This is a blog post in progress. I'm calling on all you commenters to think along, advise and correct wherever necessary.
I vaguely remembered a short discussion about discrepancies between various area and extent datasets in the comment section of SIE update 21 (perhaps somewhere else too). One of the possible causes mentioned was what I have just started referring to as the North Hole, also called the 'pole hole' by NSIDC:
Note that unlike ice extent, the Arctic values for ice area do not
include the area near the pole not imaged by the sensor, called the pole
hole. This area is (...) 0.31 million square
kilometers for SSM/I (from July 1987 to present).
IJIS also makes mention of this on their extent page:
The black dot seen at the North Pole is an area lacking data where AMSR-E cannot observe the Earth’s surface due to the limit of its observational coverage (i.e., orbit inclination of 98deg. and swath width of 1600km). Please note that this area is also counted as sea-ice cover in our estimation of sea-ice extent. We may change the policy (i.e., filling the gap with full coverage of sea ice) in the near future due to the recent drastic reduction of Arctic sea ice. We will announce this if it is implemented.
The black dot is also visible on the sea ice concentration map from Cryosphere Today at the top of the post. In previous years the ice there was practically 100% throughout the summer melt, but this year is different. Large swathes of low concentration ice have recently started to appear under the North Hole, like I noticed when commenting on the latest Race to Fram Strait animation. This could be important, because it means that for instance NSIDC area and IJIS extent are actually lower than reported (I haven't found any info yet on how other agencies, such as DMI, Arctic Roos or Uni Bremen handle the North Hole). But how much lower? That's what the 'in progress' is all about (check out the update below). Let me explain what I did so far.
One of the lessons from Patrick Lockerby's MODIS Rapidfire for Citizen Scientists-series I remembered, is that the North Pole can easily be found on the MODIS Arctic Mosaic because it's in the middle of the four quadrants on row 3 and 4, column 3 and 4.
So I decided to download these 4 quadrants from day 247 (September 4th) and combine them in one large image. So far, so good. But how to determine how big the North Hole is in that image? If I really wanted to get to the bottom of this, it was inevitable I did some calculating. Oh my, not my speciality. Bear with me, please.
As we've seen, the North Hole covers 310,000 square km according to NSIDC. If I've understood correctly each pixel in the MODIS images I downloaded is 1 square km. In other words, the total area of the North Hole circle is 310,000 pixels. The formula for the area enclosed by a circle is A = π r², so 310,000 = π r² (r being the radius of the circle). Moving π to the other side makes 310,000/π = 100,000 = r² (that's not entirely correct, but I'm not a scientist, so I can round off any way I like). The square root of 100,000 is 316, so the radius is 316. Multiplied by 2 I now have the diameter of my circle, which enables me to set the Fixed Size of my Elliptical Marquee Tool in Photoshop to a width and height of 632 pixels.
Here's what the result looks like:
That's from this combined image of the 4 quadrants:
Conclusion: I'm seeing quite a bit of dark water under the North Hole. Assuming I did everything right (maybe the circle is smaller or bigger). This means that IJIS extent is probably lower than reported. To determine exactly how much I'd probably need to do something with the colours or use a filter or something, and then count pixels. These are things I currently know nothing about.
To be continued...
UPDATE:
I've gridded the image and I don't think there are any grid cells that have ice concentrations below 15%. This means that NSIDC extent isn't overestimated. Area is another story, but whether it's relevant remains to be seen.
Here's the gridded image (25x25 pixel/square km grid cells):
And here's a grid cell with relatively a lot of blue in it:
UPDATE 2: Here's the combined image for day 250 (Sept. 7th):
Great job as always, Neven. I assume the mismatches are due to time differences in when the photos were taken. Any idea how much time elapsed between the first and last photos?.
Posted by: Steve Bloom | September 06, 2010 at 00:36
Thanks, Steve.
I wouldn't know the answer to your question. I just took what looked to me the clearest recent images of the North Hole Area. These turned out to be yesterday's Terra images.
Posted by: Neven | September 06, 2010 at 00:39
Counting pixels? Send for Goddard...
Steve: I'd guess it's a day's worth of orbits, judging only from the cloud displacements.
Posted by: Gareth | September 06, 2010 at 00:41
Great job Nevin, assuming 100% coverage was reasonable until now, (or maybe last year) I had thought of it as an upcomming issue, but you have put much more thought into it.
Tony
Posted by: Anthonywobrien | September 06, 2010 at 00:42
I hate to suggest more work for you, Neven, but not nearly enough to not suggest it. :) Anyway, it seems like it would be useful to go through the same exercise for 2007-8-9 since one could get a pretty good idea of the differences just by eyeballing.
Posted by: Steve Bloom | September 06, 2010 at 01:08
Neven, that radius is closer to 314 km so your math is good, considering the 'north hole' is unlikely to be exactly 310,000 km^2 as announced. Photoshop's Histogram function should create a reasonable estimate of the area covered with dark pixels.
I would recommend redoing this with the 'terra' Bands 3-6-7 images so the Cloud covered area can be removed from your estimate. Covert this to a 3 color image: White is cloud, Red is Sea Ice, and Black is open water. Ideally, you'll end up with a percentage of visible ocean covered by open water, then apply this estimate to the whole hole ;^)
For those interested, all of the date/time and geo-location data is published on MODIS under the 'Display metadata' link.
Posted by: Artful Dodger | September 06, 2010 at 01:08
Steve, you can have a look at my End Zone 5: sea ice concentration blog post, where there's an animation of Uni Bremen sea ice concentration maps of the last 3 years in this phase of the melting season. I hardly see any change from 100% sea ice concentration anywhere near the North Hole, except perhaps a day or two in 2008.
Lodger, thanks, it's good to know I didn't screw up my calculation too much. I've already made a 3-6-7 version and tried the Histogram thingy on both versions, but didn't think it was working. Perhaps converting to a 3 color image might help, but I fear the image is still too unclear (clouds over some pretty large areas of water).
I'm sending you the images.
Posted by: Neven | September 06, 2010 at 01:16
Hi Anthony. Welcome to the Forum! Are you with Energy Australia? Either way, we welcome your contributions.
Steve: MODIS images are available only back until March 18, 2009, but clearly a comparison with'09 would be interesting. It'd might be even more interesting to do a comparison of the Dates of minimum 'Compactness', the "CT Area per IJIS Extent" (C.A.P.I.E.) ratio.
Neven: further reading of About Arctic MODIS says this:
"Projection
The images are mapped using the Polar Stereographic projection with origin at 0 longitude and +90 latitude with a +70 standard parallel (also known as latitude of true scale). This corresponds to a scale at natural origin of 0.969858730377". So rather than using a 314 pixel radius for your circle, a better number would be 324 pixels.
Try this: before using the histogram function, reduce the color palette to custom no. of colors, set to 3. This will force PhotoShop to choose the closest match for the pixel. Out over the pole, we know it's one of: Ice, Cloud, or Water. Then, the Histogram function will be like a bar-graph, and the portion of red to black will be the Ice / Water ration, and the proportion of white will be the Cloud area to remove from the original 310,000 km^2 area of the polar hole.
Especially Neven: Thanks for your tireless effort, this is truly new and interesting. Well Done!
Posted by: Artful Dodger | September 06, 2010 at 01:58
Yes, it's probably fair to assume that the hole wouldn't include much if any open area if there's nothing going on in the immediate surroundings.
Regarding estimating from those photos, I suspect someone at NSIDC would be willing to hazard an expert guess.
Posted by: Steve Bloom | September 06, 2010 at 02:03
Especially Neven: Thanks for your tireless effort, this is truly new and interesting. Well Done!
Your Fram Strait images made me notice, so a big thanks to you too. Will try your color palette reduction trick later on. And thanks too for checking my calculation! Looks like I was conservative... :-)
Regarding estimating from those photos, I suspect someone at NSIDC would be willing to hazard an expert guess.
Steve, I've sent out some mails. I'll report if I get interesting answers.
Posted by: Neven | September 06, 2010 at 02:08
Nevin,
When IJIS calculates extent they they use a larger area than 1 km2. I think they use 25km by 25 km. If that area is 15% ice covered it counts as ice covered. I am not used to these photos but I imagine that there is enough ice so that the polar hole would still be 100% ice extent. The NSIDC area on Cryosphere Today would be high due to this problem. They might mention this in the NSIDC update for this month, they had something earlier this summer that was similar: see July 20 NSIDC update.
Posted by: michael sweet | September 06, 2010 at 02:20
Michael, thanks a lot for that link to the NSIDC update.
I am not used to these photos but I imagine that there is enough ice so that the polar hole would still be 100% ice extent.
You're right. I've gridded the image and I can't say I'm seeing any grid cells that contain less than 15% ice. I've updated the post with images.
Posted by: Neven | September 06, 2010 at 02:24
Neven, I've done the analysis of the MODIS Bands 3-6-7 image you sent. First, about 55% of the circle is Cloud covered, and about 45% shows the surface of the Ocean. For the Cloud free Area, here are the results:
Using these proportions to estimate the area of Open Water and Sea Ice for the entire N. Pole data hole, we get these results:
Posted by: Artful Dodger | September 06, 2010 at 03:38
Lodger, that's greatly appreciated as I couldn't get it done in Photoshop myself (I use it for the most basic stuff like cropping images).
So, an estimated 87K less area in the North Hole. Interesting.
Posted by: Neven | September 06, 2010 at 03:44
Yeah, I did it easily in IrfanView (free for Windows Users) since I don't own PhotoShop. I'm going to look at the results again tomorrow, as I think there can be some improvement in the Cloud boundary / shadows.
Posted by: Artful Dodger | September 06, 2010 at 03:55
In the meantime I have received some info on how they handle the North Hole over at Cryosphere Today:
In our area timeseries, we fill in the hole with an average value. The value is obtained by taking the average concentration value over all longitudes for the degree latitude circle surrounding the hole. The size of the hole changes with each sensor/platform, but if the hole is 87 deg. N to the pole, for example, we average all available concentrations in the circle 86 deg. N to 87 deg. N and fill in all points in the hole with that value.
Posted by: Neven | September 06, 2010 at 04:05
Michael: There was a discussion on this blog last week where we determined that IJIS uses a 6.25 km x 6.25 km grid for its Extent calculations. IIRC, NSIDC uses a 25x25 km grid.
Posted by: Artful Dodger | September 06, 2010 at 07:29
I read that discussion, but apparently it went out again at the back of my head.
If IJIS uses 6.25 x 6.25 it might be interesting to do another gridded image with higher (250m) resolution MODIS images. Perhaps if we're lucky, we'll get a clearer picture of the North Hole area in the coming days.
Posted by: Neven | September 06, 2010 at 07:39
Feel free Neven, but for myself I'm primarily interested in the Area results, as Extent has become deprecated by the capabilities of the current generation of space-based instruments.
This has been an interesting exercise for me, and I've revised my analysis method to allow for shadows cast by Clouds. I'm now converting to 5 colors in the Band 3-6-7 image: Cloud, Ice in Sun, Ice in Shadow, Water in Sun, and Water in Shadow.
Ice in Sun or Shadow is added together, as is Water in Sun or Shadow. Cloud is handled the same way. It's just 2 more steps to handle this in a photo-editor, so I'll revise my Area estimate as my time allows. The revision should be small but significant.
Posted by: Artful Dodger | September 06, 2010 at 08:48
Excellent that CT do "the right thing" already for their area numbers.
Posted by: Nick Barnes | September 06, 2010 at 09:34
Regrettably, my only contribution to this is a little hair splitting - Artful Dodger said: MODIS images are available only back until March 18, 2009", which is perhaps practically true, but is not literally true.
They have a patchy collection of images back to 2001 here: http://rapidfire.sci.gsfc.nasa.gov/realtime/?calendar
(unfortunately, no 2007 iamges after 24 Aug)
The problem is they are not the neat mosaics, but the original images which then get processed into the final mosaic - they are manipulated into the Polar projection and stitched together. Now, my own picture editing skills are nada (within 95% confidence limits), but Lodger at least seems a bit of a ninja, so what are the chances of being able to construct reasonable approximations for previous years?
Either way, two thumbs up, guys - another great post, Neven. The intrigue just never ends
Posted by: FrankD | September 06, 2010 at 13:07
These images are run through proprietary ortho-correction programs to correct for the position of the Satellite and the oblate spheroid which is our world -- well beyond what Photoshop can do.
I wonder if our friend Patrick Lockerby might run across some current images that capture the N. Pole itself in the rapidfire stream (flood?). You're a hero if you do, Logicman!
Posted by: Artful Dodger | September 06, 2010 at 14:11
Artful Dodger,
Thanks, I missed the earlier discussion. I had only read the information from NSIDC.
I am not surprised that Cryosphere Today is already correcting for this effect. That is their job, and they do a good job of it.
Keep up the good work Nevin. I lurk a lot here but don't generally post.
Posted by: michael sweet | September 06, 2010 at 17:27
Frank, I think it would be very difficult to determine the exact location of the Pole, whereas the mosaic images make that real easy.
Lodger, it's great what you're doing. Hopefully we get some clear images from the North Hole region in the next few days. A high could shift over there, chasing the clouds away.
Michael, thanks for the lurking and the commenting!
Posted by: Neven | September 06, 2010 at 20:54
Looking at the Bremen concentration chart today the open areas near the pole have become more extensive.
Posted by: me.yahoo.com/a/nSjChi4X3vr8X3DRw93GkY1.cerja.8nvWk- | September 07, 2010 at 06:00
Today's images from the AQUA satellite might give a clearer picture of the North Hole.
Posted by: Neven | September 07, 2010 at 13:06
Neven, your 'Aqua' link went to the PIPS forecast instead.
Posted by: Kevin McKinney | September 07, 2010 at 14:21
Fixed. Thanks, Kevin!
Posted by: Neven | September 07, 2010 at 14:23
Way to squib the challenge, guys ;-)
"Frank, I think it would be very difficult to determine the exact location of the Pole". Actually, Neven, its easy. The individual images have lat and long marked and a little blue dot for the location of the pole. But seriously, I'm not surprised trying to wrap these onto a spheroid and do a projection is out of scope, but its been a long time (and several versions) since I played with photoshop. Do you think NASA might composite a few specially if they were asked nicely?
But really, I think eyeballing the archived data will probably be as good. My gut says that few (if any) of the archived images will show anything like this hole. You don't need to projections and grids and pixel counts to say that its near 100%. And again, its easy to pick up the general area of the pole, so I looked at some. Of course, to keep it reasonable, I had to assume that anything like this is going to happen near minimum - that's not necessarily valid, but I don't have the bandwidth (let alone the patience) to check through 1000+ images per year. I've only checked a couple of years so far, but I'll post again tomorrow when I've been through the rest of the archive.
I've looked for a few days either side of the 2008 minimum and there's enough clear skies to say for sure that while there are some decent size leads, there are no patches of open water like this. Same for the last week of available images for 2007 (late August). There's more cloud around, but look at enough images and see nothing but white underneath and I think confidence is reasonable. In fact, the cap that wasn't melted to **** in 2007 looks to have been in pretty good nick. Not like this year.
Todays (Day 250) mosaic shows some BIG patches of open water at ~180 degrees. Even on the 25 km grid above, I think there'd be some squares below 15%.
Posted by: FrankD | September 07, 2010 at 15:05
2006 has some heavy cracking on the Russian side, leads a little above 85 degrees, and some loose stuff below that.
I'll keep looking further back, but I don't expect to find anything.
Posted by: FrankD | September 07, 2010 at 15:36
Frank, your effort is much appreciated, as I thoroughly dislike going through that MODIS real-time database. :-)
Meanwhile, 2 out of 4 quadrants that have appeared so far, look a bit foggy, but not cloudy. If the other 2 look just as good - and they don't necessarily have to because the ice concentration is pretty high on that side - we can make a more accurate and actual assessment of the ice area under the North Hole.
And with 'we' I mean Lodger. :-)
Posted by: Neven | September 07, 2010 at 16:07
I'm a novice ice watcher, but can we assume that the cloud cover is evenly distributed over the ice and the open water?
In other words, is the Arctic overwhelmingly dominated by weather patterns on the scale of mid-latitude cyclonic activity, or are the clouds generated by smaller disturbances on the scale of coastal fog or orographic precipitation?
Posted by: Xemasab | September 07, 2010 at 21:31
FrankD: Exciting stuff! Post some links ;^) Also, remember to look for rounded ice flows, as this is a proxy for the amount of time since breakup.
Xemasab: You don't need to assume anything, you can SEE the clouds. Just look at the 'terra' Bands 3-6-7 images to make sea ice stand out from the clouds. The rest of your post indicates that you are a WX expert, so you tell us!
Neven: Could you post another Band 3-6-7 N. Pole img when 'terra' fills in the last quadrant? I can run it through a colour reduction / histogram analysis in a few minutes...
Posted by: Artful Dodger | September 07, 2010 at 23:38
What is the current consensus on this ??
Posted by: AbbottisGone | August 06, 2015 at 23:28