How many days earlier than 2012

[johnm33]

With monthly polls it may be interesting to compare the rate of change.
For instance on http://www7320.nrlssc.navy.mil/hycomARC/navo/arc_list_arcticictn.html  the cracks across the thick coastal ice disconnecting it from the coast show up on the 21st March [ignoring those which took place before the melt season] whilst last year the first signs were on the 28th July, thats 129 days indicating a very different arctic this year.
 wanderer posted ijis dropping 200k [re arctic sea ice free (extent)] , when did that first happen last year?

[Pmt111500]

Well the CT SIA numbers say it's only 4 days ahead of 2012, then again, it's still over 2011 maximum.  I calculated for fun (by hand) the last 32 years average (to get some climatology) of the ahead/behind date at this ice area on this date from SIA data and got (somewhat to my surprise) that the 2013 melt (if one uses the area for calculating this) is only 6.25 days ahead of the long term average on the date .2739 . Of course this is the time of year when the arctic amplification is hardest to see. The largest outliers were 2006 and 2007 (well 2011 of course too) which were full 4 weeks ahead. Closest years in comparison were (difference) 2000 (0 days); 1989, 1990,1995 (1 day); 1997, 2008, 2009 ,2010 (2 days), 2002 (3 days) and 2012 (4 days),  all the 21 other years were a week or more off this years progress at this stage. Possibly there is nothing further to study here.

ah, use [square brackets] for formatting.

(modifed 16.04.2013): Lawrence Solomon of financial post 'agrees'
http://neven1.typepad.com/blog/2013/04/perception-of-the-arctic.html?cid=6a0133f03a1e37970b017c38a13c3a970b#comment-6a0133f03a1e37970b017c38a13c3a970b

(modified 18.04.2013) currently there's a funny situation with this measure, since 2013 is both 4 days ahead and 4 days behind 2012!

I might calculate this regularly to exercise my rusty maths. Other people on the site do such magnificent analysis that my best efforts would pale in comparison. Anyway, a proper calculation can now be done since the area of 2013 has dropped below 2011 maximum. Very unsurprisingly the spring progress 'ahead-days(-behind) -value' dropped to only +4,32 with the inclusion of the year 2011 maximum date.

Comparing decades to this year values are 1980s ~+15,3 ;1990s ~+3,8 ; 2000s ~-4,4 so the arctic ice area loss in early spring has been slow this year compared to 2000s but then again most of you knew that already.

[Pmt111500]

Testing LibreOffice, posting an unnecessary image of CT SIA decades, and of course 2013 is only included up to the latest date, is CT at day 108 this year (would be april 16th in my count)?

while at the spreadsheet, checked also the 11-year averages of CT SIA NH to filter out solar influence (though it's minimal in the arctic anyway, ocean, clouds and ice rule there, ok, wind too somewhat) , the image I planned to do in 2009 when one computer crashed. that doesn't look linear to me.

(modified 19.04.2013)
the third image is what it looks like 'summer arctic ice mess' :-/. Split the CT record in weeks, took the differences between consecutive weeks, plotted against years (this leaves out two or one day each year since 364/52=7), and added linear trendlines for each week. I think I saw someone do just this some two years back somewhere (Chris Reynolds?) but wanted to check myself. No accurate equations for the trendlines just yet. (modified 20.04.2013) the image modified, still looks like a mess but now at least some structure maybe be seen.

aa-aannd the fourth image and the last one today and to this message is an approximation of the weekly area loss rate during summers (~half a year centered on summer solstice) in the available record (CT SIA). I'm  not entirely sure if I got that done correctly, at least I had to do it twice over since there was an obvious error in the transfer of data from one sheet to another... here I probably should've used some better smoothing in some step. But if someone gets something similar, then maybe it's correct. The  Further cautionary advice from far back (1990s) :"Please always check how your lab equipment works before conducting a series of expensive experiments that lead to much confusion as the output of the machine is set at 90 degrees of the expected."

[ChrisReynolds]

PMT111500,

"I think I saw someone do just this some two years back somewhere (Chris Reynolds?) but wanted to check myself."

I don't think so, if I did I've forgotten (which happens).

The third graph is what's termed 'spaghetti', which is what's deterred me from posting similar rate of change graphs. However it seems to be picking out the basic form - rate of change max around June/July, due to max insolation.

The fourth graph picks out the post 2007 step increase in area loss over the summer.

If I can make some suggestions you might find interesting: Why not try using a longer time period than weeks? Months would seem an interesting approach, but rather than strict months  an approach I've not tried is the average rate of change for some longer period +/- the date of the summer solstice. One I have done is to split the melt period into pre and post the first week of June (which is when the Arctic melt really starts to get going), i.e. melt from Max to early June and early June to minimum, that's interesting. EDIT - you also might consider looking at rates of change from mid September into the early freeze season.

Never heard of Libre Office, I've got a thread on developer's corner about spreadsheeting in Excel. If you get to grips with Libre Office you might like to consider doing a thread there for people who don't have Excel. There are plenty of people who use Linux and can't do Excel spreadsheets.

[Laurent]

Libreoffice is available here :
Windows and English language
http://www.libreoffice.org/download/?type=win-x86&lang=en-GB&version=4.0.2

[Neven]

Judging from the SIC maps on the ASIG, 2013 seems to be ahead of 2012 by two weeks in the Bering Sea.

[Pmt111500]

Chris Reynolds, thank you for the suggestions, I may dig into those. I thought to make a similar graphs (to 3&4) on winter next, but maybe I'll first look on to your suggestions. Anyway, it'll be earliest late next week, the spring has finally started here and the yard needs attention. LibreOffice is an offshoot of OpenOffice, and seems to work better in converting excel files. Just installed it for a friends suggestion since no excel and some previous work-related excel files may need additional work somewhere in time. Analysis tools aren't too good in the basic version, but maybe there are some extensions available. Anyway it's been so long since I've done any extensive statistics I may have to look on some textbooks too.

[Pmt111500]

At last looked at the spreadsheet again, and after some thinking realized it might be good to rearrange data so the astronomical seasons centered on equinoxes and solstices can be easily be kept in continuous tables. Further I think it would be better to use data area equal to synodic months or their halves as the analysis base. I think the basic data tables have to be done separately for each season since (synodic) lunar year is ~354,36 days and solar ~365,25, and the lunar cycles are on different phases each sequential year. (this would leave out 11 or twelwe days a year centered on the opposite equinox/solstice). As this is daily data, thought of using alternating 29 and 30 day months as a mockup of the lunar year. This procedure might keep the lunar phase during the analysis period (season) sufficiently in the same place so the possible effect can be seen or removed... I think this makes sense but suggestions are of course welcome. This will take a while. 

(modified 26.04.2013, 14:45) These lunar calendars are hard! After trying for some 4 hours made a mod incorporating elements of arabic and chinese lunar calendars, but as of yet it's somewhat unchecked and the numbering of lunar months is a headache. I should still check how it keeps the full moons in near the same date number (trying to fix it within accuracy of 4 days...) . Additional fun was to set the number zero year of this lunar calendar to begin at december 1979 solstice, ;-).  Added also additional Lunar Leap Days and leap months as necessary (can't escape these). I may need help with this one, so I might ask a friend-astronomer to check this one.

(modified 28.04.2013) Abandoning the 19-year Metonic cycle, instead adopting a 27-year cycle which seems to keep the dates of full moons within 3 lunar month days (the orbit of the moon varies in lenght during the solar year and other astronomical disturbances). Trying (of course) to minimize the number of intercalary months and days (used to keep the lunar months near the (unnatural) solar months). moved the beginning of the calendar to 21.12.1980 (full moon at solstice) just for fun. There's now two types of leap months (3-year and 27-year) and two types of leap days (which have a zero as the date number and are out of the normal count of days...) in the system. Might need to add a third type of leap day (which would occur on century or millennial time scales) to get to the same accuracy as the solar Gregorian calendar. I'm starting to doubt this will lead to anything relevant.

[Pmt111500]

2nd of May update (actually the last data point is earlier, but hey, if we knew everything instantly we would have no need for analysis.):

CT Sea ice area for northern hemisphere:
the area decrease is on par with the 2010s average
the area decrease is 2 days ahead of 2000s average
the area decrease is 10,3 days ahead of 1990s average
the area decrease is 18 days ahead of 1980s average

There could be some sort of trend here.

the earliest year with less than 2013 on this date was 1990.
biggest outliers were 1979-1983, 1985 and 1987 with positive values of 20 days or more.

There could be some sort of trend here.

The largest negative value was for 2007 which was 10 days ahead on this date.

Finally:
Currently the melt is 9,2 days ahead of the long term average.


(Modified 5.4.2013) No news on the lunisolar calendar front, but on an initial check there seems not to be a significant correlation between the spring tides and the speed of ice loss. Probably it's just NORMAL tides AND some weather attributes combining in f.e. in the the cracking event. Of course the sun and ocean are the main drivers of the yearly loss. But there could have been some component of lunar origin. Negative result is a result too. I think I'll continue this for a bit, at least to make the statistics (and the LibreOffice functions) more clear to myself, no use in discarding good groundwork just for there was no significant results.

[Pmt111500]

The irregular update on this insignificant metric considering the thickness of the ice is as follows

CT SIA melt progress in days compared to the averages
2013 is now 7 days ahead the long term average
2013 is now 4 days behind the 2010-2012 average
2013 is now 1 day ahead the 2000s average
2013 is now 8 days ahead the 1990s average
2013 is now 16 days ahead the 1980s average

The next irregular update of this insignificant metric considering the thickness of the ice will possibly be in the end of may, once the melt has entered the arctic basin and the effect of MYI-FYI transition can be seen better in the area numbers.


(modified)
Below an image of how the CT ice area has diminshed over the years. Took weekly averages of 1979-2012 and compared that to the actual weekly values. I guess the speedup of the melt after 2007 Chris Reynolds has talked about might start on week 23 but on week 24 it's definitely notable. But anyway, on May and the first week of June the decrease looks still to behave quite linearly.

(modified again)
3 more images of the weekly deviations. The same protocol over the July-December period. July-August seems to show the same behavior as the late june. September-October might display exponential behavior from as early as 2001 and the faster decline seems to stop only on week 47, that would be late November... So, there seems to be a period of 23 weeks, in recent years at least, that has a speedier than linear decrease in ice amounts. I can't say of the confidence level of this statement though.

[ChrisReynolds]

PMT111500,

Thanks interesting way of plotting it. Week 24 seems reasonable to pick up the 'anomaly crash' around the first week of June.

When you say you 'compared to weekly values' do you mean you 'normalised' to the weekly values by dividing by the average for that week to get all the plots on the same basis?

[Pmt111500]

Chris Reynolds, just so, normalising is the proper word for that, I guess. Multiply the y-axis by hundred to get the percentages of the ice for the plotted weeks from the 1979-2012 average for the same week.

[ChrisReynolds]

Thanks,

I do get what you're doing now.

[Pmt111500]

the winter images have a milder slope and march-april is linear, january-february might have a slight increase in the rate of the decrease after 1998, this is likely connected to the marginal seas getting warmer (Ohotsk, Barents).

Just trying to picture what sort of functions would combine to produce the yearly pattern of loss of the ice, as there are at least two ways for energy to get to the Arctic (solar (CO2-influenced) and ocean overturning circulation) (add atmospheric heat transport, local atmospheric effects (methane/ozone, changes in clouds, if any?)) and the  reduction of solar energy for heating due plant activity (small and just for summer))

After seeing these, in my opinion, the function describing the behavior of sea ice area loss should have at least one second degree factor that would gradually die out for weeks 47-(winter)-22 (23), the step change could be for the sun (didn't you suggest this on some post?) reaching a more effective angle, then in autumn the atmospheric and oceanic heat transport would take over, and the winter signal would be the pure CO2 signal. However, I lack the capabilities to build such a function convincingly. y(area of sea ice) = a(heat transport)x²+b(solar angle)x₂²+c(CO₂)x₃+Dx₄(systematic changes in weather patterns??)...+/-([sigma]variability), where a=0 over the winter, b has a value over weeks 23-35, c is something about backradiation and here it ends. I just don't have the physical formulas to try to describe this better. (f.e. atmospheric and oceanic heat transport doesn't stop for winter)


(modified waaay later 10.07.2013)  would be real funny if something simple like -(x^((day#_{max n}/187.5+2)/2))+C would get you a correct answer.

[Pmt111500]

Continuing to play with weekly averages: tried to remove the yearly variation from the above weekly normalised values by dividing with the yearly normalised values. then took the inverse 1/x of this and got the plot below.

in the autumn:
the highest values are from years 2012, 2007, 2011, 2008, 2010, 2009 and 1995
the lowest values are from years 1996, 1987, 1992, 1980, 1979, 1986, 1988 and 1981

in the spring:
low anomalistic years are 2012, 2007 and 2010
high anomalistic years are 1989, 1988, 1996, 1990, 1986

the 'speed bump year' is 1995.

what the numbers actually mean, I don't exactly know, I'll let the others correct. (measure of deviation from the linear trend of the decrase in ice area, where 1=linear, or maybe the slope on these curves would be that?)

(modified 18.05.2013) continuing plotting CT SIA basics: at broad spectrum, the yearly pattern does not show much change. though timing of the fastest melts and gains might be changing a bit... (changed the color scheme to match), practising weighted averages so the making of the value for the plotted week is 0,2+0,6+0,2=1 (alas, symmetrical weights making 1, but this large spread is just a choice  to get a more stable looking plot :-P, and does not have anything to do with satellite measurements) for three consecutive weeks, this cuts the variability somewhat.

(Modified later) tried to isolate the most abnormal weeks of melt/freeze. the resulting plot added. this takes a bit of explaining: the sign of the metric depends on whether it is freezing or not, so high positive values in summer mean extreme melt, and high positive values in winter mean very fast freezing. the low values in summer mean abnormally slow melt and low values in winter mean low amount of freezing. I'm pretty certain that there is a simple mathematical 'trick' to get the values for winter properly inverted, but I've not yet found out how to do that..
Anyway, in the weekly dataset (the two extra days brutally removed) the freeze/melt stagnation weeks (also known as max/min weeks of ice) are 10 and 37. hint for those wanting to repropduce this for other datasets: |absolute values|

changed the color scheme for the 1st image, now all images have the same coloring (except for 1979 and 1980)

[Pmt111500]

I'm very likely doing repetitions of graphs made by others, but anyway, ASIG-page does not have all of these basic figures. the following is a full set of the weekly normalised values presented in the previous post (bi-monthly images) divided by yearly averages, inverted and multiplied by 100 to get percentages. (CTSIAyearlycyclechanges(weeks).png spread out and multiplied by hundred) It looks like there have been changes in the behavior of the arctic sea ice area loss/gain. These are at about weeks 975 and 1450 (september 1997 and october 2006).

(Modified) now I'd need some help from someone to confirm the second image. tried to look for the often stated claim of increasing weather variability, caused by AGW, from the CT SIA data and got this plot. this would be the 'abnormal weeks' absolute values divided in three sections, spread out 1980-2012 with the automated trendline analysis from the libreoffice. variability is huge so it is pretty untrustworthy. but, if one takes the trendlines at facevalue, it looks like the overall variability has increased a bit and the highest class in this metric has increased the most. so more wacky weather for arctic (and the rest of the globe)

(Modified 20.05.2013) Third image is about what i think is a test to check how well a linear equation would explain arctic sea ice loss/gain each year. Method is crude and after the fact (using data also from later years to calculate the R^2 for a given year), so it doesn't mean people (or university students doing their finals) should've been shouting in 1997 that assuming linear decrease of ice is blatantly false. The step change (in winter 1997-1998) in the behavior is again clear (did the huge el Nino have this effect on sea ice, demolishing the ice in the marginal seas, so Arctic Ocean would again behave linearly (or nearly so) for the next 9 years?))
Some maybe relevant reading:
http://en.wikipedia.org/wiki/Coefficient_of_multiple_correlation
http://en.wikipedia.org/wiki/Ordered_logit
http://en.wikipedia.org/wiki/Naive_Bayes_classifier
some extra course, to me at least:
http://en.wikipedia.org/wiki/Extremal_principles_in_non-equilibrium_thermodynamics
http://en.wikipedia.org/wiki/Dissipative_system
http://en.wikipedia.org/wiki/Curse_of_dimensionality
(No Monte Carlo for me yet.)
somewhat familiar subject is somehow connected to these,
http://en.wikipedia.org/wiki/Autocatalytic_reactions_and_order_creation

[Pmt111500]

Update: Now that the pack-ice is slowly relaxing itself to the driving forces of cyclones, waves and sunshine (where available) the CT SIA numbers are showing a decrease in the speed of area decrease. The steric stresses  that may have driven some new multiyear ice formation during winter have finished their job for the winter (oops, using language of chemistry for compaction of pressure ridges). Wide areas of arctic ocean coasts have coastal leads where the pack-ice may relax. Only in CAA and northern Greenland  the ice is still attached to the coast. This is of little importance since these are also the areas of thickest ice, so the driving forces should be huge to get slabs of FYI underneath it. So in my opinion the formation of potential multiyear ice has stopped. Possibly it stopped already in the fragmentation event , but I don't understand why that happened. The persistent low in the Kara Sea region may drive some FYI to attach to MYI but at the same time the churning, vertical mixing of water and the atmospheric heat transport it drives makes this of little importance of increasing the Volume of ice. In fact, many have suggested the May numbers for Volume may show a big decrease! This should show up in the CT SIA record shortly, possibly on week 22, that on my datasheet begins on 6th of June.  That week is a possible time for the next update of this metric.  Here are the values for CT SIA area decrease compared to previous years in the available record (up to May 30th, I'm still not sure if it should be 31st, but a day here or there, does it matter? even weather events take longer, when looked on this scale):

2013 is now 4 days ahead the long term average
2013 is now 7 days behind the 2010-2012 average
2013 is now 2,5 days behind the 2000s average
2013 is now 4 days ahead the 1990s average
2013 is now 14 days ahead the 1980s average

(modified) It seems that (data from wunderground) the hottest places today 1.6.2013 in Europe are in Lapland valleys and Tallinn, Estonia, recording at +31C. Looks like one has to go to southern Turkey to find hotter temperatures. In here, it's only +25C with clouds all over. Spring  lasted for about a month, and I guess this is now summer. Really a fast transition. In the news there was a Lapland fellow who complained about working conditions outside. From -25 to +25 in a month. Never know what to wear. Forecast is, cooler temperatures should return in a week or so, so he can continue his business without breaking a sweat. :-) On the news was also a japanese tourist who was disappointed of the weather because he wanted to know how it feels to be cold. :-D Wrong season for that!!!

(modfied) There is such a huge amount of data available on most things arctic nowadays so it's easy to feel oneself overwhelmed of the magnitude of things. To fight this effect of huge amounts of data somewhat, I took the Hudson Bay Modis image in full resolution, cleaned the image as much as I could, while preserving the coastline (I do not have the landmask image to do that easily). Then I reduced the size of the image to the scale 10km/pixel, to get to the level of early 1970s weather satellites. The process took almost an hour off my life but I think it was worth it. As an afterthought, I should've probably removed the colors too. So here's the early 1970's style satellite image of Hudson Bay on 6/6/2013:

and beneath it a detail of the full image, this is a bit of blue ice between (a bit north) the thick clouds in the middle.

Then an image after/before cleanup 11/06/2013 from the central area of the PAC2013 (this happens to be in r04c04 block in the arctic mosaic). This cyclone seems to have thinner clouds in the middle what ever that means. Maybe that's common for high arctic lows. The cloud bands seem to spiral towards the center (a bit to lower left). I'm not sure what i see but there might be almost open water under the thicker clouds. or maybe that's a processing artifact. when and if the PAC2013 goes away, we'll see. But this cyclone must get its energy from somewhere. open water under the lobes of cyclone might explain some of it.

[Pmt111500]

2013 is now 7,5 days ahead the long term average
2013 is now 4 days behind the 2010-2012 average
2013 is now 2 days ahead the 2000s average
2013 is now 8 days ahead the 1990s average
2013 is now 15 days ahead the 1980s average

The post-2007 june cliff started, pretty much as scheduled. High Arctic Summer enters the period when most ice ridges formed during winter have relaxed and flattened out in line with the floes. This creates small leads in their places (in fact the OBuoy#8 might be in just such a location). At the same time the big rivers discharging to arctic are shedding their final ice cover. And there's the persistent cyclone which frustrates the satellite image interpreters and possibly destroys some halocline. The persistence and the location of it make it look different from the normal extratropical cyclones, at least to me.

The unsurprising conclusion of this update is that more melt is going to happen this year. This melt should end, the latest, sometime in october.


PS. Obuoy#8 is/was somewhere in here, under the clouds (worldview): http://map2.vis.earthdata.nasa.gov/imagegen/index.php?TIME=2013159&extent=-1281661.190435,-250252.349154,-1106045.190435,-141196.349154&epsg=3413&layers=MODIS_Terra_CorrectedReflectance_TrueColor,arctic_coastlines_3413&format=image/jpeg&width=686&height=426

The sky has cleared on the buoy location!

(Modified 19.06.2013) just having fun, drew a line around the single image arctic mosaic of the area that's not been substantially fractured or turned blue, result: 2011200 km². Possibly checking the proper numbers on sunday (Midsummer hangover-day, so miscalcúlations are likely, looking forward of losing my temper modifying the spreadsheets. (If you know it'll happen then it isn't that bad.)

[Pmt111500]

Time to move this thread back to the front page. There's been quite a lot of talk on the site of the goods and bads of various data sets concerning arctic sea ice. CT SIA is though still the longest easily obtainable data set so I'll continue these irregular updates, yes it's monday and a week after than mentioned in the previous post. Current numbers of days +ahead/-behind for 2013 are:
 
2013 is now 5,5 days ahead the long term average
2013 is now 6 days behind the 2010-2012 average
2013 is now 1 days ahead the 2000s average
2013 is now 7 days ahead the 1990s average
2013 is now 15 days ahead the 1980s average

I was expecting the June cliff to continue until the last week of June and it to be a little steeper than it did. Going by the CT SIA numbers, 2013 has taken 2 days extra off the melting compared to the 2010s. Maybe the PAC2013 has prevented surface melt in the extent it has been the past three years. There's been discussion of the bottom melt http://forum.arctic-sea-ice.net/index.php/topic,398.0.html , and the predicting bottom melt remains somewhat a mystery. Anyway the pack has continued the rounding action through June, like Werther stated 29.6.2013 (http://forum.arctic-sea-ice.net/index.php/topic,92.msg8517.html#msg8517) "‘mesh-pack’ today. It is about 1,5 Mkm2 now. It has already lost 300K this season." The number I got on 19th was 2,01Mkm2 (previous message), though it's not done as rigorously or accurately as Werther's, I guess. My estimate for first year rounded floes to disappear was 55-75 days on 25.6, http://forum.arctic-sea-ice.net/index.php/topic,167.msg8116.html#msg8116 , this would place some large drops to August. I'm expecting the clouds to continue and increase their harrassement in the spectral bands measured from satellites, but this year has been a difficult one to get. (on the 5th year now, occasionally trying to statistically predict physical properties affecting ice (was it 2 on the scale mentioned in the blog?) or reproducing some of the simpler works of scientists.)

(modified): As the CT SIA wobbles at times so much with alternating drops and not-so steep drops, and a physical process such as ice melt, should be continuous (mathematically), I thought to apply weighted averaging to it. I think I'm going to change my files to use 5-day weighted averages (1x = 0,32(0)+0,225(1)+0,225(-1)+0,115(2)+0,115(-2)) since it looks like it removes most of the effects the short period processes such as melt pond draining and developing have on the CT SIA actual numbers, while still retaining some of it. Attached an unclear image of the effect smoothing has (month before June Cliff (May 6th - June 6th) on the curves. I'll just have to check how the smoothing performs on winter numbers. Of course, smoothing does not help with the effect clouds may have on the actual numbers.

Hmm, it does look a bit smoother. Having such a data set might make the relationships with some weather systems more easily seen, and anyway, it makes for prettier graphs.

[Neven]

Thanks a lot for the update, Pmt111500.

[Pmt111500]

A version of CAPIE calculated Using 5-day average CT SIA and the intermittent IJIS extent. the odd scale on x-axis should be divided at spaces of about one lunar month.

In the series 'Arctic Ice basics' I took differences from all-time daily maximums and minimums to all-time daily average and divided the results. So, on days with y-value of over 1 the maximums have diverged from the average more than minimums and the opposite is true for values under 1. X-axis scale is divided to 30,5 days (should be 30,4375, but what the heck) to emulate so called normal months. The June cliff can be noticed very clearly here. The day 61 peak is because of the leap day. umm, maybe this should be inverted.

attempted to do 'wiggle matching' to connect the stages of melt over the years. This produced no true results since autocorrelation is too high in the method selected; attaching a heavily modified graph of three year averages/day which is inverted for obvious reasons. the start of the x-axis is set on the beginning of the June cliff (8th of June) and the only notable thing is how the gap between early years and post-2007 grows ever larger during the melt period. X-axis end is August the 31st. Attempted to 'predict' 2013 behavior by the year to year increases of the decreases in ice areas on the graph and got wildly variable results depending on which periods were included (aka. the problem, 'was 2012 an exceptional year or just showing a steepening trend?'). the last one was placing some huge drops to mid-august so keeping my vote for a new record. (catching on 2012 on August 17th) (full range is an awful 3,3 - 1,9 Mkm2, too bad ice isn't behaving nicely this year.).

sorry, no update on the topic of the thread just yet (using 5-day weighted averages takes a bit more time)

[Pmt111500]

Now that 2013 has gone under some minimums in 1980s, I had to do this a bit differently from previous updates.

2013 is now 13 days ahead the long term average
2013 is now 5 days behind the 2010-2012 average
2013 is now 5 days ahead the 2000s average
2013 is now 13 days ahead the 1990s average
2013 is now 31 days ahead the 1980s average

If this loss of ice continues, I'll have to drop comparing to 1980s since the absolute minimum average for 1980s is 5,2272 Mkm2 and 2013 is just 2 days away from this at the current speed of loss. I now use 5 day averages so I'll have to wait 2 more days to see if 2013 goes under 1980s average minimum. Anyway, 2013 is now 53 days from average minimum date of 12th of september. There's been no noticeable trend in the dates of minimums as far as I can see, which might be explained by the presence of MYI that cracks to fill the voids left by FYI melting. Some very late dates are found, but for practical purposes (such as gambling) it may be advisable to stick to 2nd and 3rd weeks of september on guessing the minimum date. 2013 is starting to look like it belongs to the 2010s. I haven't done much on the spreadsheets, somehow I had lost one date number on one spreadsheet, and this may have caused some discrepancies on dates, and finding and fixing this took a while. thanks for attention.


(Modified:15:08) Added the global CT SIA figure due some curious comments elsewhere in the net. (Modified 22.07.2013) Added another Global Ice Area figure. 2013 may become the first year since 1979 that has a higher peak in the summer max than in the november max. That is pretty unusual. (I don't read much about the R2-values, they're not strictly the best measure of correlation in this case anyway.)

[Pmt111500]

Irregular update, this might be the last one comparing early years (1980s, 1990s) to this year. As was very likely expected by all the regular contributors of this forum the CT SIA numerical decline continued after the previous update and so -

Tomorrow is likely the day when 2013 5-day weighted average passes the minimum of the 1980s. The earliest minimum in the available record (CT SIA for Arctic) is on 2nd of September (1992 and 2005, 39 days to go) and the latest is the oddball year 1995 on 29th of September (66 days to go, 5 days later than anyother year). Minimum of 1990s is 4,7804 Mkm² on Sept-9th and minimum for 2000s is 3,9504 Mkm² also on Sept-9th (46 days to go)

Average daily loss from Jul-23rd (five day smooth is 2 days late) to Sept-9th has been 43153,9287 km² during the 2010s thus far. This would give 2013 a drop of 2,071388 Mkm² from the 23th of July value and this would make a minimum of 3,163104 Mkm² (the 6th smallest). 2010-2012 have hurled past the 1990s minimum in few days on average (actually this might happen in 1-3 days, so I'll have to think some other way of comparing 2013 to the early record) and passed under 2000s minimum on August 4th (possible update here is thus on Aug 6th)

2013 is now about 18¼ days ahead the long term average
2013 is now approximately 3½ days behind the 2010-2012 average
2013 is now almost 9 days ahead the 2000s average
2013 is now ~24¾ days ahead the 1990s average
2013 is now c.47 days ahead the 1980s average

Average losses/day graph. Shamelessly cherrypicking 1985 as the starting point for another graph tells that since 1985 the arctic ice has lost on average almost 306 km²/day more each year during Jul 23rd - Sept 12th. I don't know if this is a proper value for GHG-effect, anyway, this value should be calculated from the volumes.

For some light entertainment, calculated (again, last time was in 2009 if i recall correctly) what would happen if 2013 started to equal maximum losses from previous years every day from here on. Result: total loss of Ice on Sept 17th (image not included, check yourself), on the other hand, if 2013 would do maximum gains/minimum melts from now on the melt should end in two days and the value for Sept 12 would be 6,4579 Mkm² (again, no image included) and the historical maximum would be passed on Nov 7th. Also checked how many record days 2013 has thus far. This is equal, 4 maximum gain/minimum loss days vs. 4 minimum gains/maximum loss days. However the sum of these record days is -601884,531 km², so it's not like there's been no changes in the extreme value distributions.

what someone might want to check is whether land temperature developments above 30 north (ferrel-polar cells of atmospheric circulation) correlates to the average loss image, I've no idea where to find that data.

(modified) as there is a slight trend to increasing losses during this time of year did a Jim Pettit -style graph 'if 2013 behaves like recent years what'll happen?' Included the above, estimated trend in the calculus. This doesn't change the result much as can be seen. 2013 would have to start to behave like a badass 2008 to get near a new record. On the other hand, it might just behave like 2013 behaves and that might be alltogether a different thing. 2 out of six realizations put 2013 challenging second spot in minimums so at least (inserting own opinion) 33% chance of 2nd smallest minimum.

Cherrypicking the steepest trendline I dare, ;-) 2013 still does not yield a new record but ends up some 333000km2  short of it. But still, 2013 has been behaving unlike anything seen previously. (not changing my vote, voting possibly ended already?)

(Modified 26.07.2013) A dubious excercise on Global CT SIA. Ordered numbers separately for each year, then took the difference of smallest and largest value of each set of values (so this is a mix of separate dates from different years) and got the R² shown in the right bottom corner. As can be seen from the main image, ordered values of global sea ice have been decreasing, but if one looks only variance, someone (read: 'deniers') might conclude that Southern Ocean Ice balances the decrease in Arctic Ocean, which is clearly not the case.

(Modified 06.08.2013)
4th image: not sure if I did this correctly (ah, pretty sure that there's some glitch in it but anyway it shouldn't change the shape of things...), and what does this tell. Global daily values divided by daily averages, then multiplied by yearly averages and again divided by the average of the whole record. Then took 11 year means from the record where it was possible and smoothed the ends by ever shortening averages. Then shaded the ends of the graph indicating the increase in statistical uncertainty.

Now what happened in 1999? Again this is calculated after the fact, so it doesn't mean people (or just graduated students of limited statistical capabilities) should have been shouting that "Increase in Antarctic Sea Ice won't match the loss in the Arctic Sea Ice, thus the albedo will decrease!". Anyway 11 years is such a long time that an ocean current might notice it, but the sun it isn't. Can this be connected to the 1998 el Nino somehow? Did the southern ocean start to increase its heat content in 1999?

[Pmt111500]

'the stall'-update, using 5-day smooths.

2013 is now only 20,7 days ahead the long term average
2013 is now massively 11,2 days behind the 2010-2012 average
2013 is now pretty exactly 3⅓ days ahead the 2000s average
2013 is now highly suspiciously ~30,4 days ahead the 1990s average

2013 has now beaten the 1980s average minimum.  Consequently, comparisons to 'good old 1980s' will be discontinued.

The early years having a lower minimum than 2013 on Aug 3rd are:
1984 (reached Aug 3rd, 2013 value 38 days later ),
1990 (+32,7 days),
1991 (+34 days),
1993 (+15,3 days) and
1995 (+14,3 days)

1998 onwards all the years have lower minimum than 2013 on Aug 3rd.
2013 is now ~4,6 days ahead of 'post-Biggie El Nino' average

So the expected drop didn't happen yet, so this took about half an hour more to make. There's still the deep low developing on the Arctic that may make the icy waters less icy and at least more snowy and slushy. I've talked elsewhere about the stall, and what I think of it. 

No graphs nor images ready for this one.

I should be picking red and black currant in the garden, or making apple/plum jam, but maybe they can wait a couple days still, since the weather forecast promises cloudy and less sweaty weather for picking then.

Nothing further to say on this update. (forgot to update the 1990s value, correction now in place)

[Pmt111500]

Irregular update to this pretty irrelevant metric considering the stage of melt.

2013 is now close to 11 days behind the 2010-2012 average
2013 is now near 6 days ahead the 2000s average

1990s and 1980s have more Arctic sea ice at their average minimums than 2013 on Aug 9th according to CT SIA. We have pretty close to one month more melt to go. So there's been a decrease of one month on ice season on some areas in c.20 years. If this sort of seasonal change spills out of arctic, what we might expect? Month longer droughts? Month longer wet seasons? Month longer hurricane seasons? Month longer, well what ever but ice.

As there's no comparing the 2013 to 1990s or 1980s anymore I took a couple other averages.

2013 is now 21 days ahead the 1998-2006 average
but
2013 is only 3 days ahead the 1998-2012 average
so
2013 is now 9 days behind the 2007-2012 average

Those who remember 1990s, might enjoy this:

1995 has probably lost to 2013 already, 2001's a loser too. It's only a matter of few days more data to 1998-2006 to lose too. One might be tempted to say that on late melt season there's been a three week decrease in ice season on some areas in just 10 years, but some denialist might take offense and say he likes to kill Santas. Well, they may like to do it anyway.

This may well be the last update for this pretty irrelevant metric since discussion on the prediction thread. So, clouds may well have an effect to arctic temperatures that helps to preserve ice, without causing anymore upwelling than currently present. Anyway, 2013 looks become a member of the post-2007 years looking only at the minimums. For this, 2013 has to reach at least 3,70Mkm2 and thus lose some 600k of CT SIA in a month. Can 2013 do it? *Suspenseful drumming*

(modified 14.08.2013) the daily CT SIA of 2013 has dropped below the 5-day average minimum of 1998-2006! 573968 km2 to go to the above mentioned limit. There still might be some talk of recovery, but it'll be recovery compared to the last 5 years - Not to the 1980s, not to the 1990s, but maybe to poor sort of 2000s levels. Everybody having memories of over 23 years old are thus either demented or dishonest if they claim otherwise (not meant to irritate anyone, just showing my trust on CT SIA numbers). updates might continue if there are further developments.

(Modified 18.08.2013) 4,212497 Mkm2 (value for 14.08.2013) means that 2009 is behind 2013. So are every year before 2007.

full record:
2013 is 11 days behind from the 2007-2012 average.
2013 is 15,6 days behind 2012.
2013 is 13,6 days behind 2011
2013 is 8,5 days  behind 2010
2013 is 1,5 days ahead of 2009
2013 is 7,6 days behind 2008
2013 is 12,7 days behind 2007

This is getting exiting.

[Pmt111500]

Since all the pre-2007 years have lost to 2013 by now it might be ok to compare the areas directly and not days.
2013 is now 1,680350Mkm ² ahead of the 1980s average
2013 is now 1,069645 Mkm² ahead of the 1990s average
2013 is now 0,284784 Mkm² ahead of the 2000s average
2013 is now 0,711273 Mkm² behind of the 2010-2012 average.

It'll be interesting to see what sort of cycles may be proposed to explain this 'recovery' to ~2009 levels. And equally interesting to see what sort of methane levels are to be found on the arctic atmosphere when the temperatures drop to sub-photosynthesis levels.

(Modified 24.08.2013) The 5-day average CT SIA is now 3,908799224 Mkm² (for the 21st of August), with 19 days to go the average drop to get to the arbitrarily chosen 3,7Mkm² limit of post 2007 years is now 11000km²/day. (oh, actually it looks like the limit should be 3,64Mkm² and still no one has shouted foul. thanks.)

[Pmt111500]

Back to the subject of the thread. The graph is an approximation on how 2013 performed compared to 2012. The 'analysis' isn't on par with scientific standards so I advise using some other source or making one yourself if you find this interesting. Tried to put some of the items discussed here on the graph on correct times of the year, please check also those before showing this to anyone. The last date is August 21st since that was about the time I lost faith in 2013 to ever catch up on 2012 measured by this pretty irrelevant metric. If one doesn't have a proper standard dataset to compare to, the results are flimsy, hence the name of the graph. There might be another update shortly on this thread, since 2013 might cross the dividing line between 1998-2006 vs. 2007-2012. But, if there is an early turnaround, this year fits right in between the 'rainbowgraph' a couple of messages back.

If I'd have to say something about this, it doesn't tell anything about which was more exceptional, 2012 or 2013. There's been talk about the 'long tail' of arctic sea ice disappearance, which could be driven f.e. by the increased fresh water content on the surface of the arctic ocean easing and increasing the freezing of the surface. I do not know if the late summer/autumn inflows have increased in Siberia or Arctic Canada. The other reason for this could be the increased fogginess/cloudiness (increasing the albedo) due the destruction of the Polar Vortex, which usually keeps the Arctic a bit more clear of clouds. The increase in cloudiness could also be from the melting of the ice and increased evaporation where it can happen. The third reason for this years apparent slowness, I can come up with is the scary one, if the heat transport to arctic has remained somewhat constant compared to the post-2007 years, this heat would have to be located in deeper ocean, where it may interact with yedoma (was it the correct word) thus making the methane concentrations above arctic highly interesting come the start of the freezing. It'll take at least 2 more days of data of 2013 to make the decision whether this year belongs firmly to post-2007 years, confirming (in my mind) that the linear assumption of sea ice disappearance is still wrong. Update on a separate message possibly then.

[Pmt111500]

Now that 2013 has shown it belongs to the group of post-2007 years, looking at minimums, I made a bold move and omitted all data prior 2007. Made a graph to look on how much these years deviate from the averages until sept-8. As can be seen, 2012 is about a quarter more "out there" than 2013, but to the other direction. 2010 might be used as a guideline for this new regime that might be prevalent now. One has to remember though, that large deviations from linearity arise before a chaotic collapse to a new regime, so this year is a pain in the ass for the statisticians who like to use various non-linear functions to predict (or guess) the future behavior of CT SIA numbers. more analysis might follow after this year's minimum has been confirmed.

[jdallen]

My sense is, that we are at the start of a period which will go through serious oscillation in SIA and SIE. I think they are going to lose their importance as indices to volume and sensible heat in the Arctic Ocean.

On an annual basis, I think we will consistently see return to maxima in line with what we saw last winter for some time to come; there won't be enough surplus heat either stored or entering the arctic to prevent it, but we may start seeing slow reduction of average thickness at maximum.

What happens in summer will be wildly variable until sea water temps rise.  There may be some predictable heat content threshold we could compute.  Year over year, for some time the wildcard will be cloud cover. It may be a sufficiently powerful negative feedback it will ruin some people's (and my) prediction of an ice free summer by 2020.  I think it will continue until after we reach the sea heat conten threshold I'm imagining. 

[Pmt111500]

this is what I meant about 2013 being a PITA (pain in the a**e) for statistical approach. Guess what the confidence for continued exponential decrease after 2013 now is. This year is such an outlier in many of the basic measures, that I'm tempted to believe that 2007 was a tipping point (or a step change, oh dear) to steeper linear decrease. The trend is still for continued decrease, but as of now, looking at post jan-2007 values only, it looks like the decrease in area might still be linear (from 2007 onwards), no matter what it is on the volume.

The first image is so noisy I decided to check averages, it looks like taking bimonthly averages smooths weather noise out, and allows for speculation of climatic causes affecting sea ice, this is though only my own opinion. translation: weekly-7 biweekly-15 monthly-29 bimonthly-61 seasonal-91 halfyearly-183 yearly-366 day averages. Take more than seasonal averages and you begin to lose too much detail, take shorter than monthly averages you get weather.

[jdallen]

this is what I meant about 2013 being a PITA (pain in the a**e) for statistical approach. Guess what the confidence for continued exponential decrease after 2013 now is. This year is such an outlier in many of the basic measures, that I'm tempted to believe that 2007 was a tipping point (or a step change, oh dear) to steeper linear decrease. The trend is still for continued decrease, but as of now, looking at post jan-2007 values only, it looks like the decrease in area might still be linear (from 2007 onwards), no matter what it is on the volume.

The first image is so noisy I decided to check averages, it looks like taking bimonthly averages smooths weather noise out, and allows for speculation of climatic causes affecting sea ice, this is though only my own opinion. translation: weekly-7 biweekly-15 monthly-29 bimonthly-61 seasonal-91 halfyearly-183 yearly-366 day averages. Take more than seasonal averages and you begin to lose too much detail, take shorter than monthly averages you get weather.

Your starting point is 2007-01-01?

I'm having a hard time agreeing with your conclusion I'm afraid, looking just at your graph.  Looking at the noise, I'm not convinced the SIA deviations over time are giving us a clear enough signal to do that.  The amplitude in those waves just seems too high, even with the smoothing.

[Pmt111500]

Your starting point is 2007-01-01?

Yes, it was easy to split from the full table. I should have likely used october or september-2006 instead as the starting point for the possible new regime. There are many things possibly affecting SIA originating outside the Arctic and correlations with these might be better found using the 1 month - 3 month smoothed data, that's all I meant by 'climatic causes'.

[jdallen]

Your starting point is 2007-01-01?

Yes, it was easy to split from the full table. I should have likely used october or september-2006 instead as the starting point for the possible new regime. There are many things possibly affecting SIA originating outside the Arctic and correlations with these might be better found using the 1 month - 3 month smoothed data, that's all I meant by 'climatic causes'.

Smoothing makes sense for sure; 1 month or 3 month, I'd say that's a toss up.  I agree anything less than that is lost in noise; anything more loses too much definition.

I'm still having a hard time deciding what the pattern implies.  Not sure it is a good reflection of a trend.

[Pmt111500]

"there won't be enough surplus heat either stored or entering the arctic to prevent it, but we may start seeing slow reduction of average thickness at maximum." my opinion too.

on the subject of the thread (freezing period),
2013 is now 11 days ahead of 2007-2012 average, but
2013 is now 9 days behind the 1998-2012 in freezing. 
there are not yet comparisons to other periods, since they got more ice at the minimum than 2013 now has.

one potentially interesting idea for a table came to mind, 'the days under previous years' minimums.' this'll have to wait until CT SIA 2013 reaches 5,5Mkm2 (1980 minimum), this value should be reached at about Oct 20th, if 2013 continues this subduing attitude towards cAGW. then 2013 would have spent 92 days under the 1980 minimum. 1980 was a cold year at least in here (as far as I can remember), so perhaps that's not a surprise.

I've not even begun the (half-promised) analysis of this year, other things have been on my mind. Like, should I try to add some things arctic to Foster & Rahmstorf (2011), of course it wouldn't be on the scientific level of things, but some incomplete science-as-a-hobby type of thing. This would require seeking (at least) monthly values of several variables, and making pretty-advanced-for-me type of stats on those. This should be done separately to NH and SH, too and possibly also to under-Hadley vs under-extratropical-atmosphere data sets so it would be a huge undertaking. I'm very likely too lazy for that.

Meanwhile the arctic ice has begun to grow and there's no reason not to occasionally check how many days 2013-2014 is off from various values from previous years. So there'll likely be irregular updates here still.

[Bob Wallace]

Meanwhile the arctic ice has begun to grow

What's the evidence that the ice has begun to grow (freeze) as opposed to simply spreading out over a larger area?

[Pmt111500]

Meanwhile the arctic ice has begun to grow

What's the evidence that the ice has begun to grow (freeze) as opposed to simply spreading out over a larger area?

CT SIA numbers ;-).

[Bob Wallace]

There are two explanations for increasing CT SIA numbers.  Freezing is one.  Spreading of the pack is another (as long as the ice doesn't get spread so thin that it falls below the 15% threshold).

Look at Wipneus's regional extent and area graphs.  And at his maps.  What you'll see is expansion in areas where it does not seem cold enough to freeze salt water.  It looks more as if the ice is simply spreading out.

As if someone took the fan away from the punch bowl, the fan that was forcing the ice to one side.  Now pieces are free to bounce off each other and spread themselves out a bit.

[ChrisReynolds]

Bob,

Air temperatures across the whole of the Arctic Basin are now well below zero deg C (273K). Link - slow link direct to NCEP/NCAR.

It's air temperature that counts as this drives the heat loss that forms surface ice.

[Pmt111500]

this just to say that 'growth of ice' and 'growth in numbers measuring the area of ice' aren't necessarily the same, but if there's a number one can do some maths with it.

most or possibly all graphs on this thread are based on datasets http://arctic.atmos.uiuc.edu/cryosphere/timeseries.anom.1979-2008 and http://arctic.atmos.uiuc.edu/cryosphere/timeseries.global.anom.1979-2008
or versions of thereof (5-day-smooth),
if it hasn't been mentioned clearly enough previously. oh, there was CAPIE too, http://www.ijis.iarc.uaf.edu/seaice/extent/plot.csv is needed to reproduce that one.

I've not tried to do anything with the volume numbers produced by PIOMAS yet, though might some day. Likely not in this thread though, even though the headline is 'How many days earlier than 2012', which might include PIOMAS-numbers also. That might be a topic for another thread in 2014, if the PIOMAS daily values are available somewhere.

[jdallen]

Bob,

Air temperatures across the whole of the Arctic Basin are now well below zero deg C (273K).

It's air temperature that counts as this drives the heat loss that forms surface ice.

I'm not sure about the first, but agree with the 2nd.  I'm not sure what that link you posted is showing, Chris.  Other sites like Koeln don't show things any where near that cold across most of Siberia and sub-arctic Canada.