Climate control of sea-ice edge phytoplankton blooms in the Hudson Bay system

Abstract

The Hudson Bay System (HBS), the world’s largest inland sea, has experienced disproportionate atmospheric warming and sea-ice decline relative to the whole Arctic Ocean during the last few decades. The establishment of almost continuous positive atmospheric air temperature anomalies since the late 1990s impacted its primary productivity and, consequently, the marine ecosystem. Here, four decades of archived satellite ocean color were analyzed together with sea-ice and climatic conditions to better understand the response of the HBS to climate forcing concerning phytoplankton dynamics. Using satellite-derived chlorophyll-a concentration [Chla], we examined the spatiotemporal variability of phytoplankton concentration with a focus on its phenology throughout the marginal ice zone. In recent years, phytoplankton phenology was dominated by two peaks of [Chla] during the ice-free period. The first peak occurs during the spring-to-summer transition and the second one happens in the fall, contrasting with the single bloom observed earlier (1978–1983). The ice-edge bloom, that is, the peak in [Chla] immediately found after the sea-ice retreat, showed substantial spatial and interannual variability. During the spring-to-summer transition, early sea-ice retreat resulted in ice-edge bloom intensification. In the northwest polynya, a marine wildlife hot spot, the correlation between climate indices, that is, the North Atlantic Oscillation and Arctic Oscillation (NAO/AO), and [Chla] indicated that the bloom responds to large-scale atmospheric circulation patterns in the North Hemisphere. The intensification of westerly winds caused by the strong polar vortex during positive NAO/AO phases favors the formation of the polynya, where ice production and export, brine rejection, and nutrient replenishment are more efficient. As a result, the winter climate preconditions the upper layer of the HBS for the subsequent development of ice-edge blooms. In the context of a decline in the NAO/AO strength related to Arctic warming, primary productivity is likely to decrease in the HBS and the northwest polynya in particular.

Full text

RESEARCH ARTICLE
Climate control of sea-ice edge phytoplankton
blooms in the Hudson Bay system
Lucas Barbedo
1,
*, Simon Be
´langer
1
, and Jean-E
´ric Tremblay
2,3
The Hudson Bay System (HBS), the world’s largest inland sea, has experienced disproportionate atmospheric
warming and sea-ice decline relative to the whole Arctic Ocean during the last few decades. The
establishment of almost continuous positive atmospheric air temperature anomalies since the late 1990s
impacted its primary productivity and, consequently, the marine ecosystem. Here, four decades of archived
satellite ocean color were analyzed together with sea-ice and climatic conditions to better understand the
response of the HBS to climate forcing concerning phytoplankton dynamics. Using satellite-derived
chlorophyll-a concentration [
Chla
], we examined the spatiotemporal variability of phytoplankton
concentration with a focus on its phenology throughout the marginal ice zone. In recent years,
phytoplankton phenology was dominated by two peaks of [
Chla
] during the ice-free period. The first peak
occurs during the spring-to-summer transition and the second one happens in the fall, contrasting with the
single bloom observed earlier (1978–1983). The ice-edge bloom, that is, the peak in [
Chla
] immediately found
after the sea-ice retreat, showed substantial spatial and interannual variability. During the spring-to-summer
transition, early sea-ice retreat resulted in ice-edge bloom intensification. In the northwest polynya, a marine
wildlife hot spot, the correlation between climate indices, that is, the North Atlantic Oscillation and Arctic
Oscillation (NAO/AO), and [
Chla
] indicated that the bloom responds to large-scale atmospheric circulation
patterns in the North Hemisphere. The intensification of westerly winds caused by the strong polar vortex
during positive NAO/AO phases favors the formation of the polynya, where ice production and export, brine
rejection, and nutrient replenishment are more efficient. As a result, the winter climate preconditions the
upper layer of the HBS for the subsequent development of ice-edge blooms. In the context of a decline in the
NAO/AO strength related to Arctic warming, primary productivity is likely to decrease in the HBS and the
northwest polynya in particular.
Keywords: Marginal ice zone,Ocean Color Radiometry,Climate indices,Phytoplankton phenology,Polynya
1. Introduction
The Arctic Ocean and its surrounding seas are facing the
most pronounced climatic changes on the Earth. Several
regional positive feedback processes amplify the warming
of the Arctic (Overland et al., 2004; Serreze et al., 2009),
such as the sea-ice albedo (Screen and Simmonds, 2010),
decline of sea ice in all seasons (Stroeve and Notz, 2018),
and the lapse rate and the thermal radiative balance
(Pithan and Mauritsen, 2014). Those changes also increase
the frequency of extreme climatic events in the northern
latitudes (Pithan and Mauritsen, 2014). As a consequence,
the timing of Arctic-wide sea-ice melt season has advanced
at a rate of 5 days per decade since 1979 (Stroeve et al.,
2011, 2014). Arctic sea-ice volume has declined at a rate
of –513 km
3
y
–1
and –287 km
3
y
–1
during winter and fall
season, respectively, between 2002 and 2018 (Kwok,
2018). Climatic changes have been even more accentuated
in the Hudson Bay System (HBS), a complex ecosystem
embracing Hudson Bay, James Bay, Foxe Basin, and Hud-
son Strait, which forms the world’s most extensive inland
sea (1.24 10
6
km
2
). For example, Stroeve and Notz
(2018) showed that Hudson Bay sea-ice cover in Septem-
ber has decreased at the rate of –1,046 km
2
y
–1
between
1979 and 2018, a loss of –93.6%relative to the average of
45.2%for the whole Arctic. As reported by Hochheim et
al. (2011), sea-ice concentration (SIC) losses range from –
15.1%to –20.4%per decade since 1980 in the northwest
and southwest sectors of Hudson Bay (HB), respectively. As
a consequence, the duration of the open water season
increased by 12 days per decade between 1980 and
2005, which is almost twice the rate observed in the Arctic
Ocean (i.e., 6.4 days per decade; (Markus et al., 2009).
1
De
´partement de Biologie, Chimie et Ge
´ographie, Groupes
Que
´bec-Oce
´an et BORE
´AS Universite
´du Que
´bec a
`Rimouski,
Rimouski, Que
´bec city, Que
´bec, Canada
2
Takuvik Joint International Laboratory, Laval University,
Que
´bec city, Que
´bec, Canada
3
De
´partement de Biologie et Que
´bec-Oce
´an, Universite
´Laval,
Que
´bec city, Que
´bec, Canada
* Corresponding author:
Email: lucasbarbedo@gmail.com
Barbedo, L, et al. 2020. Climate control of sea-ice edge
phytoplankton blooms in the Hudson Bay system.
Elem Sci Anth
,8:1.
DOI: https://doi.org/10.1525/elementa.039

Finally, the mean annual sea surface temperature trend is
about 3.7 C since the postindustrial era, which is much
larger than the trend observed over the Arctic Ocean
(approximately 2 C; (Brand et al., 2014).
The ongoing changes in the sea-ice cover extent and
thickness have impacted the primary producers of the
arctic and subarctic marine ecosystems (Kahru et al.,
2016). Changes in the cryosphere have resulted in an
imbalance in primary production between ice algae,
under-ice, and open water phytoplankton in the Arctic
Ocean and adjacent polar seas (Arrigo et al., 2012; Leu
et al., 2015). This imbalance has been driven by decreasing
sea-ice coverage and time shifts in seasonal sea-ice dynam-
ics (i.e., ice and snow thickness, melt onset, pond onset,
etc.; Arrigo et al. 2014; Horvat et al. 2017). Recent field
observations (Arrigo et al., 2012) and model simulations
(Horvat et al., 2017) suggest that the warming trends at
high latitudes may have increased the occurrence of
under-ice blooms as a consequence of the rise in under-
ice solar radiation during the spring season due to thinner
first-year ice, early melting, a significant fraction of leads,
and melt ponds at the ice surface (Mundy et al., 2009,
2014; Palmer et al., 2014; Leu et al., 2015; Assmy et al.,
2017; Horvat et al., 2017).
High phytoplankton biomass and primary production
rates in the marginal ice zone (MIZ), as defined hereafter
as the area along the edge of the ice pack that is affected
by open ocean processes, have been reported in the liter-
ature over several decades (Barber et al., 2015, and refer-
ences therein). Physical, chemical, and biological processes
occurring at the MIZ trigger important ecological succes-
sions in the marine ecosystem and impact the coupling
between sympagic, pelagic, and benthic realms (Leu et al.,
2015). Satellite ocean color observations have proven
a handy tool to map chlorophyll-a concentration [Chla],
a proxy for sea surface pelagic phytoplankton biomass and
abundance, in open waters found in the edge zone of the
MIZ (see Barber et al., 2015) in both poles since the launch
of the Coastal Zone Color Scanner (CZCS) (Maynard and
Clark, 1986, 1987; Mitchell et al., 1991). Recently, Perrette
et al. (2011) developed a method to assess the phytoplank-
ton bloom along the edge of the ice pack, which is
referred to as ice-edge bloom hereinafter, at the pan-
Arctic scale using the Sea Wide Field-of-view Sensor (Sea-
WiFS). Based on the year 2007, they concluded that the
ice-edge blooms were ubiquitous in the northern latitudes
(>66.6 N). Also using SeaWiFS, Lowry et al. (2014) re-
ported that ice-edge blooms on the Chukchi Sea shelf
depend on the timing of sea-ice retreat and further spec-
ulated that massive under-ice blooms were widespread in
nutrient-rich waters of Pacific origin. Renaut et al. (2018)
applied the method of Perrette et al. (2011) to 11 years of
chlorophyll-a observation of the Moderate-Resolution
Imaging Spectrometer Sensor (MODIS) and reported an
intensification and a northwardexpansionoftheice-
edge bloom in many subarctic seas. Altogether, these re-
sults suggest that nonlinear processes regulate ice-edge
blooms. Among them, sea-ice production and snow cover
thickness affect convective mixing and nutrient replenish-
ment during winter, light availability at the onset of the
growing season, and the amount of meltwater available
for the subsequent emergence of meltwater stratification.
Are ice-edge blooms a recurrent feature throughout
the HBS? To our knowledge, ice-edge blooms in the HBS
have not been examined explicitly, using satellite ocean
color or with field observations. The HBS was excluded
from recent ocean color analyses conducted in the Arctic
(Perrette et al., 2011; Renaut et al., 2018). Except in river
plumes and some upwelling spots in Hudson Strait and
near the Foxe Peninsula and the Belcher Islands, the HBS
has been assumed to be a low-productivity oligotrophic
system (Ferland et al., 2011; Tremblay et al., 2019). A few
studies investigated under-ice phytoplankton and pelagic
community interactions during the spring season in a few
coastal locations (i.e., near the Belcher Islands and Great
Whale River in the southeastern coastal HB; Michel et al.,
1988, 1993; Runge et al., 1991; Monti et al., 1996). These
land-fast ice camp-based studies reported many critical
ecological processes occurring during the transition from
the under-ice to a pelagic-dominated system: photo-
adaptation of ice-algal communities (Michel et al., 1988);
ice-algal communities seeding the pelagic ecosystem (Mi-
chel et al., 1993); the coupling between under-ice grazers,
interfacial, and pelagic communities (Runge et al., 1991);
and the impact of river plume dynamics on algal commu-
nity composition (Monti et al., 1996). None of these stud-
ies found evidence of under-ice phytoplankton blooms
during the melting season at the Hudson Bay basin scale.
Model-based investigations by Sibert et al. (2010, 2011),
however, reported that under-ice production timing is
mainly controlled by surface melting (snow and ice) pro-
cesses that determine the light levels, even under condi-
tions of sufficient nutrient availability. A bay-wide
assessment of [Chla] is needed to obtain a large-scale per-
spective of the phytoplankton productivity of the HBS.
In summary, there are many indications that the MIZ is
a biologically productive feature in the HBS. Still, no sys-
tematic observations of ice-edge phytoplankton blooms
have been reported in this subarctic region. The main
objective of this study was to assess the temporal and
spatial variability of the ice-edge bloom based on a system-
atic analysis of available satellite time series. Therefore, we
analyzed two decades of satellite remote sensing data
combining ocean color observation of [Chla] and passive
microwave imagery for SIC. First, we compared the phy-
toplankton phenology assessed from archived observa-
tions of the CZCS (1978–1983) to the modern
phenology obtained from MODIS-Aqua (2002–2014).
Next, we characterized phytoplankton dynamics based
on temporal evolution of satellite [Chla] and timing of
sea-ice retreat to obtain climatological conditions, trophic
categories, and phytoplankton phenological types in mar-
ginal ice zones. We documented the interannual variability
of the surface [Chla] at the MIZ and examined how it is
impacted by the timing of sea-ice retreat and winter air
temperature. Finally, we examined the influence of large-
scale climate variability patterns, particularly the Arctic
Oscillation (AO) and North Atlantic Oscillation (NAO), on
oceanographic processes and ice-edge bloom intensity in
the HBS.
Art. 8(1), page 2 of 25 Barbedo et al: Ice-edge Blooms in the Hudson Bay System

2. Materials and methods
2.1. Satellite data
2.1.1. Ocean color
Multi-sensor merged [Chla] Level-3 (i.e., binned and
mapped) 8-day composites from the Globcolour Project
(http://www.globcolour.info/) were used as a proxy for
phytoplankton biomass. Globcolour products have a spa-
tial resolution of 4.63 km and cover the 1998–2018
period. The merged product was selected to improve the
spatial-temporal coverage diminishing gaps due to cloud
cover and sea-ice coverage (Maritorena et al., 2010). The
binning methodology combines the normalized water-
leaving radiances from different ocean color sensors
whenever they are available, which includes SeaWiFS
(1998–2010), MODIS-Aqua (2002–2018), Medium-
Resolution Imaging Spectrometer (MERIS: 2002–2011),
and Visible Infrared Imaging Radiometer Suite (VIIRS:
2012–2018). [Chla] was estimated from normalized
water-leaving radiances merged using the Garver-Siegel-
Maritorena (GSM) semi-analytical model (Garver and Sie-
gel, 1997; Maritorena et al., 2002). GSM also yields particle
backscattering (bbp) and colored detrital matter (CDM)
coefficients at 443 nm.
Colored dissolved organic matter (CDOM) and non-
algal particles can deteriorate the performance of ocean
color algorithms because, as phytoplankton pigments,
they absorb the blue part of the visible spectrum (Bricaud
et al., 1981). Consequently, in Arctic and sub-Arctic seas,
[Chla] overestimates (underestimates) in the lower (high-
er) range of [Chla] have often been reported (Cota et al.,
2004; Hirawake et al., 2012; International Ocean Colour
Coordinating Group [IOCCG], 2015). CDOM is known to be
a dominant optical component in most of the HBS, mak-
ing this region optically complex (Granskog et al., 2007;
Mundy et al., 2010; Gue
´guen et al., 2011; Xi et al., 2013;
Burt et al., 2016; Heikkila et al., 2016). As discussed by Ben
Mustapha et al. (2012), Be
´langer et al. (2013), and Lewis
and Arrigo (2020), the GSM algorithm can minimize the
impact of other optical constituents on Chla retrievals
because the more mechanistic GSM algorithm is able to
take into account seasonal and regional variability of bio-
optical properties compared to empirical algorithms for
Polar shelves. Therefore, the GSM was selected because it
can better represent the optically complex waters of the
HBS (Xi et al., 2013, 2014, 2015), and data from depths
shallower than 50 m were excluded from the analysis to
avoid turbid or CDM-rich waters and river plumes.
We performed a cluster analysis (see below) on 8-day
composites (multiyear averages) of Chla to detect poten-
tial changes in phytoplankton phenology between the
1980s and the 2000s. We used the CZCS for the period
of 1978–1983 and MODIS for 2002–2014. For consistency
in the data processing for these two ocean color missions,
Chla was calculated using the case-1 waters standard
empirical Ocean Color algorithm CZCS OC3 and MODIS
OC3v5 (O’Reilly et al., 1998, 2000). The use of band ratio
algorithms allows a better interconnection between dis-
tinct sensors (Antoine et al., 2005; McClain, 2009). We
tested the sensitivity of the cluster analysis to the choice
of Chla algorithms and available ocean color products.
Specifically, we used the Globcolour merged chlorophyll-
a products obtained using both GSM (as above) and empir-
ical algorithms for the same period as MODIS OC3v5 alone
(2002–2014). We found some spatial discrepancy in the
cluster distribution between products, but the main con-
clusion drawn from the analysis remained unchanged (i.e.,
increased occurrence of double bloom, see Results section).
2.1.2. SIC
SIC was obtained from the National Snow and Ice Data
Center. It is based on daily passive microwave radiometry
processed using the Bootstrap algorithm (Comiso, 2000)
at 25 km resolution. The Bootstrap technique clusters the
multichannel passive microwave sensors: Scanning Multi-
channel Microwave Radiometer on the Nimbus-7 satellite,
Special Sensor Microwave/Imager and Special Sensor
Microwave Imager/Sounder from the Defense Meteoro-
logical Satellite Program’s satellites, and the Advanced
Microwave Scanning Radiometer (Comiso et al., 1997). SIC
was interpolated onto the same Chla grid using the near-
est neighborhood scheme implemented in Matlab.
2.2. Climate index and reanalysis
The Arctic Oscillation is a climate pattern characterized by
winds circulating counterclockwise around the Arctic at
around 55 N latitude. The AO index is calculated from
the first component of the empirical orthogonal function
of monthly anomaly variations in sea-level atmospheric
pressurenorthof40N, which explains 22%of total
variance (Thompson and Wallace, 1998). The North Atlan-
tic Oscillation index is extracted from the difference
between monthly anomaly variations in sea-level atmo-
spheric pressure over Greenland (low pressure) and The
Azores Island (high pressure) (Hurrell et al., 2001, 2003;
Qian et al., 2008). NAO/AO data were obtained from the
Climatic Prediction Center/the National Oceanic and
Atmospheric Administration (http://www.cpc.ncep.noaa.
gov). Here, we used the average index for the winter
months of January, February, and March.
We obtained the monthly air temperatures from the
National Centers for Environmental Prediction/National
Center for Atmospheric Research (NCEP/NCAR) Reanalysis
Project. Anomalies of air temperature were calculated as
the difference between each monthly average and its cor-
responding monthly climatology and normalized by the
standard deviation, both of which were calculated using
data from the 1948 to 2018 period.
2.3. Data analysis
2.3.1. Phytoplankton phenology during the ice-free season
We investigated the potential changes in phytoplankton
phenology between the 1980s and 2000s using a K-means
cluster analysis applied on the annual time series of Chla.
The approach used here is similar to that adopted by
D’Ortenzio and D’Alcala
`(2009) in the Mediterranean Sea,
D’Ortenzio et al. (2012) at the global scale, Ardyna et al.
(2017) in the Southern Ocean, and Marchese et al. (2017)
at regional Arctic scales.
Briefly, we analyzed the seasonal variability (or shape)
of the annual Chla time series (i.e., a multiyear average of
Barbedo et al: Ice-edge Blooms in the Hudson Bay System Art. 8(1), page 3 of 25

8-day Chla available for a given period, hereafter referred
to as climatology) to determine the number of phytoplank-
ton blooms during the annual cycle, as well as their tim-
ing. Therefore, we focused on the relative change in Chla
over the year rather than the absolute values of Chla. In
addition, we standardized the individual climatologies,
Chlclimðx;y;tÞ, for each 4.60-km pixel (x;y), as:
ChlNðx;y;tÞ¼Chlclimðx;y;tÞChlðx;yÞ
sChlðx;yÞð1Þ
where tisthetimereferringtothe8-daycomposite;
ChlNðx;y;tÞis the normalized time series; and Chlðx;yÞ
and sChlðx;yÞare the annual mean and standard deviation
values of climatological time series for the pixel (x;y).
Chlclimðx;y;tÞwas calculated with a temporal resolution
of 8 days for CZCS and MODIS-Aqua. Data applied to
phenology analysis were also filtered by a spatial-
temporal median filter (3 33), and the gaps were
filled using the data interpolating empirical orthogonal
functions scheme implemented in R (DINEOF; Beckers
and Rixen, 2003). The ideal number of clusters was deter-
mined by Davies-Bouldin criterion, and cluster stability
wasaccessedbyJaccard,Bootstrap,andNoisedcoeffi-
cients (Hennig, 2007; D’Ortenzio and D’Alcala
`, 2009;
D’Ortenzio et al., 2012).
2.3.2. Phytoplankton phenology at the marginal ice zone
To assess the impacts of sea-ice retreat timing on MIZ
phytoplankton blooms (also refers to phytoplankton
spring blooms or ice-edge blooms), we analyzed both Chla
and SIC variability in parallel. The method is similar to that
of Perrette et al. (2011), which was also adopted by Lowry
et al. (2014) and Renaut et al. (2018). The sea-ice retreat,
tR, is defined as the day at which SIC is below 10%for at
least 24 days. This time interval is longer than the 20 days
applied by Perrette et al. (2011) and Renaut et al. (2018)
and the 14 days by Lowry et al. (2014) because we used 8-
day composites instead of daily maps. However, to avoid
sub-pixel contamination in ice-infested regions near the
ice edge (Be
´langer et al., 2013), we opted to be more
conservative by applying a 10%threshold on SIC, as did
Perrette et al. (2011) and Renaut et al. (2018) instead of
50%as applied by Lowry et al. (2014).The maximum Chla
observed in the MIZ was extracted for each pixel for each
year, yielding one map of MIZ Chla per year.
We recognized that sea-ice cover hides an important
component of phytoplankton phenology that is out of the
reach of ocean color satellites (Arrigo et al., 2012). To
address this problem, we developed a trophic predictor
of under-ice and pelagic phytoplankton phenology
throughout the MIZ by analyzing a seasonal time series
of satellite-derived SIC and Chla. This trophic predictor is
defined by the temporal evolution of Chla since the sea-
ice retreat (tR).
Figure 1 illustrates five typical Chla phenologies ex-
pected during the spring-to-summer transition at the MIZ,
assuming that the temporal evolution of Chla during phy-
toplankton blooms has a Gaussian-like curve (Jo
¨nsson and
Eklundh, 2004; Platt et al., 2009). Each of these phenolo-
gies is defined in terms of three metrics: (1) the very first
Chla observation after the sea-ice retreat (½ChlaðtRÞ), (2)
the maximum Chla reached within five weeks after the
sea-ice retreat (½Chlaðt1;2;3;:::Þmax ), and (3) the minimum
reach within five weeks after the sea-ice retreat
(½Chlaðt1;2;3;:::Þmin ). Table 1 presents the criteria that
define each type of MIZ phenology. The bloom threshold,
B, was set at 0.5 mg m
–3
as in Perrette et al. (2011). The
oligotrophy threshold, Ol,wassetat0.2mgm
–3
.We
applied these criteria to each pixel and for each year,
yielding one thematic map per year.
2.3.3 Correlation analysis between MIZ Chla and environ-
mental forcing
To investigate the impact of climate and environmental
forcing on ice-edge blooms, we calculated the Pearson
Figure 1. Schematic of phytoplankton phenologies throughout marginal ice zones. The five possible phenologies are
consistently low biomass indicating undetectable (old) under-ice bloom or an oligotrophic set up (I: blue line or
dotted), probable (recent) under-ice bloom (II: red line), consistently high biomass indicating efficient nutrient
replenishment (III: green line), bloom triggered in ice-free waters after the sea-ice retreat (VI: orange line), and
bloom triggered under-ice that develop a peak just after the sea-ice retreat (V: cyan line). DOI: https://doi.org/
10.1525/elementa.039.f1
Art. 8(1), page 4 of 25 Barbedo et al: Ice-edge Blooms in the Hudson Bay System

correlation coefficient (r) of the linear relationships
between the MIZ Chla and (1) the annual values of the
AO, (2) the sea-ice retreat day (tR), and (3) the winter
(December–January–February, DJF) air temperature
anomaly. The analysis was made on each grid cell of the
HBS and between 1998 and 2018. We standardized the
maximum Chla values in the MIZ (Perrette et al., 2011)
using climatology mean and standard deviation as above.
Only significant slopes, obtained from a least-squares fit at
an interval of confidence of 95%(P> 0.05) and for pixels
having more than 10 valid years between 1998 and 2018,
were mapped.
3. Results
3.1. Air temperature and phytoplankton phenology:
CZCS versus MODIS era
Warmer air temperatures in the HBS began around 1998
(Figure 2), which coincides with the beginning of the
modern era of ocean color radiometry (i.e., the launch
of SeaWiFS in August 1997). This recent period con-
trasted with the CZCS era (1978–1983), which was char-
acterized by the predominance of a negative air
temperature anomaly (Tair) relative to the average air
temperature calculated over the 1948–2018 period. One
extreme cold event was observed at the beginning of the
CZCS era in 1978. During the MODIS observational era,
Tair was mainly positive, but colder winters were often
seen, as in 2002, 2003, 2004, 2011, 2014, 2015, 2016,
and 2018. In contrast, most summers showed positive
anomalies, except 2004, 2007, 2011, and 2018. The latter
stands out in the modern era with a noticeable negative
Tair during most of the year.
The distinct Tair between the CZCS and MODIS eras
coincided with remarkable changes in phytoplankton phe-
nological patterns in the HBS (Figure 3). Our analysis
resulted in three statistically distinct bioregions character-
ized by the shape and the number of observable Chla
peaks, hereafter referred to as phytoplankton “blooms,”
over an annual cycle. The spring–summer bloom (SB) type
(red pixels and curve in Figure 3) exhibited only one Chla
peak, which reached its maximum a few weeks after the
summer solstice (June 21). The double bloom (DB) type
(green pixels) was characterized by a first Chla peak occur-
ring in late May/early June, and a second peak at the very
end of the open water season, when ocean color observa-
tions were still available (i.e., mid-October). The last type,
the fall bloom (FB) (blue pixels), was initiated in mid-
summer, peaked in mid-September, and declined until the
sea-ice recovered (Figure 3C). FB contrasted with the sec-
ond peak of the DB phenology, in which Chla increased
continually until the sea-ice recovery.
Three types of bloom were observed during the CZCS
era, but with no distinct spatial patterns (Figure 3A). Of
65,873 valid observations over that period, the SB, DB,
and FB occupied 28%,38%, and 24%, respectively. During
the MODIS era, for which 70,972 valid pixels were found,
SB remained at about 28%, but DB increased to 54%, and
FB decreased to 18%of the total surface area. This result
suggests that a remotely sensed spring or summer bloom
is a common feature in the HBS. Whether this increase in
Chla at the beginning of the open water season is an
actual ice-edge bloom or not is examined further in the
next sections.
Single Chla peak phenology (SB and FB) might be seen
as a variation of DB phenology. The start date and the
peakamplitudeofsingleSBoccurredshortlyafterthe
spring bloom of DB phenology. Similarly, the single FB
peak occurred before the fall bloom of the DB phenology.
The fact that the DB phenology has become a predomi-
nant feature in the HBS implies that an earlier spring–
summer bloom is now occurring. We examined whether
or not the duration of the open-water-season (OW),
which was defined as the numberofdayswithSICbelow
10%during an annual cycle, could explain the fact that
more DB are now observed. The OW duration increased
from 15 to 39 days between CZCS and MODIS eras, reach-
ingabout125days(4months)inmoderntimes
(Figure 3D and Table 2).TheregionswhereSBinthe
CZCS era was replaced by DB in the MODIS era had an
increase in OW of 34 days (Table 2). The OW duration in
areaswhereFBwasreplacedbyDBorremainedasDB
had increased by 21 and 23 days, respectively. However,
these increases in OW were not more significant than for
pixels showing SB in both CZCS and MODIS eras (39
days). Therefore, no clear link was identified between the
OW duration and the change (or the lack of change) in
the phytoplankton phenology patterns.
Table 1. Remote sensing criteria defining the types of
phytoplankton phenology throughout marginal ice zones.
DOI: https://doi.org/10.1525/elementa.039.t1
Phenology Type
a
Criteria
b
I—Oligotrophic or old
under-ice bloom
½ChlaðtRÞ <B
½ChlaðtRÞ <½Chlaðt1;2;3:::Þmax <B
II—Probable (recent)
under-ice bloom
Ol <½ChlaðtRÞ <B
½Chlaðt1;2;3:::Þmax <½ChlaðtRÞ
III—Mesotrophic/nutrients
replenished
½ChlaðtRÞ  B
½Chlaðt1;2;3:::Þmax B
½Chlaðt1;2;3:::Þmin B
VI—Bloom triggered in ice-
free waters
½ChlaðtRÞ <½Chlaðt1;2;3:::Þmax 
½ChlaðtRÞ <B
½Chlaðt1;2;3:::Þmax B
V—Bloom triggered under-
ice
½ChlaðtRÞ <½Chlaðt1;2;3:::Þmin 
½ChlaðtRÞ  B
½Chlaðt1;2;3:::Þmin <B
a
Defined based on the time evolution of Chla after the sea-ice
retreat (with Chla in mg m
–3
).
b
Bis the bloom threshold (B), here set to 0.5 mg m
–3
;the
oligotrophic threshold (Ol) was set to 0.2 mg m
–3
;tRis the day
of sea-ice retreat defined as the day at which sea-ice concentra-
tion (SIC) is below 10%for at least 24 days; t1;2;3::: refers to weeks
after the sea-ice retreat.
Barbedo et al: Ice-edge Blooms in the Hudson Bay System Art. 8(1), page 5 of 25

Figure 2. Surface air temperature anomalies and ocean color satellite coverage. Monthly anomalies of air temperature
(Tair) spanning the period from 1948 to 2018 (negative anomalies in blue and positive in red bars). The temporal
coverage of CZCS (1978–1983), MODIS (2002–2014), and Globcolour (1998–2018) Chla observations are marked.
During the modern era of ocean color radiometry, positive values for Tair were often observed during the spring-to-
summer transition. DOI: https://doi.org/10.1525/elementa.039.f2
Figure 3. Phytoplankton phenology changes between CZCS (1978–1983) and MODIS eras (2002–2014). Cluster-derived
maps of phytoplankton phenological regimes for the (A) CZCS era (1978–1983) and the (B) MODIS era (2002–2014).
(C) The plot shows the three types of phenology obtained using the K-means cluster analysis. (D) Open water season
duration (number of days with sea-ice concentration lower than 10%) for each phenological domain in the CZCS era
(blue boxplots) and MODIS era (yellow boxplots). The boxplots show the median (red lines), 25th and 75th percentiles
at the hinges, and the whiskers extend to show +1standarddeviation.DOI:https://doi.org/10.1525/
elementa.039.f3
Art. 8(1), page 6 of 25 Barbedo et al: Ice-edge Blooms in the Hudson Bay System

3.2. Marginal ice zone blooms between 1998 and
2018
Figure 4A shows the climatology of the 8-day bins of Chla
for the whole HBS from the Globcolour data set (1998–
2018), together with the average daily percentage of OW
in the HBS (blue bars). Interestingly, we observed two
distinct peaks in Chla: one in spring around day-of-year
(DOY) 75 (mid-March) and a second around DOY 150
(beginning of June). The first peak, reaching values around
0.7 mg m
–3
, is observed when sea ice covers the bay
almost entirely (90%of the area). During that time, we
noted early phytoplankton blooms in very restricted loca-
tions known as polynyas, such as the northwestern part of
Hudson Bay (Ferguson et al., 2010). The second increase in
Chla begins as soon as the OW area starts to increase and
can be considered as an ice-edge bloom, which is a typical
feature along the marginal ice zone during the spring-to-
summer transition (Mitchell et al., 1991). During this tran-
sition period, the Chla reached a relatively high level (1.0
mg m
–3
) with lower variability, as depicted by the lower
standard deviation (gray-shaded area on Figure 4A) com-
pared to the rest of the year. This bloom is probably pro-
ductive, as it is synchronized almost perfectly with the
peak of incident photosynthetic available radiation
(PARð0þÞ, where 0þindicates PAR just above the sea
surface) (Figure 4B). The end of the OW season is charac-
terized by a continuous increment of Chla, referred to as
the fall bloom, which ended when sea ice recovered in late
October (DOY 300). The high standard deviation during
this season indicates extremes of low and high Chla,
pointing toward high spatial variability.
To illustrate the recurrence of ice-edge blooms in the
HBS, we plotted the frequency distribution of Chla (N
approximately 10
6
pixels) as a function of number of days
after the sea-ice breakup between 1998 and 2018 (Fig-
ure 5). Higher Chla values are generally restricted to the
first 8-day period after the sea-ice retreat, with Chla values
ranging from 1 to 3 mg m
–3
. In contrast, observations of
high chlorophyll-a later during the open water season were
relatively low. These results suggest that ice-edge blooms,
which are characterized by an increase in phytoplankton
abundance propagating along the ice edge, could be
observed frequently by ocean color satellites in the HBS.
However, spatial analysis is required to confirm whether
ice-edge blooms are present everywhere in the HBS or
whether they are restricted to some parts of the Bay, where
ocean conditions are more favorable, as suggested by the
phenology analysis based on the climatology (Figure 3).
Panels A and B of Figure 6 show maps of the average
and the standard deviation of the maximum Chla, respec-
tively, observed in the ice-edge zone of the MIZ between
1998 and 2018. The statistics were calculated using the 21
annual maps depicted in Figure S1. The limits of the sub-
regions are those used by Landy et al. (2017) who reported
satellite-based sea-ice thickness and snow depth in the
HBS. In general, we noted higher Chla in the surrounding
of the Bay, with the highest values found in the Hudson
Strait (HS) and Southeastern (SE) HB. These two regions
also show a higher standard deviation (STD) in the maxi-
mum Chla at the MIZ, indicating relatively high
Table 2. Changes in the number of days (Dt) of open
water season between the CZCS (1978–1983) and MODIS
(2002–2014) eras, according to phenological cluster. DOI:
https://doi.org/10.1525/elementa.039.t2
Cluster
a
MODIS SB MODIS FB MODIS DB
CZCS SB 39 24 34
CZCS FB 27 15 21
CZCS DB 24 18 23
a
SB indicates spring bloom; FB, fall bloom; and DB, double
blooms.
Figure 4. Climatologies of Chla, percentage of open water, and photosynthetic active radiation (PAR). (A) Chla
climatology from Globcolour (GSM; merged product) and open water fraction (blue area) as a function of day of
year (DOY). (B) PARð0þÞ climatology (Be
´langer et al., 2013; Laliberte
´et al., 2016) (black lines) their respective
standard deviation (shaded) observed between 2003 and 2010 in HBS. Sea-ice retreat was typically followed by
a pelagic bloom during the spring-to-summer transition with a broad peak in early June (DOY 152) that was
preceded by a smaller peak in February (DOY 120) and followed by a peak in the fall (DOY 243). DOI: https://
doi.org/10.1525/elementa.039.f4
Barbedo et al: Ice-edge Blooms in the Hudson Bay System Art. 8(1), page 7 of 25

interannual variability. We also observed a high STD in the
northwestern (NW) polynya. Southwestern (SW) and
Northeastern (NW) HB have moderate to high Chla
(around 1 mg m
–3
) in the MIZ and relatively low interan-
nual variability. The central HB shows both low Chla (<0.5
mg m
–3
) and STD interannual variability, suggesting that
ice-edge blooms are not a dominant feature offshore. Fig-
ure 6C and D illustrates the average and standard devi-
ation of sea-ice retreat date (tR)forthesameperiod
(1998–2018) based on annual maps shown in Figure S2.
The central HB had the latest tR, ranging from DOY 180 to
210 (month of July), relatively low interannual variability
(STD < 10). The highest (lowest) interannual variability of
tRoccurred in the coastal (offshore) areas of SW and NE
Hudson Bay. Interestingly, the tRand Chla were somewhat
very similar, suggesting that the magnitude of the ice-
edge bloom is linked to the timing of sea-ice retreat, as
explored further below.
Table 3 presents the statistics and frequency of occur-
rence of the surface extension (in km
2
and %of the total
surface area) for four subcategories of Chla, which is
a good proxy for the trophic state of the MIZ. On average,
an ice-edge bloom (0.5 to 2.0 mg m
–3
) or major ice-edge
bloom (here major bloom is defined as Chla > 2 mg m
–3
)
occupied 46.3%of the Bay, but blooms can range from
27.9%(1999) and 56.2%(2008) of the total area (see also
Figure S1). High interannual variability was also observed
in the area of major blooms ranging from 2.3%(2010) to
9.0%(2015) (Figure 7). During the study period (1998–
2018), the surface area of the ice-edge blooms increased
by 10.5%(i.e., 0.5%per year times 21 years). In 2015,
a major ice-edge bloom occupied as much as 7,200 km
2
(or 9.0%of the HBS surface) but was found in the poly-
nyas (Figure 7). In 2018, the timing of the sea-ice breakup
showed large spatial variability with an early breakup in
the NW and a late breakup in the NE (Figure 7). The ice-
edge bloom occupied a large portion of the NW, but with
modest Chla compared to 2015. This result suggests that
the magnitude of the ice-edge bloom is explained not only
by the timing of the sea-ice retreat.
Figure 7 shows annual maps of the maximum Chla
observed during the MIZ time window for selected years
(2008, 2011, 2015, and 2018), showing interesting fea-
tures in terms of ice-edge blooms. Several key observa-
tions can be made from these maps. First, the Chla in
the central HB remained relatively low (<0.5) after the
breakup for most of the years, except in 2005, 2008,
2012 and 2016, which means that no ice-edge blooms
occurred in the central HB for most of the years analyzed.
Moderate to high values of Chla were observed during the
MIZ period along the surrounding of the bay with obvious
variability in both northwestern and northeastern parts of
the bay, which sometimes showed opposite patterns (e.g.,
compare 2015 and 2016; Figure S1). Higher Chla in the
MIZ was observed systematically in the Hudson Strait
where mixing and nutrient replenishment are more
intense relative to the rest of HB (Ferland et al., 2011).
However, the largest ice-edge bloom was observed in the
NW HB in 2015 with Chla higher than 2.0 mg m
–3
.
To gain insight into the variability in the MIZ Chla
(Figure 7), we considered the timing of the sea-ice retreat.
Figure S2 depicts the DOY of the sea-ice breakup or, in
other words, the beginning of the MIZ period in the HBS
for each year between 1998 and 2018. The breakup
occurred within a wide range of 4 months, that is, from
DOY 120 to 240. An early breakup can be found either in
the NW or NE HB, where polynyas are common. Sea ice
often accumulated in the southern part of HB delaying the
breakup to the end of the summer, as observed in 2000,
2004, 2008, 2009, and 2015. A visual comparison of the
annual maps presented in Figures 7, S1, and S2 suggests
that, at least in some parts of Hudson Bay, early sea-ice
breakup resulted in more intense ice-edge blooms observ-
able from space. This relationship, however, is far from
applicable to the whole HBS, which we examine further
in the next section.
Figure 8 shows maps of the frequency of occurrence of
the four trophic states in the ice-edge zone defined above
(Table 3). These maps provide an additional point of view
about the spatial and interannual variability of ice-edge
blooms in the HBS between 1998 and 2018. Oligotrophy
occurrence (<0.2 mg m
–3
) was more frequent offshore,
covering a vast region in the central part of HB in 30%
of the years analyzed. On the other hand, the occurrence
of oligotrophy conditions across the bay was less than 5%
(Figure 8A). As observed in Figure 8B, moderate oligo-
trophy (0.2 Chla < 0.5 mg m
–3
) conditions were more
frequent in the central HB with a frequency higher than
60%covering a large proportion of the bay (almost half of
it). Ice-edge bloom (0.5 < Chla 2.0 mg m
–3
) occurrence
in the surrounding of the bay and HS generally surpassed
65%. A presence higher than 80%was reached in the
domains of the NW polynya, proximal areas of the most
significant river plumes, the eastern bay, and HS
(Figure 8C). Recurrently productive hot spots (>50%of
Figure 5. Frequency distribution of Chla after the sea-ice
retreat. Frequency distribution of Chla as a function of
thenumberofdaysafterthesea-iceretreat.The
marginal ice zone (red box) is defined by the first 24
days after the sea-ice retreat (Perrette et al., 2011). The
subsequent days define the open water season.
Relations are extracted from annual time series of
Chla for all pixels and all years between 1998 and
2018 in the HBS. DOI: https://doi.org/10.1525/
elementa.039.f5
Art. 8(1), page 8 of 25 Barbedo et al: Ice-edge Blooms in the Hudson Bay System

occurrence; Figure 8D) were found in relatively small
regions, south of Belcher Island, some places in the north-
ern HS, and river-influenced coastal areas. We also
observed occurrence above 10%in the NW HB polynya
and south of Southampton Island.
3.3. Phytoplankton phenology at the MIZ
We performed a time-series analysis of SIC and Chla at the
pixel level to assess the potential that under-ice blooms
escaped satellite range. We considered four scenarios for
the temporal evolution of Chla at the MIZ, as shown in
Figure 1. Applying the criteria of Table 1 resulted in 21
thematic maps presented in Figure S3. To synthesize the
results, we calculated the frequency of occurrence of each
type based on 21 years of observations. Figure 9 shows
the frequency of occurrence of the five phenology types
between 1998 and 2018.
Consistently low biomass observed in the MIZ after
the ice retreat (Figure 9A) could be due to an under-ice
bloom that escaped the satellite range entirely or to an
oligotrophic state at the sea surface. Similarly, the prob-
able (recent) under-ice bloom (Figure 9B) may also
indicate that oligotrophic domains remain in the ice-
edge zone. However, these domains involve moderate
concentrations between 0.2 and 0.5 mg m
–3
after the
sea-ice retreat followed by a decline in Chla. Oligotro-
phy/old under-ice bloom (Type I) and probable (recent)
under-ice blooms (Type II) can be considered similar
features, though with distinct magnitudes and bloom-
timing. This scenario (Types I and II) dominated the
whole central HB, extending north of Southampton Is-
land with a frequency of occurrence ranging from 50%
to 100%. In contrast, the situation characterized by
Chla consistently above 0.5 mg m
–3
(Figure 9C), sug-
gesting mesotrophic conditions where surface waters
are efficiently replenished in nutrients, was frequent
only in coastal waters or at the ends of the Hudson
Strait. The last scenario, that is, when the bloom is
Figure 6. Climatic maps of maximum Chla in the ice-edge zone and of sea-ice retreat timing. Average (A) and standard
deviation (B) of the maximum Chla observed during the 24-day period after the sea-ice retreat. The statistics were
calculated using the annual maps between 1998 and 2018 (n¼21 for each grid cell) shown in Figure S1. Similar
statistics (C and D) were made for the DOY of sea-ice retreat (tR) shown in Figure S2. DOI: https://doi.org/10.1525/
elementa.039.f6
Barbedo et al: Ice-edge Blooms in the Hudson Bay System Art. 8(1), page 9 of 25

triggered in ice-free waters, was infrequent, except at
the margin between the coastal domains and the cen-
tral HB, from Churchill to NE HB (Figure 9D). Finally,
the typical ice-edge bloom scenario, that is, when Chla
peaks just after the ice breakup and then vanishes to
low concentration, was relatively frequent (30%–70%
of the time) all around HB and in the central part of
HS (Figure 9E).
Figure 7. Maximum Chla in ice-edge zone and ice-retreat timing during extreme years and the BaySys expedition.
Annual map of maximum Chla following the sea-ice retreat (left panels) and the timing (DOY) of sea-ice retreat (right
panels) for selected years (same as Figures S1 and S2). Top panels are for 2008 and 2000, which showed the largest
and smallest ice-edge bloom extension, respectively, as reported in Table 2. Bottom panels are for 2015, the year
showing the largest ice-edge bloom in the NW polynya, and 2018, the year of the BaySys expedition. DOI: https://
doi.org/10.1525/elementa.039.f7
Table 3. Annual extension of distinct trophic states in the ice-edge zone estimated using Chla thresholds and sea-ice
retreat between 1998 and 2018 in the HBS. DOI: https://doi.org/10.1525/elementa.039.t3
Ice-edge Zone Trophic State Extension in 10
4
km
2
(%
a
) and Extreme Years
b
Categories Chla (mg m
–3
) Median Max., year Min., year Trend
c,d
RMSE
e
Oligotrophy <0.2 3.7 (4.6%) 18.2 (22.7%), 1999 0.5 (0.6%), 2008 –0.4 (–0.4%) 4.3
Moderate
oligotrophy
0.2–0.5 28.8 (35.9%) 36.4 (45.4%), 2000 13.6 (17.0%), 2015 –0.2 (–0.3%) 4.9
Bloom 0.5–2.0 33.1 (41.3%) 45.0 (56.2%), 2008 22.3 (27.9%), 1999 þ0.4 (þ0.5%) 5.6
Major bloom >2.0 4.0 (5.0%) 7.2 (9.0%), 2015 1.8 (2.3%), 2010 0.0 (0.0%)1.5
Sea-ice retreat
f
183 194, 2004 174, 2006 þ0.1 5.7
a
%of total area of Hudson Bay and Hudson Strait deeper than 50 m: 80 10
4
km
2
(Amante and Eakins, 2009);
b
Extreme events are reported according to year of maximum and minimum coverage;
c
In an interval of confidence of 95%(P< 0.05) for the extension of trophic states;
d
In an interval of confidence of 99%(P< 0.01) for the day of sea-ice retreat;
e
Root mean square error;
f
First day of the year when sea-ice concentration is below 10%.
Art. 8(1), page 10 of 25 Barbedo et al: Ice-edge Blooms in the Hudson Bay System

3.4.The magnitude of the ice-edge bloom and timing
of sea-ice breakup
As shown above, a high spatiotemporal variability charac-
terized the dynamic of the MIZ, in terms of both Chla and
sea-ice breakup in HBS (Figures 7, S1, and S2). Here, we
examined the potential link between these two variables
by aggregating data in the seven subregions presented
above (Figure 10): northwestern (NW), southwestern
(SW), central and offshore (Central), narrows at the Hud-
son Bay entrance (Narrows), Hudson Strait (HS), northeast-
ern (NE), and southeastern (SE). The boxplots of Figure 10
show the Chla in the MIZ as a function of the date of the
sea-ice retreat (tR). The frequency occurrence of tRis also
shown on top of each boxplot. These distributions allow
an understanding of the dynamic of the ice-edge bloom in
each subregion of the HBS.
The western part of the bay (NW and SW; Figure 10)
showed a similar relationship between the timing of the
ice breakup and the magnitude of the ice-edge bloom. The
sooner the sea ice breaks up, the higher the Chla in the ice
edge. The NW subregion, known to host wind-driven poly-
nyas (Landy et al., 2017), shows more frequent sea-ice
retreat in June (86.4%of the time), though it was also
often observed in May (21.7%of the time). The median
Chla values were >1 mg m
–3
when the breakup was in
May, and approximately 0.7 mg m
–3
if it occurred during
the first half of June. We observed oligotrophic conditions
whenever the ice melt happened after June 15, with Chla
well below the threshold of 0.5 mg m
–3
. In the SW in May,
however, ice-edge blooms only occurred 4.2%of the time.
Sea-ice breakup happened mainly between mid-June and
mid-July (60.3%of the time) and showed ice-edge Chla
very close to the 0.5 mg m
–3
threshold. The influence of
the Churchill and the Nelson rivers may explain the larger
Chla in summer compared to the NW subregion where
river inputs were minimum in the North. The increase in
Chla in the ice edge after August 1 only represented 0.3%
and 2.1%of the time in the NW and SW, respectively, and
thus was negligible.
The Narrows of HB entrance and the central HB (Fig-
ure 10) showed a similar relationship between the timing
of the breakup and the magnitude of the ice-edge bloom.
First, the sea-ice retreat happened mainly in late June
(approximately 50%of the time) or early July (approxi-
mately 30%of the time) but never before mid-May. Sec-
ond, when the sea-ice retreat between mid-May and mid-
Figure 8. Frequency of occurrence of four trophic stages in the ice-edge zone. Maps of frequency occurrence of four
categories of chlorophyll-a concentration in the ice-edge zone based on 21 years (1998–2018) of ocean color
observations defining MIZ trophic conditions: (A) oligotrophy, (B) moderate oligotrophy, (C) bloom, and (D) major
bloom. DOI: https://doi.org/10.1525/elementa.039.f8
Barbedo et al: Ice-edge Blooms in the Hudson Bay System Art. 8(1), page 11 of 25

June (approximately 11%of the time), moderate phyto-
plankton blooms characterized the ice edge with a median
Chla around 1.0 mg m
–3
. Third, when the sea-ice retreat
occurred after mid-June, it was followed by Chla approx-
imately 0.5 mg m
–3
. Finally, slightly higher Chla was found
in the ice edge when the sea ice retreated in late July, but
this scenario represents only 7%or 8%of cases.
In the eastern part of the bay (NE and SE; Figures 10),
Chla remained well above the bloom threshold regardless
of the date of the sea-ice retreat. We found higher Chla in
the SE, compared to NE, just south of the Belcher Islands
(Figure 6A). Similarly, the Hudson Strait should be con-
sidered as a highly productive system because the median
Chla in the ice-edge zone remained close to 1.5 mg m
–3
when sea ice retreated between mid-May and the end of
July. There, early ice retreat (before mid-May) resulted in
lower Chla compared to the rest of the season.
In summary, the timing of sea-ice retreat seems to play
a crucial role in setting the magnitude of the ice-edge
bloomobservedfromspaceinmostoftheHBS(NW,
Figure 9. Frequency of occurrence of five phytoplankton phenological types throughout the marginal ice zone. Maps of
frequency of occurrence for the five types of MIZ phenology between 1998 and 2018, as depicted in Figure 1: (A)
Oligotrophy/old under-ice blooms, (B) probable (recent) under-ice bloom, (C) mesotrophic/nutrient replenished, (D)
bloom triggered in ice-free waters, and (E) bloom triggered under-ice. DOI: https://doi.org/10.1525/elementa.039.f9
Art. 8(1), page 12 of 25 Barbedo et al: Ice-edge Blooms in the Hudson Bay System

SW, Central, and Narrows), except in the eastern part of
the bay (SE, NE, and HS). Significant ice-edge blooms were
more restricted to May in the NW and SW subregions
where the sea-ice retreat was earlier compared to the rest
of the HBS (Figure 6C).
The response of ice-edge blooms to the timing of sea-
ice retreat showed some distinct spatial patterns.
Figure 11A shows the slope qChlaIEZ
qtRof the linear
regression between the maximum Chla in the ice-edge
zone and tRon a pixel basis. The negative (positive) slope
indicates that the sooner (later) the ice retreat, the higher
the Chla in the ice-edge zone. Interestingly, we found
strong negative slopes in the Northwest HB, the Narrow
at the HB entrance, and in the central HS. Negative slopes
were also found in about half of pixels in the NE subre-
gion. Some coastal spots in the south presented positive
slopes, which are associated with river discharge. The
Figure 10. Influence of sea-ice retreat timing on Chla along the ice edge in Hudson Bay subsystems. Boxplots of
maximum chlorophyll-a concentration (½Chlamax) in the marginal sea-ice zone (MIZ) in relation to day of the year of
sea-ice retreat (tR) in each subregion of the Hudson Bay System. The central red line mark is the ½Chlamax median, the
edges of the blue boxes are the 25th and 75th percentiles, the whiskers extend to the most extreme data points
between +0.26 sconsidering a normal distribution and the percentage of valid pixels for each boxplot in relation to
each subregion is marked above the respective upper whisker. The significant differences between these boxes were
ensured by the ANOVA statistical test at a confidence interval of 99.9%. Chla above the threshold of 0.5 mg m
–3
(green line) defines the marginal ice bloom occurrence (Perrette et al., 2011). DOI: https://doi.org/10.1525/
elementa.039.f10
Barbedo et al: Ice-edge Blooms in the Hudson Bay System Art. 8(1), page 13 of 25

increment of Chla with delays in the sea-ice retreat was
also observed southeast of Belcher Islands, where tidal mix-
ing is particularly intense (Webb, 2014). However, these do-
mains were also associated with relatively high root mean
square errors (RMSE above 1.00 mg m
–3
), suggesting large
uncertainty in the relationship. Finally, we noted a neutral
response of ice-edge blooms to tRin most of the central HB.
3.5. The magnitude of the ice-edge bloom and
atmospheric forcing
We computed a winter temperature index using the nor-
malized air temperature anomalies relative to the 1948–
2018 mean (T0
air). A similar approach was used by Hoch-
heim and Barber (2014) to investigate the impact of sur-
face air temperatures on sea-ice breakup in the HBS. Here,
the response of ice-edge bloom magnitude is discussed in
terms of the positive, negative, and neutral slope of the
relationship between the winter air temperature anomaly
and the maximum Chla in the ice-edge zone
(qChlaIEZ =qT0
air;Figure 12A). The positive (negative) slope
indicates that the warmer (colder) the winter, the higher
(lower) the Chla in the ice-edge zone in the spring–sum-
mer season. Based on this analysis, we found a positive
relationship between air temperature and ice-edge bloom
in the Hudson Bay entrance and NE (red pixels in
Figure 12A), with NW and SE subregions appearing more
sensitive than the rest of the bay (blue pixels). This positive
relationship means that higher magnitude ice-edge
blooms are expected after a cooler winter. Some spots, for
example, near large river plumes (Churchill, Nelson), had
a negative relation between ice-edge blooms and T0
air, but
were also associated with the highest RMSE. This result
may be related to river discharge variability (timing or
volume). In contrast to the HB, the Hudson Strait pre-
sented a complex spatial distribution concerning the
relation between ice-edge Chla and T0
air, with strong neg-
ative slopes observed in most of the area.
We subsequently examined the impact of climatic tele-
connections, such as the North Atlantic Oscillation and
Arctic Oscillation, on the phytoplankton dynamic in MIZ,
by correlating the maximum Chla in the ice-edge zone
with the NAO/AO index (Thompson and Wallace, 1998).
As illustrated in Figure 13A and B,themainpattern
observed is a positive correlation between NAO/AO and
ice-edge zone Chla in the NW and south of Belcher Islands
polynyas, where sea ice usually retreats earlier (Landy et
al., 2017). As reported in the Figure 13C (D), the time
series of normalized ice-edge Chla and NAO (AO) index
correlated significantly, with a Pearson correlation coeffi-
cient of 0.27 (0.29) in a confidence interval of 95%for the
whole HBS and 0.46 (0.41) for the NW polynya. The cor-
relation maps between ice-edge blooms, AO (Figure 13A),
and NAO (Figure 13B) showed a similar pattern: a vast
region of high correlation controlled by the NW polynya
dynamic, negative correlation in the HB and NE bay, and
high correlation south of Belcher Island. The maximum
valueofbothChla,AO,andNAOindicesduringthe
1998–2018 period occurred in 2015 when we observed
a massive phytoplankton bloom in the whole HBS and the
NW polynya (Figures 7 and 13C, D). In contrast, the
lowest NAO/AO indices occurred in 2010, preceding a neg-
ative anomaly in ice-edge Chla in the HBS by one year but
simultaneous with a local minimum in the NW polynya
(Figure 13C vs. 13D). The primary mismatches with NAO/
AO indices occurred in 2005 and 2017 (Figure 13C) when
considering the whole HBS and 2000 and 2017 in the NW
polynya (Figure 13D).
4. Discussion
We first discuss the inherent limitations of ocean color
data for describing the phytoplankton dynamic in an
Figure 11. Relationships between the timing of sea-ice retreat and the maximum Chla along the ice edge. Linear
regressions between tRand the maximum Chla in the ice-edge zone (i.e., ½ChlaIEZ ðx;yÞ¼slopeðx;yÞ
tRðx;yÞþoffsetðx;yÞ) were performed for each grid cell (x;y) using the annual maps shown in Figures S1 and S2
(n¼21 for each grid cell). The slope of the regressions is presented in A and its associated root mean square error
(RMSE) in B. DOI: https://doi.org/10.1525/elementa.039.f11
Art. 8(1), page 14 of 25 Barbedo et al: Ice-edge Blooms in the Hudson Bay System

Figure 12. Relationships between anomalies of winter air temperature and maximum Chla in the marginal ice zone.
Linear regressions between the winter air temperature anomaly (T0
air) and the maximum Chla in the ice-edge zone (i.e.,
½ChlaIEZ ðx;yÞ¼slopeðx;yÞtRðx;yÞþoffsetðx;yÞ) were performed for each grid cell (x;y) using the annual map shown
in Figures S1 (n¼21 for each grid cell). The slope of the regressions is presented in A and its associated root mean
square error (RMSE) in B. Note that a single value of T0
air (FigureS4) was used for the whole HBS due to the lack of
spatial variability. DOI: https://doi.org/10.1525/elementa.039.f12
Figure 13. Teleconnection between climatic indices (NAO/AO) and phytoplankton ice-edge blooms. Map of correlation
coefficients (P>95%) between Chla maxima in the ice-edge zone and AO (A) and NAO (B) between annual 1998 and
2018 generated using the Chla GSM algorithm from Globcolour and climatic indexes obtained from CPC/NOAA.
Pearson’s correlation with t-test smaller than 95%confidence interval was removed. A 95%confidence interval was
applied to the time series of climatic indexes AO and NAO and normalized mean values of annual Chla maxima in the
ice-edge zone for the whole HBS (C) and in the Northwest Hudson Bay polynya (D). DOI: https://doi.org/10.1525/
elementa.039.f13
Barbedo et al: Ice-edge Blooms in the Hudson Bay System Art. 8(1), page 15 of 25

ice-infested region under the influence of significant riv-
erine inputs. Then, we explore the variability of Chla in the
ice-edge zone in terms of potential physical forcings that
control nutrient availability and incoming light in the
upper layer and consequently the magnitude of the
bloom. We further examine the impact of climatic changes
on the phytoplankton dynamic throughout the MIZ. To
achieve this objective, we explore the relationships
between ice-edge bloom intensity and climate indices
(NAO/AO), winter air temperature, and sea-ice retreat. Cli-
mate indices are used to describe the planetary telecon-
nections between the northern polar vortex. Finally, we
focus on the HB NW polynya, an essential hot spot for
marine life, as it shows high sensitivity to large-scale cli-
mate forcing.
4.1. Satellite ocean color of the HBS
Satellite remote sensing provides high revisit frequency,
global coverage, and continued operational monitoring
capability for access-limited regions such as HBS (Babin
et al., 2015; Lee et al., 2015). Despite inherent limitations
(IOCCG, 2015), that is, restricted vertical range to the first
optical depth (1=Kd), lack of data under clouds or sea ice,
presence of terrigenous optical components and sea-ice
contamination of ocean color products, satellite data have
proven efficient to describe the phytoplankton dynamic at
high latitudes, including in the MIZ (Maynard and Clark,
1987; Perrette et al., 2011; Lowry et al., 2014; Renaut et al.,
2018).
Notwithstanding, standard empirical Chla algorithms for
global ocean data, and even Arctic-adapted ones, can be
biased (IOCCG, 2015), calling into question the choice of
a fixed threshold to detect phytoplankton blooms. In the
Arctic and sub-Arctic seas, empirical algorithms tend toover-
estimate (underestimate) Chla in the lower (higher) range
(Cota et al., 2004; Hirawake et al., 2012; IOCCG, 2015).
In particular, CDOM and non-algal particles can dimin-
ish the performance in these algorithms because, as phy-
toplankton pigments, they absorb the blue part of the
visible spectrum (Bricaud et al., 1981). CDOM is known
to be a dominant optical component in most of the HBS,
making this region optically complex (Granskog et al.,
2007; Mundy et al., 2010; Gue
´guen et al., 2011; Xi et al.,
2013; Burt et al., 2016; Heikkila et al., 2016). The GSM
algorithm can better deal with the presence of CDOM
compared to empirical algorithms and was selected to
minimize the impact of other optical constituents on Chla
retrievals (Ben Mustapha et al., 2012). The nearshore
region where river inputs contribute more to the water
reflectance variability than phytoplankton was avoided by
masking shallow water pixels (i.e., depth < 50 m). Gran-
skog et al. (2009) found that river runoff and CDOM are
constrained within the coastal domain (100–150 km from
the shore), where they are transported along the coast
counterclockwise around the HBS. Xi et al. (2013) reported
that phytoplankton was the main optically significant
component driving the seasonal variability of light absorp-
tion in the HBS, while CDOM remained relatively stable
over the annual cycle. These results provide confidence in
the present Chla time-series analyses. However, we cannot
exclude an influence of riverine-derived CDOM on Chla, in
particular along the eastern coast of HB, where most river
runoff is transported before its exit through the Hudson
Strait (Saucier et al., 2004). A high CDOM background in
the east of HB (Xi et al., 2013) explains the relatively high
Chla observed in the ice edge regardless of the sea-ice
breakupintheeasternpartofHB(NEandSEinFig-
ure 10). An evaluation of the GSM algorithm based on
in situ measurements in the HBS beyond the scope of this
study but would be necessary to evaluate the potential
bias in Chla and consequently the choice of 0.5 mg m
–3
as a threshold for detecting pelagic blooms.
Here, we adopted a fixed Chla-based threshold of 0.5
mg m
–3
to detect the ice-edge bloom (Perrette et al., 2011)
instead of dynamic methods like, for example, rate of
change method, cumulative biomass-based thresholds
(Brody et al., 2013), or model fitting (Ardyna et al., 2014;
Marchese et al., 2017). Because the ice-edge bloom can
reach its peak just after the sea-ice retreat (Figure 1 and
type V in Figure 9), the pelagic cycle often began with the
highest annual values, without a well-defined Gaussian-
like seasonal evolution in Chla. Therefore, we could not
apply the Gaussian fitting method to the HBS.
4.2. Winter air temperatures and ice retreat impact
on ice-edge blooms
What role do winter air temperatures (T0
air) and sea-ice
retreat (tR) play in the magnitude and variability of ice-
edge blooms? Can these relationships be affected by cli-
mate change? We used 21 years of SIC and Chla to address
these questions.
As illustrated in Figures 11 and 12,tRand T0
air can
have a very strong but local impact on the balance
between under-ice and pelagic production throughout the
MIZ (Palmer et al., 2014). This impact can be explained by
the interactions of various oceanographic processes in the
MIZ, such as tidal resonance (Webb, 2014), water mass
exchange throughout the Hudson Strait (Sutherland et
al., 2011), polynya dynamics (Landy et al., 2017), and fresh-
water inputs (St-Laurent et al., 2011). Notably, the rela-
tively strong negative correlation between Chla and tRin
the northern part of the HBS illustrated in Figure 11 is
very similar to the Arctic water intrusion pattern in the
system depicted by Wang et al. (1994), Saucier et al.
(2004), and Piecuch and Ponte (2015).
In winter, primary production in the HBS is limited by
photosynthetic available radiation, as confirmed by the
absence of dinoflagellate cysts (a proxy for under-ice pro-
duction) reported by Heikkila et al. (2016) in two coastal
sites in both west and east HB. On the other hand, physical
processes controlling the cryosphere, that is, brine produc-
tion and vertical mixing, are essential to bringing deep
nutrient-rich waters to the upper ocean. These processes
can be critical in the HBS because it is a relatively shallow
sea (Stewart and Lockhart, 2005; Granskog et al., 2011).
Also, compared to other polar domains of the Arctic, the
sea ice is thinner, which may result in less brine rejection
and relatively weak vertical mixing and nutrient replen-
ishment during winter.
Art. 8(1), page 16 of 25 Barbedo et al: Ice-edge Blooms in the Hudson Bay System

Winter air temperature is a good proxy for sea-ice pro-
duction and the overall winter severity. Hochheim and
Barber (2014) reported that a positive anomaly of only
1C decreased the sea-ice coverage by 14%and advanced
the breakup by one week. However, this generalization
can be ambiguous concerning its effect on phytoplank-
ton productivity because winter air temperature can
impact many different key processes. Those include the
nutrient replenishment due to sea-ice or brine produc-
tion (Landy et al., 2017) or winter convection in the open
waters (Stewart and Lockhart, 2005; Granskog et al.,
2011; St-Laurent et al., 2011), the PAR attenuation by sea
ice and snow cover (Horvat et al., 2017), the timing and
length of the open water season (Ardyna et al., 2014;
Hochheim and Barber, 2014), and the riverine input
(De
´ry and Wood, 2004; De
´ry et al., 2005). The latter
controls the amount of CDOM and nutrients (Granskog
et al., 2007), which have a contrasting impact on phyto-
plankton production.
The complexity of air, sea, and ice interactions at high
latitude makes the interpretation of the relationship
between ice-edge Chla and winter air temperature diffi-
cult (Figure 12A). Indeed, various feedbacks are in play
between polynya dynamics, air temperature, vertical mix-
ing, brine production, and phytoplankton blooms (Chaud-
huri et al., 2014). For example, Saucier et al. (2004)
reported that the NW polynya dynamic causes significant
sensible heat loss at a rate of 100 W m
–2
in winter. The
presence of a polynya during the cold season tends to
intensify the sea-to-air heat flux, which potentially warms
the atmosphere locally (Saucier et al., 2004). Similarly, the
Weddell Sea polynya, Antarctica, can effectively heat the
atmosphere at the regional scale (Campbell et al., 2019).
However, the polynya may only affect local air tempera-
tures, which may be difficult to assess, given the coarse
resolution of the NCEP/NCAR reanalysis data (Winsor and
Bjo
¨rk, 2000).
The strong negative relationship between ice-edge Chla
and winter air temperature in the HB polynyas (i.e., NW
and Belcher Island) reflects the intensification of polynya
dynamics, which leads to efficient nutrient replenishment
in the euphotic zone (Figure 12). Atmospheric cooling of
the ocean also influences the amount of ice and brine
production, which further increases vertical mixing and
weakens the under-ice stratification (Stewart and Lock-
hart, 2005; Granskog et al., 2011). Consequently, the nutri-
ent stock available at bloom onset may be greater (Nguyen
et al., 2009; Barthe
´lemy et al., 2015). Indeed, thick sea ice
stocks a large quantity of freshwater, which provides low-
density water and stabilizes the upper ocean upon sea-ice
melting (Galbraith and Larouche, 2011).
In summary, an inheritance effect of winter sea-ice
production could indirectly impact the level of phyto-
plankton primary production that will occur in the
spring–summer season. Therefore, we conclude that
cooler air temperatures intensify the wind-driven poly-
nyas of the HB due to steady atmospheric cooling of the
ocean, which supports ice (and brine) production, vertical
mixing, and deep nutrient replenishment during winter.
As a result, as soon as the stratification is reestablished
and light becomes available in spring, phytoplankton can
reach higher biomass.
4.3. Ice-edge phytoplankon blooms response to
global and local forcing
A warmer climate has already been established in the HBS
due to warmer air and less frequent cold and dry polar air
masses (Leung and Gough, 2016). Based on air tempera-
ture analysis between 1948 and 2018 (Figure 2), positive
anomalies have been predominant in most seasons since
1998. Impacts on the cryosphere dynamics have been
observed: thinner sea-ice cover, early sea-ice retreat, late
ice recover, and longer open water season (Hochheim and
Barber, 2014; Kowal et al., 2017). These changes in the sea-
ice dynamics may have driven significant changes in the
marine ecosystems, with potential cascading impacts on
all trophic levels in the HBS (Hoover et al., 2013; Keller et
al., 2014; Andrews et al., 2017).
Mechanisms controlling the spring–summer and fall
blooms are distinct. Our results suggest that pelagic pro-
duction may have increased in the HBS since the open
water season has increased in duration and double blooms
are more frequent (Figure 2). Sea-ice dynamics and
incoming PAR in the upper ocean layer control the
spring–summer bloom. Kahru et al. (2011) reported a pro-
nounced trend toward earlier phytoplankton blooms in
about 11%of the North Polar seas area between 1997
and 2009, including the northern part of the HBS. The
latter can be related to earlier ice breakup trends (Kowal
et al., 2017). In contrast, the lengthening of the open
water season and the increase in nutrient pumping by
wind-driven turbulence (Ardyna et al., 2011) and the inver-
sion of heat fluxes drive the fall blooms. Notwithstanding,
the increase in annual phytoplankton primary production
is correlated with the length of the open water season
(Arrigo and van Dijken, 2015).
The HBS is located in a transition between subpolar
and polar domains. However, its complex topography and
its semi-confined configuration by the continental shelf
result in relative isolation with regard to global ice
exchange and ocean circulation (Sutherland et al., 2011).
Its primary connection to the global ocean takes place
through Hudson Strait. Therefore, HBS trends and dynam-
ics are especially impacted by local atmospheric processes
(Hochheim and Barber, 2010, 2014; Hochheim et al.,
2011) and by climatic teleconnections (De
´ry and Wood,
2004). Therefore, climatic forcing and global scale telecon-
nections (e.g., NAO/AO) can impact both sea-ice retreat
and recovery in HBS (Hochheim and Barber, 2010; Hoch-
heim et al., 2011).
Climate indices express global or hemispherical tele-
connections (Carmack et al., 2006). Part of the atmo-
spheric and oceanic variability of the HBS has been
correlated with both AO (De
´ry and Wood, 2004; Hoch-
heim and Barber, 2010; Hochheim et al., 2011) and NAO
indices (Qian et al., 2008). NAO is recognized as a signifi-
cant descriptor of the winter-to-winter variability over
northeastern America (Wallance, 2007; Qian et al.,
2008). It accounts for the direct North–South dipole pat-
terns between the North Atlantic and the Arctic oceans.
Barbedo et al: Ice-edge Blooms in the Hudson Bay System Art. 8(1), page 17 of 25

Meanwhile, AO (annular paradigm) also includes the
Pacific sector processes, such as the Aleutian High (Wang
et al., 2005; Wallance, 2007). Nevertheless, there is an
overlap between NAO and AO patterns in the Atlantic
sector. Therefore, NAO and AO are highly correlated and
related to similar phenomena (Ambaum et al., 2001;
Wallance, 2007), especially during winter (Rogers and
McHugh, 2002). Both NAO and AO indices are associated
with the intensity of the north polar vortex (Wallance,
2007), with a positive phase reflecting the strengthening
of the winter-time polar vortex and westerly intensifica-
tion between 50 Nand70N (Moritz et al., 2002). For
example, Hurrell et al. (2001) reported that positive NAO
strongly affects the Atlantic Ocean, causing substantial
changes in surface wind patterns with stronger westerly
winds encircling the North Pole. In contrast, Arctic warm-
ing and a wavier polar jet resulting in an intensification
of southward advection of polar air masses in North
America were observed during negative NAO/AO phases
(Baldwin and Dunkerton, 1999; Ding et al., 2014; Francis
and Vavrus, 2015; Leung and Gough, 2016; Meleshko et
al., 2016). We found that the ice-edge zone Chla and
NAO/AO indices are relatively well correlated, especially
in the NW polynya (Figure 13). The extreme event of the
highest (lowest) NAO/AO in 2015 (2010) corresponded
to a strong (weak) ice-edge bloom in NW HB polynya
(Figure 13). These results suggest that global atmo-
spheric circulation patterns, depicted by the NAO/AO,
strongly influence phytoplankton blooms in the NW
HB polynya.
Figure 14 illustrates schematically how the strength of
westerlies affects the NW HB polynya processes control-
ling ice-edge blooms during winter and spring-to-summer
transition. As discussed by Saucier et al. (2004) and Landy
et al. (2017), the NW polynya is maintained by strong
westerly winds opening up areas of water along the
northwestern coast by coastal divergence and enhancing
ice production. In winter, the NW polynya acts as an “ice
factory,” where ice growth is favored by thermodynamic
processes (ocean cooling) and then the ice is exported by
winds (Landy et al., 2017). Brine rejection due to sea-ice
production plays a fundamental role in vertical stratifica-
tion and deep convection (Prinsenberg, 1988) and results
in the nutrient replenishment of the euphotic zone
through vertical mixing. Meanwhile, we found that early
sea-ice retreat results in more intense ice-edge blooms
(NW in Figures 10 and 11). This finding suggests that
nutrient stocks set up in winter and the timing of polynya
expansion in the spring-to-summer transition are the
main drivers of the balance between pelagic and under-
ice production throughout the MIZ. Notably, NAO/AO
events can control the feedback between westerlies, the
NW HB polynya dynamic, and the ice-edge blooms
because initially brine rejection forced by ice production
and export sets up the preconditions of nutrients stocks in
the euphotic zone. After that, early polynya expansion due
to intensification of westerlies results in ice-edge bloom
intensification during the spring-to-summer transition
(Figure 10-NW).
De
´ry and Wood (2004) highlighted that AO could
explain 70%–90%ofthevarianceinHBSriverdischarge
anomalies. According to De
´ry and Wood, the origin of
dominant air masses over the HBS controls this pattern.
Duringthepositive(negative) AO index, north easterly
(north westerly) winds that advect relatively warm (cool),
moist (dry) air from the Labrador Sea to the HBS increase
theriverdischarge(De
´ry and Wood, 2004; Hochheim and
Barber, 2014). We found weak correlations between AO
and ice-edge zone Chla in the southern HB, where river-
ine input into the HBS is more important (St-Laurent et
al., 2011). We also found no significant correlations
between the maximum Chla in the ice-edge zone and
Figure 14. Conceptual model of teleconnections of NAO/AO to phytoplankton dynamics in the NW HB Polynya. DOI:
https://doi.org/10.1525/elementa.039.f14
Art. 8(1), page 18 of 25 Barbedo et al: Ice-edge Blooms in the Hudson Bay System

riverine discharge, even at local scales (results not
shown), suggesting a marginal role of rivers in the pri-
mary productivity of the HBS.
4.4. Potential role of under-ice blooms
Recent observations have suggested that under-ice phyto-
plankton blooms can draw down the surface nutrient pool
before the open water season starts (Mundy et al., 2009;
Arrigo et al., 2014). Then, the subsurface chlorophyll-
a maximum sinks near the nitracline before or just after
the ice retreat, allowing the phytoplankton to achieve
a balance between PAR and nutrients, as illustrated in
Figure 14. As a result, the ocean color satellites detect
a low Chla in the ice-edge zone (Horvat et al., 2017; blue
and red curves in Figure 1). The frequency occurrence
map depicting this situation (Figure 9A and B) suggests
that a vast portion of the central-western part of the HBS
is experiencing under-ice blooms. However, this scenario
may also reflect the oligotrophic nature of the system.
Only in situ observations or model simulations can
help to resolve the above situation (under-ice bloom vs.
oligotrophy). Most field observations before the BaySys
projectwerebasedonicecamps,whichfocusedmoreon
ice algae or open waters later in summer. Michel et al.
(1993) and Michel et al. (1988) reported a maximum of
vertically integrated ice-algal Chla of 23.6 mg m
–2
in
April, while Monti et al. (1996) reported algal concentra-
tionsattheice/waterinterfaceashighas100.6mgm
–2
between April and May. High ice-algal concentrations
were followed by an increase in water column Chla reach-
ing moderate concentrations (i.e., 2 and 4 mg m
–3
)
in May during the ice melt (Runge et al., 1991; Michel
et al., 1993), suggesting an initiation of the bloom under
ice (type V in Figure 1). Model simulations also suggest
that nutrient draw-down in the surface waters begins in
May, but the maximum diatom biomass is reached after
theiceretreatinJuneorJuly(Sibertetal.,2011).
Recently, however, Tremblay et al. (2019) concluded that,
except in river plumes and some upwelling spots (e.g.,
Foxe Peninsula, Belcher Islands and Hudson Strait), the
HBSshouldbeconsideredasanoligotrophicsystem.
More observations in the MIZ during the spring-to-
summer transition are still needed.
5. Conclusions and perspectives
We showed that phytoplankton phenology in the HBS is
subject to substantial spatiotemporal variability and
closely linked to large-scale atmospheric forcings. In
recent years, the phenology has been characterized by two
peaks in Chla, one after the sea-ice breakup in the mar-
ginal ice zone (May–June) and one in the fall. Here, we
focused our effort on the ice-edge bloom. In the western
part of the HB, the magnitude of the ice-edge bloom
depends on the timing of ice breakup, with more intense
blooms occurring when ice retreats early. The northwest-
ern polynya stands out as a significant feature in the HBS
in terms of primary production, and its variability is highly
coupled to the northern hemisphere climate variability
(NAO and AO indices). At the bay scale, we found no direct
evidence that river discharge, which can supply the surface
waters with nutrients and impact the vertical stratifica-
tion, influenced the MIZ.
The relative contributions of the under-ice, MIZ, and
fall blooms to total annual primary production remain to
be established. Regionally tuned satellite-derived primary
production models (Ardyna et al., 2013; Be
´langer et al.,
2013; Lee et al., 2015) could be used at least to assess the
contribution of the MIZ and fall blooms. The combination
of in situ observations, satellite monitoring (e.g., ocean
color in synergy with sea-ice thermodynamic stages and/
or albedo), and 3D ocean models coupled to biological
models will be most promising to predict how all compo-
nents of the HBS marine ecosystem will be impacted by
climate change, including primary productivity.
Supplemental files
The supplemental files for this article can be found as
follows:
Figure S1. Ice-edge blooms between 1998 and 2018 in
the Hudson Bay System. Ice-edge blooms detected by the
maximum Chla during the 24 days after sea-ice retreat
(first day of continuous SIC < 10%) (Perrette et al.,
2011) between 1998 and 2018. Days of the year and their
respective calendar date: 120 (Apr 30), 135 (May 15), 150
(May 30), 165 (Jun 14), 180 (Jun 29), 195 (Jul 14), 210 (Jul
29), 225 (Aug 13), and 240 (Aug 28).
Figure S2. Sea-ice retreat between 1998 and 2018 in
the Hudson Bay System. Sea-ice retreat is detected when
SIC (Comiso, 2000) reaches the threshold of 10%(Perrette
et al., 2011).
Figure S3. Maps of four phytoplankton phenological
categories throughout marginal ice zone between 1998
and 2018. Annual classification maps of MIZ phenology
based on evolution of Chla following sea-ice retreat (tR).
The 4 categories are oligotrophic state or old under-ice
bloom (blue); probable (recent) under-ice bloom (red);
mesotrophic system where efficient nutrient replenish-
ment is in place (green), bloom triggered in ice-free waters
(orange); and bloom triggered under ice (cyan).
Figure S4. Winter air temperature anomalies in the
Hudson Bay System. Winter anomalies of surface air tem-
perature (T0
air) calculated using air temperatures from
National Centers for Environmental Prediction/Atmo-
spheric Research (NCEP/NCAR) Reanalysis Project as the
difference between each winter (January, February, and
March) average and its corresponding climatology and
normalized by the standard deviation, using data from the
1948 to 2018 period.
Data Accessibility Statement
The day of the year of Sea-ice retreat (tR) and maximum
chlorophyll-aconcentration in the sea-ice edge zone pro-
duced in this research are made available through the
BaySys data repository (open access).
Acknowledgments
This work is a contribution to the Natural Sciences and
Engineering Council of Canada (NSERC) Collaborative
Research and Development (CRD) project titled BaySys.
Funding for this research was graciously provided by
Barbedo et al: Ice-edge Blooms in the Hudson Bay System Art. 8(1), page 19 of 25

Manitoba Hydro, NSERC, Amundsen Science, and the Ca-
nada Research Chairs program. In addition, this research
contributes to the ArcticNet Networks of Centers of Excel-
lence and the Arctic Science Partnership (ASP). We thank
the NSIDC, NASA-OBPG, and ESA-GlobColour for provid-
ing satellite data freely.
The authors would like to express their gratitude to Prof.
Christopher Horvat, Brown University (Rhode Island/USA),
because his review and directions were very important to
improve this study.
Additional information: http://umanitoba.ca/
faculties/environment/departments/ceos/research/
BaySys.html.
Funding information
LBDF is supported by the BaySys grant as well as UQAR
and Que
´bec-Ocean grants. This study was also supported
by individual grants from ArcticNet and NSERC (355774-
2009 and RGPIN-2014-03680 to S.B.).
Competing interests
The authors have no competing interests to declare.
Author contributions
Contributed to conception and design: LBF and SB.
Contributed to analysis and interpretation of data: all
authors.
Drafted and/or revised the article: all authors.
References
Amante, C, Eakins, BW. 2009. Etopo1 1 Arc-Minute
global relief model: Procedures, data sources and
analysis. Boulder, CO: National Geophysical Data
Center, NOAA.
Ambaum, MHP, Hoskins, BJ, Stephenson, DB.2001.
Arctic oscillation or North Atlantic oscillation? J
Clim 14(16): 3495–3507. DOI: http://dx.doi.org/
10.1175/1520-0442(2001)014<3495:AOONAO>2.
0.CO;2.
Andrews, J, Babb, D, Barber, DG. 2017. Climate change
and sea ice: Shipping in Hudson Bay, James Bay,
Hudson Strait, and Foxe Basin. Elem Sci Anth
6(19). DOI: http://dx.doi.org/10.1525/elementa.
281.
Antoine, D, Morel, A, Gordon, HR, Banzon, VF, Evans,
RH. 2005. Bridging ocean color observations of the
1980s and 2000s in search of long-term trends. J
Geophys Res Oceans 110(C6). DOI: http://dx.doi.
org/10.1029/2004JC002620.
Ardyna, M, Babin, M, Gosselin, M, Devred, E, Be
´langer,
S, Matsuoka, A, Tremblay, J-E
´. 2013. Parameteri-
zation of vertical chlorophyll-a in the Arctic Ocean:
Impact of the subsurface chlorophyll maximum on
regional, seasonal, and annual primary production
estimates. Biogeosci 10: 4383–4404. DOI: http://dx.
doi.org/10.5194/bg-10-4383-2013.
Ardyna, M, Babin, M, Gosselin, M, Devred, E, Rainville,
L, Tremblay, J-E
´. 2014. Recent Arctic Ocean sea ice
loss triggers novel fall phytoplankton blooms.
Geophys Res Lett 41(17): 6207–6212. DOI: http://
dx.doi.org/10.1002/2014GL061047.
Ardyna, M, Claustre, H, Salle
´e, JB, D’Ovidio, F, Gentili,
B, van Dijken, G, D’Ortenzio, F, Arrigo, KR.2017.
Delineating environmental control of phytoplank-
ton biomass and phenology in the Southern Ocean.
Geophys Res Lett 44(10): 5016–5024. DOI: http://
dx.doi.org/10.1002/2016GL072428.
Ardyna, M, Gosselin, M, Michel, C, Poulin, M, Trem-
blay, JE
´. 2011. Environmental forcing of phyto-
plankton community structure and function in the
Canadian High arctic: Contrasting oligotrophic and
eutrophic regions. Mar Ecol Prog Ser 442: 37–57.
DOI: http://dx.doi.org/10.3354/meps09378.
Arrigo, KR, Perovich, DK, Pickart, RS, Brown, ZW, van
Dijken, GL, Bates, NR, Arrigo, KR. 2014. Phyto-
plankton blooms beneath the sea ice in the Chukchi
sea. Deep Sea Res 105: 1–16. DOI: http://dx.doi.
org/10.1016/j.dsr2.2014.03.018.
Arrigo, KR, Perovich, DK, Pickart, RS, Brown, ZW, van
Dijken, GL, Lowry, KE, Mills, MM, Palmer, MA,
Balch, WM, Bahr, F, Bates, NR, Benitez-Nelson,
C, Bowler, B, Brownlee, E, Ehn, JK, Frey, KE, Gar-
ley, R, Laney, SR, Lubelczyk, L, Mathis, J, Matsuo-
ka, A, Mitchell, BG, Moore, GWK, Ortega-
Retuerta, E, Pal, S, Polashenski, CM, Reynolds,
RA, Schieber, B, Sosik, HM, Stephens, M, Swift,
JH. 2012. Massive phytoplankton blooms under Arc-
tic Sea Ice. Science 336(6087). DOI: http://dx.doi.
org/10.1126/science.1215065.
Arrigo, KR, van Dijken, GL. 2015. Continued increases in
Arctic Ocean primary production. Prog Oceanogr
136: 60–70. DOI: http://dx.doi.org/10.1016/j.
pocean.2015.05.002.
Assmy, P, Ferna
´ndez-Me
´ndez, M, Duarte, P, Meyer, A,
Randelhoff, A, Mundy, CJ, Olsen, LM, Kauko,
HM, Bailey, A, Chierici, M, Cohen, L, Doulgeris,
AP,Ehn,JK,Fransson,A,Gerland,S,Hop,H,
Hudson,SR,Hughes,N,Itkin,P,Johnsen,G,
King, JA, Koch, BP, Koenig, Z, Kwasniewski, S,
Laney, SR, Nicolaus, M, Pavlov, AK, Polashenski,
CM, Provost, C, Ro
¨sel, A, Sandbu, M, Spreen, G,
Smedsrud, LH, Sundfjord, A, Taskjelle, T, Tatar-
ek, A, Wiktor, J, Wagner, PM, Wold, A, Steen, H,
Granskog, MA. 2017. Leads in Arctic pack ice
enable early phytoplankton blooms below snow-
covered sea ice. Sci Rep 7(January): 1–9. DOI:
http://dx.doi.org/10.1038/srep40850.
Babin, M, Be
´langer, S, Ellingsen, I, Forest, A, Fouest,
VL, Lacoura, T, Ardynaa, M, Slagstad, D.2015.
Estimation of primary production in the Arctic
Ocean using ocean colour remote sensing and cou-
pled physical-biological models: Strengths, limita-
tions and how they compare. Prog Oceanogr 139.
DOI: http://dx.doi.org/10.1016/j.pocean.2015.08.
008.
Baldwin, MP, Dunkerton, TJ. 1999. Propagation of the
Arctic Oscillation from the stratosphere to the tro-
posphere. J Geophys Res 104(D24): 30937–30946.
DOI: http://dx.doi.org/10.1029/1999JD900445.
Art. 8(1), page 20 of 25 Barbedo et al: Ice-edge Blooms in the Hudson Bay System