The future of Arctic sea ice is a concern for various reasons. As seen in global climate model experiments, such as those conducted at GFDL, the Arctic is a region that is projected to warm more rapidly than the global average. [reference: Winton (2006a)]. Arctic sea ice changes can in turn influence aspects of the global climate. And while beyond the scope of these model simulations, other research suggests that Arctic sea ice changes can affect a broad range of factors - from altering key elements of the Arctic biosphere (plants and animals, marine and terrestrial, including polar bears), to opening shipping routes and influencing other human activities. In December 2006, the the Fish and Wildlife Service announced a proposal to list the polar bear as a threatened species under the Endangered Species Act. The associated Dept. of the Interior News Release listed the primary threat to polar bears is the decrease of sea ice coverage due to climate change.

Though there are uncertainties associated with both the 21st century projections of greenhouse gas and sulfate aerosol levels and the coupled general circulation models themselves, these GFDL model results suggest that the Arctic is a region where one can look for potentially dramatic climate change signals in the 21st century. On this page, we show some results from coupled climate model experiments conducted at NOAA's Geophysical Fluid Dynamics Laboratory (GFDL) in Princeton, New Jersey. using GFDL's current state-of-the-art climate model known as the GFDL CM2.1 model.

Note: Not all of the figures presented here have been published in the peer-reviewed literature. However, all of the model experiments from which the figures were derived have been documented in peer-reviewed scientific journals. We request that if you have any questions about these climate model results, please contact us. (Graphics from an older climate model known as the GFDL R30 coupled model can be seen by following this link.)

GFDL CM2.1 model results

Here we present some GFDL CM2.1 model results for the period 1861-2100. The feature shown in the figures immediately below is the sea ice concentration simulated by the global coupled climate model averaged over August, September and October of each year (the months of the year when Northern Hemisphere sea ice concentrations generally are at a minimum). Three years (1885, 1995 and 2085) are shown below to illustrate the trend simulated by the model. A clear reduction of sea ice extent over time is visible, with the rate of decrease being greatest during the 21st century portion of the simulations. (For a related discussion, see Mike Winton's 2006 Geophysical Research Letters paper, Does the Arctic sea ice have a tipping point?) The modeled summertime sea ice trend is not a smooth one, as there is a good deal of year to year variability in the sea ice concentrations. That variability is evident in an animation of the 240 years [mpg 0.7MB].
These abrupt changes are of particular concern for human and ecosystem adaptations and are a subject of much current research. [reference: Winton (2006b)].

• you may view a larger version of year 1885 panel (1920x1080 pixels, 1.6MKB, opens in another browser window)
• you may view a larger version of year 1985 panel (1920x1080 pixels, 1.6MKB, opens in another browser window)
• you may view a larger version of year 2085 panel (1920x1080 pixels, 1.6MB, opens in another browser window)

• Click here for a CM2.1 mpg Sea Ice animation of the 260 year time series 1861-2100 for the CM2.1 Sres A1B simulation [640 x 360 pixels, 0.7MB mpg]
• Click here for a somewhat larger CM2.1 mpg Sea Ice animation of the 260 year time series 1861-2100 for the CM2.1 Sres A1B simulation [2.0MB mpg]

More about the model & the climate forcing scenario used in the simulation

For the years up to year 2000, the model simulation is designed to include most of the major climate forcing factors that were observed to change in the real world (things like changes in atmospheric greenhouse gas levels, volcanic aerosols, soot, tropospheric sulfate aerosols, land surface changes, etc.). The GFDL CM2.1 model has been shown to be credible at reproducing the decade to decade variations in global mean surface air temperature observed during the 20th century, though it tends to exhibit less warming than was observed in the high northern latitudes [reference: Knutson et al. (2006)]. The CM2.1 model simulated Arctic sea ice extent is close to the observed on an annual mean basis, but it somewhat underestimates the summer extent [reference: Delworth et al. (2006)].

In order to explore a range of "If ... Then" future scenarios, several different 21st century emissions scenarios have been used when conducting model simulations at GFDL and other climate research centers. In the CM2.1 figures displayed above, we show results from what is known as the SRES A1B emissions scenario - one with a mid-level increase in 21st century greenhouse gas levels [SRES reference]. We display results from the A1B scenario not because it is considered any more or less likely to resemble the emissions scenario that actually will occur in the coming decades, but rather because, even as a "middle of the road" emissions scenario, the model's summertime Arctic sea ice extent exhibits a fairly dramatic climate change response that is clearly visible in the graphics.

The three panels above and the CM2.1 sea ice animation show how the Northern Hemisphere sea ice concentrations averaged over August, September and October vary in time. In this model simulation, by the late 21st century the Arctic almost becomes ice free during the summer months. But even at year 2100, during the long Arctic winters the sea ice grows back in our simulations (not shown) though it is thinner than that which was simulated for the 20th century.

Note that the CM2.1 model figures shown above depict sea ice concentration (not sea ice thickness). Sea ice concentration is a measure of how much of the area is covered by sea ice, and is shown here for CM2.1 in part because it looks most like what a satellite sees. The colors on the globe range from dark blue (ice free) to white (100 percent ice covered).

The graph on the right side of the frames shows how the Northern Hemisphere summer sea ice extent varies over time in this model simulation. The thick red curve extends from the start of the simulation to the year shown on the globe. (The example displayed to the right was taken from year 2020.) A value of 1.0 on the vertical axis (horizontal line) corresponds to the average Arctic sea ice extent the model simulated for August through October during the twenty year period 1981 to 2000. A year with a value of 1.4 indicates that 40% more of the model's Arctic was covered with sea ice that year than was the case for the average from 1981 to 2000. Similarly, a value of 0.6 indicates a 40% reduction in sea ice extent compared to the 1981 to 2000 average. Note that by the end of the 21st century, the modeled summer sea ice extent in the Arctic is less than 20 percent of the 1981 to 2000 average.

The GFDL CM2.1 coupled model used to conduct the simulations is representative of the state-of-the-art in global climate modeling. This model became GFDL's workhorse model for studies of decadal to century time scale climate variability and change in the spring of 2004, and will remain our workhorse model for a few years.

GFDL CM2.1 Model References & Data Sets

The model experiments from which the figures were derived have been documented in peer-reviewed scientific journals (see below). However, the specific CM2.1 sea ice concentration figures and animations shown above have not appeared in the peer reviewed literature.

Some references:

• Assessment of Twentieth Century Regional Surface Temperature Trends using the GFDL CM2 Coupled Models, Knutson, et al., Journal of Climate, (2006), Vol. 19, No. 9, pages 1624-1651. [4.9MB pdf]
• GFDL's CM2 global coupled climate models - Part 1: Formulation and simulation characteristics, Delworth et al., Journal of Climate, (2006), Vol. 19, No. 5, pages 643-674. [7.2 MB pdf]
• Does the Arctic sea ice have a tipping point?, Winton, Geophysical Research Letters, (2006) 33, L23504, doi:10.1029/2006GL028017 [7.2 MB pdf]
• IPCC Special Report Emissions Scenarios - Summary For Policymakers, IPCC, (2000). [1.2MB pdf]

A more complete list of CM2.1 journal papers documenting the model and its results can be found on our CM2.x references web page. Links are provided that allow one to download copies of several of these papers.

One may download netCDf format model output files from the GFDL Data Portal by following links provided on the GFDL CM2.X Coupled Climate Models web page. The filenames containing the variable plotted above begin with variable name sic (sea ice concentration from Table O1c). Plotted model output for years 1861 to 2000 are drawn from the CM2.1U-D4_1860-2000-AllForc_H2 simulation (also known as climate of the 20th Century experiment 20C3M-run2). Year 2001 to 2100 results are from experiment CM2.1U-H2_SresA1B_X1 (also known as SRESA1B-run1, with CO2 increasing to 720 ppm).

links to some related materials & background info

Note: The United States Government does not endorse any of the non-Federal websites that may be listed on this page.
When you click on any of the links presented below, you may be leaving the GFDL website.

artic sea ice trends: observations & model results

some recent publications & news items

• In a November 20, 2006 NOAA Magazine article titled New Data Show Downward Trend In Arctic Sea Ice the chief scientists of the U.S. National Ice Center is quoted as saying that their newly assembled observational sea ice data sets "show shrinkage in the Arctic Ocean summer ice cover of more than eight percent per decade" over the period 1972 to 2004.
• The National Snow and Ice Data Center has a web page summarizing the State of the Cryosphere: Sea Ice
• On December 12, 2006, the New York Times published an article by Andrew Revkin titled By 2040, Greenhouse Gases Could Lead to an Open Arctic Sea in Summers. That article was based upon climate model results published by Marika Holland of the National Center for Atmospheric Research (NCAR). There are similarities between the NCAR model results and the GFDL CM2.1 results shown here.
• Recent research that combined elements of model results and satellite observations concluded that observations show a statistically significant decrease of Arctic sea ice extent during the past three decades.
  A Model Assessment of Satellite Observed Trends in Polar Sea Ice Extents, Vinnikov, et al., (2006): Geophysical Research Letters, 33, L05704, doi:10.1029/2005GL025282.   [figure correction].
• Evaluation of the Simulation of Arctic and Antarctic Sea Ice Coverages by Eleven Major Global Climate Models, Parkinson, et al., (2005)(2006 correction), J. Geophys. Res., 111, C07012, doi:10.1029/2005JC003408.   [correction].
  
  The two figures above from Parkinson et al. (2006) show the areal distributions of Northern Hemisphere March and September sea ice covers, averaged over 1979 to 2004. On the left (their Fig. 2a) as observed from satellite data. on the right (their Fig. 2l) as computed from the GFDL CM2.1 model. In each case the late summer sea ice distribution (September) is depicted in dark shading, and the late winter sea ice distribution (March) extends over both the dark and light shaded regions.

some earlier publications

• See Thinning of the Arctic Sea-Ice Cover, by D.A. Rothrock, Y. Yu, and G.A. Maykut, Geophysical Research Letters, Vol. 26, No. 23, pp 3469-3472, Dec. 1, 1999.
  Additional information related to this GRL article can be found at the Rothrock et al. Web Page at U. Washington's Polar Science Center.
• The arctic ice thickness anomaly of the 1990s: A consistent view from observations and models, by DA Rothrock, J. Zhang, and Y. Yu appeared in the Journal of Geophysical Research, 108(C3), 3083, doi: 10.1029/2001JC001208, 2003.
• In the 30 August 2002 issue of Science, Richard A. Kerr wrote a news piece entitled Whither Arctic Ice? Less of It, for Sure [PDF] that sought to summarize the level of understanding understanding regarding 20th and 21st century Arctic sea ice changes.