Guest post by Willis Eschenbach
A few days ago, Steve Goddard put up a post called “Does PIOMAS Verify?” In it, he compared the PIOMAS computer model estimate of the Arctic ice volume with the SIDADS satellite measured Arctic ice area. He noted that from 2007 on, the two datasets diverge.
Intrigued by this, I decided to compare the PIOMAS ice volume dataset with the Cryosphere Today (CT) Arctic ice area dataset. Here is that data:
Figure 1. Arctic ice area (red line) from Cryosphere Today. Black line is a 6 year Gaussian average.
When I compared the two datasets, I expected to find something curious happening with the PIOMASS dataset. Instead, I found a puzzle regarding the CT dataset.
I compared the CT area dataset with the PIOMAS dataset, and I found the same thing that Steve Goddard had found. The datasets diverge at about 2007. So I took a hard look at the two datasets. Instead of an problem with the PIOMAS volume dataset, I found the CT area dataset contained something odd. Here is a plot of the CT daily data with the daily average variations removed:
Figure 2. Cryosphere Today daily ice area anomaly. Average daily variations have been removed.
The oddity about the data is what happens after 2007. Suddenly, there is a strong annual signal. I have put in vertical black lines to highlight this signal. The vertical lines show the end of September of each year. Before 2007, there is only a small variation in the data, and it does not have an annual signal. After 2007, the variation gets large, and there is a clear annual aspect to the signal. The area in September (the time of minimum ice) is smaller than we would expect. And the area in March (the time of maximum ice) is larger than we would expect.
I considered this for a while, and could only come to the conclusion that there was some kind of error in the CT dataset. So I decided to look at another dataset, the NOAA SIDADS dataset.
Again, I removed the monthly signal, leaving only the anomaly. Here is that result:
Figure 3. SIDADS monthly ice area anomaly. Monthly variations have been removed.
Again we see the same oddity after the start of 2007, with a large annual variation where none existed before 2007. In the SIDADS dataset the variation is even more pronounced than in the CT data.
So that is the puzzle. What has changed? Are they using a new satellite? If so, has the changeover been done properly? Since the smallest of the data has gotten smaller and the largest of the data has gotten larger, is the average data still valid? Just what the heck are we looking at here?
Despite searching, I have not been able to find the answer to this question. However, I have great faith that the assembled masses of the WUWT readership will find it very quickly. (And then some of the readers will likely tell me that this shows I am a layman and a fool, and that I should have been able to find the answer easily on my own … so sue me.)
Well, Willis,
As I always tell people, data doesnt falsify a theory. When data contradicts a theory we have a choice: question the data ( like you choose), modify the theory, throw out the theory.
So, What leads you to believe the a new algorithm is the cause? That seems like saying divergence is due to some anthropogenic cause.. special pleading?
When do you see the annual signal starting.. Perhaps a test is in order
Carl McIntosh says:
June 1, 2010 at 10:12 am
Joe Lalonde says:
“Another thing is that our species CAPTURES and holds water when no other species had in the past and this is TRILLIONS of gallons a day.”
Not even beavers?
*
*
Shaved, or unshaved? 🙂
Willis,
As I said before, you can center an anomaly anywhere you want. I, like you, think about it as a departure from normal … and “normal’ for the ice record is not zero, it is the long term average.
Actually, if normal is the long term average then normal should be removed from the anomaly, ergo centered around zero. Did you back out monthly or daily averages from each years average or the long term average?
Heres an experiment – graph each months departure from normal from normal by the long term average monthly departures. I think that you will again see the annual cycle show up for the last three years, however this time you will be looking at the ratio of each months departure to the long term average.
What I believe it will show is that the last three years have had a much larger departures from normal – and that it only happens just either side of the melt – again because that’s the only time that it can. You really can’t have too much departure at max or during the melt or freeze only during the min can the departure vary much from the long term average. Which is exactly what happened when the ice flushed in ’07. The annual signal shows up because the opportunity for a large departure only occurs on an annual basis due to the confined nature of the arctic basin.
Willis Eschenbach says:
June 1, 2010 at 1:24 pm
899 says:
June 1, 2010 at 3:12 am
So Willis, what was the Sun doing in 2007?
Taking a nap?
*
*
Could’a been.
As Ulric Lyons pointed out earlier:
The solar wind velocity was very generous through 2006/7 winter:
http://www.solen.info/solar/coronal_holes.html
When all of this was being discussed a few years ago here, I was wont to take Frank Sinatra’s mood piece ‘Summer Wind’ and slightly modify the lyrics to say ‘Solar Wind.’
See here:
http://www.lyrics007.com/Frank%20Sinatra%20Lyrics/The%20Summer%20Wind%20Lyrics.html
It actually fits quite nicely, all things considered …
I would look at wind patterns in the same number of years before the period under question as a control, and then the wind pattern during the period under question, as a start. I would center my wind data over Fram Strait in order to narrow the scope to the most effective area where wind can flush or not, depending on its parameters. Would it also be possible to overlay anomalous wind patterns to this graph? Or maybe the AO (though it may be too much of an overall statistic with the wind patterns we want buried in its noise).
Willis Eschenbach says:
June 2, 2010 at 12:00 am
Steven mosher says:
June 1, 2010 at 10:46 pm
“Dunno, looks like an issue for the null hypothesis. ice cycling like its never cycled before.”
It would … if I thought it was real. I don’t, I think it is from the known change in the satellite and the way that is being dealt with by means of a new algorithm.
w.
But a known change in the satellite that Spencer uses for his tropospheric temperature not the satellite that is used to produce the sea ice extent.
Also a similar fluctuation occurred in the Antarctic sea ice in the 70’s following a marked decrease in sea ice coverage there, the fluctuation has since petered out.
Phil.
yes. I suspect we are seeing a real change in the way ice melts and reforms. Also changes in the character of the ice, changes that MAY have never been see before, like rotten ice. Didn’t some guy take a research ship into the ice and report back on a change in the character of ice..
The Null aint looking so good.
Ah.. I see that toby has posted the video.
Willis:
“Me, I think this new pattern reflects a change in satellites, or a change in procedures, or something like that. But hey, I’ve been wrong before”
on what evidence? you have the data. It indicates a divergence from past normal behavior. It challenges the Null. I wouldn’t think the first response is to challenge the instrument on no colorable basis. but hey, its climate science. Now surely, we must accept the data and in the absence of any concrete evidence that the instrument is bad, we cant speculate that the instrument might be bad. As I pointed out, that would be exactly like the special pleading that Briffa did WRT his divergence problem.
Willis:
“I think we may have a winner. From AIRS/AMSU/HSB Version 5 Modification of Algorithm to Account for Increased NeDT in AMSU Channel 4 (emphasis in original).”
Confirmation Bias. You assumed it had to be a sensor problem because the data falsifies the Null.
Checking the metadata for the data you showed.. gimme a second
SIDADS, which you refer to says this:
“The Sea Ice Index provides a quick look at sea ice changes in spatial and historical context, and gives a consistent, up-to-date source of sea ice extent and sea ice concentration values and images. The NSIDC Near-Real-Time DMSP SSM/I Daily Polar Gridded Sea Ice Concentrations and the Sea Ice Concentrations from Nimbus-7 SSMR and DMSP SSM/I Passive Microwave Data data sets are used to generate the monthly records of sea ice extent and sea ice concentration for the Arctic and Antarctica from satellite passive microwave data.”
http://nsidc.org/data/g02135.html
Instrument
* SMMR Scanning Multichannel Microwave Radiometer
* SSM/I Special Sensor Microwave/Imager
This data set is generated from brightness temperature data derived from Nimbus-7 Scanning Multichannel Microwave Radiometer (SMMR) and Defense Meteorological Satellite Program (DMSP) -F8, -F11 and -F13 Special Sensor Microwave/Imager (SSM/I) radiances at a grid cell size of 25 x 25 km. The data are provided in the polar stereographic projection
Nimbus-7
http://www.geog.ucsb.edu/~jeff/115a/history/nimbus7.html
not operational in 2007.
DSMP: F8 is operational: F11 blew up in 2004. F13 lost
“Notice to Data Users: Processing of this data set is temporarily suspended due to the loss of the DMSP F-13 satellite. NSIDC is working with our data source, Remote Sensing Systems (RSS), to acquire data for this product from the DMSP F-17 satellite. We hope to have the data processed and available by late Spring 2010.”
http://nssdc.gsfc.nasa.gov/nmc/spacecraftDisplay.do?id=1987-053A
f13:
http://nssdc.gsfc.nasa.gov/nmc/spacecraftDisplay.do?id=1995-015A
http://microwave.msfc.nasa.gov:5721/source_documents/dmsp_f13.html
Should be a simple matter to contact them and ask the question.
magicjava is talking about a different platform
Jim Cripwell says:
June 1, 2010 at 6:49 am
Interestingly enough, total sea ice extent, Arctic plus Antarctic still stays remarkably constant.
Well, as their minimums occur at different months (September / February), that’s not true. When plotting each year’s minimums, one gets this chart, http://www.columbia.edu/~mhs119/UpdatedFigures/Storms_Fig20.gif (Warm season sea ice area in the Arctic and Antarctic, Fig 20, Makiko Sato’s website).
D Gallagher says:
June 2, 2010 at 3:01 am
While that is possible, if that were the case we’d see the largest swing in 2007, followed by progressively smaller and smaller swings in 2008 and 2009 as the ice recovered from the 2007 flushing event.
But that’s not the case, see Figs. 2 & 3 above.
Phil. says:
June 2, 2010 at 6:54 am
The change was in the Aqua AMSR-E satellite, which as you point out, is used by Spencer. It is also used for Arctic sea ice analysis.
Steven mosher says:
June 2, 2010 at 9:28 am
On what evidence? Well, the change in the satellite algorithm seems like evidence. My first response, however, was not to challenge the instrument. It was to pose the question to the assembled masses.
Steven mosher says:
June 2, 2010 at 10:02 am
Good find, Mosh. I’ll take a look and see what’s going on there. I’ve written to Dr. Bilitza to see what I can find out.
Willis Eschenbach says:
June 2, 2010 at 1:22 pm
The change was in the Aqua AMSR-E satellite, which as you point out, is used by Spencer. It is also used for Arctic sea ice analysis.
No the change was in AMSU channel 4 not in the AMSR-E.
“we (Spencer and Christy) don’t even use NOAA-15 right now…we are using the AMSU data from NASA’s Aqua satellite in the final UAH product.”
Sheesh, David, you don’t sound too confident in your prediction! 🙂
Eric Anderson,
No, I am not amazingly confident. But that is more to do with my personality than my prediction. 🙂 I am not actually a betting kind of guy. Willis made the challenge; I took it up. Then I start to panic. 😉
Willis wrote:
“I have never seen an annual pattern suddenly emerge after a quarter century of no annual pattern.”
Well, we’ve not seen such a change in the sun in quarter of a century.
“I think we may have a winner.
Previously processed AIRS/AMSU data during the period from October 1, 2007 to March 2, 2008 will be reprocessed with the modified algorithm and the data products replaced.”
That can’t be the culprit. The anomalies started about 3 months prior to Oct 1, 2007.
As I mentioned before, Malaga View, I find this very interesting. Until now, I had no idea the the satellite data wasn’t perfect! (Silly me.) Now I see a video of a scientist cruising along at an “open water” speed where the satellite says “ice” and you point out links showing serious difficulties with the data coming in. (By the way, thanks for the links.)
Why are most people discussing this as if there is a solid base for discussion?
Ulric Lyons says:
June 1, 2010 at 2:16 pm
899 says:
June 1, 2010 at 3:12 am
So Willis, what was the Sun doing in 2007?
Willis Eschenbach says:
June 1, 2010 at 1:24 pm
“Taking a nap?”
______________________________________________________
That was Maunder yes?
*
*
Back then, when all the hoopla was making the daily nooze, I gave consideration to taking liberty with Frank Sinatra’s mood piece ‘Summer Wind,’ and employ the term ‘Solar Wind.’
A slight change in a few of the other lyrics would have made for an interesting assessment of the realities.
See this for the Summer Wind lyrics:
http://www.lyrics007.com/Frank%20Sinatra%20Lyrics/The%20Summer%20Wind%20Lyrics.html
Maybe an enterprising and budding artist might produce something interesting there …
Summary so far of reasons to be skeptical that the recent pattern is an artifact
1. Willis’ best lead only affects data from 1st Oct 2007, while new pattern starts in Aug 2007.
2. Willis’ argument of ‘pattern is suspect because it is unprecedented’ doesn’t hold because the input pattern ( new magnetic behaviour of sun ) is unprecendented, for ice satellite data.
3. The pattern is also observed by SSM/I. For ‘artifact’ hypothesis to hold, SSM/I is required to have broke at same time as AMSR, and in such a way as to produce the same pattern.
You know, from visual inspection of SC graphs vs ice extent graphs alone it’s easy to see a corelation between arctic ice area & extent deviation and solar magnetic activity.
It strongly suggests
Low solar mag activity -> high deviation of arctic sea ice
High solar mag activity -> Low deviation of arctic sea ice
For both the 1987 and 1996 solar minimum, there were spikes in the ice graphs.
Anyone going to do a proper corelation analysis?
Willis
“On what evidence? Well, the change in the satellite algorithm seems like evidence.”
but Wrong Satellite.
Now that you are contacting the scientist in charge of the correct one, I’m looking forward to his response. if its an adverse result ( to the theory that its a bad algorithm) then I expect that adverse result will be disclosed.
I’m surprised you feel the need to say this.