contributions to
total runoff were up to 18 %. These calculations confirm that
algal growth is an important factor in the contribution of the
GrIS to global sea level rise. This contribution will increase
if biologically darkened areas expand or a greater proportion
of the ice is covered by high-biomass blooms under warmer
climates. These observations therefore indicate that the omis-
sion of biological growth is leading current models to under-
estimate future GrIS contributions to sea level rise.
3.8 Interannual variability and potential positive
feedback
MODIS data (Fig. 4) indicate that 2017 was a particularly
high-albedo year when the dark zone was especially small
and bright, whereas 2016 was a particularly low-albedo year
where the dark zone was wider and darker than most years
(Fig. 4a and b and Tedstone et al., 2017). Previous field ev-
idence (Williamson et al., 2018) demonstrates that the ice
was darkened by high concentrations of algae in 2016. In
our Sentinel-2 remote-sensing tile (22WEV) the bare-ice
zone was wider in 2016 (6758 km2) than in 2017 (6205 km2)
and a larger area was covered with algae (on 25 July 2016,
3919 km2was covered by algae compared to 3653 km2on
28 July 2017). While the proportional total algal cover-
age was similar between the 2 years (57.99 % in 2016 and
58.87 % in 2017), the proportion of the algal ice that was
classified Hbio was much higher in 2016 (8.35 %) compared
to 2017 (2.54 %). The mean albedos and their standard de-
viations were very similar for each ice surface class in both
years (Table 4). The runoff from the south-western GrIS bare
ice (albedo <0.6) was 94.1 Gt in 2016, of which 67.6 was
from dark ice (albedo <0.39). We estimate that 8.8–12.2 Gt
of this runoff was attributable to the growth of algae, repre-
senting 9 %–13 % of the total runoff from bare-ice sectors.
The absolute values for runoff are therefore much higher but
the proportion of the bare-ice total attributed to algae was
approximately the same between the 2 years.
The snow line retreated further and earlier in 2016 com-
pared to 2017, creating a wider bare-ice zone that existed for
longer and was not transiently covered by summer snowfall
events, whereas in 2017 a smaller bare-ice area was exposed
later and was covered by 5–10 cm of snow several times dur-
ing the summer (Fig. 4c, d). The more prolonged exposure
of a larger bare-ice zone in 2016 enabled Lbio surfaces to
extend to higher elevations and biomass to accumulate to
greater mass concentrations at lower elevations in summer
2016, explaining the greater Hbio coverage. This indicates
that the intensity of the algal bloom is a function of exposure
time, as postulated by Tedstone et al. (2017) and Williamson
et al. (2018). More prolonged exposure of larger ablation ar-
eas under a warming climate (Stroeve et al., 2013; Shimada
et al., 2016; Tedesco et al., 2016; Tedstone et al., 2017) are
likely to be prone to more spatially expansive, darker algal
blooms that enhance melt rates, leading to a potential pos-
itive feedback that is not currently accounted for in surface
mass balance models, whereby earlier exposure of bare ice
leads to enhanced algal coverage, which will be able to ac-
cumulate higher biomass and accelerate melting. Melting in
turn stimulates algal growth by liberating nutrients and liquid
water.
4 Conclusions
Our measurements and modelling demonstrate that the
growth of algae on the GrIS accelerates the rate of melt-
ing and increases the GrIS contribution to global sea level
rise. Field spectra show a dramatic depression of the sur-
face albedo in the visible wavelengths for surfaces contami-
nated by algae. Derivative analysis of the same spectra show
uniquely biological absorption features and an inverse rela-
tionship was observed between biomass and surface albedo.
We employ a novel radiative-transfer model to show that
this albedo decline cannot be attributed to local mineral
dusts. Radiative forcing calculations and an energy balance
model predict that melting of glacier ice can be accelerated
by 21.62 ±5.07 (SE) % for Lbio surfaces and 26.15 ±3.77
(SE) % for Hbio surfaces. We demonstrate that the growth
of algae occurs over a large proportion of the ablating area
of the south-western GrIS by identifying algal blooms in
remote-sensing data from a UAV and Sentinel-2, finding
78.5 % of the surface within a 200 ×200 m sample area at
our field site to be covered by algae. Using Sentinel-2 we de-
tected algae covering 57.99 % of the Kangerlussuaq region in
2017 and 58.87 % of the same region in 2016. The spatial res-
olution of the sensor makes these conservative estimates, es-
pecially for Hbio surfaces. Runoff modelling informed by our
field measurements and remote-sensing estimate between 4.4
and 6.0 Gt of runoff from the south-western ablation zone
could be attributed to the growth of algae in summer 2017,
representing 10 %–13 % of the total. Because 2017 was a par-
ticularly high-albedo year for the south-western GrIS, we
also ran our analysis for the particularly low-albedo 2016
melt season. In 2016 a wider bare-ice zone was exposed for
longer, and there was a concomitant increase in the extent
www.the-cryosphere.net/14/309/2020/ The Cryosphere, 14, 309–330, 2020

326 J. M. Cook et al.: Glacier algae accelerate melt rates
of the algal bloom, more of which was classified as Hbio
(high biomass). The percentage algal contribution to south-
western GrIS runoff was approximately the same as in 2017
(9 %–13 %), but the absolute volume was much higher (8.8–
12.2 Gt). This interannual comparison indicates the existence
of a feedback because in years where snow retreats further
and earlier, there is a larger and more prolonged area for
algal bloom development where melting is enhanced, stim-
ulating further algal growth. This study therefore demon-
strates that algae are important albedo reducers and cause a
melt-enhancing feedback across the south-western GrIS. The
omission of these critical biological albedo feedbacks from
predictive models of GrIS runoff is leading to underestima-
tion of future ice mass loss and contribution to global sea
level rise. This is particularly significant because larger abla-
tion zones and longer growth seasons are expected in a future
warmer climate.
Code and data availability. Codes and datasets used in this study
are available at the following DOIs.
–BioSNICAR-GO code and data:
https://doi.org/10.5281/zenodo.3564517 (Cook et al., 2020a).
–Ice surface classification codes:
https://doi.org/10.5281/zenodo.3564529 (Cook et al., 2020b).
–Spectra processing codes:
https://doi.org/10.5281/zenodo.2598219 (Cook, 2020).
–Field and associated data:
https://doi.org/10.5281/zenodo.3564501 (Cook et al., 2020c).
Supplement. The supplement related to this article is available on-
line at: https://doi.org/10.5194/tc-14-309-2020-supplement.
Author contributions. JMC developed the measurement protocol,
gathered field measurements, analysed the data, wrote the main
code, curated the data repository, produced the figures and wrote
the manuscript. OMcA was instrumental in building and testing the
UAV. AJT, CW, JM and SH gathered field data. CW provided advice
regarding microscopy and biological sampling protocols, helped
with experimental design, and led the empirical measurements of
glacier algae pigmentation and absorption coefficients. AJT wrote
the code for radiometric calibration of multispectral imagery from
the UAV and post-processed the UAV images, derived 2016 and
2017 dark-ice extent from MODIS imagery, analysed MAR snow
depth outputs, produced Fig. 4, translated the energy balance model
into Python, and made significant contributions to the manuscript
writing and experimental design. JM and SM provided cleaned min-
eral dust and PSD data to feed into the radiative-transfer model,
and JM provided useful discussions regarding experimental design.
MS provided DISORT modelling and estimates of mineral dust re-
fractive indices. MF helped develop the bio-optical model. RB pro-
vided advice regarding field spectroscopy and helped measure min-
eral dust refractive indices in the laboratory. AJH helped develop
the experimental design. AJH, JR and AM all provided advice on
UAV remote sensing. JR, DvA and AJH modelled runoff from the
GrIS dark zone. AD provided microscopy images from field sam-
ples. Other authors contributed to field work and/or sample prepa-
ration and commented on the style and content of the final paper.
Competing interests. The authors declare that they have no conflict
of interest.
Acknowledgements. Joseph M. Cook, Andrew J. Tedstone, An-
drew J. Hodson, Christopher Williamson, Archana Dayal, Stefan
Hofer, Andrew McGonigle, Alexandre M. Anesio, Tristram D. L.
Irvine-Fynn, Edward Hanna, Marian Yallop and Martyn Tranter
acknowledge funding from UK National Environmental Research
Council large grant “Black and Bloom”. Joseph M. Cook grate-
fully acknowledges the Rolex Awards for Enterprise, National Ge-
ographic and Microsoft (“AI for Earth”) and NERC Standard Grant
“MicroMelt” . Liane G. Benning, Jenine McCutcheon and James
B. McQuaid acknowledge funding from the UK National Envi-
ronmental Research Council large grant “Black and Bloom”, and
Liane G. Benning and Sathish Mayanna acknowledge funding from
the German Helmholtz Recruiting Initiative. Thomas Gribbin ac-
knowledges the Gino Watkins Memorial Fund and Nottingham Ed-
ucation Trust. Greenland Analogue Project (GAP) weather station
data are made available through the Programme for Monitoring of
the Greenland Ice Sheet (http://www.promice.dk/home.html, last
access: January 2020). MAR v3.8.1 regional climate model out-
puts used to estimate mean snow depth were provided by Xavier
Fettweis. We thank Stephen Warren his for helpful comments that
improved the manuscript.
Financial support. This research has been supported by the
UK National Environmental Research Council (grant no.
NE/M021025/1), the UK National Environmental Research Coun-
cil (grant nos. NE/M020770/1, NE/S001034/1), and the German
Helmholtz Recruiting Initiative (grant no. I-044-16-01).
Review statement. This paper was edited by Marco Tedesco and re-
viewed by Stephen Warren and one anonymous referee.
References
Bamber, J., Westaway, R. M., Marzeion, B., and Wouters, B.:
The land ice contribution to sea level during the satellite era,
Environ. Res. Lett., 13 063008, https://doi.org/10.1088/1748-
9326/aac2f0, 2018.
Benning, L. G., Anesio, A. M., Lutz, S., and Tranter, M.: Bi-
ological impact on Greenland’s albedo, Nat. Geosci., 7, 691,
https://doi.org/10.1038/ngeo2260, 2014.
Blankenship, R. E., Tiede, D. M., Barber, J., Brudvig, G. W., Flem-
ing, G., Ghirardi, M., Gunner, M. R., Junge, W., Kramer, D. M.,
Melis, A., Moore, T. A., Moser, C. C., Nocera, D. G., Nozik, A.
J., Ort, D. R., Parson, W. W., Prince, R. C., and Sayre, R. T.:
Comparing photosynthetic and photovoltaic efficiencies and rec-
The Cryosphere, 14, 309–330, 2020 www.the-cryosphere.net/14/309/2020/

J. M. Cook et al.: Glacier algae accelerate melt rates 327
ognizing the potential for improvement, Science, 332, 805–809,
2011.
Bøggild, C. E., Brandt, R. E., Brown, K. J., and Warren, S. G.: The
ablation zone in northeast Greenland: ice types, albedos and im-
purities, J. Glaciol., 56, 101–113, 2010.
Box, J. E., Fettweis, X., Stroeve, J. C., Tedesco, M., Hall, D. K.,
and Steffen, K.: Greenland ice sheet albedo feedback: thermo-
dynamics and atmospheric drivers, The Cryosphere, 6, 821–839,
https://doi.org/10.5194/tc-6-821-2012, 2012.
Brock, B. W. and Arnold, N. S.: A spreadsheet-based (Mi-
crosoft Excel) point surface energy balance model for glacier
and snow melt studies, Earth Surf. Proc. Land., 25, 649–
658. https://doi.org/10.1002/1096-9837(200006)25:6<649::aid-
esp97>3.0.co;2-u, 2000.
Cook, J. M., Edwards, A., Irvine-Fynn, T. D. I., and Takeuchi, N.:
Cryoconite: Dark biological secret of the cryosphere, Prog. Phys.
Geog., 40, 66–111, https://doi.org/10.1177/0309133315616574,
2015a.
Cook, J. M., Edwards, A., and Hubbard, A.: Biocryomorphol-
ogy: Integrating Microbial Processes with Ice Surface Hydrol-
ogy, Topography, and Roughness, Front. Earth Sci., 3, 78,
https://doi.org/10.3389/feart.2015.00078, 2015b.
Cook, J. M., Hodson, A. J., Taggart, A. J., Mernild, S.
H., and Tranter, M.: A predictive model for the spectral
“bioalbedo” of snow, J. Geophys. Res.-Earth Surf., 122, 434–
454, https://doi.org/10.1002/2016JF003932, 2017a.
Cook, J. M., Hodson, A. J., Gardner, A. S., Flanner, M., Tedstone,
A. J., Williamson, C., Irvine-Fynn, T. D. L., Nilsson, J., Bryant,
R., and Tranter, M.: Quantifying bioalbedo: a new physically
based model and discussion of empirical methods for characteris-
ing biological influence on ice and snow albedo, The Cryosphere,
11, 2611–2632, https://doi.org/10.5194/tc-11-2611-2017, 2017b.
Cook, J. M.: Spectra Processing Codes, Zenodo,
https://doi.org/10.5281/zenodo.2598219, last access: January
2020.
Cook, J. M., Williamson, C., Tedstone, A. J., Mc-
Cutcheon, J., and Flanner, M.: BioSNICAR-GO, Zenodo,
https://doi.org/10.5281/zenodo.3564517, last access: January
2020a.
Cook, J. M., Tedstone, A. J., Williamson, A. J., and
McCutcheon, J.: Ice surface classifiers, Zenodo,
https://doi.org/10.5281/zenodo.3564529, last access: January
2020b.
Cook, J. M., Tedstone, A. J., Williamson, C., and Mc-
Cutcheon, J.: Field and associated data, Zenodo,
https://doi.org/10.5281/zenodo.3564501, last access: January
2020c.
Dauchet, J., Blanco, S., Cornet, J.-F., and Fournier, R.: Calculation
of radiative properties of photnthetic microorganisms, J. Quant.
Spectrosc. Ra., 161, 60–84, 2015.
Dee, D. P., Uppala, S. M., Simmons, A. J., Berrisford, P., Poli,
P., Kobayashi, S., Andrae, U., Balmaseda, M. A., Balsamo, G.,
Bauer, P., Bechtold, P., Beljaars, A. C. M., van de Berg, I., Biblot,
J., Bormann, N., Delsol, C., Dragani, R., Fuentes, M., Greer, A.
J., Haimberger, L., Healy, S. B., Hersbach, H., Holm, E. V., Isak-
sen, L., Kallberg, P., Kohler, M., Matricardi, M., McNally, A. P.,
Mong-Sanz, B. M., Morcette, J.-J., Park, B.-K., Peubey, C., de
Rosnay, P., Tavolato, C., Thepaut, J. N., and Vitart, F.: The ERA-
Interim reanalysis: Configuration and performance of the data
assimilation system, Q. J. Roy. Meteorol. Soc., 137, 553–597,
https://doi.org/10.1002/qj.828, 2011.
Dial, R., Ganey, G., and Skiles, S. M.: What colour should
glacier algae be? An ecological role for red carbon in
the cryosphere, FEMS Microbiol. Ecol., 94, fiy2007,
https://doi.org/10.1093/femsec/fiy007, 2018.
Egan, W. G. and Hilgeman, T. W.: Optical properties of inho-
mogenous materials: applications to geology, astronomy, chem-
istry and engineering, Academic Press, San Diego, USA, ISBN:
0122326504, 1979.
Enderlin, E. M., Howat, I. M., Jeong, S., Noh, M.-J., van Angelen,
J. H., and van den Broeke, M. R.: An improved mass budget
for the Greenland ice sheet, Geophys. Res. Lett., 41, 866–872,
https://doi.org/10.1002/2013GL059010, 2014.
Fettweis, X., Box, J. E., Agosta, C., Amory, C., Kittel, C., Lang, C.,
van As, D., Machguth, H., and Gallée, H.: Reconstructions of the
1900–2015 Greenland ice sheet surface mass balance using the
regional climate MAR model, The Cryosphere, 11, 1015–1033,
https://doi.org/10.5194/tc-11-1015-2017, 2017.
Flanner, M. G., Zender, C. S., Randerson, J. T., and Rasch,
P. J.: Present-day climate forcing and response from
black carbon in snow, J. Geophys. Res., 112, D11202,
https://doi.org/10.1029/2006JD008003, 2007.
Ganey, G. Q., Loso, M. G., Burgess, A. B., and Dial, R.
J.: The role of microbes in snowmelt and radiative forc-
ing on an Alaskan icefield, Nat. Geosci., 10, 754–759,
https://doi.org/10.1038/ngeo3027, 2017.
GDAL/OGR contributors: GDAL/OGR Geospatial Data Abstrac-
tion software Library, Open Source Geospatial Foundation, avail-
able at: https://gdal.org/ (last access: January 2020), 2019.
Green, R. O., Dozier, J., Roberts, D., and Painter, T. H.: Spectral
snow-reflectance models for grain size and liquid water fraction
in melting snow for the solar reflected spectrum, Ann. Glaciol.,
34, 71–73, 2002.
Greene, C. A., Gwyther, D. E., and Blankenship, D. D.: Antarc-
tic Mapping Tools for Matlab, Comput. Geosci., 104, 151–157,
https://doi.org/10.1016/j.cageo.2016.08.003. 2017.
Haardt, H. and Maske, H.: Specific in vivo absorption coefficient of
chlorophyll a at 675 nm, Limnol. Oceanogr., 32, 608–619, 1987.
Hanna, E., Navarro, F. J., Pattyn, F., Domingues, C. M., Fet-
tweiss, X., Ivins, E. R., Nicholls, R. J., Ritz, C., Smith,
B., Tulaczyk, S., Whitehouse, P., and Zwally, J.: Ice sheet
mass balance and climate change. Nature, 498, 51–59,
https://doi.org/10.1038/nature12238, 2013.
Hillebrand, H., Dürselen, C.-D., Kirschtel, D., Pollingher, U., and
Zohary, T.: Biovolume calculation for pelagic and benthic mi-
croalgae, J. Phycol., 35, 403–24, 1999.
Howat, I.: MEaSUREs Greenland Ice Mapping Project
(GIMP) Land Ice and Ocean Classification Mask, Ver-
sion 1., NASA National Snow and Ice Data Center Dis-
tributed Active Archive Center, Boulder, Colorado USA,
https://doi.org/10.5067/B8X58MQBFUPA (last access: August
2018), 2017.
Howat, I. M., Negrete, A., and Smith, B. E.: The Green-
land Ice Mapping Project (GIMP) land classification and
surface elevation data sets, The Cryosphere, 8, 1509–1518,
https://doi.org/10.5194/tc-8-1509-2014, 2014.
Hu, W.: Dry weight and cell density of individual algal
and cyanobacterial cells for algae research and develop-
www.the-cryosphere.net/14/309/2020/ The Cryosphere, 14, 309–330, 2020

328 J. M. Cook et al.: Glacier algae accelerate melt rates
ment, PhD thesis, University of Missouri-Columbia, avail-
able at: https://mospace.umsystem.edu/xmlui/bitstream/handle/
10355/46477/research.pdf (last access: September 2019), 2014.
Huovinen, P., Ramirez, J., and Gomez, I.: Remote sensing of
albedo-reducing snow algae and impurities in the Maritime
Antarctic, ISPRS J. Photogramm., 146, 507–517, 2018.
Jäger, C., Fabian, D., Schrempel, F., Dorschner, J., Hen-
ning, T., and Wesch, W.: Structural processing of en-
statite by ion bombardement, Astron. Astrophys. 401, 57–65,
https://doi.org/10.1051/0004-6361:20030002, 2003.
Jonsell, U., Hock, R., and Holmgren, B.: Spatial and temporal vari-
ations in albedo on Storglaciären, Sweden, J. Glaciol., 49, 59–68,
https://doi.org/10.3189/172756503781830980, 2003.
Kirk, J. T. O.: A theoretical analysis of the contribution of algal cells
to the attenuation of light within natural waters III. Cylindrical
and spheroidal cells, New Phytol., 77, 341–358, 1976.
Knap, W. H., Brock, B. W., Oerlemans, J., and Willis, I. C.: Compar-
ison of Landsat TM- derived and ground-based albedos of Haut
Glacier d’Arolla, Switzerland, Int. J. Remote Sens., 20, 3293–
3310, 1999.
Langen, P. L., Fauso, R. S., Vendecrux, B., Mottram, R. H., and
Box, J. E.: Liquid water flow and retention on the Green-
land Ice Sheet in the regional climate model HIRHAM5:
local and large scale impacts, Front. Earth Sci., 4, 110,
https://doi.org/10.3389/feart.2016.00110, 2017.
Lee, E. and Pilon, L.: Absorption and scattering by long and ran-
domly oriented linear chains of spheres, J. Opt. Soc. Am., 30,
1892–1900, 2013.
Leya, T.: Fedlstudien und genetische Untersuchungen zur Ky-
rophilie der Schneealgen Nordwestspitzbergens, Shaker Verlag,
Aachen, 2014.
Liang, S.: Narrowband to broadband conversions of land
surface albedo I. Remote Sens. Environ., 76, 213–238,
https://doi.org/10.1016/S0034-4257(00)00205-4, 2001.
Lutz, S., Anesio, A. M., Jorge Villar, S. E., Benning, L.
G.: Variations of algal communities cause darkening of a
Greenland glacier, FEMS Microbiol. Ecol., 89, 402–414,
https://doi.org/10.1111/1574-6941.12351, 2014.
Lutz, S., McCutcheon, J., McQuaid, J. B., and Benning, L.
G.: The diversity of ice algal communities on the Greenland
Ice Sheet revealed by oligotyping, Microb. Genom., 4, 1–10,
https://doi.org/10.1099/mgen.0.000159, 2018.
Masojídek, J., Torzillo, G., and Koblížek, M.: Photosynthesis
in Microalgae, in: Handbook of Microalgal Culture, edited
by: Richmond, A. and Hu, Q., John Wiley and Sons, Ltd.,
https://doi.org/10.1002/9781118567166.ch2, 2013.
McCutcheon, J., Lutz, S., Williamson, C., Cook, J. M., Tedstone, A.
J., Vanderstraeten, A., Wilson, S. A., Stockdale, A., Bonneville,
S., McQuaid, J. B., Tranter, M., and Benning, L. G.: Mineral
phosphorous drives glacier algal blooms on the Greenland Ice
Sheet, in preparation, 2020.
Micasense: Red-Edge camera radiometric calibration model,
available at: https://support.micasense.com/hc/en-us/articles/
115000351194-RedEdge-Camera-Radiometric- Calibration-Model
(last access: November 2018), 2019.
Morel, A. and Bricaud, A.: Theoretical results concerning light ab-
sorption in a discrete medium, and application to specific absorp-
tion of phytoplankton, Deep-Sea Res., 28, 1375–1393, 1981.
Muller, F. and Keeler, C. M.: Errors in short term ablation measure-
ments on melting ice surfaces, J. Glaciol., 8, 91–105, 1969.
Ngheim, S. V., Hall, D. K., Mote, T. L., Tedesco, M. Al-
bert, M. R., Keegan, K., Shuman, C. A., DiGirolamo, N.
E., and Neumann, G.: The extreme melt across the Green-
land Ice Sheet in 2012, Geophys. Res. Lett., 39, L20502,
https://doi.org/10.1029/2012GL053611, 2012.
Nitsche, R. and Fritz, T.: Precise determination of the com-
plex optical constant of mica, Appl. Opt., 43, 3263,
https://doi.org/10.1364/ao.43.003263, 2004.
Noël, B., van de Berg, W. J., Machguth, H., Lhermitte, S., Howat,
I., Fettweis, X., and van den Broeke, M. R.: A daily, 1 km
resolution data set of downscaled Greenland ice sheet surface
mass balance (1958–2015), The Cryosphere, 10, 2361–2377,
https://doi.org/10.5194/tc-10-2361-2016, 2016.
Nordenskiöld, A. E.: Cryoconite found 1870, 19–25 July, on thein-
land ice, east of Auleitsivik Fjord, Disco Bay, Greenland, Geol.
Mag., 2, 157–162, 1875.
OCDB: Optical Constants Database, Laboratory Astrophysics
Group of the AIU Jena, Stubachtal Olivine dataset, avail-
able at: https://www.astro.uni-jena.de/Laboratory/OCDB/data/
silicate/crystalline/oliv_vis.txt (last access: November 2019),
2002.
Osborne, M.: Ardupilot Mission Planner (v1.3.48), available at:
https://ardupilot.org/ardupilot/index.html, 2017.
Painter, T. H., Duval, B., and Thomas, W. H.: Detection
and quantification of snow algae with an airborne imag-
ing spectrometer, Appl. Environ. Microbiol., 67, 5267–5272,
https://doi.org/10.1128/AEM.67.11.5267-5272.2001, 2001.
Pollack, J. B., Toon, O. B., and Khare, B. K.: Optical properties of
some terrestrial rocks and glasses, Icarus, 19, 372–389, 1973.
Remias, D., Schwaiger, S., Aigner, S., Leya, T., Stuppner,
H., and Lutz, C.: Characterization of an UV- and VIS-
absorbing, purpurogallin-derived secondary pigment new
to algae and highly abundant in Mesotaenium berggrenii
(Zygnematophyceae, Chlorophyta), an extremophyte liv-
ing on glaciers, FEMS Microbiol. Ecol., 79, 638–648,
https://doi.org/10.1111/j.1574-6941.2011.01245.x, 2012.
Rothman, L. S., Rinsland, C. P., Goldman, A., Massie, S. T., Ed-
wards, D. P., Flaud, J.-M., Perrin, A., Camy-Peyrey, C., Dana,
V., Mandin, J.-Y., Schroeder, J., McCann, A., Gamache, R. R.,
Wattson, R. B., Yohino, K., Chance, K. V., Jucks, K. W., Brown,
L. R., Nemtchinov, V., and Varanasi, P.: The HITRAN molecu-
lar spectroscopic database and HAWKS (HITRAN atmospheric
workstation): 1996 edition, J. Quant. Spectrosc. Ra., 60, 665–
710, https://doi.org/10.1016/S0022-4073(98)00078-8, 1998.
Roush, T., Pollack, J., and Orenberg, J.: Derivation of midinfrared
(5-25 µm) optical constants of some silicates and palagonite,
Icarus, 94, 191–208, 1991.
Ryan, J., Hubbard, A., Irvine-Fynn, T. D., Doyle, S. ., Cook, J. M.,
Stibal, M., and Box, J. E.: How robust are in situ observations
for validating satellite-derived albedo over the dark zone of the
Greenland Ice Sheet?, Geophys. Res. Lett., 44, 6218–6225, 2017.
Ryan, J., Hubbard, A., Irvine-Fynn, Cook, J. T., Smith, L. C.,
Cameron, K., and Box, J. E.: Dark zone of the Greenland
Ice Sheet controlled by distributed biologically-active impuri-
ties, Nat. Commun., 9, 1065, https://doi.org/10.1038/s41467-
018-03353-2, 2018a.
The Cryosphere, 14, 309–330, 2020 www.the-cryosphere.net/14/309/2020/

J. M. Cook et al.: Glacier algae accelerate melt rates 329
Ryan, J., van As, D., Cooley, S. W., Cooper, M. G., Pitcher, L.
H., and Hubbard, A.: Greenland Ice Sheet surface melt ampli-
fied by snow line migration and bare ice exposure, Sci. Adv., 5,
eeav3738, https://doi.org/10.1126/sciadv.aav3738, 2018b.
Sanna, L. and Romeo, A.: Mineralogy and geochemistry of cry-
oconite sediments in Eqip Sermia glacier (central-west Green-
land), J. Mediterr. Earth Sci., 10, 159–166, 2018.
Seager, S., Turner, E. L., Schafer, J., and Ford, E. B.: Vegetation’s
red edge: a possible spectroscopic biosignature of extraterrestrial
plants, Astrobiology, 5, 372–390, 2005.
Shepherd, A., Ivins, E. R., Barletta, V. R., Bentley, M. J., Bettad-
pur, S., Briggs, K. H., Bromwich, D. H., Forsberg, R., Galin, N.,
Horwath, M., Jacobs, S., Joughin, I., King, M. A., Lenaerts, J.
T. M., Li, J., Ligtenberg, S. R. M., Luckman, A., Luthcke, S.
B., McMillan, M., Meister, R., Milne, G., Mouginot, J., Muir,
A., Nicolas, J. P., Paden, J., Payne, A. J., Pritchard, H., Rig-
not, E., Rott, H., Sohn, H.G. Rensen, L. S., Scambos, T. A.,
Scheuchl, B., Schrama, E. J. O., Smith, B., Sundal, A. V., van
Angelen, J. H., van de Berg, W. J., van den Broeke, M. R.,
Vaughan, D. G., Velicogna, I., Wahr, J., Whitehouse, P. L., Wing-
ham, D. J., Yi, D., Young, D., and Zwally, H. J.: A Reconciled
Estimate of Ice-Sheet Mass Balance, Science, 338, 1183–1189,
https://doi.org/10.1126/science.1228102, 2012.
Shimada, R., Takeuchi, N., and Aoki, T.: Inter-annual and geograph-
ical variations in the extent of bare ice and dark ice on the Green-
land ice sheet derived from MODIS satellite images, Front. Earth
Sci., 4, 43, https://doi.org/10.3389/feart.2016.00043, 2016.
Skiles, S. M., Painter, T. H., and Okin, G. S.: A method to retrieve
the spectral complex refractive index and single scattering optical
properties of dust deposited in mountain snow, J. Glaciol., 63,
133–147, https://doi.org/10.1017/jog.2016.126, 2017.
Smeets, C. J. P. P. and Van den Broeke, M. R.: Temporal and spa-
tial variations of the aerodynamic roughness length in the abla-
tion zone of the Greenland ice sheet, Bound.-Lay. Meteorol., 128,
315–338, https://doi.org/10.1007/s10546-008-9291-0, 2008.
Stibal, M., Box, J. E., Cameron, K. A., Langen, P. L., Yallop, M.,
Mottram, R. H., Khan, A. L., Molotch, N. P., Chrismas, N. A. M.,
Quaglia, F. C., Remias, D., Smeets, C. J. P., van den Broecke,
M. R., Ryan, J. C., Hubbard, A., Tranter, M., van As, D., and
Ahlstrøm, A. P.: Algae drive enhanced darkening of bare ice on
the Greenland Ice Sheet, Geophys. Res. Lett., 44, 11463–11471,
2017.
Stroeve, J., Box, J. E., Wang, Z., Schaaf, C., and Barett,
A.: Re-evaluation of MODIS MCD43 Greenland albedo ac-
curacy and trends, Remote Sens. Environ., 138, 99–214.
https://doi.org/10.1016/j.rse.2013.07.023, 2013.
Takeuchi, N., Dial, R., Kohshima, S., Segawa, T., and
Uetake, J.: Spatial distribution and abundance of red
snow algae on the Harding Icefield, Alaska derived
from a satellite image, Geophys. Res. Lett., 33, L21502,
https://doi.org/10.1029/2006GL027819, 2006.
Tedesco, M., Foreman, C., Anton, J., Steiner, N., and Schwartzman,
T.: Comparative analysis of morphological, mineralogical and
spectral properties of cryoconite in Jakobshavn Isbræ, Green-
land, and Canada Glacier, Antarctica, Ann. Glaciol., 54, 147–
157, https://doi.org/10.3189/2013AoG63A417, 2013.
Tedesco, M., Doherty, S., Fettweis, X., Alexander, P., Jeyaratnam,
J., and Stroeve, J.: The darkening of the Greenland ice sheet:
trends, drivers, and projections (1981–2100), The Cryosphere,
10, 477–496, https://doi.org/10.5194/tc-10-477-2016, 2016.
Tedstone, A. J.: Python implementation of point surface energy bal-
ance model, Zenodo, https://doi.org/10.5281/zenodo.3228331,
2019.
Tedstone, A. J., Bamber, J. L., Cook, J. M., Williamson, C. J., Fet-
tweis, X., Hodson, A. J., and Tranter, M.: Dark ice dynamics of
the south-west Greenland Ice Sheet, The Cryosphere, 11, 2491–
2506, https://doi.org/10.5194/tc-11-2491-2017, 2017.
Tedstone, A. J., Cook, J. M., Williamson, C. J., Hofer, S., Mc-
Cutcheon, J., Irvine-Fynn, T., Gribbin, T., and Tranter, M.:
Algal growth and weathering crust structure drive variability
in Greenland Ice Sheet ice albedo, The Cryosphere Discuss.,
https://doi.org/10.5194/tc-2019-131, in review, 2019.
Toon, O. B., McKay, C. P., Ackerman, T. P., and San-
thanam, K.: Rapid calculation of radiative heating rates
and photodissociation rates in inhomogeneous multiple scat-
tering atmospheres, J. Geophys. Res., 94, 16287–16301,
https://doi.org/10.1029/JD094iD13p16287, 1989.
Uetake, J., Naganuma, T., Hebsgaard, M. B., Kanda, H.,
and Kohshima, S.: Communities of algae and cyanobacte-
ria on glaciers in west Greenland, Polar Sci., 4, 71–80,
https://doi.org/10.1016/j.polar.2010.03.002, 2010.
United States Geological Survey (USGS): Unmanned Air-
craft Systems data post-processing: Structure from mo-
tion photogrammetry, available at: https://uas.usgs.gov/nupo/
pdf/PhotoScanProcessingMicaSenseMar2017.pdf (last access:
May 2018), 2017.
van As, D., Van den Broeke, M. R., Reijmer, C. H., and Vande
Wal, R. S. W.: The summer surface energy balance of the
high Antarctic Plateau, Bound.-Lay. Meteorol., 115, 289–317,
https://doi.org/10.1007/s10546-004-4631-1, 2005.
van As, D., Hubbard, A. L., Hasholt, B., Mikkelsen, A. B., van
den Broeke, M. R., and Fausto, R. S.: Large surface meltwa-
ter discharge from the Kangerlussuaq sector of the Greenland
ice sheet during the record-warm year 2010 explained by de-
tailed energy balance observations, The Cryosphere, 6, 199–209,
https://doi.org/10.5194/tc-6-199-2012, 2012.
van As, D., Bech Mikkelsen, A., Holtegaard Nielsen, M., Box, J. E.,
Claesson Liljedahl, L., Lindbäck, K., Pitcher, L., and Hasholt, B.:
Hypsometric amplification and routing moderation of Greenland
ice sheet meltwater release, The Cryosphere, 11, 1371–1386,
https://doi.org/10.5194/tc-11-1371-2017, 2017.
van den Broeke, M. R., Enderlin, E. M., Howat, I. M., Kuipers
Munneke, P., Noël, B. P. Y., van de Berg, W. J., van Meijgaard,
E., and Wouters, B.: On the recent contribution of the Greenland
ice sheet to sea level change, The Cryosphere, 10, 1933–1946,
https://doi.org/10.5194/tc-10-1933-2016, 2016.
van Diedenhoven, B., Ackerman, A. S., Cairns, B., and Fridlind,
A. M.: A flexible paramaterization for shortwave optical
properties of ice crystals, J. Atmos. Sci., 71, 1763–1782,
https://doi.org/10.1175/JAS-D-13-0205.1, 2014.
Wang, S., Tedesco, M., Xu, M., and Alexander, P. M.: Mapping ice
algal blooms in southwest Greenland from space, Geophys. Res.
Lett., 45, 11779–11788, https://doi.org/10.1029/2018GL080455,
2018.
Warren, S. G.: Optical properties of snow, Rev. Geophys., 20, 67–
89, https://doi.org/10.1029/RG020i001p00067, 1982.
www.the-cryosphere.net/14/309/2020/ The Cryosphere, 14, 309–330, 2020

330 J. M. Cook et al.: Glacier algae accelerate melt rates
Warren, S. G. and Brandt, R. E.: Optical constants
of ice from the ultraviolet to the microwave: A re-
vised compilation, J. Geophys. Res., 113, D14220,
https://doi.org/10.1029/2007JD009744,2008, 2008.
Wharton, R. A., McKay, C. P., Simmons, G. M., and Parker, B. C.:
Cryoconite holes on glaciers, BioScience, 35, 499–503, 1985.
Wientjes, I. G. M. and Oerlemans, J.: An explanation for the dark
region in the western melt zone of the Greenland ice sheet, The
Cryosphere, 4, 261–268, https://doi.org/10.5194/tc-4-261-2010,
2010.
Wientjes, I. G. M., Van de Wal, R. S. W., Reichart, G. J., Sluijs,
A., and Oerlemans, J.: Dust from the dark region in the west-
ern ablation zone of the Greenland ice sheet, The Cryosphere, 5,
589–601, https://doi.org/10.5194/tc-5-589-2011, 2011.
Williamson, C. J., Anesio, A. M., Cook, J., Tedstone, A.,
Poniecka, E., Holland, A., Fagan, D., Tranter, M., and Yal-
lop, M. L.: Ice algal bloom development on the surface of
the Greenland Ice Sheet, FEMS Microbiology Ecology, fiy025,
https://doi.org/10.1093/femsec/fiy02, 2018.
Williamson, C. J., Cameron, K. A., Cook, J. M., Zarsky,
J. D., Stibal, M., and Edwards, A.: Glacier Algae: A
Dark Past and a Darker Future, Front. Microbiol., 10, 524,
https://doi.org/10.3389/fmicb.2019.00524, 2019.
Yallop, M. L., Anesio, A. J., Perkins, R. G., Cook, J., Telling, J., Fa-
gan, D., MacFarlane, J., Stibal, M., Barker, G., Bellas, C., Hod-
son, A., Tranter, M., Wadham, J., and Roberts, N. W.: Photophys-
iology and albedo-changing potential of the ice-algal community
on the surface of the Greenland ice sheet, ISME J., 6, 2302–2313,
2012.
The Cryosphere, 14, 309–330, 2020 www.the-cryosphere.net/14/309/2020/