‘Science at its best’ is illustrated by climate modelers’ continuing efforts to fully understand the differing changes in Arctic and Antarctic sea ice in a warming climate.
In his recent book The Climate Casino, environmental economist William Nordhaus writes, “We are rolling the climate dice, the outcome will produce surprises, and some of them are likely to be perilous.” Surprises are already taking place in Earth’s Polar Regions, with human-caused climate change still in its infancy.
Arctic sea ice is melting faster than had been foreseen just a few decades ago. Antarctic sea ice is increasing as time goes by, though questions have recently been raised about the accuracy of the data and whether the trend is as large as generally thought.
These aren’t the only surprises happening in the climate system — indeed, the current “hiatus” in surface temperatures is one surprise few had foreseen, one that this time happens to be in humanity’s favor — but these three are among the most prominent. The immediate availability of the previous day’s measurements of Arctic and Antarctic sea ice extent is perhaps akin to a horse race, but the phenomena will be unfolding over decades and even centuries.
Different Geographies Matter
There is no inherent reason that the Arctic and Antarctic should react similarly in a changing climate. Besides greenhouse gases, sea ice is affected also by changes in winds and ocean currents. In addition, the two regions have fundamentally different geographies: the Arctic is a large ocean mostly surrounded by land, and the Antarctic a large land mass surrounded by ocean.
In the Arctic, amplified warming is expected as the highly reflective sea ice melts and as the darker ocean water absorbs more heat (the “ice-albedo effect”), and these developments have been observed over the last several decades:
By comparison, the Antarctic is a large land mass — it has the highest average elevation of any continent — surrounded by an ocean, which acts as a heat sink. At its yearly maximum, sea ice more than doubles Antarctica’s effective size. Climate models predict only modest melting there compared to melting in the Arctic, but they have mostly done a poor job of simulating the observed changes.
Over 40 percent (by volume) of Arctic sea ice has melted since satellites began monitoring it 35 years ago. Sea ice in the Arctic has been in retreat almost since the beginning of the satellite era (1979); its extent has dropped about 1.5 million square kilometers since then — one-seventh, an area almost three times that of Texas. Although 2013’s minimum sea ice extent was significantly larger (51 percent) than 2012’s record low, there is a “very clear and obvious trend” in Arctic sea ice, says Axel Schweiger of the Polar Science Center at the University of Washington.
(Sea ice can be measured by area or extent; extent is always larger. Scientists mostly discuss extent, because area can be biased low by summer melt ponds, which appear to the satellites as open water rather than as water atop ice.)
Arctic sea ice extent, area, volume, and thickness are all undergoing large reductions, with negative trends for all months. The Arctic’s annual minimum sea ice extent in recent Septembers, for instance, is declining by an average of 84,000 square kilometers a year, an area the size of Utah. That’s a full 14 percent per decade, and that pace is accelerating. The annual maximum extent in late winter has been losing 2.6 percent per decade.
Let’s put that in context: On average, a volume of Arctic ice that would fill Lake Ontario is lost every six years.
Observed sea ice decline seems to be somewhat faster in recent years than what the collection of models predict, but it is difficult to be sure of the discrepancy. “Maybe the trend over the last 30 years has been accelerated by natural variability,” says Schweiger.
In particular, the annual sea ice minimum in September is declining faster than predicted by even the latest models — those from the fifth version of the Coupled Model Intercomparison Project (CMIP5). A 2012 study led by Julienne Stroeve of the National Snow and Ice Data Center (NSIDC) in Colorado found that while CMIP5 models predict lower sea ice than the earlier CMIP3 models, even they are falling behind the observed rate of melting, with 64 percent statistically different from the observed trend (compared to CMIP3’s 85 percent).
In the figure, the thick red line is the observed September sea ice extent, and the thick black line is the average of CMIP5 models — they cross in about 2007, with observations already one standard deviation (thick black line) below predictions. “Trends in Arctic sea ice extent from CMIP5, CMIP3 and observations,” Julienne C. Stroeve et.al. Copyright 2012, Geophysical Research Letters, American Geophysical Union.
While some of the CMIP5 models incorporate more sophisticated sea ice processes, or include methods for allowing melt ponds, Stroeve and co-authors found that conclusions drawn from CMIP5 models differ only marginally from those of the earlier CMIP3 models. Approximately 60 percent of the observed rate of decline since 1979 is from external forcings like greenhouse gases, with the rest resulting from natural variations.
When will summer Arctic sea ice disappear, which scientists usually define as less than one million square kilometers? A few models project as early as 2020, but 32 percent of 56 CMIP5 model runs find it happening by the end of this century. Arctic sea ice cover in 2007 may have passed a tipping point, according to a 2013 paper by Valerie Livina and Tim Lenton, in the sense of an abrupt and persistent change to much greater seasonal variation.
The Antarctic: Without Models … No Data
The story of Antarctic sea ice is nearly the opposite of that of the Arctic: It has famously been increasing for years, at a rate about one-third that of the Arctic sea ice.
This result poses a long-standing puzzle, but just recently questions have been raised about how accurately the data model captures the true extent. (Yes, the data require a model. As Paul N. Edwards writes in his book A Vast Machine: Computer Models, Climate Data, and the Politics of Global Warming, “without models there are no data.”)
The Southern Hemisphere Extent Anomalies (January 2014).
The current dataset from NOAA shows that the annual maximum extent of Antarctic sea ice has been increasing at a rate of about 19,000 square kilometers a year — about 1 percent per decade. The annual minimum is growing at about 5 percent per decade. Ice volume is growing too.
But climate models of southern hemisphere sea ice have generally predicted a decrease. Over the years, various explanations of the increase have been put forward, such as a shift in winds (and thus ice drift) or changes in ocean dynamics. But a slight increase in Antarctic sea ice was predicted in 1991 by the pioneering climate modeler Syukuro Manabe, as an increased supply of water at the near-Antarctic ocean surface slightly lowers sea surface temperatures in his model, leading to more ice (see page 795).
Most models, however, inevitably contain some limitations. A 2013 study in the Journal of Climate by John Turner and colleagues of the British Antarctic Survey concluded “many of the models have signiﬁcant problems in their simulation of the sea ice minimum… .” The difference is even more puzzling, however, because sea surface temperatures in the Southern Ocean have been increasing since the middle part of the last century.
This large discrepancy between observations and models led Ian Eisenman of the Scripps Institution of Oceanography in La Jolla, California, to take a closer look at the data. “I was starting to look into a possible mechanism for the expansion,” says Eisenman, “and was looking into the data as part of the background for this work when I stumbled on this strange inconsistency between papers published before 2007 and those published afterwards.”
Eisenman noticed that in the mid-2000s, the post-1979 trend in Antarctic sea ice was quite low, statistically indistinguishable from zero. But the increases in measured sea ice in the last several years have made the long-term trend positive and statistically significant.
Eisenman started digging into the algorithms used to calculate sea ice. Satellites don’t directly measure the area of sea ice; instead, they receive electromagnetic microwaves from the surface, whether sea ice or sea water. Sea ice emits a greater intensity of microwaves than does an ocean surface at the same temperature. But complicating the data is the fact that a warmer ice-free ocean surface emits more intensely than a cold ice surface.
Fortunately, these emissions vary with the frequency of the microwaves, and with their polarization — the orientation of the electromagnetic wave. These differences allow a distinction between sea ice and open water, but there are many complications resulting from weather effects, melt ponds, thin ice, and other factors.
Over the years, two algorithms, each developed in the 1980s by the NASA Goddard Space Flight Center, have been most responsible for converting microwave data into sea ice concentrations. The most widely used is the “Bootstrap” algorithm developed and implemented by NASA’s Josefino Comiso, which is the basis for discussions in the IPCC’s Fourth and Fifth Assessment Reports, and the one Eisenman and his colleagues focused on.
Digging deep into the data, Eisenman found that the increase in the Antarctic sea ice trend in the mid-2000s was primarily the result of an undocumented change from 1992 in the processing via the Bootstrap algorithm, instead of a real physical increase in sea ice. That was the same year a new satellite was launched with a microwave sensor to replace the old one.
Published by Copernicus Publications and licensed under the Creative Commons Attribution 3.0 License.
However, Eisenman’s analysis cannot determine whether that change introduced an error in the data, or removed one. In their paper he and his colleagues write:
Although our analysis does not definitively identify whether this undocumented change introduced an error or removed one, the resulting difference in the trends suggests that a substantial error exists in either the current dataset or the version that was used prior to the mid-2000s, and numerous studies that have relied on these observations should be reexamined to determine the sensitivity of their results to this change in the dataset.
The science that has been put forth over the years to explain the Antarctic’s sea ice increase will need to be reexamined too:
Furthermore, a number of recent studies have investigated physical mechanisms for the observed expansion of the Antarctic sea ice cover. The results of this analysis raise the possibility that this expansion may be a spurious artifact of an error in the satellite observations, and that the actual Antarctic sea ice cover may not be expanding at all.
NASA’s Josefino Comiso told The Yale Forum he agrees the data are inconsistent before and after 1992. He said that inconsistency helps explain why the earlier trend in Antarctic sea ice is weaker than in a paper he wrote with Fumihiko Nishio in 2008. Comiso says the dataset presented in that paper “should be trusted and not regarded as suspicious,” with a trend using the bootstrap algorithm of +1.44 ± 0.20 percent per decade.
In fact, with that correction Comiso finds a very similar positive trend for Antarctic sea ice in another algorithm sometimes used, the “NASA Team” algorithm. Comiso said he concludes Eisenman et.al. “cannot make the conclusion that is implied in the title of their paper” — that Antarctic sea ice really is expanding.
Eisenman says that some criticisms of the title “raise a fair point,” and that his paper is still in discussion and not yet formally published. He says he plans to adjust the text to acknowledge the likelihood of positive (but lower) trend, in accordance with comments his paper has received since it first appeared as a discussion draft: Some say the paper has brought forth important issues, and others say the trend in southern hemisphere sea ice is still positive, though lower.*
Climate modelers, whether modeling the data or modeling the sea ice’s future in a changing climate, will continue to grapple with these complexities and more, trying as ever to wring “the truth” out of the fundamental observations.
Science at its best.
*This paragraph, third to last, was lightly edited on March 21, 2014.
Sea ice photo credit: National Oceanic and Atmospheric Administration. Photographer: Commander Richard Behn, NOAA Corps.