I made a small and somewhat inaccurate contribution to another WUWT thread, this time on Arctic Ice volume, and a figure quoted by the Guardian which I found surprising. My error is explained here; Stoat's account of the matter is here.
Anyway, it made me realise that I should pay more attention to ice volume. Although I criticise the Guardian's 75% reduction since 1979 as a mild exaggeration, the mean reduction of 66% is still an impressive figure. So I downloaded the PIOMAS data, and put it through the anomaly visualisation program of my previous post.
Again there are two anomalies calculated - without and with trend adjustment. The trend this time is much larger, so you need to watch the units.
Here's the analogue of the familiar Jaxa plot with the fitted periodic annual curve. It isn't an anomaly, and there is no detrending. You can see the steady downtrend, and I've shown the y-axis based at zero, so you can see it isn't so far off.
2012 is marked in black.
Here is the anomaly plot with trend, shown over the years from 1979:
You can't now see where ice-free is, but the persistent trend is evident, with some recent acceleration.
Now here is the same data plotted over a one year interval. Again 2012 is black. The ice volume is pretty much as last year, which is bucking the trend a bit.
Plotting the anomaly after allowing for trend gives a similar picture, but note the lower y-range. Over the last two years, mid-year has been a marked minimum of the anomaly, but this year (as at end April) it has not yet started that dive.
Interesting, thanks. I shall link to you.
ReplyDeleteYou should have probably done a sanity check on PIOMAS before you wasted your time.
ReplyDeleteCryosat-2 saw almost 50%* more volume back in March and at the same time IceBridge observed vast expanses of 4m ice while PIOMAS imagines almost none that thick.
*Yes; as much as there was in 1979.
Anon,
DeleteI don't think looking at PIOMAS is a waste of time. There's obviously a lot of uncertainty. What I do with these plots is to visualize, not affirm, what they are saying.
Anonymous,
DeleteCheck your data again.
PIOMAS provides gridded data with .heff files being the effective thickness, which is the thickness to which the sub-grid thickness parameterisation is then applied. PIOMAS had very little ice over 3m thick even as far back as the 1980s. When you check ice bridge you'll find that the 4m thick (and thicker) peaks are localised readings of ridge thickness, these would not register in terms of the PIOMAS effective thickness (they don't occupy entire grid cells, but they would make an appearance in the application of the sub grid parameterisation.
Bottom line: By submarine ULS, ICESat, Cryosat, and localised readings from expeditions, the ice has thinned massively since the 1980s and 1990s.
Maybe references for those claims anonymous? But see my comment in the previous thread on large uncertainties.
ReplyDeleteClimate Science appears to be unique in not needing error bars. (Or that's the impression I get when I see graph after graph with no uncertainty estimates associated with them. Try getting those published in any other field.)
Critics of climate science often employ a rather expanded scope for the term. I would have called this glaciology or some such.
DeleteBut I don't think you will often see error bars on highly modelled output. Especially as here, where compute times likely rule out Monte Carlo. Basically, as Anon suggests, the main uncertainty is whether the model is right, and it's hard to put a number on that.
Nick, in my field an empirical model that failed to provide uncertainty bounds would likely prove unpublishable.
ReplyDeleteThis isn't a criticism of you that you don't provide what they fail to provide, or that I don't "like" their answer.
It's completely bizarre to me this is ever acceptable, or that acceptance of this behavior is so prevalent across an entire field of study.
In empirical science, the absence of knowledge of the uncertainty in the measurement equates to the absence of a measurement. Measurements are a central value plus bounds. No bounds, it's not a measurement, it might as well have been plucked from the sky.
I also dispute whether you could put bounds on the range of value. The fact they don't suggests to me the bounds are rather embarrassingly large and they'd rather not show it, for whatever reason.
The fact that it is an empirically based model also doesn't relieve them of their responsibility. After all, a mercury filled thermometer is based on a model of thermal expansion. It isn't itself a true temperature measurement. The number you use from the thermometer is itself the result of a model of thermal expansion, and (on a scientific grade thermometer) there are calibration figures that go with the thermometer that tells you how accurate it is (and the marks on side of the thermometer tell you its precision).
ReplyDeleteThe same requirements exist for models computed on a computer.
Anonymous wrote:
ReplyDelete"You should have probably done a sanity check on PIOMAS before you wasted your time.
Cryosat-2 saw almost 50%* more volume back in March and at the same time IceBridge observed vast expanses of 4m ice while PIOMAS imagines almost none that thick."
I don't know what the basis of your comment is (Steven Goddard?), but here's a quote from a BBC article after the ESA revealed new thickness maps:
Tuesday's release shows a complete seasonal cycle, from October 2010, when the Arctic Ocean was beginning to freeze up following the summer melt, right through to March 2011, when the sea ice was approaching peak thickness. Cryosat found the volume (area multiplied by thickness) of sea ice in the central Arctic in March 2011 to have been 14,500 cubic kilometres.
This figure is very similar to that suggested by PIOMAS (Panarctic Ice Ocean Modeling and Assimilation System), an influential computer model that has been used to estimate Arctic sea ice volume, and which has been the basis for several predictions about when summer sea ice in the north might disappear completely.
Recently the IceBridge mission has released sea ice thickness data through NSIDC measured by planes (well, actually, the EM-bird hanging from the plane) from end of March/early April. Here's my blog post on that, with a map made by a commenter, and NSIDC mentions it at the end of their latest monthly analysis.
Even if PIOMAS is off by 25% (good luck with proving that), things still look very serious. And to be frank, I don't think it's off by much, in either direction.
Sorry to others - I'm busy this weekend but can't let nonsense stand.
DeleteNeven you quote the BBC saying;
"sea ice [per Cryosat2] in the central Arctic in March 2011 to have been 14,500 cubic kilometres"
What is the central Arctic the BBC refer to? It can't be what cryosphere today, among others, call the "Arctic basin" because that would need an average 5.5m thick ice there which PIOMAS certainly doesn't (and apparently couldn't) show.
The European Space Agency certainly do say total March 2011 Arctic Sea Ice averaged 2.3m which would be about 33,000km^3 according to NSIDC extent figures (~50% more than the Pan-Arctic Ice Ocean Modeling and Assimilation System imagines).
Cryosat-2 has been extensively validated by surface and airborne observations and has already taken measurements for longer than the fey Icesat's faulty altimeters (which some claim validate PIOMAS) intermittently did.
I'm quite sure that the European Space Agency will stop pissing about and step up at Rio 20 to declare that the vanishing Arctic ice scare was a load of cobblers just like the wikivandal has prophesied. And just as you and their paymasters would want, eh, Neven?!
"What is the central Arctic the BBC refer to?"
DeletePerhaps they should've been more specific, but I believe it is the ice in the Arctic Ocean, without Hudson Bay and the Bering and Okhotsk seas.
Like I wrote at the time ESA released a second set of graphs (actually they didn't, but I captured them from their live presentation and published them in this blog post):
"In that animation we also see average sea ice thickness to be well over 2 meters in April/May. According to my crude calculations based on PIOMAS volume and CT SIA average thickness was a little over 1.6 meters on April 1st 2011. But again, this difference could be due to CryoSat not taking fringe ice outside of the Arctic Ocean into account."
Last year I wrote a blog post about that first ice thickness map ESA had put out, of which Duncan Wingham, the Principal Scientists for the CryoSat Satellite Mission, said during the last presentation that they had cobbled it up quickly so that they'd have something to show during the Paris Air and Space show. Obviously this crude map wasn't very accurate and it differs from the new map which I captured.
It's a pity that ESA hasn't even put these graphs on their website, but at least they released the video that can be watched in the BBC article.
"I'm quite sure that the European Space Agency will stop pissing about and step up at Rio 20 to declare that the vanishing Arctic ice scare was a load of cobblers just like the wikivandal has prophesied. And just as you and their paymasters would want, eh, Neven?!"
The spectacular decline in Arctic sea ice seems to annoy you a lot.
"What is the central Arctic the BBC refer to?"
DeletePerhaps they should've been more specific, but I believe it is the ice in the Arctic Ocean, without Hudson Bay and the Bering and Okhotsk seas.
Like I wrote at the time ESA released a second set of graphs (actually they didn't, but I captured them from their live presentation and published them in this blog post):
"In that animation we also see average sea ice thickness to be well over 2 meters in April/May. According to my crude calculations based on PIOMAS volume and CT SIA average thickness was a little over 1.6 meters on April 1st 2011. But again, this difference could be due to CryoSat not taking fringe ice outside of the Arctic Ocean into account."
Last year I wrote a blog post about that first ice thickness map ESA had put out, of which Duncan Wingham, the Principal Scientists for the CryoSat Satellite Mission, said during the last presentation that they had cobbled it up quickly so that they'd have something to show during the Paris Air and Space show. Obviously this crude map wasn't very accurate and it differs from the new map which I captured.
It's a pity that ESA hasn't even put these graphs on their website, but at least they released the video that can be watched in the BBC article.
"I'm quite sure that the European Space Agency will stop pissing about and step up at Rio 20 to declare that the vanishing Arctic ice scare was a load of cobblers just like the wikivandal has prophesied. And just as you and their paymasters would want, eh, Neven?!"
The spectacular decline in Arctic sea ice seems to annoy you a lot.
Neven, Cryosat only measures the central arctic ice volume, as your quote says (as a view of the website shows). Wouldn't "same value" indicate PIOMAS is underestimating volume?
ReplyDelete(Still it would be interesting to look at a plot of Cryosat volume versus time. I've been looking. The ESA is probably the worst agency in the world in terms of timely data release.)
How did you get the 25% uncertainty number?
I didn't mean it as an uncertainty number. I just meant to say that even if PIOMAS volume numbers are off by 25% (just a random number) things in the Arctic still look serious. But I don't actually think their numbers are off by 25%, because the model is validated against various ground measurements (subs, etc).
DeleteThe CryoSat numbers are out there somewhere, but I don't know where as I wouldn't know what to do with them. You can ask at the Arctic Sea Ice blog. I believe some people there know where to find them.
Neven, I'm approaching this from "first you choose the right methodology, then you draw your conclusions from that."
DeleteI've 9 weeks until my next data deployment, so I won't have much free time between now and then. Otherwise this would be a fun project to work on.
PIOMAS does have some information on validation here.
ReplyDeleteFrom what I can tell their thickness model is very simplistic, does not explain very much of the variance in the measured data, and definitely appears to underestimate the thickness of multiyear ice.
See e.g. this figure. The bias appears to be about 50% of true value for 5-m thick ice for example.
Sure, PIOMAS isn't perfect, but it seems to be the best we've got. And we need as much as we can have, because there's a lot of evidence showing that the Arctic sea ice hasn't finished declining yet.
DeleteIf you'd like to know more about PIOMAS you could read Schweiger et al. 2011.
Actually it's the only one that updates daily. Schweiger 2011 compares their model against others. As I was suspecting they seem to be low-ball.
DeleteAlso, if you believe their own verification results, this suggests that the bias should be higher in September than March—there is much less thin ice surviving in September, since if they are underestimating the thickness of multiyear ice, the largest systematic bias in their model will be near ice minimum.
ReplyDeleteYou need to look at late summer against other measurements to compare, not peak ice, where the effect of bias is diluted (and even offset, because they appear be biased high for one-year ice).
Marginally relevant, but I noticed that while this year's ice had been holding up well in both JAXA extent and PIOMAS, JAXA has recently taken a dive.
ReplyDeleteIt's taking a dive everywhere, Nick, even Arctic ROOS. That's because of that surplus of ice in the Bering Sea disappearing and Hudson Bay starting to melt out quickly (I'll have a blog post on those regions later this week). And then there are the polynyas in the Beaufort and Laptev Seas, which are quite big for the time of year.
DeleteIf you look at the anomalies, there's some interesting structure in the JAXA record.
ReplyDeleteDo you mean the dive? It sure shows there.
DeleteBut did you mean to link an anomaly plot?
It's a deannualized plot rather than an anomalized plot (basically the same thing except the subtracted annual cycle has zero mean). I'm using 1980-1999inclusive , or if you prefer the interval [1980, 2000), as the baseline here.
ReplyDeleteWhat I was noticing was the loss from 5.2-5.6 followed by a recovery to 6.1. There's a similar but noisier version of this from about 4.5 to 5.2. I figure part of what we're seeing is wind related events of compaction and expansion of the ice, that's superimposed on a larger-scale decline.
This is what that same plot looks like over the entire period of measurement. (It's a combined JAXA GSFC extent plot.)
You can see pretty clearly that after subtraction of the averaged 1980-1999 annual cycle, I'm getting a residual annual component. Either the amplitude has gotten larger or there is a phase shift, or both.
Anyway, with that background, my main interest was in the influence the arctic waves have on the measured sea ice extent.
(Obviously, if we could measure it reliably enough, volume shouldn't show variability associated with mechanical stretching and shrinking of the arctic, as long as you don't get mechanical transport of ice out of the region of measurement of course.)
NB: This is before coffee.
ReplyDeleteer
On a related note, I think this demonstrates the need for a high sampling rate of climate than once per month.
By monthly averaging you lose details.
If you look at a higher resolution than monthly, the high-frequency noise hasa pattern of spikes in it. (This is deannualized AQUA channel 5.)
These spikes are called "Intra-Seasonal Oscillations" and Roy describes them as:
[...] temperature swings are mostly the result of variations in rainfall activity. Precipitation systems, which are constantly occurring around the world, release the latent heat of condensation of water vapor which was absorbed during the process of evaporation from the Earth’s surface.
Monthly averages are nothing more than a really crappy low-pass filter followed by resampling to 12 samples per year. So not only do you lose temporal information by doing that, it also does bad things to your noise floor.
there's a related comment thread you may have missed on Lucia's blog on the ISOs.