ArticlePDF Available

H. Epstein, U. Bhatt, M. Raynolds, D. Walker, J. Pinzon, C. J. Tucker, B. C. Forbes, T. Horstkotte, M. Macias-Fauria, A. Martin, G. Phoenix, J. Bjerke, H. Tømmervik, P. Fauchald, H. Vickers, R. Myneni, T. Park, and C. Dickerson 2018. Tundra greenness. Bulletin of the American Meteorological Society (BAMS) 99, 8: S165-S169.



Content may be subject to copyright.
5. THE ARCTICJ. Richter-Menge, M. O. Jeffries, and
E. Osborne, Eds.
a. IntroductionE. Osborne, J. Richter-Menge, and M. O. Jeffries
Annual average Arctic air temperatures (above
60°N) in 2017 continued to increase at twice the
rate of the rest of the world, with the annual average
surface air temperature second highest (2016 ranked
first) since the year 1900. Extreme warm conditions
were particularly prevalent in Alaska at the end of
2017 when the atmospheric circulation drove warm
southern air masses into the Pacific Arctic region.
The same wind pattern, along with high sea sur-
face temperatures, slowed the southward advance
of the sea ice edge, leading to a month-long delay
in autumn freeze up in the Chukchi Sea and Bering
Strait regions of the Pacific Arctic, setting another
new record for the satellite era (1978–present). Ear-
lier in the year, on 7 March, the Arctic sea ice winter
maximum extent measured by satellite was the lowest
on record (since 1979), covering 8% less area than the
1981–2010 mean. The 2017 sea ice minimum on 13
September was the eighth lowest on record and cov-
ered 25% less area than the long-term mean. Ten of
the lowest September sea ice minimum extents have
been recorded in the last eleven years. Continued loss
of thick multiyear ice (evidenced by <1% multiyear
ice present in March 2017 relative to 16% in 1985)
also contributes a positive feedback to ice loss, as the
majority of today’s sea ice is thin first-year ice prone
to breakup and melt.
As summer sea ice extents continue to shrink
back, seasonal buildup of upper ocean heat in ice-free
regions is increasing. In August 2017, sea surface tem-
perature (SST) records were broken for the Chukchi
Sea, with some regions as warm as +11°C, or 3° to
4°C warmer than the long-term mean (1982–pres-
ent). Most other boundary regions and marginal seas,
which are typically ice free during summer months,
also had anomalously warm SSTs in 2017. As in winter
2016/17, the delayed freeze up in the Pacific Arctic in
late 2017 extended the exposure of the upper ocean
in the Chukchi Sea to the sun’s heat. Mean SSTs from
1982–present show statistically significant warming
trends over much of the Arctic.
After a rapid start to the Greenland ice sheet melt
season in early April, moderate to below-average melt
persisted for much of the remainder of the season.
As a result, summertime area-averaged albedo for
the entire Greenland ice sheet was the third high-
est value since 2000. Glaciers and ice caps outside
of Greenland continue to show declining trends
in cumulative mass balance. Long-term terrestrial
snow cover estimates show dramatic declines in the
Arctic since 2005. In 2017, snow cover extent was
again below the 1981–2010 average across the North
American Arctic, driven by earlier snow melt across
the Canadian Arctic.
Terrestrial permafrost, a critical component of
the Arctic landscape, supports much of the built in-
frastructure in the region (e.g., buildings, highways,
airstrips, pipelines) and continues to experience no-
table change. Climate variables, such as atmospheric
temperature, rain events, and snow depths, are driv-
ing higher permafrost temperature and increasing
active layer thickness (surface soil layer that thaws
and refreezes seasonally). In 2017, five of six per-
mafrost observatories on the North Slope of Alaska
reported record warm permafrost temperatures. In
the same region, tundra greening, or an increase in
above-ground vegetation, has been linked to changes
in the permafrost active layer thickness, the warming
Arctic climate, the extended growing season, and
even reductions in sea ice cover. Over the 35-year
observational time series, tundra greenness has in-
creased throughout the majority of the circumpolar
Arctic, most notably on the North Slope of Alaska,
Canadian low Arctic tundra, and eastern Siberia.
Another phenomenon, tundra browning, is emerging
in the relatively sparse regions of western Alaska, the
Canadia n Archipelago high Arctic, and northwestern
Siberia and may be attributed to winter warming
events and perhaps even insect outbreaks. The Arctic
tundra is also impacted by wildland fires, which are
increasing as a result of warming climate conditions.
While 2017 was an average wildfire season in Alaska
(652 904 acres burned), significantly warmer and
drier conditions in the Upper Yukon zone of north-
east Alaska resulted in high fire danger for much of
the season and accounted for more than half of the
acres burned in the United States.
High above the Arctic, atmospheric ozone con-
centrations in winter 2016/17 were unremarkable
and well above previous record minima in 2010/11
and 2015/16. UV radiation, which depends on at-
mospheric ozone concentrations and other factors,
varied in time and space across the Arctic.
While observational time series are central to
monitoring Arctic change, paleoclimate reconstruc-
tions based on fossil records can help scientists place
the rates and magnitudes of modern change into a
long-term, geological context. Arctic paleoceano-
graphic records indicate that the magnitude and
sustained rate of declining sea ice trend observed
in the modern era is unprecedented in any existing
high resolution paleoclimate sea ice reconstruction
for at least the last 1450 years. Similarly, according
to paleoclimate studies, today’s abnormally warm
Arctic air and sea surface temperatures have not been
observed in the last 2000 years. Indigenous knowl-
edge gathered by Arctic Peoples over many millennia
is another means to holistically understand Arctic
change beyond instrumental records. Coproduction
of knowledge can bring together knowledge systems
of scientists and indigenous knowledge–holders to de-
velop suitable sustainability and adaptation practices
to address issues arising from the changing Arctic
system (see Sidebar 5.2).
b. Surface air temperature—J. Overland, E. Hanna,
I. Hanssen-Bauer, S.-J. Kim, J. E. Walsh, M. Wang, U. S. Bhatt,
and R. L. Thoman
Arctic surface air temperature is an indicator of
both regional and global climate change. Although
natural variability contributes to year-to-year and re-
gional differences in air temperature, the magnitude
of the long-term temperature trend across the entire
Arctic is an indicator of global climate change and the
impact of increasing greenhouse gas concentrations
(Overland 2009; Notz and Stroeve 2016).
After a warm Arctic-wide autumn 2016, early
2017 had notable short-term, regional temperature
anomalies in response to a highly variable jet stream.
Spring and summer 2017 had near-average air tem-
peratures relative to the 1981–2010 climatology. The
spring and summer conditions were reminiscent of
those occurring before the long-term, above-average
temperature increases began in the 1990s. Rather
than higher sea level pressure extending over much
of the Arctic, as observed in many recent years, weak
low pressures were seen in 2017—a return to a wind
forcing typical from a decade ago. The atmospheric
forcing in spring and summer 2017 is consistent with
a year when some Arctic indicators ran counter to the
recent long-term trend over the previous decade. For
example, Eurasian spring snow extent was above aver-
age for the first time in over a decade (see Section 5i).
At +1.6°C, the mean annual 2017 surface air
temperature (SAT) anomaly for land stations north
of 60°N is the second highest value (after 2016) in
the record starting in 1900 (Fig. 5.1). Despite near-
average temperatures during spring and summer
months, extreme heat during autumn and winter,
particularly over the Chukchi Sea and extending
northward to the pole, contributed to near-record
breaking warm conditions in 2017 (Fig. 5.2). Cur-
rently, the Arctic is warming at more than twice the
rate of lower latitudes.
The greater rate of Arctic temperature increase,
compared to the global increase, is referred to as
Arctic amplification. Mechanisms for Arctic am-
plification include: reduced summer albedo due to
losses of sea ice and snow cover; the increase of total
water vapor content in the Arctic atmosphere; a sum-
mer decrease and winter increase in total cloudiness
(Makshtas et al. 2011); the additional heat generated
by newly sea ice–free ocean areas that are maintained
later into the autumn (Serreze and Barry 2011); and
the lower rate of heat loss to space in the Arctic,
Fig . 5.1. Arctic (land stations north of 60°N) and
global mean annual land surface air temperature (SAT)
anomalies (°C, 1981–2010 base period) for 1900–2017.
Note that there were few stations in the Arctic, par-
ticularly in northern Canada, before 1940. (Source:
CRUTEM4 dataset.)
Fig. 5.2. Seasonal anomaly patterns for near-surface air
temperatures (°C, 1981–2010 base period) for 2017 in
(a) JFM, (b) AMJ, (c) JAS, and (d) OND. Temperatures
are from slightly above the surface layer (925 mb) to
emphasize large spatial patterns rather than local
features. (Source: NOAA /ESRL.)
Fig. 5.3. Arctic Mar 2017 air temperature anomalies
Fig. 5.4. Arctic mean sea level pressure field (hPa) for
su mme r 2 017.
relative to the subtropics, due to lower mean surface
temperatures in the Arctic (Pithan and Mauritsen
2014). Recent reductions in air pollution in Europe
are reducing the relative rate of Arctic warming due
to decreased downward longwave radiation, coun-
tering other mechanisms that contribute to Arctic
amplification (Acosta Navarro et al. 2016).
Seasonal air temperature variations in 2017 are
divided into winter (January–March, JFM), spring
(April–June, AMJ), summer (July–September, JAS),
and autumn (October–December, OND; Fig. 5.2).
These seasonal SAT divisions are chosen to coincide
with the seasonal cycles of key Arctic variables. For
example, the summer sea ice minimum occurs in
September, and autumn cooling continues through
On a seasonal basis, winter was unremarkable
in terms of major features (Fig. 5.2a). However,
there were notable short-term, regional temperature
anomalies in response to highly variable jet stream
shapes. For instance, Iceland experienced a record
high maximum temperature of 19.1°C in February,
exceeding the previous February (1998) record of
18.1°C by a full degree (Trausti Jonsson, Icelandic Met
Office, 2017, personal communication). March 2017
had major warmth across Siberia (Fig. 5.3) including
eastern Asia.
Spring showed some positive temperature anoma-
lies in the East Siberian Sea (Fig. 5.2b), a continuation
of a warm feature observed in March. This regional
warming supported early sea ice loss in the Chukchi
Sea (see Section 5d). May saw anomalous high pres-
sure extend between Greenland and Norway, with
relatively warm but unexceptional temperatures over
Similar to summer 2016, neutral temperature
anomalies occurred across the central Arctic in
summer 2017 (Fig. 5.2c), in contrast to the warm
conditions observed during much of the previous
decade. The summer 2017 conditions did not support
continued overall extreme summer sea ice loss (see
Section 5d). Mean coastal Greenland temperatures
were near climatological levels, in contrast to some
summers in the recent decade.
Alaska/northwestern Canada was the only region
with above-average summer surface air temperatures.
Several locations in the interior of Alaska had the
warmest calendar month of record in July. On a more
local and short-term basis, many stations in the north
and east of Iceland reported record high temperatures
for September.
Summer sea level pressure was characterized by
negative anomalies in the central Arctic (Fig. 5.4).
This pattern prevented extra heat in the midlatitudes
from penetrating into the central Arctic. Such added
heat from outside the Arctic is associated with low
sea ice summers (Parkinson and Comiso 2013). This
sea level pressure pattern was accompanied by cloud
cover that limited the solar heating of the lower at-
mosphere in the central Arctic.
A broad swath of extreme warm temperature
anomalies (> +4°C) stretched across the central Arc-
tic in autumn (Fig. 5.2d). The warmest temperature
extremes, north of the Bering Strait and north of
Svalbard, were due to heat stored in the upper Arctic
Ocean (see Section 5c) and to advection of warm
air from the south (generated from the Pacific and
Atlantic Oceans).
December 2017 had extreme warm temperatures
in Alaska and cold temperatures in the central and
eastern U.S., with incidences of snow as far south
as Mississippi (Fig. 5.5a). This temperature pattern
is associated with large north–south meanders of
the tropospheric jet stream (Fig. 5.5b). Because the
extratropical mid-troposphere wind direction ap-
proximately follows the contour direction of geopo-
tential heights, Fig. 5.5b shows warm winds from the
southwest extending into Alaska and cold air moving
southeast from Canada in December. Warm air is less
dense and supports rising geopotential heights. Thus,
warm temperatures over Alaska can help maintain
the persistence of this North American weather pat-
tern. Contributing to the relatively warm tempera-
tures in Alaska in autumn was the delayed freeze-up
of sea ice in Alaskan waters. Freeze-up lasted well
into December and set a new record for the satellite
era beginning in 1978 (see Section 5d).
c. Sea surface temperature—M.-L. Timmermans, C. Ladd, and
K. Wood
Summer sea surface temperatures (SST) in the
Arctic Ocean are determined mainly by absorption
of solar radiation into the surface layer. In the Barents
and Chukchi Seas, there is an additional contribu-
tion from advection of warm water from the North
Atlantic and North Pacific Oceans, respectively. Solar
warming of the ocean surface layer is inf luenced by
the distribution of sea ice (with more solar warming
in ice-free regions), cloud cover, water color, and
upper-ocean stratification. River inf luxes influence
the latter two, as well as provide an additional source
of warm water. SSTs are an essential indicator of the
role of the ice–albedo feedback mechanism in any
given melt season; as the area of sea ice cover de-
creases, more incoming solar radiation is absorbed
by the ocean and the warmer ocean in turn melts
more sea ice.
SST data presented here are from the NOAA Opti-
mum Interpolation (OI) SST Version 2 product (OIS-
STv2), which is a blend of in situ and satellite measure-
ments (Reynolds et al. 2002, 2007). Compared to in
situ temperature measurements, the OISSTv2 product
showed average correlations of about 80%, with an
overall cold SST bias of −0.02°C (Stroh et al. 2015).
August SSTs provide the most appropriate repre-
sentation of Arctic Ocean summer SSTs because they
are not affected by the cooling and subsequent sea
ice growth that typically takes place in the latter half
of September. Mean SSTs in August 2017 in ice-free
regions ranged from ~0°C in some regions to as high
as 11°C in the Chukchi and Barents Seas (Fig. 5.6a).
Compared to the 1982–2010 August mean (note the
monthly SST record begins in December 1981), most
boundary regions and marginal seas had anomalously
high SSTs in August 2017 (Fig. 5.6b). Particularly
high anomalies (around 3°–4°C above the 1982–2010
average) were observed in the Beaufort, Chukchi,
and southern Barents Seas. SSTs in the boundary
regions and marginal seas, which are mostly ice free
in August, are linked to the timing of local sea ice
retreat, which facilitates the direct solar heating of
the exposed surface waters.
Fig . 5.5. Dec 2017 fields show the cause of warm
temperatures in Alaska and simultaneous cold
temperatures in the central and southern U.S. (a)
925-hPa air temperature anomalies (°C) and (b)
corresponding 500-hPa geopotential height field (m),
showing the strong wave tropospheric jet stream
pattern extending north into Alaska and south into
eastern North America.
In August, regions off the west and east coasts of
Greenland and in the southern Barents Sea were mark-
edly cooler (by up to 3°C) than in August 2016 (see
Timmermans 2017). It is notable also that compared
to August 2012 (the summer of lowest minimum sea
ice extent in the satellite record, 1979–present), Au-
gust 2017 SSTs in the Chukchi Sea region were up to
3°C higher (Fig. 5.6c). This illustrates the significant
interannual and spatial variability in summer SSTs.
Cooler SSTs in August 2012 (compared to August 2017)
in the Chukchi Sea were related to the persistence of
sea ice in that particular region (even while the main
ice pack retreated) and a strong cyclonic storm in the
region that brought cool conditions late in the summer
season (see Timmermans et al. 2013).
Mean August SSTs from 1982 to 2017 show statis-
tically significant linear warming trends over much
of the Arctic Ocean (Fig. 5.6d); the cooling trends
in the Laptev and northern Barents Seas are notable
exceptions. Warming trends coincide with declining
trends in summer sea ice extent (including late-season
freeze-up and early melt, e.g., Parkinson 2014; see sec-
tion 5d), increased solar absorption (e.g., Pinker et al.
2014), release of stored ocean heat (e.g., Timmermans
2015), and milder air temperatures (see Section 5b).
Mean August SSTs for the entire Chukchi Sea region
exhibit a statistically significant warming trend of
about +0.7°C decade−1, based on a linear fit.
d. Sea ice coverD. Perovich, W. Meier, M. Tschudi, S. Farrell,
S. Hendricks, S. Gerland, C. Haas, T. Krumpen, C. Polashenski,
R. Ricker, and M. Webster
1) Sea ice extent
Arctic sea ice (1) acts as a barrier between the
underlying ocean and the atmosphere, (2) limits
the amount of absorbed solar energy due to its high
albedo, (3) provides a habitat for biological activity,
and (4) limits human access to the Arctic Ocean and
adjacent seas. The extent of the Arctic sea ice cover
varies substantially over the course of a year, with the
end-of-winter ice cover generally two to three times as
large as at the end of summer. The months of March
and September are of particular interest because they
are the months when the sea ice typically reaches its
maxi mum and minimum extents, respectively. Figure
5.7 shows the monthly average Arctic sea ice extents
in March 2017 and September 2017, measured by
satellite-based passive microwave instruments.
Sea ice extent is the total area covered by at least
15% concentration of sea ice. Based on data from the
Fig. 5.6. (a) Mean SST (°C) in Aug 2017. White shad-
ing is the Aug 2017 mean sea ice extent (shown in
each panel) and gray contours indicate the 10°C SST
isotherm. (b) SST anomalies (°C) in Aug 2017 relative
to the Aug 1982–2010 mean (dotted black contour in-
dicates zero anomaly). Black line indicates the median
ice edge for Aug 1982–2010. (c) SST anomalies (°C) in
Aug 2017 relative to Aug 2012. Black line indicates the
median ice edge for Aug 2012. (d) Linear SST trend
(°C yr1) for Aug of each year from 1982–2017. Trend
is only shown for values that are significant at the 95%
confidence interval; the region is gray otherwise. Black
line indicates the median ice edge for Aug 1982–2010.
(Sources: SST data are from the NOAA OISSTv2; sea
ice extent and ice-edge data are from NSIDC Sea Ice
Index, Version 3, Fetterer et al. 2017.)
Fig. 5.7. Average monthly sea ice extent in (a) Mar
(left) and (b) Sep (right) 2017 illustrate the respective
winter maximum and summer minimum extents. The
magenta line indicates the median ice extents in Mar
and Sep, respectively, for the period 1981–2010. Maps
are from NSIDC at
(Fetterer et al. 2017).
National Snow and Ice Data Center (NSIDC) sea ice
index (Fetterer et al. 2017), the sea ice cover reached
a maximum extent of 14.42 million km2 on 7 March,
which was 8% below the 1981–2010 average. This
is the lowest maximum value ever observed in the
satellite record.
On 13 September, the sea ice extent reached a sum-
mer minimum value of 4.64 million km2. This is the
eighth lowest extent in the satellite record. While the
2017 minimum extent represents a modest increase
from the 2016 minimum, it was 25% less than the
1981–2010 average minimum ice extent. The 10 lowest
September extents have occurred in the last 11 years
(Parkinson and DiGirolamo 2016).
In 2017, sea ice extent showed decreasing trends
in all months and virtually all regions, except for
the Bering Sea during winter (Meier et al. 2014).
The September (typical Arctic sea ice minimum)
monthly average trend for the entire Arctic Ocean
is now −13.2% decade−1 relative to the 1981–2010
average (Fig. 5.8). Trends are smaller during March
(typical Arctic sea ice maximum), at −2.7% decade−1,
but the decrease is statistically significant. Both the
September and March trends are significant at the
99% confidence level.
Freeze-up in the Chukchi Sea was extremely slow,
and the sea ice extent in the region at the beginning
of December 2017 was the lowest in the satellite
record. It was not until the end of December that
the region was substantially frozen over, a month
later than normal. Upper ocean heat accumulated
during the summer, through the absorption of solar
radiation, likely slowed ice growth in the Chukchi
region (see Section 5c). Anomalous southerly winds
during October–December also played a significant
role by advecting warm air and ocean waters into the
region through the Bering Strait (see Section 5b) and
preventing southward advancement of the ice edge.
2) age of the ice
The age of sea ice is another key descriptor of
the state of the sea ice cover. Compared to younger
ice, older ice tends to be thicker, stronger, and more
resilient to changes in atmospheric and oceanic forc-
ing (i.e., changes in atmospheric circulation patterns
and ocean heat). The age of the ice is measured us-
ing satellite observations and drifting buoy records
to track ice parcels over several years (Tschudi et al.
2010; Maslanik et al. 2011). This method has been
used to provide a record of the age of the ice since
1985 (Tschudi et al. 2015, 2016).
Very old ice (>4 years old) continues to be a dimin-
ishingly small fraction of the Arctic ice pack in March
(Fig. 5.9). The extent of the oldest ice has declined
from 2.54 million km2 in March 1985 (representing
16% of the total ice pack) to 0.13 million km2 in March
2017 (0.9% of the total ice pack). The distribution of
ice age in March 2017 was similar to that of March
2016, although there was a decrease in the fractional
coverage of the oldest ice, from 1.2% in March 2016
to 0.9% in March 2017. First-year ice dominates the
winter sea ice cover, comprising ~79% of the ice cover
Fig. 5.8. Time series of sea ice extent anomalies (%) in
Mar (the month of maximum ice extent) and Sep (the
month of minimum ice extent). Anomaly value for each
year is the percent difference in ice extent relative to
the 1981–2010 mean. The black and red dashed lines
are least squares linear regression lines.
Fig. 5.9. (a) Arctic sea ice age coverage by year, ex-
pressed as the fraction of the total ice area, 1985–2017.
Sea ice age coverage maps for (b) Mar 1985 and (c)
Ma r 2017.
in March 2017, compared to ~55% in the 1980s. The
thinner, younger ice is more mobile and susceptible
to mechanical wind forcing, and it is vulnerable to
complete melting in the summer and contributes to
the observed decrease in summer sea ice extents by
enabling more heat to be absorbed by the upper ocean.
3) Sea ice thickneSS and Snow depth
Satellite remote sensing and regular airborne sur-
vey programs continued to record changes in Arctic
sea ice thick ness and volume. These survey programs
derive ice thick ness and volume by observing the free-
board of the ice cover, which is the distance between
the surface of the ocean and the top of the ice. During
this past year the ESA CryoSat-2 radar altimeter mis-
sion completed its seventh year of operation, provid-
ing sea ice thickness estimates between October and
April (Laxon et al. 2013). The CryoSat-2 freeboard
measurements expand the data record of satellite
and submarine-based observations that document
the decline in sea ice thickness since 1958 (Kwok and
Rothrock 2009; Lindsay and Schweiger 2015).
In spring 2017, CryoSat-2 products from the Al-
fred Wegener Institute indicated a spatially variable
pattern of ice thickness (Fig. 5.10a), which is typical.
The April 2017 thickness anomaly, compared to the
period 2011–16 (Fig. 5.10b), shows below-average
thicknesses in the multiyear ice region north of the
Queen Elizabeth Islands of the Canadian Arctic Ar-
chipelago, the Chukchi Sea, and the shelf regions of
the East Siberian Sea. Above-average thicknesses were
observed in the Beaufort Sea and the eastern part of
the central Arctic Ocean.
Sea ice volume estimates were generated from
Cryosat-2 observations for 2011–17 for the months of
October through April. Results for the central Arctic
Ocean show a decline from 2011 to 2013, an increase
in 2014, followed by a steady decline from 2014 to
2017. The April 2017 sea ice volume (13.19 ± 1.15 ×
103 km3) ranks as the third lowest spring volume after
April 2012 (13.14 ± 1.27 × 103 km3) and 2013 (12.56 ±
1.21 × 103 km3) in the CryoSat-2 data record (2011–17).
The difference between the three lowest volume es-
timates lies within the observational uncertainties
of the instrument. For more information regarding
instrument uncertainty see Ricker et al. (2014).
Fig. 5.10. Apr 2017 (a) sea ice thickness (m) derived from CryoSat-2 radar altimeter data and (b) sea ice
thickness anomaly (m; base period 2011–16). (c) Snow depth (m) on Arctic sea ice at the end of winter,
prior to melt onset; recent in situ measurements (stars), made in 2015 and 2017, and airborne observa-
tions (multiple airborne survey lines), made in Mar and May in 2009–12 and 2014–15, are overlaid on
the long term mean snow depth for the months of Mar and Apr (adapted from Warren et al. 1999).
Black line and arrows in (c) designate the western Arctic.
At present , the Arctic Ocean is experiencing changes
in ocean surface temperature and sea ice extent that are
unprecedented in the era of satellite observations, which
extend from the 1980s to the present (see Sections 5c,d).
To provide context for current changes, scientists turn
to paleoclimate records to document and study anthro-
pogenic inuence and natural decadal and multidecadal
climate variability in the Arctic system. Paleoceanographic
records extend limited Arctic instrumental measurements
back in time and are central to improving our understand-
ing of climate dynamics and the predictive capability of
climate models. By comparing paleoceanographic records
with modern observations, scientists can place the rates
and magnitudes of modern Arctic change in the context
of those inferred from the geological record.
Over geological time, paleoceanographic reconstruc-
tions using, for instance, marine sediment cores indicate
that the Arctic has experienced huge sea ice uctuations.
These uctuations range from nearly completely ice-free
to totally ice-covered conditions. The appearance of ice-
Fig. SB5.1. The oldest known paleoclimate evidence of sea ice in the Arctic are (a) fossilized remains of sea ice
dwelling diatoms (Synedropsis spp.) and (b) ice rafted debris that date back to 47 million years ago (Stickley
et al. 2009). (c) Global compilation of paleoclimate records indicates that cooling ocean temperatures (°C) and
declining atmospheric CO2 (ppm) coincide with major NH sea ice development (data: Beerling and Royer 2011;
Zhang et al. 2013; Anagnostou et al. 2016). Global ocean temperature anomalies are determined by millions of
stable oxygen isotopic measurements of fossilized calcite benthic foraminifera shells. Arrows indicate cooling
temperature and declining CO2 concentrations through the greenhouse to icehouse transition. Red and orange
“+” on the right y-axis indicate the CMIP5 multimodel mean projected temperature and atmospheric CO2,
respectively, in the year 2050 and 2100.
rafted debris and sea ice-dependent diatoms in Arctic marine
sediments indicate that the rst Arctic sea ice formed approxi-
mately 47 million years ago (St. John 2008; Stickley et al. 2009;
Fig. SB5.1), coincident with an interval of declining atmospheric
carbon dioxide (CO2) concentration, global climate cooling,
and expansion of Ear th’s cryosphere during the middle Eocene.
The development of year-round (i.e., perennial) sea ice in the
central Arctic Ocean, similar to conditions that exist today, is
evident in sediment records as early as 14–18 million years ago
(Darby 2008). These records suggest that transitions in sea ice
cover occur over many millennia and often vary in concert with
the waxing and waning of circum-Arctic land ice sheets, ice
shelves, and long-term uctuations in ocean and atmosphere
temperatures and atmospheric CO2 concentrations (Stein et
al. 2012; Jakobsson et al. 2014).
Over shorter time scales, shallow sediment records from
Arctic Ocean continental shelves allow more detailed, higher-
resolution (hundreds of years resolution) reconstructions
of sea ice history extending through the Holocene (11 700
years ago to present), the most recent interglacial period.
Fig. SB5.2. (a) Atmospheric CO2 concentrations (ppm),
(b) paleoclimate reconstructions of summer Arctic sea
ice extent (km2; Kinnard et al. 2011), and (c) annual
atmospheric temperature anomalies (°C; McKay and
Kaufman 2014) and sea surface temperature anoma-
lies (°C; Spielhagen et al. 2011) spanning the last 1500
years. Atmospheric (red solid line: 5-yr mean and light
gray: annual mean) and upper-ocean (dark gray with
circles indicating individual data points) temperature
anomalies are plotted together to show respective
temperature trends. Vertical dashed line indicates the
onset of the Industrial Revolution. Atmospheric CO2
concentrations [shown in (a)] are from the Law Dome
ice core record (Etheridge et al. 1996, 1998) and mod-
ern observations from the Mauna Loa observatory [Dr.
Pieter Tans, NOAA/ESRL (
/ccgg/trends/ ), and Dr. Ralph Keeling, Scripps Institu-
tion of Oceanography (].
A notable feature of these records is an early Holocene sea
ice minimum, corresponding to a thermal maximum (warm)
period from 11 000 to 5000 years ago, when the Arctic may
have been warmer and had less summertime sea ice than
today (Kaufman et al. 2004). However, it is not clear that the
Arctic was ice-free at any point during the Holocene (Polyak
et al. 2010). High-resolution paleo–sea ice records from the
western Arctic in the Chukchi and East Siberian Seas indicate
that sea ice concentrations increased through the Holocene
in concert with decreasing summer solar insolation (sunlight).
Sea ice extent in this region also varied in response to the
volume of Pacic water delivered via the Bering Strait into the
Arctic Basin (Stein et al. 2017; Polyak et al. 2016). Records from
the Fram Strait (Müller et al. 2012), Laptev Sea (Hörner et al.
2016), and Canadian Arctic Archipelago (Vare et al. 2009) also
indicate a similar long-term expansion of sea ice and suggest sea
ice extent in these regions is modulated by the varying inux
of warm Atlantic water into the Arctic Basin (e.g., Werner et
al. 2013). Taken together, available records support a circum-
Arctic sea ice expansion during the late Holocene.
A notably high-resolution summer sea ice history (<5-year
resolution) has been established for the last 1450 years using
a network of terrestrial records (tree ring , lake sediment,
and ice core records) located around the margins of the
Arctic Ocean (Kinnard et al. 2011). Results summarized in
Fig. SB5.2 indicate a pronounced decline in summer sea ice
extent beginning in the 20th century, with exceptionally low
ice extent recorded since the mid-1990s, consistent with the
satellite record (see Section 5d). While several episodes of
reduced and expanded sea ice extent occur in association with
climate anomalies such as the Medieval Climate Warm Period
(AD 800–1300) and the Little Ice Age (AD 1450–1850), the
magnitude and pace of the modern decline in sea ice is outside
of the range of natural variability and unprecedented in the
1450-year reconstruction (Kinnard et al. 2011). A radiocar-
bon-dated drif twood record of the Ellesmere ice shelf in the
Canadian High Arctic, the oldest landfast ice in the Northern
Hemisphere, also demonstrates a substantial reduction in ice
extents over the 20th century (England et al. 2017 ). A suppor t-
ing sediment record indicates that inowing Atlantic water in
Fram Strait has warmed by 2°C since 1900, driving break up
and melt of sea ice (Spielhagen et al. 2011). Complementary
mooring and satellite observations show the “Atlantication”
of the eastern Arctic due to enhanced inow of warm saline
water through Fram Strait (Nilsen et al. 2016) and nutrient-rich
Pacic water via the Bering has increased by more than 50%
(Woodgate et al. 2012), further driving sea ice melt and warm-
ing seas. Similar high-resolution proxy records from Arctic
regions also indicate that the modern rate of increasing annual
surface air temperatures has not been observed over at least
the last 2000 years (McKay and Kaufman 2014). Scientists con-
clude that broad-scale sea ice variations recorded in the paleo
Snow plays several critical roles in the growth and
melt of Arctic sea ice. These roles include insulating
the ocean from the atmosphere, dampening heat
fluxes, reducing sea ice growth, ref lecting more than
80% of the incoming sunlight, delaying ice melt, and
contributing to melt pond formation (Granskog et al.
2017). Prior to the 1990s, observations of snow on
Arctic sea ice were limited to in situ measurements.
Warren et al. (1999) compiled many of these obser-
vations into a long-term record. New approaches to
measure snow depth have since emerged, including
improved instruments for in situ and autonomous
observations and remote sensing. Field observations
from recent years underscore significant regional
and interannual variability in snow on Arctic sea ice.
Figure 5.10c shows the historical snow depth record,
plus a compilation of airborne snow depth measure-
ments collected between March and May in 2009–12
and 2014–15, and in situ measurements made in 2015
and 2017. The recent mean snow depths range from
0.05 to 0.55 m. Compared to the record published by
Warren et al. (1999) there has been an overall decrease
in snow depths of 37% ± 29% in most of the western
Arctic (aka North American Arctic) at the end of
winter (Fig. 5.10c). The trend in the mean anomalies is
−0.27 cm yr−1 with 99% signif icance. This decrease is
potentially associated with later sea ice formation and
thus later onset of snow accumulation in the previous
autumn (Webster et al. 2014; Kurtz and Farrell 2011;
Blanchard-Wrigglesworth et al. 2015).
e. Greenland ice sheet—M. Tedesco, J. E. Box, J. Cappelen,
R. S. Fausto, X. Fettweis, K. Hansen, M. S. Khan, S. Luthcke,
T. Mote, I. Sasgen, C. J. P. P. Smeets, D. van As,
R. S. W. van de Wal, and I. Velicogna
The Greenland ice sheet (GrIS) plays a crucial role
in the climatological, hydrological, and ecological
cycles at regional and global scales. The high albedo
of the ice sheet contributes to a modulation of the
amount of solar energy absorbed by Earth, and the
location and topography of the ice sheet affects atmo-
spheric circulation. The GrIS is also a major contribu-
tor to current and projected sea level rise, through
surface runoff and iceberg calving. The summer of
2017 over the Greenland ice sheet was characterized
by below-average (1981–2010) melt extent and above-
average surface albedo, with the net ablation being
below the 2008–17 average at many test sites but still
above the average for the 1961–90 reference period
when the ice sheet was in steady equilibrium. Overall,
total mass loss in 2017 was close to the average of the
1) Surface melting
Estimates of melt extent across the GrIS are
obtained from brightness temperatures measured
by the Special Sensor Microwave Imager/Sounder
(SSMIS) passive microwave radiometer (e.g., Mote
2007; Tedesco et al. 2013). These estimates point to
a rapid start of the melting season in 2017, similar to
2016, with melt extent in early April reaching an area
once typical of early June (Fig. 5.11a). From mid-June
through mid-July 2017, however, melt extent was
persistently below the 1981–2010 average. The spatial
extent of melt for summer 2017 (June–August, JJA)
was above average on 15 of 93 days (16%) and reached
its maximum extent of 32.9% of the ice sheet area
on 26 July. The maximum extent of surface melt in
2017 was lower than the average maximum extent
of 39.8% for the period 1981–2010 and was the low-
est maximum extent since 1996. There was regional
variability in the characteristics of the summer melt.
Most of the western and northeast ice sheet margins
had more days than average with melt (relative to
record were predominantly driven by changes in basin-
scale atmospheric circulation patterns, uctuations in air
temperature and strength of incoming solar radiation,
and changes in the inow of warm water via Pacic and
Atlantic inows (Polyak et al. 2010).
There is general consensus that ice-free Arctic sum-
mers are likely before the end of the 21st century (e.g.,
Stroeve et al. 2007; Massonnet et al. 2012), while some
climate model projections suggest ice-free Arctic summers
as early as 2030 (Wang and Overland 2009). Paleoclimate
studies and observational time series attribute the decline
in sea ice extent and thickness over the last decade to
both enhanced greenhouse warming and natural climate
variability. While understanding the interplay of these fac-
tors is critical for future projections of Arctic sea ice and
ecosystems, most observational time series records cover
only a few decades. This highlights the need for additional
paleoceanographic reconstructions across multiple spatial
and temporal domains to better understand the drivers and
implications of present and future Arctic Ocean change.
1981–2010), while the southeast margin had fewer
days than average. The magnitude and evolution of
surface melt in 2017 were consistent with the state
of the dominant atmospheric circulation pattern, as
defined by the Arctic Oscillation and North Atlantic
Oscillation, both of which were strongly positive
(Tedesco et al. 2017).
2) Surface maSS ba lance
Consistent with the low-to-moderate surface melt-
ing described above, the August 2016–August 2017
surface mass balance (SMB) year along the K-transect
at 67°N in west Greenland (Fig. 5.11b; van de Wal et
al. 2012) was characterized by moderate mass loss
over the ablation region (Tedesco et al. 2017). The
SMB along the transect line, which has been continu-
ously monitored for 28 years, was approximately one
standard deviation below the 1990–2017 mean. The
equilibrium line altitude (defined as the elevation at
which mass losses balance mass gain, i.e., SMB = 0)
in 2017 was around 1490 meters, which is 40 m below
the 28-year mean. The mass balance gradient was 3.4
mm w.e. (water equivalent) m−1 yr−1, which is about 6%
lower than the average (Tedesco et al. 2017).
Due to the relativelylow summer temperatures,
net ice ablation averaged over the PROMICE sites
(Fig. 5.11b), distributed around Greenland in the
ablation zone, was about 20% (or 0.6 standard devia-
tions)lower in 2017 than compared to the 2008–17
average. The largest ablationanomaly values, more
than one standard deviation below average, occurred
at the southwest and northwest margins. The largest
absolute ablation of 5.5 m of ice was measured at
the southern tip of the ice sheet. More details can
be found in Tedesco et al. (2017). While the surface
mass balance observations indicate that surface melt
was relatively moderate in 2017, compared to that
observed in the last decade, it was still higher than
observed prior to 1990. When referencing the values
to the 1961–90 climatological standard period (Van
As et al. 2016), all eight low-elevation PROMICE
station sites experienced above-average ablation
anomalies in 2017 (Fig. 5.11b). However, only three
stations were beyond the estimated uncertainty:
KPC_L (+96% ± 49%), SCO_L (+15% ± 14%) and
KAN_L (+48% ± 35%).
3) albedo
The area-averaged albedo (the fraction of incident
solar radiation reflected by a surface) for the entire
Greenland ice sheet for summer 2017 was 80.9%, as
determined using data from the Moderate Resolu-
tion Imaging Spectroradiometer (MODIS; after
Box et al. 2017; Fig. 5.11c). This is the third highest
summer albedo value, after those of 2000 and 2013,
during the 2000–17 MODIS period. Positive albedo
anomalies are consistent with reduced melting in 2017
and snowfall events during the summer. The highest
2017 summer albedo anomalies occurred along the
western margins of the ice sheet (Tedesco et al. 2017).
Fig. 5.11. (a) Spatial extent of melt, derived from the
satellite product, as a percentage of the ice sheet area
during 2017 (red line) and the 1981–2010 mean spatial
extent of melt (dashed blue line). Light and dark gray
areas represent the interdecile and interquartile
ranges, respectively. (b) 2017 ablation anomalies (%
of average, relative to 1961–90) at lower PROMICE
(Programme for monitoring of the Greenland ice sheet
weather station sites in the Greenland ice sheet) abla-
tion area, using historical coastal temperature records.
(c) Distribution of albedo anomalies (%, 2000 –09 refer-
ence period) for summer 2017, derived from MODIS .
Area within the rectangle in (c) indicates the location
of the K-transect.
4) total maSS balance
GRACE satel lite gravity estimates obtained follow-
ing Velicogna et al. (2014), Sasgen et al. (2012), and
Luthcke et al. (2013) and available since 2002, indicate
that between April 2016 and April 2017 (the most
recent 12-month period of reliable data) there was a
net ice mass loss of 276 ± 47 Gt (gigatonnes; Fig. 5.12).
This is 144% greater than the April 2015–April 2016
mass loss (191 ± 28 Gt) and close to the average April-
to-April mass loss (255 ± 7 Gt) for 2003–17 (Sasgen
et al. 2012). The updated trends of total ice mass
loss for the 15-year GRACE period are 264 Gt yr−1
(Velicogna et al. 2014) and 270 Gt yr1 (Sasgen et al.
2012); the different values reflect the slightly different
computational approaches adopted in the two stud-
ies. The GRACE mission came to an expected end in
October 2017. No further data will be available from
this important source. It is anticipated that gravity
measurements from space will resume and ice mass
estimates will be possible again when the GRACE
follow-on mission is launched. At the time of writing,
the expected launch window is in spring 2018.
5) marine-terminating glacie rS
Marine-terminating glaciers are the outlets by
which the Greenland ice sheet discharges ice mass
to the ocean. Glacier area measurements from
LANDSAT and ASTER, available since 1999 (Box
and Hansen 2015) for 45 of the widest and fastest-
flowing marine-terminating glaciers, reveal a pattern
of continued relative stability since 2012/13 (Fig. 5.13).
Among the surveyed glaciers, 22 retreated, 10 were
stable, and 13 advanced. Overall, the annual net area
change of the 45 glaciers at the end of the 2017 melt
season, which started in June a nd ended in September,
was −102.8 km2. This is ~80% of the 18-year survey
period average of −126.6 km2 year−1. The largest area
losses were in eastern Greenland, where the Helheim
and Kangerdlugssauq glaciers lost, respectively,
11.6 k m2 and 9.9 km2 in area. The largest advance was
observed at Petermann glacier, northwest Greenland,
where the area increased by 11.5 km2.
6) Surface air temperatureS
Measurements at 20 weather stations of the Danish
Meteorological Institute (Cappelen et al. 2018; Table
5.1) indicate widespread above or near-average air
temperatures for 2017, relative to the period 1981–
2010. The exception was during spring 2017 (March–
May, MAM) in coastal northeast Greenland and the
start of July in western Greenland, when many sites
experienced relatively cool temperatures. Looking
in more detail, during winter 2016/17 (December–
February, DJF) a new seasonal record high was set
in Aputiteeq, located in eastern Greenland. February
in Aputiteeq was particularly warm, with a new
monthly record set. At Kap Morris Jesup, along the
northern coast, the winter season was the second
warmest (only exceeded in 2011), with December 2016
matching the record warmth of December 2009. April
2017 was generally colder than average at many sites,
compared to April 2016 when record breaking high
temperatures were recorded. In autumn (September–
November, SON) some individual months were
record setting at Danmarkshavn, Daneborg, and
Ittoqqortoormiit. At Danmarkshavn, Daneborg, and
Fig. 5.12. Change in the total mass (Gt) of the Green-
land ice sheet between Apr 2002 and Jun 2017, es-
timated from GRACE measurements. (Due to the
decommissioning of the GRACE satellite, no data
are available after Jun 2017.) Data are based on an
unweighted average of JPL RL05, GFZ RL05, and CSR
RL05 solutions, which reduce noise in the GRACE data
for 2017 (Sasgen et al. 2012).
Fig. 5.13. Glacier area change (km2) from LANDSAT
and ASTER imagery available since 1999 for 45 of the
widest and fastest-flowing marine-terminating glaciers
(after Box and Hansen 2015).
Table 5.1. Seasonal and annual surface air temperature anomalies (°C) relative to the 1981–2010 average
at 20 Danish Meteorological Institute weather stations in Greenland, where observations have been
made for a minimum of 30 years. Seasons are winter (DJF 2016/17); spring (MAM 2017); summer (JJA
2017); autumn (SON 2017). Highlighted cell indicates a new seasonal record. The year that observations
began is given, together with the station name, geographic coordinates, and elevation.
Station Name, Start Year;
Latitude, Longitude, Elevation
Pituffik/Thule AFB
194 8;
76.5°N, 68.8°W,
77 m a.s.l.
Anomaly (°C) 1.1 0.5 0.2 0.2 1.4
Max Ye ar 2010 1986 1953 1957 2010
Min Year 1992 1949 1992 19 96 196 4
72 . N, 56.1° W,
126 m a.s.l.
Anomaly (°C) 1. 2 0.7 1.4 0.0 0.7
Max Ye ar 2010 1947 1932 2012 2010
Min Year 1887 1983 1896 1873 1917
68.7°N, 52.8°W,
43 m a.s.l.
Anomaly (°C) 0.9 0.8 0.6 0.3 0.8
Max Ye ar 2010 2010 2016 2 012 2010
Min Year 198 3 198 4 1993 1972 1986
69.2°N, 51.1°W,
29 m a.s.l.
Anomaly (°C) 0.4 0.1 0.1 0.5 0.5
Max Ye ar 2010 1929 1847 1960 2010
Min Year 1863 1863 1813 18 63 1837
67°N, 50.7°W,
50 m a.s.l.
Anomaly (°C) 0.6 0.7 0.4 0.3 0.7
Max Ye ar 2010 1986 2 016 1960 2010
Min Year 198 4 1983 1993 19 83 1982
70°N , 53. W,
10 m a.s.l.
Anomaly (°C) 1. 2 1.2 0.6 0.4 1.2
Max Ye ar 2010 2010 2010 2012 2010
Min Year 198 4 198 4 1983 19 72 198 2
178 4 ;
64. N , 51.7 ° W,
80 m a.s.l.
Anomaly (°C) 0.6 0.6 0.1 0.2 0.6
Max Ye ar 2010 2010 1932 2 012 2010
Min Year 1818 1818 1802 1819 1 811
62° N, 49.7 °W,
36 m a.s.l.
Anomaly (°C) 0.9 1.3 0.2 0.0 1.0
Max Ye ar 2010 2010 2005 2010 2010
Min Year 198 4 198 4 1993 19 69 19 82
61.2°N, 45.4°W,
27 m a.s.l.
Anomaly (°C) 1.4 1.4 1.3 0.2 0.9
Max Ye ar 2010 2010 2010 2016 2010
Min Year 1884 198 4 19 89 1873 18 74
60.7°N, 46.1°W,
32 m a.s.l.
Anomaly (°C) 0.7 1.0 0.3 0.1 0.7
Max Ye ar 2010 2010 1932 19 29 2010
Min Year 1884 1863 1811 1 811 1874
Kap Morris Jesup
198 0;
83.7°N, 33.4°W,
4 m a.s.l.
Anomaly (°C) 1. 5 5.2 0.8 0.4 0.4
Max Ye ar 2 011 2 011 2014 1995 2016
Min Year 198 5 198 8 1985 1997 199 0
Aputiteeq the autumn season was second warmest,
exceeded only by 2016.
At Summit, the highest elevation of the GrIS,
winter 2016/17 was the fourth warmest, with Febru-
ary 2017 second warmest after February 2005. May
was the second warmest since 1991, after May 2010. A
new July record-breaking low temperature of −33.0°C
was measured at Summit on 4 July. On 28 July, a new
record high July temperature of 1.9°C was measured
at Summit.
f. Glaciers and ice caps outside GreenlandM. Sharp,
B. Wouters, G. Wolken, L. M. Andreassen, D. Burgess, L. Copland,
J. Kohler, S. O’Neel, M. S. Pelto, L. Thomson, and T. Thorsteinsson
The Arctic is the world’s third most heavily glaci-
ated region, after Antarctica and Greenland. Though
the total mass of glaciers and ice caps in the region
is significantly less than that of the Antarctic and
Greenland ice sheets, ice loss from Arctic glaciers
and ice caps has become a significant contributor to
current global sea level rise as a result of recent sum-
mer warming (Gardner et al. 2011, 2013; Jacob et al.
2012; Millan et al. 2017).
Table 5.1. (cont.)
Station Name, Start Year;
Latitude, Longitude, Elevation
Station Nord
81.6°N, 16 .7°W,
36 m a.s.l.
Anomaly (°C) 1.0 2.7 1.8 0.4 2.2
Max Ye ar 2016 2011 2006 2003 2016
Min Year 1968 1967 1961 1970 1989
76.8°N, 18.7°W,
1 m a.s.l.
Anomaly (°C) 1.1 0.6 2.1 1.0 4.4
Max Ye ar 2016 2005 1976 2016 2016
Min Year 1983 1967 1966 1955 1971
74.3°N, 20.°W 2,
44 m a.s.l. .
Anomaly (°C) 0.5 0.3 3.1 0.1 4.8
Max Ye ar 2016 2005 1996 2016 2016
Min Year 1968 1975 1961 1985 1971
70.5°N , 2 W,
70 m a.s.l.
Anomaly (°C) 1.0 2.5 0.9 0.2 3.6
Max Ye ar 2016 2014 1996 2016 2 016
Min Year 1951 1966 1956 1955 1951
67. 8 °N, 32 .3°W,
13 m a.s.l.
Anomaly (°C) 1.6 4.4 1.4 0.2 2.2
Max Ye ar 2016 2017 1974 2016 2016
Min Year 1973 1969 1969 1967 1973
Tasi i l aq
65. 6°N , 37.6° W,
53 m a.s.l.
Anomaly (°C) 1.2 2.3 1. 3 0.2 1.6
Max Ye ar 2016 1929 1929 2016 19 41
Min Year 1899 1918 1899 1983 1917
61.9°N, 42°W,
39 m a.s.l.
Anomaly (°C) 0.1 1.1
Max Ye ar 2003 2 011 1999 1961 2010
Min Year 1983 1976 1967 1983 19 69
Prins Chr. Sund
60.1°N, 42.2°W,
88 m a.s.l.
Anomaly (°C) 0.5 0.6 0.2 0.2 1.3
Max Ye ar 2010 2010 2005 2010 2010
Min Year 1993 1993 1989 1970 19 82
72 .6°N , 38 .W,
3202 m a.s.l.
Anomaly (°C) 0.6 1.4 0.6 −0.6 2.7
Max Ye ar 2010 2010 2016 2012 2002
Min Year 1992 1993 1992 1992 2009
The state of glaciers, ice caps, and ice sheets is
often described by their mass balance. The annual
climatic mass balance of a glacier (Bclim) is defined as
the dif ference between the annual snow accumulation
on the glacier and the annual mass loss by surface
melting and runoff. For the purposes of calculation,
a “mass balance year” is usually taken as the period
between the ends of successive summer melt seasons.
Variations in the mass of most monitored Arctic gla-
ciers and ice caps are controlled largely by changes in
their climatic mass balance. However, those glaciers
that terminate in the ocean [e.g., Devon Ice Cap NW
(Arctic Canada), and Hansbreen and Kongsvegen
(Svalbard); Table 5.2; Fig. 5.14] or in lakes can also
lose mass by melting below the waterline. However,
this mass balance term is rarely routinely measured.
Here, Bclim measurements made in 2015–16 and
2016–17 at individual glaciers monitored across the
Arctic region are reported (Table 5.2; Fig. 5.141). All
Bclim data are from the World Glacier Monitoring
Service (WGMS 2018). Positive (negative) annual Bclim
values indicate that a glacier gained (lost) mass over
the course of the mass balance year that includes a
winter accumulation season, when snow deposition
typically exceeds meltwater runoff (positive mass
balance), followed by a summer ablation season,
when the opposite is the case (negative mass bal-
ance). The timing and duration of the accumulation
and ablation seasons vary from region to region and
from year to year, but in most cases, net accumula-
tion occurs from late autumn to late spring, and net
ablation from late spring to late autumn. At the time
of writing, estimates for the 2016–17 mass balance
year were available for only 16 glaciers [two in Alaska,
nine in Iceland (nine measurement locations at seven
glaciers), three in Svalbard, and two in Norway] of
the 27 that are regularly monitored (Fig. 5.14). So
that a complete cycle of results can be reported, Bclim
measurements for the 2015–16 mass balance year are
also reported.
Relative to the long-term (1985–2015) mean Bclim
values, 20 of the 24 values reported for 2015–16 were
more negative than the mean, and four were more
positive. Five of the 18 annual net balances reported
for 2016–17 were more negative than the 1985–2015
mean, and 13 were more positive. The mix of posi-
tive and negative anomalies in 2016–17 contrasts
1 Table 5.2 lists 25 glaciers and ice caps by name while Fig.
5.14 shows the location of 27 sites where Bclim is measured.
The difference in numbers is accounted for by Hofsjökull,
Iceland, where Bclim is measured at three different sites on a
single ice cap (no. 9 in Table 5.2).
with the tendency for predominantly negative mass
balance anomalies over the past decade. However,
the long-term tendency of the cumulative Bclim since
the mid-1990s continues to be toward more negative
cumulative balances in all five regions (Fig. 5.15),
indicating continuing mass loss. With the exception
of Svalbard (where there has been no obvious recent
acceleration of mass loss rates; Fig. 5.15), rapid mass
loss across the five regions typically began during
the 1990s.
New data on the length of the summer melt season
at Wolverine and Gulkana glaciers in Alaska (Fig.
5.16) show that since measurements began in 1966
the summer melt season has increased by about 18
days (14%) at Wolverine Glacier, located in a maritime
climate, and 24 days (24%) at Gulkana Glacier, located
in a more continental climate. These data suggest that
increases in summer melt played a significant role
in generating more negative annual mass balances
in this region.
Bclim measurements for the 2015–16 mass balance
year are from 24 glaciers: three in Alaska, four in
Arctic Canada, nine in Iceland, four in Svalbard, two
in northern Norway, and two in northern Sweden
(Table 5.2). All these glaciers had a negative annual
Bclim in 2015–16. At Austre Broggerbreen and Midtre
Fig. 5.14. Locations of the 27 sites on 25 Arctic glaciers
and ice caps (black circles) that have long-term obser-
vations of annual climatic mass balance (Bclim). Areas
outlined in yellow are the Randolph Glacier Inventory
(RGI) regions of the Arctic (Pfeffer et al. 2014). Some
individual glaciers are too close for identification and
can be identified by the numbers shown at the edge of
the RGI region. They can also be referenced in Table
5.2. Red shading indicates glaciers and ice caps, includ-
ing ice caps in Greenland outside the ice sheet.
Lovenbreen in Svalbard, Bclim was the most negative
ever recorded. This is attributed to relatively low snow
accumulation in winter 2015–16 and high summer
melt in 2016, especially in the record warm and rainy
month of July. Of the 18 glaciers with measurements
for both 2015–16 and 2016–17, 16 (two in Arctic
Canada, all nine in Iceland, three in Svalbard, and two
in northern Scandinavia) had a more positive annual
Bclim in 2016–17 than in the previous year, while two
(both in Alaska) had a more negative annual Bcl im
than in the previous year. In Svalbard, the positive
mass balance on Kongsvegen in 2016–17 is linked
to above-average winter snowfall, which delayed the
onset of ice melt in summer 2017.
Table 5.2. Measured Bclim climatic mass balance of 25 glaciers in Alaska (3), Arctic Canada (4), Iceland (7),
Svalbard (4), and Northern Scandinavia (7) for 2015/16 and 2016/17, together with the 1985–2015 mean and
standard deviation for each glacier [(Hofsjökull (Iceland) is treated as a single glacier, although measure-
ments are made in three different sectors of this ice cap)]. (* Indicates one or more years of data missing
from the record). Negative (positive) values for Bclim indicate mass loss (gain). Data are from the World
Glacier Monitoring Service (WGMS 2018), with updates for Alaska from S. O’Neel and M. Pelto, White
Glacier from L. Thomson, Svalbard from J. Kohler, and mainland Norway (Engabreen and Langf jordjokulen)
from L. M. Andreassen. Numbers in column 1 refer to the glaciers located in Fig. 5.14. Results for 2016/17
may be based on measurements made before the end of the melt season and may be subject to revision.
Region Glacier
(record length, years)
Bclim Mean
(kg m2 yr1)
Bclim Std. dev.
(kg m2 yr1)
(kg m2yr1)
2015 16
(kg m2 yr1)
20 16 –17
1Wolverine (52) 603 1016 400 116 0
3Lemon Creek (65) 640 798 120 0 148 0
2Gulkana (52) 778 721 1400
Arctic Canada
7Devon Ice Cap (NW)
(56) 204 205 483
5 Meighen Ice Cap (55) 26 397 775
4Melville South Ice Cap
(52) 418 477 792
6 White (54) 308 316 268
8 Langjökull S. Dome (19) 1288* 855 1677
9 Hofsjökull E (25) 545* 871 1120 650
9 Hofsjökull N (26) 565* 754 830 490
9 Hofsjökull SW (25) 802* 1017 450 80
10 Köldukvislarjökull (22) 475* 738 642 —
11 Tungnaarjökull (24) 112 8 * 830 196
12 Dyngjujökull (18) 146* 806 M —
13 Brúarjökull (23) 258* 683 342 —
14 Eyjabakkajökull (24) 709* 839 930
17 Midre Lovenbreen (49) 379 330 1200 420
16 Austre Broggerbreen (50) 486 363 1450 530
15 Kongsvegen (31) 114* 360 320 40
18 Hansbreen (28) 397* 474 1078
Although some of the 2016–17 mass balance mea-
surements are provisional, 12 of the reporting glaciers
(two in Alaska, one in Arctic Canada, six in Iceland,
two in Svalbard, and one in northern Scandinavia)
had negative annual balances, and six (Meighen
Ice Cap, Canada; Hofsjokull SW, Brúarjökull, and
Dyngjujökull, Iceland; Kongsvegen, Svalbard; and
Engabreen, Norway) had positive balances (Table 5.2).
Estimates of regional scale ice mass changes since
2011 can be derived from CryoSat-2 radar altimetry,
which measures glacier surface elevation (Wouters
et al. 2015). This approach provides regional mass
change estimates for Iceland, Svalbard, the Russian
Arctic, and the Canadian Arctic (Fig. 5.17). Cryo-
Sat-2 estimates for the period 2011–17 identify the
Canadian Arctic as the most important of these four
regional sources of glacier mass loss (7-year mean:
−60.19 Gt yr−1), followed by Svalbard (−18.95 Gt yr−1),
the Russian Arctic (−13.46 Gt yr−1), and Iceland (−2.36
Gt yr−1). Estimates for Alaska and northern Scandi-
navia are not available.
Fig. 5.16. Length (days) of the annual ablation season
at Gulkana (red) and Wolverine (blue) glaciers, Alaska,
showing the mean rate of change (days yr1) over the
1966–2017 observation period at each site. Coefficients
of determination (r2) determined by least squares
linear regression are 0.133 for Wolverine Glacier
(p = 0.008) and 0.08 for Gulkana Glaciers (p = 0.04).
(Source: S. O’Neel, USGS.)
Fig. 5.15. Cumulative climatic mass balance (Bclim in
kg m2) for glaciers and ice caps in f ive regions of the
Arctic, and for all monitored glaciers and ice caps (Pan-
Arctic). Average annual climatic balances for each
region are calculated for each year using the measured
annual mass balances for all monitored glaciers in the
region which are then summed over the period of
record to produce the cumulative Bclim. Note that the
monitoring periods vary between regions and that the
number and identity of glaciers monitored in a region
may vary between years.
Table 5.2. (cont.)
Region Glacier
(record length, years)
Bclim Mean
(kg m2 yr1)
Bclim Std. dev.
(kg m2 yr1)
(kg m2 yr1)
2015 16
(kg m2 yr1)
20 16 –17
Northern Scandinavia
19 Eng abreen (48) 127 1024 230 1250
20 Langfjordjokulen (27) 948* 737 1660 270
21 Marmaglaciaren (24) 460* 550 370
22 Rabots Glaciar (31) 465* 659 — —
23 Riukojetna (26) 592* 785 — —
24 Storglaciaren (71) 153 760 240
25 Tarfalaglaciaren (19) 198* 111 8 — —
Rapid changes occurring within the Arctic heighten
the need to understand the many causes of the changes
and their cumulative impacts. Most importantly, to better
understand Arctic change a holistic view is needed that can
only be achieved by bringing together multiple knowledge
systems and scientic disciplines. This includes Arctic
Indigenous Peoples and their knowledge.
Arctic Indigenous Peoples have been an integral part
of the Arctic ecosystem from time immemorial and have
acquired and built upon a unique knowledge system—an
indigenous knowledge—shaped by that environment. It
is a systematic way of thinking, which is applied to phe-
nomena across biological, physical, cultural, and spiritual
systems. It includes insights based on evidence acquired
through direct and long-term experiences and extensive
and multigenerational observations, lessons, and skills.
Indigenous knowledge has developed over millennia
and is still developing in a living process, including knowl-
edge acquired today and in the future, and it is passed
on from generation to generation (Inuit Circumpolar
Council-Alaska 2016). Indigenous knowledge stresses the
importance of understanding interconnecting systems,
that is, ecological, physical, cultural, and social systems,
the relationship between those components, and the
need to understand cumulative impacts (Inuit Circum-
polar Council-Alaska 2015). This world view and way of
understanding will aid in gaining a holistic understanding
of the Arctic and the changes that are occurring there.
To gain a truly holistic understanding of the chang-
ing Arctic, it is necessary to bring together indigenous
knowledge and science through a coproduction of
knowledge process. Such a process offers opportunities
to bring together different knowledge systems to develop
adaptation policies and practices for sustainability, and to
address biodiversity conservation and ecosystem-based
management in a changing Arctic.
The coproduction of knowledge process brings to-
gether indigenous knowledge holders and scientists to
work in partnership from the inception of a project, for
ex amp le, iden ti cat ion of research ques tio ns an d hy pot h-
eses, through analysis and output. Equity is a cornerstone
of the process, ensuring fairness and the opportunity to
engage in all aspects of a project. All participants have a
fair and equal chance of succeeding. The coproduction
of knowledge process requires culturally appropriate
methodologies in data collection and analyses to be agreed
upon by all knowledge holders.
Successful coproduction of knowledge fosters an
environment of trust and respect , works toward empow-
erment and capacity building, and recognizes indigenous
knowledge sovereignty; it is important to recognize the
sovereign rights of indigenous peoples, including those
related to their own knowledge. This includes indigenous
peoples fully understanding the risks and opportunities of
participating in a research project, having authority over
how data and information are shared, and the right to not
participate in a research project. The principles of free,
prior, and informed consent are critical to the coproduc-
tion of knowledge process (UN General Assembly 2007).
Successful coproduction of knowledge is built upon
lo ng- ter m par tne rsh ips . A go od rst step is an unde rst and -
ing of the lay of the land in indigenous homelands. Just as
scientists understand the importance of networks in their
research, so indigenous peoples also live in complex social
and governance systems, allowing the opportunity to
leverage existing indigenous networks, institutions, and
organizations. It is important to understand partnership
building from an indigenous perspective and to know that
partnership and participation are not synonymous. Clear,
transparent, culturally appropriate terms of reference are
recommended to ensure there are no misunderstandings
and to help with relationship building.
Indigenous knowledge and modern science have
different approaches, methodologies, analyses, and
validation processes. The coproduction of knowledge
process requires respect for each knowledge system and
avoiding translation of one knowledge system into the
other, that is, trusting an indigenous knowledge holder’s
ability to analyze their own information and respect that
each person at the table comes with the credentials
needed to be there. While some credentials are built from
academic degrees and publications, others come from
holding and demonstrating a body of knowledge through
a lifetime of hunting, shing, gathering, or being an elder.
Many Arctic science projects have aimed to build
partnerships with indigenous communities, but few
have used a true coproduction of knowledge process
that brings together indigenous knowledge holders and
scientists equitably from the inception of the project. An
exemplar that demonstrates the process is the Ikaaġvik
g. Terrestrial permafrost—V. E. Romanovsky, S. L. Smith,
K. Isaksen, N. I. Shiklomanov, D. A. Streletskiy, A. L. Kholodov,
H. H. Christiansen, D. S. Drozdov, G. V. Malkova, and S. S. Marchenko
Permafrost is an important component of the
Arctic landscape, inf luencing hydrological systems
and ecosystems, and presenting challenges for built
infrastructure, for example, buildings, roads, rail-
ways, airports, and pipelines. Permafrost temperature
and active layer thickness (ALT) are key indicators
of changes in permafrost conditions. Permafrost is
defined as earth materials (e.g., soil, rock) that exist at
or below 0°C continuously for at least two consecutive
Sikukuun (Ice Bridges) project in Kotzebue in northwest
Alaska (Mahoney et al. 2017). This four-year (2017–20)
project, which aims to understand fundamental processes
underlying the mechanisms and impacts of changing coastal
se a ice , r st brou ght togethe r in digenous know le dge holder s
with scientists from different disciplines to develop the
research focus and questions, decide on a methodolog y, and
then agree on a plan for implementing the project together.
Indigenous knowledge will also inform the synthesis and
dissemination of the results of the project.
The success of a coproduction of knowledge process
must be dened by both the indigenous knowledge holders
and scientists involved in a project. Experts from both
knowledge systems must agree that a coproduction of
knowledge occurred and it will hold all of the basic elements
presented here. These include recognizing and working
toward equity through building capacity, empowering
indigenous partners, fostering an environment for trust
and respect, building a collaborative process that involves
multiple steps and continuous evaluation and which is
dened by all those involved in a project, and respecting
indigenous knowledge sovereignty.
Fig. 5.18. Location of the permafrost temperature
monitoring sites shown in Fig. 5.19 superimposed on
average sur face air temperature anomalies (°C) dur-
ing 2000–16 (with respect to the 1981–2010 mean)
from the NCEP-reanalysis (Kalnay et al. 1996). Data
provided by the NOAA /ESRL Physical Sciences Divi-
sion (ww Sites shown in Fig. 5.19
for (a) Barrow (Ba), West Dock (WD), KC-07 (KC),
Duvany Yar (DY), Deadhorse (De), Franklin Bluffs
(FB), Galbraith Lake (GL), Happy Valley (HV), Norris
Ck (No); (b) College Peat (CP), Old Man (OM), Chan-
dalar Shelf (CS), Birch Lake (BL), Coldfoot (Co), Nor-
man Wells (NW), Wrigley 1 and 2 (Wr), Healy (He),
Gulkana (Gu); (c) Eureka EUK4 (Eu), Alert BH1, BH2,
and BH5 (Al), Resolute (Re), Arctic Bay (AB), Pond In-
let (PI), Pangnirtung (Pa); (d) Janssonhaugen ( Ja), Bay-
elva (Ba), Kapp Linne 1 (KL), Urengoy #15- 06 and #15-
10 (Ur), Juvvasshøe ( Ju), Tarfalaryggen (Ta), Polar Ural
(ZS), Bolvansky #56, #59, and #65 (Bo), Iskoras Is-B-2
(Is). Information about these sites is available at http:
/sites_map, and
Fig. 5.17. Cumulative regional glacier mass anomalies
(in, Gt) for Iceland, Arctic Canada, Arctic Russia, and
Svalbard, derived using data from CryoSat2 radar al-
timetry (2011–17) (B. Wouters, Utrecht University).
Cumulative mass anomalies in each region are defined
relative to the glacier mass measured in the region at
the start of the measurement period. Trend lines and
average annual rates of mass change (Gt yr1) in each
region are shown. Annual cycles in the accumulation
and removal of mass are evident in each region.
years. The active layer is the seasonally thawed layer
above the permafrost. Permafrost temperatures, at a
depth where seasonal temperature variations are neg-
ligible, are powerful indicators of long-term change.
On the other hand, the active layer responds to
shorter term f luctuations in climate and is especially
sensitive to changes in summer air temperature and
precipitation. Warming and thawing of permafrost
in the Arctic are reported here.
Changes up to 2017 (most recent data available)
in mean annual permafrost temperatures and ALT
are summarized for a number of sites throughout
the Arctic (Fig. 5.18). Recent long-term changes in
permafrost temperature are driven mostly by air
temperature trends (Romanovsky et al. 2015). Other
important influences on
permafrost temperature,
such as snow depth and
density, vegetation charac-
teristics, and soil moisture,
can affect the observed
permafrost temperature
trends at the local scale.
In general, the increase in
permafrost temperatures
observed since the 1980s
is more significant in the
higher latitudes where the
largest increase in air tem-
perature is observed (Fig.
1) permafroSt temper a-
Temperatures in the
cold continuous perma-
frost of northern Alaska,
Northwest Territories
(Canada), and northeast-
ern East Siberia continue
to rise (Fig. 5.19a). In 2017
on the North Slope of
Alaska, record high tem-
peratures at 20-m depth
occurred at all perma-
frost observatories (Bar-
row, West Dock, Franklin
Bluffs, Happy Valley, and
Deadhorse; Fig. 5.19a) with
the exception of Galbraith
Lake. The permafrost tem-
perature increase (+0.1° to
+0.2°C) between 2016 and
2017 was substantial and comparable to the highest
rate of warming observed in this region, which oc-
curred during 1995–2000 (Fig. 5.19a). Since 2000,
permafrost temperature increase at 20-m depth in
this region has ranged from 0.21° to 0.66°C decade−1
(Fig. 5.19a; Table 5.3).
In Interior Alaska, following the slight cooling of
2007–13, permaf rost temperatures increased and were
higher in 2017 than in 2016 at all sites (Coldfoot, Old
Man, College Peat, Birch Lake, Gulkana, and Healy
in Fig. 5.19b). The largest changes, at Birch Lake and
Old Man, were associated with new record highs in
2017 for the entire 33-year measurement period (Fig.
5.19b; Table 5.3).
Fig. 5.19. Time series of mean annual ground temperature (°C) at depths of 9
to 26 m below the surface at selected measurement sites that fall roughly into
priority regions of the Adaptation Actions for a Changing Arctic Project (AMAP
2015): (a) cold continuous permafrost of northern Alaska , Northwest Territo-
ries (Canada), and NE East Siberia; (b) discontinuous permafrost in Interior
Alaska and northwestern Canada; (c) cold continuous permafrost of eastern
and High Arctic Canada (Baffin Davis Strait); and (d) continuous to discontinu-
ous permafrost in Scandinavia, Svalbard, and Russia/Siberia (Barents region).
Temperatures are measured at or near the depth of zero annual amplitude
where the seasonal variations of ground temperature are negligible. Note that
the temperature scales are different in each graph. Data are updated from
Christiansen et al. 2010; Romanovksy et al. 2017; Smith et al. 2015, 2017; Ednie
and Smith 2015; Boike et al. 2018.
In northwestern Canada, the temperature of
permafrost in the central Mackenzie Valley (Nor-
man Wells and Wrigley in Fig. 5.19b; Table 5.3) has
generally increased since the mid-1980s (Smith et
al. 2017). Although less warming has been observed
since 2000, permafrost temperatures in 2017 at these
sites were the highest recorded. Greater recent warm-
ing has been obser ved in the colder permafrost of the
northern Mackenzie region (Norris Ck, KC-07 in
Fig. 5.19a and Table 5.3; Smith et al. 2017), with the
highest temperatures during the observation period
occurring in 2016/17.
In northeastern Canada, the 2016/17 mean perma-
frost temperatures in the upper 25 m of the ground
at Alert, northernmost Ellesmere Island in the high
Arctic, were among the highest recorded since 1978
(Fig. 5.19c). Although permafrost at Alert has gener-
ally warmed since 1978, permafrost temperatures
have increased at a higher rate since 2000 (Table 5.3;
Smith et al. 2015), consistent with air temperature
trends (Fig. 5.18). There has been little change at Alert
since 2010 (Fig. 5.19c), which coincides with a period
of lower mean annual air temperatures. At other sites
in the Queen Elizabeth Islands (Resolute and Eureka)
and on Baffin Island (Pangnirtung, Pond Inlet, and
Table 5.3. Change in mean annual ground temperature (°C decade1) for sites shown in Fig. 5.19. For
sites where measurements began prior to 2000, the rate for the entire available record is provided as
well as the rate for the period after 2000. The names of the stations with record high temperatures in
2017 are shown in red. Note that some records only began after 2007, as shown in Fig. 5.19.
Region Sites Entire Record Since 2000
Alaskan Arctic plain West Dock (WD), Deadhorse (De),
Franklin Bluffs (FB), Barrow (Ba) +0.36 to +0.8 +0.44 to +0.65
Northern foothills of the
Brooks Range, Alaska Happy Valley (HV), Galbraith Lake (GL) +0.3 to +0.42 +0.34 to +0.47
Southern foothills of the
Brooks Range, Alaska
Coldfoot (Co), Chandalar Shelf (CS),
Old Man (OM) +0.08 to +0.35 +0.14 to +0.25
Interior Alaska College Peat (CP), Birch Lake (BL),
Gulkana (Gu), Healy (He) +0.07 to +0.22 +0.03 to +0.1
Central Mackenzie Valley Norman Wells (NW), Wrigley (Wr) Up to +0.1 <+0.1 to +0.2
Northern Mackenzie Valley Norris Ck (No), KC-07(KC) +0.5 to +0.9
Baffin Island Pangnir tung (Pa), Pond Inlet (PI),
Arctic Bay (AB) +0.5 to +0.7
High Canadian Arctic Resolute (Re), Eureka (Eu) +0.4 to +0.7
High Canadian Arctic Alert (Al) at 15 m
+0.3 to +0.4
+0.7 to +0.9
North of East Siberia Duvany Yar (DY) +0.3
North of West Siberia Urengoy 15-06 and
15-10 (Ur) +0.31 to +0.47 +0.1 to +0.19
Russian European North Bolvansky 56, 59, and 65 (Bo),
Polar Ural (ZS-124) +0.18 to +0.46 +0.1 to +0.83
Svalbard Janssonhaugen (Ja), Bayelva (Ba),
Kapp Linne 1 (KL) +0.7 +0.6 to +0.7
Northern Scandinavia Tar f ala rgge n ( Ta), Iskoras Is-B-2 (Is) +0.1 to +0.4
Southern Norway Juvvasshøe (Ju) +0.2 +0.2
Arctic Bay), permafrost temperature measurements
since 2008, at 10- to 15-m depth, indicate an overall
warming (Fig. 5.19c; Table 5.3). A decrease in perma-
frost temperature since 2012 appears to be associated
with lower mean annual air temperatures over the last
few years in the region.
Increases in permafrost temperature over the last
30–35 years in northern Russia have been similar
to those in northern Alaska and the Canadian high
Arctic (Drozdov et al. 2015). In the Russian European
North and western Siberian Arctic, temperatures at
10-m depth have increased by ~0.4° to 0.6°C decade−1
since the late 1980s at colder permafrost sites (Fig.
5.19d, sites Bolvansky #59, Urengoy #15-06 and #15-
10) and increased less in warmer permafrost sites
(Table 5.3; Fig. 5.19d, sites Bolvansky #56 and Urengoy
#15-06; Drozdov et al. 2015). In these regions, there
are differences in permafrost temperature (cold vs.
warm) because surface conditions such as vegetation,
surface wetness, and soil moisture vary according to
landscape types, while climatic conditions are largely
independent of surface condition and landscape.
In the Nordic region, where the temperature at
20-m depth has increased between 0.1° and 0.7°C
decade−1 (Fig. 5.19d; Table 5.3) since 2000, warm-
ing and thawing of permafrost have been observed
recently (Christiansen et al. 2010; Isaksen et al. 2011;
Farbrot et al. 2013). Lower rates of warming occur
where permafrost temperatures are close to 0°C
and latent heat effects related to melting ground ice
are important. Greater warming occurred in colder
permafrost on Svalbard and in northern Scandinavia
(Table 5.3). In the discontinuous permafrost zone of
southern Norway, permafrost warmed between 2015
and 2017, following a period of cooling between 2011
and 2014 (Fig. 5.19d).
2) active layer thickneSS
In 2017, standardized, mechanical probing of
ALT was conducted at 76 Circumpolar Active-Layer
Monitoring (CALM) program sites in Alaska and
Russia. Each site consists of a spatial grid varying
from 1 ha to 1 km2 in size and is representative of
the regional landscape (Shiklomanov et al. 2012).
Additional active-layer observations, derived from
thaw tubes (Duchesne et al. 2015), are available from
25 Canadian sites located in the Mackenzie Valley,
northwestern Canada.
The average ALT in 2017 for 20 North Slope of
Alaska sites was 0.52 m, which is 0.06 m (or +12%)
higher than the 1996–2005 mean and is one of the
highest in the 22-year data record. Previous maxima
occurred in 1998, 2013, and 2016 (Fig. 5.20). The
interior of Alaska is characterized by a pronounced
ALT increase over the last 22 years (Fig. 5.20). How-
ever, after reaching the 22-year maximum of 0.77 m
in 2016, the 2017 ALT decreased by 10% to 0.69 m.
Records from the sites with thaw tubes in the
Mackenzie Valley indicate that there has been a gen-
eral increase in ALT in this region since 2008 (Fig.
Fig. 5.20. Long-term annual active-layer thickness change (m) in six different Arctic regions for 2017 as observed
by the CALM program relative to the 2003 –12 mean. Positive (negative) anomaly values indicate the active
layer was thicker (thinner) than average. Thaw depth observations from the end of the thawing season were
used. Only sites with at least 20 years of continuous thaw depth observations are shown.
5.20; Duchesne et al. 2015; Smith et al. 2017). ALT in
2016 (most recently available data) was on average
0.06 m greater than the 2003–12 mean, similar to the
previous peak value in 2012.
A decrease in ALT from 2016 to 2017 was reported
for all Russian regions. In West Siberia, the average
2017 ALT was 1.25 m, which is 0.2 m (or 14%) smaller
than the 20-year maximum observed in 2016. In the
Russian European North, the 2017 ALT was 1.08 m
compared to 1.24 m in 2016. A 2017 ALT of 0.69 m
was reported for East Siberia, which is 0.1 m smaller
than the regional average 2016 ALT value. The small-
est decrease was reported in the Russian Far East
(Chukotka), where the ALT in 2017 was 0.03 m (or
5%) less than that reported in 2016.
In the Nordic region, active layer records (1999–
2017) indicate a general ALT increase of 0.10 to 0.30
m since 1999. The particularly warm summer of 2014
in the Nordic region contributed to the thickest active
layer measured so far at some places.
h. Tundra greenness—H. Epstein, U. Bhatt, M. Raynolds,
D. Walker, J. Pinzon, C. J. Tucker, B. C. Forbes, T. Horstkotte,
M. Macias-Fauria, A. Martin, G. Phoenix, J. Bjerke, H. Tømmervik,
P. Fauchald, H. Vickers, R. Myneni, T. Park, and C. Dickerson
Vegetation in the Arctic tundra has been respond-
ing to environmental changes over the course of
the last several decades, with the tendency being an
increase in the amount of above-ground vegetation,
that is, “greening” (Bhatt et al. 2010). These vegetation
changes vary spatially throughout the circumpolar
Arctic in both direction and magnitude, and they
are not always consistent over time. This suggests
complex interactions among the atmosphere, ground
(soils and permafrost), vegetation, and animals of
the Arctic system. Changes in tundra vegetation can
have important effects on permafrost, hydrology,
carbon and nutrient cycling, and the surface energy
balance (e.g., Frost et al. 2017; Kępski et al. 2017), as
well as the diversity, abundance, and distribution of
both wild and domesticated herbivores (e.g., Fauchald
et al. 2017; Horstkotte et al. 2017). We continue to
evaluate the state of the circumpolar Arctic vegeta-
tion, to improve our understanding of these complex
interactions and their impacts on the Arctic system
and beyond.
The reported controls on tundra greening are
numerous and varied. They include increases in sum-
mer, spring, and winter temperatures and increases
in growing season length (Bhatt et al. 2017; Fauchald
et al. 2017; Horstkotte et al. 2017; Myers-Smith et
al. 2018; Vickers et al. 2016), in part controlled by
reductions in Arctic Ocean sea ice cover (Bhatt et al.
2017; Macias-Fauria et al. 2017; see Section 5d). Other
controls on tundra greening include increases in
snow water equivalent (see Section 5i) and soil mois-
ture, increases in active layer depth (see Section 5g),
changes in the patterns of herbivore activity, and even
a reduction in the human use of the land (Fauchald
et al. 2017; Horstkotte et al. 2017; Martin et al. 2017;
Westergaard-Nielsen et al. 2017).
Using Earth-observing satellites with subdaily
return intervals, Arctic tundra vegetation has been
continuously monitored since 1982. Here, data are
reported from the Global Inventory Modeling and
Mapping Studies (GIMMS) 3g V1 dataset, based
largely on the AVHRR sensors aboard NOAA satel-
lites (Pinzon and Tucker 2014). At the time of writing,
the GIMMS3g V1 dataset was only available through
2016. The GIMMS product (at 1/12° resolution for
this report) is a biweekly, maximum-value compos-
ited dataset of the normalized difference vegetation
index (NDVI). NDVI is highly correlated with above-
ground vegetation (e.g., Raynolds et al. 2012), or
“greenness,” of the Arctic tundra. Two metrics based
on the NDVI are used: MaxNDVI and TI-NDVI.
MaxNDVI is the peak NDVI for the year (growing
season) and is related to yearly maximum above-
ground vegetation biomass. TI (time-integrated)
NDVI is the sum of the biweekly NDVI values for
the growing season and is correlated with the total
above-ground vegetation productivity.
Examining the overall trend in tundra green-
ness for the now 35-year record (1982–2016), it is
apparent that the MaxNDVI and the TI-NDVI have
increased throughout most of the circumpolar Arctic
tundra (Fig. 5.21). Regions with some of the greatest
increases in tundra greenness are the North Slope of
Alaska, the low Arctic (southern tundra subzones) of
the Canadian tundra, and eastern Siberia. However,
tundra greenness has declined (i.e., the tundra has
been “browning”) on the Yukon–Kuskokwim Delta
of western Alaska, in the high Arctic of the Canadian
Archipelago, and in northwestern Siberia. Regions of
greening and browning, measured by NDVI increases
and decreases, respectively, tend to be consistent
between MaxNDVI and TI-NDVI.
Following 2–3 years of successive declines prior to
and including 2014, the NDVI or greenness of Arctic
tundra increased in 2015 and 2016 for both indices
(MaxNDVI and TI-NDVI) and both continents
(North America and Eurasia), exhibiting substantial
recovery from the previous years of “browning.”
(Fig. 5.22). One exception was the TI-NDVI for
North America, which continued to decrease in
2015. MaxNDVI and TI-NDVI for the entire Arctic
increased 6.0% and 9.3%, respectively, between 2015
and 2016. MaxNDVI in North America increased by
6.3% compared to 5.4% in Eurasia. The first substan-
tial annual increase in TI-NDVI for North America
since 2010 occurred in 2016, potentially due to the
high growing season temperatures that year.
All NDVI values for 2016 were greater than
their respective mean values for the 35-year record.
MaxNDVI values ranked second, third, and first for
the Arctic, Eurasian Arctic, and North American
Arctic, respectively. TI-NDVI values ranked first,
first, and second for the Arctic, Eurasian Arctic,
and North American Arctic, respectively. Based on
remotely-sensed land surface temperatures (LST)
from the same sensors as those providing the NDVI
values, the summer warmth index (SWI: sum of mean
monthly temperatures >0°C) for the Arctic as a whole
and for the Eurasian Arctic was greater in 2016 than
in any other year of the satellite record (since 1982).
For the North American Arctic, the 2016 SWI was
the second highest on record (very close to the high-
est value in 1994).
Even though the past two years have seen large
increases in tundra NDVI, there are still regions of
the Arctic that have experienced browning over the
length of the satellite record. There have also been
substantial periods of tundra browning even within
a general greening trend. While research on tundra
browning is still relatively sparse, there has recently
been greater attention given to this phenomenon.
Bjerke et al. (2017) report on extensive vegetation
dieback in northern Norway (including Svalbard) in
2014 and 2015. They attributed this dieback largely to
Fig. 5.22. (a) MaxNDVI and (b) TI-NDVI for Eurasia
(top), the Arctic as a whole (middle), and North
America (bottom) for 1982–2016.
Fig. 5.21. (a) Magnitude of the trend in (a) MaxNDVI
and (b) TI-NDVI for 1982–2016
Despite the low temperatures and short
growing seasons of northern ecosystems,
wildland re is the dominant ecological dis-
turbance in the boreal forest, the world’s
largest terrestrial biome. Wildland re also
affects adjacent tundra regions. This sidebar,
with a focus on the 2017 Alaska re season,
addresses the history and variability of re
disturbance in Alaska (US) and Northwest
Territories (Canada), outlines how short-term
weather conditions (temperature, precipita-
tion, convection, and wind) inuence area
burned, and discusses projections for future
tendencies in re susceptibility.
Beyond immediate threats to lives and
property, re impacts include compromised
human health and limited visibility due to
smoke. Fire disturbance affects terrestrial
ecosystems at multiple scales, including car-
bon release through combustion (Kasischke
et al. 2000). About 35% of global soil carbon
is stored in tundra and boreal ecosystems
(Scharlemann et al. 2014) that are potentially
vulnerable to re disturbance (Turetsky et
al. 2015). Other impacts include interactions
with vegetation succession (Mann et al. 2012;
Johnstone et al. 2010), biogeochemical cycles
(Bond-Lamberty et al. 2007), energ y balance
(Rogers et al. 2015), and hydrology (He. Liu
et al. 2005). Combustion of the insulating
surface organic layer can destabilize underly-
ing permafrost. Because permafrost impedes
drainage and ice-rich permafrost settles upon
thawing (thermokarst), accelerating degrada-
tion of the permafrost may have large conse-
quences for northern ecosystems (Jorgenson
et al. 2010; Jones et al. 2015).
Weather is a dominant control of re ac-
tivity on a year-to-year basis. Over the longer
term, high-latitude re regimes appear to be
responding rapidly to environmental changes
associated with the warming climate. Although
highly variable, area burned has increased since the 1960s in
much of boreal North America (Kasischke and Turetsky 2006;
Gillett et al. 2004). Over that time, both the number and size
of individual re events has increased, contributing to more
freq uent la rge re yea rs in nor thwes tern No rth Amer ica (Ka -
sischke and Turetsky 2006). Figure SB5.3 shows area burned
each year since 1980 in Alaska and Nor thwest Territories,
including both boreal and tundra regions.
Although highly variable, high-latitude re seasons gen-
erally begin and end earlier than in more temperate areas
Fig. SB5.3. Annual area burned (ha) each year since 1980 in (a) Alaska
and (b) Northwest Territories (Canada), including both boreal and
tundra regions. Note that high fire years are not coincident in these
subregions, indicating the importance of local weather and other
conditions (e.g., fuels, ignition). Category definitions used here are
from the f itted log-normal distribution to the observed 1980–2017
area burned; below normal is the 0–33rd percentiles, near normal is
the 33rd–66th percentiles, above normal is the 66th–90th percentiles,
much above is greater than the 90th percentile.
Fig. SB5.4. Average (gray line) and climatological range (gray shading) of
BUI between 1 Apr and 30 Sep in Alaska’s boreal interior for 1994–2017,
compared to the 2017 average (solid purple line) and the 2017 predictive
service area AK02 (Upper Yukon and surrounding uplands, centered
around the Arctic Circle; dashed purple line). While the boreal interior
average BUI for 2017 (purple line) was similar to the historic average BUI
(gray line), the Upper Yukon Zone (dashed purple line), where the major-
ity of the hectares burned in the territory in 2017, showed a significant
elevation in BUI from mid-Jun to mid-Aug.
(Fig. SB5.4). Depending on weather, re
danger can increase as soon as areas are
snow-free in April and May; season-ending
rains typically fall in July or August, but
their absence can extend the season into
September, as in the record years of 2004
(2.67 million ha) and 2005 (1.88 million
ha) in Alaska. Recent large re seasons in
high latitudes include 2014 in Northwest
Territories (Fig. SB5.3), where 385 res
burned 3.4 million ha, and 2015 in Alaska
(Fig. SB5.3), where 766 fires burned 2
million ha—the latter was more than half
the total area burned in the entire United
States (NWT 2015; AICC 2015). North-
ern communities threatened or damaged
by rec ent wil dr es in clude Fort McMurray,
located in the boreal forest in Alberta,
Canada, where 88 000 people were evacu-
ated and 2400 structures were destroyed
in May 2016 (Kochtubajda et al. 2017). The
2007 Anaktuvuk River Fire is the largest
(104 000 ha) and longest-burning (almost
3 months) re known to have occurred on the North Slope
of Alaska and initiated widespread thermokarst development
(Jones et al. 2015).
Most area burned in northern ecosystems occurs during
sporadic periods of high re activity. Half of the area burned
in Alaska from 2002 to 2010 was consumed over just 36 days
(B arr ett et al. 2016 ). Rec ent anal yse s have iden ti ed a te mpe ra-
ture threshold in Alaska with a much greater likelihood of re
occurrence within a 30-year period at locations where mean
Ju ly te mpe rat ure s excee d 13.4°C (Young et al. 2017). Lar ge r e
even t s re qui re the co nu enc e of warm and dr y wea the r condi -
tions with a source of ignition (often lightning from convective
thunderstorms) and fuels that can carry re. High latitude
ecosystems are characterized by unique fuels, in particular,
fast-drying beds of mosses, lichens, and accumulated organic
material (duff) that underlie resinous shrubs and dense, highly
ammable conifers. These understor y fuels dry rapidly during
periods of warm, dry weather and the long day lengths of June
and July. Consequently, extended drought is not required to
increase re danger to extreme levels.
Historically, lightning is responsible for the majority of the
ac re a ge burn ed in hig h lat itu des , as li ght nin g-i gnited re s occur
in more remote locations and thus are subject to lower levels
of suppression than human-started incidents. Veraverbeke et al.
(2017) showed that lightning ignitions have increased in boreal
North America since 1975 and were a major contributor in
the extreme 2014 Northwest Territories and 2015 Alaska re
seasons. In addition, Partain et al. (2016) found that human-
induced climate change—manifested as a combination of high
surface air temperatures, low relative humidity, and low pre-
cipitation—increased the likelihood of the extremely dry fuel
conditions seen in Alaska in 2015 by 34%–60%.
The snow-free season has increased by approximately 5
days decade−1 in Alaska since 1979 (Liston and Hiemstra 2011).
In resp onse, in 2006 Alask a’s re ma nag eme nt agencie s shi f ted
the statutory start of re season ahead by a month, from 1
May to 1 April, to better prepare for early season events. In
addition to adapting to long-term trends, managers in Alaska
and Canada must track day-to-day variability in threats to dis-
persed populations with limited resources. Managers in both
regions use the Canadian re weather index (FWI) system on
a daily basis to estimate the spatial and temporal distribution of
wildre potential from observed and forecast weather condi-
tions (Lawson and Armitage 2008). Among the FWI indices,
the buildup index (BUI), based on cumulative scoring of daily
temperature, relative humidity, and precipitation, represents
seasonal variability in fuel availability and ammability (Fig.
SB5.4). A BUI threshold of 80 has been identied as a critical
indicator of re growth potential in Alaska (Ziel et al. 2015).
In 2017, the typical area burned in Alaska (264 221 ha;
Fig. SB5.3) was reected in a fairly normal BUI across the
boreal region that essentially paralleled the climatological
average (Fig. SB5.4). However, the impact of a “normal”
season can fall disproportionately on specic areas in a
landscape this large. In 2017, while there were no signi-
cant peaks in the BUI, local conditions in the Upper Yukon
zone in northeast Alaska were signicantly warmer and
drier. Consistent with the Upper Yukon BUI trend (Fig.
SB5.4), the re season was extended and fairly severe in
that large region of the state, with periods of high re
danger (BUI �80) from mid-June to mid-August near and
north of the Arctic Circle. More than 160 000 ha (63% of
the 2017 Alaska total) burned in the Upper Yukon area
during this period.
Under a range of climate change scenarios, analyses
using multiple approaches project signicant increases (up
to four-fold) in area burned in high latitude ecosystems by
the end of the 21st century (French et al. 2015; Young et al.
2017; Yue et al. 2015, and references therein). In addition,
annual lightning frequency is projected to increase by 12%
± 5% per °C of warming in the contiguous United States
(Romps et al. 2014) and may increase correspondingly
in high latitudes. Because specic re events depend on
multiple interacting factors, the resulting changes in high
la tit ude re regi mes will var y greatly over sp ace and ti me,
but all evidence indicates that northern ecosystems will
become increasingly susceptible to burning.
changes in winter weather, specifically reductions in
snow cover areal extent due to winter warming events,
which left the ground exposed to subsequent freezing
and desiccation (Vikhamar-Schuler et al. 2016). Insect
outbreaks were identified as a secondary contributor
to vegetation mortality (Bjerke et al. 2017).
i. Terrestrial snow cover in the Arctic—C. Derksen, R. Brown,
L. Mudryk, K. Luojus, and S. Helfrich
Satellite-derived estimates of snow cover extent
(SCE) over Arctic land areas date back to 1967 and
have revealed dramatic reductions since 2005. These
changes are important to the Arctic system because
spring snow cover over land areas significantly
influences the surface energy budget (snow is
highly ref lective of incoming solar energy), ground
thermal regime (snow is an effective insulator of the
underlying soil), and hydrological processes (the
snowpack stores water in solid form for many months
before spring melt). Changes in snow cover also have
the potential to impact fauna living above, in, and
under the snowpack, vegetation, biogeochemical
activity, and exchanges of carbon dioxide and other
trace gases (Brown et al. 2017).
Spring (April–June) SCE anomalies for the Arctic
(land areas north of 60°N) were regionally computed
for North America and Eurasia using the NOAA snow
chart climate data record, which extends from 1967
to present (maintained at Rutgers University; Estilow
et al. 2015;; Fig.
5.23). For the first time in over a decade, 2017 Eur-
asian Arctic spring SCE was above average relative to
the 1981–2010 reference period. April and May SCE
anomalies were positive, including the second high-
est May SCE over the period of satellite observations.
These are the first positive SCE anomalies observed
in May over the Eurasian Arctic since 2005; June SCE
anomalies were positive across the Eurasian Arctic
for the first time since 2004. SCE anomalies over the
North American Arctic were negative all spring but
did not approach the series of record-breaking low
SCE values observed in recent years.
Snow cover duration (SCD) departures were
calculated from the NOAA Interactive Multisensor
Snow and Ice Mapping System (IMS; Helfrich et al.
2007) product to identify differences in the onset of
snow cover in fall and melt of snow cover in spring
relative to a 1998–2010 reference period. While there
was evidence of earlier snow cover onset over much of
midlatitude Eurasia in autumn 2016 (consistent with
cold surface air temperature anomalies), Arctic land
areas (with the exception of Alaska) had near-normal
snow onset timing (Fig. 5.24a). Later-than-normal
snow melt onset across Eurasia (Fig. 5.24b), also
reflected in the positive SCE anomalies (Fig. 5.23),
was consistent with colder-than-normal surface air
temperatures across this region (especially in May
and June). Spring snow melt across the Canadian
Fig. 5.24. SCD anomalies (%) from the NOAA daily
IMS snow cover product (Helfrich et al. 2007; rela-
tive to 1998–2010 base period because of the shorter
available time series but higher spatial resolution IMS
data). IMS data record for the (a) 2016 autumn season
and (b) 2017 spring season. Snow depth anomaly (% of
1999–2010 average) from the CMC snow depth analysis
for (c) Apr and (d) Jun 2017.
Arctic was slightly earlier than normal, coincident
with warmer-than-average surface temperatures in
May and June. Snow depth anomalies derived from
the Canadian Meteorological Centre daily gridded
global snow depth analysis (Brasnett 1999) showed
predominantly positive anomalies over high latitude
regions of Siberia and North America in April (Fig.
5.24c) and mainly negative anomalies outside the
Arctic. By late spring (June), the anomalies exhibited
contrasting continental patterns, with Eurasia char-
acterized by extensive positive snow depth anomalies,
while the North American Arctic was dominated by
negative snow depth anomalies (Fig. 5.24d), consis-
tent with the region of earlier snow melt (Fig. 5.24b).
Four independent products were integrated to
generate a multidataset snow water equivalent (SWE;
the amount of water stored in solid form as snow)
anomaly time series (1980–2017) for April (typically
the month of maximum SWE across the Arctic; Fig.
5.25). The datasets were derived from: (1) modern
atmospheric reanalysis (the Modern-Era Retrospec-
tive Analysis for Research and Applications version
2; Reichle et al. 2017); (2) reconstructed SWE driven
by ERA-Interim meteorology described by Brown et
al. (2003); (3) the physical snowpack model Crocus
driven by ERA-Interim meteorology (Brun et al.
2013); and (4) the European Space Agency GlobSnow
product derived through a combination of satellite
passive microwave measurements and climate station
observations (Takala et al. 2011). While there is a high
degree of interannual variability in the multidataset
SWE anomalies, they predominantly show a negative
trend since 2000 (Fig. 5.25). North American Arctic
SWE was again negative in 2017 while Eurasian SWE
anomalies were positive, indicating a deeper-than-
Fig. 5.23. (a) Monthly SCE anomaly (× 103 km2) for Arctic land areas (> 60°N) from the NOAA snow char t CDR
for (a) Apr, (b) May, and (c) Jun from 1967 to 2017. Anomalies are relative to the average for 1981–2010 and
standardized (each observation is differenced from the mean and divided by the standard deviation and is thus
unitless). Solid black and red lines depict 5-yr running means for North America and Eurasia, respectively. Solid
symbols denote anomalies for 2017.
Fig. 5.25. Mean Apr SWE anomalies for Arctic land
areas calculated from four independent products for
North American (black) and Eurasian (red) sectors of
the Arctic. Anomalies are relative to the average for
1981–2010 and standardized (each observation is dif-
ferenced from the mean and divided by the std. dev.
and is thus unitless). Solid black and red lines depict
5-yr running means for North America and Eurasia ,
respectively, and shading indicates the interdataset
anomaly spread (± 1 std. dev.). Solid symbols denote
anomalies for 2017.
average snowpack in early spring was a precursor to
the above-average snow extent that followed later in
the season.
Despite the long-term decline in Arctic spring SCE
driven by increasing temperature trends, negative
snow anomalies are not consistently observed in every
season (nor in all regions). Off-trend anomalies, such
as those observed in the Eurasian Arctic in 2017, are
driven by natural variability in atmospheric circula-
tion patterns which drive regional temperature and
precipitation anomalies. The rebound in Eurasian
SCE during May and June 2017 was consistent with
winter and spring season circulation patterns which
generally favored colder surface temperatures, en-
hanced precipitation, and above-average snow ac-
cumulation across northern Eurasia.
j. Ozone and UV radiation—G. H. Bernhard, V. E. Fioletov,
J.-U. Grooß, I. Ialongo, B. Johnsen, K. Lakkala, G. L. Manney,
and R. Müller
This report emphasizes the November 2016 to
April 2017 period because chemically-induced loss
of polar ozone occurs predominantly during winter
and spring (WMO 2014). Chemical processes that
drive ozone depletion are initiated at temperatures
below about 195K (−78°C) in the lower stratosphere
(altitude of approximately 15 to 25 km), which lead
to the formation of polar stratospheric clouds (PSCs).
These clouds act as a catalyst to transform inactive
forms of chlorine-containing substances (e.g., HCl
and ClONO2) to active, ozone-destroying chlorine
species (e.g., ClO).
Temperatures in the Arctic stratosphere between
late November and late December 2016 were about
5°C higher than the average temperature of the
observational record (1979–2015); temperatures in
late November 2016 were near the highest values on
record for this period. Temperatures dropped below
the threshold for the formation of PSCs only in late
December. (The onset of PSC formation is typically
in early December with the earliest onsets observed
in mid-November.) Temperatures remained low
enough to sustain PSCs through mid-February 2017.
Starting in late December, modest chlorine activa-
tion was measured by the Aura Microwave Limb
Sounder (MLS). From late January to mid-February
2017, active chlorine (ClO) concentrations were, on
average, 45% higher than the mean concentration
calculated from the MLS data record (2005–16) be-
cause stratospheric temperatures during this period
were about 4°C below average. Ozone decreases via
destruction by activated chlorine started in late Janu-
ary and continued through mid-March 2017. After
mid-March, chlorine was deactivated and chemical
ozone destruction ceased.
Between December 2016 and mid-January 2017,
ozone mixing ratios (a measure of ozone concentra-
tions) were close to the upper limit of values from the
observational record (2004–17) (Fig. 5.26). At the end
of January, mixing ratios started to decline and fell
below average in March and April 2017. However, in
Fig. 5.26. Average ozone mixing ratios (ppmv) mea-
sured by Aura MLS at an altitude of ~18 km for the
area bounded by the polar vortex. Data from 2016/17
(red), 2015/16 (green), and 2010/11 (blue) are compared
with the average (solid white) and minimum/maximum
range (gray shading) from 2004/05 to 2014/15, exclud-
ing 2010/11. Gaps in the record for 2010/11 are due to
missing data.
Fig. 5.27. Area-averaged minimum total ozone col-
umn (DU) for Mar that are calculated poleward of 63°
equivalent latitude (Butchart and Remsberg 1986).
Open circles represent years in which the polar vortex
broke up before Mar, resulting in relatively high values
due to mixing with lower latitude air masses and a
lack of significant chemical ozone depletion. Red and
blue lines indicate the average TOC for 1979–2016
and 2005–16, respectively. Data are adapted from
Müller et al. (2008) and WMO (2014), updated using
ERA-Interim reanalysis data (Dee et al. 2011a). Ozone
data from 1979–2012 are based on the combined to-
tal column ozone database version 3.0 produced by
Bodeker Scientific (
/total-column-ozone). Data for 2013–17 are from OMI.
comparison to 2010/11 and 2015/16 (the years with
the largest chemical ozone loss in the observational
record), mixing ratios in 2016/17 remained well above
values observed in those record years.
The evolution of the Arctic total ozone column
(TOC; i.e., ozone amounts integrated from the
surface to the top of the atmosphere) is used here to
compare 2017 measurements to the observational
record (1979–2016). Specifically, March TOC is evalu-
ated because chemically induced Arctic ozone loss
is typically largest in this month (WMO 2014). The
minimum Arctic daily TOC measured by satellites in
March 2017 was 345 Dobson units (DU), which was
7.7% (29 DU) below the average of the observational
record (374 DU) and 5.4% (20 DU) below the 2005–16
average when MLS data are also available (Fig. 5.27).
Spatial deviations of monthly average TOCs from
historical (2005–16) averages (Figs. 5.28a,b) were
estimated w ith measurements from the Ozone Moni-
toring Instrument (OMI), which is co-located with
MLS on the Aura satellite. Average TOCs for March
2017 were up to 15% higher over the Norwegian Sea,
Greenland, and northern Canada, and up to 20%
lower over northern Siberia relative to the long-term
average (Fig. 5.28a). This spatial pattern is similar to
a recently described Eurasia–North America dipole
mode for the month of February, consisting of a shift
to negative ozone anomalies over Eurasia and positive
anomalies over North America (Zhang et al. 2018).
Monthly average TOCs for April 2017, the month
when the polar vortex (the low-temperature cyclone
in which most of the springtime chemical destruction
of ozone occurs) broke up and air from high and mid-
latitudes started to mix, departed by less than ±10%
from the historical average, and ozone anomalies for
May through November 2017 were unremarkable.
UV radiation is quantified with the UV index
(UVI), which is a measure of the ability of UV ra-
diation to cause erythema (sunburn) in human skin
(WHO 2002). In addition to its dependence on TOC,
the UVI depends on the sun angle, cloud cover, and
surface albedo (Weatherhead et al. 2005). In the Arc-
tic, the UVI scale ranges from 0 to about 7, with sites
closest to the North Pole having the smallest peak
radiation and UVI values < 4 all year. UVI values ≤
5 indicate low to moderate risk of erythema (WHO
2002). UVI anomalies are assessed using satellite in-
struments (OMI) and ground-based measurements,
with the former providing the better spatial coverage
and the latter providing greater regional accuracy
(Bernhard et al. 2015). Figures 5.28c,d quantify the
spatial differences in monthly average noontime
UVIs from historical (2005–16) averages and are
based on OMI measurements. Figures 5.28c,d also
indicate anomalies calculated from ground-based
measurements at ten research stations located
throughout the Arctic and Scandinavia.
Compared to the historical mean, average noon-
time UVIs for March 2017 were larger by up to 25%
over northern Siberia and smaller by up to 20% over
Greenland and the Davis Strait (Fig. 5.28c). Areas
with high UVIs roughly match areas with low TOCs
and vice versa, but UVI anomalies have a larger spa-
tial variability because of their added dependence
on cloud cover. While relative UVI anomalies can
be high, absolute anomalies remained below 1 UVI
unit because solar elevations in March in the Arc-
tic remain low. Anomalies derived from OMI and
ground-based measurements agree to within ±7%.
Anomalies for April 2017 differed by less than ±15%
from the historical average (not shown), except at the
western coast of Alaska and the Bering Strait, where
OMI measured anomalies of up to 50%.
Ground-based UV measurements at all sites var-
ied within historical bounds from July to November.
However, UVIs at Alert, Eureka, and Resolute in
northernmost Canada and at Summit, Greenland,
were unusually high between 15 May and 15 June
despite only small negative TOC anomalies (Fig.
5.28b). At Alert, Resolute, and Summit, positive UVI
anomalies of between 5% and 10% measured at the
ground were in good (±3%) agreement with the satel-
lite data. At Eureka, heavy snowfall in mid-May led
to high surface albedo and high UVIs until mid-June.
Measurements from the ground indicated a positive
Fig. 5.28. (a) Anomalies of TOC (%) and (c) noontime UVI (%) for Mar 2017. (b) and (d) as in (a) and (c) but for
15 May–15 Jun. Anomalies are relative to 2005 –16 averages. Maps are based on OMTO3 Level 3 total ozone
product (Bhartia and Wellemeyer 2002). (c) and (d) also compare UVI anomalies from OMI (first value in
parenthesis) with ground-based measurements at 10 locations (second value presented). Gray shading indicates
areas where no OMI data are available.
UVI anomaly of 25%, while OMI reported a negative
anomaly of −40%. This large inconsistency can be
attributed to systematic errors in the OMI dataset,
which are caused by a mismatch of the actual high
surface albedo and the albedo climatology (Tans-
kanen et al. 2003) used by the OMI UV algorithm. Be-
cause of this mismatch, the high reflectivity observed
from space due to snow was misinterpreted as cloud
cover, resulting in erroneously low UVIs reported
by OMI. The relatively large difference (12%) of UVI
anomalies derived from OMI and ground-based
measurements at Finse, Norway, for the same period
can be attributed to snow cover disappearing 20 days
earlier in 2017 compared with the average snow disap-
pearance date for 2005–16 (Fig. 5.28d). Differences
between satellite and ground-based measurements
at Eureka and Finse illustrate that UV estimates
from space require verification with ground-based
measurements, in particular during months when
snowmelt occurs.
ResearchGate has not been able to resolve any citations for this publication.
ResearchGate has not been able to resolve any references for this publication.