San José State University
Thayer Watkins
Silicon Valley
& Tornado Alley

The Decline and Recovery of
Summertime Arctic Sea Ice

There is currently great concern about what may be happening to the Arctic sea ice. The record of area based upon satellite images indicate a downward trend. Al Gore chose to make a dramatic announcement at the Copenhagen Meetings that the Arctic sea ice in the summer could be gone in five years. The operative word in his statement is could. He did not say that the sea would be gone in five years; he said it could be gone. Such statements involving could and might may be true but are generally vacuous scientifically. Their contradictions are also true but vacuous; i.e., the Arctic sea ice in the summer could last forever. Other climate pundits are saying the sea ice will be gone in the summer in 20 to 30 years. Such predictions involve extrapolation of trends. The extrapolation of short term trends is generally invalid but especially so for temperature and other variations like the extent of sea ice which are the cumulation of randomly varying quantities.

Another matter is the obvious flaw in saying that higher atmospheric temperatures are responsible for the melting of the sea ice. If the heat in the atmosphere were responsible for the melting of ice it would not be above normal; it would have been lowered as a result of the transfer to heat from the air to the ice. If the temperature is higher in an area and the ice has melted, it melted as result of something other than the air temperature. Using the amount of heat required for melting ice had to have come from the ocean water rather than the air. The air in contact with the ice simply does not have enough heat energy to supply the amount needed to melt a couple of meters of ice.

Before looking at the statistics on Arctic sea ice in the summertime it is highly informative to look at two pictures from the National Aeronautic and Space Administration (NASA).

The situation on the left likely represents a maximum extent of the summer sea ice because the Earth had just experienced a global cooling episode from about 1940 to 1975. Note that in 1979 there was a gap between the sea ice and the Alaskan and northwestern Canadian coasts of the Arctic Ocean. Such a gap did not exist in 1979 between the north Siberian coast and the sea ice. In 2005 there was such a gap and that was about the only difference between the 1979 condition and the 2005 condition. In order to see this more clearly the following image was created by imposing the 2005 image over the 1979 image and enlarging the result. The sea ice that existed in 1979 but did not exist in 2005 or vice versa shows up as gray.

Although the primary difference was the disappearance of the sea ice off the north coast of Asia there were some slight gains in sea ice around the rest of the perimeter of the sea ice in 2005 over 1979. What is abundantly clear is that there was not a disappearance of sea ice all around the perimeter but specifically the Asian coast. The fact that the sea ice did not disappear around its perimeter indicates that the phenomenon was not due to an increase in global temperature.

It was already clear that the decline in summer sea ice in the Arctic was not due to global temperature because the same thing was not happening in the Antarctic. That left the possibility that the decline was due to hemispheric causes. Some suspected the exhaust from jet aircraft taking arctic flight paths. What the images indicate is that the cause is not due to northern hemispheric developments but something specifically North Asia.

Some possible causes could be changes in the flow of fresh water into the Arctic Ocean from rivers in Siberia. There could be reduced flows from water withdrawal for agriculture or urban use. Or the freshness of the river water could be changing due to pollutants put into the water by industry, urbanizations and/or agriculture.

A North Asian factor that could be involved is that central Siberia has produced an inordinately large share of the increase in average global temperature, particularly at night in the winter. Another entirely speculative possibility is that it might have something to due with the rapid industrial growth in China since 1979.

The energy balance for the Arctic is a unique situation. Near the summer solstice the North Pole receives more radiant energy than any place on Earth. This is because the Sun is relatively high in the sky and shines 24 hours per day. However the Arctic has cloud cover about 70 percent of the time. The clouds reflect away about 90 percent of the incoming radiation. The Sun's radiation that makes it to the surface is reflected away in large part. The ice and snow have a high albedo (reflectivity) and the open water a much lower albedo. Changes in cloud cover could drastically affect the tendency of the sea ice to melt.

The Future of the Decline

Will the trends continue to the point when the summer sea ice disappears altogether? A gap exists between the North American coast and the sea ice but that is not changing. The gap for the North Asian coast has reached about the same dimensions as the North American gap so there is reason to expect that the decline will end in the near future.

From the latest data apparently this has already occurred.

As the graph legend indicates the violet line is the record for 2007, the year in which the september sea ice reached its minimum. The green line is for 2008 and the blue for 2009. For both of those years the sea ice area exceded that of 2007. The black line is for the average for the period 1978 to 2006. The gray area encloses the average plus or minus one standard deviation. Unfortunately that standard deviation is for all months combined rather than the graph showing the variation in standard deviation in the different months. The red line shows the data for 2010 which is available. Note that the red line has crossed the 1979-2006 average. That does not guarantee that the minimum for 2010 will be above the average but it strongly suggests that the low values for 2006 and 2007 were not an indication of an irreversible trend.

For information on the variability of the sea ice area throughout the year see Sea Ice Variability.

The image below is from the National Snow and Ice Data Center (NSIDC) and shows the ice extent for April 20th, of 2010 along with the median limit for April 20th sea ice for the period 1979 to 2000. As can be seen there is not a significant difference between the two limits.

On May 4th, 2010 the NSIDC reported:

During April, Arctic sea ice extent declined at a steady pace, remaining just below the 1979 to 2000 average. Ice extent for April 2010 was the largest for that month in the past decade.

If the decline in the sea ice were a result of the increase in global temperature then a decline in global temperature would bring an end to the decline in summer sea ice in the Arctic Ocean. There is a discernable cycle in average global temperature that goes back to the beginning of the temperature record in 1850. This cycle involves upswings lasting about thirty years and downswings also lasting about thirty years. There is additionally a long term trend of about 0.5°C per century. The global average temperature went through a 35 year decline from about 1940 until about 1975. Then there was an upswing of the cycle and that upswing ended about 2005. The Earth is now in a downswing that will last into the 2030's. For more on this see Cycles and Trends in Average Global Temperature and Their Projection.

The Inappropriateness of
Extrapolating Short Term Trends
of Variables That Are the Cumulative
Sum of Stochastic Terms

The rate of change of the temperature of a body generally is proportional to the net heat energy inflow to that body. When the temperature is at a critical level such as the melting/freezing point or the boiling/condensation point the net heat energy inflow will result in a change in phase within the body instead of a change in its temperature. So the net amount of a substance that changes phase (melts or freezes) is like a temperature change.

The temperature change over a period of time is the cumulative sum of stochastic terms. Such variables appear to have short term trends even when there is no long term trend. The display below illustrates this. Each time the REFRESH icon is clicked the computer generates 60 random terms which have an expected value of zero.

Thus instead of extrapolating the statistics on the extent of the ice what should be analyzed is the statistics on the change in the extent of the sea ice.

In this display there is no indication that there was any significant shift in the structure of the variable in recent years. A regression analysis of the change on time gives a negative coefficient but the t-ratio for that coefficient is only 0.12 and not statistically significant at the 95 percent level of confidence. The coefficient of determination (R²) for the regression is only 0.0006. This indicates that there is no reason to say there is a trend in the annual change in the global sea ice. However the average value for the annual changes over the period is negative, but the t-ratio for that average value is -0.11, which is not statistically significant at the 95 percent level of confidence. For more on the statistical problems for variables which are the cumulative sum of stochastic terms see Statistics.

Update for July of 2010

Although the extent of arctic sea ice in April 2010 was reaching the average for 1979-2000 the statistics since then show a surprising decline.

There is an explanation for how such a dramatic change in the seasonal pattern for the extent of arctic sea ice. Note the statement that the graph shows the area of ocean with at least 15% sea ice. If the proportion of the sea ice in any area drops below 15% the ice in that area drops out of the statistics. Winds from any direction can cause such a change. If winds blow from the north the ice can be dispersed into open ocean where the proportion of sea ice drops below 15%. If the wind blows from the south the ice can be driven north and area of sea ice drops without any melting occurring. Such a wind from the south is likely to increase the temperature but that increase in temperature would have nothing to do with the statistical disappearance of the sea ice. Nevertheless one often sees the statistical decrease attributed to a higher atmospheric temperature. This is falacious for two reasons. First if the air melted ice the air would decrease in temperature. The fact that the air temperature is higher is evidence that the air was not responsible for any melting of ice. Second the amount of heat energy that could defuse into the ice is minuscule compared to the heat of fusion required to melt two meters of ice.

What happened after April is that a differential pressure developed which caused winds to blow from northern Asia driving the sea ice north and coincidentally raising the atmospheric temperature. Thus the supposed decline in arctic sea ice is statistical artifact having to do with winds and the way the extent of sea ice is measured. It casts doubt on the relevance of the official statistics for assessing climate changes.

The pictures of the extent of ice are informative. Two recent ones shown below are for July 12th and July 24th.

Note the disappearance of the ice in a region off the north coast of Asia but not much change off the north coast of North America. If the disappearance had to do with temperature change in the Arctic or the extent of first-year ice there would be similar changes in both areas. If the statistical disappearance had to do with a south wind from Asia blowing dispersed ice floes north then there would be changes off the coast of Asia but not off the coast of North America. Since it is the latter that is occurring the evidence indicates not the melting of arctic sea ice but the compacting of dispersed ice.

Note also the absence of ice in Hudson Bay. The line for the median extent of ice indicates that Hudson Bay is included in the statistics. Now consider the picture for June.

This picture shows Hudson Bay nearly filled with sea ice. First of all it is perplexing that Hudson Bay ice and ice between the Canadian islands is included with the Arctic Ocean sea ice. The statistics on the area of arctic sea ice are presented to the public as though they represent strictly Arctic Ocean ice. What is indicated about is that the official statistics on the area of arctic sea ice are subject to influences that have very little to do with the Arctic Ocean sea ice.

The variability of the extent of arctic sea ice varies with the season. The graphs from the NSIDC display ±2stanadard deviations limits based upon the annual variability. Shown below are the 95 percent confidence limits (±2stantdard deviations) throughout the year, starting with July 1st.

As shown below, even the record low of the year 2007 did not deviate from the long term average by more than two standard deviation units.

This means the levels for 2010 are less than two standard deviation units from the long term average and thus not statistically different from the long term average at the 95 percent level of confidence.


Something is happening to the statistics on the extent of Arctic sea ice in the summer but it is not global and it is not hemispheric. The ice is not melting around the perimeter of the Arctic Ocean ice cap. Instead it is sometimes declining off the north coast of Asia and at other times Hudson Bay. The source of the statistical decline is therefore likely located sometimes in northern Asia and sometimes in northern Canada. However the future decline is not likely to go much further. The decline in 2005 opened a gap along the north coast of Asia that is comparable to the gap that has always existed along the western part of the north coast of North America in the summertime.

Since 2007 the extent of arctic sea ice has been increasing and, as of April of 2010, it was near the average for the years 1979 to 2000. Winds are currently changing the area with at least 15% sea ice without necessarily affecting the amount of sea ice in the Arctic Ocean.

HOME PAGE OF applet-magic
HOME PAGE OF Thayer Watkins