Latest PIOMAS Volume update

[Neven]

And so I expect PIOMAS to put July 2015 at a volume loss of at least 6500 km3, which will put it around 1000 km3 below 2014, and close the gaps with 2011, 2012, 2013.

And a drop of 6659 km3 it is. 2015 is now 809 km3 behind 2013, down from 1223 km3 last month. Oh, and 971 km3 below 2014.

[seaicesailor]

And so I expect PIOMAS to put July 2015 at a volume loss of at least 6500 km3, which will put it around 1000 km3 below 2014, and close the gaps with 2011, 2012, 2013.

And a drop of 6659 km3 it is. 2015 is now 809 km3 behind 2013, down from 1223 km3 last month. Oh, and 971 km3 below 2014.

lol. Isn't it funny, that PIOMAS usually triggers a lot of reactions immediately? It is as if nobody was surprised of this, the way things are unfolding right now.

[plinius]

Rather surprised that the model did not produce a larger decrease.

[oren]

And so I expect PIOMAS to put July 2015 at a volume loss of at least 6500 km3, which will put it around 1000 km3 below 2014, and close the gaps with 2011, 2012, 2013.

And a drop of 6659 km3 it is. 2015 is now 809 km3 behind 2013, down from 1223 km3 last month. Oh, and 971 km3 below 2014.

You nailed it.

[Wipneus]

The monthly gridded data are available, still waiting for the daily values.

Here are the thickness maps of July 2014 and 2015 and the differences.

Again: these are average monthly thicknesses.

[Wipneus]

Same as above: thickness 2012 and the difference between July 2015-2012

[Nightvid Cole]

Same as above: thickness 2012 and the difference between July 2015-2012

The "2012" July thickness map looks just like the 2014 map. Did you copy the 2014 hyperlink again accidentally?

[greatdying2]

Rather surprised that the model did not produce a larger decrease.

Ditto.

[Neven]

lol. Isn't it funny, that PIOMAS usually triggers a lot of reactions immediately? It is as if nobody was surprised of this, the way things are unfolding right now.

Oh, I'm quite surprised to see how much July can make up for May and June.

Here's the latest PIOMAS update over on the ASIB, BTW.

[ChrisReynolds]

I need to see the daily data to see how things have evolved over July. But here is the series of monthly differences from within the Central Arctic for the 2015 - 2012 difference.



In the July average most of the 2012 to 2015 volume difference is still in the Central Arctic, specifically from grid boxes in that region with an effective thickness of between 3.0 and 3.9m.

[plinius]

Again, without the adhoc corrections, an assessment of the model isn't really feasible. And PIOMAS is not making those public.

[Wipneus]

Same as above: thickness 2012 and the difference between July 2015-2012

The "2012" July thickness map looks just like the 2014 map. Did you copy the 2014 hyperlink again accidentally?

Yes, I should be more careful changing code just before shutting down for the day.

Thanks, corrected now.

[plinius]

lol. Isn't it funny, that PIOMAS usually triggers a lot of reactions immediately? It is as if nobody was surprised of this, the way things are unfolding right now.

Oh, I'm quite surprised to see how much July can make up for May and June.

Here's the latest PIOMAS update over on the ASIB, BTW.

Neven, you cannot assume that this is genuine melt. If there is no ice left in the data, my take is that PIOMAS simply zeroes it out. Which means that what you consider melt will to some part be just model adjustments.

[Benje]

With help from Dugeonmasrter and Neven (for which many thanks) I can show how I have at last added the Greenland ice loss (from year low to year low) to the Sea ice loss (year low to year low). It is not a profound concept...it is just the change of Ice in the Artic....and it is not well done (I could only estimate by eye the greenland loss from the Danish site but the errors would at least not be cumulative.

But even this simple graph shows the overall position in a different way. Obviously the Greenland loss is (in general but not including 2013 and 2014) increasing rapidly and in 2015 the direction of travel will likely resume.

I know there are many skeptics about the concept of Greenland ice being as it were interchangeable with sea ice...or to look at it the other way round when the sea ice is gone that Greenland will melt as much as the sea ice would have if it had been there. Nevertheless, the history is real and as to the future, we will see.
Anyway, I hope it is of interest to some
Benje

[JayW]

With help from Dugeonmasrter and Neven (for which many thanks) I can show how I have at last added the Greenland ice loss (from year low to year low) to the Sea ice loss (year low to year low). It is not a profound concept...it is just the change of Ice in the Artic....and it is not well done (I could only estimate by eye the greenland loss from the Danish site but the errors would at least not be cumulative.

But even this simple graph shows the overall position in a different way. Obviously the Greenland loss is (in general but not including 2013 and 2014) increasing rapidly and in 2015 the direction of travel will likely resume.

I know there are many skeptics about the concept of Greenland ice being as it were interchangeable with sea ice...or to look at it the other way round when the sea ice is gone that Greenland will melt as much as the sea ice would have if it had been there. Nevertheless, the history is real and as to the future, we will see.
Anyway, I hope it is of interest to some
Benje

I had trouble with the pdf, so I turned it into a jpeg so others could see it.

[Neven]

Thanks, Benje and JayW.

[Benje]

Thanks JayW

[ChrisReynolds]

lol. Isn't it funny, that PIOMAS usually triggers a lot of reactions immediately? It is as if nobody was surprised of this, the way things are unfolding right now.

Oh, I'm quite surprised to see how much July can make up for May and June.

Here's the latest PIOMAS update over on the ASIB, BTW.

Neven, you cannot assume that this is genuine melt. If there is no ice left in the data, my take is that PIOMAS simply zeroes it out. Which means that what you consider melt will to some part be just model adjustments.

Plinius,

I don't see the sea ice concentration assimilation as being as problematic as you do. Do you view the assimilation of weather as being a problem?

[ChrisReynolds]

Gice daily and heff daily are now available. Frankly, they're a bit boring. Losses from 1 June to 1 July are high, but not exceptional on the whole.

Because they're boring I can't be bothered putting my other plans aside to blog tonight, but here are some raw data if anyone wants to see it. Volume for regions in km^3.

Posted as formatted text in a txt file, should copy into a spreadsheet OK. Table of 30 June volumes, 31 July volumes, and table of loss between those dates. All by region.

[plinius]

I don't see the sea ice concentration assimilation as being as problematic as you do. Do you view the assimilation of weather as being a problem?

Not the concentration assimilation per se. With my incomplete understanding, there should be a lot of real information about model performance in the residuals that come from the assimilation. In simple terms: Ice melts out and this melt-out contrasts with predictions from the model, when it should happen, i.e. a running validation (or invalidation) of the model, that would tell at any time about tendency to under- or overestimate the ice volume in the concerned regions.

[Richard Rathbone]

http://psc.apl.uw.edu/research/projects/arctic-sea-ice-volume-anomaly/validation/

This sort of thing.

[ktonine]

Not the concentration assimilation per se. With my incomplete understanding, there should be a lot of real information about model performance in the residuals that come from the assimilation. In simple terms: Ice melts out and this melt-out contrasts with predictions from the model, when it should happen, i.e. a running validation (or invalidation) of the model, that would tell at any time about tendency to under- or overestimate the ice volume in the concerned regions.

Running validation?  Ummm .... if we had running data on thickness we wouldn't have to rely on models

We have a few buoys, obviously, but we can't really claim they're representative of the ice in their vicinity - much less extrapolate to the entire sea ice area.  It's expensive, time-consuming, and pretty much unrealistic to expect a running validation. 

[crandles]

Not the concentration assimilation per se. With my incomplete understanding, there should be a lot of real information about model performance in the residuals that come from the assimilation. In simple terms: Ice melts out and this melt-out contrasts with predictions from the model, when it should happen, i.e. a running validation (or invalidation) of the model, that would tell at any time about tendency to under- or overestimate the ice volume in the concerned regions.

Note I am a complete amateur, this is all just speculation and am probably getting this completely wrong.

Comparing piomas thickness to actual thickness is difficult because the actual measurements are too sparse and have their own uncertainties. Assimilation corrections are different to this and it may be interesting to see information on these.

There will be lots of unders and overs in every region. If there was a lot of information in this e.g. a high ratio of one way to the other, do you think the modellers wouldn't have used that information to tweak the parameters of the model? I would assume they have in effect done that with optimising their hindcasts. After this is done, the unders and overs would tend to approach similar levels in each region so I doubt there would be much real information at this point.

However, if the system has been running for two or three years since the last time a hindcast and twiddle with the parameters to minimise those assimilation errors (and other comparisons to thickness data) was done, then there could be new information emerging. I wouldn't assume this will be a vastly important, the data might well look like lots of random pluses and minuses that is very difficult to distinguish whether any changes in ratio of +s' to -s' is all just random fluctuations or if there is any emerging trend contained therein.

While you may well not be able to deduce anything from such data, I think there might be circumstances where interesting information could emerge:
Suppose the PIOMAS model has a tendency to change ice volume towards say 2009 ice volume levels. Running a hindcast up to say 2011, most years in the hindcast would be years when the starting volume was higher than where the model tends to want to go so the model would naturally reduce volume as happened. There would be relatively few years with the current situation of volume less than 2009 levels where the model tends to increase or keep the ice volume fairly level. If this isn't what the ice is actually doing, then there might be radically different assimilation errors in the recent years than in the hindcast training period. These radically different assimilation errors might in these sort of situations be very interesting. However reality may not be close to this scenario which I created to try to explain a circumstance where the assimilation errors might be interesting.

Perhaps other people think differently to this?
Note I am a complete amateur and am probably getting this completely wrong.

[epiphyte]

Running validation?  Ummm .... if we had running data on thickness we wouldn't have to rely on models

That's true - but I think inapposite to plinius' point. If PIOMAS is forcing modeled thickness to zero when observed thickness drops to zero, it would be very interesting to know where and when it was doing that, and what the modeled thickness was before it was zeroed out. That would let us know if there are systematic errors which are being systematically eliminated whenever they would otherwise come to light...

[Richard Rathbone]

Running validation?  Ummm .... if we had running data on thickness we wouldn't have to rely on models

That's true - but I think inapposite to plinius' point. If PIOMAS is forcing modeled thickness to zero when observed thickness drops to zero, it would be very interesting to know where and when it was doing that, and what the modeled thickness was before it was zeroed out. That would let us know if there are systematic errors which are being systematically eliminated whenever they would otherwise come to light...

You don't know when it drops to zero. The measurements are too noisy.  If you actually did do this sort of forcing you'd be driving the model with the noise (like the Navy were).

[crandles]

You don't know when it drops to zero. The measurements are too noisy.  If you actually did do this sort of forcing you'd be driving the model with the noise (like the Navy were).

http://psc.apl.washington.edu/research/projects/arctic-sea-ice-volume-anomaly/

Model and Assimilation Procedure

PIOMAS is a numerical model with components for sea ice and ocean and the capacity for assimilating some kinds of observations. For the ice volume simulations shown here, sea ice concentration information from the NSIDC near-real time product are assimilated into the model to improve ice thickness estimates

So they are doing that and they obviously believe it helps or they wouldn't be doing it.

What does your '(like the Navy were)' comment mean?

[Jim Hunt]

What does your '(like the Navy were)' comment mean?

I assume Richard is referring to this?

http://forum.arctic-sea-ice.net/index.php/topic,1320

[ChrisReynolds]

I don't see the sea ice concentration assimilation as being as problematic as you do. Do you view the assimilation of weather as being a problem?

Not the concentration assimilation per se. With my incomplete understanding, there should be a lot of real information about model performance in the residuals that come from the assimilation. In simple terms: Ice melts out and this melt-out contrasts with predictions from the model, when it should happen, i.e. a running validation (or invalidation) of the model, that would tell at any time about tendency to under- or overestimate the ice volume in the concerned regions.

I understand that, and yes a trend or shifts to greater adjustments might be instructive. However it might not tell us just about model bias. Consider a grid cell, concentration is applied from satellite data and this causes the model to deduct a certain volume. Has this volume been lost because the model physics are wrong? Has it been lost because of difference between atmospheric forcings and actual weather? Here small changes in NCEP/NCAR cloud fraction may play a large role. What of ocean heat? How much ocean data is available to identify what role that plays.

In the end the assimilation of concentration, and atmospheric data play a role of keeping the physics in check and instead of the model free-running, pulling it more in line with a key observational measure of sea ice (concentration). My suspicion is that the bias is always towards loss when melt is underway and towards gain when growth is underway.

'Training' of the PIOMAS model was done in the past, mainly using DRA submarine transect data and ICESat data from Ron Kwok. As I understand it, the datasets were broken down in time and area. One set was used to adjust unkown parameters to better match the data. Then the model was verified against the second set to see how well it tracked those observations in hindcast. The switch to PIOMAS V2 was not such an adjiustment, they had a similar problem to that experienced by ACNFS this year, but the impact was very small.

Have you asked the PSC team or Dr Zhang if this data can be made available?

[Wipneus]

Daily gridded thick was also released. Here is the July melt as a sequence.

(click req'd)

[ChrisReynolds]

Summary of July losses.
http://dosbat.blogspot.co.uk/2015/08/piomas-volume-loss-over-july.html

Data derived from gridded PIOMAS is now updated.
http://dosbat.blogspot.co.uk/2013/12/regional-piomas-volume-data.html

[Tor Bejnar]

I just looked at past (crossed out) data and graphs at the top of this thread (page 1, introductory comment). Last year the minimum PIOMAS ice volume was about 2.7M less than the July 31 volume.  A 2.7M reduction for 2015 would bring volume to below 6.0M km³. On the first graph, this would put a point 1.5M above the suggested trend curve (projecting ice free in 2020), halving the discrepancy of the 2014 minimum.

There are feedbacks that will cause a long tail (ice free much later) and feedbacks that are pushing for a quick melt.  The show goes on!

[Nick_Naylor]

Daily gridded thick was also released. Here is the July melt as a sequence.

Looking at July 31, it looks like PIOMAS is overestimating the ice in the Beaufort region, showing thick ice in some areas where the concentration was very low. Makes me wonder how quickly/frequently it assimilates concentration. Those sparse floes would have to be unbelievably thick for the average thickness to exceed two meters in the area i highlighted.

[oren]

Daily gridded thick was also released. Here is the July melt as a sequence.

Looking at July 31, it looks like PIOMAS is overestimating the ice in the Beaufort region, showing thick ice in some areas where the concentration was very low. Makes me wonder how quickly/frequently it assimilates concentration. Those sparse floes would have to be unbelievably thick for the average thickness to exceed two meters in the area i highlighted.

It made me wonder too. Looking at that big area of open water between the highlighted area and the CAA, on PIOMAS it's not showing up at all. What do the experts say about this?

[Neven]

Has anyone been wondering yet what monthly decrease PIOMAS will report next week? I know I have. 

Just like last month I've been trying to get a feel for what's possible by looking at average August weather conditions in previous years. I made this temperature+sea level pressure map, with years following a ranking based on PIOMAS monthly loss (the loss in km3 is noted below the year in question), with 2015 at the bottom:



When it comes to SLP, 2015 resembles 2011 most, but 2011 had higher average temperatures over the CAB. 2011 lost 2372 km3 during August, but at the same time 2011 had 1625 km3 less volume at the start of August compared to this year.

Temeprature-wise 2015 looks similar to 2009, but 2009 had much stronger cyclone activity during August. Mind you, 2015 runs only up to August 26th, so the recent cyclone might still change the average somewhat, although not much I expect. 2009 lost 2578 km3 during August, but also had 1209 km3 more volume than this year at the start of August.

Given all this I don't expect August to see a loss of much over 2500 km3, unless PIOMAS has some melting momentum factor from July that bleeds through into August. Of course, a lot of MYI has disappeared during the month, not just the 'arm', but perhaps also in the region where ice is oldest and thickest (north of the CAA), where there has been widespread fragmentation and dispersal in the past couple of weeks. I don't know if PIOMAS has a resolution that picks up on that.

So yeah, I'd say around 2500 km3, not enough to dip below 2013, even though with 2221 km3 it lost relatively little volume during August.

[Wipneus]

PIOMAS updated the daily volume data. I updated my graphics, see the top post

Monthly gridded data also updated (but not the daily), I will post the graphics later today.

[werther]

Thanks Wipneus!

Nothing on the PIOMAS site yet, but you bring us the essential. The running monthly average, which has been growing for a while, is dipping again since July. And even steeper during August.

[Wipneus]

Look again Werther! I noticed the update on the PIOMAS site first.

[crandles]

The chances of minimum volume in 2013, 2014 and 2015 being above minimum volume of each of 2010, 2011 and 2012 seems pretty slim if the trend is an exponential decline. It could still be noise but this is surely looking unlikely now?

[slow wing]

Thanks Wipneus, that's very interesting!


With those winds around, I predict it could drop some more in September and the daily minimum could fall below that for 2013.


I can't wait to see the gridded data to see where all the ice remains... 

[Wipneus]

The daily gridded thickness data is now available as well. Here is an animation of the thickness development during August.

(Yes, must click to start animation)

[Wipneus]

Average monthly thickness of August 2015 compared with years 2006-2014. The ice  north of Ellesmere  is anomalous thick compared with all of these years, while the ice right at the pole is comparatively thin.

Click for the wider picture.

[slow wing]

Interesting that the 2015 melt season started with all that thick multi-year ice on the Pacific side but retains no memory of that in the current ice thickness map, relative to previous years.

Instead, the remaining thick ice is all in the traditional 'thick ice sanctuary' off the Canadian coast. This sanctuary ice is even thicker than in past recent years.

Since I don't have Wipneus' original, here's a screenshot of the thickness map at the end of August...

[Adam Ash]

As I see it tho, the remaining MYI is in a very different place.  It is now in among the pack, subject to whatever wind and current do with it, and wherever those force send it.

In previous years it was anchored to the land, and comparatively immobile.

So winter circulation of the Beaufort Gyre is likely to rotate this MYI 90 degrees - from the CAA to near the Chukchi, where it is in line to get the same battering ice received there this year. 

And isn't that same rotation of the gyre simply going to pull young ice into that area north of the CAA from north of Greenland.  That's just plain mush, and while you may get some second year ice out of it, the gyre will just roll it on into the Beaufort the year after without a chance to get much thicker?

[Peter Ellis]

The Arctic ice is pack ice, not fast ice, and is never "anchored" to land.

[crandles]

With those winds around, I predict it could drop some more in September and the daily minimum could fall below that for 2013.

2015 is above 2013 on day 243 by 401km^3

Recent year drops from day 243 to minimum have been
Year Fall Day243
2007 194 6652
2008 751 7823
2009 396 7235
2010 256 4838
2011 305 4607
2012 259 3932
2013 182 5574
2014 405 7217

So 2013 dropped unusually little after day 243.

Getting below 2013 minimum would require a drop of 401+182=583 km^3

Only one year 2008 dropped more than 583 km^3 from day 243 so maybe not impossible.

However, from 1979 to 2006 largest drop is 452, so 2008 look like an unusual outlier.

Also in the recent years listed, the highest three falls are from highest three values on day 243. This year's day 243 value is not as high at 5975. Maybe there is no such relationship and a big drop this year in September is possible.

In short, not impossible but it looks pretty unlikely.

[Neven]

Curious as I am, I'm in an Internet cafe right now (I don't have a connection at my holiday address), trying to open the PIOMAS data file, but somehow I can't unpack the .gz file with my usual extraction programme, Extractnow, nor with 7-zip, etc. Anyone experienced the same problems?

Never mind, it suddenly works.

See you next week...

[Neven]

Given all this I don't expect August to see a loss of much over 2500 km3, unless PIOMAS has some melting momentum factor from July that bleeds through into August.

2629 km3 it is. Interesting... 

[iceman]

Given all this I don't expect August to see a loss of much over 2500 km3, unless PIOMAS has some melting momentum factor from July that bleeds through into August.

2629 km3 it is. Interesting... 

It looks like the dip in the anomaly trace came about 23rd-26th August, which would implicate the cyclone (and subsequent accelerated bottom melt) in the larger-than-expected volume decline.

[seaicesailor]

Given all this I don't expect August to see a loss of much over 2500 km3, unless PIOMAS has some melting momentum factor from July that bleeds through into August.

2629 km3 it is. Interesting... 

It looks like the dip in the anomaly trace came about 23rd-26th August, which would implicate the cyclone (and subsequent accelerated bottom melt) in the larger-than-expected volume decline.

Accelerated bottom melt means PIOMAS has some melting momentum factor from August that bleeds through into September.

[oren]

Given all this I don't expect August to see a loss of much over 2500 km3, unless PIOMAS has some melting momentum factor from July that bleeds through into August.

2629 km3 it is. Interesting... 

It looks like the dip in the anomaly trace came about 23rd-26th August, which would implicate the cyclone (and subsequent accelerated bottom melt) in the larger-than-expected volume decline.

Accelerated bottom melt means PIOMAS has some melting momentum factor from August that bleeds through into September.

When I looked at that drop my layman's intuition said that PIOMAS had some thick ice there, but when extent dropped to zero due to the cyclone t had to delete it and thus the sharp volume drop. No idea if this has any merit.