Since 1998, satellite-based NPP in the AO
has increased by 57% (Fig. 2C and Table 1), far
outpacing previously estimated rates ( 2 – 4 , 6 ).
The most dramatic increases were on the eastern
interior shelves (Siberian, Laptev, and Kara sub-
regions), the inflow shelves (Chukchi and Barents
subregions), and the Basin subregion (Table 1).
The inflow shelves together contributed 70%
of the AO NPP increase (Table 1), which is
consistent with past studies that noted the
importance of the Barents and Chukchi seas to
AO NPP ( 2 , 3 , 18 ). A more modest yet statistically
significant increase was seen on the western
interior shelf (Beaufort subregion) (Table 1).
The outflow shelves (Nordic, Canada, and Baffin
subregions) exhibited the smallest percent in-
crease in NPP, with the Nordic and Canada
subregions showing no statistically significant
trend (Table 1).
Historically, greater OW area and longer
OW duration associated with sea ice decline
have been the primary drivers of increased
spatially integrated NPP across the AO ( 2 – 5 ).
We found that although there were significant
regional increases in OW duration (Table 1),
there was no significant trend in OW duration
across the entire AO, and OW duration was
not a significant predictor of changes in AO
NPP (Table 1). Although OW area significantly
increasedintheAOandmostofitssubregions
(Table 1), these increases have slowed in re-
cent years (Fig. 2A and Table 1). This recent
deceleration in sea ice decline is likely due to
internal climate variability temporarily masking
human-induced changes ( 19 ). Regardless of
the cause, although OW area explained 74%
of the variance in AO NPP between 1998 and
2008 when sea ice was declining rapidly (Table 2),
the relationship became less significant be-
tween 2009 and 2018, when rates of sea ice
loss slowed, explaining only 20% of the variance
in NPP (Table 2). This indicates that the recent
increases in AO NPP were not driven by in-
creases in OW area alone.
We found that increased Chlaexplained
80%ofvarianceinAONPPbetween2009and
2018 compared with only 26% during the pre-
vious decade, when changes in OW area con-
trolled the trend in NPP (Table 2). Within the
Barents subregion, which contributed more
than any other region to AO NPP, significant
increases in Chlasince 1998 sustained greater
NPP despite the slowing of OW expansion
(Table 1). Clearly, changes in phytoplankton
biomass over the past decade were largely
responsible for the sustained increase in NPP
across the AO (Fig. 2C), particularly in the in-
flow shelves (Fig. 1B), despite the slowing of
sea ice loss.
There are a few possible causes for the
observed increase in phytoplankton Chlaover
the past decade. Photoacclimation can lead
to altered cellular Chlaconcentrations, but
incident light, mixed-layer depth, and light
attenuation within the water column did not
change sufficiently during our study period
to significantly alter C:Chlaratios in areas
where Chlaincreased (fig. S1), so this possibil-
ity can be eliminated. Earlier phytoplankton
blooms ( 4 ) could intensify the mismatch be-
tween grazing and phytoplankton growth, re-
sulting in higher Chlaconcentrations in recent
years. However, this possibility is diminished
by annual changes in Chlabeing the same in
spring (April through June) as they were in
summer (July through September) (fig. S2),
when grazing rates would be expected to be
highest. Interannual variability in atmospheric
factors such as cloud cover (fig. S3), the state of
the North Atlantic Oscillation and AO (fig.
S4), wind speed (fig. S5), and the number of
upwelling-favorable wind days per year (fig.
S6) were also unrelated to the measured in-
creases in Chla.In the AO, nitrogen availability limits max-
imum phytoplankton biomass ( 7 ), so the in-
creases in Chlabeing restricted to the inflow
shelves and to the central Arctic, where sea ice
had receded back from the shelf break, points
to nutrients playing a role. This is supported
by seasonal maximum Chlaconcentrations,
which would be especially sensitive to ad-
ditional nutrient input, having increased at
more than three times the rate of mean Chla
concentrations from 2009 to 2018 in the AO
and the Barents and Chukchi seas (Table 1).
Increased advection of Atlantic Ocean waters
into the Barents Sea ( 20 ) and Pacific Ocean
waters into the Chukchi Sea ( 21 ) may supply
enough additional nutrients to sustain the
higher biomass observed on these inflow shelves
(Fig. 1B and Table 2). The Pacific inflow through
the Bering Strait, which provides most of the
nutrients that fuel Chukchi Sea summer blooms
( 22 ), has increased by ~50% from 1999 to 2015
( 21 ). The“Atlantification”of the Barents Sea
may be associated with increased advection
of phytoplankton and nutrients ( 19 , 23 – 26 ).
In addition, the incoming warm Atlantic water
reduces stratification, which has not increased
since 1979 ( 27 ), and promotes vertical mixing
( 28 ) that increases nutrient availability to sus-
tain substantial increases of phytoplankton
biomass and production ( 18 ). Last, decreased
sea ice cover and increased frequency and
intensity of storms at high latitudes ( 10 ) result
in episodic nutrient injections from the his-
torically inaccessible deep water beneath
the pycnocline through increased internal
wave mixing ( 11 ) and storm-induced upwell-
ing ( 5 , 28 , 29 ) throughout the shallow shelves.
Increased Chlaalong the interior shelf break
(Fig. 1B) may be fueled by upwelling events that
pull“new”nutrients from deep basin reser-
voirs into the nutrient-depleted upper eupho-
tic zone and that are increasingly common200 10 JULY 2020•VOL 369 ISSUE 6500 sciencemag.org SCIENCE
1998 – 2018 (% change) 1998–2018 (change year−^1 ) 1998–2008 (change year−^1 ) 2009–2018 (change year−^1 )SST (°C)............................................................................................................................................................................................................................................................................................................................................
Arctic............................................................................................................................................................................................................................................................................................................................................13.4* 0.021* 0.021 0.040*
Chukchi............................................................................................................................................................................................................................................................................................................................................24.4 0.029 0.140 0.089
Barents............................................................................................................................................................................................................................................................................................................................................14.6* 0.027* –0.009 0.083
Siberian 153.3 0.038 0.151 0.097*
............................................................................................................................................................................................................................................................................................................................................
Laptev 157.6* 0.078* 0.074 0.019
............................................................................................................................................................................................................................................................................................................................................
Kara 126.5* 0.085* 0.156* 0.041
............................................................................................................................................................................................................................................................................................................................................
Beaufort 68.7 0.074 0.111 –0.085
............................................................................................................................................................................................................................................................................................................................................
Baffin 47.4* 0.052* 0.073 –0.0120
............................................................................................................................................................................................................................................................................................................................................
Canada 78.4* 0.037* 0.051* –0.033
............................................................................................................................................................................................................................................................................................................................................
Nordic 13.7* 0.029* 0.031 0.046
............................................................................................................................................................................................................................................................................................................................................
Basin –40.8 –0.012 –0.013 0.006
............................................................................................................................................................................................................................................................................................................................................
*Significant trend;P<0.05.RESEARCH | REPORTS