Arguments
Error identified in satellite record may have overestimated Antarctic sea ice expansion
Posted on 1 August 2014 by John Abraham
There has been a lot of attention on ice at the southern pole of the Earth. To be clear, the Earth’s climate is changing and the Earth is getting warmer. This means that the oceans are warming, the atmosphere is warming, sea levels are rising, and ice is melting. In fact, the Earth’s ice is melting almost everywhere. In the Arctic, sea ice is in a long-term retreat, the Greenland ice sheet is melting, so is the Antarctic ice sheet as are the world’s glaciers. But, there is a perplexing anomaly. The sea ice (ice floating on water) that surrounds the Antarctic appears to be growing. Scientists want to know why.
There are many hypotheses, and my colleague Dana Nuccitelli has written about this recently, but here I add a few emerging points. For instance, we know that there is an enormous amount of ice atop the Antarctic ice sheet that is melting each year. Since ice is much fresher than sea water (less salty), the resulting freshwater is creating a fresher zone of water surrounding the continent. The presence of fresh water affects how easily ice can form.
Another view has looked at the quality of the measurements themselves. Could some of the increase be a spurious trend in the measurements themselves? This view was investigated in a very recent publication by Ian Eisenman and colleagues. What the authors found was that a change in sensor calibration caused a shift that has been interpreted as ice acceleration. In the abstract, the authors state,
Specifically we find that a change in the intercalibration across a 1991 sensor transition when the data set was reprocessed in 2007 caused a substantial change in the long-term trend. Although our analysis does not definitively identify whether this change introduced an error or removed one, the resulting difference in the trends suggests that a substantial error exists in either the current data set or the version that was used prior to the mid-2000s… 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 much of this expansion may be a spurious artifact of an error in the processing of satellite observations.
First, we should understand how challenging this problem is. Satellites can measure microwave emission from the ground level. The energy emitted by sea ice differs from the energy emitted by open water. Furthermore, the energy emitted by a surface changes with temperature. As a consequence, it becomes difficult to distinguish between cooler ice temperatures and warmer water temperatures, for example. So, different algorithms have been developed to make these distinctions and the algorithms have changed over time. The image below shows that moving from one algorithm to another (red to blue) caused a shift in 1991 and a significant difference of ice extent from the IPCC AR4 (black square in right image) to the more recent AR5 (black circle in right image).
A change in algorithms leads to a potential spurious increase in sea ice. Figure from Eisenman, Meier, and Norris, distributed under the Creative Commons 3.0 License.
An easier way to view the impact of the algorithm switch is shown below which plots the difference between the methods. There, the 1991 jump is clearly evidence.
Time evolution of differences in ice extent from two algorithms. Figure from Eisenman, Meier, and Norris, distributed under the Creative Commons 3.0 License.
Nice article, but could be much better if NASA's Comiso comment:
"That error has already been corrected and the expansion being reported now has also been reported by other groups as well using different techniques."
was discussed, too. I find these disagreements between scientists great as you can learn a lot.
Thank you to John Abraham for this post. It is a highly relevant topic to address.
I am studying the development of SST of the Southern Ocean(60S-90S). Until now, I been looking at data from GISS, NOAA, Berkeley, NCDC,… and I'm surprised how different trends in SST are. Some point to a decreasing SST other to growing SST.
Additional, I have been looking at Sea Ice Cover(SIC) and beyond the possible errors mentioned by John Abraham, there is an error in 2009. Both errors are calibrating errors by shift of satellite. The latter error has not been corrected on all dataset I can find at KNMI.
What’s my point up to now: - you can’t trust SST and you can’t trust SIC!
I would be pleased if some could point me to the best data for analyzing SST and SIC for Southern Ocean.
Bojan D,
You are mistaken if you think that the comment you linked to is best characterised as a rational contribution to a scientific disagreement.
Note the first substantive paragraph at your link:
"The massive growth of Antarctica’s ice sheets has confounded scientists for years now, as global warming was expected to shrink the polar ice caps. But while the Arctic has shrank some, Antarctic sea ice coverage has broken hundreds of records this year alone. On July 25, the South Pole sea ice reached 436,681 square miles above the 1981 to 2010 average — the 127th daily record for the year."
The more obvious clues that this is propaganda rather than science are the failure to distinguish between land ice and sea ice in the first sentence; the readiness of the author to dismiss Arctic ice loss non-quantitatively as ''shrank some" in contrast to the eager documentation of the square miles of loss of Antarctic sea ice; and what appears to be a fundamentally misleading approach to the tallying of broken records, such that a high extent of ice that persists is counted as multiple daily records. Throughout the rest of the article there is an obvious lack of effort on the author's part to put the modest extent gains into the context of the quantitatively greater losses throughout most of the cryosphere, or to consider, for instance, that melting Antarctic land ice and other climatic changes could be a major contributor to the sea ice extent growth.
You would be well advised to read the many useful links on this site that - in contrast to your link - consider the full planetary budget of ice.
@Leto,
Not sure who are you responding to. I think you are responding to the link I provided, not to the point I tried to make. I provided the link only to back up the citation of NASA scientists.
I'll put this in simple terms to avoid further misunderstanding:
- a group of scientists claim they have found the error,
- apparently, another scientist says the error is known, was already accounted for and doesn't change the results much.
So 'apparently' is the key. What is the story behind this so called refutation? Is it real or another denialist spin? This is what distinguishes great articles from merely good ones.
Leto @3.
Indeed so. That comment @1(along with other more meaty quotes by Josafino Comiso has been cut and pasted quite extensively round the deniosphere. The source of the comment is an item at Live Science which unsurprisingly gives a somewhat different perspective of the Comiso/Eisenman issue.
This Live Science link gives much more background, but I'm struggling to comprehend why the perspective is somewhat different. If anything, it is different not in favour of Eisenman et al.
The full Eisenman et al (2014) paper is available on-line complete with "Supplemental Discussion and Figures."
BojanD @6.
If within your struggles you find yourself able to articulate the substance of that difference "not in favour of Eisenman et al," it may be worthy of further discussion. For myself, I see a significant difference between coverage that quotes Comiso as saying no more than Eisenman's 'error' is a non-issue having been already corrected and coverage that presents the difference between Comiso & Eisenman as but the likelihood Eisenman places on the potential for a 'post-2007' error being easily possible and the likelihood according to Comiso being non-existent. Further, that Comiso describes Eisenman et al (2014) as "misinformation" points to a bit of history between the two of them.
Note, editing mishap in my comment at 3... Should read 'square miles of gain'. The point remains the same. It is clearly a propaganda piece with no attempt to weigh the evidence.
OK guys, all bs aside; Has Antarctic ice extent reached record levels or not? I am getting bombarded by some of my friends that are delialists and I need to be able to make my claim acurately and defend it!
Thx,
Jen.
[JH] Exactly what "BS" are you referring to? The OP?
(Rob P) - It is indeed possible that winter Antarctic sea ice (extent) has reached record levels, but given that the decrease in the volume of summer Arctic sea ice is, on average, 10 times greater over the last several decades, and the combined loss of ice volume from ice sheets, icecaps, glaciers and sea ice in the last decade over 60 times greater, I wonder what they are trying to get at?
jenna @9.
If you are truly after making an accurate claim that you can defend, you will be needing more than a "yes" or a "no" answer. You could help yourself by giving the Eisenman et al paper a quick read. The final figure S11 looks quite fun - ominous for folk who want a simple "Antarctic Sea Ice is expanding" message. Antarctic Sea Ice levels are both rising and falling depending where you look which is why the net Antractic anomaly can shift +1,000,000 sq km in just a couple of weeks. And a quick perusal would also allow you to see the effect that Eisenman et al are agruing is a single step change of ~160,000 sq km which is small beer when compared with recent anomalies. That is probably what you think you're after as it is news on transitory records you're asking about.
It would be fair to say that over the period of continual satellite coverage Antarctic Sea Ice has been increasing. The beef of the Eisenman et al paper is that the 160,000 sq km step change in 1991/2 makes a non-trivial change to calculations of linear trend, which is true. But the trend is still positive and that trend is not so linear in form. It increased in rate from ~2007, just as the negative Arctic trend increased in rate also from ~2007.
Now what has got your denialist chums all aflutter lately is the recent year or so which has seen a very strong positive trend in the Antarctic coinciding with a period of high concentration suppressing negative anomalies in the Arctic. However if you want to rain on their parade, you simply point to the world not having started in November 1978. There are some satellite data prior to that which, with surface observations, have been used to compile HadISST back to 1870. There are a lot of gaps in that record which are bridged by straight lines, but it looks to me pretty conclusive that, just as the 160,000 km doesn't amount to a hill of beans in the grand scheme of things, the recent high Antarctic anomalies are also chicken-feed within the long view. The graph below is compiled from HadISST and taken from Tamino's Open Minds' site. It plots Extent in sq km.
Tamino also did a pre-release item on Eisenman at al (2014). Withn this account Tamino analyses the relative geographical significance of the Arctic and Antarctic trends since 1978. Such a form of analysis was suggested by Eisenman (2010) and shows the Arctic decline over six-times greater than the Antarctic rise, something which may give the odd denialist pause for thought.
jenna @9, the Antarctic sea ice extent has reached a record level for the period 1979-2014. That is often, and incorrectly described as the satellite era. In fact, satellite observations of antarctic sea ice began in 1973, although due to change of instruments and methods, NASA does not show the pre-1979 record alongside the later record. There are further records going back to earlier in the 20th century from whaling records, and also from ice proxies such as the one shown below:
(Source)
The ice core proxy is from a compound released by algae that grow one the underside of the sea ice, and therefore represent a good proxy for sea ice extent, in this case extending back to 1841. The red line is the satellite record from 1973-1994. The sea ice extent has now exceeded the value in the early 1980s, thus representing a record over the satellite era. It does not represent a record over the period since 1841. Indeed, it does not even come close.
A very informative report on antarctic sea ice, including an update of the above graph (6:37 on the video) has recently become available from the ABC. A review of ice core proxies of sea ice extent is also available, and shows relative antarctic sea ice extent for the Holocene (fig 3) and the last 800,000 years (fig 5).
I should note that the updated figure to the one I show above does not include the correction discussed in the OP, and so may overstate the current sea ice extent relative to periods prior to 1994. As such, it may not actually be a record even for the satellite era - but that is rather irrelevant given a fuller record.
Tom@11
You mistakingly linked to Abram 2013 on you local Windows machine (file:///C:/Users/Tom/Downloads). The weblink to it is vailable here.
@MA Rodger, it was exactly this 'misinformation' part that tilted me slightly into "not in favour of Eisenman" perspective.
Even if this 'misinformation' is highly subjective, specially if there is a history between these two researcher, which I was not aware of, there is a case to be made against Eisenman: different methods yielded similar results, at least according to the Comiso (unless they all use his algorithm). Don't we usualy tell deniers that hockey stick is a pattern obtained by different methods and proxies and that it is therefore robust? When Eisenman identifies and provides objective criteria to check whether this error is real, we can definitely have further discussion. But we can all probably agree that even if nothing turns out, Antarctic sea ice is not that terribly important for the whole picture of AGW.
BojanD @14.
Digging out the meat from accounts of this Comiso/Eisenman dispute did have a feel of obscure histiography about it. And the descriptive requirements are quite high for this subject, so slips easily occur. The Live Science account, for instance, presents a sentence that goes way beyond where it should, saying - "According to this view, the trend toward expansion might be an error introduced by a recalibration of how satellite data is processed."
Note that, contrary to your assertion @14, there is no dispute that the error is real. The dispute concerns the potential location of it within the latest algorithms.
My entire point with the "misinformation" comment is that it is a very strong accusation to make against the Eisenman et al paper. And the only support provided for such a strong accusation (Essentially My work is correct because it yields the same result as the work of others unaffected by this dispute [a point you make @14], and anyway it's water under the bridge now the trend has increased.) is not a scientific defence. Thus the accusation is wholly unwarranted and that does need some explanation. Thus I consider that the harsh words could result from a level of animosity existing already between the two of them.
Elsewhere the paper has been described as “an excellent piece of scientific vigilance" and note that account also makes plain that Comiso's calibration error was "inadvertantly introduced" and "corrected ... unknowingly" and the presence of an error only identified by dint of Eisenman et al.
And concerning the water-under-the-bridge dismissal of EIsenman et al (2014), the effect of this undocumented error/correction, while diminishing as new data accumulates, remains still quite large and would have constituted the majority of the trend had not there been a genuine increase in trend to mask it - and note it is a 'masking process' that has grown a lot since the paper was written.
Ok, I see the source of our misunderstanding. You were focusing on the scientific process per se (the fact that error was undetected and even that the fix was undetected, too), I was focusing on the fact that mainstream result is probably still correct, hence my remark about error being real or not. Now I'm not saying that your position is not important. :)
However, now I'm faced with another enigma. If Comiso made an error and the fix inadvertantly, how come the other researchers didn't detect any of this stuff before if they were using independant methods. Surely they didn't independantly made an error and the fix at the same time. Or is the data really that noisy?
☺ On what do you base your view that it is a “fact that (the) mainstream result is probably still correct”? Note that the implication of using the word “still” suggests it has remained correct. Also what levels of probability do you consider to be factual (as to not place some bound on it would make the statement meaningless)?
☻ I also note that you seem to be presuming that I am signed up to the error (which I think we can agree exists) as having been fixed. On what do you base such a presumption?
You'll likely agree it best to clear this matter up before becoming embroiled in the explanation of an actual enigma with all that that may entail.
As for the mainstream result, I was of course refering to the 'corrected' data sets, so the word 'still' was indeed unfortunate. It's tough to get to probabilities, but I can do this.
In favour:
not in favour:
IMO it's likely (more than 68%) that current mainstream results reflect the real trend or that the error, if found, will turn to be a minor one.
The NSIDC discusses this error in their current sea ice report (scroll to the end). They say their data was unaffected by the reported error. The NSIDC seems to feel that it was a small error and that other data, including their own, confirm the amount of Antarctic sea ice. Antarctic sea ice ahs increased over the past three years. The quesiton is whether it is a long term gain or just a short term fluxuation.
BojanD @18.
I am a bit mystified by your comment that it is possible the "the error, if found, will turn to be a minor one." Surely the size of the error in question is not in doubt. Or do you think otherwise?
Beyond that, consider where you place yourself w.r.t. Eisenman & Comiso. Eisenman gives no preference for the error being within either BootstrapV1 or BootstrapV2. Thus his position could be characterised as 50:50. Comiso insists the error is within BootstrapV1 so his position could be characterised as 100:0. The neutral position between the two would thus be 75:25. Your stated position (>68%) could perhaps be considered as centered on 84:16, closer to the neutral position than the Comiso position.
Does that make sense?
Assuming, then that it does make sense @20, that leaves BojanD's "enigma" @16 - Why didn't anybody notice the inadvertent error-&-fix? Weren't all the "other researchers" paying attention?
The IPCC AR4 & AR5 are the two accounts that Eisenman is pointing to as presenting Antarctic trends without noticing. But they don't appear to present these trend figures as being very important.
That is AR4 WG1 Section 4.4.2 doesn't do a great job on describing this trend. 4.4.2.2 says regarding both poles "Different estimates, obtained using different retrieval algorithms, produce very similar results for hemispheric extent" but AR4 is here more concerned with the difference between the Arctic & Antarctic trends than with the level of that 'similarity of result' from the different algorithms when applied to Antarctica. Thus it presents for both poles only the Cosimo data (what we are calling here the BootstrapV1) with the trend from the data to 2005 described "the antarctic results show a small positive trend of 5.6 ± 9.2 × 103 km2 yr–1 (0.47 ± 0.8% per decade), which is not statistically significant ... (at) 90% confidence". What AR4 describes as a "similar result" would include the NASA Team data that I calculate as having a 1979-2005 trend of 13.4 ± 10.0 × 103 km2 yr–1. (pretty much the same as Eisenman et al (2014) fig S6D shows).
However, AR4 provides here no more than a short account with the take-away being the difference between Arctic & Antarctic rather than the nailing of the Antarctic average trend.
AR5 WG1 Section 4.2.3.1 again presents only the Cosimo trend (now the BootstrapV2 trend) but then turns to discuss seasonal trends, and within FAQ 4.1 reasons for the Antarctic trend and the regional differences within the Antarctic.
However both within AR5 WG1 Suplementary Material Section 4.SM.1 (where our BootstrapV1 is given the acronym SBA and our NASA Team is called NT1) and within Eisenman et al (2014) Supplimenatary Discussion & Figures there is surely ample detail to drown any thought of "enigma" concerning the unnoticed change in trend from using BootstrapV1 to BootstrapV2. There is so much more other stuff going on.
@michael sweet, thanks for the link.
@MA Rodger, when you say that the size of error is not in doubt, I'm sure you mean the size of transition from V1 to V2 (or vice versa). However, I was alluding that there might not be a dichotomy, that even if V2 is much better than V1, there could still be a minor mistake lurking in it and I was merely hedging against it.
As for the figure, you're making some unwarranted assumptions about my position regarding the Eisenman vs. Comiso. I've never stated or implied my absolute position, only relative one, a shift in position. In the beginning (starting with my first post) my position was almost aligned with Eisenman, but then gradually moved towards Comiso. And based on this NSIDC link by Michael, even 68% figure, which was very conservative to begin with, is stale, since it's stated in report: "Using the newer version of the algorithm, Antarctic extent trends agree much more closely with the trends from the NASA Team algorithm used by NSIDC." So my 'enigma' is pretty much explained by this and my new figure is probably more like 90%.
BojanD @22.
The transition from V1 and V2 (or V2 to V1) does indeed constitute the error in question. When you say there may be "a minor mistake lurking," do you really mean to suggest there is still something 'lurking' beyond the chosen method being properly applied, another abet minor V1toV2/V2toV1 mistake? I mention this as I can but assume such a harsh meaning is not intended. Yet it is still made.
Regarding your "enigma," it is your creation but its definition as you have described it down this thread has now become itself 'enigmatic'. (And I would add that the Nature article linked @15 had effectively made the "much more closely" comment which was also illustrated in the Supplimentary Discussion & Figures of Eisenman et at (2014) data for SIE (although not the SIA data), so the NSIDC post should not come as some recent revelation.)