Lennart, ASLR, sidd, and icefest ask above for comment on the new Gogenini paper at
http://www.igsoc.org/journal/60/223/j14j129.pdf . This is a beautifully written paper that provides significant new information on the Greenland radar program. It would take several posts to cover the ground.
I first puzzled over why the paper used only radar data from 2005-09 for Jakobshavn, whereas a great many overflights with presumably better technology took place over the subsequent years (first image below shows 2010-14 tracks).
It has to do with pretenders to the ice-penetrating radar throne (emphasis added):
'Since 1990, several attempts have been made to sound Jakobshavn and other fast-flowing glaciers, with very limited success. Fully coherent radars operating over the frequency range 1–450 MHz are used for these measurements (Gogineni 2001, 2012; Dall 2012; Morlighem 2014). We first succeeded in sounding Jakobshavn Isbræ ... in 2005.'
Chasing this down, 'Dall 2012 P-band radar sounding in Antarctica' concerns a threat from satellite radar, which could provide more comprehensive coverage of big continents if only the surface clutter issue could be resolved. That radar (ESA Polaris) on airplanes not only reached bedrock in 3300 m ice on the Greenland summit ridge (possibly detecting basal melt as well) in 2008 but also obtained spectacular results (images below) in Antarctica using a remarkably effective enhancement (first derivative of nadir return power) to better get at internal stratifications.
As you might expect, ImageJ is all over gradients, laplacians and hessians of digital imagery, eg
http://bigwww.epfl.ch/thevenaz/differentials . Gimp offers Filters → Artistic → Van Gogh (line integral convolution), various edge detectors, Difference of Gaussians, or roll your own algorithm (bump one pixel up and down, subtract grayscales, take max).
However derivatives would be better taken directly, possibly even in hardware, on incoming 12 bit return data -- in fact the whole Cresis effort should be integrated from antenna design right through digital enhancement product -- too often the latter is an afterthought or salvage effort after the goal post moves (eg from bedrock to radar stratigraphy, internal deformations and water pockets).
Which is exactly what happened here: five years after electrical engineers design the radar instrument and IceBridge conducts the flights, they bring in some poor computer science guy (11th author
http://www.csce.uark.edu/~jgauch/) to recover the latent data via histogram equalization, adaptive median and bas relief edge-detection filtration, all of which are standard menu items in Gimp.
Morlighem 2014 combined sparse radar ice-thickness data with high-resolution ice motion interferometery in a computer-intensive mass conservation optimization model, finding 'deeply incised submarine glacial valleys beneath the Greenland ice sheet' (doi: 10.1038/ngeo2167, paywalled, have copy) to determine ice thickness and bed topography along the entire periphery of the Greenland ice sheet, detecting widespread overdeepenings in the faster moving ice streams that extend 'significantly deeper below sea level and farther inland than
previously thought'.
Omg, 'previous thinkers' must mean Gogenini and co-authors of his many papers?!? The response establishes that while Cresis did not invent ice-penetrating radar, they certainly built ever-improving scientific instruments, pioneered their use on a survey-level scale (unlike Polaris) and succeeded in reaching bedrock observationally in very difficult deep ice by 2006 (whereas Morlighem 2014 merely models it).
Note however this modelling does fill in the blanks of bedrock polygons bounded by inevitably sparse radar tracks (see post above), so it's worth a more detailed look as interpolative method.
The 2014 model paper chose to cite a 2001 paper for purposes of invidious comparison (ref 7: Bamber, Gogenini 2001). It is difficult to comprehend how the Rignot group could be unaware of the intervening updates, notably the widely cited replacement (Bamber, Gogenini 2013 doi:10.5194/tc-7-499-2013) given Rignot is listed as co-author. That went online 15 Nov 2012, whereas the Morlighem article was submitted a
full year later, 20 Nov 2013.