This post concerns our paradigm of the Arctic Ocean and its Sea Ice. My view, despite years of watching the waxing and waning of ice extents was subconsciously wrong, and others may share the same misconception.

I owe my enlightenment to a great book by Russian scientists from the Arctic and Antarctic Research Institute (AARI) in St. Petersburg. It’s entitled Climate Change in Eurasian Arctic Shelf Areas, by Ivan Frolov et al. The ebook is behind a paywall, but Dr. Bernaerts graciously provided me a hard copy from his library.

The book is small in volume, but rich in information and insights, so I am taking the time to digest. In reading Chapter 4 I came upon a section entitled: Changes in ice exchange between the Arctic basin, marginal seas and the Greenland Sea. Now I was well aware the export of ice through the Fram Strait and knew of the great 2012 storm that so affected extents that year. But then I read this:

There is extensive sea ice exchange between the Arctic Basin and its marginal seas, which are the major sources of new ice for the Arctic Basin. The Arctic Basin serves as a reservoir for the marginal seas; it both receives large ice masses exported from the seas and supplies the seas with thicker multiyear ice. The direction and intensity of ice exchange depends to a great extent on the wind regime. However, local winds alone do not completely determine this exchange of ice. Ice export from the ice cover of marginal seas depends on sea ice conditions in the central Arctic because the sea ice originating from the marginal seas must have some ability to replace the central Arctic ice cover. Thus, the marginal seas depend to some degree on the intensity of ice export from the Arctic Basin to the Greenland and other subarctic seas. However, ice flow from the basin to the seas during onshore winds is strongly restricted by the shoreline and landfast ice, and ocean circulation also influences this ice exchange.

The Great Arctic Cyclone August 2012

It’s Not an Ice Cap, It’s an Ice Blender

Frolov et al made me realize that all our observations of Arctic ice are in fact snapshots of an ice blender constantly moving ice around the Arctic ocean. When we observe and measure extent in one of the seas, that particular ice was not there previously, and will be gone in the near future, replaced to some extent by ice coming from elsewhere. That is the full implication of Arctic ice lacking a land anchor (like Greenland or Antarctica) and existing as “drift ice”.

Figure 4.12. Mean resulting ice-drift pattern for summer (a) and winter (b) during the warm epoch and the difference between ice-drift vectors during the warm and cold epochs for summer (c) and winter (d).

Frolov et al. Provide the statistics regarding the annual dynamics. In the wintertime the shelf seas form “fast ice”, that is ice locked onto the coastlines. Additional ice has nowhere to go but go with the flow north toward the pole or to the neighboring sea. In the summer the flow reverses and the Arctic basin, which received ice from the marginal seas, now sends ice back to replace losses there.

Belugas were observed among West Greenland sea ice. Credit: Kristin Laidre/University of Washington

Which seas get more ice and which get less ice depends mostly on whether the prevailing circulation is cyclonic or anticyclonic. The diagrams show that where there is a strong low pressure area, a cyclonic air flow develops, which moves water and drift ice in a counter-clockwise direction (seen from above). A strong high-pressure system acts in the opposite direction. I like this image the best, but the labels are in French

Considering the Arctic as a whole, a large-scale cyclone such as the massive one in August 2012, breaks up ice, moves it away from western Russian seas, and flushes great chunks of ice south through the Fram Strait into Greenland Sea where they melt. That storm was exceptional in its strength and size, but storms are always at work in the Arctic, and over multiyear periods, we can observe regimes favoring one or the other type of storm.

Frolov et al. point out:

Recent analyses of wind-driven circulation in the Arctic Ocean show that wind-driven ice motion and upper ocean circulation alternate between anticyclonic and cyclonic regimes. Shifts between regimes occur at 5-year to 7-year intervals, resulting in 10-year to 15-year periods. Based on these analyses, these authors proposed an Arctic Ocean Oscillation (AOO) index showing alternation of the cyclonic and anticyclonic regimes.

Table 4.2. Changes in the ice cover area in August from the beginning to the
end of the circulation cycles in Arctic Ocean regions (in 10^3 km2)

Table 4.2 shows that in 88% of cases during anticyclonic regimes, sea ice extent increases in the North European Basin and decreases in the Siberian Arctic Seas, while cyclonic circulation has the opposite effect. The absolute value of changes in the Siberian Arctic Seas is more than 5 times higher than in the North European Basin.

Frolov et al. Summarize:

An increase in the recurrence of cyclonic pressure fields over the Arctic Basin at the transition from a cooling to a warming epoch leads to changes in ice cover deformation processes. The cyclonic systems of the multiyear ice drift contribute to ice cover divergence. This process is most prevalent in summer, whereas in winter, especially in relatively thin ice zones, ice compacting is usually observed. Anticyclonic SLP fields have the opposite effect.

According to Gudkovich and Nikolayeva (1963), in a year that westerly and southwesterly winds increase over the eastern Barents Sea during October— December, the setup they create in the Kara Sea increases ice export from this sea toward the north. Dominant easterly and northeasterly winds produce the opposite result. This study also shows that ice export from the eastern East Siberian Sea and the southwestern Chukchi Sea during the period considered is strongly influenced by wind field vorticity in the vicinity of Wrangel Island. Anticyclonic vorticity increases the ice export, and cyclonic vorticity results in additional ice flow from the north.

Ice transported through the Fram Strait.

Frolov et al:

Table 4.4. Correlation coefficients between the long period fluctuations of the
area of ice exported through Fram Strait (October-August) and total ice area of the Arctic Seas Asian shelf in August for the period 1931-2000 at different time lags.

As shown in Figure 4.14, ice export fluctuations slightly precede corresponding sea ice extent changes in the Arctic Seas. The cross correlation function between the smoothed values of ice export and total sea ice extent exhibits the highest correlation coefficients at time lags (sea ice extent after export) of 4, 5, and 6 years (Table 4.4). Following decreased ice export through Fram Strait in the early 1990s, a tendency for its increase was observed. Based on the time lags shown in Table 4.4, a transition to the phase of increased sea ice extent in the Arctic Seas would be expected at the beginning of the twenty-first century, as confirmed by Figure 4.14a.

figure 4.14a (a) Interannual fluctuations of the total ice area of the Siberian shelf seas in August, and (b) areas of ice exported from the Arctic Basin through Fram Strait. The values of the bold curves are smoothed by a polynomial to the power of 6.

Figure 4.14a shows long-period changes in the total area of ice export through Fram Strait from October of one year to August of the next year for 1931-2000. An approximation of data by a polynomial to the power of 6 (bold curve) indicates the cyclic character of these changes, with the cycle lasting about 60 years. Figure 4.14a shows that the fluctuations of total sea ice extent of the Arctic Seas of the Siberian shelf (from the Kara to the Chukchi Seas) have a similar character.

It is remarkable that increased ice export through Fram Strait is accompanied by increased sea ice extent in the Arctic Seas, contrary to the opinions of those who assume that ice export to the Greenland Sea increases during climate warming, accompanied by a decrease in sea ice extent in the Arctic Seas.

The average drift and current speed in Fram Strait for the preceding year influences the ice exchange between the Arctic Basin and the Laptev, East Siberian, and Chukchi Seas in winter (October—March). The increased ice export to the Greenland Sea contributes to the increased ice export from these seas to the Arctic Basin, and its decrease results in the opposite effect (Gudkovich and Nikolayeva, 1963).

Summary

The estimates above show that, on average, about 1 million km2 of the ice cover is transported annually from the Arctic Seas to the Arctic Basin, which is comparable to current estimates of the area of ice exported annually from the Arctic Basin to the Greenland Sea. (e.g., Koesner, 1973; Mironov and Uralov, 1991; Vinje, 1986). Given a typical ice thickness value, we can estimate the volume of ice exported to the Arctic Basin during a winter to be approximately 1500-2000 km3. This value is about half as large as the available estimates of ice export to the Greenland Sea in winter (Vinje and Finnekasa, 1986; Alekseev et al., 1997), which can be accounted for by ice growth, ice ridging, and other processes that occur during transport of the ice to Fram Strait.

It is a mistake to think of the Arctic as an ice cap that shrinks and grows in extent.  In fact Arctic ice is constantly in flux, more like a kalidiscope than an solid sheet. And the natural forces within the climate system cause fluctuations on a quasi-60yr oscillation

NASA’s Aqua satellite captured this natural-color image of the storm in the Arctic on August 7, 2012. The storm – which appears as a swirl – is directly over the Arctic in this image. NASA image by Jeff Schmaltz, LANCE/EOSDIS Rapid Response.

It’s official–This leap year February is complete and we can now look at the annual Arctic Ice Extent situation at day 60 with two months in the books.

The Resurgence of Arctic ice is continuing in MASIE, the most accurate dataset, but in SII, the remote sensing dataset, not so much.

The MASIE graph shows an extent matching the ten-year average. At 15.02 M km2, 2016 exceeds 2015 annual maximum of 14.91 recorded on day 62, and this year’s peak ice may well go higher.

This table shows comparisons between MASIE and SII

It is readily shown that SII is severely underestimating this year’s growth of ice compared both to SII 2015 and to MASIE. A monthly differential of nearly 600k km2 has opened up due to SII showing a large decline while MASIE shows a gain compared to last year.

Below is a comparison from MASIE regarding the NH seas comprising the NH statistics.

In the table 2016 shows two seas on the Atlantic side lower than this date last year, Barents and Greenland Seas, while the Baltic is much higher, though a smaller size sea.  Barents had grown to almost 600k km2 by day 20, then lost 150k up to day 55, but has now regained half of that loss.

Baffin Bay is down some, but not a large %, while CAA is the same extent.

On the Pacific side, Okhotsk which was the lowest in the last 10 years in 2015 has much more ice now, nearly the highest in 10 years. Bering is also up, so it may be a case of “Goodbye Blob, Hello Normal.”

So is the Winter ending and stopping the ice growth?

Here is my local observation:

Montreal Suburb Street on March 1, 2016

That’s the snowpack on our street seen from my driveway. And I went cross-country skiing today, an activity normally precluded in March by lack of snow cover and temperatures above freezing. With fresh snowfall last night and -13C this morning, it was one of the best days this season. With a blizzard warning and more snow expected tonight, I’m likely to be back out later this week.

So the report from here: The Siberian Express is on time and going strong.

What’s happening with Arctic ice?

It depends on whose measurements you look at. Before you decide, make sure you have read NOAA is Losing Arctic Ice.

MASIE: “high-resolution, accurate charts of ice conditions”
Walt Meier, NSIDC, October 2015 article in Annals of Glaciology.

Something strange is happening in the reporting of sea ice extents in the Arctic. I am not suggesting that “Something is rotten in the state of Denmark.” That issue about a Danish graph seems to be subsiding, though there are unresolved questions. What if the 30% DMI graph is overestimating and the 15% DMI graph is underestimating?

The MASIE record from NIC shows an average year in progress, with new highs occurring well above the 2015 maximum:

Meanwhile from NOAA’s Sea Ice Index (SII) dataset we get this:
The monthly average January 2016 sea ice extent was the lowest in the satellite record, 90,000 square kilometers (35,000 square miles) below the previous record low in 2011.

Why the Discrepancy between SII and MASIE?

The issue also concerns Walter Meier who is in charge of SII, and as a true scientist, he is looking to get the best measurements possible. He and several colleagues compared SII and MASIE and published their findings last October. The purpose of the analysis was stated thus:

Our comparison is not meant to be an extensive validation of either product, but to illustrate as guidance for future use how the two products behave in different regimes.

The Abstract says:
Passive microwave sensors have produced a 35 year record of sea-ice concentration variability and change. Operational analyses combine a variety of remote-sensing inputs and other sources via manual integration to create high-resolution, accurate charts of ice conditions in support of navigation and operational forecast models. One such product is the daily Multisensor Analyzed Sea Ice Extent (MASIE). The higher spatial resolution along with multiple input data and manual analysis potentially provide more precise mapping of the ice edge than passive microwave estimates. However, since MASIE is based on an operational product, estimates may be inconsistent over time due to variations in input data quality and availability. Comparisons indicate that MASIE shows higher Arctic-wide extent values throughout most of the year, largely because of the limitations of passive microwave sensors in some conditions (e.g. surface melt). However, during some parts of the year, MASIE tends to indicate less ice than estimated by passive microwave sensors. These comparisons yield a better understanding of operational and research sea-ice data products; this in turn has important implications for their use in climate and weather models.

http://www.igsoc.org/annals/56/69/a69a694.pdf

The whole document is informative and worth the read.
For instance MASIE is described thus:

Human analysis of all available input imagery, including visible/infrared, SAR, scatterometer and passive microwave, yields a daily map of sea-ice extent at a 4 km gridded resolution, with a 40% concentration threshold for the presence of sea ice. In other words, if a gridcell is judged by an analyst to have >40% of its area covered with ice, it is classified as ice; if a cell has <40% ice, it is classified as open water.

The fact that MASIE employs human judgment is discomforting to climatologists as a potential source of error, so Meier and others prefer that the analysis be done by computer algorithms. Yet, as we shall see, the computer programs are themselves human inventions and when applied uncritically by machines produce errors of their own.

The passive microwave sea-ice algorithms are capable of distinguishing three surface types (one water and two ice), and the standard algorithms are calibrated for thick first-year and multi-year ice (Cavalieri, 1994). When thin ice is present, the algorithms underestimate the concentration of new and thin ice, and when such ice is present in lower concentrations they may detect only open water. The underestimation of concentration and extent of thin-ice regions has been noted in several evaluation studies. . .Melt is another well-known cause of underestimation of sea ice by passive microwave sensors.

The paper by Meier et al. is a good analysis, as far as it goes. In this post I will show the gory details and bring the comparison up to date.

Detailed Comparison between SII and MASIE

Here is a graph comparing SII and MASIE over the last decade and in the last year:

The green line shows the SII deficit to MASIE each month averaged over the last 10 years. The red line shows a cumulative surplus of ice reported by MASIE running through the 12 months, averaged over the last 10 years. Clearly, the graph shows SII underestimates ice extent most of the year, but by September the discrepancy is minimal. Then a huge surplus of ice is reported by SII each October, which results in SII reporting a higher annual extent than MASIE.

But look at what is happening recently.

The blue line shows the SII monthly deficit to MASIE in 2015, while the purple line shows the MASIE surplus during 2015. SII last year underestimated extents more than previously, and with a smaller correction in October, MASIE shows an annual surplus, the cumulative divergence for 2015 is about 2M km2 above the 10 year average.

And in 2016, SII results are increasingly untrustworthy. January 2016 is 450k km2 down, and February (so far) is 600k km2 less than MASIE.

Conclusions

It is unwise to rely on NOAA’s Sea Ice Index as a sole measurement of Arctic ice extent.

The October SII readings are unbelievable, and resemble an adjustment applied to bring the annual results into line.

The MASIE record is good enough for Walter Meier to analyze it with the objective of making SII closer to MASIE’s accuracy.

Addendum:

Here are the tables so you can see the numbers:

Recently when accessing MASIE for the daily ice extent update, I noticed this statement in the left margin of the home page:

NSIDC has received support to develop MASIE ice extent but not to maintain MASIE. We are actively seeking support to maintain the product over the long term. If you find MASIE helpful, please let us know with a quick message to NSIDC User Services.

So I sent them a message of appreciation:

To: nsidc@nsidc.org<nsidc@nsidc.org>;

Update below February 22, 2016

Needless to say, “Ice Free” never happened.  It is true that in the last ten years, August and September monthly extents declined slightly, but the other ten months have increased more than twice as much.  So the over all trend has been slightly upward.

Here is the current image from NASA:

Figure 2: Color-coded map of the daily sea ice concentration in the Northern Hemisphere for the indicated recent date along with the contours of the 15% edge during the years with the least extent of ice (in red) and the greatest extent of ice (in yellow) during the period from November 1978 to the present. The extents in km2 for the current and for the years of minimum and maximum extents are provided below the image. The different shades of gray over land indicate the land elevation with the lightest gray being the highest elevation. Source: NASA

For comparison, here is the ice chart from MASIE:

The comparable MASIE image is showing about 500k km2 more ice than the NASA image. Through mid February, 2016 is following the average winter ice growth over the last ten years, and is greater than 2015 which had a maximum below average. The NSIDC Ice Index is running behind MASIE by about 600k km2.

It remains to be seen in March how this year’s maximum will compare to other years.

Update February 22, 2016

Some additional information on the MASIE ice product:

MASIE: Human analysis of all available input imagery, including visible/infrared, SAR, scatterometer and passive microwave, yields a daily map of sea-ice extent at a 4 km gridded resolution, with a 40% concentration threshold for the presence of sea ice. In other words, if a gridcell is judged by an analyst to have >40% of its area covered with ice, it is classified as ice; if a cell has <40% ice, it is classified as open water.

The passive microwave sea-ice algorithms are capable of distinguishing three surface types (one water and two ice), and the standard algorithms are calibrated for thick first-year and multi-year ice (Cavalieri, 1994). When thin ice is present, the algorithms underestimate the concentration of new and thin ice, and when such ice is present in lower concentrations they may detect only open water. The underestimation of concentration and extent of thin-ice regions has been noted in several evaluation studies

Melt is another well-known cause of underestimation of sea ice by passive microwave sensors.

Meier et al. How do sea-ice concentrations from operational data compare with passive microwave estimates? Implications for improved model evaluations and forecasting

Click to access a69a694.pdf