Cryosphere Today 2013 Arctic SIA daily minimum: May poll

[Neven]

THERE HAVE BEEN SOME CHANGES TO THE POLL, WITH NOW MORE OPTIONS BETWEEN 0 AND 1 MILLION KM2; RE-CAST YOUR VOTE!

Same text as last month's poll:

It should be 'fun' and interesting to hold a poll each month, to see how opinions change according to sea ice rhythm and pace. These monthly polls can then also function as a dedicated thread for everything concerning Cryosphere Today sea ice area. Daily updated data can be found here.

Again, this poll is for Cryosphere Today sea ice area daily minimum (unlike the other popular measure: NSIDC sea ice extent monthly minimum), so we're looking for the lowest total Arctic sea ice area number on any given day at the end of the 2013 melting season.

Here's how things are currently looking based on data up to April 28th:



These are the daily minimums for the last 7 years (in millions km2):

    2005: 4.092
    2006: 4.030
    2007: 2.919
    2008: 3.004
    2009: 3.425
    2010: 3.072
    2011: 2.905
    2012: 2.234

Again, try to use this thread to discuss CT SIA mainly. There are other threads for the other topics, and if there isn't, feel free to open one.

Poll closes at the end of the month, and I will then open a new thread for the 2013 CT SIA minimum June poll. Keep voting. It'll be interesting to see if there are shifts in voting patterns as the melting season progresses.

[Neven]

I'm very busy at the moment, but could open another poll if there is sufficient interest.

[Jim Williams]

Poll closes at the end of the month, and I will then open a new thread for the 2013 CT SIA minimum June poll. Keep voting. It'll be interesting to see if there are shifts in voting patterns as the melting season progresses.

I'm sticking with complete melt-out until July.  Then I will re-evaluate my guess.

I will admit that so far it's looking like a repeat of last year.

[Pmt111500]

Sticking with the same as in previous poll. Might re-evaluate near the end of the period, depending on how the weather develops. This note applies at least to the summer solstice, when (in my opinion) different sort of atmospheric conditions start (decreasing solar radiation leads to more clouds on average).

[Richard Rathbone]

Same again for me. Maybe by June there will be enough development to think about picking a new number, but I didn't see anything in April.

[Neven]

I also voted "Between 2.0 and 2.25 million km2" again.

[wanderer]

SIA down to 12,03882 million sq. km - another 100k, seems like area doesn't give a damn about extent these days!

http://arctic.atmos.uiuc.edu/cryosphere/IMAGES/seaice.recent.arctic.png
sea ice anomaly -0,694

Lowest area on record on day 120 was in 2011: 11,70815 million sq. km
http://arctic.atmos.uiuc.edu/cryosphere/arctic.sea.ice.interactive.html

[deep octopus]

Between 1.25 and 1.5. I also don't like the fundamentals of the ice's health. It won't diverge too far from the trend (as it rarely does) until June, but I predict extent to fall below 2007/2011 levels within a couple of weeks, with especial weakness coming from the western Siberian side (Barents, Kara, Laptev), at least in the near term.

[Artful Dodger]

Hi Neven,

I'd like to see the bins changed next poll. Again we see 4 empty bins at that top of this poll, and a bimodal distribution at the bottom because the bottom bin is 4 x larger than the others.

I think we can do better! 

[SATire]

Again we see 4 empty bins at that top of this poll, and a bimodal distribution at the bottom because the bottom bin is 4 x larger than the others.

That is the main reason why I took that 4-in-one-bin again like in April. My estimate is still 2.0 +- 2, so the 0-1 bin gives max. chance for me.

[Sigmetnow]

SIA down to 12,03882 million sq. km - another 100k, seems like area doesn't give a damn about extent these days!

Lots of ice loss near Alaska recently.  Which begs a newbie question:  What geographical areas does the “CT arctic sea ice area” include?

[crandles]

SIA down to 12,03882 million sq. km - another 100k, seems like area doesn't give a damn about extent these days!

Lots of ice loss near Alaska recently.  Which begs a newbie question:  What geographical areas does the “CT arctic sea ice area” include?

I think it is NH though



doesn't seem to separate Baltic, and various lakes that freeze over

[wanderer]

Area now down from
day 112: 12,79345
to
day 122: 11,80377

Nearly one million sq. km in 10 days!

[SteveMDFP]

Area now down from
day 112: 12,79345
to
day 122: 11,80377

Nearly one million sq. km in 10 days!

Indeed.  Now, if the central arctic ice pack were thick and healthy, we could chalk this up to action in the marginal seas, and not consider this necessarily a harbinger of the melt to come.  But the central arctic is virtually a slush pond in comparison to past years.  The central basin under June's solar onslaught will fare no better than the Bering (etc.) under Aprl/May's spring thaw.

My sense is that we're seeing effects of positive feedbacks that aren't modeled, because we don't collect necessary data.

In partiular, I suspect the halocline stratificatioin of the arctic is steadily breaking down under the effects of increased ice motion at the surface, increased winds, and increased storms.

It's possible that the cyclone of August (GAC-2012) served to mix warmed surface waters into lower strata tens of meters down, that heat to percolate up throughout the winter, leading to thin ice over the entire arctic basin this spring.

How to test this hypothesis?  We'd need a flotilla of profiler buoys all over the arctic to make year-on-year comparisons.  We have, what, a dozen?  And a third seem to flush out the Fram every year.  Without such data, we're reading tea leaves, not modeling physics.

[Neven]

And a 142K drop, not bad. Not bad at all. Interestingly IJIS extent is more conservative, although 2013 is already near the lower years. But I think I'm seeing plenty of day-to-day changes on the UB SIC maps to justify the drops.

For CT area April 2013 had the lowest daily average since 2005:

2005: -31,277
2006: -37,327
2007: -30,747
2008: -41,979
2009: -37,793
2010: -48,659
2011: -44,019
2012: -46,190
2013: -51,292

[RaenorShine]

Sticking with the same as last month 1.5-1.75 in a stick the tail on a donkey approach.

The CT Global Sea Ice Area anomaly has gone negative in the past couple of days in is falling rapidly also.

[ChrisReynolds]

Same as last month for same reasons, 1.75 to 2M km^2. Nothing in the data has changed my opinion so far. Will re-evaluate as needed, but not before end of June.

[Agres]

Several snow events - increase albedo, but indicative of WV/ latent heat transport into Arctic.

In the next 10 days we will see rain events on sea ice.  Then, area, extent, and volume will start to fall.

[FrankD]

Around Day 100, 2013 was around 15th lowest in the catalogue (I wasn't counting exactly, someone might like to be more exact). 24 days later, 2013 is in 3rd place on CT area.

Just spitballing here, but if it continues at the current rate (more or less - the decline is hardly likely to be metronomic), it looks like we'll pass 2006 in a little under two weeks, and 2011 one-two weeks after that. I expect record territory by Day 150 (May 30th), if not before, if Agres weather forecast is correct.

That said, I'm personally expecting a flatter bottom on the curve than last year, and at this stage I'm doubtful of a record minimum in September.

[crandles]

Hi Neven,

I'd like to see the bins changed next poll. Again we see 4 empty bins at that top of this poll, and a bimodal distribution at the bottom because the bottom bin is 4 x larger than the others.

I think we can do better! 

Neven has asked me to do some analysis later. I agree with this and so suggest for the June poll (if we can have 16 options):

0.00-0.25
0.25-0.50
0.50-0.75
0.75-1
1.00-1.2
1.2-1.4
1.4-1.6
1.6-1.8
1.8-2
2.0-2.2
2.2-2.4
2.4-2.6
2.6-2.8
2.8-3
3.0-3.25
3.25+

July poll may need to go down to 0.1 gaps in the expected range.

A month seems a long time when a lot can happen so it might be hard to know whether differences arise from different views or different information. So, would it be better to shorten the poll period to half a month?

[frankendoodle]

Voted for 1.75 to 2M km2 again, reinforced seeing as we already lost 450K km2 in May. In fact, if we loose sea ice area at the rate we have the past ten days, we'd have zero sea ice mid September.

[frankendoodle]

Last year saw SIA decline by an average of 79K km2 per day between May 1st and August 31st. If we have a repeat performance this year will see SIA at 2.28 million km2 at the end of August and a daily minimum of just about 2 million km2.

[Neven]

Neven has asked me to do some analysis later. I agree with this

Again: Thanks, crandles!

and so suggest for the June poll (if we can have 16 options):

0.00-0.25
0.25-0.50
0.50-0.75
0.75-1
1.00-1.2
1.2-1.4
1.4-1.6
1.6-1.8
1.8-2
2.0-2.2
2.2-2.4
2.4-2.6
2.6-2.8
2.8-3
3.0-3.25
3.25+

July poll may need to go down to 0.1 gaps in the expected range.

We can have as many options as we like, but how about I keep the gap 0.25 and add the options all the way to zero?

A month seems a long time when a lot can happen so it might be hard to know whether differences arise from different views or different information. So, would it be better to shorten the poll period to half a month?

It might be better, but doing just one a month is enough of a job for me (which I mustn't forget). I'm also going to open a NSIDC September average minimum poll soon, and want to keep doing that every month. But twice a month......

[Juan C. García]

We can have as many options as we like, but how about I keep the gap 0.25 and add the options all the way to zero?

I agree.

A month seems a long time when a lot can happen so it might be hard to know whether differences arise from different views or different information. So, would it be better to shorten the poll period to half a month?

It might be better, but doing just one a month is enough of a job for me (which I mustn't forget)

Instead of making several polls on the same topic in a month, the SIA and SIE polls can start the 10th of the month (or the 15th the latest) and end the last day of the month. By example, I am waiting for the May 20th to give my vote on Cryosphere SIA.

[crandles]

We can have as many options as we like, but how about I keep the gap 0.25 and add the options all the way to zero?

A month seems a long time when a lot can happen so it might be hard to know whether differences arise from different views or different information. So, would it be better to shorten the poll period to half a month?

It might be better, but doing just one a month is enough of a job for me (which I mustn't forget). I'm also going to open a NSIDC September average minimum poll soon, and want to keep doing that every month. But twice a month......

Keeping it at monthly is fine with me, I just thought I would air it as a possibility. You could always delegate starting the polls to me as well.

Keeping the gap at 0.25 and add the options all the way to zero is also fine at the moment. But as we get closer to the minimum then I think 0.25 becomes quite a large range. So having 16 options and everyone choosing 3 or 4 seems silly. Hence I was trying to establish the idea that the gaps needs to get smaller in the expected range and the gaps to grow (preferably slowly at first) as you get further from expected range.

[Neven]

Keeping the gap at 0.25 and add the options all the way to zero is also fine at the moment.

Okay, done. I've added an announcement to the OP that people can re-cast their votes.

I think this is fine for June and perhaps July as well, but August will most probably need smaller gaps.

Solvitur ambulando!

[Neven]

NSIDC 2013 Arctic SIE September minimum: May Poll is now up too!

The dice are rolling...

[sofouuk]

so far we're at 63 responses, of whom almost exactly 2/3 have gone for between 1.5 and 2.25. keep the bins the same size through august and i doubt you'll see much of a change from this kind of distribution - there'll be a hard core of doom mongers who'll expect a late and low minimum, and an equally hard core telling them to calm down. all that might change will be, the same 2/3 will be voting between 1.25 and 2.0 if june/july melts quickly, or between 1.75 and 2.5 if it's slow. might be different come september, but by that stage it's hardly a prediction

would changing bin size from 2.5 to 2.0 really be worth the effort?, and doesn't going down to 1.0 seem over the top? given that this is just a bit of fun, i think the bins are just fine now

[NeilT]

I've had a bit of time today as I don't fly out of Stockholm till this afternoon.  It's the most day time I've had for 3 months.

So I decided to try and put together the consolidated set of thoughts I've had around these melting events.

First is that continuing increased in greenhouse gasses retain heat in the atmosphere which causes global warming and climate change.  Average CO2 is around 2ppm per annum.  Up from about 1.5ppm per annum.  But the real story is in the accumulated volume of CO2 in the atmosphere

OK so I have to ask "what heat".  Well, in reality, there is only one heat source.  The heat generated by solar flux in the day time.

OK so I have to go and find the solar output over the satellite period on which we are basing our predictions.  Sadly I can't find a chart which shows cycle 24 because it's not finished yet.  However we know that the cycle 23-24 low was the lowest for 100 years.  We also know that cycle 24 is much lower in solar output than the previous 3 cycles.

Now there is one more element to this picture.  So called "natural" variability.  The very largest of this is the El nino/La nina pacific SST cycles.  I'm willing to bet that this is not variable at all but actually modellable if we just had enough data and knew enough about the heat flows through the oceans.

OK so CO2 is rising.  This means that every solar cycle the system becomes more sensitive to solar input.  Less sensitive to natural variations and that the heatsinks of the planet have to absorb the trapped heat and try and balance the climate.

So we should see loss of ice on a cyclical basis.  With regrowth during the solar low and loss during the solar peaks.  Also one point I have noted is that the solar peak and the fall off into the solar low tend to have a cumulative impact.  I would expect this with CO2 as the heat generated by the cycle is retained until it finally hits the low and the effect is cumulative.

Then we come to the final part.  Variability.  So I need to get the SST cycle chart and overlay it on the other two.  Where El Nino and solar peak coincide, we should see more loss of ice.  Where low and La Nina  coincide we should see significant regrowth.

On the opposite side, if a La Nina and a peak coincide we should see less loss and if a low and an El Nino coincide we should see less regrowth or even, in extreme cases, a continuance of the ice loss.

Over all of this, the continuing increase of the CO2 input should create a constant degradation in the ice pack as the heat retained, decade on decade, should continue to impact ice regrowth, especially in winter and in solar lows.

So if I overlay the Ice extent chart, with the solar cycle chart, with the El Nino/La Nina chart, for the satellite record.  I should get a correlation of the ice loss results.   Granted I have not allowed for changes in Ice Mechanics or for weather changes due to the large increase of heat into the Arctic.  But these are relatively new phenomena and data simply does not exist for it.

Apologies, I'm now wizard with images so I simply made them all 50% transparent and saved it out as a PNG.  The source images are in the same album which is open if you have the link



It doesn't all map out cleanly, certainly cycle 22 was odd.  However the predictions by Hansen Et Al, that we transitioned to a new base in 2,000 and that CO2 mechanics finally took over, looks realistic in terms of ice loss.

OK so looking at that, what should it predict for 2013?

CO2 input in 2012 was almost at record highs and it continues to be high
We are at solar maximum.
El Nino kicked off at the end of 2012.
The ice is thin
Loss is faster than normal in the early part of the season
Weather, so far, is not favouring ice growth.
For 2011/12 the cycle is upside down.  (relatively), Low solar, La Nina and high ice loss.  2010 is understandable, I take 2011 and 12 to mean a breakdown in the integrity of the ice.

Personally I’d say all bets are off.  But we do have to make some kind of reasoned answer and predict some kind of change.

[NeilT]

Of course adding in Mt Pinatubo would help too...

[ChrisReynolds]

Neil,

Why are you using sea ice extent? Area suffers from the same problem.

The driver of changes in extent and area at minimum is the volume loss. Volume loss thins the ice and makes it easier for a given melt to open up areas of open water. The volume loss also represents an increasing energy imbalance in the system. But area/extent at minimum are a result of volume reduction.

If you use volume at minimum then the picture is one of decline despite all those changes.

Average annual extent/area shows a better relationship with CO2 because the annual average is not so strongly impacted by conditions within the Arctic, where most of the volume loss has occurred. Instead the annual average factors in the recession of the ice edge due to marginally warmer conditions due to anthropogenic GW in seas outside the arctic. In such areas the equilibrium establishing the ice edge is being affected by global warming and CO2 forcing.

Johannessen shows the relationship between CO2 and sea ice extent. First graphic of this page:
http://dosbat.blogspot.co.uk/2012/05/co2-causes-arctic-sea-ice-loss.html

[NeilT]

Ah, yes, put it on the wrong thread.....

That'll teach me to look more closely before posting.  I should have put it on the SIE thread.

[icebgone]

I am voting for 2.1 SIA.  I think volume is going to drop below 2,000km3.  Thinner ice tends to spread out increasing SIE over 2012's but decreasing SIA slightly.  I still think it will be 2017 before we see a month long ice free arctic but 2015 is looking like at least a few days ice free.  That is really a scary thought!  I am now more inclined to believe we will see a long thin tail of persistent ice for several years before all the ice vanishes during summer because irradiance peaks this year and then begins a long decline.

[ChrisReynolds]

Ah, yes, put it on the wrong thread.....

That'll teach me to look more closely before posting.  I should have put it on the SIE thread.

My point applies to both extent and area, for summer minimum both are the result of changes in area.

[ChrisReynolds]

CT Area anomalies since 1980, and the post 2007 change.
http://dosbat.blogspot.co.uk/2013/05/ct-area-anomalies.html

[TerryM]

CT Area anomalies since 1980, and the post 2007 change.
http://dosbat.blogspot.co.uk/2013/05/ct-area-anomalies.html

Wow!


Terry
BTW for some reason was unable to comment on your site.

[ChrisReynolds]

Don't know why you couldn't comment, but this can be discussed here.

"Wow", yes it is impressive, I've been sitting on this for over a year, should have posted previously.

I'll be blogging on this shortly, which should tie in some posts I've had in progress for months.

1) After 2007 there's a significant increase in seasonal melt this is why the CT Area shows a step drop after 2007.
http://arctic.atmos.uiuc.edu/cryosphere/IMAGES/seaice.area.arctic.png
On average this increased seasonal melt leads to just under 1M km^2 extra area lost through the melt season. (Anomalies can go lower due to open water at the end of the season while baseline average area would have been increasing.

2) 2007, 2010, 2011, and 2012 show large drops in early June as the ice edge enters into the Arctic Ocean.

3) 2007, 2011, 2012 show extremely rapid growth rates in the autumn, other years show rapid growth. The deeper the drop in anomalies the more rapid the autumn growth, apart from 2008. In 2008 there was a thermodynamic rebound due to 2007 taking the ice so much lower than the equilibrium at that point.

This is all happening because of the step thinning and reduction of MYI following 2007. See point 2, as there was another step thinning and loss of MYI in 2010.

[Vergent]

Chris,

Great graphic! "The tale of the tape" does not show the seasonal pattern, yours does. Great work!

Verg.

[ChrisReynolds]

CT's anomaly plot does sort of show it:
http://arctic.atmos.uiuc.edu/cryosphere/IMAGES/seaice.anomaly.arctic.png
It's the post 2007 'wild behaviour', but putting year on year seems to me to be better.

[NeilT]

My point applies to both extent and area, for summer minimum both are the result of changes in area.

Actually my point was that the changes in area are as a result of multiple factors.

CO2
Solar output
El Nino/La Nina
Volcanic activity
Weather.
Stability of the pack year on year

If you put them all together, the observed Area changes in Arctic SIA are quite logical.  Look at 2007, CO2 continues to rise, 3 out of 4 years with El Nino, tail end of a strong solar cycle, no significant volcanic activity and, finally, perfect weather conditions.

Only when you put the factors together can you see a picture which you can estimate.  The big game changer is 2011/12.  Relatively low solar output, strong and mild La Nina respectively, no exceptional volcanic activity and relatively benign weather except for the big storms in 2012.

Something has to drive this.  It doesn't just happen.  So if the driver is still weak, but the change accelerates, then we must be reaching a point of change.  Any estimate must factor in that change.  Or it will be wrong.

Much is said, on a daily basis, about the useless and linear predictions of the GCM's in modelling the change in the Arctic.  So if we don't factor in how the Arctic is reacting to the inputs into it's environment, then we'll make the same mistake.

As far as I can see, the state of the Arctic ice is such that it now only requires minimal energy input to destabilise and melt it.

So what happens when we go back to normal input?  Or the next 10ppm CO2, or another perfect storm of melt and export conditions.  In fact it appears that the melt and changes to the pack are actually driving more export.

Also for me, the big WOW on the Anomaly charts is not the huge drop, you can get that from the "tale of the tape" on CT and quickly running it from left to right.  The big one for me was that the Anomaly for 2012 had approx. 140 days at -2M anomaly or greater.

This was the point I was trying to make.  It's not loss of SIA driving melt, it's environmental conditions melting the ice which causes loss of SIA.  Which in turn, accelerates loss of SIA...

If I had a chart the same as the Extent, for Area, I'd have used it instead.  But the point I'm making is the same and reflects in the same way.

[SATire]

I have removed my vote because the 4-in-one-bin was cut to a quarter.

[Ice Cool Kim]

Of course adding in Mt Pinatubo would help too...

Doubt it. Volcanic effects are much over rated. Mostly coincidence with some other variations. Cooling attributed to both El Chichon and Mt P both started before the actual events that were supposed to have caused them.

Volcanism may have added to the drops but they are generally confounded with other 'natural' variations that were happening around the same time.

[Chuck Yokota]

I noticed that the percentage drop in both area and extent from 2011 to 2012 exceeded the percentage drop in volume, for the first time in the last 7 years ( I haven't checked earlier years).  This says to me that ice pack thinning has gone as far as it can, and I expect area and extent to drop proportionately with volume.

[Ice Cool Kim]

That's an interesting observation, Chuck.

The large recent extent of one-year ice must affect this. If all ice was 1 year ice then %age of volume change and area change would be the same: 1:1.

It would seem to follow from that, that your observation that % volume change in 2012 was less than % area change, implies an increase in the amount of multi year ice.

Now I'm sure someone will correct me if that logic is wrong but if the ratio has dropped there must be greater area of thicker ice.

[icebgone]

2.1 by September 9.  Variable weather takes us to 2.6.  A busy fram results in 1.6-1.8. Hence the median response.  By July 15, we'll know for sure.

[Chuck Yokota]

Ice Cool Kim,

I think that the differences are small enough that I would chalk it up to "weather" unless there developed enough of a trend to look for an explanation.  For the two years from 2010 to 2012, the differences in total percentage loss was within one percent between volume (26.4%), area (27.3%), and extent (26.8%).

[ChrisReynolds]

Chuck, I don't see it at all.

If you take the year on year % change in volume and area you get the following

   volume   area   % Diff
2001   11.18   8.75   27.78
2002   -11.39   -11.00   3.49
2003   -5.11   2.65   -292.95
2004   -3.51   3.41   -202.75
2005   -7.31   -4.46   63.70
2006   -1.81   -1.83   -0.96
2007   -28.19   -27.32   3.17
2008   9.51   2.88   229.98
2009   -2.53   14.02   -118.06
2010   -35.76   -10.29   247.45
2011   -9.28   -5.45   70.35
2012   -18.82   -23.09   -18.50

Surely if there is such an effect it should be visible on the year to year basis, which is how that's been calculated. %Diff is the percentage difference between the above percentages.

Plotting September averages of Volume and Area, Volume as a function of Area, shows that as of 2012 there is no significant deviation from a linear trend, although the underlying form is more likely to be an exponential.
http://farm9.staticflickr.com/8470/8380352351_1c0a3be3f3_o.jpg

Specifically the best exponential fit is 0.7231X^1.7969.

[Chuck Yokota]

Chris,

I see I wasn't clear in what I was trying to say.  Not a percentage difference between the percent declines, but an arithmetic comparison of the two-year decline:

                    Volume       Area         Extent
                     (Kkm^3)  (Mkm^2)  (Mkm^2)

2010             4.428         3.072       4.93
2012             3.261         2.234       3.61

%decrease   26.4%        27.3%     26.8%

eg. volume  (1 - 3.261/4.428)*100% = 26.4%

[crandles]

Chuck,

Taking a longer look than last 7 years, I would suggest it is noise about a trend:



http://farm9.staticflickr.com/8273/8752107222_4992f0cb10_b.jpg

I think it was Wipneus who was pointing this out this proportionate decline in area and thickness. It took the 2012 data to bring me around to accept it wouldn't curve below the line before curving back towards the origin.

[ChrisReynolds]

Chuck,

I see what you're doing now. Here is the same process for all the area and volume data (sorry I don't do extent). Differences of less than 1% (volume minus area) are in bold.

   volume   area   Diff
1981   -25.31   -6.60   18.71
1982   -17.00   -6.69   10.31
1983   19.76   8.68   -11.08
1984   8.10   -8.62   -16.72
1985   -3.99   -7.32   -3.32
1986   10.05   14.61   4.56
1987   4.84   5.93   1.09
1988   -6.79   -4.40   2.39
1989   -3.47   -8.94   -5.47
1990   -7.90   -10.03   -2.12
1991   -8.01   -7.38   0.62
1992   8.66   8.59   -0.07
1993   -9.22   0.28   9.50
1994   -8.43   -4.19   4.24
1995   -8.57   -1.40   7.17
1996   0.76   8.76   8.00
1997   17.82   11.10   -6.72
1998   -16.06   -18.63   -2.57
1999   -17.16   -14.19   2.98
2000   -4.85   -2.20   2.65
2001   11.57   7.83   -3.74
2002   -1.48   -3.22   -1.74
2003   -15.92   -8.65   7.28
2004   -8.44   6.15   14.59
2005   -10.56   -1.20   9.35
2006   -8.99   -6.21   2.78
2007   -29.49   -28.65   0.84
2008   -21.36   -25.23   -3.87
2009   6.74   17.30   10.57
2010   -37.39   2.28   39.67
2011   -41.72   -15.18   26.54
2012   -26.36   -27.28   -0.93
I don't see it in this data.