Yesterday, as covered by WUWT, NSIDC announced that Arctic sea ice melt had turned the corner on September 10th with a value of 4.14 million square kilometers:
XMETMAN writes of his discovery of a discrepancy between what NSIDC announced yesterday, and what their data actually says. I’ve confirmed his findings by downloading the data myself and it sure seems that the minimum was on September 7th, and not the 10th:
Source: ftp://sidads.colorado.edu/DATASETS/NOAA/G02135/north/daily/data/NH_seaice_extent_nrt_v2.csv
He says on his blog:
The Arctic sea ice looks to have reached its minimum on the 7th September, which is four days earlier than average. The sea ice extent bottomed out at 4.083 million square kilometres making it the second lowest since records started in 1978 – well that’s according to the data file that I’ve just downloaded!
Strangely, according to the data that I download from the National Snow and Ice Data Center [NSIDC] the minimum occurred three days later on September 10th. As I said in my introduction on the 7th the value was 4.083, but according to the news item that I’ve included below, the value on the 10th was 4.14 million square kilometres and tied it with the year 2007, which according to the data file is third.\
…
All these daily values translate into the following chart with the minimum occurring on the seventh and not the tenth of September.
It is a puzzle. Perhaps whoever wrote the NSIDC press release looked at their 5 day average value in their Chartic interactive graph instead of the raw data? When using that tool, the data rounds up to 4.14 as you can see:
NSIDC is closed for the weekend, so perhaps we will get an answer to this puzzle on Monday.
Bill Illis:
Thanks Bill. No reason the suspect there’s a problem with the filter but I have checked over the whole process and it seems to be doing what is expected.
Here I have plotted my results against those of 5d NDISC points that I lifted from their interactive graph. I subtracted 2.5 to remove the phase shift of the trailing average.
Now my results for the last few years are about a day earlier. For the pre-88 data they are very similar. I think this is because Nimbus-7 had a 6 day repeat cycle so was actually averaging or blurring the data anyway.
For the middle section there is a notable discrepancy, Two spike years disappear. This is the kind of thing I was expecting having removed short term weather distruption. This really was the aim of what I was doing.
If we are believe the data, what this shows is that weather disruption tends to lead to an extra late breakup of ice in some years but does not tend to produce early minima. That is that there is an asymmetric bias introduced by taking short term averages which do not remove weather ‘noise’ from the data.
Having looked at this again, I realised that the drift to earlier dates as the filter length is increased is simply a reflection of the asymmetry of the annual cycle. Refreezing tends to be rapid once it starts, so smoothing the data leads to a progressive shift to ealier dates.
Bill’s Illis comment that there was ‘something wrong with the filter’ was incorrect, it is doing exactly what it is supposed to but the issue raised was enlightening. This needed explaining.
Thanks for pointing this out Bill.
Turning point? Hmmmh…
For those who don’t automatically think « Wow! That data sure is flawed from top to bottom! »
Of course: Antarctica isn’t warm! It’s august, i.e. “february” there, Europe’s often coolest winter month.
But it is warmer. What is really unusual is that the dark red anomalies aren’t where I supposed to find them (around the Peninsula). They are located in the central area. Hmmmh…
“Shouldn’t happen”, some Lisp machines once said just before experiencing a hard shutdown 🙂
maybe you need to say what were are looking at since you provide it out of context with no explanation.
What data is it ? satellite, surface, model output?
Nice pics are useless unless we know what is bring shown.
Sorry Greg, it was of course not my intention to hide it away.
It is here: https://climate.copernicus.eu/
This is certainly surface reanalysis, with a rather strong similarity to Nick Stokes’ TempLS output:
https://moyhu.blogspot.de/2016/09/big-rise-021c-in-surface-temperature-in.html
Data doesn’t adjust itself . So who are the actual people at NOAA , NASA and Environment Canada that cherry pick weather station data and ” adjusting” numbers to claim “records ” ?
More likely NOAA announces the newest fudged numbers since fudged numbers started 10 years ago .
Why did the CRU in the UK purposely destroy data records ?
Right there red flags should be flying .
“Why did the CRU in the UK purposely destroy data records ?
Right there red flags should be flying .”
Because there was no need to keep it.
It wasn’t their data.
The data is still available at the original source.
http://www.nytimes.com/gwire/2009/10/14/14greenwire-scientists-return-fire-at-climate-skeptics-in-31175.html
The analysis of variation in date of melting has now been written up and posted at Climate Etc.
https://judithcurry.com/2016/09/18/is-the-arctic-sea-ice-spiral-of-death-dead
so – nobody checked how NSIDC calculate extent data and minimums?
As set out here?
http://nsidc.org/arcticseaicenews/faq/#diff-min-values
“In April 2012, NSIDC updated its method of calculating daily values for the Arctic sea ice extent minimum from a 5-day centered average to a 5-day trailing average. The new calculations show, for example, that the record minimum occurred on September 18, 2007, which was two days later than we originally reported (September 16). In addition, NSIDC updates extent values, calculated initially with near-real-time data, when final processed data becomes available. These final data, processed at NASA Goddard, use higher quality input source data and include additional quality control measures. The recalculations show a 2007 record low extent of 4.17 million square kilometers (1.61 million square miles). Our originally published value was 4.13 million square kilometers. In the final data, the date of the minimum may also change for some years.
For more information on calculating daily sea ice extent values, see the Sea Ice Index documentation.”
In the final data, the date of the minimum may also change for some years. All their so-called ‘data’ is always changing, The past is never what it used to be. Winston Smith is hard at work, but he is your friend Griff.
Well, it doesn’t strike me to be a big deal in regard to what the low was or is. The only time it would become a concern if someone tried to use their averaging method and then pick a day as a low that wasn’t a low. This could result in losing track of trends relating to when lows are occurring and when a recovery starts. If they would say “5 day running….” when they make the announcement, I don’t understand the big deal.