Arctic Sea-ice Extent at a Record Low

Press release from

Figure 1: Mean sea-ice extent in the Arctic in October 2019.

21 November 2019

Although the summertime minimum for the Arctic sea-ice extent has come and gone, and the freezing season has begun again, sea-ice growth is nevertheless at a historical low. According to analyses from the University of Bremen’s Institute of Environmental Physics, the mean sea-ice extent in the Arctic this October was only 5.44 million km², which is more than 443,000 km² below the previous minimum year, 2012 (Figure 1). Other analytical algorithms, like those used by the National Snow and Ice Data Centre (NSIDC), can vary slightly (see the article from 2018). This is an area the size of Morocco or 1.24 x the size of Germany. As Christian Haas, Head of the Sea Ice Division at the Alfred Wegener Institute, explains: “Since the beginning of continuous sea-ice observation, we’ve never seen so little sea ice in the Arctic at this time of year. It’s a troubling process, most likely caused by the increased heat input from the ocean and the atmospheric circulation in the Arctic.”

Figure 2: Mean October sea-ice extent in the Arctic from 1979 to 2019.
Figure 3: Difference in mean ice margin position in October 2019 compared to the mean value for October 2012. Regions marked in blue have more sea ice than the reference period; those marked in red have less.
Figure 4: Daily sea-ice extent in the Arctic to 20 November 2019 (red). For the sake of comparison, the respective extents in 2007, 2012, 2016, 2017 and 2018 are also depicted, together with the long-term average for 1981-2010 (grey) and the span of two standard deviations (light grey).

The sea-ice extent in the Arctic is an important indicator for the effects of climate change in the polar regions. Here the sea ice is manifesting a negative climatological trend that reflects the increasing warming of the Arctic. Since the beginning of continuous satellite observation, the October sea-ice extent has declined by ca. 34 percent, and reached its lowest extent this year (Figure 2). Especially in the northern Canada Basin, in the Chukchi, Siberian and into the Laptev Sea, there was considerably less ice than in the comparison year 2012, the year with the lowest summertime sea-ice extent to date (Figure 3). The observed changes in comparison to other years are discussed in more detail below. As can be seen in Figure 4, in the years 2016 and 2018, too, the sea-ice growth began late, producing low sea-ice extents in October / November. Potential causes for the weak ice growth this year include the unusually high ocean surface temperatures, which were up to 4 °C above the long-term average (1971 – 2000) in major expanses of the Arctic Ocean (Figure 5). Here it will take a prolonged cooling phase to reduce the stored heat and drop the ocean surface temperature to the freezing point. In addition, atmospheric temperatures up to 6 °C higher than the long-term average were dominant in the Beaufort Sea (see Figure 5). To the north of Greenland the air was also unusually warm, which slowed ice growth. In Figure 6, an integrated, vertical depiction of atmospheric temperatures in the Arctic across all lines of longitude, we can clearly see how the high ocean surface temperatures contributed to a warming of the air masses above. This is especially apparent in the area between 73°N and 80°N, where the temperature anomaly (deviation from the climatological mean, 1971 – 2000) was ca. 6 °C. This area largely corresponds to the position of the ice margin, especially in the Beaufort Sea and Chukchi Sea. Since the humidity was also especially high in these areas (see Figure 7), the formation of new ice was made far more difficult.

Figure 5: Sea surface temperature anomaly (left), pressure anomaly at sea level (middle), and temperature anomaly in °C at 925 hPa (right) for October 2019 in the Arctic, in comparison to the long-term average for 1981-2010. Source:
Figure 6: Integrated zonal temperature anomaly (in °C) in the Arctic from the surface / ocean surface to the tropopause (reference period 1971-2000). Source:
Figure 7: Integrated zonal specific humidity anomaly (in %) in the Arctic in October (reference period 1971-2000). Source:

Analysis and Significance of Sea-ice Formation in October

When discussing individual years in comparison to climatological changes, the considerable influences of year-to-year changes (interannual variability) must be taken into account. Especially in high latitudes, these variations can be so much more pronounced than the actual climatic changes that climate changes essentially become invisible. This effect is especially prominent in the Arctic – due in part to the ice-albedo feedback. In this regard, climate researchers speak of a poor ‘signal-to-noise’ ratio – the climate changes (the signal) are large, but so are the interannual and decadal variations (produced by ‘internal variability’; the background noise). This often produces ostensibly contradictory conclusions when examining changes in individual years.

An analysis of the sea-ice conditions in October that takes into account the ‘signal-to-noise’ ratio has been conducted by Dr Frank Kauker, a sea-ice physicist at the AWI. Using model studies, Kauker compared e.g. the number of non-freezing days (NNFD) from 1 September to 31 October with a climatological mean, and with the number in selected individual years. In this regard, the NNFD also roughly represents the time at which freezing begins (1 September plus NNFD): however, the latter is difficult to define, since early in the winter, freezing phases often alternate with warm phases, until prolonged wintry conditions finally set in. Principally speaking, the NNFD can also be calculated using the changes in the observed ice concentration; however, the basis of the calculation is the change in sea-ice extent. In the high-resolution sea-ice model employed at the AWI, not only the sea-ice extent but also the change in ice volume – which is used here to determine the NNFD – is simulated. This parameter can’t be determined using observations, since Arctic-wide monthly ice information can only be gathered from October to April. Figure 9 shows this number for 2019 as an anomaly in comparison to the climatology for the period 1980 – 2019, and for the years 1990, 2007, 2012 and 2015 – 2018.

Figure 8: Sea-ice age from 22 to 28 October 1985 (upper left) and 2019 (upper right), together with a time series of age groups for the corresponding week from 1985 to 2019 (below) from the NSIDC’s EASE-Grid Sea Ice Age, Version 4. (Source:
Figure 9: Difference in the number of non-freezing days (NNFD) in the period 1 September to 31 October 2019, compared to the climatology of the years 1980 to 2019 (upper left ‘mean’), and to the years 1990, 2007, 2012 and 2015-2018.

Compared to the climatology (Figure 9, upper left-hand corner), we can see higher NNFD values throughout the Arctic in 2019, particularly in / to the north of the Chukchi Sea, Siberian Sea and Laptev Sea, where the NNFD reached up to 40. Yet a comparison with 1990 shows that the NNFD isn’t higher everywhere in 2019. 1990 was selected because it had the highest mean NNFD for the entire Arctic Ocean prior to 2005 (Figure 2 also shows a low value for the sea-ice extent in October). With regard to the changes in sea ice age shown in Figure 8, the year 1990 shows a significant reduction in the amount of multiyear ice, from which the sea ice didn’t recover. The next year considered is 2007, in which the sea-ice extent in September was more than 1 million km² below the long-term average, and remained extremely low in October (see Figure 2); but the sea ice age (Figure 8) shows another major loss of multi-seasonal ice in 2007, after which the level remained low. In comparison to 2019, the area responsible for the low sea-ice extent in 2007 was to the north of the Chukchi Sea and Siberian Sea. Yet this area is characterised by a significantly higher NNFD than 2019 (Figure 9) – all other areas show a lower NNFD than in 2019. 2012 shows a similar distribution to Figure 3; to the north of the Chukchi Sea, Siberian Sea and Laptev Sea, the NNFD was higher. The comparison with the years 2015 and 2017 is fairly unspectacular in this regard, because the NNFD is only higher than in 2019 in very small areas. A comparison with 2018 reveals an east-to-west distribution – whereas in (very small) eastern areas, the NNFD was higher than in 2019, the NNFD was substantially lower in the Beaufort Sea. In other words: it’s not the Chukchi Sea, Siberian Sea and Laptev Sea that make the difference; rather, it’s the Beaufort Sea.

What explains the persistent low Arctic sea-ice extent in October? A comparison with the climatology doesn’t yield any clear answers: in 2019, the NNFD was far above the long-term average, but that’s also true for nearly all recent years. A comparison with 2012 shows the Chukchi Sea, Siberian Sea and Laptev Sea to be the areas with a higher NNFD (i.e., with delayed ice growth). Yet a comparison with 2018 shows that the conditions in these areas were very similar in 2018 and 2019 – which tells us that the Beaufort Sea is the decisive factor.
Sea ice age

Sea Ice age

Sea ice that survives after the September minimum is referred to as second-year or multiyear ice. Figure 8 shows that the majority of the ice is second-year ice (‘1-2 years old’). It was formed in the fall/winter 2018/19 and survived this summer. In addition, there are roughly one million km2 of ice that formed after the September minimum (‘0 to 1 year old’). It can be clearly seen that there is far less older ice than in the past: the area of ice that is at least two years old is only a third as large as in the mid 1980s, and only half as large as in the mid 2000s.(Source: NSIDC)

Figure 10: The research ship RV Akademik Fedorov while working on an ice floe in October 2019. (T. Krumpen)

It’s been nearly two months since the MOSAiC expedition departed from Tromsø, and the first everyday routines have now set in at the drift camp. Dr Thomas Krumpen, expedition leader for the research ship RV Akademik Fedorov, which accompanied RV Polarstern at the beginning of the expedition, has now returned to Bremerhaven and reports on the lengthy search for the ideal floe for Polarstern’s drift, and on the first steps in installing the monitoring instruments for the ‘distributed network’ all around the icebreaker. Describing the search efforts on site, the sea-ice physicist recalls: “The ice conditions in the planned starting region for the drift experiment were extremely difficult, since there were only very few suitable floes that matched our requirements for thickness and size.” Nevertheless, the distributed network encircling RV Polarstern was set up in just a few days, with the aid of helicopters and the hard work of everyone involved. Surveys of the ice conditions, which involved gathering ice cores and using sledge-mounted electromagnetic (EM) sensors, indicated a modal ice thickness of 70-80 cm, although in many cases the solid ice was only 30-40 cm thick, and rested atop a layer of rotten ice, permeated by seawater.

In keeping with the original plan, extensive observatories were set up on a total of three larger ice floes, while smaller monitoring units and autonomous buoys were installed on eight smaller floes. In addition, 53 GPS buoys were deployed around the MOSAiC base in a radius of 40 to 100 kilometres to monitor the large-scale drift of the ice and observe deformations in the ice field, which can produce shearing and the formation of new leads or ice hummocks. In the second half of November, the first new team will set off to relieve the colleagues currently on site following the exhausting initial setup phase. The animation below shows the first results from the buoys deployed (see Figure 11).

Figure 11: Animation of changes in position for the ice buoys deployed near the MOSAiC camp (centre of image). Source: Animation prepared by Daniel Watkins.

Prof Christian Haas (AWI)
Dr Thomas Krumpen (AWI)
Dr Frank Kauker (AWI)
Dr Renate Treffeisen (AWI)
Dr Klaus Grosfeld (AWI)

You have any further questions?

Date of publication: 22 Nov 2019

Link to sender’s website

Announcements are published as a service to readers. The sender is responsible for all content.

Announcements for publication can be submitted to

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.