The DMSP satellite is still operating, but the SSM/I sensor is not
Regular readers will recall that on Feb 16th I blogged about this graph of arctic sea ice posted on the National Snow and Ice Data Center sea ice news page. The downward jump in the blue line was abrupt and puzzling.
Click for larger image
Today NSIDC announced they had discovered the reason why. The sensor on the Defense Meteorological Satellite Program (DMSP) satellite they use had degraded and now apparently failed to the point of being unusable. Compounding the bad news they discovered it had been in slow decline for almost two months, which caused a bias in the arctic sea ice data that underestimated the total sea ice by 500,000 square kilometers. This will likely affect the January NSIDC sea ice totals.
Figure 1. High-resolution image Daily Arctic sea ice extent map for February 15, 2009, showed areas of open water which should have appeared as sea ice. Sea Ice Index data. About the data. Please note that our daily sea ice images, derived from microwave measurements, may show spurious pixels in areas where sea ice may not be present. These artifacts are generally caused by coastline effects, or less commonly by severe weather. Scientists use masks to minimize the number of “noise” pixels, based on long-term extent patterns. Noise is largely eliminated in the process of generating monthly averages, our standard measurement for analyzing interannual trends. Data derived from Sea Ice Index data set.
Figure 2.
High-resolution imageDaily total Arctic sea ice extent between 1 December 2008 and 12 February 2009 for Special Sensor Microwave/Imager SSM/I compared to the similar NASA Earth Observing System Advanced Microwave Scanning Radiometer (EOS AMSR-E) sensor. —Credit: National Snow and Ice Data Center
Dr. Walt Meier of NSIDC had planned to do a guest post here on WUWT, but this evening, with the magnitude of the problem looming, he’s asked to defer that post until later. I certainly can’t fault him for that. He’s got his hands full. Hopefully they have a contingency plan in place for loss of the sensor/space platform. I applaud NSIDC for recognizing the problem and posting a complete and detailed summary today. I’ve resposted it below in its entirety. Note that this won’t affect other ice monitoring programs that use the Advanced Microwave Scanning Radiometer (EOS AMSR-E) sensor, which is on an entirely different platform, the AQUA satellite.
UPDATE: 2/19 Walt Meier writes with a clarification: “One detail, though perhaps an important [one]. I realize that it is bit confusing, but it is just one channel of the sensor that has issues. And it isn’t so much that it “failed”, but that quality degraded to the point the sea ice algorithm – the process to convert the raw data into sea ice concentration/extent – failed on Monday.” – Anthony
From NSIDC Sea Ice News:
As some of our readers have already noticed, there was a significant problem with the daily sea ice data images on February 16. The problem arose from a malfunction of the satellite sensor we use for our daily sea ice products. Upon further investigation, we discovered that starting around early January, an error known as sensor drift caused a slowly growing underestimation of Arctic sea ice extent. The underestimation reached approximately 500,000 square kilometers (193,000 square miles) by mid-February. Sensor drift, although infrequent, does occasionally occur and it is one of the things that we account for during quality control measures prior to archiving the data. See below for more details.
We have removed the most recent data and are investigating alternative data sources that will provide correct results. It is not clear when we will have data back online, but we are working to resolve the issue as quickly as possible.
Where does NSIDC get its data?
NSIDC gets sea ice information by applying algorithms to data from a series of Special Sensor Microwave/Imager (SSM/I) sensors on Defense Meteorological Satellite Program (DMSP) satellites. These satellites are operated by the U.S. Department of Defense. Their primary mission is support of U.S. military operations; the data weren’t originally intended for general science use.
The daily updates in Arctic Sea Ice News & Analysis rely on rapid acquisition and processing of the SSM/I data. Because the acquisition and processing are done in near-real time, we publish the daily data essentially as is. The data are then archived and later subjected to very strict quality control. We perform quality control measures in coordination with scientists at the NASA Goddard Space Flight Center, which can take up to a year. High-quality archives from SSM/I, combined with data from the earlier Scanning Multi-channel Microwave Radiometer (SMMR) data stream (1979–1987) provide a consistent record of sea ice conditions now spanning 30 years.
Data error sources
As discussed above, near-real-time products do not undergo the same level of quality control as the final archived products, which are used in scientific research published in peer-reviewed journals. However, the SSM/I sensors have proven themselves to be generally quite stable. Thus, it is reasonable to use the near-real-time products for displaying evolving ice conditions, with the caveat that errors may nevertheless occur. Sometimes errors are dramatic and obvious. Other errors, such as the recent sensor drift, may be subtler and not immediately apparent. We caution users of the near-real-time products that any conclusions from such data must be preliminary. We believe that the potential problems are outweighed by the scientific value of providing timely assessments of current Arctic sea ice conditions, as long as they are presented with appropriate caveats, which we try to do.
For several years, we used the SSM/I sensor on the DMSP F13 satellite. Last year, F13 started showing large amounts of missing data. The sensor was almost 13 years old, and no longer provided complete daily data to allow us to track total daily sea ice extent. As a result, we switched to the DMSP F15 sensor for our near-real-time analysis. For more information on the switch, see “Note on satellite update and intercalibration,” in our June 3, 2008 post.
On February 16, 2009, as emails came in from puzzled readers, it became clear that there was a significant problem—sea-ice-covered regions were showing up as open ocean. The problem stemmed from a failure of the sea ice algorithm caused by degradation of one of the DMSP F15 sensor channels. Upon further investigation, we found that data quality had begun to degrade over the month preceding the catastrophic failure. As a result, our processes underestimated total sea ice extent for the affected period. Based on comparisons with sea ice extent derived from the NASA Earth Observing System Advanced Microwave Scanning Radiometer (EOS AMSR-E) sensor, this underestimation grew from a negligible amount in early January to about 500,000 square kilometers (193,000 square miles) by mid-February (Figure 2). While dramatic, the underestimated values were not outside of expected variability until Monday, February 16. Although we believe that data prior to early January are reliable, we will conduct a full quality check in the coming days.
Sensor drift is a perfect but unfortunate example of the problems encountered in near-real-time analysis. We stress, however, that this error in no way changes the scientific conclusions about the long-term decline of Arctic sea ice, which is based on the the consistent, quality-controlled data archive discussed above.
We are actively investigating how to address the problem. Since we are not receiving good DMSP SSM/I data at the present time, we have temporarily discontinued daily updates. We will restart the data stream as soon as possible.
Some people might ask why we don’t simply switch to the EOS AMSR-E sensor. AMSR-E is a newer and more accurate passive microwave sensor. However, we do not use AMSR-E data in our analysis because it is not consistent with our historical data. Thus, while AMSR-E gives us greater accuracy and more confidence on current sea ice conditions, it actually provides less accuracy on the long-term changes over the past thirty years. There is a balance between being as accurate as possible at any given moment and being as consistent as possible through long time periods. Our main scientific focus is on the long-term changes in Arctic sea ice. With that in mind, we have chosen to continue using the SSM/I sensor, which provides the longest record of Arctic sea ice extent.
For more information on the NSIDC sea ice data, see the following resources on the NSIDC Web site:
- Sea Ice Index
- Why is the Sea Ice Index product used to study sea ice?
- Do your data undergo quality control?
Dr. Walt Meier of NSIDC
“Some people might ask why we don’t simply switch to the EOS AMSR-E sensor. AMSR-E is a newer and more accurate passive microwave sensor. However, we do not use AMSR-E data in our analysis because it is not consistent with our historical data. Thus, while AMSR-E gives us greater accuracy and more confidence on current sea ice conditions, it actually provides less accuracy on the long-term changes over the past thirty years.”
His comment reminded me of the classic discussion of ‘accuracy’ versus ‘precision’.
accuracy – The ability of a measurement to match the actual value of the quantity being measured.
precision – The ability of a measurement to be consistently reproduced.
I took his statement to mean that the exact value of ice extent (accuracy) was less useful than the repeatability (precision) in looking for long term trends. If you have good precision in your data, it is easier to detect trends or external influences.
As we have seen here at WUWT in an earlier thread on Arctic ice, there is a ‘circle’ at the pole that is not covered by a given satellite. If your goal is to monitor the changes at the edges of the ice, consistently leaving the ‘circle’ out of the total probably does not affect the precision, but would affect the accuracy of the actual value of total ice. Leaving the circle out of the total DOES become important if you switch satellites, and the diameter of the ‘circle’ changes. As covered in that earlier thread, comparing satellite 1 to satellite 2 data affects the apparent trend, unless you know to correct for the size of the ‘circle’. Similarly, switching sensors would require ‘adjustments’ to previous data to allow an on-going comparison. And we all know the hazards that can arise from that slippery slope of adjusting historical data.
As noted by others, I applaud the efforts to obtain and publish quality data, and wish Dr Meier and his colleagues all the best in resolving this problem.
I don’t know if someone already asked but what about the Antarctic sea ice?
Do they have 2 different satellites: 1 for the arctic and 1 for antarctica, or is it the same satellite and was Antarctic sea ice data also wrong the last 45 days?
so does this affect ijis or not? if it does, then the ice will be looking better than all but 1 of the past 7 years and about equal to that (2003)…….
in addition to that, if you check out the graphs on this page:
http://arctic.atmos.uiuc.edu/cryosphere/
and add 500,000 to them, you will have a fairly unimpressive “downtrend” over the past 30 years.
in fact, that would bring the ijis data to just above 2003, or the highest it has been for this time of year since the ijis graph began….it would also give this year an excellent chance of surpassing the high ice extent reached that year….
Oops, I guess this is worth blogging about!
I can’t think of a better example of why great websites like this, and the hard work of Anthony and the WUWT crew, are desparately needed. How long would the data drift (failure) gone unnoticed? How long would it have taken the AGW crowd to start using it to push their agendas? Another hockey stick in the making -and just as accurate.
Great job Anthony!
Clive (22:54:23) :
Anthony,
Thank you for this revelation.
In mid December, I noticed that sea ice growth had flatlined and thought that strange since it had been very cold in N Canada and N Russia. I wrote to NSIDC on December 21 and received the reply below from David Korn at NSIDC. I still am not sure why the sea ice would not have grown when there was no sunlight and it was darned cold. That is obviously highly subjective on my part … but still.
If the mid December levels are indeed accurate (they may well be) then growth rates have little to do with air temperatures but can only be wind and or ocean currents.
Obviously it has to be cold enough but beyond that the actual extent is strongly dependent of drift and wind. Once the basin itself is covered with ice the growth has to be into the oceans where influx of warmer water can stop advancement of the ice.
The good doctor made another mistake… he forgot credit WUWT readers and Anthony for bringing the error to his attention.
Hmm, as a former “icepik” with the Coast Guard International Ice Patrol, I wonder if it’s time to go back to the good old days of human observation by ship and plane.
There’s a reason the the Ice Patrol still tracks icebergs by plane ratrher than relying on satellite data.
@F Rasmin
Here at this link is the Canadian weather and temps.
http://www.weatheroffice.gc.ca/canada_e.html
As a person with more than 30 years experience in remote earth sensing I wanted to comment on some back stage things that customarily go on in the day-to-day business. Our company supports NASA, NOAA, and the Not-to-be-Named other alphabet soup organizations with both on-orbit and ground based systems used to collect, transmit, transport, and array spacecraft-derived data.
Launching satellites is way cool so almost all of the organizational management attention is focused on launch activities, maybe 85%. Ten percent might be devoted to building the spacecraft. That leaves only a tiny bit of attention to turn towards actually operating the spacecraft and a miniscule amount of that goes to analyzing the derived data.
Among the worker bees Rocketeers are the stars of the business. Troglodytes, like my team, sitting in the darkened control rooms and serving as payload controllers are not glamorous at all. They monitor health and safety of the spacecraft, not the quality of the data. They spend all day looking at screens of data and responding to alarm conditions and developing work arounds for failed pieces of the system. (And there are always lots of failed pieces)
The scientists who are the recipients and actually are looking at the data are not at all part of the system and are effectively on their own. None of them got to their position by being programmatic hotshots able to deal with the intricacies of planning, budgeting, executing, overseeing and managing complex programs. They are usually not at all interested in those details and do not have the staff, expertise, time, or money to worry about ash and trash details like “is the sensor working properly?” They work in locations far from the payload controllers, don’t have access to the telemetry, and certainly don’t have a detailed understanding of how things work.
Lastly, almost all of the software on the systems still being used today are extremely unwieldy concantenations of uncommented multiple patches made by generations of programmers many of whom aren’t even alive today.
It ain’t always a pretty picture.
Cheers, John
Dave (07:15:02) :
I don’t know if someone already asked but what about the Antarctic sea ice?
Do they have 2 different satellites: 1 for the arctic and 1 for antarctica, or is it the same satellite and was Antarctic sea ice data also wrong the last 45 days?
It’s the same satellite in a polar orbit. The antarctic isn’t affected, if you compare the SSM/I and AMSR-E maps they are virtually identical, one reason being that there’s hardly any sea ice left to ‘miss’.
mark (07:16:15) :
so does this affect ijis or not? if it does, then the ice will be looking better than all but 1 of the past 7 years and about equal to that (2003)…….
No because they use AMSR-E.
in addition to that, if you check out the graphs on this page:
http://arctic.atmos.uiuc.edu/cryosphere/
and add 500,000 to them, you will have a fairly unimpressive “downtrend” over the past 30 years.
Why would you do that?
Re Phil’s query about the origin of the quoted description of the incident as a “catastrophic failure”: literally 5 seconds of Googling tells me it came from http://nsidc.org/arcticseaicenews.
I reckon it’s news after all.
I guess this extends my list of examples from the other day about just why it’s important for funny-looking items in publications used to support AGW to be investigated in a public manner – though again, this problem is obviously not a “poster” example as NSIDC were apparently looking at the oddity already and would have likely found the cause just as fast without outside prodding. But the same certainly couldn’t be said for equally unbelievable kinks in a Mann, Schmidt, or Steig graph, even if the underlying causes were as honest a defect as this sensor failure.
Accuracy vs precision…..hmm.
If the measuring device has a linearity scale degradation, I fail to understand how it can be used to determine long term trends unless the error and uncertainty is a known value.
RSS still has a step change in their data in 1992. It sticks out like a sore thumb, yet has not been corrected; its trend is still notably higher than UAH.
Is there a comparison of data between the two satellites for say the last 5 years?
Frank wrote:
” First, I applaud Dr. Meier for owning up to this and initiating a thorough investigation of the problem. I hope a resolution is possible in the near term.”
Must be a sign of the times. I always used to think it was a responsibility to own up and to correct things. Now we are expected to congratulate people for doing that.
That it took 45 days or more to find out something was “catastrophically” amiss to me shows there’s something organisationally wrong. The NISDC in my view is a managerial mess.
Have they know QM system?
Aren’t their instruments calibrated and checked?
Have they been blinded by a previous belief?
I don’t feel compelled to congratulate anyone for admitting this “Catastrophic failure”. If my boss found out I’ve been screwing things up for 45 days, I’d be happy to get off with a reprimand.
Speculating – had the error been in the other direction (contrary to prevailing climate doctrine) it would have been discovered in 45 minutes, and not 45 days.
Phil. (07:01:26) :
“Much of the data” isn’t in question, more hyperbole like your headline, where does the quotation about “catastrophic failure” come from? At least two organizations are continuing to use the SSM/I data without obvious difficulty so “catastrophic” is obviously an exaggeration.”
Phil., please see the posted message from the NSIDC “News” site (reproduced by Anthony above, for your convenience):
“The problem stemmed from a failure of the sea ice algorithm caused by degradation of one of the DMSP F15 sensor channels. Upon further investigation, we found that data quality had begun to degrade over the month preceding the catastrophic failure.”
Also, who are the two organizations still using the SSM/I data? Links?
Karl Heuer (06:32:52) :
VG-
Note that Mr. Will stipulated
** Global ** sea ice levels
The Arctic is but one part of the global sea ice total.
Yes he stipulated global sea ice from CT, but as CT posted their global sea ice values which showed Will to be wrong I’m not sure what your point is?
“our data shows that on February 15, 1979, global sea ice area was 16.79 million sq. km and on February 15, 2009, global sea ice area was 15.45 million sq. km. Therefore, global sea ice levels are 1.34 million sq. km less in February 2009 than in February 1979.”
According to some sites, the Antarctic sea ice extent is very large this year.
Well I’d ditch those sites then because they’re wrong!
http://iup.physik.uni-bremen.de:8084/amsr/antarctic_AMSRE_nic.png
I suspect that the sum of Arctic and Antarctic sea ice levels for Jan 2009, are very equivalent to the combined (global) total for 1979.
Fortunately we have the data so we don’t need your suspicions!
@frank K (06:47:36):
“I suppose it’s like having a miscalibrated thermometer which reads 75 F when it’s really 65 F.”
See, once you know whether the miscalibrated thermometer is consistently either +10 F above whatever “real” value you arbitrarily use as a reference or about 1.15 times this reference value (i.e. 75/65=1.15 or so), then you can reliably use that miscalibrated thermometer. I guess this ties back to the “precision” vs “accuracy” issue that another poster mentioned.
If the AMSR-E results can be mathematically related (i.e. correlated) to the SSM/I values – which is an exercise I haven’t done – then it should be possible to match both sets of data so that the AMSR-E sensor can now be reliably used.
People looking at Canadian or Sibrerian temps are on the wrong path. It’s been frozen over there for weeks and weeks.
Look where ice could yet still develop:
1) Barent’s Sea
2) East of Greenland
3) Ochotsk (sp?) Gulf.
http://www.ijis.iarc.uaf.edu/cgi-bin/seaice-monitor.cgi
The last few days have shown much new ice development
Oh how I wish all the folks involved in “the great global warming debate” were as classy and professional as Dr. Meier.
Let me add a defense of his choice not to use the AMSR-E data. The NSIDC is a raw data source. I’d like to be able to view the data from the two separate sources and draw my own conclusions.
Someone should merge the two data sets and I think a good normalization procedure can be devised. However, this is a separate task and should be kept clearly separate from the two raw data sets. Otherwise, there is risk of the kind of errors Mann made (pun intended).
Leon Brozyna (22:49:26) (and Pierre)
In electronics, you always hope that when something fails, it fails “hard” and does not become intermittent or “drift”. These types of failures are much harder to spot, and diagnose.
I believe that since there are very few platforms to measure against, it is harder to spot this kind of drift until, in this case the sensor failed “hard”. After all, could one tell if the AMSR-E was reading high, or DMSP reading low?
The problem is compounded by the remoteness of the sensor. You can’t exactly waltz out to it and check the calibration, although I sure that the owners would like the capability to do that.
Dr. Walt Meier and NSIDC deserve the round of applause that they are receiving from most on this site for their openness and the care that they are taking in correcting the data. After all, I’m pretty sure that their satellite is nowhere near a BBQ pit 🙂
[j/k]
So, according to the satellite info, we really WILL have an ice free North Pole this summer!
[j/k]
Thanks for every bit of work by everybody on this, as it is fascinating to follow.
“However, we do not use AMSR-E data in our analysis because it is not consistent with our historical data. Thus, while AMSR-E gives us greater accuracy and more confidence on current sea ice conditions, it actually provides less accuracy on the long-term changes over the past thirty years. There is a balance between being as accurate as possible at any given moment and being as consistent as possible through long time periods.”
That’s scientific gibberish, imo, except that without even looking at AMSR-E data or having read “Bill Illis (04:32:26)”, it’s possible to reasonably predict that the AMSR-E data says something different from what the NSIDC prefers – and probably prefers purely on subjective grounds.
How about looking at what the conditions were like in 1922?
Monthly Weather Review, Volume 51, Issue 3 (March 1923), page 142
DISTRIBUTION OF ICE IN ARCTIC SEAS, 1922
You can compare the description with Sea Ice Extent in the summers of this century using maps Cryosphere Today. For locations, see here and here.
Selected excerpts:
“In July the whole west coast of Nova Zembla was clear, and in August Franz Josef Land was probably accessible by open sea.”
“[In summer] Some sealers circumnavigated Spitzbergen, a feat that is not possible in most years.”
“[…] and though the east coast of Greenland does not appear to have been clear of ice, open water touched the coast in about latitude 74° N.”
“On the Newfoundland Banks … July was clearer than usual”
What is listed as exceptional in 1922, is the rule(!) in this century.
yeah phil….never mind, i wasn’t keeping the different sources of data clear. i got confused. it will be fun to see the NSIDC data corrected……500,000 is a lot of movement in the graphs i have grown to love!