Jakobshavn Isbræ / Ilulissat Isfjord / West Greenland

[A-Team]

pick a name from Inuit

What about Quuasuuaqtakqassermiauuaqtnunatakquaqtakqu? Means rocky headland between two ice streams, fairly easy to pronounce (rhymes with Nioghalvfjerdsfjordenbrae).

The 2 m DEM of I Howat, produced from stereopairs of 0.5 m WorldView, is an interesting object if you like large files in 32 bit 'tiled tif' format (which can be imported into ImageJ using its 'Bio Formats' plugin). Below are various products made from crops of the calving front and the island between the branches.

Note the calving front has quite a bit of height variation in transects perpendicular to flow lines. The component stream in the depression along the north bank of the south channel is still quite prominent in 2015 Landsats.

The crevasse pattern coming off the island region is probably more indicative of surface ice gradient than of underlying bedrock. Movement here is very very slow compared to adjacent regions to the east that have become unbuttressed for more of the season since main calving front retreat reached its minimum in mid-August rather than late September as in 2013-14.

It's not so easy to determine whether Jakobshavn's acceleration continued in 2015 as it would be hard to find matching times for those three years in cloud-free Landsat-8.

The first image needs as click as it had to be quite wide to retain Howat's original resolution. The second image comes from Google Earth and has had a shadowing convolution applied from the northwest. The false colorings are a sampler from ImageJ lookup LUT tables applied to the 16-bit DEM. Some are more instructive than others.

[AbruptSLR]

I think that the following open access reference hasn't yet been cited in this thread:

Bondzio, J. H., Seroussi, H., Morlighem, M., Kleiner, T., Rückamp, M., Humbert, A., and Larour, E. (2015), "Modelling the dynamic response of Jakobshavn Isbræ, West Greenland, to calving rate perturbations", The Cryosphere Discuss., 9, 5485-5520, doi:10.5194/tcd-9-5485-2015.

http://www.the-cryosphere-discuss.net/9/5485/2015/tcd-9-5485-2015.pdf

Abstract: "Calving is a major means of ice discharge of the Antarctic and Greenland Ice Sheets. The breaking off of icebergs changes the ice front configuration of marine terminating glaciers, which affects the stress regime of their upstream areas. Recent observations show the close correlation between the ice front position and the behaviour of many outlet glaciers. However, modelling of a glacier subject to calving poses various challenges. No universal calving rate parametrisation is known, and tracking of a moving ice front and the related boundary conditions in two or three spatial dimensions is non-trivial. Here, we present the theoretical and technical framework for a Level-Set Method, an implicit boundary tracking scheme, which we implemented into the Ice Sheet System Model (ISSM). The scheme allows us to study the dynamic response of a drainage basin to user-defined front ablation rates. We apply the method in a suite of experiments to Jakobshavn Isbræ, a major marine terminating outlet glacier of the western Greenland Ice Sheet. The model robustly reproduces the high sensitivity of the glacier to frontal ablation in form of calving. We find that enhanced calving is able to trigger significant acceleration of the ice stream. Upstream acceleration is sustained through a combination of various feedback mechanisms. However, lateral stress and ice influx into the trough are able to stabilise the ice stream. This study contributes to the present discussion on causes and effects of the continued changes occurring at Jakobshavn Isbræ, and emphasises that the incorporation of seasonal calving and dynamic lateral effects is key for realistic model projections of future global sea level rise on centennial time scales."

[nukefix]

I think that the following open access reference hasn't yet been cited in this thread:

Bondzio, J. H., Seroussi, H., Morlighem, M., Kleiner, T., Rückamp, M., Humbert, A., and Larour, E. (2015), "Modelling the dynamic response of Jakobshavn Isbræ, West Greenland, to calving rate perturbations", The Cryosphere Discuss., 9, 5485-5520, doi:10.5194/tcd-9-5485-2015.

Figure 4 is highly interesting:

[nukefix]

According to the Figure 4 the "island" with the long name is actually ice grounded below sea-level...so why is it not moving faster? No back-pressure?

[A-Team]

Figure 4 the "island" with the long name is actually ice grounded

Indeed it seems impossible that 'island ice' would not have moved since the 1985 photograph above. It's totally stagnant, right down to the fine details -- it's the best ground control point around, see below. The animation shows the amazing stability of the 'island' as the ice flows around it everywhere to the east and north. The frames here are seven path 10, row 11's for 2015 from day 50 to day 226. I'll add the 2014'w with this viewing geometry in a bit.

They don't make clear where Fig 4 DEM data comes from. Fig 1 velocities are from 2008/2009. The calving front positions in Fig 2 don't give the month. In model development papers they just need to lay out any old heuristic assumptions; accurate facts don't matter so much because they're seeking the most general case (even though there JI is unique), in fact fact acquisition can be downright inconvenient so ok to defer that to a later model run down the road.

this open access discussion paper hasn't yet been commented on in this thread

Nice spotting, AbruptSLR. It is primarily a modeling exercise of an abstracted glacier that has prospects to be useful if only the physical and geometrical properties of Jakobshavn were better known. This is not a case of 'garbage in, garbage out' by any means but the authors themselves note:

Bathymetry of the ice-choked fjord of Jakobshavn Isbræ is difficult to measure and currently poorly known. As a first order estimate, we apply a parabolic profile of 800 m depth along the ice fjord, fitted via spline interpolation to known topography data.

But that is only the start of the intractable problems with Jakobshavn that go back for decades. There are huge issues with meltwater discharge and calving front cavities, hydrofracturing of crevasses, basal meltwater surges overwhelming meltwater conveyance, thick and possibly deformable till facies, the unobserved temperature with depth profile (eg ice rheology) and maybe temperate ice or even liquid water in some areas yet seismic crystal orientation fabric inconsistent with this. The data we have shows 13 cycles of prograde and retrograde bed -- yet that is invariably modeled for simplicity as straight retrograde.

It's one thing to say "enhanced calving is able to trigger significant acceleration of the ice stream... however, lateral stress and ice influx into the trough are able to stabilise the ice stream" and another to find quantitative support in Landsat or Sentinel for the massive extended event of the last two months (which obviously the paper had no opportunity to consider).

However it's not clear the model has the capability to have predicted this event and the glacier's reaction to it, nor to illuminate changed terminus behavior if that is what this event represents. That is, according to the model, what should we expect to see from analyzing satellite imagery: how much acceleration for how long how far upglacier?

Note however the fifth section of the paper has a quite insightful discussion of the issues; it is quite feasible to skip over the esoteric method sections and read this for qualitative insights.

I'm coming around to the perspective of the drunk looking for his car keys under the street lamp: there's a lot more experimental illumination at Russell Glacier to the south and Store to the north. Jakobshavn may be more important for sea level rise but the observational window really closed 30 years ago (before the big acceleration). It's just too deep, too crevassed, too fast moving and too inaccessible to radar and seismic to get the physically descriptive parameters that would allow on-topic modeling. Meanwhile Safire is taking cores to bedrock at Store, not even a pipe dream at JI any more. (The last attempt at hot water drilling took place in 1989.)

Jakobshavn Glacier, west Greenland: 30 years of spaceborne observations [a 17-year old paper discussing research at Jakobshavn 1962-1992 that could have been written yesterday]
HG Sohn et al DOI: 10.1029/98GL01973

Early 1960's reconnaissance satellite images are compared to more recent image and map data in an interannual and seasonal study of West Central Greenland margin fluctuations. From 1962 to 1992, ice sheet margins to the north and south of Jakobshavn Glacier retreated despite a decline in average summer temperatures. The retreat may be reversing along the southern flank of the ice stream where regional mass balance estimates are positive. From 1950 to 1996, the terminus of Jakobshavn Glacier seasonally fluctuated ∼2.5 km around its annual mean position. The total calving flux during the summer is more than six times that during winter. We identified that summer melting and the break-up of sea ice and icebergs in the fjord are important in controlling the rate of iceberg production. If correct, calving rates may be expected to increase should climate become warmer in the near future.

[nukefix]

Looks like there must be a rock mound or something that is providing enough pressure to stop the whole thing toppling over.

[sidd]

Re: Bondzio et al.

On a first reading, this leapt out at me:

"The modelled glacier response due to enhanced calving is in good agreement with
observations. However, the reversibility of the ice front position is in stark contrast to
Jakobshavn Isbræ’s ongoing observed retreat. Therefore, the actual calving rates must
have stayed increased, as our results suggest that the glacier would have readvanced
otherwise."

The calving rate increased, and remained high, but the model suggested otherwise. I shall have to read carefully as time permits, there are many places the model could have gone awry, and the authors do point to shortcomings.

[Espen]

The shade formed by the Sentinel image correspond to rock islands elsewhere, so my bet it is an island

[A-Team]

I wish these clowns would actually write about the real Jakobshavn Isbrae sometime...

[A-Team]

The first image below shows some new ground control points to the immediate SW of the calving front. These are locatable on every Landsat-8 from 2013-2015 regardless of snow, ablation and illumination conditions. These are quite difficult to find but there does not seem any way to automate finding consistently fixed features over many images over the whole 260x260 km image.

It turns out even for the identical path,rows of Landsat, images of the same date are not necessarily of the same pixel dimensions, nor are they directly superimposable (co-located) relative to ground points. This necessitates some degree of manual cropping and rectilinear shifting. They do not need dimensional rescaling but if path,rows differ they will need a rotation in addition to shifting. In theory there would be some way of doing all this automatically off the metadata.

The animation below has 3 dates from 2013, 7 from 2014 and 7 from 2017. All of these are path 10, row 11. The earliest date is day 50 and the latest day 303. It's quite challenging to bring them all up to 'matching' contrast as the sun angle differs, there may be fresh snowfall on some and thin clouds on others. This results in a certain level of residual flickering.

Below, the global histograms were first normalized, contrast locally equalized, and then gamma adjusted. This won't be optimum for every region of interest because say, if it includes dark rocks of no interest, those still throw off adjustments unless they are masked out. In other words, a standardized global process will not be optimal for a smaller region of interest, for example side flow from the south or the island.

The animation below is not about the calving front, we have done enough on that already. Instead the interest is in regional velocity differences and the partitioning of the overall ice stream to contributing sources. It's apparent that only a fraction of the calving front ice originates in the inland.

Note too that the motion of the ice on the margins is quite complicated -- it won't do to merely specify the magnitude of the velocity (speed) because the velocity vector field does not have a uniform direction, ie has a non-trivial curl and divergence.

You would never know any of this from reading a journal article on Jakobshavn. There the ice flows straight so a single central flowline suffices, despite all the physics in advecting the majority of ice from the sides and getting it up to speed.

Very rarely someone will display velocity by arrow lengths or as unit arrows over a speed coloration but the arrows can't be too dense. It's also feasible to represent compass direction with a color wheel and use brightness or saturation for magnitude. I've only seen a single article with any analysis of vector field properties.

The print legacy curse results in very infrequent use of scientific animations. The pdf of the article will not support raster animations as it is essentially a portable print format. Sometimes a flaky .mov is provided in supplemental but this only makes sense online.

Even web-only journals still use rigid print layouts that don't accommodate high resolution imagery -- Landsat-8 has been out there for 3 full seasons now yet I've never seen a single high resolution still, much less a decent animation in a scientific paper. As far as I know, this forum is about the sole source of multi-frame Landsat or Sentinel glacier animations.

However the work finally appears, constructing animations is absolutely key to developing any insight into the actual motion of ice at Jakobshavn and what features of it are stable or changing over time.

[A-Team]

These animations are fascinating, who knew?

I am posting these by section rather than one big big animation because each sub-area has to be optimized for its special issues such as contrast, cloud-free dates, optimal milliseconds of frame delay and so forth. Maybe Jakobshavn should have its own youTube channel?

The one below centers on a sleepy little region just above Jakobshavn's elbow region. The translucent histograms show the current status of my efforts to improve consistency of contrast across nine different dates, here from day 43 (12 Feb 15)  to day 251 (08 Sep 15) all from path 9, row 11. These come at 16 day intervals but several were too cloudy. Their omission causes slight lurches in velocities.

By selecting a small region of interest (ROI) in an especially clear mid-season image, it is feasible to adjust all the others to match its histogram. (Note the target is still an overall summary that loses positional information but is localized to a ROI which improves on adjusting to the global histogram which might contain dark rocks and cloud shadows that would throw the process off.)

To do this graphically, note the contrast adjustment tool in Gimp etc seems to draw b-splines. Thus adjusting a given histogram to the target's involves neutralizing their difference, bumping up where too dark and bumping down where too bright in proportion to the parameter defining the difference in splines.

In terms of matched pairs of sun angles, we are looking for dates evenly spaced about the summer solstice which for path 9, row 11 amounts to 187±n*16 of which only 123,251 is available for this path,row. These may provide the most favorable case for accurate velocity determination.

171   203
155   219
139   235
123   251
107   267
91   283
75   299
59   315
43   331

[oren]

A-Team you have my vote. Amazing animations.

It's scary how the whole ice is moving, not just the ice stream. I guess it should have been obvious to me. But when the ice stream recedes, it means the rate of ice discharge might accelerate by having the sides of the stream contribute more and more.

[A-Team]

It's scary how the whole ice sheet is moving, not just the ice stream.. ice discharge might accelerate by having the sides of the stream contribute more and more

My thoughts as well. Attached at two 'entrainment' animations that give an overview of where the ice being discharged is originating from. One of these should display immediately, the other has better resolution but needs a click. Full 15 m resolution (not provided) is a large file that overloads most graphic cards as it tries to show 150 ms frames.

Somehow I had gotten the misapprehension that Jakobshavn Isbrae is, as the name suggests, an ice stream -- a distinct feature with sharp shear zones that moves much faster than the enveloping ice sheet creeps along. Maybe it was in years gone by.

Here it looks like very little of the calving front originates in the smooth narrow ice stream but rather that most of it comes from the surrounding ice sheet that has been 'gotten up to speed' some 15-70 km back as it joined the ice stream. It would be challenging though not impossible to make quantitative comparisons to 2014 and 2013 to see if ice sheet entrainment has increased.

Landsat just shows an overhead view of surface ice. It is a bit puzzling to see how this works out as depth because the incoming ice sits over much higher bedrock and so is much thinner than the deep narrow channel, yet this latter has to be filled completely.

The other surprising feature is seen in the southeast corner. It appears that the (smooth white) upper ice sheet above the ablation zone is shedding waves of ice that reach the ice stream in a year or two. The third animation shows a piece of this at full resolution.

[Shared Humanity]

It is awesome watching the ice sheet fracture and crevasse as it moves over what must be an underlying topographical feature and then slump as it completes its journey over this feature and meets the main ice stream. Given the northern wall that is being unzipped is far longer than the main calving face, I would be shocked if it did not eventually contribute far more ice than the main ice stream.

[Shared Humanity]

When you look at the larger animation just posted, you can clearly see the portion of the northern wall that is grounded below sea level. This portion of the sheet is moving far faster than the rest and on the still photo it looks to be at a lower elevation.

On the topographical map below which shows the underlying topography, it is clear that the land mass that is causing the fracturing as the sheet moves over it is the western tip of the large island that sits north of the  main ice stream.

[Espen]

Here is an image from NASA July 14 2001, where I think the now hidden island is located

And below the first image is an image clearly showing the level difference (island), the northern branch to the left and southern in front (date unknown)

[nukefix]

Kick-ass amazing animations!

[Espen]

Yes very nice amimations, would have liked a pause /break in the first animation?

[A-Team]

Seems like that pre-2004 image Espen posted above should be credited to Henrik Højmark Thomsen.
Bevægelsen af is mod fjorden foregår igennem en isstrøm som her ses fra luften.
http://extra.geus.info/cet/Ilulissat_dk/voii02-dk.html

One of the great curiousities is that we know the deep channel goes all the way back to the deep interior of the island yet there is no surface manifestation of that channel for two-thirds of the distance. It is just barely possible to see the three tributaries but there seems to be non-use of two of them. What made the deep channels if not the ice -- paleo stream flow?

like pause /break in the first animation?

Ok, i have put in animation breaks and slowed them down slightly. It is really best to save the image to disk, then open in Gimp etc and adjust speed to liking in animation player. Too bad gif format does not provide a little shell with step-through and speed controllers.

I'm looking forward to some amazing winter animations from Sentinel IW. It would be a blessing to have images on schedule every 12 days regardless of weather. Sentinel sees things differently so it may be possible to track movement much farther into the interior.

Actually it seems like we have 43 Sentinel IW's already for the calving front area. Some of them seem to go a ways inland from the calving front. It would be quite a crunch to collect and re-project all of this (or even a key subset, say May to Oct). It would almost be better to store these elsewhere at an online archive at full size and let people whittle that down for their individual projects.

S1A   IW   GRDH   1SSH   2014   11   24
S1A   IW   GRDH   1SSH   2014   11   28
S1A   IW   GRDH   1SSH   2014   12   06
S1A   IW   GRDH   1SSH   2014   12   10
S1A   IW   GRDH   1SSH   2014   10   11
S1A   IW   GRDH   1SSH   2014   10   19
S1A   IW   GRDH   1SSH   2014   10   23
S1A   IW   GRDH   1SSH   2014   10   31
S1A   IW   GRDH   1SSH   2014   11   04
S1A   IW   GRDH   1SSH   2014   11   12
S1A   IW   GRDH   1SSH   2014   11   16
S1A   IW   GRDH   1SSH   2014   12   18
S1A   IW   GRDH   1SSH   2014   12   22
S1A   IW   GRDH   1SSH   2015   01   03
S1A   IW   GRDH   1SSH   2015   01   11
S1A   IW   GRDH   1SSH   2015   01   15
S1A   IW   GRDH   1SSH   2015   02   04
S1A   IW   GRDH   1SSH   2015   02   08
S1A   IW   GRDH   1SSH   2015   02   16
S1A   IW   GRDH   1SSH   2015   02   20
S1A   IW   GRDH   1SSH   2015   02   28
S1A   IW   GRDH   1SSH   2015   03   04
S1A   IW   GRDH   1SSH   2015   03   12
S1A   IW   GRDH   1SSH   2015   03   16
S1A   IW   GRDH   1SSH   2015   03   24
S1A   IW   GRDH   1SSH   2015   03   28
S1A   IW   GRDH   1SSH   2015   04   05
S1A   IW   GRDH   1SSH   2015   04   09
S1A   IW   GRDH   1SSH   2015   04   17
S1A   IW   GRDH   1SSH   2015   04   21
S1A   IW   GRDH   1SSH   2015   04   29
S1A   IW   GRDH   1SSH   2015   05   03
S1A   IW   GRDH   1SSH   2015   05   11
S1A   IW   GRDH   1SSH   2015   05   15
S1A   IW   GRDH   1SSH   2015   05   27
S1A   IW   GRDH   1SSH   2015   06   08
S1A   IW   GRDH   1SSH   2015   07   26
S1A   IW   GRDH   1SSH   2015   08   07
S1A   IW   GRDH   1SSH   2015   08   19
S1A   IW   GRDH   1SSH   2015   08   31
S1A   IW   GRDH   1SSH   2015   09   12
S1A   IW   GRDH   1SSH   2015   09   24
S1A   IW   GRDH   1SSH   2015   10   18

[nukefix]

Would it be a lot of work to run the stack through imgraft to see what the speed is?

http://imgraft.glaciology.net/

[oren]

It's scary how the whole ice sheet is moving, not just the ice stream.. ice discharge might accelerate by having the sides of the stream contribute more and more

My thoughts as well. Attached at two 'entrainment' animations that give an overview of where the ice being discharged is originating from. One of these should display immediately, the other has better resolution but needs a click. Full 15 m resolution (not provided) is a large file that overloads most graphic cards as it tries to show 150 ms frames.

Somehow I had gotten the misapprehension that Jakobshavn Isbrae is, as the name suggests, an ice stream -- a distinct feature with sharp shear zones that moves much faster than the enveloping ice sheet creeps along. Maybe it was in years gone by.

Here it looks like very little of the calving front originates in the smooth narrow ice stream but rather that most of it comes from the surrounding ice sheet that has been 'gotten up to speed' some 15-70 km back as it joined the ice stream. It would be challenging though not impossible to make quantitative comparisons to 2014 and 2013 to see if ice sheet entrainment has increased.

Landsat just shows an overhead view of surface ice. It is a bit puzzling to see how this works out as depth because the incoming ice sits over much higher bedrock and so is much thinner than the deep narrow channel, yet this latter has to be filled completely.

The other surprising feature is seen in the southeast corner. It appears that the (smooth white) upper ice sheet above the ablation zone is shedding waves of ice that reach the ice stream in a year or two. The third animation shows a piece of this at full resolution.

The large animation should be textbook material.
Has anyone ever measured the speed of the ice sheet in various locations around JH and its behavior over time? I'm willing to bet there's been acceleration there, especially the parts that have lost buttressing in the last few years due to calving front retreat. I would presume that ice sheet acceleration follows stream acceleration with a delay, and that the whole ice sheet in that area was much more stable before the loss of the tongue in the early 2000s.

[nukefix]

Has anyone ever measured the speed of the ice sheet in various locations around JH and its behavior over time? I'm willing to bet there's been acceleration there, especially the parts that have lost buttressing in the last few years due to calving front retreat. I would presume that ice sheet acceleration follows stream acceleration with a delay, and that the whole ice sheet in that area was much more stable before the loss of the tongue in the early 2000s.

Yes, historical measurements at JH are available for example over here:

http://products.esa-icesheets-cci.org/

The product is question is greenland_jakobshavn_timeseries_2002_2010.zip

IV = Ice Velocity

We could process the optical/radar stacks with Imgraft and compare with the older numbers..

[Shared Humanity]

If you look at the large animation and the associated still image, you can see a large section of the ice sheet south of the main ice stream that is heavily crevassed and moving rapidly north,flowing into the  main ice stream. If you again look at the topographical map of this region, it is clear that this section of the ice sheet is being fractured as it moves up and over elevated land that lies south of the main stream and north of a shallow inland sea.

[A-Team]

We could process the optical/radar stacks with Imgraft and compare with the older numbers

Yes, we need to 'up our software game' at some point. There is only so far we can go without using something like Imgraft. Sounds like they've made quite a few improvements in it. It would also be ideal for processing drone photography (if we ever get to that next summer).

We present a flexible, open source Image GeoRectification And Feature Tracking toolbox (ImGRAFT). ImGRAFT incorporates all the processing steps needed to transform monoscopic, terrestrial, oblique images into a velocity field. ImGRAFT assimilates the rectification of the images and subsequent feature tracking into one toolbox.
http://www.geosci-instrum-method-data-syst.net/4/23/2015/gi-4-23-2015.pdf

For example, the big calving event of 15 Aug 15 might have removed so much buttressing that the ice stream then surged. As a practical matter, this means comparing velocities of Landsat pairs days 210/226 with days 212/228 just to get at days 226/228 that bracketed the event. That probably requires more attention to photogrammetry details to have any confidence in the overall accuracy.

The glacier was moving right along between days 212/228 below (multiply pixel separations by resolution and divide by 16 day orbital return to get at average velocity). Note the striking shear lines implied by rapid lateral fall-off in velocities.

It would also be desirable to measure some of the slower velocities in more remote areas which we could do accurately using longer time spans. The bedrock surely influences surface ice as Shared Hum points out but this effect will fall off farther from the coast -- when the ice gets up to 3,000 m thick, it hardly notices a blip in the bedrock. You see the same effect along an ordinary stream ... where shallow, there are riffles over cobbles, where deep the water surface is smooth.

[A-Team]

The animation below shows a preliminary analysis of the colors in the vertical iceberg calved in May 2014. The color scanner moves from right to left, picking a color every 10 pixels. (It is set to average the colors over a radius of 5 pixels to minimize noise.)

The gif format doesn't display the result too well since it wants to cut 24-bit color down to the 8-bit, losing all the subleties of color gradients. However the HSV and RGB numbers are still accurate. The palette of swatches shows the previous 12 selections. It emerges that even the white ice is already a bit bluish.

The second animation takes a succession of 15 parallel 'drill cores' one pixel wide and 700 deep. These are expanded to 200 x700 to bring out color variation as the 'drilling rig' moves to the next spot over (which is up as the iceberg lies sideways).

Because the vertical profile of ice at Jakobshavn has been reported to retain both age stratigraphy (paleo volcanic ions) and crystal orientation fabric expected in proportion to depth, bottom freeze-on of ice does not play a significant role here. (Bottom freeze-on was once put forward as an explanation for upheaval features in extreme northern Greenland but that has no experimental support and was later replaced by other hypotheses.)

In summary, these calved blue bergs and the sediment carrying pieces of these events would provide an excellent substitute for drill cores on Jakobshavn (which are infeasible because of crevasses and rapid movement). Ice from this particular event is long gone and with it, the opportunity. However there were two 'sapphire blue' events this summer and no doubt more next summer.

These are best studied with instrumented drone overflights as it would be very dangerous to land a survey crew (because of potential roll-overs or swamping from subsequent calving tsunamis). I've not heard of drones being outfitted with sample grabbers but that certainly seems feasible.

[nukefix]

Here's the S-1 EW from 24.10 in UTM22 with 15m pixel size. It has been oversampled from the EW-product with 40m pixel size.

edit: looks like I started with thw wrong input product (GRDM instead of GRDH)...fixing it now

[oren]

The rock island at the split of the north and south branches is very visible here.

[A-Team]

looks like I started with the wrong input product (GRDM instead of GRDH)...fixing it now

Be of interest to retain both outputs to illustrate the difference in resolution.

The rock island at the split of the north and south branches is very visible here.

We're all agreed there is no exposed rock? Yet. It's sill covered with 135 meters of ice. Question is, why are certain regions in the Sentinel showing up as different colors (as polarizations HH, HV, blue for RGB?). Possibly because stagnant ice is older and smoother at the scales relevant to reflection at the radar wavelengths, it being primarily a polarization effect.

In the attached animation shows the rock island and its average color as insets in the first frame, upper left. Subsequent frames find this average color throughout the picture with decreasing degrees of stringency (the radius out from the average color increases from 0, 5, 10, 15 to 20). That is, if the island were showing exposed rock, there would be a lot more exposed rock to the east which we know from the NASA-supplied accurate colorizations is not the case.

[nukefix]

The rock island at the split of the north and south branches is very visible here.

Yes, but the area is still ice covered - I wonder if there are any reliable estimates on where the bedrock is at that point.

Here's the image again from EW GRDH, upsampled to 15m UTM22:

[oren]

I did not mean to imply it's actually exposed rock... I was just surprised that the hidden island surmised a few posts back by Espen was so visible here as if it was actually exposed rock.

[A-Team]

I wonder if there are any reliable estimates on where the bedrock is at that point.

Normally the ice-penetrating radar would clearly show depth to bedrock. However, a few posts back, I could only find a single overflight during the last 22 years and it was unsatisfactory. It is possible that something better has been done with airGrav data.

Bed topography of Jakobshavn Isbræ, Greenland from high-resolution gravity data
Lu An, UC Irvine AGU2015 abstract [I wrote to see if details could be made available, have not heard back]

... little is known about the depth of the glacier at its grounding line and upstream of the grounding line and the sea floor depth of the fjord is not well known either.

Here we present a new approach to infer the glacier bed topography, ice thickness and sea floor bathymetry near the grounding line using high-resolution airborne gravity data from AirGRAV collected in August 2012 from a helicopter platform. The data was combined with radio echo sounding data, discrete point soundings in the fjord and the mass conservation approach on land ice.

AirGRAV acquired a 500m spacing grid of free-air gravity data at 50 knots with sub-milligal accuracy, i.e. much higher than NASA Operation IceBridge’s 5.2km resolution at 290 knots. We use a 3D inversion of the gravity data combining our observations and a forward modeling of the surrounding gravity field, and constrained at the boundary by radar echo soundings and point bathymetry. We reconstruct seamless bed topography at the grounding line that matches interior data and the sea floor bathymetry. The results reveal the true depth at the elbow of the terminal valley and the bed reversal in the proximity of the current grounding line.

Then there was this 2011 discussion paper that did not make it into publication but explains the methods and has some relevant illustrations in the neighborhood of the island and calving front. The island is between the T0 and T14 cross-sections so not quite what we need.

Geophysical evidence for soft bed sliding at Jakobshavn Isbrae, West Greenland
www.the-cryosphere-discuss.net/5/339/2011/

We analyze new gravity and magnetic profiles across the glacier, extending from the mouth of the outlet fjord to 64 km inland of the 2008 grounding line. Our results provide new insights into Jakobshavn Isbrae's geologic underpinnings and controls on the basal velocities. Earlier studies of basal processes minimized basal slip as a fast flow mechanism. Currently, velocities are up to double those considered in these studies, necessitating a reanalysis of the basal conditions. The gravity field along the glacier's main trunk cannot be attributed to the gravitational effect of bed topography and the overlying ice sheet. We interpret the remaining gravity signal as evidence of up to 2400 m of low density sediment beneath the main trunk. Examining recent velocities, we find basal slip is a major contributor to ice flow along most of the sediment filled trough. Within 54 km of the grounding line, only isolated 1–3 km wide regions have velocities that possibly result solely from internal deformation of the ice. We conclude soft bed sliding over the thick sediment wedge beneath Jakobshavn Isbrae is the dominant mechanism of fast flow.

[Shared Humanity]

A-Team...

Do you know how thick the ice is in the heavily crevassed field that lies south of the main branch which is flowing north?

From the topography map, it appears as if the bedrock quickly climbs from below sea level (the shallow inland sea) to approximately 1000 meters. Given this rapid rise, it would seem to be sufficient to fracture quite thick ice.

[Shared Humanity]

Looking at the large animation of the crevassed ice sheet that is moving north towards the main branch, there appear to be melt lakes forming and draining just prior to the point where the ice begins to crevasse. The close up animation shows these more clearly. These melt lakes, if that is what they are, would drain towards the shallow inland sea that lies south of the elevated bedrock which is causing the fracturing. I suspect that there is permanent, perhaps shallow, accumulated melt water that has collected beneath this section of the ice sheet.

Would this melt water have any effect on the speed of the ice sheet?

[nukefix]

From the topography map, it appears as if the bedrock quickly climbs from below sea level (the shallow inland sea) to approximately 1000 meters. Given this rapid rise, it would seem to be sufficient to fracture quite thick ice.

Ice under pressure is pliable like honey, crevasses are formed in the top-part but they should not go down to - or even close to - the bottom, as far as I know.

[A-Team]

Some of the crevasse fields result from ice being bent over a near-surface bedrock feature faster than its plastic response, others from dynamic thinning and lateral shearing. Some of the meltwater lakes form in the same standing wave hollow year after year.

I've been intending to do south of the south branch to see where the ice divide is between Russell and Jakobshavn drainages. Also, to see how far out to the east ice sheet movement can still be resolved (and where it first decouples from the bedrock channel).

Because even the Landsats of the same path, row do not co-register initially -- and because others like 8,11 are unacceptably rotated in addition to displaced, I haven't been able to just align the images once and for all and then just crop out individual animations (or tile the whole area with animations at full resolution). If bad local clouds are discarded, tiled animations get out of synch, if clouds are retained the animations become distracting..

Below the 9,11 and 10,11 are aligned to give a better 17-frame time series for the north bank of the south branch. Note the finger coming into the calving front from the northeast. For what it is worth, the second 26-frame animation throws in the 8,11 not co-located except at the tip of the island as well as some cloudy frames.

Here is the list of clear Landsat's for 2015, with another (the last?) expected in about 4 hours. The 3rd column calculates the interval in days between scenes, which you can see is fairly irregular for purposes of smooth motion.

2015  043  00  09  11  LC80090112015043LGN00
2015  050  07  10  11  LC80100112015050LGN00
2015  052  02  08  11  LC80080112015052LGN00
2015  059  07  09  11  LC80090112015059LGN00
2015  075  16  09  11  LC80090112015075LGN00
2015  091  16  09  11  LC80090112015091LGN00
2015  107  16  09  11  LC80090112015107LGN00
2015  114  07  10  11  LC80100112015114LGN00
2015  123  09  09  11  LC80090112015123LGN00
2015  132  09  08  11  LC80080112015132LGN00
2015  146  14  10  11  LC80100112015146LGN00
2015  148  02  08  11  LC80080112015148LGN00
2015  155  07  09  11  LC80090112015155LGN00
2015  162  07  10  11  LC80100112015162LGN00
2015  180  18  08  11  LC80080112015180LGN00
2015  187  07  09  11  LC80090112015187LGN00
2015  194  07  10  11  LC80100112015194LGN00
2015  196  02  08  11  LC80080112015196LGN00
2015  210  14  10  11  LC80100112015210LGN00
2015  212  02  08  11  LC80080112015212LGN00
2015  226  14  10  11  LC80100112015226LGN00
2015  228  02  08  11  LC80080112015228LGN00
2015  235  07  09  11  LC80090112015235LGN00
2015  251  16  09  11  LC80090112015251LGN00
2015  267  16  09  11  LC80090112015267LGN00
2015  283  16  09  11  LC80090112015283LGN00

[Espen]

That island, and I am pretty sure there is one under that pile of ice, could easily be 150 - 300 meters at the top judging from different images / photos. That is also the reason the ice on the top is not moving at all, if the bedrock at this point was below sea  / calving front level it would a least move somehow!

But what annoys me the most is that an amateur like me is the first to ever mention the word Island? 

[A-Team]

Right. I guess if it's just below sea level, it's a virtual nunatak?

I started in on ice sheet movement far to the east of the calving front. There's definitely a 'trust issue' because even sticking to the same path,row 8,11 the ground control points are far to the west. Everything else may be moving.

There are additionally issues with seasonal changes -- the 15 m animation (needs a click to start) compares an unmistakable triangular feature at about the first tributary between day 52 and day 228. The farther the days are apart, the more motion that can be detected! The features are merely co-registered here, no motion is shown. Yet. They are shown boxed in white in the first image for context.

[Espen]

Yes it is amazing, Jakobshavn "probably the most stared at glacier" (reminds me of a beer commercial btw.) got´s own Tobias Ø: https://da.wikipedia.org/wiki/Tobias_%C3%98

[A-Team]

Den 28. april 2001 rejste GEUS og ASIAQ henholdsvis Dannebrog og det grønlandske flag på øen, og øen betragtes nu som dansk territorium.[2] Suverænitetshåndhævelsen forventes at have stor betydning, når det i de kommende år skal afgøres, hvem der har ret til Ishavet (Nordpolen)

No way Tobias Ø is Danish. I visited there earlier on the 27th and declared it UN property, no hunting or drilling allowed within a radius of 200 nautical miles ever under the conservation easement.

Deep interior, due west ~200 km of JI's island I'm still seeing 9 pixels of westward motion between day 052 and day 212 for 0.84 meters per day or 308 meters per year (365*9*15/170). This does not max out Landsat sensitivity but errors will grow farther east.

This awaits a couple of quality controls because the ground control points are all by the calving front: (1) verification with multiple dates that the motion is consistently and proportionally greater with elapsed time and (2) verification in a tall vertical strip that other ice sheet motion is inward at other angles.

[nukefix]

Sentinel-2 should have 10m absolute positioning accuracy so it should be possible to just stack images and perhaps make max 1 pixel adjustments. That should simplify things a lot as even after map projection data from different tracks should align nicely.

edit: this should be a lot better than Landsat GeoCover images which appear to have a positional accuracy of less than 50m.

[A-Team]

Sentinel-2 should have 10m absolute positioning accuracy so possible to just stack images and  make max 1 pixel adjustments. That should simplify things a lot as even after map projection data different tracks align nicely. a lot better than Landsat GeoCover images which appear to have a positional accuracy of less than 50m.

Right.  The nitty gritty of Landsat Level I, GeoCover and potential future products becomes important at some point. I filled out their survey the other day, not real sure what all the checkboxes meant but what harm can come from requesting refined products.

The U.S. Geological Survey is working to improve the quality of Landsat products, in response to the evolving needs of the scientific research and applications communities.  Toward this end, we have created a brief survey that seeks to determine preferred levels of Landsat data processing, as well as associated quality assessment information at the pixel level. This information will allow future Landsat products to more readily satisfy research and application requirements. The survey can be accessed at http://landsat.usgs.gov/intro_eolps.php. Your participation is strictly voluntary and is greatly appreciated.

It is feasible to manually co-register a couple dozen Landsats to ground control points in Gimp but inconvenient to do so in ImageJ. Unfortunately, only the latter can do contrast adjustments on 16-bit files. Contrast is almost an insurmountable issue in scenes covering both the (dark) ablation and (white-out) accumulation zones.

Supposing a master stack could be made at manageable file size, cropping each region-of-interest down to blog width (700-1400 pxls) will have to be followed by individual contrast adjustment of each layer. Some of that, like normalization gamma, and local histogram equalization might be scriptable but many scenes will still require intervention. However desirable an optimal motion series might be, it is not necessary for just visualizing or measuring feature movement.

It looks like 30 tiles would do it, a quarter of that if tile size were doubled to 21 km on a side. While the 27" AAPL monitor displays 5120×2880 so 14,745,600 pixels, a typical Landsat has 17,481 x 17,581 or 307,333,461 pixels so it would take a 4x5 bank of 20 monitors to display an image properly, though 2x1 would work ok restricted to the area east of the island.

Features vary greatly over melt lake season. There will be some standing wave objects that are not moving downslope amidst the objects that are traveling routinely. Automatic feature recognition tools will not work satisfactorily over the range of imagery involved, as is obvious from the animation a post or two back.

Here is the available cloud-free data for the most favorable Landsat scene 08 11 for studying motion east of the calving front. The 3rd column gives the total number of elapsed days since the first 2013 image. This determines the overall sensitivity.

That is, if the earliest and latest scenes are compared and 1 pixel (15 m) of motion is (optimistically) detected over the 816 days, the velocity pencils out to 18.4 mm per day or 6.7 m per year. Peak reported velocity at the calving front is 52 m/day or 18980 m per year (neglecting seasonal variation), which is 2832 times faster. The summit ridge pole arrays move ~ 1 m/year.

Some of this is attributable to thinning/creep/rheology but at higher elevations it would largely represent rigid body bulk ice sheet movement. There are prospects too for determining not just average velocities between scenes but also (just barely) year-on-year acceleration using day 0, 416 and 816.
 
2015   228   816   8   11   LC80080112015228LGN00
2015   212   800   8   11   LC80080112015212LGN00
2015   196   784   8   11   LC80080112015196LGN00
2015   180   768   8   11   LC80080112015180LGN00
2015   148   736   8   11   LC80080112015148LGN00
2015   132   720   8   11   LC80080112015132LGN00
2015   052   640   8   11   LC80080112015052LGN00
2014   305   528   8   11   LC80080112014305LGN00
2014   209   432   8   11   LC80080112014209LGN00
2014   193   416   8   11   LC80080112014193LGN01
2014   161   384   8   11   LC80080112014161LGN00
2014   097   320   8   11   LC80080112014097LGN00
2014   081   304   8   11   LC80080112014081LGN00
2013   302   160   8   12   LC80080122013302LGN00
2013   142   000   8   11   LC80080112013142LGN01

[P-maker]

A-team:

”[no Eemian ice in west-central Greenland]”

I totally concur!

What about encouraging some of your friends (with amble budgets and plenty of flying time) to spread out a series of ground control points. How about some 1 sq. km fields painted black by dust and sitting next to cleared white (1 sq. km) blocks along one of the slower flow lines? The field work would be interesting to follow as the “black blocks” dig themselves down and the “white blocks” stand tall above the ice surface.

Cheers P

[A-Team]

series of ground control points. How about some 1 sq. km fields painted black

In the early '90's, K Echelmeyer was putting dark fabric blocks on seracs to track motion, making mid-Jakobshavn traverses on foot, and climbing hollow ice blisters, see Surficial Glaciology of JI part 1 J Glaciology 37(121)1991. Without nunataks, nothing is really horizontally stationary east of JI until Summit. Landsat metadata allows orthorectification of images (given the high precision DEM) without literal ground control points. In theory. What is available for download isn't.

The remnant Eemian ice reaches almost down to Jakobshavn; however it is east-central (where the bottom ice has very slow horizontal motion) rather than coastal. Minute 2:02 of  http://svs.gsfc.nasa.gov/cgi-bin/details.cgi?aid=4249

The Renland ice cap project (RECAP) reached Eemian ice well before bedrock at 584.1 m this June. This isolated dome is also a bit north of Jakobshavn overlooking an east coast fjord. It was not included in the above animation as it lacked ice-penetrating radar at the time JA MacGregor pulled together the isochrons islandwide.

http://news.ku.dk/all_news/2015/06/new-greenland-ice-core-drilled-through-the-renland-ice-cap/ overview
http://recap.nbi.ku.dk/field_diaries/2015/?p=4 well-done field diaries

G Tsoflias has helpfully just released the slides underlying two 2011 AGU talks to his ResearchGate page; no journal articles surfaced but the first remains exceedingly interesting. They reprocessed ice-penetrating radar Jakobshavn grid lines flown in 2008 using seismic interpretation software (SMT Kingdom Suite $$$), obtaining 10 isochron surfaces that are simply not visible in the Cresis radargrams as posted, though age structure can still be seen in transects farther south (4th image).

They did not tie their isochrons into summit ridge ice core dates but that is easily done today -- the 'three sisters' triple (middle of Last Ice Age) is immediately recognizable whereas the 94 kyr 'two brothers' doublet is not. The preservation of isochrons in a central flowline in the main channel of Jakobshavn proves that ice layering has survived the fast flow regime above the elbow and that calved blue bergs may retain them as well.

[nukefix]

Interesting-looking related paper here: http://onlinelibrary.wiley.com/doi/10.1002/2014JF003215/full

[A-Team]

Interesting-looking related paper here: http://onlinelibrary.wiley.com/doi/10.1002/2014JF003215/full

Indeed it's interesting, I have written about that article maybe 25 times, including in the very last post "ice-penetrating radar at the time JA MacGregor pulled together the isochrons islandwide."

Just looking now at who has been citing that article, I came across a very interesting thesis that purports to resolve the Petermann upheaval issue in terms of geological shear. Mandel dates the "triple_mid_bed.line" (middle sister) to 46.5 kyr. Most of the work --  which bears an uncanny resemblance to many Petermann posts and illustrations (forum not cited!) -- uses fit-for-purpose $$$ geological modeling software called MOVE. http://www.mve.com/software/move

http://epic.awi.de/38445/1/BachelorThesis_FelicitasMundel.pdf

a 3D model of internal ice layers of the Greenland Ice Sheet is proposed. It was created with the structural geologic software MOVE, using Radar Depth Sounder profiles of the catchment of Petermann Glacier, located in North-West Greenland. The model visualizes the large-scale folding of the ice layers and shows that basic geologic processes are mainly responsible for the creation of these cylindrical folds.

The funnel-shaped channel of Petermann Glacier leads to compression across flow direction, while increasing ice velocity causes local tensile stress along flow. In deep ice layers, sheath folds are present. Due to decreasing ice velocity towards the bottom of the ice column, a strong shear stress acts on the ice layers, which leads to the formation of sheath folds.

Recall the non-committal review of upheavals in MacGregor et al (which restricted itself to auto-traceable features away from the coast):

For several radargrams, the structure of the basal layer qualitatively resembles that of sheath folds in rock, which can form due to flow in the vicinity of rheologically weak layers or slip surfaces [Reber 2012]. These folds are three-dimensional structures whose vertical cross section (as they would be observed by ice-penetrating radar) varies depending on the cross section’s orientation with respect to the main axis of the fold. Sheath folds can generate characteristic “eye” structures, which are observed occasionally near the base of the GrIS.

If these features are indeed sheath folds, then these observations appear to support the hypothesis presented by NEEM  explaining the formation of disrupted basal ice. They hypothesized that the interface between rheologically weaker ice from the LGP and stiffer ice from the Eemian period leads to the development of a slip surface there. This slip surface is equivalent to a false bed, which has also been identified in the West Antarctic Ice Sheet at Siple Dome.

Reflections within and bounding these structures are often indistinct and difficult to trace except by purely manual methods, so we did not map them exhaustively. The apparent slopes of these reflections sometimes exceed that which can be imaged by airborne radar sounding so their resemblance to sheath folds may be partly due to aliasing of backscatter from these complex features. The possibility of sheath folds within the GrIS does not exclude the possibility of widespread basal freeze-on beneath the GrIS inferred from these same radar data [Bell 2014]

[A-Team]

Here is a bit more from that 2011 poster of Arntsen/Tsoflias characterizing isochron depth east of Jakobshavn from a dense radar grid flown earlier, again using proprietary seismic software. The graphics in the pdf were very erratically sized making the second animation jump around. These slides show that older iso-surfaces of constant age reflect bedrock elevational features to a greater extent.

[A-Team]

The animation below shows the calving front and 42 km to the east at 30 m resolution over 14 cloud-free days for band 8 of Landsat-8 path,row 08,11. At the original resolution, this would be 2800 x 1400 pixels which is too big for most monitors and beyond the ability to smoothly animate so the overview file below is reduced to 1400 x 700. This will display ok with a click though the first loop is slow (frames have to be cached).

The post below tiles this original into four pieces each 1400 x 700 at the native 15 m resolution of Landsat panchromatic. There are a lot of contrast issues going out farther towards the summit ridge as the ablation zone is generally quite dark but the higher accummulation zone very white.

[A-Team]

The four animations tile the one above, bringing it to the full 15 m resolution of the satellite. There is a whole lot of complex motion in the surrounding ice sheet as it is entrained into the main ice stream (which only contributes a small portion of the calving front).

I've not seen a published model of Jakobshavn which acknowledges this motion, much less accommodates the frictional heat and shearing forces that it brings in. While it's understandable why someone would elect to model the ice stream a simple single central flowline, it's equally clear that this has zero prospects for predicting future behavior of Jakobshavn.

On the technical side, I looked into three potential improvements: (1) automating and optimizing contrast adjustment via a secondary masking image provided by the all-important adaptive contrast tool in ImageJ (CLAHE menu), (2) rectifying at least locally other path,rows of Landsat to provide smoother cloud-free time series, and (3) an online service that could provide these animation on demand for user-supplied animation corners and date ranges anywhere in coastal Greenland or Antarctica, if a seamless zoom into a tiled pre-compute wasn't satisfactory.

This is easy enough to do since EarthExplorer already provides a bulk download tool, corner cropping is readily scripted in ImageJ and Gimp, cloud cover is provided in metadata, #2 is not anything new and #1 is feasible (which is not to say that some user intervention will still be helpful, say selecting frames for refinement from the low res pre-compute). In effect Landsat in dimension 2+1. Specifying band combinations does not add significant additional complexity to the service. Sentinel1-A, which has a different contrast issue but not a cloud problem,  is already set up for a very similar process via the Toolbox provided.

[oren]

Very useful animations. Good luck to anybody trying to model all this complex flow.
In general, JH flux acceleration should be measured not just by flow speed at the main channel plus retreat, but also by measuring all the additional minor calving fronts that get opened up. Although these are slower, they have the potential to be much wider in total. And since most of that stuff is currently buttressed by the main channel, serious front retreat might remove the traffic jam and result in significant acceleration of the side flows.

[A-Team]

currently buttressed by the main channel, serious front retreat might remove the traffic jam and result in significant acceleration of the side flows

That's my view too, all ice flows have to be considered if buttressing is to be addressed holistically.

To get at the flows farther up the ice sheet, something has to be done in terms of optimizing Landsat contrast in white-on-white regions. Below are a couple animations illustrating the problem and what can be done about it. The last frame in the 2nd animation uses some secret sauce that works quite well in refining radargrams as well. It takes a couple hours per Landsat scene of processing time.

It turns out there are a whole lot of subtle features out there in what initially appear to be very bland regions. Whether these are standing waves, wind-blown snow, or surficial representation of ice flow strain could be resolved with time series... if contrast can be consistently repaired across different illumination and position of melt season.