First Forecasts 2: PIOMAS

Commenter Gas Glo has just reported that Jinlun Zhang of the University of Wachington's Polar Science Center has issued a first forecast for the September 2011 average sea ice extent: 5.0 million square kilometers.

As it says on Dr. Zhang's webpage: The ensemble forecasting system is based on a synthesis of a model, NCEP/NCAR reanalysis data, and satellite observations of ice concentration. The model is the Pan-Arctic Ice-Ocean Modeling and Assimilation System (PIOMAS, Zhang and Rothrock, 2003).

Here's a nifty animation of the forecast:

To put this forecast of 5.0 million km2 into perspective, here are the NSIDC September SIE values for the past several years:

• 2005: 5.57
• 2006: 5.92
• 2007: 4.30
• 2008: 4.68
• 2009: 5.36
• 2010: 4.90

This forecast isn't final. It's updated every month as more data is put into the model. Last year Dr. Zhang was pretty much on the ball with his final prediction of 4.8 million square kilometers for the July Sea Ice Outlook.

Mind the caveat on Dr. Zhang's webpage:

The purpose of the seasonal predictions of arctic sea ice is for scientific research and education only. There are many uncertainties with the predictions and the results must be viewed with caution. The work is conducted at the Polar Science Center, Applied Physics Laboratory, University of Washington. Support is provided by NSF. Neither the Government nor the University of Washington nor any individual person makes any warranty, expressed or implied, or assumes any responsibility or liability for the use of the prediction results or any other results in this web site.

 

Comments

[dorlomin]

O/T but there is a cracking post on Greenland over at SkSci
http://www.skepticalscience.com/Variations_in_Greenland_Glaciers.html

[Peter Ellis]

Interesting. I wonder why their forecast is higher for this year than last year, given that we're starting from a lower initial level. As I understand it, there are two plausible reasons for this:

1) Different atmospheric forcing (i.e. simulated weather across the summer period) used in the two prediction runs. The forcing they use is an average of the last 7 years' data. Given that last year's Arctic summer was particularly cloudy, it may be that including this leads to a predicted forcing that favours less melt.

2) Different start conditions - although the ice extent is lower than this time last year, perhaps the thickness is greater? It's hard to credit that explanation though, given the current PIOMAS volume plot.

It would be very interesting to see last year's April forecast animation next to this year's.

[Neven]

Thanks for the link, dorlomin. A piece by the Yooper and Mauri Pelto is absolute top notch.

Peter, I think it has to do with 2. If the rest of April and start of May show a great extent decline, the forecast will change accordingly.

Unfortunately I was late to the game last year, but I did write a piece on Zhang's (and Lindsay's) prediction for September 2010 SIE in June.

One thing I noticed about the current animation is that the Northern Sea Route opens at the very end of the melting season, and the NWP remains clogged.

[Gas Glo]

Peter,

Why is the following not a plausible reason:

While ice extent and volume are lower, increased proportions are over 1 year and over 2 years old per

http://farm6.static.flickr.com/5222/5593055103_7b479d6bc2_b.jpg

from NSIDC updates.

Not only is there a greater proportion of older ice, the distribution of this older ice might be such that it is likely that less older ice will get advected though Fram Strait.

Nares Strait also appears to be more blocked than last year further reducing opportunities for advection.

I would suggest this is more plausible than your number 2 explanation. However, I am unsure whether this could be significant enough.

[Peter Ellis]

Advection and blockage of Nares strait may well be a factor in practice, but I don't think they're relevant to the point I'm making about the difference between the PIOMAS forecasts, since the forecasts are run using predicted average conditions. Given that, the model runs for last year and this year should presumably have used very similar forcings - the computer won't "know" the Nares is more blocked this year.

Turning to the question of ice age, the proportion of first year ice is almost identical to 2010. Given the lower starting extent, that means there's less older ice in total in the Arctic this year.

It's true that within the "older ice" section, there is slightly more third-year ice and slightly less second-year ice than 2010, which may make some of the difference, if PIOMAS incorporates ice age as well as thickness, which I don't know. Older ice is generally thicker and less saline than younger ice, however as I pointed out, from the current PIOMAS volume anomaly, we can tell that thickness isn't much greater than 2010 (if at all), so we can rule that aspect out at least. Salinity, possibly?

If the volume estimates are right, what it tells us is that the "three-year ice" currently in the Arctic looks different to "three-year ice" in previous years - it's thinner, for a start! Maybe it represents ice floes that almost melted out in 2009/2010; i.e. rather than getting down to ~1 metre by the end of summer and thickening up from there, it was down to 0.5 metres and had a lot more catching up to do. That leaves you with ice that's technically three years old, but physically much more like younger ice - i.e. Dr Barber's "rotten ice".

[Gas Glo]

Thanks Peter.

Looking at http://neven1.typepad.com/blog/2010/06/predictions-lindsay-and-zhang.html

Lindsay and Zhang's forecast from end of April was 5.12 +/- 0.42 million sq km. Guessing a bit but if PIOMAS model is initialised monthly at start of month, that 5.12 may be the equivalent of this year's 5m km^2 from the runs initialised with data at start of April.

So it looks like the change in the forecast could well be actually downward as we would expect from extent, area and thickness data.

[quote]The model uses the observed air temperature, wind, clouds, and precipitation to estimate maps of the ice motion, ice thickness distribution, and ocean temperatures and currents for past years, up to and including the most recent month.[/quote]

Nares strait was open at beginning of April 2010 but not 2011. I am guessing that if the model 'models ice motion' then the model will be initialised with such ice data as where ice is rather than using "predicted average conditions". I am also guessing "predicted average conditions" refers to atmospheric weather that cannot be predicted more than a couple of weeks out.

Maybe you know the model resultion is not enough to model ice motion through Nares strait? (Or maybe we disagree on interpretation of what has been written about what the model does. Or maybe we have different knowledge of what the model does.)

[Peter Ellis]

The Lindsay / Zhang forecasts do not actually use the PIOMAS model - they are a regression analysis without an underlying physical model (much like the Gompertz curves above). The method takes a one-off input of the current (e.g. April) thickness field and then projects the September extent based on parameters such as the percentage of ice in various thickness categories. They're thus not directly comparable to the forecast covered in this post.

I don't know if there was a full PIOMAS prediction made in April last year: memory says yes, but not what the number actually was. There were definitely June/July/August predictions made of 4.7/4.8/4.8 respectively: labelled "Zhang" in the SEARCH outlook archive.

As regards advection etc., I think we are disagreeing based on what's been written about what the model does.

Yes, it will model ice motion through Nares based on what's happened in previous years - e.g. wind in that direction plus water currents in that direction gives x amount of ice movement. However, as Patrick Lockerby covered on his blog, ice bridge formation is a stochastic event that depends on whether ice blocks happen to line up in the right way to form an arch-like structure that can then consolidate into a bridge. I don't think that's the sort of thing that can be reliably forecast (i.e. it's "weather-like" rather than "climate-like").

I believe that PIOMAS does incorporate ice movement data from buoys, which could potentially indicate the presence/absence of a bridge in any given year, but unless there's one passing through Nares right now (or within the last few of weeks at least), the data simply isn't there to feed into the model.

[Peter Ellis]

Additionally, the bridge points in the Nares strait are about 2-6 pixels across (there are some zoomed regions on the main page that show this). Pretty sure that's not enough to reveal presence/absence of a bridge.

There's an interesting picture near the bottom of their page - shows the error margin on their prediction. For a very substantial part of the Arctic they're forecasting ~0.5m ice with ~0.3m standard deviation. In the recion blocking the Northern Sea Route, it's 0.75m ice with 0.6m standard deviation. Similarly, there are large areas they've predicted as ice-free with uncertainty up to half a metre.

So I think the take home lesson is that at this point it's very much up in the air. :-)

[Gas Glo]

Thanks Peter.

I probably am assuming too much about what the model does.

"e.g. wind in that direction plus water currents in that direction" are obvious factors derived from what has happened in previous years and I agree such would be built in. I also agree ice bridge formation is a stochastic event that cannot easily be forecast. But bridge formation not easily forecast does not necessarily mean no effects built in at all.

I am thinking endogenous variables should be used where possible if there is potentially significant effect. So while I am not saying ice bridge formation should be modeled, I am thinking:

Ice movement is more than just wind in that direction plus water currents in that direction. It will be more complex to build in other effect such as that ice movement is also slowed down by presence of nearby coastline(s) (and water shallower than ice thickness), and the more ice area (& volume?) there is, the more that coastline slowdown effect is increased.

If such things are built in then it is far from perfect modelling of ice bridges but it isn't no modelling of effect of ice bridges either and this could produce a difference in the forecasts.

I am only guessing that such things are built in.

[Peter Ellis]

I am only guessing that such things are built in.

I too would hope these effects are built in to some extent. However, they're also built into the real world!

What I mean is that we know that in reality the flow through Nares can vary year-on-year from A (not much) to B (a lot) depending on ice bridge formation. The other variables you mention of water depth, coastline etc. stay the same from year to year*. So, no matter how well you write the model to handle those parameters, you still can't tell whether you have an A year or a B year without including the ice bridge itself.

* Leaving aside major landslides, Godzilla attacks, and so on. :-)

[L. Hamilton]

Last year, Jinlun Zhang and I sent in similar forecasts (4.8) for the August SEARCH outlook.
http://www.arcus.org/search/seaiceoutlook/2010/august
His forecast was based on much more detailed analysis that ought to work better in the long run.

While he starts higher this year, I'm starting lower: 4.4
http://neven1.typepad.com/blog/2011/04/trends-in-arctic-sea-ice-volume.html

[Rob Dekker]

Hi guys,

Just checking in after two weeks of absense. My wife gave birth to a very healthy and strong 3940 gr baby girl on March 27.
We've been busy, but very happy since then. Mom and baby are doing great. We are blessed.

On this subject of the sea ice minimum forecast by Lindsay and Zhang, I agree with Peter Ellis.
The Lindsay and Zhang forecast of 5 million km^2 does not seem to match the PIOMAS forecast at all.
In fact, the PIOMAS animation suggest something closer to 4 million km^2 ice minimum (eyeballing the extent at the end of animation).
Did anyone look in detail at the PIOMAS forecast and how it differs from Lindsay and Zhang's ?

Regarding this web site : Neven, what have you done ?
The comment section is jumping up and down (about half a line) every 2 seconds or so, and I had to write this response in a separate (Textpad) application and paste it. Just typing in my response would lock up Internet Explorer after 10 seconds or so.
Did you make any changes in the past few weeks that could have caused this ?
FYI, I'm using IE 7.0.

I was also interested to see if William had accepted by $ 10,000 bet back at stoat.
But it seems that he is still analysing dayly versus monthly data, or is not interested in betting seriously.
Either way, I'll respond there shortly.

Besides the baby, and the bets and the changes in Neven's blog, it looks like the melt season started for real in the Arctic...
I'm fastening my seat belt....

Rob

[Neven]

Congratulations with your baby girl, Rob!

I have IE8 and am not experiencing the problem you describe. Anyone else have IE7 and experiencing the jittering Rob describes? The last thing I have changed is the RSSfeed widget in the right hand bar. The news items scroll up every 3 seconds I believe.

BTW, it's just Zhang's prediction. I don't know if he'll do one with Lindsay as well this year.

[Rob Dekker]

Thanks for the congrats, Neven.

Regarding the jittering problems, I understand that you don't have any problems on IE8, but for IE7, it is really serious.
There seem to be 3 different 'jitters' going on simultaneously :
1. Every 5 seconds, your news post rolls. No problem with that.
2. Every 3 seconds, the 'comment' section jumps up and down about half a line. That's annoying.
3. Also every 3 seconds, but out of sync with jitter 2, your site connects to internet. It seems it is trying to update itself or so.
That jitter is also annoying, since IE temporarily 'stalls' for a few hundred milliseconds when it updates.
So this last jitter also interrupts scrolling and moving around on your pages.

Besides these jitter problems (which make your site feel very nervous), the really bad thing is that IE completely locks up or kills itself (and I have to kill it and re-start it) when I'm typing in a response into the comment section.
Seems that IE cannot handle getting keyboard input and doing a web-update (jitter 3) at the same time.
So I hope that you can do something with this info, and if not, then I would much prefer it if you would re-instate your Feb/March version. Because at this moment, for an IE7 user like me, your site is almost impossible to use.

There must be a better way to include your rolling-news bar, since I don't see the same problems on other sites with similar features.
Sorry, but you make it very hard to use your site with IE7..

[Neven]

Rob, I'm sorry things don't work well visually. The bit of experience I have is that things will always look different or (not) work on different browsers. There's even a difference between my desktop and laptop, even though I use Mozilla FF on both.

I have removed that scrolling news widget (and replaced it with something better). Ideally this would help, but in practice it often doesn't. Let me know if it helps, and if not, I'll remove another widget to see if it helps. But there's only so much I can do, as I'm dependent on TypePad's platform.

In the meantime, would it possible for you to compare with another browser (to see if the problem persists) or upgrade IE?

[L. Hamilton]

I get very jittery pages in IE8, as well. Seems to refresh and jump up every second or two.

[Rob Dekker]

Hi Neven,
I don't think your news banner is the problem, and I don't think the browser version is the issue either.

I think the biggest problem is that your site is running a script that is connecting to internet and probing something very frequently.

I did a bit more investigation, and this is what I found :

On a side page like this, updates happen every two seconds or so, but on you main page, internet probes happen almost continuously.
In fact, your front page (http://neven1.typepad.com/blog/) attempts so many updates that you actually takes up about 50-60 % of the CPU time on my computer.

Meanwhile, the icon on the page tab (left top on IE) is blinking almost constantly, indicating that it's connecting to internet.
Also, the 'status report' bar (bottom bar on IE) normally shows "Done" when it's done loading a normal web-page.
In your case, apart from "Done", I see it the flashing the addresses it accesses. The flashes are very short (10 milliseconds or so) indicating that access is fast, so I had a hard time 'reading' what it is trying to access.
But after some staring I see the following links coming by very frequently :
http://static.all.fbcdn.net/xd_proxy.php/....various unreadable subaddresses

Now, "fbcdn.net" is facebook.

I went to the HTML source of your front page, and see that indeed you are trying to do something with facebook, but I don't know HTML good enough to understand which refences are probed continuous, and which ones you use for facebook login checking.

Either way, your pages feel very nervous, accesses internet frequently (within 100msec - 2sec depending on where on your site you are), takes up considerable CPU on a client computer, and most interestingly : It was not like this a few months ago.
Please check out what you have changed, and again, it does not look like the news banner is the issue at all. I actually like the news banner.

The jumping comment section is annoying too, although I'm not sure if that has anything to do with the facebook accesses. Also it took me 3 tries just to post this comment (since IE locks up).

Please fix. I can't use your site like this. Thanks !

[Dave Eater]

Don't know if this will help, Rob... but try using "compatability mode" in IE8. I'm afraid I don't know how to turn that on (I don't use IE8, or IE in general - they stopped supporting the Mac years ago).

[Rob Dekker]

Neven,
The problem seems resolved !
Within hours after my post, your site stopped accessing facebook and anything else it probed, the jitter in the comment section disappeared, and I can now freely type in this comment without IE locking up on me.

Thank you !

Just a question : what did you change ?

Thanks

Rob

[Neven]

Rob, to tell the truth I hadn't changed anything yet (except for replacing the scrolling RSSfeed widget with a TypePad sidebar RSSfeed). I was actually about to ask TypePad support to tell me what to do. Maybe it had to do something with TypePad after all and they changed something after people over at other blogs complained?

Right now in FireFox I don't see the 'Arctic Sea Ice News' RSSFeed anymore. Weird.

Anyway, I'm very glad things seem to work better for you now. Let me know if this changes.

ps I fixed the link in your previous comment.