Greenland subglacial topography

[A-Team]

Shared asks if the Greenland bedrock is really bedrock under and near the Grand Canyon or some version of till and porewater and how that might be experimentally determined.

First, since the Canyon passes very close by the NEEM and NGRIP cores-to-bedrock, it's worth reviewing what those cores found at the bottom. Neem was highly folded but still frozen, NGRIP was dirty mush and had that never-pursued pine needle. However the ice sheet is hardly moving along the summit ridge so bottom till might not be well developed or representative of the Canyon below Petermann.

It is possible to reprocess the radar using advanced techniques (alternatively, curie depth) to get global maps of bottom temperate ice and geothermal gradient but these have generated little enthusiasm. Massive bottom freezeups would bring till closer to the surface but internal reflectors there have not yet been cored or even interpreted.

On middle NEGIS, researchers supplemented radar with 'active source' seismic (explosives) to characterize bottom till in a trio of 2014 articles, notably www.the-cryosphere-discuss.net/8/691/2014/. Two of the main authors there, Alley and Anandakrishnan, wrote the 2007 Whillans paper on the Antarctic!

Seismic AVO analysis [amplitude vs offset, see wikip] allows us to determine elastic properties of the basal material (density, compressional-wave velocity, and shear-wave velocity), an indicator of likely basal material. Our AVO analysis indicates that ∼ 10 m dilatant till is present near the the central part of the ice stream (site C). The subglacial sediment is somewhat more consolidated at sites B and D, one in each shear margin, with well-consolidated sediment at sites A and E outboard of the shear margins. The pattern of radar basal reflectivity and hydropotential supports and extends the seismic data. The radar data indicate that the central portion of the ice stream bed is wet, with water flow generally oriented along-flow.

[A-Team]

The image below shows 5 radar flights crossing the Canyon between NEEM and NGIP plus one flight segment 20110506_02_013 (orange arrows) that actually follows the Canyon downstream for some distance. The little pushpins show the bottom of the Canyon within the radargrams.

The thick green line connects up the bottom points in a linear fashion. Of course there is no way of knowing what route the trough actually takes between data points, which are separated by as much as 24.4 km.

Shared H asked if there is enough information in these radar transcripts to determine the extent of water or till at the bottom of the Canyon. I would say not.

The 2011 down-canyon segment exhibits the richest substructure (shown full-size from the 'Three Sisters' to bedrock in the 2nd image) but its interpretation is rather up for grabs.

[A-Team]

The breakthrough new paper by JA MacGregor -- literally 21 years in the making -- on englacial Greenland is introduced over at the 'What's new" forum, http://forum.arctic-sea-ice.net/index.php/topic,1090.msg43840.html#msg43840

I'll just repeat the minimum here and get on with a one-by-one description of the figures and animations. It will take the rest of the month to get through this paper and its data repositories.

Radiostratigraphy and age structure of the Greenland Ice Sheet
JA MacGregor, MA Fahnestock, GA Catania, JD Paden, S Goginen, SK Young, SC Rybarski, AN Mabrey, BM Wagman, M Morlighem

We present a comprehensive deep radar stratigraphy of the Greenland Ice Sheet from airborne deep ice-penetrating radar data collected over Greenland by U Kansas between 1993 and 2013 [2014 added very little]. To map this radiostratigraphy efficiently, we developed new techniques for predicting reflector slope from the phase recorded by coherent radars. This radiostratigraphy provides a new constraint on the dynamics and history of the Greenland Ice Sheet.

When integrated along-track, these slope fields predict the *stratigraphy and simplify semi-automatic reflection tracing. The stratigraphy was dated via synchronized depth–age relationships for the six deep Greenland ice cores [Camp Century, NEEM NGRIP, GRIP, GISP2, DYE3].

Additional reflections were dated by matching reflections between transects and by extending depth–age relationships using the local effective vertical strain rate [see http://gravity.ucsd.edu/pub/2004_elsberg.pdf].

The oldest reflectors (Eemian) are found mostly [but not entirely] in the northern part of the ice sheet. Reflections do not conform to the bed topography within the onset regions of fast-flowing outlet glaciers and ice streams. Disrupted  stratigraphy is also observed in a region north of NEGIS that is not presently flowing rapidly.

Dated reflections are used to make a 3D gridded age product for the ice sheet and to determine the depths of key climate transitions that were not observed directly.

283 MB. main article pdf, figure pdfs, animations mp4 [Download copy-edited version from journal.]
  ftp://ftp.ig.utexas.edu/outgoing/joemac/gris_strat_rev2.zip

many MB. raw layer data MATLAB HDF5 [Don't download, wait for NSIDC to host update.]
  ftp://ftp.ig.utexas.edu/outgoing/joemac/Greenland_radiostratigraphy.mat

many MB netCDF viewable in Panoply freeware, age volumes, isochron depths of 11.7, 29, 57 and 115 kyr bands. [Don't download, wait for NSIDC to host update.]
  ftp://ftp.ig.utexas.edu/outgoing/joemac/Greenland_age_grid.nc

Fig.1: Part a shows all possible Cresis radar tracks colored by year (I've made and posted this several times); part b shows the subset of radar tracks deemed suitable for this project, colored by number of trackable 1-km segments. Note   the ice margins were unsuitable for tracking isochrons (black lines). Thick lines in the image do not scale to actually display flight number details; this needs to be redone in kml for Google Earth.

Fig.2: Key to understanding the paper: a very helpful and detailed flow chart of methods for tracing, dating, and normalizing stratigraphy.

Fig.8: The distribution of oldest ice on Greenland one of several conditions is bundled here with Fig.1b

[sidd]

the .mat file is 1.5 G , nc is 2.4 G, latter opens fine in qgis, former crashes octave, but qgis or grass should be coaxable into reading.

fascinating

[A-Team]

sidd, o that i had time

Time? Seems like you can do a lot with very little -- hope you can make some cool products for us. And thanks for the reports on file sizes and open source software that effectively opens them. I will continue to integrate that information into earlier posts as it comes in.

I looked at the 3 animations in MacGregor (2015) which have links from the journal. These mp4's are not small at 150, 36 and 57 MB but they open and run smoothly in QuickTime without any hitches. Which is good because you need to watch them many times over. I'm looking at some schemes for re-animating them in Gimp at forum-compatible sizes (starting with QT's iPhone export option).

anim1.mpg  http://tinyurl.com/k3w8csv
anim2.mpg  http://tinyurl.com/oolbkek
anim3.mpg  http://tinyurl.com/ooa9teo

Animation 1 is a giant pan along all the radar isochrons they could follow for longish stretches and assign dates to by tying into the NGRIP ice core. They've rescaled everything uniformly which requires a minimum of four datable isochrons, the NGRIP depth–age relationship and a second-order polynomial best fit as described in the article pdf.

The color coding is fairly self-explanatory: vertical magenta pipes if one of the six ice cores is in view, vertical dashes to separate distinct isochron segments, and green for bedrock. Instead of a sea level line, they provide an NGRIP bottom line. The drill site being at 2917 m elevation and ice core 3085 m long gives a sea level line 168 m higher than the furnished line (which I suspect they used because nothing below it had reliable dating). The white boxes show Cresis ID segments. Oddly no lat,lon or Greenland thumbnail is provided so you'd have to look up the Cresis ID to see where you are.

I stopped the animation at a few interesting places for screenshots, cleaned them up a bit and reduced from 1360 default QT width by 50% to get within forum width at 700 pixels. QT lets you mouse the image along, meaning you're not stuck with discrete frame cuts at the junctions. The QT time allows lets you return to a location; I couldn't get the QT controller out of the way except by going to full screen (authors made bottom buffer slightly too small).

[A-Team]

Below are 4 frames from the second MacGregor 2015 paper, link above. This shows Greenland from a southern perspective with a moving vertical face -- colored to show the ice's age, made by progressively removing the surface and gradually exposing bedrock.

The ice-age color scale looks logarithmic but isn't: it is is piece-wise linear within the three different periods. The vertical exaggeration is 100:1. The top surface is textured by a MODIS mosaic (Haran 201). This mouses up nicely without the movie controller (ie it's a smooth pan over the fixed bedrock, which could potentially be removed to just show the east-west ice face sections).

The upper left shows older ice near Petermann is almost all gone despite the far northern latitude; the upper right suggests that a sub-optimal (ie wrong) location was picked for the NEEM drill site; the lower left suggests that the summit ridge location for NGRIP wasn't the best either; the lower right shows a pocket of pre-Eemian ice is found surprisingly far to the south, below GISP2 and Summit.

[Martin Gisser]

Indeed it is the mid-depth isochrons, in conjunction with ice thickness overhead, that are best suited to refining bedrock topography by providing intermediate caps. In this view, the unusual data structure of Greenland calls for something different than 'ordinary kriging', more along the lines of incorporating polynomial trends as in 'universal kriging'.

Oh, kriging! That's another of millions of stuff I'd love to learn... Methinks it's not so bad as you put it. It's even spline-ish, with the additional benefit of estimating errors, cf. N.Cressie, "Statistics for spatial data" (xerox pdf online), who likes it. Plus, I guess you can do the universal kriging several times and adjust the parametrized trend (e.g. correcting for the Peterman canyon). And my hunch is, with the Bayesian and Gaussian ansatz one can do some very slick gerenalizations of the method, like, "tensor field kriging", but now I'm halluzinating...

http://www.jstor.org/discover/10.2307/2290837?sid=21105134201141
Journal of the American Statistical Association 1994
Kriging and Splines: An Empirical Comparison of Their Predictive Performance in Some Applications
Abstract:
(...) (...) There has been some debate as to which of kriging and splines is better--a debate that has centered largely on operational issues, because the two methods are based on different models for the process. In this article the debate is turned to where it ultimately matters--namely, the precision of prediction based on real data. By dividing data sets into modeling sets and prediction sets, the two methods may be compared. In the cases examined, kriging sometimes outperforms splines by a considerable margin, and it never performs worse than splines. Various configurations of data show that the sampling regime determines when kriging will outperform splines.

[A-Team]

The press conference and related stories for MacGregor 2015 should appear today at 9:00 am Eastern -- they got sandbagged yesterday by the melt lake collapse stories in SE Greenland and the far NE Flade Isblink.

I'm now recommending not to download those big raw data files until NSIDC is ready to host them at their urls. They are still undergoing revision -- best if everyone works off the same improved data. I'm hoping those will have the proprietary matlab format issue resolved as well.

Meanwhile, I took a look at 3rd animation of MacGregor 2015 which is hosted at the journal as open source, final version. This is my favorite of the three, offering very fine-grained stepping through the east-west slices. This is really three coupled animations running simultaneously: predictions of age-depth from an update of Nye theory, observed from radar striations, and observed vs theory.

Below, I cut out five frames out of the several hundred in 'observed vs theory' near our favorite ice streams and summit cores to give a sense of how utterly inadequate even modified Nye modelling is (red regions in the vertical slices), not that this inconvenient truth will stop the flow of those papers.

Nye was an contemporary of Glen and is (rarely in Greenland glaciology papers) co-credited with the Glen-Nye flow law that is the basis of glacier dynamic modelling. The 1950's papers are, amazingly, online. It's fair to say that there is nothing in them physics-wise that could not be found in a late 19th century continuum mechanics textbook but that had to be adapted to the unique properties of glacial ice under deformational stress.

http://en.wikipedia.org/wiki/Ice-sheet_dynamics
http://go.owu.edu/~chjackso/Climate/papers/Glen_1958_The%20flow%20law%20of%20ice.pdf
http://www.researchgate.net/publication/224962689_The_mechanics_of_glacier_flow/links/004635243f41372c4e000000.pdf

[A-Team]

Oh, kriging! it's maybe not so bad as you put it.

I'm still looking into exactly what gets done as a practical matter in the most basic krige, then get to what exactly was done in Bamber 2013 and what kept the Pleiades supercomputer so busy in MacGregor 2015.

It seems necessary to distinguish between how a single point is interpolated from the initial data set (by IDW, inverse data weighting -- known data contributes inversely according to its distance, inverse power variable) from embedding the data in a sweeping statistical model universe in which, if the assumptions are met, gives not only the same interpolation value but an estimate of the error wanted in science (but mostly seen in model 'science').

In other words, you get an identical value for each interpolated point by IDW as by ordinary kriging. It's just that IDW doesn't produce an error map. The question is, if the assumptions of kriging are not valid, what use is its error calculation?

For those blank 10,000 km² polygons in NE Greenland -- given a fine-grained scale of variation nearby at a much smaller scale -- it is living in a fool's paradise to think that physically uninformed numerology will somehow detect what is under all that ice.

On the western flank of Greenland, I don't need splines or kriging, just a cylindrical sheet. This is not just a splines vs kriging debate, this is kriging vs smooth hilbert spaces.

From my perspective, glaciologists are struggling to get away from gridded sections into the underlying platonic surfaces that they are forever vertically sectioning and pixelating -- like the seven wise men looking at the elephant from different perspectives but never putting it together.

Analytically described iso-surfaces, their curvatures, and simple bedrock topography driven dimple operaters acting on them can provide a vastly more efficient and insightful description of the data -- and with it greatly improved computational ice dynamics. That is, do the heavy lifting in math, leave numerical approximation to the very end.

Ultimately the issue with kriging is falsifiability of its predictions. Suppose someone interpolates radar section data and provides a product. Suppose further additional radar flights, designed to cut the biggest empty polygons in half, take place in 2015. This new data resolves the very weakest part of the interpolation: centers of empty polygons that were farthest removed from actual data.

Comparing now interpolation to fact, under what circumstances do we say the predictions of kriging failed? Looking at the colossal uncertainties in the kriging bedrock error map, i don't foresee any such circumstances. They'll point at statistics theorems that assert it has to work -- optimally. Maybe so, given a large population of ice sheets; here we just have Greenland and its idiosyncratic history.

We don't actually need to wait for more radar flights to compare interpolative predictions of methods A, B and C. There is a very favorable situation in the Petermann grid where a 2013 flight, which came too late for inclusion in kriging, cut diagonally across a square grid cell. I'll provide the grid cell data shortly but hold back the diagonal data. Then we can put kriging to the test.

[A-Team]

After looking at several kazillion Greenland radargrams, it becomes apparent that the englacial isochron story -- like
Shakespeare in a nutshell at Cliff Notes -- can be simplified to a few key stratigraphic surfaces out of the hundreds visible when image resolution limits are really pushed in years of favorable radar design.

The six surfaces that seem most instructive are at intermediate depth and easily recognized over almost the whole of Greenland. I call them the Lower Lid, the Three Sisters and the Blurry Brothers. They are very useful as markers for upward inelastic deformations (very rarely breached). The latter amount to the oldest of the easily tracked layers (ie provide a consistent cap over Eemian upheavals.

These surface presumably correspond to layers of volcanic ash laid down over the then-exposed surface of Greenland by massive stratovolcanos. These may have induced climatic hiccups at the time but do not represent major climate transitions.

However we can hope that dating in MacGregor 2015 can accommodate all the major northern hemisphere events whether or not they left a mark in the isochrons and ice core layers. Events such as Bond B6, Younger Dryas, Dansgaard-Oeschger, Older Dryas, Dansgaard-Oeschger, Heinrich H1 etc that I tabulated provisionally earlier but now by SO Rasmussen et al Dec 2014 at 2nd link below.

http://forum.arctic-sea-ice.net/index.php/topic,984.msg36237.html#msg36237
http://tinyurl.com/lsn5hkk Rasmussen 2014
http://tinyurl.com/opwrmw4 Seierstad 2014
http://tinyurl.com/k46z598 Blockley 2014
http://www.iceandclimate.nbi.ku.dk/data Data Archive

In very favorable circumstances, starting from a carefully reprocessed image from MacGregor 2015 posted at NASA SVS/Goddard  and piling on additional focusing and contrast enhancements, it can be seen that the Three Sisters and the Blurry Brothers isochrons actually harbor numerous other layers that would not be readily visible in 99% of the stock Cresis imagery. However these are strictly conformal and thus susceptible to later infill.

Thus for the pre-Holocene ice, it seems that almost all the low hanging fruit could be picked from a handful of isochrons. This may be enough to fix Greenland ice sheet modelling, though note the contorted ice below the older Blurry Brother remains very poorly understood (even as it controls basal resistance to flow).

Note in this image (whose origin and scale aren't provided), it appears that the ice flow is (more or less) left to right with hold-up of flow by the basal obstruction increasing linearly with depth (dotted orange lines). Note too how the curvatures (reciprocal radii of nesting circles) vary systematically with depth and capture, along with the center line angles and layer thinning, the effect of the basal deformation on the ice layers.

Thus a very few geometric parameters suffice to describe this radargram completely. It does not take much more to extend this from 2D vertical sections to the larger isosurfaces being sampled by the flight lines.

Ice this deep being incompressible, the ice between any two lines (their age-volume) is presumably conserved. That means it is squished out to the two sides, possibly more so to the stoss side than the lee. In many cases, these internal deformations have no overt manifestation on the ice surface, yet they must unless the surface rise is overwritten by new accumulation of snow.

At some sites drillers encountered brittle ice that made for difficult core recovery, I've not seen any brittle fracture (discontinuities in radar isochrons) even over the most extreme deformations, suggesting a process slow with respect  rates of ice ductile adaptation.

Probably the most puzzling feature of isochrons over deformations is the lack of notable reflector diminution. At the time of tephra deposition, a fixed amount of salt (~ permittivity) was deposited, creating the reflector as it became buried. Given the mass conservation above, it seems that squeezing ice off to the sides at the top of the deflector would lower the constituents of conductivity to the point it seriously weaken the radar echo. The molar amounts would be reduced but not the molarity and so not the permittivity contrast.

There's a whole lot of ice sheet history in these isochrons -- they leverage the six isolated cores out to all of Greenland. It is not worth theorizing about future Greenland melt until existing experimental data has been subsumed into understanding the past. Models to date have not made any real use of the 21 years of radar data -- but that has to change.

[A-Team]

Below is a first effort at unwinding the history of the Greenland ice sheet ... still needs some tweaks to maintain age volume and retain ellipsoid level, the idea here being to model the basal upheaval life cycle and its effects overhead.

The second image is made using the FeatureJ suite in Fiji. The yellow arrows point to some briefly trackable isochrons in EFZc, the red arrows point to dark center lines in the Three Sisters that define unwarping functions for G'MIC online, and the green arrows show that the deepest ice still has trackable horizons. Whether these are conventional ice core linked isochrons or from a different time system intrinsic to basal freezeups is not yet resolved.

[GeoffBeacon]

A-team

These recent posts of yours really impress me.

However, you follow one thread of scientific methodology in saying

Ultimately the issue with kriging is falsifiability of its predictions

Surely any comparison of modeling and measurement will not be true or false but will vary over a continuum from "nearly spot on" to "hopeless". Policy makers need "best we can get" with some estimate of reliability.  I see you are waiting for improved data. I hope soon we can know your expert judgment on the story as reported in, say, Time magazine "Two Massive Lakes Under the Greenland Ice Sheet Drained Away in Weeks". The article's headline:

The discovery signals a "catastrophic" environmental shift

http://time.com/3677814/greenland-ice-sheet-lakes-drainage/

P.S. This is not intended to promote an off-topic discussion of scientific methodology on this thread.  It's just a plea for help.

[A-Team]

Geoff, good questions. The subglacial lakes are near-surface, so not much off-topic for the subglacial forum (though we have a full plate already with bedrock and radar isochrons, ie deep englacial) but the discussion of them so far has been mostly over at 'What's New in Greenland'. If someone would like to pursue the two dozen or so melt papers of 2014-25, probably best to put it all in a dedicated forum.

I would suggest Time is on the right side of the issue but is somewhat ahead of developments. There is definitely a concern that too-rapid climate warming will bring too much meltwater too soon -- and with it latent heat, greater susceptibility of warmer ice to stress, and bedrock lubrication -- into the glacier faster than it can be safely dissipated.

Whether that will bring catastrophic effects such as jökulhlaups on a <0.8% slope remains to be seen. However these are positive feedbacks that certainly must be factored in to modelling Greenland's accelerating contribution to sea level rise. I suspect they haven't been. The same can be said for the condition of basal Eemian ice or the 100 kyr of historic constraints emerging from MacGregor 2012.

I've been moving forward on developing a tangible example that will keep us focused on measured vs interpolated, see below. Petermann is the most favorable case to discuss because a very tight rectilinear grid was flown.

In a nutshell, suppose you have two opposite walls of a grid cell that both intersect the Grand Canyon (framed in blue below, 2nd image). Because the intersections are oblique with respect to the bottom of the canyon, it will appear at very different places on the two radar transects. Meanwhile, the other two opposite walls may miss altogether and not even be aware of the canyon.

The continuity of the canyon (and monotonicity of its depth profile) is a self-evident proposition to us but not to interpolative software. We would connect up the canyon's appearance on opposite walls with a diagonal bottom that matches up at opposite walls.

The software won't, it is simply using local data de-weighted by its distance. Consequently it will give two dimpled depressions (basins) near where the edges intersect the canyon. Water won't flow over the hump dividing the two basins.

In fact you can see that very clearly in Bamber 2013 -- patches of color that are too high (3rd image below). In this situation, ordinary kriging under-performs what someone would infill using common sense and putting in elevation with a Photoshop brush. We'll see this confirmed by real data (held-back purple flight line).

Islandwide, I am proposing that mid-level radar isochrons can be more reliably interpolated than bedrock, indeed isochronals surface curvatures arising flow deformation over bedrock can give us better infill that bedrock interpolation alone.

[A-Team]

Here is one of the Petermann grid cells cut open. The left edge matches the right. Interpolation seeks to fill the interior at every level, from bedrock to surface including all the radar stratigraphy.

The bottom image wraps up this Petermann flight line polygon and provides a view from the northeast corner. Here the Holocene has been removed; the orange trace represents the Lid to the Three Sisters; the red trace what we know about the bedrock.

[A-Team]

To really predict anything about Greenland's near-term contribution to sea level rise, it is necessary to pile a climate model on top of the ice sheet model, which itself combines models for firn, meltwater response, internal temperature and strain rate, paleo ice sheet response to warming, and bedrock attributes (topography, pressure melt, geothermal gradient, till)l.

Naturally the more direct experimental input and the less ungrounded interpolation and speculation about the ingredients, the better the output. Ironically, despite progress on all fronts, the best scientific judgement of 2007 (rejected by a secondary IPCC panel) looks very much like that of today. However there's been one surprise after the other in Greenland's response to warming, so  and that predictional stability could represent lucky cancellations of error rather than minor improvements converging to a final sea level rise estimate.

The ice-penetrating radar program was largely targeted at ice thickness and bedrock topography. Very little was done with the paleo-stratigraphy that it picked out as a byproduct. That all changed with MacGregor 2015.

Isochronal surfaces, unlike bedrock, are sensitive to the direction and speed of ice flow which are accurately known, and also to paleo warming epochs. They are also bedrock-aware in the sense of softened (fourier filtered) deflections over bedrock highs and lows.  These three advantages, in conjunction with comprehensive availability of traced isochronal lines provided in MacGregor 2015, open the door to improving several model ingredients above.

I'm looking here at the feasibility of improving the extraction of isochronal surfaces and from them, bedrock interpolation between the radar lines. This doesn't require any number crunching as it can all be done in graphics software.

It's quite feasible to determine all the intersections of all the radargrams in a spreadsheet, to pare the sides of these polygons down as shown below to an isochron capping bedrock with elevation shown by grayscale gradient, and then to extrude these islandwide, enabling a Google Earth 'street view' that can pan around inside the experimental data. Subsequent infill of our isochronal foliation is then guided by the three considerations above, plus some landscape logic for further improvement of bedrock.

[Lennart van der Linde]

Nice NASA-video on GIS in 3D:
https://www.youtube.com/watch?v=u0VbPE0TOtQ&feature=youtu.be

[A-Team]

Nice NASA-video on GIS in 3D

Right, I posted this at the time it came out last Friday. Speaking of the pre-Eemian ice whose thickness is mapped in Fig.10 of MacGregor 2015, it suggests first of all that NEEM was a very unfortunate choice of drill site being way too low on the wrong fork of the summit ridge.

Below I propose two plausibly better drill sites -- if the goal is obtaining a longer climate record for the Northern Hemisphere from undisturbed ice with the thickest possible annual layers at depth.

The first emphasizes the thickest available ice below the 115 kyr isochron (reddish in Fig.10) that lies at an orthogonal intersect of two recent radar overflights. The latter provides assurance that the bedrock is well-behaved in 2D (ie not drilling onto a steep-walled cliff an earlier site).

The recommended lat,lon for a new Greenland deep drill core site is 76.777, 43.867 which is located where 20130419_01_056 and 20110329_01_028 meet at a surface elevation of 2894 m. The second image shows those transects with contrast enhancement -- note the isochrons are quite flat and the bedrock is featureless in both.

I'll be curious to see in the data files of MacGregor 2015 whether they found a way of locating very old isochrons in these, as the wide variety of enhancements I tried did not pull out any overt lines in the deepest ice (which is the norm, too little starting contrast).

More adventurous drillers might consider the bottom of the Grand Canyon (at the 20130419_01_056 summit crossing). The animation below shows various enhancements of the radargrams there. It would be feasible to drill to the canyon  bottom (recalling that real drill holes are generally off-vertical), with the advantage that the till and streambed rock itself can now be cored with newer ice drills.

The canyon at this latitude (76.600º) is some 6.6 km rim to rim with a depth of only 228m (ie, it's a broad valley without the slightest resemblance to any transect of the real Grand Canyon). Still, the Greenland feature is deep and steep relative to the prevailing profile of interior bedrock.

It's not so clear whether layering of pre-Eemian ice in the very lowest layers of the canyon would be protected (or  worsened) by ice flow overhead, depending on its orientation across or along the canyon's direction. The bottom ice may have the thickest annual layers yet the added depth in places could bring it below the pressure melting point.

[Lennart van der Linde]

Right, I posted this at the time it came out last Friday.

Ah, I missed that one in the What's New folder. Hard to stay up to date!

[A-Team]

Nice NASA Visualization Lab video on englacial GrIS in 3D based on MacGregor 2015 data

A little cross-posting never hurt. As the posts pile up, I can hardly recall myself where I put what, having made ~ 45 posts on this one paper alone (and its AGU2014 predecessors). I have not seen any substantial coverage at any of the "other" climate science forums; they are mostly caught up in the meltlake melodrama.

However it will scarcely be possible to publish a paper in 2015 on Greenland climate change without citing this paper and incorporating its data.

At first it seems that surface melt (funneled down through moulins to bedrock and channeled down to sea) is little concerned with englacial stratigraphy. However it will prove practically impossible to physically model the future distribution of meltwater betweeen firn retention, laminar lubricating bedrock flow, and efficient channelized discharge without latent heat deposition.

It's far more instructive to skip the dodgy modelling and just look at the paleo response of Greenland to previous episodes of warming. That history, a book waiting to be read, provides the bottom line resulting from earlier rounds of global warming, some of which lasted longer, or ramped up faster, or reached greater extremes, though none of these can provide an exact replica of today as initial conditions won't match.

Consequently we need to keep up with climatic geochronology, the ice core ¹⁵N temperature revision and englacial memory, as we're very fortunate to have major new papers on all three.

Something went very very wrong in setting Greenland scientific priorities starting in the late 1990's. I've been trying to trace that down and -- since the same people are largely in charge today -- verify that research has a balanced trajectory today (after 15 years of https://en.wikipedia.org/wiki/Rip_Van_Winkle).

I've attached a map of 1999 radar coverage. You can see the flight plan makes better sampling sense than many subsequent years, that it provides better internal stratigraphic contrast than many of later designs, that the oldest 500 m of ice can be seen to have a very squishy interface with bedrock, that the huge upheavals below Petermann had already been detected, and the Grand Canyon was already in clear view.

The radargram below was enhanced in ImageJ2 using FeatureJ vertical differentiation and 'glow' color-indexing (a look-up table in the Image menu). The second image shows the U Kansas Cresis program provided very extensive coverage just in 1999, not to mention the contributions of earlier years.

Over 27 track segments captured profiles of the upper Grand Canyon in 1999 alone. Others revealed thick undisturbed layers of pre-Eemian ice, crossing right between the two proposed drill sites described in the previous post. In the case of 19990518_01_003 below, that is approximately at the 76.650º, 44.218º marking.

[A-Team]

Glaciologists have long wondered whether the position of Greenland's summit ridge has always been where it is today, if  orographic effects on weather have changed, and how persistent the current pattern of annual accummulation (markedly less for the Zachariae snowshed) has been.

http://bprc.osu.edu/wiki/Greenland_Accumulation_Grids

Can paleo summit position be extracted from englacial stratigraphy? Ice sheets have very long memories -- if say the paleo summit ridge was further west, there should be a discrepancy between the peak of the older stratigraphic lines and a more easterly peak of surface ice today.

Even if we could map the summit ridge in space and time, these would not necessarily be of cosmic significance. Still, every bit is important if we are not to draw the wrong lessons from reconstructing the ice sheet history under past climatic regimes.

Greenland barely has a 1º slope crossing west to east so this mismatch would be very subtle, but perhaps that could be drawn out by extreme horizontal compression of radar imagery crossing the summit ridge. Operationally this just means re-sizing an image with uncoupled width and height scale changes. This greatly exaggerates the curvatures of the isochronal horizons.

While I see some signs of this, the isochrons can also be deflected by troughs and hills in bedrock topography (or bottom freeze-on upheavals). These can have a horizontal component due to flow (see angle lines in a post above) but near the slow flowing summit, the deflection is almost entirely vertical.

The Grand Canyon passes under the summit ridge (coincidence or consequential?) for ~ 150 km between NEEM and NGRIP, providing a serious subsidence of horizons above. The animation below floats the bedrock layer upwards into a grieviously squeezed radargram to facility comparison to higher layers.

Rather than draw a line representing bedrock (which involves an element of subjectivity), that portion of the radargram can be lifted (in Gimp) by a bound box from which lighter colored pixels are removed by the contiguous color picker tool, whose selection can be nicely augmented by the 'grow' commend in the Selection menu.

[A-Team]

Here is a virtual flight up Greenland's Grand Canyon starting below Petermann.

The 18 transects amount to about 60% of what is available down to midway between NEEM and NGRIP; I may add more of them later. Each image has to be individually rescaled in both horizontal and vertical directions. They are not optimized for contrast here but could be.

The gray cross-hairs are centered over the deepest part of the canyon and mark 2000 m ice thickness. I may add sea level, lat,lon and distances between frames later, along with stream gradient. Note that most transects will not be orthogonal to the course of the canyon but rather oblique. The Cresis frame identifiers have also been supressed here to reduce clutter.

The horizontal bar represents 10 km; the vertical 1 km (ie the vertical exaggeration is 10x). At 1:1 scale, the profile of the canyon is not at all dramatic (2nd image shows a frame at other scales). For comparison, see a real Grand Canyon profile at Lava Falls (western end) at 1:1 scale. The Greenland feature is not a canyon as that term is used in geomorphology; however it does stand out as a remarkable depression relative to the overall extreme flatness of interior Greenland bedrock.

The collected transect data is the totality of what we currently know about this canyon. Kriging can manufacture additional virtual transects via interpolation but these do not provide any additional real data; indeed they introduce artifactual closed basins as explained above and could not conceivably predict canyon wall profile details.

Thus it's better to stick to the data in searching for clues to how and when the canyon valley formed -- such as the shape and asymmetry of the walls, especially on outside curves. One of the animation frames shows what must be an island (or a mound left un-eroded as a paleo river changed channels).

It's not clear if the canyon has any significance to ice sheet flow today. While it seemly provides a natural conduit for glacier flow, its alignment is not particularly correlated with surface velocity. The isochrons are fairly undisturbed over the canyon, exhibiting moderate and variable dip but in general the deflection expected over any basal trough or hummock.

Note even where paleo ice flow has passed across (not along) the canyon, the ice would dip rather than shear as it passed overhead, meaning only the deepest pre-Eemian ice in the canyon bottom could be protected from deformation, ie not represent stagnant ice.

The blogware won't animate the animation, it is about 1.8 MB. This link shows it properly: http://tinyurl.com/kla7rp8. I've added an all-in-one composite of 18 canyon transects. They are ordered north to south with column one near Petermann and column three on the summit ridge midway between NGRIP and NEEM.

[A-Team]

Here is a great project for a bright undergraduate or citizen scientist wanting to contribute something to our knowledge of Greenland. It requires nothing more than an internet connection and familiarity with picture software.

I've written out the steps for this in terms of a follow-along example. These just need to be repeated on ~40 other examples and pulled together. The results can then enhanced with ice layer dates and depths using the open database compiled by MacGregor 2015. See also somewhat different 'glow' enhancement of this same example in #68 above.

Finding flight segments that cross the Canyon:

* Go to the U Kansas storage repository https://data.cresis.ku.edu/data/rds/

* Open Google Earth (freeware) and add the locations of the main ice cores:

NEEM	77.450°	-51.060°
CCentury	77.167°	-61.133°
NGRIP	75.100°	-42.320°
GRIP	72.587°	-37.642°
Summit	72.582°	-38.490°
Dye3	65.183°	-43.817°

* Click to load the Cresus kml track files for 2011radar data
  https://data.cresis.ku.edu/data/rds/2011_Greenland_P3/kml_good/Browse_Data_20110507_01.kml
  https://data.cresis.ku.edu/data/rds/2011_Greenland_P3/kml_good/Browse_Data_20110429_01.kml

* Add the 1999 kml files 19990507_01, 511_01, 512_01, 513_01, 514_01, 517_01, 519_01, 525_01

* Open all tracks and zoom into to northern Greenland

* The lower Canyon crosses the north and east edges of the closely spaced grid just below Petermann Glacier. 

* Select a flight segment, open the corresponding pdf at U Kansas and identify the frame that cuts across the Canyon.

* Save the _1echo.jpg of this frame for processing.

Processing the image:

* Follow along as we process a frame crossing the summit ridge in central Greenland, namely 19990518_01_003.
    https://data.cresis.ku.edu/data/rds/1999_Greenland_P3/images/19990518_01/19990518_01_003_1echo.jpg

* The horizontal resolution: 1165 pixels to display 150.91 km, or 7.7198 pixels per km.

* The vertical resolution: 721 pixels to display 3000 m of depth, or 240.33 pixels per km.

* The vertical scale exaggeration is thus the ratio of vertical to horizontal, or 31.132 (greatly distoring the Canyon).

* A good choice of re-scaling: width normalized to 50 km and 5:1 vertical exaggeration. Because the original image is marked of in fifths (for lat,lon and km), that scale will transfer as 10 km units.

* But first, note the poor contrast at depth (too white). Adjusting the contrast now (to enhance the pre-Holocene isochrons) will preserve features after re-sizing.

* Now isolate the region of interest, capturing the frame and vertical and horizontal scales (two markers) by copy-paste-move. The blog is limited to 700 pixel width,  so here a region of 232 pixels will expands to 700 after horizontal rescaling.

* Rescale image width:height first by 150.91/50 = 3.0182:1, then rescale vertically by 1:0.1606 = 5/31.132, best done in one step as 3.0182:0.1606 using bicubic  interpolation

* To keep order, have latitudes decrease across the image left to right (ie north to south); otherwise flip the image horizontally fixing text to retain readability.

* Flight segments known to cross the Canyon include:

19990511_01_004   20110502_01_012 19990511_01_013   20110502_02_024 19990511_01_014   20110507_01_013 19990511_01_015   20110507_01_014 19990514_01_006   20110507_01_015 19990514_01_023   20110507_01_022 19990519_01_022   20110507_01_023 20100324_01_012   20110507_01_030 20110429_01_012   20110507_01_031 20110429_01_013   20110507_01_038 20110429_01_020   20110507_02_001 20110429_01_021   20110507_02_018 20110429_01_028   20110507_02_018 20110429_01_032   20120508_07_002 20110429_01_033   20120514_01_020 20110429_02_004   20130419_01_016

Main Results:

* Determine the width of the Canyon to the side-spillovers (top of first wall encountered) using horizontal dimensions of a selection box. In the example these are 126 x 92 pixels which converts (since 232 pixels~ 10 km) to 5.43 km left of center and 3.97 right of center for a wall-tp-wall width of 9.4 km.

* Determine the depth of the Canyon using the vertical dimension of the same selection box. Here, these are 59 x 45 pixels which convert using the vertical scale to 509 m and  387 m, with the average 448 m being representative.

* Determine the lat,lon of the lowest point by proportionality to pixel scales. Here the latitude ranges from 76.709 to 76.650 over 742 pixels so .059º/742 degrees per pixel. The center of the trough is 306 pixels in from the left, so its latitude is 76.685. A similar calculation for longitude results in the lat,lon of the canyon bottom here as 76.685, 44.833 which can be entered as a way point in a Google Earth line tracing out the canyon bottom over the flight lines.

* Determine the lat,lons demarcating the Canyon bounds and enter as a Google Earth line segment. Unlike the low point of the canyon, here the (to be determined later) angle the flight line makes with the Canyon axis matters.

* Add sea level, bedrock depth of NGRIP, and dates of the radar reflection horizons, deferring to MacGregor 2015 on these, though note only a few distinct horizons can be seen on this image in addition to the informally named Holocene Lower Lid, Three Sisters, and Blurry Brothers that are seen throughout north-central Greenland.

* Add a surface velocity vector (ice flow) over a compass to the image at the lat,lon of the Canyon low point. The first issue here is the Google Earth projection, which has mouse-over lat, lon, is not so easy to overlay on ice velocity polar stereographic projection, which does not. The second issue is extreme variability of ice velocity, from 1 km/yr to 18,000. This might best be handled with concentric circles, initially linear for interior Greenland (2000 m contour) but then logarithmic, as indicated on the graphic below.

* After all the Canyon overflights have been processed, the profiles can be connected up in succession by piecewise linear interpolation, splines or kriging. Widths can now be corrected where the flight line is oblique to Canyon direction. The downhill gradient amounts to differences in consecutive bottom elevations, though this represents today's geometry, not pre-isostasic conditions when water last flowed down the hypothesized paleo river. Note a river would differentially erode the inside bends of a curve, making the wall gradient steeper, proportionally to paleo gradient.

[A-Team]

One of the most instructive flights of the Cresis ice penetrating radar took place 15 years ago, 19990514_01. That went from Camp Century to the summit, passing over a much diminished Canyon in 19990514_01_06, then circled around and came back north passing over GRIP and NGRIP before transecting the Canyon again at 19990514_01_23 on its way back to Thule.

A later flight 20110506_01 connected up Camp Century, NEEM and NGRIP. It intersected 19990514_01 along the summit ridge. These two flights allow borehole data such as isochron horizon dates, pressure melting points, and bedrock depth below sea level to be transferred at intersections to 19990514_01 which initially only had Camp Century.

A second round of chaining allows almost any flight segment in northern Greenland to be tied into the six main ice cores. The Google Earth screenshot below shows how simple it is to keep track of all this information -- the pushpins show the location of the Canyon bottom at the two transects mentioned above.

The radar view of the Canyon at 19990514_01_23, shown at ~30 x vertical exaggeration below, is worth a close look for the odd behavior of the Three Sisters which are weakly deflected in some places in fading correspondence with bottom topography and elsewhere by dramatic basal upheavals (which are not near enough to the Canyon for an association), yet in between these isochrons seem to hit bedrock.

In the situation that multiple other overflights meet 19990514_01 at oblique angles where it transects the Canyon, the opportunity arises for effective tomographic local 3D reconstruction. Here it is useful to show the span of the Canyon, not just the deepest point. The graphic shows two other flight segments (19990514_01_004 and 19990514_01_014) intersect the span of 19990514_01_23.

[Laurent]

Already posted the video of the latest NASA video
Here’s What the Inside of Greenland Looks Like, in 3D
http://www.climatecentral.org/news/heres-what-the-inside-of-greenland-looks-like-in-3d-18630

[A-Team]

Here’s What the Inside of Greenland Looks Like

Dr. MacGregor kindly sent me these videos two weeks ago and I posted them at that time to the Greenland news forum, along with analysis and commentary. However there is considerable news value in story quotes from Joaghin and MacGregor collected by the reporter, so good catch.

Joaghin said he had not gotten into the data yet which is understandable -- it is massive and complex to evaluate, and subject to near-term refinement.

After 21 years of half-åssed efforts at doing something with the radar data, I say three cheers for MacGregor's team for even embarking on such a risky and laborious project:

“We did not envision being able to do this at the scale that we did,” he said. But once “it occurred to me that I could do it . . . I kinda had to try.”

MacGregor said that claim for very extensive Eemian ice was “bound to be one of the more controversial aspects of our work” because they had to extrapolate more there and make educated guesses on ice ages based on the layers around it. 

This is the part that I'm curious about. As you know, Iceland has been showering Greenland with volcanic ash since like forever, leaving hundreds of island-wide conductive annual layers in the six cores and corresponding radar-reflective horizons.

However, there are three very puzzling echo free zones (EFZs) that each span many millennia, yet according to three major new consensus geochronology papers on Greenland tephra (which have fallen on deaf ears at Greenland forums), these were by no means quiescent periods for Iceland.

The EFZs may have different explanations. The upper two do show reflector substructure in favorable regions such as 19990514_01_023_1echo.jpg. The question really is why these details seem lost elsewhere.

 The lowest EFZ faces the geothermal gradient, deformation by basal freeze-up, and strain shearing from ice sheet slumping. This may have churned the permittivity to the point that good reflectors became too diluted -- except at precious flight lines over the rare undisturbed relic ice.

MacGregor 2015 could have used these where available, but since they mostly weren't, relied instead on always-available local age-depth curves which can be extended to explain (date) the ice below the last datable isochron. That's reasonable enough given little evidence for ice accumulation anomalies.

The prevalence of these faint oldest isochrons depends on the sophistication used in radargram enhancement, ie something more than slope-driven noise reduction and adaptive contrast normalization. My sense is a bit more can and will shortly be done before hitting the wall (signal:noise). However if a horizon is just not there, there's no way it can be brought out.

I'm attaching a small piece of a very unusual radargram, namely 19990511_01_014, that appears to have 2-3 traceable isochrons below the oldest common horizons (the Blurry Brothers). This flight segment is also unusual for slicing high across the west sidewall of the Grand Canyon, then a few km later tracking down the long axis of the Canyon for ~50 km.

[A-Team]

To further help our imaginary bright undergraduate pursue the so-far neglected experimental description of Greenland's Grand Canyon and Petermann Upheaval Grid, I provide below a plug 'n' play spreadsheet that magically assembles all the overflight intersections making up the corners of the 400-odd relevant polygonal cells.

The one slight technical hassle here is that as the plane flies the grid, it continues to collect data around the curves. Thus it won't work to use the start-stop lat,lon provided on Cresis images in a formula for the intersection of two straight lines.

However the Cresis kml files contain many internal waypoints in (lon,lat,elevation) format that you can view by saving from the browser and opening in a plain text editor or excel. Only five of the ~40 are shown on the radargram and those with the last 3 decimals and elevation discarded. These provide a piecewise linear description of the curved flight.

Because the plane turns with a fixed wing angle, its path has constant curvature meaning these segments don't just look like circular arcs, they are circular arcs. Consecutive waypoints then describe chords upon which the radargram arc can be radially projected.

<name>19990507_01_017</name>
         <LineString>
            <coordinates>
            -53.799259,78.234722,2328
            -53.921063,78.219810,2216
            -54.043893,78.204753,2217
            -54.167942,78.189534,2183
            -54.293387,78.173887,2147
            -54.420392,78.158031,2202
            ...
            </coordinates>
         </LineString>

Most of these short line segments don't intersect anything but it involves a substantial computation to find those that do on an all-Greenland, all-flights scale. The graphic below shows the small scale machinery needed to locate the Grand Canyon on three flight segments intersecting in a triangular polygon southeast of Petermann.

Knowing the flight line intersections amounts to re-casting kml waypoint data into polygonal vertices and edges, which has the effect of slicing the radargrams into their component faces. It is only at these intersections where we have any prospects for tomographic reconstruction.

The bottom line, after rescaling the radar imagery according to orientation and vertical and horizontal scales, is how many pixels over from the left define the intersection lines. Auxilliary metadata is carried as a secondary transparent layer co-registered to the radargram but not interfering with its enhancement. It is not rocket science to script the meta layer image using BMP format.

In the example, the third radargram intersects the Canyon swaths of the other two but high on the walls making it difficult to see its Canyon data.

As noted earlier, it takes some major mickey mouse to find the intersection of two lines whose equations are not known but given two lat,lon points on each. None of the many online calculators, including Wolfram, can do this; wikipedia gives the formula but inconveniently as a graphic which I re-cast as text below:

Find intersection of two lines where:
Line A is defined by (LAT1, LON1) and  (LAT2, LON2)
Line B is defined by (LAT3, LON3) and  (LAT4, LON4)

The point of intersection (LAT, LON) is given by
LAT = (lat1lon2-lon1lat2)(lat3-lat4)-(lat1-lat2)(lat3lon4-lon3lat4)/(lat1-lat2)(lon3-lon4) -(lon1-lon2)(lat3-lat4)
LON = (lat1lon2-lon1lat2)(lon3-lon4)-(lon1-lon2)(lat3lon4-lon3lat4)/(lat1-lat2)(lon3-lon4) -(lon1-lon2)(lat3-lat4)

In spreadsheet cell terms:
LAT = ((B2*C3-C2*B3)*(B4-B5)-(B1-B2)*(B4*C5-C4*B5))/((B1-B2)*(C4-C5)-(C1-C2)*(B4-B5))
LON = (B1*C2-C1*B2)*(C4-C5)-C1-C2)*(B4*C5-C4*B5)/(B1-B2)*(C4-C5)-(C1-C2)*(B4-B5)

[A-Team]

After listening to so many Greenland glaciologists bitch and moan about how so much radar data accrued that, like, how could anyone expect them to analyze it (despite being paid to do just that), I am about ready to radio-collar some of them to track how their days are spent.

Actually I already know where the time goes: teaching, grading, office hours, grant writing, paper writing, answering email, parking lot committee, ppt presentations, meeting travel, mowing the lawn, appeasing spouse and kids -- research is the last thing anyone in academia has time for.

Better just to point a model algorithm at a stack of assumptions and let an unattended computer grind through parameter space and spew out the next paper. It's hardly going to be rejected in peer review when the reviewers are doing the same thing themselves. Even though they all knew this passage from Cuffey & Paterson:

Events in the past decade gave a harsh reminder of the need to maintain skepticism about models; ice sheet models did not predict the dramatic recent increase of ice outflow from either Greenland or Antarctica.

I'm not seeing a whole lot of testable predictions (ie 2015-2025) being published today. The weather is blamed; climate change is 2115, four generations off and the researcher long gone, no feedback there to distinguish a brilliant from botched model. Yet what does weather have to do with anything above the Parca flux gate?

Models became the tail wagging the dog in terms of setting the agenda. They didn't call for englacial depth-date isosurfaces or melt/accumulation history of the Greenland ice sheet so those fell below the radar screen so to speak, whereas a model already has fatal shortcomings if it doesn't begin by encompassing all the experimental data). 

By fall of 1999, there was sufficient data -- but not yet an overwhelming amount -- to make a decent pass at englacial Greenland. Indeed the bedrock topography was first interpolated in 2001. With 0.1 FTE of grad student time, isochron analysis too could have been updated each season.

Bedrock is wanted by the models but interior Greenland is so very flat at 1:1 scale that bumps in terrain have minor effects compared to dilantant till, basal temperate ice and geothermal flux, which remain almost totally unknown. Without these, bedrock depth alone is just a trip to DisneyWorld. The models have no room for that bottom 500 m of warmed up, churned up ice so it didn't happen (even though it's directly observable).

But it's going to get worse: it appears now that the englacial temperature profile has been determinable all along from the radar data.

So the best bet here may be going forward off paleo: the ice cores give us the actual local climate of the past (during which all manner of extreme climate change took place) to which the ice sheet responded in a determinable way. Provided someone has done something with the isochrons.

However a fair amount of present ice loss in both Antarctica and Greenland involves a handful of special cases. Sites like middle and lower Jakobshavn Isbrae don't have any paleo left to speak of, leaving us with Holocene moraines, troughs to the edge of continental shelf and profound overdeepenings.

Cuffey & Paterson are on this too, bashing the glaciology of

"general principles that by themselves do not answer the questions that matter. The fate of the ice sheets in a warming climate depends on the outflow along a few dozen outlet glaciers and ice streams, each one of which needs to be separately understood."

Glaciology has a reputation for being long on theory but short on data. That is changing due to open source satellite (most recently Grace, Sentinel and WorldView 0.6 m), comprehensive ice-penetrating radar and rapid drilling techniques.

The issue in Greenland though is not lack of data but rather under-analysis of data. Modelling under those circumstances will inevitably be sub-optimal. I'm not aware of scientists in other disciplines walking away from a poker game leaving so many chips on the table.

[A-Team]

Setting up to do the Petermann Grid ... it is quite remarkable in its central area but plain layer cake elsewhere. It takes 2-3 hours of work to set up the files, so despite the importance of this glacier it's no wonder it has not been done before  -- Greenland glaciology had only ~300 people x 2000 hrs/yr/person x 4 yrs of data availability = 2,400,000 hours.

To process this many files efficiently, a couple of Google Earth and Gimp tricks come into play: selectively open subsets of the Cresis-supplied pdfs in Gimp for pulling flight path segments and their images for bulk cropping; layers are readily re-ordered and grouped; filmstrip seamlessly tiles only the particular subset of layers checked (eg upheaval, canyon route, bedrock upwarping).

The animation is just a teaser, I will tie it into the grid tomorrow or so. The velocity overlay (thx to NASA Visualizations) of the grid is just heuristic for now -- not so easy to co-register an oblique view with unstated parameters with an overhead GoogE view from an unrecorded elevation. Next post will overlay Bell 2014 basal freeze-ups on this grid and provide all the intersection coords.

These Petermann radargrams record very unusual ice sheet behavior ... especially for the paleo climate of 80º north. Note first that these features -- and the isochron deformations above -- look less dramatic at !:1 scale. Second, west of the basal upheaval, a very stable trio of younger isochrons is dipping 700 m down to bedrock.

It's a waste of time to model the ice sheet's future in the northern third of Greenland until the relationship between paleo climate  and the ice physics of this region is understood.

[Lennart van der Linde]

Time for a publication?

And is that Eemian ice at the  bottom?

[A-Team]

Time for a publication? And is that Eemian ice at the bottom?

I had the same questions!

It's hard to say if an update to Bell 2014 is in the works somewhere by somebody, not seeing anything at AGU2014. That article was very coy ... no substantive supplemental data (eg listing of flight segments containing freeze-ups, no screenshot archive), a subjective visual survey with no definition of qualifying freeze-ups (eg, a cut-off for what isn't), only one rock check and that over on the northeast coast, so it all has to be re-done from scratch, just like the Bamber 2013 Grand Canyon.

MacGregor 2015 will have most of the older ice here dated as Eemian or pre-Eemian because the standard six-reflector pattern of mid-age ice sits a few hundred meters above relatively undisturbed older ice (between upheavals).

The upheavals themselves are 'explained' as freezing on of basal meltwater, migrating in from parts unknown, from an unspecified source, by one of several mechanisms, over an unspecified time frame. It's not known whether these are fossil features tied to some paleo climate event or are still growing (alt. dissipating) today; some may have associations with Petermann Glacier flow but others -- huge ones -- occur off the summit ridge in regions where the ice has hardly moved in 100 kyr.

The freeze-up ice is allochthonous (relocated material): if the Petermann Grid area were somehow were pressure-melted at the bottom or sitting over a geothermal flux excursion, what then motivates re-freezing? However other materials (till, soil, pine needles?) carried up with it from the bottom are likely autochthonous, possibly including an upper layer of displaced Eemian ice.

The meltwater might be of recent meteoric origin, having reached bedrock via collapsing meltlakes, crevasses or moulins and then migrated sideways to re-establish local hydrostatic equilibrium. Most of the surface melt action is well to the south though one englacial lake has been reported on the west coast at approximately this latitude.

The other option is pressure-melt at depth, as seen at the bottom of NGRIP but not nearby NEEM. I've not seen any freeze-ups so far situated over the Grand Canyon -- instead the stratigraphy is just conformally slumped as expected over any bedrock depression -- so that seems implausible as a meltwater conveyance. The same might be said for sea water intrusion (too much permittivity) even in view of the ice shelf and past glacial overdeepening.

Clearly one or more features needs to be drilled, minimally just for the temperature profile. Each of the scenarios above plays out differently in terms of stable isotope patterns.

In 2010, the DC8 flew straight down the whole length of Petermann, some 258 km from Nares Strait to the south end of the Grid. See https://data.cresis.ku.edu/data/rds/2010_Greenland_DC8/images/20100324_01/ and look at 028-31.

The highlights are posted below: nothing, overdeepening, more nothing, finally some datable older ice (likely Eemian) at the southern end of the grid  (N5, W16) in the grid coordinate system laid out above. There is some basal turmoil in the last frame but it's not clear whether this should be classified as having some freeze-up.

Yet looking at its orthogonal intersection (at its very far right end) with the east-west line W16 in the grid, it's clear that the line from the glacier terminus is just getting into some rather heavy upheaval territory. It all depends on where the slice happens to fall, here a bit off from the two big upheaval centers.

After whole experimental enchilada is assembled, we could then ask what is going on in the interiors of the grid cells (using held-back diagonal overflights to test interpolation methods), and maybe then begin to speculate about what it all represents.

[A-Team]

Since it takes all of 10 minutes (table below) to collect all the vertices of the Petermann Grid using GooglE's 'path' feature, I wondered why these haven't been previously tabulated as it provides the basis for programmatic joining of radargrams.

My sense of wonderment grew in comparing the resolution (image below) of the only published map of Greenland-wide freeze-ups to a commemorative US Postal Service stamp (the one shown is 40 mm x 31 mm). The freeze-ups portions of flight tracks seem hand-drawn lines yet Cresis offers kml files online with all the internal waypoints needed to describe the segments. Vector maps scale without loss of resolution and are easily distributable. The N4,W16 location above is barely locatable as things stand.

Wonderment gave way to astonishment searching for a working definition of freezeup in article text. Surely no two observers would make the same annotation calls on these features, which were first discovered not in this article but (like the Grand Canyon) in the 1997-1999 overflights. Below is a flight segment from 1997 clearly showing both:

https://data.cresis.ku.edu/data/rds/1997_Greenland_P3/images/19970527_01/19970527_01_004_1echo.jpg

However the problem is deeper than just remedying belated anecdotal depictions: even an objectively defined procedure might cull the incipient, weak, and decaying freeze-ups that could provide critical information on the conditions of origination and overall life cycle.

Since the signature of a basal upheaval is deformation of the stratigraphy above (beyond what bedrock topography might do), freeze-ups might be taken quantitatively as radius of curvature (ImageJ plugin) relationships between smoothed bedrock and several ubiquitious dated isochron (second image). Somewhere in here conservation of non freeze-up ice volume, direction of ice flow, and paleo changes in that flow have to play a major role.

Rather than black/white binning, continuous coloring can show a nuanced freeze-up status for the the Petermann Grid. This is self-deprecating (sidesteps cutoff).

Interpolating the color into the interior of the cells by a variety of methods, those can be objectively evaluated using  held-back oblique flights from other years. After our work-study student completes this in the second week of work, the analysis can be repeated over in the Zachariae basin, which essentially finishes off Greenland.

The contribution of bedrock topography to stratigraphic deformation is of great interest in itself but not much of an issue on this grid. The fine structure of that topography is rapidly damped out in younger stratigraphy in a manner reminiscent of fourier filtration, though flares remain as though it were really the first derivative of the topography (slope) that mattered.

Scientists have an insider joke around LPUs. That stands for Least Publishable Unit. The joke is on you, the funder. Research is conducted and then broken into as many small pieces as are standalone publishable, maximizing the number of publications (or impact factor/future citation count) and so the appearance of productivity. That risks someone else barging in, so be sure not to pass the baton in supplemental. 

I don't have any illusions about a few blog posts changing the custom-and-culture of Greenland glaciology. Let's just say I find it very peculiar. (Note: many do work very very hard, especially those in the field.) 

	N1		N2		N3		N4		N5		N6		N7		N8		N9		N10		N11		N12		N13		N14		N15		N16		N17		N18
W1	59.301	80.006	58.856	80.054	58.471	80.095	58.052	80.139	57.627	80.18	57.201	80.223	56.771	80.265	56.336	80.304	55.882	80.349	55.458	80.39	55.013	80.43	54.553	80.469	54.17	80.504	0	0	0	0	0	0	0	0	0
W2	59.04	79.952	58.652	79.994	58.223	80.039	57.815	80.079	57.393	80.122	56.967	80.166	56.524	80.209	56.12	80.251	55.669	80.293	55.241	80.329	54.797	80.372	54.276	80.413	53.986	80.469	0	0	0	0	0	0	0	0	0
W3	58.816	79.894	58.414	79.934	57.987	79.979	57.582	80.02	57.148	80.064	56.724	80.107	56.295	80.147	55.864	80.186	55.427	80.223	54.974	80.261	54.503	80.3	54.042	80.337	53.619	80.369	0	0	0	0	0	0	0	0	0
W4	58.579	79.834	58.181	79.876	57.769	79.92	57.355	79.964	56.934	80.005	56.527	80.05	56.058	80.09	55.684	80.128	55.225	80.172	54.788	80.213	54.347	80.25	53.917	80.29	53.468	80.324	0	0	0	0	0	0	0	0	0
W5	58.358	79.778	57.955	79.817	57.534	79.863	57.126	79.905	56.726	79.947	56.292	79.988	55.857	80.034	55.435	80.072	55.069	80.108	54.592	80.151	54.139	80.191	53.702	80.232	53.228	80.259	0	0	0	0	0	0	0	0	0
W6	58.135	79.718	57.724	79.76	57.302	79.803	56.905	79.845	56.476	79.889	56.063	79.93	55.666	79.97	55.217	80.015	54.783	80.053	54.366	80.09	53.914	80.127	53.47	80.161	53.009	80.197	52.534	80.231	52.106	80.262	0	0	0	0	0
W7	57.911	79.658	57.503	79.703	57.101	79.744	56.705	79.788	56.274	79.829	55.862	79.872	55.446	79.913	55.016	79.953	54.585	79.996	54.168	80.031	53.719	80.071	53.307	80.109	52.871	80.148	52.431	80.187	51.987	80.222	51.538	80.26	51.064	80.294	0
W8	57.691	79.599	57.291	79.642	56.88	79.686	56.472	79.729	56.061	79.771	55.64	79.814	55.226	79.854	54.804	79.895	54.378	79.935	53.925	79.97	53.516	80.01	53.052	80.054	52.711	80.092	52.199	80.129	51.749	80.169	0	0	0	0	0
W9	57.469	79.538	57.071	79.585	56.664	79.627	56.254	79.668	55.854	79.712	55.426	79.752	55.018	79.792	54.608	79.834	54.183	79.873	53.742	79.911	53.333	79.949	52.911	79.989	52.46	80.026	52.029	80.062	51.596	80.099	51.166	80.135	50.751	80.167	0
W10	57.251	79.481	56.873	79.522	56.465	79.566	56.054	79.608	55.64	79.649	55.241	79.691	54.822	79.733	54.4	79.772	53.975	79.813	53.563	79.851	53.131	79.891	52.697	79.929	52.268	79.966	51.833	80.003	51.396	80.039	51.008	80.071	50.617	80.101	0
W11	57.043	79.422	56.651	79.463	56.255	79.508	55.855	79.548	55.453	79.59	55.033	79.632	54.616	79.671	54.214	79.713	53.783	79.75	53.349	79.791	52.929	79.829	52.506	79.867	52.08	79.904	51.646	79.941	51.206	79.977	50.77	80.014	50.337	80.047	0
W12	56.835	79.36	56.434	79.401	56.031	79.446	55.625	79.488	55.217	79.53	54.824	79.57	54.421	79.612	53.993	79.65	53.573	79.69	53.16	79.729	52.734	79.769	52.305	79.805	51.892	79.843	51.455	79.88	51.036	79.916	50.614	79.95	50.198	79.984	0
W13	56.626	79.298	56.251	79.342	55.835	79.385	55.446	79.427	55.045	79.471	54.631	79.511	54.214	79.551	53.803	79.592	53.39	79.63	52.964	79.669	52.535	79.708	52.123	79.746	51.706	79.78	51.267	79.819	50.855	79.853	50.461	79.886	50.032	79.92	0
W14	56.423	79.238	56.023	79.283	55.628	79.324	55.231	79.366	54.831	79.408	54.423	79.45	54.018	79.489	53.61	79.528	53.192	79.569	52.771	79.61	52.356	79.646	51.929	79.683	51.511	79.72	51.085	79.756	50.647	79.792	50.246	79.826	49.825	79.861	0
W15	56.214	79.18	55.832	79.22	55.419	79.266	55.022	79.308	54.643	79.347	54.231	79.387	53.827	79.428	53.419	79.469	53.02	79.507	52.586	79.547	52.18	79.584	51.752	79.62	51.331	79.658	50.885	79.694	50.469	79.729	50.049	79.765	49.604	79.799	49.178
W16	56.002	79.117	55.623	79.16	55.232	79.203	54.848	79.248	54.441	79.288	54.041	79.328	53.647	79.365	53.232	79.407	52.824	79.445	52.402	79.483	51.978	79.524	51.584	79.56	51.154	79.597	50.731	79.633	50.306	79.669	49.877	79.703	49.413	79.74	48.989
W17	55.816	79.061	55.408	79.102	55.018	79.145	54.624	79.182	54.26	79.223	53.841	79.265	53.45	79.304	52.926	79.358	52.702	79.376	52.211	79.425	51.8	79.459	51.365	79.497	50.968	79.533	50.548	79.57	50.124	79.606	49.709	79.638	49.311	79.672	48.857
W18	55.611	79.001	55.224	79.04	54.913	79.073	52.022	79.361	51.613	79.401	51.211	79.433	50.795	79.472	50.367	79.506	49.945	79.544	49.488	79.58	49.006	79.62	48.563	79.656	0	0	0	0	0	0	0	0	0	0	0
W19	55.418	78.939	55.033	78.978	51.856	79.297	51.429	79.335	51.028	79.372	50.604	79.409	50.198	79.443	49.767	79.483	49.312	79.522	48.929	79.549	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0

[A-Team]

The first image below considers the neighborhood of the track 20110429_01_15, a component of the west-east line W16 in the Petermann Grid. To the west, in the direction of Humboldt Glacier, the widest tidewater glacier in Greenland at 110 km width the bedrock is quite flat, there are no upheavals, and the stratigraphy is undisturbed layer cake (indicated by the grayscale gradient.

This flight segment is augmented by two extensions, six orthogonal intersections, and one 1997 diagonal (held-back). These provide opportunities to determine for example, if a minor upheaval candidate is actually the toe of a much larger feature or fades out over the scale of grid cells (~8 km). There is no bedrock topography anywhere in this scene, nor any unaccounted-for mid-age horizon deformations; hence invoking GRACE gravity mapping to rule out radar artifacts from side-scatter off rocky hillsides is gratuitous.

The second set of images extends the west-east line W16 to the east, past the Grand Canyon for a total length of ~125 km and does the same for the flanking tracks W15 and W17 and so on northward. Note that the 'same' three upheaval features extend at least 45 km north to south, the direction of today's ice flow.

In the next post or so, I'll be looking for indications that Petermann has captured ice that once flowed to Humboldt at some point in the past, along the lines of Bungenstock in Antarctica (DOI: 10.1002/2014JF003291).

These radar transects could not be reduced below 1165 pixel width with retention of detail and thus will not display properly on the forum without an additional click.

[sidd]

"To the east, in the direction of Humboldt Glacier ..."

Humboldt is west of Petermann ?

[sidd]

I was just (finally) reading at the Karlsson and Dahl-Jensen paper from the cryosphere discuss list, which has a very accessible treatment of NEGIS.

http://www.the-cryosphere-discuss.net/9/719/2015/tcd-9-719-2015.html

I just realized (I'm slow but i get there) there must be nonconformities at the shear margins at the edges of fast flow features like NEGIS, where the layercake has been advected downstream.

Apart from that its a nice paper, perhaps might be discussed in the NEGIS thread(s). Basal hydro switching was discussed by Livingston I think, especially in the context of Petermann/Humboldt.

sidd

[A-Team]

I fixed it, thx sidd. It's a struggle with both coasts of Greenland east of me. Hawaii, now that's west.

Yes, several very good papers on NEGIS the last 12 months. Good point about non-conformities at the shear lines -- even if the stratigraphy matches, it doesn't mean their trajectories did: streamlines vs streaklines vs pathlines vs timelines https://en.wikipedia.org/wiki/Streamlines,_streaklines,_and_pathlines

MS Pelto posted a nice review of Humboldt, dates from Sept 2010. Much better radar coverage arrived summer 2013 when U Kansas flew a very dense grid out of Thule, inland to 125 km. Flipping hastily through the pdf with all that imagery, I'm not seeing any big surprises, maybe small upheaval at 0130420_01_017, a small deformation at 0130420_01_028 but on the whole indistinct stratigraphy and not too much of it.

It look like mostly Holocene ice, which raises the question of what happened to all the older ice (79.8º N here). Petermann ice at a similar distance inland (depending how measured) is ~1200 m deeper and ~90 kyr older. Humboldt today seems to be moving rather slowly per Joughin 2010 with ice discharge mainly in the NE corner surveyed. It's not so clear how it got to be the widest marine terminating glacier in the northern hemisphere -- plowed out by some long-gone major glacier or just a fortuitous lack of topography (which better fits the lack of continental shelf features).

https://glacierchange.wordpress.com/2010/09/04/humboldt-glacier-retreat-greenland/
https://data.cresis.ku.edu/data/rds/2013_Greenland_P3/pdf/20130420_01.pdf

[A-Team]

Going by the transects shown in post #81 above, it appears the Petermann Grid upheavals are coherent over at least 80 km in the north-south direction and 40 km west-east. Indeed the changes between say W09, W10, W11 takes a second look to even see.

This is very very different situation from the usual incoherent jumble of oddly shaped and randomly oriented upheavals in terms of the underlying explanatory physics (which has escaped us so far) -- these are not hummocks but rather Basin and Range-like ridges oriented along the flow.

How far into central Greenland does this feature complex go, where is the water coming from, how many cubic km are involved, where did the displaced younger ice go (it's incompressible), what forces are lifting features up into the thousands of meters of ice overhead on flat or even reverse slope terrain, why here and not everywhere else (notably Humboldt)?

To be continued for sure, but I'll be looking for a pressure ridge mechanism that doesn't involve advection of off-site material or affect bedrock other than isostatic adjustment. Note how the Three Sisters deform but never fracture, and how the Oldest Sister sometimes dips down to bedrock even in a region with 500 m of Eemian ice.

Of the published explanation to date, W Pauli might say "das ist nicht nur nicht richtig, es ist nicht einmal falsch!"
That 1930's quote has morphed into a whole web site today: http://www.math.columbia.edu/~woit/wordpress/

The animation is still heuristic but gives an idea of how scene tiling, rescaling, cropping, aligning, and combining with north-south sectioning should ultimately be done. The background grid centers at 1500 m depth with 500 m increments and  ~10 km horizontal width units.

I might add some dotted lines at intersections so folks viewing on their iphones can print these out, cut halfway with scissors and assemble into a grid like cardboard box dividers of a wine case.

[A-Team]

It would be interesting to estimate the total volume displaced by the Petermann upheavals. For reference, the overall surface mass balance loss for all of Greenland is something like 739 GT/yr which pencils out to 806 km³ of ice using its lesser density.

For an initial estimate, the grid flown can be assumed along and across the long axis of these features which are fortunately aligned more or less along the direction of contemporary ice flow (that motivate flight planning). Looking at the 9 evenly spaced cross sections above, some way is needed to paint and count upheaval pixels (of known area).

The first image below shows an effort along these lines for 20110429_01_015. It's not entirely subjective but still unsatisfactory (high error) because different people might not pick out the same areas as upheavals. It might work better to take the area under the Middle Sister isochron because this provides a consistent cap to the upheaval volume.

Since the Petermann Grid is surrounded by level layer cake in which the Middle Sister and bedrock topography are flat, the area between them can be taken as non-upheaval and that constant subtracted off to measure regions of upheaval. This may be directly available in MacGregor 2015 but is easy enough just to do right off the Cresis radar imagery.

Note troughs and basins (negative numbers) are also quantified by this procedure and equally important to physical interpretation. The minimum occurs where this isochron meets the bedrock (see above for example occurrences) -- this is telling us something about the big melt-off of the Eemian (or maybe just mass-conservation ice borrowed by the upheaval process).

Having obtained the area of upheaval at 9 cross sections, since these are consistent with a stable ridge structure, the volume can be estimated by summing the average area between a pair of consecutive cross sections times their separation in km. Ordinary kriging won't be effective here because troughs between the ridges will weigh in on the interpolation.

The second image, at original radar scale, shows that the average thickness between the Middle Sister and bedrock is 120.4 pixels over the 1090 pixel width of 20110429_01_015 which corresponds, according to the vertical scale bar of 500 m per 116 pixels, to 519 m times the width of 52.39 km for an area of 27.75 km² of which maybe 2/3 will prove upheaval area after I get around to measuring and subtracting off adjacent layer cake thickness on  20110429_01_016.

This 18.5 km² time the ~80 km extent northward give ~1500 km³ for the upheavals transected by this one track, which might be a third of the total upheavals in the Petermann Grid and its less-flown connecting regions to the south. This comes in at a respectable half the annual SMB volume loss. That 4500 km³ is a lot of ice.

A bit of statistics is needed to capture variation in bedrock and upper isochron but gimp does these in a flash (third image) which could give some sense of error associated with upheaval volume estimates. But first, next post, it's better to get some control on north-south interpolation by looking at transects that happen to intersect 20110429_01_015 at a peak in upheaval deformation.

[Espen]

Fantastic work A-Team, by the time of spring you will know all the Polar-Mouse holes up there?

[A-Team]

It would get your attention if the younger ice were peeled off and a 1000 m tower of unexpected ice were revealed, dwarfing the tallest skyscraper in Dubai (Burj Khalifa, 830 m)? Maybe if you saw the whole 80 km ridge, of which the tower is only a representative slice.

Overflights orthogonal to contemporary flow have a quasi twofold symmetry centered on each feature whereas (next post or so) flights parallel to flow do not have this symmetry but lopsided nested lobes instead, on the downhill side that suggest extensional flow in that direction. In the Petermann Grid upheaval zone, there are no significant deflections attributable to bedrock because it is too featureless. (No upheavals are located over the Grand Canyon.)

Although the ice was laid down in sedimentary layers, I'm not completely thrilled with use of rock geology terms (anticlines, synclines, recumbent folds, sheath folds) to describe ice deformations. There may be a better analogy to gravity-driven salt bed flow (eg Paradox Formation, US) though dissolution of salt by water does not quite correspond to ice melt.

I'm beginning to wonder if we almost lost the Greenland Ice Sheet during the Eemian. I'm not seeing the slightest hint of deformations in more recent layers. There have back-and-forth views on how much of the sea level should be attributed to Antarctica. These views affect allocation of research resources.

I've annotated various features in the last frame of the animation. The vertical exaggeration is about 10x. May have to click to see it run at full size. A portion at 2x enlargement is shown in the second image. As usual, things look quite different at 1:1 scale (third image).

The fourth image shows one-off diagonal transects from other years. It's really difficult for me to imagine how any working Greenland glaciologist then or now could be unaware of the 1997-99 overflights. Zachariae is also significantly impacted.

For the convenience of this yet-to-surface bright undergraduate, I sorted through the entire Cresis archive for flight segment names relevant to studying the Petermann Grid -- these can be grepped into a kml file that Google Earth will load in one step.

Some of these are just flanking layer cake needed to define upheaval volume (see preceding post). It's important to recall that NEEM drilled through a double fold that cannot be seen on radar, probably for lack of adequate reflectors. It follows that upheavals must be far more widespread than what we currently can see:

20100324_01_007, 20100324_01_008, 20100324_01_009, 20100324_01_010, 20100324_01_011, 20100324_01_012, 20100324_01_013, 20100324_01_014, 20100324_01_015, 20100324_01_016, 20100324_01_017, 20100324_01_018, 20100324_01_019, 20100324_01_020, 20100324_01_021, 20100324_01_022, 20100324_01_023, 20100324_01_024, 20100324_01_025, 20100324_01_026, 20100324_01_027, 20100324_01_028, 20100324_01_029, 20100324_01_030, 20100324_01_031, 20100324_01_032, 20100324_01_033, 20100324_01_034, 20100324_01_035, 20100324_01_036, 20100324_01_037, 20100324_01_038, 20100324_01_039, 20100324_01_040, 20100324_01_041, 20100324_01_042, 20100324_02_001, 20100324_02_002, 20100324_03_001, 20100324_03_002, 20100324_04_001, 20110429_01_001, 20110429_01_009, 20110429_01_010, 20110429_01_011, 20110429_01_012, 20110429_01_013, 20110429_01_014, 20110429_01_015, 20110429_01_016, 20110429_01_017, 20110429_01_018, 20110429_01_019, 20110429_01_020, 20110429_01_021, 20110429_01_022, 20110429_01_023, 20110429_01_024, 20110429_01_025, 20110429_01_026, 20110429_01_027, 20110429_01_028, 20110429_01_029, 20110429_01_030, 20110429_01_031, 20110429_01_032, 20110429_01_033, 20110429_01_034, 20110429_02_001, 20110429_02_002, 20110429_02_003, 20110429_02_004, 20110429_02_005, 20110429_02_006, 20110429_02_007, 20110429_02_008, 20110429_02_009, 20110429_02_010, 20110429_02_011, 20110429_02_012, 20110429_02_013, 20110429_02_014, 20110429_02_015, 20110429_02_016, 20110429_02_017, 20110429_02_018, 20110429_02_019, 20110429_02_020, 20110429_02_021, 20110507_01_011, 20110507_01_012, 20110507_01_013, 20110507_01_014, 20110507_01_015, 20110507_01_016, 20110507_01_017, 20110507_01_018, 20110507_01_019, 20110507_01_020, 20110507_01_021, 20110507_01_022, 20110507_01_023, 20110507_01_024, 20110507_01_025, 20110507_01_026, 20110507_01_027, 20110507_01_028, 20110507_01_029, 20110507_01_030, 20110507_01_031, 20110507_01_032, 20110507_01_033, 20110507_01_034, 20110507_01_035, 20110507_01_036, 20110507_01_037, 20110507_01_038, 20110507_02_001, 20110507_02_002, 20110507_02_003, 20110507_02_004, 20110507_02_005, 20110507_02_006, 20110507_02_007, 20110507_02_008, 20110507_02_009, 20110507_02_010, 20110507_02_011, 20110507_02_012, 20110507_02_013, 20110507_02_014, 20110507_02_015, 20110507_02_016, 20110507_02_017, 20110507_02_018, 20110507_02_019, 20110507_02_020, 20110507_02_021, 20110507_02_022

Flight segment names for diagonals to the Grid:
19970527_01_003, 19970527_01_004, 19970527_01_009, 19970527_01_019, 19970528_01_005, 19970528_01_006, 19970528_01_009, 19970528_01_010, 19970528_01_011, 19990507_01_003, 19990507_01_004, 19990507_01_015, 19990507_01_016, 19990510_01_014, 19990510_01_015, 19990510_01_016, 19990511_01_003, 19990511_01_004, 19990511_01_006, 19990511_01_007, 19990511_01_013, 19990511_01_014, 19990511_01_015, 19990512_01_002, 19990512_01_003, 19990512_01_019, 19990513_01_002, 19990513_01_003, 19990514_01_023, 19990517_01_008, 19990517_01_009, 19990517_01_010, 19990519_01_022, 19990525_01_022, 20030513_01_019, 20030513_01_020, 20110502_01_010, 20110502_01_011, 20110502_01_012, 20110502_02_024, 20110502_02_025, 20110502_02_026, 20110509_01_011, 20110509_01_012, 20110509_01_013, 20120503_03_056, 20120503_03_057, 20120503_03_058, 20120503_03_059, 20120514_01_016, 20120514_01_017, 20120514_01_018, 20120514_01_019, 20120514_01_020, 20120514_02_045, 20120514_02_046, 20120514_02_047, 20120514_02_048, 20120514_02_049, 20120514_02_050, 20120514_02_051, 20120514_02_052, 20120516_01_014, 20120516_01_015, 20120516_01_016, 20120516_01_017, 20120516_01_018, 20120516_01_076, 20120516_01_077, 20120516_01_078, 20120516_01_079, 20120516_01_080, 20130426_01_011, 20130426_01_012, 20130426_01_013

[Shared Humanity]

Still trying to follow all of your posts with a  degree of difficulty.

Correct me  if I've  misunderstood.

The three sisters (upheavals) consist of older ice than the ice surrounding them?

[A-Team]

Still trying to follow all of your posts with a  degree of difficulty.

These upheaval features someday will be a big yawn. At this point in time, they are a difficult research topic (only because they have never been an experimental target). However my sense is they are low-hanging fruit, once the tree is located.

That is, the Petermann grid sub-optimally overlies upheaval alignments. Now that their positions are known, this area could be re-flown specifically to determine the 3D structures of these features. They aren't skyscrapers, they are mountain ranges elongated in the direction of flow.

It can be assumed that some group will drill one or more of these upheavals this summer. That would be hugely informative. Thule, with commercial flights and a full-service hotel, is a short hop away.

The 'Three Sisters' are a triplet pattern of late Pleistocene radar horizons that are easily recognized by eye and seen all over Greenland. That is because they occur in an extensive echo free zone, ie are dark bands against a white background. It's just a name I made up (borrowed from a mountain range of that name east of my home town), not something used in journals.

However layer cake is a term used by Greenland glaciologists so perhaps I could have chosen another baked good. Upside-down cake is best reserved for re-folded recumbent folds; bagels might reference cross-sectional eyes of sheath folds, pound cake for radar reflectors a mixture of four ingredients (till, sea salt, meltwater, ice), and jelly rolls have been used by reporters to describe the upheavals.

Here is glaciologist Kirsty Tinto:  “When you’re flying over this flat, white landscape people almost fall asleep it’s so boring—layer cake, layer cake, layer cake,” says Tinto. “But then suddenly these things appear on the screen. It’s very exciting. You get a sense of these invisible processes happening underneath.”

Properly speaking, the Three Sisters should be replaced with their MacGregor 2015 dates and/or tied to Icelandic volcanic eruption names from that new chronology and/or tied to their tephra or conductivity band in the six main cores to bedrock. Scientific nomenclature gets geeky in hurry since hundreds of bands have to be named, so I opted for something blog-friendlier.

Layer cake is ubiquitous and provides the default stratigraphy for modeling (but that seldom considers anything beyond surface elevation, slope, ice thickness and bed topography). However it is these extremely complicated ice penetrating radar scans have the most to tell us about the history of Greenland Ice Sheet during warmer epochs.

I looked around for an exceedingly boring stretch of layer cake that displays the Three Sisters concept without a lot of distractions: here is one between Petermann and Humboldt glaciers. Below that, the 4th frame of the animation better viewed in #83 above.

The last image is a montage of 4 consecutive flight segments showing upheavals sandwiched between flanking layer cake (which appears wavy here because of the great horizontal compression necessary to bring the image into 700 pixel blog width maximum).

[A-Team]

Here is an instructive progression of transects in the center of the Petermann Grid. Note that the flight lines did not quite follow the axis of the feature over the 46.8 km -- it moves to the left during the animation (the flights needed to be 12º more east-westerly).

[Espen]

It looks very dramatic.

[Espen]

Can I suggest this for background music?:

[A-Team]

Can I suggest this for background music?

Maybe save that for the north-south radar soundings.
With Grieg vocals, is there concern this Norwegian soprano could melt Greenland ice?

[A-Team]

Here are three west-east extensions north of the Petermann Grid set shown above and three more to the south. These are somewhat more complicated not only to make but also to bring into the comparison as multiple flight segments have to be combined as the features being tracked shift westward, which also raises offset issues.

However you can still see that features extend continuously over all 12 tracks which span 94.4 km (according to GooglE's path ruler). They do fade out eventually and none reach the active region of Petermann Glacier.

The bottom line is coherence rather than skyscraper jumble. The vertical scale exaggeration creates visual artifacts, the upheavals are actually quite gentle at 1:1 scale. Note too that the flight grid was presumably laid out parallel/orthogonal to the local glacier velocity but the features themselves are oblique to these lines.

We are hitting the wall in terms of display width vs available Cresis radar resolution. The latter is retained below but the images will have to be viewed at full size rather than the 700 pixel width blog limit. Yet even there to view the whole assembled grid, a large retinal grade monitor will ultimately be needed.

[A-Team]

Recall the Petermann flight grid was aligned by design to the velocity vector field of the glacier. That turns out not to be quite so well aligned with the upheaval features. That means the west-east transects are slightly oblique to the features and that the north-south flight lines do not run down their long axis.

While we wait for this to be re-flown with the upheavals as top priority, it's worth looking at some of the one-off diagonal flights from earlier years to see if any by happenstance went down a feature axis. These would also be sites of multiple crossings which are favorable for 3D tomographic reconstruction which so far is lacking.

As mentioned earlier, the radar imagery is sliced, diced, and re-processed in various inconvenient ways, including but not limited to unspecified tiling overlaps, arbitrary vertical offsets, different horizontal and vertical scales on every segment, and lack of optimal (or even uniform) contrast.

I marked up individual flight segments with alternating colors so stop-start coverage can be seen in GooglE (first image) and looked at a promising area with multiple crossings (second image), later realizing if the west-east Petermann Grid lines were rotated to horizontal that miniature versions of upheavals could be placed and rescaled in situ without having to delve into their details (third image).

This is the best way I've come across to display the whole archive of grid files in one small graphical overview. The third image shows an 8% piece of that project, west-east only (north-souths would comprise another layer in gimp).

[A-Team]

Here is a more favorable case of feature tracking that turns up five one-off oblique flight lines that shed additional light on the 3D structure. Note the feature itself is oblique to the Petermann Grid but is easily recognizable as a rather invariant linear 'ridge' upheaval (subject to 1:1 hor:ver rescaling later).

Note not only is the peak shape rather well conserved, the internal 'substructure' also carries over consecutive transects which are separated by ~ 8 km. It's not known what the radar reflectors are in these upheavals (till?) nor if the stratigraphy timeline is upside-down (ie oldest on top). Different features may have different origins in time, different life cycles, and different underlying physical mechanisms.

What's shown is 72.4 km that tracks ice movement fairly well, though that need to be revisited with PARCA stakes and satellite pair coorelation. The feature fades out 35 km from the mouth of the glacier (~ first surface rocks). The second image tracks its southern extent (with top row repeating bottom row of first) -- this is a huge continuous feature on the order of ~120 km long, 5 km wide and up to 1 km high.

Note the flanking features are quite remarkable ... and have very different properties.

[Timothy Astin]

A-Team,

I have greatly enjoyed your many posts on the ground radar investigation of the Greenland Ice Sheet, and will continue to enjoy them as you work away at the archive and report recent publications.

In my former university life I used to teach GPR and its close cousin reflection seismology, and if still in that field would have risen to your challenge to have some student project(s) based on the public archive.

There are complications with interpreting ground radar data where time is the vertical axis, though I haven't looked in to how much the radar velocity might vary with depth in an ice sheet, one anticipates that velocity likely increases with depth.

So, as you will know, the true geometry of these upheavals is distorted a bit in the time section data, and would require reprocessing for "migration" and depth conversion to give a more accurate geometry.   It won't will affect the basics of interpretation, but there might be some local surprises in the geometries.

The upheavals are interesting. Clearly there is a density instability, with a significant density contrast between upheavals and overlying ice towards the base of the sheet. This contrast becomes zero by the mid-range depths, stopping the upheavals rising further.

Can the density contrast be as simple as a temperature contrast? In regions of the sheet without convective heat transfers through groundwater movement, basal heat flow into the ice sheet should cause the ice to be warmer at the base than at the top of the sheet.  As ice flows readily, and given that any temperature contrast can be maintained over long periods, the observed upheavals suggest that the expected small temperature contrast is enough.

Again, the upheavals seem to be largely of Eemian ice, so there may well be interesting factors like the differential consolidation of Eemian and later ice, with different insulating (thermal conductivity) properties within the ice-sheet, controlling temperature distribution over time.

The elongation of upheavals along the ice flow direction is compatible both with propagating temperature anomalies laterally as the ice sheet flows, and also with expected convection vortices within a flowing sheet.  Clearly the presence of recumbent folds in the ice associated with upheavals shows the importance of the lateral flow occurring at the same long time as the upheavals develop.  (Compare the elongation of stone stripes developed on intermittently frozen ground when there is a small amount of down-slope creep.)

"Proper" science would try to quantify likely heat flow, resultant temperature gradients and consequent density variations to test this. That is another sort of student geophys project to add to the ones looking at as the GPR data itself.

[A-Team]

Hello Timothy, welcome aboard and thanks for a most stimulating post! You've arrived just in time for the very first upheaval structural assemblies. And that site above with the yellow arrows has enough tomography (5 slices, axially interpolatable) to put it together in 3D.

radar velocity likely increases with depth in an ice sheet, so the true geometry of these upheavals is distorted a bit and would require reprocessing for "migration" and depth conversion to give some local surprises in the geometries.

The right-hand ordinates on the Cresis radargrams do provide experimentally recorded radar round trip times in microseconds, from which the left-hand depth axis is derived by assuming a medium of constant dielectric (radar velocity) for ice, after discounting for firn and plane-to-surface air.

Thus the data is there to re-process the whole archive if only we had image-driven 'instructions' for the variable vertical compression. Here we might ask what sort of local conditions would give the most notable geometric effects.

Radar of many designs has flown over the six summit cores to bedrock numerous times -- where the depth, temperature profile, ice fabric, conductivity, and other annual ice core properties are known. For example GRIP2 has a long section of brittle ice, NGRIP temperate basal ice whereas NEEM does not. Despite these opportunities to fine-tune the radar velocity profile, Cresis has never seen a good reason to alter the assumption of constancy (perhaps it is overwhelmed by other error sources).

That's not the end of the story however; there has been new work on extracting temperate ice distribution as well as the islandwide temperature depth profile using phase and other esoteric properties of radar return.

The upheavals are interesting.

Indeed there is nothing resembling these spectacular deformations anywhere else on the planet. (The Gamburtsev ice upheavals are coming off a steep mountain range.) There will be interesting physics at scale here, but not new physics. The question is, which physics.

Any explanation for their occurrence must also explain their non-occurrence. These features are overwhelmingly concentrated in the far north, and there in regions of unexceptional bedrock topography and slow flow, mostly in the Petermann and Zachariae basins but also with dramatic exceptions in the deep interior and as far south as Epiq.

We don't know if the monkey at the keyboard who typed the perfect Shakespeare sonnet went on to type a second. Now that I've stumbled onto a practical workaround for all the unfortunate choices made in radar database structuring, it will be interesting to compare and contrast the Zachariae Grid with that of Petermann.

Clearly there is a density instability, with a significant density contrast between upheavals and overlying ice. This contrast becomes zero by the mid-range depths, stopping the upheavals rising further. Can the density contrast be as simple as a temperature contrast?

Excellent suggestions. The 5-6 hypotheses put forward to date have invoked either gravitational (buoyancy, isostatic equilibrium, hydrostatic equilibrium, density instability, downslope sliding) or electromagnetic forces (aka chemistry: volumetric changes in the ice Ih <--> liquid water phase diagram, supercooled water, viscosity of ice with temperature).

Here we need to make a table of what each predicts (in advance of drilling one), or more broadly how might they be experimentally distinguished. Why now? Because modellers will leave themselves too much wiggle room in terms of free parameters -- not be disprovable whatever the outcomes on core isotopic, temperature and conductivity.

I don't recall any of these theorists saying in the 1990's, go fly these northern glaciers, you will see spectacular upheavals.

geothermal or basal frictional heat flow into the ice sheet should cause the ice to be warmer at the base than at the top of the sheet. As ice flows readily, and given that any temperature contrast can be maintained over long periods, the observed upheavals suggest that the expected small temperature contrast is enough... there may well be differential consolidation of Eemian and later ice, with different insulating (thermal conductivity) properties within the ice-sheet, controlling temperature distribution over time.

We'll have lots more temperature profiles shortly from Sommers et al. The counter-intuitive aspect of the heat equation is that 2.5 million years of cold ice sitting on the warmish pre-Cambrian bedrock has cooled it to great depth, overwhelming the upwelling geothermal gradient to a certain extent. Any rebound of the gradient after the warming of the Eemian would be quite  sluggish, lagging by thousands of years. There remains, as you say, a great gulf between the time scales of diffusive vs conductive, convective and advective heat transfer. 

The elongation of upheavals along the ice flow direction is compatible both with propagating temperature anomalies laterally as the ice sheet flows, and also with expected convection vortices within a flowing sheet. Clearly the presence of recumbent folds in the ice associated with upheavals shows the importance of the lateral flow occurring at the same long time as the upheavals develop. Compare the elongation of stone stripes developed on intermittently frozen ground when there is a small amount of down-slope creep. Student can quantify.

Well put, no coincidence there with the recumbent and sheath folds spilling out to the sides. The curvature of the Petermann basin forces flow lines to converge, as does Zachariae's, whereas the flat western flank of central Greenland does not. I've not seen an explanation for why the summit ridge has this prominent spur heading west to NEEM, then Camp Century and the coast. Nor how long it has had it.

At the end of the day, what does all this have to do Greenland's contribution to sea level rise? I would say south-central ice is definitely toast, the western flank will suffer from melt, whereas surprises may come from the north where Petermann and Zachariae are awakening, even if the glaciologists on them are mostly not.

The upheavals may also provide our best insights into what actually happened in the Northern Hemisphere during the last inter-glacial. There's clearly something very wrong with the notion of Greenland just brushing off the extended warming of the Eemian.