« Retrospective 2 | Main | SIE 2011 update 14: two steps back, one step... »


Feed You can follow this conversation by subscribing to the comment feed for this post.

Paul Klemencic

Good work, and very interesting. It confirms what I see as well. The final extent and shape of the ice pack depends on where the ice is located. The ice pack up against the Canadian Archipelago and the north coast of Greenland will be the last to go; the ice pack north of 80N is generally going to last the melt season, and the lower latitude ice pack elsewhere is going to be decimated (lose about two-thirds of the ice) by the minimum.

I think your "model" as it is, could be improved if you could somehow adjust for the rotation of the pack, and the transport through the Fram Strait. The current "model" assumes the ice remains relatively close to its location at the end of July. In reality, the ice will move, and the ice pack will "spread" as the season progresses.

As AGW keeps weakening the pack in future years, eventually the pack at the end of July will be small enough (less than 5.6 million sq km versus 6.35 million sq km at the end of July 07) for much more rapid ice movement, spreading, and transport. When the end of July melt falls below 5.6 million sq km, the melting rate in August and September will be much higher. Then a model based on estimating the health of the pack using ice temperature will have bigger errors than we see in these diagrams.

For example, this year if Franz Josef islands and Svalbard don't buttress the ice pack, the loss of ice into the Fram and through the Olga Strait will increase, and the edge regions of the pack in that direction could slide toward the Fram. The "red" bump in the pack next to Severnaya Zemyla shown in your 2011 chart could disappear, as that ice ends up west of 45E longitude toward the Fram.

It seems that breaking down the pack into sections, and modeling each section of the ice pack could improve forecasts significantly.

Russell McKane

I couldn't agree with you more Paul, thanks for the comments. Ice movement particularly convergence and divergence in the final weeks does make it impossible to be accurate with this method as it is correlating observational material six weeks apart. Rotation can be allowed for in part as it is more predictable. Thus I have been conservative with the shaping of the Beaufort Gyre. I would like to be able to do this regionally but I don't think the resolution of the maps allows this. Ultimately it comes down to conoiseurship - like juging a good wine - experiance and deep understanding of the neuances. This is something I feel from your comments that you have.

A post script for those who have not seen these ice maps before. This puzzled me for the first few years of observing. The colour of the ice that is melting has a temperature of 245 degrees Kalvin - well below melting temperature - infact 18 degrees below - if you go to the website and check out the beginning of the melt times, May, you will see a predominance of yellow in old ice areas and red in first year ice areas - these when melt starts become hot pink (one degree below melting) then suddenly they go the cyan blue. This is due to the heat exchange of melting ice. it takes 18 degrees of heat exchange to change states from solid to liquid and this is registered on the satalite data as cold ice. Thus the areas of drop in colour concentration - mixing with black of background are the more rapidly melting ice and are good for telling regional areas to watch. As the colour returns to red as it will toward in late August it is a sign that the end of the melt is here and any further loss will only be in those areas still faded cyan. The nice thing about this site is you can back track through the data and retrieve any number of maps and even overlay different types of data onto the current image.

Seke Rob

Thanks for the instructions how to read these charts, new to me too. Minor point, but 245K is -28C, or were you talking Fahrenheit... yes you were... -18.7F :D


This is a really nice idea for prediction, I'm really surprised no one has mentioned it before.

I am having a lot of problem with cyan though. Do you know what the satelite is reading that gives the reading of temperature so low?

To me, I don't understand how readings this low can be melt. I am not arguing or saying that it is wrong, my mind just keeps saying infroming me that I do not undersand it. The only thing that makes sense to me is that it is an artifact of the sensing process rather than an actual equiibrium temperature.



Wow, i sure wish it was possible to edit posts, a lot of unclarity and mispellings. Sorry about that but if anyone has an answer, I'd sure like to know.


North West Passage:

It seems like the Amundsen route with small changes is now open, if you take the trip through the Bellot Strait and south of King William Island, instead of the original route.




A very interesting prediction, and well inside the 75 north parallel.


Russell McKane

Thanks for the comments, now for the Cyan problem. Thanks Seke Rob, Firstly I am now doubting my 18 degrees(I was thinking C or K) butyou are right we are ten degrees Kelvin lower than that. This 18 degrees came from my memory from something I read which said arhh that's the answer to why the cyan. Process is right but detail wrong The following snippet is rom the famous Wikipedia.
'When ice melts, it absorbs as much energy as it would take to heat an equivalent mass of water by 80 degrees C. During the melting process, the temperature remains constant at 0 degrees C. While melting , any energy added breaks the hydrogen bonds between the ice(water) molecules. Energy becomes available to increase the thermal energy(temperature) only enough hydrogen bonds are broken that the ice can be considered liquid water.'

Now obviously we don't have a drop of 80 degrees here either - The water stays at 0 degrees c until it has absorbed enough heat (80cal per gram- 1cal is what it takes to heat 1 gram of liquid water 1 degree C)

This web page describes the process well. http://www.school-for-champions.com/science/heat_ice_steam.htm

This is an endothermic process - absorbing enery - now the satellite processes are not putting physical thermometers in water but are measuring energy levels or exchanges from space. Thus I am assuming that they are picking up this drop in energy as it is absorbed during the process of change of state of water and feeding it back as a temperatre reading on a Kelvin scale.

The key thing from these images is that the process of melting shows up as a distinct colour change and allows us to identify clearly areas undergoing the process ofmelting even if they don't actually get to a state of total liquid.
It also helps us identify older ice that is too thick to move to this transition state.
I have asked Neven to include another image from June 29 2011 which shows this process in action. Notice the blue in the Laptev sea area before the heat exhange - in the next few days it turns to cyan. the process also helps to identify polymers(spelling?) developing from melting not just ice transport related ice free areas.

Interestingly the reverse of this heat exchange process is evident in the refreeze of October which has led to anomalous high air temperatures over the arctic as the exothermic exchange pumps heat into the arctic air and delays the refreeze somewhat. This has been commented on a bit over the last few refreeze periods.

I hope this helps Fred as you said - 'the only thing that makes sense to me is that it is an artifact of the sensing process rather than an actual equilibrium temperature.'
You are right it is the sensing process bu picking up real heat exchange - so not sure if it's an artifact or a very useful happenstance.

Chris Biscan

Jaxa and UB do not seem to be on the same page.

Regardless the weather is definitely helping the ice and that is a good good thing.

I have asked Neven to include another image from June 29 2011 which shows this process in action. Notice the blue in the Laptev sea area before the heat exhange - in the next few days it turns to cyan.

I've added this image to the article.

Rob Dekker

Russell, this is a great post, but allow me to clarify one thing :

Even though I'm not an expert on AMSR, I know wnough about it to state that with AMSR, talking about the ice 'temperature' is very misleading.

Microwave emissions from any surface are extremely dependent on frequency where you measure, the polarization you measure, and a spectrum of factors having to do with the structure of the surface you are looking at.

For example, the crystaline structure of ice and snow emits far more microwave radiation than liquid water.

This is why AMSR is so good in finding the boundary between ice and water and because clouds are transparant for most microwave bands, AMSR can make the great images of ice extent and concentration that we are looking at every day (NSIDC, UBREMEN, JAXA).

Any way, I may be on a limb here, but I think that the 'temperature' as calculated on the images you show do reflect the real temperature that the surface has ASSUMING that it is solid ice and snow.

So what you are actually looking at is not the 'heat absorption' during the melting process, but in fact the 'water-content' or 'water-concentration' of the surface.

So all the 'blueish' color (and darker) are areas where there is more water in the surface than pure ice.

Still, it is very interesting that there seems to be a reasonable correlation between the 30% ('water content' I guess) of the ice in July with the final ice extent. Very interesting stuff !

Russell McKane

Thanks Rob,
I realise that it is an algorithm applied to a spectral data set. In posting I was hoping some collective brains and knowledge will tease out what I have been observing. The Ti2 data is for Ice temperature and I understand the issue of water and ice mixing particularly with melt ponds. When you compare the daily extent maps with the Ti2 map there is a distinct gap between the Ti2 and the position of the ice edge which I assume is because of the presence of large amounts of water - sea and melt- the reverse occurs in their Ocean temperature (To) where ocean temperature does not show near the ice edge. It's actually this gap which alerted me to a time delay factor between Ti2 edge and actual edge. In May and June this gap is about ten days so it really helps to work out what areas will be melting next and how much.
As to Cyan as it definately can't be measuring a temperature of 245K in this region - all further sugestions - knowledge welcome.

Rob Dekker

Russell, I think the Cyan (I called it 'blueish') simply suggest there is some water content on the surface. Most likely the cyan in the center comes from wet snow.
So 'cyan' is the color of wet snow.
Once freezing starts happening you will see the pack turn pink/purple.

Rob Dekker

I did not have much time to find a really good ref to what AMSR images really show, but this NSIDC link may be helpfull :


Important thing to note is that the "ice temperature" plot you show relates only to real (sensible) ice temperature if we are looking at ice. So during winter time, the image will actually reveal the temperature of the ice (and cyan would suggest that it is 30 C below freezing). During the melting season however, temperature of the ice is close to -1.5 C (melting temp of sea water ice), and the AMSR "ice temperature" thus reveals the "water content" of the surface.

You used amrs.C 0.3, which suggests that the image it cut-off (black) when the water content exceeds 30%. Since snow cannot hold much more that 30% water until it turns into a melting pond, that suggests that everything black on your images has significant melting ponds and/or leads.

So I thinl what your observation shows is that the areas that contain significant melting ponds and/or leads are likely to melt out completely before the summer ice minimum.

And by itself that is a very interesting observation.

Chris Reynolds

Russell McKane,

I've done a blog post on your technique here, in short (sorry) but I'm not convinced:
It needed to be a seperate post because of the number of images involved. However I'm happy to discuss here.

Re the AMSR.n.T2 temperature. Are you absolutely sure this is ice temperature?

The energy required for phase change from ice to water would not appear in AMSR as AMSR measures brightness in the microwave bands to estimate temperature. Phase change from ice to water would merely peg temperature at zero until the ice is melted, or the cold season starts. So all AMSR would register would be a warming from below zero, and a leveling at around zero.

The T2 channel shows temperature at -28degC in the light blue. There is no way the Arctic sea ice is that cold in the summer. Furthermore with regards the May image you post, note that around the periphery of the sea-ice the temperature falls to light blue (-28degC), whereas the pink/red area in the centre of the pack is just below 0degC.

I'd have some sympathy with Rob Dekker's suggestion that we're seeing water content. Perhaps the blue is where the pack is above a threshold of wetness, the surrounding black is wetter?

AMSR isn't something I've read many papers on, and I won't have the time to investigate further for some time.

Peter Ellis


"AMSR.n .Ti" is defined as "Ice temperature calculated by IOMASA optimal estimation algorithm".

From this, one might guess that "AMSR.n .Ti2" is a second ice temperature measurement by a different algorithm. If so, it really isn't very accurate, for the reasons Chris gives.

Here's a PDF giving some of the considerations involved in deriving useful information from the raw microwave data. If you can make head or tail of it, you're a much better man than I!


Marc McNaught

Niven....thanks for this wonderful blog page. I check it four or five times a day.

I see the latest ECMWF/GFS model develops a high over Greenland in the first week of August and pushes ridges to the north and east. Thickness levels also warm as a result. What effect do you see this pressure field having on this years melt. The AO also looks to be trending negative.


Hi, Marc. Thanks for the kind words. I have just published SIE update 14, which contains my view on current developments.

Chris Reynolds

Peter Ellis,

From that paper you linked to I think the key statement is:
"Ice emissivity is a function of many factors – surface roughness, brine…"
They truncated it at brine.

Without complicated analysis using other data it seems the implied temperatures are non-sensical. I suspect that AMSR.n.T2 is the raw 6GHz brightness. Arcticio just informed me that it's 6GHz. That's basically using a radio receiver at 6GHz (6,000MHz) and measuring the emission from the surface and intervening atmosphere - damn sensitive kit.

As for whether the technique here is of value I don't think this side-issue has a direct bearing. If T2 is a good predictor then it is, and the details of how can be worked out afterwards. I'll leave it to others to make up their minds about the usefulness of this prediction method.

Russell McKane

Chris R
Thanks for your comments and useful side blog. Food for constructive thought. Not had the time I would like yet to process and respond. Ultimately if it is a good predictor it is. That's why I put it out in the public sphere, to hold my thinking accountable.
In a real sense what I am suggesting is a qualitative approach in contrast to a quantitative one. It may be able to be reduced to a qualitative method by working out edges and area then perhaps averaging 7 days of the last week in July and taking into account that the end of melt also varies by a week or two but I certainly don't have the tools to do this.

Another thought - If at the start of the melt the difference between actual edge and the Ti2 edge is around 10 days and as we progress to the minimum this difference grows to around six weeks (42 days) and then beyond (in days) but it never gets to this as the winter refreeze calls a halt to the process and returns it to being close to the same again. Then at some time the final minimum edge will be passed through. I think Chris's series in his bog shows this critical transition. So what I am postulating is that the pattern shows us a predictive time frame about the last week in July for a minimum at around the end of the first week of September (around six weeks) it is a shape pattern and an extent pattern that will be affected by rotation and transport and to a bigger extent a spread ice pack of MYI as is common in the Beaufort. But that it can still give us a reasonable near term prediction of the minimum. If this is statistically supported or supportable is another matter. Will we have enough melt seasons in this time frame of like conditions for a few years of validation is a mute point for all of us though.


Hello Russell,

Sorry but I don't get this paragraph:
"Another thought - If at the start of the melt the difference between actual edge and the Ti2 edge is... ....at some time the final minimum edge will be passed through. "

With regards the T2 as a predictor. As I've said upon seeing it I was initially persuaded. My main line of reasoning was that T2 was somehow showing us the thicker, probably multi-year (MY} ice, that which is harder to melt. As I mentioned, I used to follow Quikscat when it covered Arctic sea-ice, and found that the ice retreated back to the MY ice. Quikscat used to be obscured by June, so it was basically an effort to track where the MY ice had moved and get a guesstimate of where the minimum area would be. I have to say not always an accurate one, weather always plays a major role.

I've looked at my reply and still stand by what I said about the dangers of spotting correlations where there aren't any because of a restricted dataset (restricted in the frequency domain, or in terms of general shape). That's not intended as any kind of personal criticism, the Wunsch study I referred to was criticising a large range of papers by professional scientists.

Chris R

Rob Dekker

Chris, I appreciate your reservations of the correllation of Russell's observations. However, since the plots he shows seem to relate to the 'wet-ness' of top surface, there may very well be a physical explanation for the correlation he observed.

As I mentioned before, the AMSR plots Russell presents show that the areas that contain significant melting ponds and/or leads (at the end of July) are likely to melt out completely before the summer ice minimum (in September).

Russell McKane

Chris R
You almost have me convinced Chris , but as I have been looking at this over a three year period in the early melt time right through to what i presented here I also want to test what you are saying against these other aspect. It's related to the paragraph you didn't understand - a lot in my head there not yet made explicit. I am going to have to wait till the weekend before I can work on a conscidered reply. And I take nothing personally in this - infact I am enjoying the stimulation and challenge. More coming and please challenge it. Russell

Chris Reynolds

Rob Dekker,

It's that aspect that interests me most. I suspect we're seeing thicker ice that's wet on the surface, or has been wet and refrozen. That there is so much day-to-day variation suggests a rapid function such as ice wetness may have a role: Perhaps black is wet, blue is re-frozen?


I'll pop back here and check later in the weekend, sorry for not geting back but I've been busy reading papers in preparation for a series of new posts - it takes a lot of time.

Rob Dekker

Chris Perhaps black is wet, blue is re-frozen?

I doubt that blue is 'refrozen'. For starters, in July, the temperature at the ice surface is a few degrees above freezing, and not the 245 K that 'blue' suggests. Besides, the radiative balance in July across much of the Arctic prehibits freezing to occur. If the sun does not set, it's hard to 'freeze' anything.

By now, I'm convinced that 'blue' indicates the wet snow that covers much of the Arctic sea ice in summer, and black simply suggests that it's wetter-than-wet, or IOW, leads and melting ponds dampen microwave emissions over solid ice/snow.


Does this


The tropospheric temperatures show no trend, while the channel 3 data shows an upward trend. In the Arctic, the channel 3 data now also become sensitive to sea ice, which causes a warm signature compared to the cold ocean background. If you don’t believe ice can cause a “warm” signal, check out today’s NOAA-15 channel 3 imagery, which shows warm signatures in the Arctic ocean and in the sea ice area surrounding Antarctica:

This is because microwave radiometers measure brightness temperature, which is a product of the temperature of an object and its microwave emissivity (1 minus its reflectivity).

To make things even more difficult in the interpretation of ch. 3, more snow cover on sea ice will cause an anomalously cold signal, rather than warm signal, just as it does over land — unless the snow is melting, in which case it switches to an anomalously warm signal. The point is that it is difficult to interpret what the upward trend in the Arctic channel 3 data is due to. Stratifying the data by season would probably give some insight.

help at all?

Russell McKane

Chris , Rob Dekker,
A weekend has come and gone but no time to work on this. Thanks for reference crandles - it shows the problem but not the interpretation for polarview site- we don't see switching signal between warm and cold just to the anomolously cold- I'm not sure if it is the same instramentation -

This is a known issue with passive microwave measurements, and is the basis for snow cover, snow depth, and snow water equivalent retrievals which are used by NASA and NOAA, primarily from the AMSR-E instrument

But looks like you are on the right track Rob.

Chris Reynolds

No problem Russell,

When you do post here just drop me a quick line on my blog in case I forget to pop back here. I'm happy to debate this in a leasurely manner.


It's a different instrument (unless I'm mistaken), but could be a similar sort of process. Thanks.

Lauren Jonczak

Great post Russell. I have been looking up the different kinds of temperature measurement devices for a paper that I am writing. Thanks so much for sharing this great information, it is very helpful!

Verify your Comment

Previewing your Comment

This is only a preview. Your comment has not yet been posted.

Your comment could not be posted. Error type:
Your comment has been posted. Post another comment

The letters and numbers you entered did not match the image. Please try again.

As a final step before posting your comment, enter the letters and numbers you see in the image below. This prevents automated programs from posting comments.

Having trouble reading this image? View an alternate.


Post a comment