NSIDC vs. NANSEN vs. AMSR-E

Guest post by Steven Goddard

In my May 1st piece, Dr. Walt Meier at NSIDC hypothesized that differences in algorithm between NSIDC and NANSEN (NORSEX) were causing the gap between the NSIDC interpretation of normal and the NANSEN interpretation of normal, as seen below.  They use different baseline periods which introduces some difference – but the discrepancy should go the other way due to the fact that the NANSEN base period (1979-2007) includes more low ice extent years from the current decade.

So I tried an experiment to test this out, where I overlaid NANSEN on top of NSIDC for the entire winter – and found that they are nearly identical.  This would tend to discount the theory that differences in the algorithm are to blame.  It appears from this more likely that one or the other has an error in their historical database which is affecting the interpretation of “normal.”  Dr. Meier has stated that he is confident about the accuracy of the NSIDC database.

Using a third reference point, I tried another experiment comparing NSIDC (blue) vs. AMSR-E (red) and did see something interesting. Starting in late March, NSIDC (in blue) began to show more ice than AMSR-E (in red) – which uses a different satellite.

Last winter, the SSM/I satellite used by both NSIDC and NANSEN began to degrade, as reported by NSIDC and WUWT (of course.)

http://nsidc.org/images/arcticseaicenews/20090217_Figure2.png

The degradation caused the ice extent to appear lower than AMSR-E, but now the problem seems to be going the other way – with SSM/I showing more ice than AMSR-E. What does it all mean? Given that NSIDC and NANSEN seem identical this year, I don’t think this explains the discrepancy in their baseline.  It does appear that there is still an error in the SSM/I data however.

UIUC has quit posting their SSM/I images because the quality has become so bad.

February 25, 2009 – The SSMI images for many days in 2009 were bad enough that we removed them from this comparison display (see note below and the NSIDC website). There is enough interest in these side-by-side comparison images that we will try to replace them with corresponding images from the AMSR-E sensor in the coming weeks.

AMSR-E has only been around during the current decade, so they are not able to provide long term means.  However, current ice extent is highest on record for the date.

http://www.ijis.iarc.uaf.edu/seaice/extent/AMSRE_Sea_Ice_Extent.png

Advertisements

53 thoughts on “NSIDC vs. NANSEN vs. AMSR-E

  1. Makes me wonder what other satalites are breaking down?
    Any developing problems with the SST measuring satalite (like the notorious bering sea hotspot in the NOAA charts), what about the one that gives us global average temps?
    If you can’t trust GISS and the current satalites start to have problems like the one that measures sea ice, maybe it’s time we have new cutting edge technology take the stand and get the best possible data, what do you think?

  2. Since were looking at the various lines wondering where they come from….
    I have one question. What is casuing the bump in each trend at the start of June?

  3. My own calculations from the raw gridded data showed that the arctic is currently well above the mean. My mean included the whole 79-09 record.
    http://noconsensus.wordpress.com/2009/05/03/global-sea-ice-nears-record-high/
    My data is considered preliminary from the NSIDC but the NOAA satellite causing the problem is removed and NOAA 13 is in its place. I don’t understand how my own mean can be that much higher but I believe it’s correct.
    To my knowledge not too many others outside of paid climatology have worked independently from the raw gridded sets.

  4. NSIDC released their update on April Arctic Sea Ice Extent today:
    http://nsidc.org/arcticseaicenews/
    It seems reasonably factual and starts out with the obvious, “Arctic sea ice extent declined quite slowly in April; as a result, total ice extent is now close to the mean extent for the reference period (1979 to 2000).” it also mentions that, “The decline rate for the month of April was the third slowest on record.”
    It seems that they may be hoping that, “The thin spring ice cover nevertheless remains vulnerable to summer melt.”, but they may be reading the updates by the Catlin team versus the PAM-ARCMIP Polar 5 mission. Time will tell who’s right…

  5. Adam from Kansas (18:46:55) : If you can’t trust GISS and the current satalites start to have problems like the one that measures sea ice,
    Since GISStemp use simulated temperatures for the arctic that are based on sea ice estimates, I would speculate that the satellites screwing up the sea ice estimates also screw up GISStemp anomalies…

  6. Barry L 18:50:55
    That is the standard time for switching to a slightly different algorithm for calculating the values, because the summer has a greater amount of standing water on top of ice, which can confuse the sensors.
    =========================================

  7. What I’d like to know is why they count ice in the Sea of Okhostk as “Arctic Sea Ice Extent” and is not inside the Arctic Circle or connected in any way to the Arctic or even the Bering Sea… This is really a huge trick they use to gin up melting of the “arctic” when it really isn’t. What can we do to make a stink about this?

  8. Just for grins, I did a Google search of “2009 record snow” and got:
    Results 1 – 10 of about 11,000,000 for 2009 record snow. (0.24 seconds)
    Here, you try it. Somehow I don’t think we’re having a snow shortage…
    Even if some percentage of those 11 million pages are things where the 2009 is not a year marker, there is still a lot of snow falling somewhere…

  9. When Dr. Meier mentioned “differences in algorithm,” I assumed it mean -historical- differences, not current differences. Because the current data are very visually similar.
    IOW: How does the 1979 data for the two methods compare?

  10. Here is a nice comparison of where the NH sea ice extent in 2009 is compared to other years (from 1979 to 2009) and the two lowest minimum sea ice extent years (2008 and 2007) and the two highest minimum sea ice extent years (1980 and 1996).
    This chart uses the NasaTeam algorithm which is slightly different than the current algorithm used by the NSIDC but it still gives a nice overall picture of the situation.
    2009 is pretty close to the average right now (but there is still a lot of time left before the minimum extent happens around September 10th).
    http://img377.imageshack.us/img377/6448/dailyseimay3.png

  11. What are the possibilities that certain satelite detectors are more susceptible to cosmic rays?

  12. There are a lot of factors that make it likely that the May report will show less melting than average
    1) The report discusses the average for the month. All reports of this kind are inherently stale — by the time the data is published you know whether the month finished strong or weak. This gives you a lot of advance notice on the following month. We already know the data through May 3rd and May got a really high start.
    2) The sea ice deficit (relative to the 79-00 average) in the sea of Okhotsk will disappear and will not affect the sea ice in other areas.
    3) The weather forecast for Barrow, AK is NOT GOOD for melting sea ice.
    http://www.weather.com/outlook/health/skin/tenday/USAK0025?from=36hr_topnav_skin
    and, you’re welcome for the cute beach scene at the top. Really makes you think of Barrow, eh?
    Since were on the topic of comparing graphs, the AMSR-E graph clearly shows that the April AVERAGE sea ice for 2003 was higher than 2009. (Compare the dark blue line to the red)
    http://www.ijis.iarc.uaf.edu/en/home/seaice_extent.htm
    But NSIDC says other wise.
    http://nsidc.org/images/arcticseaicenews/20090504_Figure3.png
    WUWT?

  13. Dr. Meier needs to check his data. In the published .txt data , there are missing data points (not many, but a couple) it skips from 1987-1989 at least twice. They need to go back and validate what they publish.

  14. Briefly, from my earlier comment on the Mayday – May Day! post:
    http://wattsupwiththat.com/2009/05/01/mayday-may-day/#comment-124916
    Perhaps, in December 2008, when NANSEN seems to have adjusted their algorithm and posted a change to their ice data effective 22 October, they also massaged the data back to 11 September and then stopped. Might it be possible that if they had taken the time to go back through all their older data that their average would be lower than it is now?
    In other words, what if the NANSEN 1979-2007 average itself is the problem, that it is based on a faulty way of measuring the ice that wasn’t corrected until December 2008, and only the data back to 11 September was corrected and none earlier? Seems a stretch I know, but what if no one really looked at the way they were doing things until last year when the NANSEN ice extent line was coming up on their mean and only then did they notice a problem with their data processing procedures.
    Speculation is easy; only the folks in Norway know for sure.

  15. By the way. Dr. Meier, needs to look at the trend calculation. Last month…it showed a “-2.7% decline per decade”…an ever increasing gain the past year… Yet all of a sudden…with an above average daily GAIN the past month…now he is showing a -2.8% decline for March. He needs to review the data. How can bouncing ABOVE last year’s data for the month of March, now equate to losing the upward trend?
    Am I missing something? Please correct me by pointing me to the rigth data series they publish as .txt files. I converted to Excel to plot my own trend which does not agree with theirs.

  16. Frederick Michael (20:09:19) I assume you mixed your dates and meant that 2009 is greater than 2003, and it appears so in both graphs you link to.

  17. Mike Strong (20:31:04) :
    … How can bouncing ABOVE last year’s data for the month of March, now equate to losing the upward trend?

    The April trend should be less steep in 2009 than it was last year, as the new point is a high one. But comparing April to March isn’t so simple. Expect the trend to get steeper as we go into summer. The melting the last few years was greater in the summer — and the denominator in the % calc is smaller then too.

  18. Mike Lorrey (19:36:30) :

    What I’d like to know is why they count ice in the Sea of Okhostk as “Arctic Sea Ice Extent” and is not inside the Arctic Circle or connected in any way to the Arctic or even the Bering Sea… This is really a huge trick they use to gin up melting of the “arctic” when it really isn’t. What can we do to make a stink about this?

    Perhaps you should insist they be more consistent in their terminology. At http://nsidc.org/data/seaice_index/daily.html they use the phrases “Northern Hemisphere (Arctic)” and “Southern Hemisphere (Antarctic)”. In the 1970s sea ice formed between Cape Cod and the islands of Nantucket and Marth’s Vineyard one year. In order to have that counted the next time it happens, I suggest you insist that NSIDC measure all Northern Hemisphere ice. Since that’s what they do, everyone will be happy.

  19. Leon Brozyna (20:15:35) :
    Briefly, from my earlier comment on the Mayday – May Day! post:
    http://wattsupwiththat.com/2009/05/01/mayday-may-day/#comment-124916
    Perhaps, in December 2008, when NANSEN seems to have adjusted their algorithm and posted a change to their ice data effective 22 October, they also massaged the data back to 11 September and then stopped. Might it be possible that if they had taken the time to go back through all their older data that their average would be lower than it is now?

    They didn’t change their data at all, here’s the before and after graphs (it blanks back and forth between the graphs):
    http://wattsupwiththat.files.wordpress.com/2008/12/nansen_sea_ice_extent1-520.gif
    However on the second graph which compares this graph with the mean the latter part of the graph was in error (and therefore didn’t match the above curve). At the date in question the latter graph was corrected and thereafter matched it’s partner. It’s unclear to me why an error in fall 2008 would effect the 1979-2007 average in any case, however since the base data was unchanged that’s moot.

  20. Steve Keohane (20:41:56) :
    Frederick Michael (20:09:19) I assume you mixed your dates and meant that 2009 is greater than 2003, and it appears so in both graphs you link to.

    I make lots of mistakes but not this time. For the month of April, the dark blue line starts way above the red line but ends below. On average, the blue line (2003) is higher. However, the NSIDC comparison (of averages) shows 2009 just a hair higher. That’s why I wondered what’s up.
    By the way, the differences are so tiny, it may not be all that mysterious.

  21. Regarding the comparison of NSIDC, Nansen and JAXA current values are ~13.8, 13.1 and 13.0 respectively which reflects the differences that have existed through the winter. That is Nansen and JAXA are very close and NSIDC somewhat higher. The real outlier is DMI which has a much smaller range, currently ~10.3.

  22. Phil,
    There is a fixed Y-offset between NSIDC extent and NANSEN extent. That has been normalized in the graphs above. Since you can’t make direct numerical comparisons between them, the point of this exercise is to show the relative deviations (or lack of them) over time.

  23. Frederick Michael 21 36
    You are right. Using the data from the JAXA site gives an average 2009 April extent of 13.58 m sq km compared with average 2003 April extent of 13.65 m sq km.

  24. Oh! No sunspots today.
    So much for the Watts Effect.
    I guess it only works for Gore.

  25. Well, NSIDC’s May update essentially shows April trend flat since 1989. But then it did last year too, before the “1st year ice” phenom hit in summer. I’ve still got my money on Arctic minimum being right around the 2005 line, which would still be Good News Going in the Right Direction (unless you’re a panic-monger, of course).
    I find it interesting they are now claiming +/- 2 standard deviation as their claimed “natural deviation” standard. Have they done that before? The suspicious side of me is wondering if they are doing so both to avoid noting that the current line is actually within 1 standard deviation, and also to give themselves more room on the upside to still claim “natural deviation” if/when the current trend line crosses above the long-term trend line.

  26. Is there an error in the first graph? The NANSAN data uses the ice area up to ’07 so it should have a lower mean. But the 1st graph shows NSIDC having a lower mean.

  27. Forgive me, but how can we accept the data from the NSIDC as uncorrupted and reliable? Consider these gems from their website:
    “Arctic sea ice has declined dramatically over at least the past thirty years, with the most extreme decline seen in the summer melt season. ”
    Their data do not support this assertion. I see no visible trend line from 79 – 97. The decline was certainly not ‘dramatic’. We do not have complete records of arctic sea ice prior, so we cannot comment on what happened earlier more than 30 years ago; doing so is misleading. There was a substantial negative trend 1998 through 2008; with substantial recovery in the past year. From the data, that’s pretty much what we can say.
    Here’s another one: “For sea ice extent data, the standard deviation is computed for each day of the year from the extent on that day over the 22 years of the average climatology, 1979-2000.” This is akin to saying, lets take only men taller than 5’8″ and shorter than 6’0″, and compute the mean and stnd deviation using this pool. As a result, anyone 5’7″ or 6’1″ will almost surely be two standard deviations outside the norm. But the chosen pool is entirely arbitrary and the standard deviation is meaningless.
    How can they expect me to accept the impartiality of the data or algorithms when their statistical technique is so poor and their bias so strong? Is the University of Colorado really this feeble?

  28. Steven Kopit: Is the University of Colorado really this feeble?
    That could be it. It could also be that your interpretation is lacking. Is that an option for you?

  29. The two datasets, NSIDC and AMSR-E for ice extent are very close, show you how close in a moment.
    In a recent email I said
    “Making a rash prediction based on known amsr-e daily data so far this
    month, April, the month northern extent data you will be producing will
    be 14.55 +-1%”
    And I see the published figure is 14.58
    I’ve not updated AMSR-E data here yet, maybe it is closer.
    Here is a very wide plot of the two datasets as monthly plotted on top of each other. Making that clear, raw AMSR-E daily data and NSIDC monthly compensated by a fiddle factor for each month.
    The compensation could be improved but this is good enough for now.
    http://www.gpsl.net/climate/data/sea_ice/northern-ice-extent-combined-raw-2009-04-28a.png
    13.58
    —- = 14.55 or of course reversed
    0.933
    To 3 dec places these are the monthly compensation figures for all the dataset, is consistent over the overlap years. The graph assumes this is so over all time for the longer dataset. Valid? No way of knowing or do you have a way of cross checking?
    Nov 0.932
    Dec 0.922
    Jan 0.942
    Feb 0.939
    Mar 0.925
    Apr 0.933
    May 0.926
    Jun 0.926
    Jul 0.933
    Aug 0.990
    Sep 1.000
    Oct 0.878
    Meier has said in reply to an email that the dataset match problem is for ice area, extent is not the problem. This makes sense.

  30. Geo (05:29:43)
    Good points you make there. I noticed the same thing on the NSIDC May 4 posting.
    1-They seem to have toned down the ‘spin’ in the way they wrote up the report.
    2-Mr. Goddard might be having an impact as they included a graph with a ±2 StdDev shaded area, similar to the ±1 StdDev that’s been shown at NANSEN. And the skeptic in me also wonders why ±2 StdDev instead of ±1 StdDev? Is it to give themselves some breathing room in case the ice extent goes much above the 1979-2000 mean?
    3-While the April Extent graph for the period still shows a downward trend, I think it’s noteworthy that it also shows that this year’s level is the highest in seven years, as is apparent in the AMSR-E graph.
    Now all that remains is to see what the ice looks like in September; whatever happens between now and then will be the result of storms and variations in wind and ocean currents moving the melting ice about.

  31. You know that graph is the inverse of the integral of the heat flux into the Arctic?

  32. And the skeptic in me also wonders why ±2 StdDev instead of ±1 StdDev?

    Because this corresponds to the 95% confidence interval which has been standard in statistical analysis for a very, very long time.

  33. Steven Kopits (05:37:36) :
    Forgive me, but how can we accept the data from the NSIDC as uncorrupted and reliable?

    Dr. Meier’s comments reflect his observation a steady decline in arctic sea ice for quite a while and he isn’t yet convinced that the recent recovery will reverse the trend. I don’t recall him ever speculating on the cause of the trend, so he shouldn’t get lumped in with the CO2 crazies. His data passes every check I know of for being legit and his recent mea culpa was textbook perfect.
    Real data has warts — sometimes embarrassing ones. Phony data is just too perfect. Real data from multiple sources is easy to check and thus holds to a higher standard.
    You can trust the sports scores in the NY Times.

  34. E.M.Smith (19:40:38) :
    Just for grins, I did a Google search of “2009 record snow” and got:
    Results 1 – 10 of about 11,000,000 for 2009 record snow. (0.24 seconds)

    Well, you did it without the quotes, not as you indicate here. That means any article with any of the words, which obviously give you 11 million hits. But if you try it as shown, the number of hits are considerably less:
    Try this instead
    Results 1 – 10 of 10 for “record snow 2009”. (0.39 seconds)
    Which doesn’t prove anything, except you have to do such “Google surveys” carefully….

  35. “dhogaza (08:20:50) :
    And the skeptic in me also wonders why ±2 StdDev instead of ±1 StdDev?
    Because this corresponds to the 95% confidence interval which has been standard in statistical analysis for a very, very long time.”
    Agreed; but there might be a reason to choose another:
    1) If you travel to some place and wonder if you have enough gas.
    2) If you wonder about surviving an operation.
    Answer: 1) by car 1 SD; by Skyhawk 100% 2) for cancer: maybe less; to remove a mole: 3 SD

  36. It would be interesting if someone in contact with Dr. Walt M. would ask for a someone more rigorous comparison of different years of ice thickness. Tho I recognize that, say, 3rd year ice in a multi-year warming trend might only be on average as thick as 2nd year ice in a multi-year cooling trend.
    Still, something like “2nd year is usually ~100% thicker than 1st year ice. 3rd year ice is usually ~50% thicker than 2nd year ice. 4th year ice is ~25% thicker than 3rd year ice. 5th year and greater is pretty much indistinguisable.”
    Something like that would be useful to know.

  37. “Still, something like “2nd year is usually ~100% thicker than 1st year ice. 3rd year ice is usually ~50% thicker than 2nd year ice. 4th year ice is ~25% thicker than 3rd year ice. 5th year and greater is pretty much indistinguisable.”
    Something like that would be useful to know”
    The ice is floating and moves… there is no permanent ice, it just moves on out of the Arctic and melts. So it is more a matter of ice which manages to be located in lucky places and gets to live longer.
    (twiddle)
    Here we go, have fun
    http://cersat.ifremer.fr/news/scientific_results/global_mapping_of_arctic_sea_ice_drift_a_unique_database

  38. I have a question: Whats the maintenance record of the Mauna Loa CO2 concentratometer and what are the decay rates of the components in it? If it can be shown the supposed CO2 curve is really reflective of the decay in the value of a given electronic component in it, that would be a huge blow to the alarmists.

  39. Anthony,
    It appears NSIDC started time stamping their arctic sea ice extent figures. I believe that has been a topic of concern on WUWT before. Sorry if this is old news!
    REPLY: Nope, you are the first to notice. I suppose it would be too much for them to credit WUWT for suggesting this on more than one occassion, but it is nice to know they read. – Anthony

  40. NCEP Ice product looking bad for the Arctic in the last few days.
    http://iabp.apl.washington.edu/maps_daily_ncepice.html
    This doesn’t appear to be from data originating with the Defense satellite F-15 which has been experiencing SSMI problems.
    “NCEP Ice Concentration (MODIS Ancillary Data)”
    “The NCEP global Ice Concentration product is produced once daily from
    the Special Sensor Microwave/Imager (SSMI)”
    http://gcmd.nasa.gov/records/GCMD_SEA_ICE.html
    “MODIS (or Moderate Resolution Imaging Spectroradiometer) is a key instrument aboard the Terra (EOS AM) and Aqua (EOS PM) satellites”
    http://modis.gsfc.nasa.gov/about/

  41. Now, Anthony. . . you know what Ronald Reagan used to say: “There’s no limit to what you can accomplish if you don’t care who gets the credit”.

Comments are closed.