I looked at one of several approaches to automating internal ice penetrating radar striation finding, that of Patton. This paper as its example processes an absolutely critical flight line, that between NGRIP and NEEM on 6 May 2011. Those two ice cores have been analyzed for every conceivable parameter, so not only provide precise dates for radar striations (which will not necessarily correspond to visible ice layers) but also clues to why the radar echo is there at all and what the lines mean.
The ice cores thus possess a rather subtle property not initially appreciated, one that forms a coherent hypersurface over a million sq km of Greenland ice sheet in the case of the WHT (Wisconson Holocene Transition). That follows from identification of the WHT down to sea level at FOXX and GULL steam holes (at 515 m and 600 m depths) along a flowline down from Swiss Camp, as well as DUCK right there on the Jakobshavn Isbrae channel.
The key point here is drill holes are points (expensive ones) but radar transects are planar sections whereas Greenland is volume (2.9 million cubic km of ice). We will never have enough drill holes to extrapolate to isochronal hypersurfaces of O18, pH, conductivity, sulfate, methane, beryllium, smoke, ash etc etc. However we do have enough flight lines. Thus correlation of radar striations with ice core dates and properties allows the transects to push out the core properties.
For lower Jakobshavn Isbrae south channel, striations are most evident in the adjacent ice sheet. However seismicity triggered by calving but propagating upstream has been associated with striation overthrusting at the big curve.
The Patton paper begins by clever sharpening of radar striations with a slant gaussian adapted to local slope (discovered by monte carlo sampling of angle to maximal contrast in the fourier transform). That is quite effective (Fig.2-3). The echograms are then processed with contour snakes (ImageJ plugin not used). I found the contiguous color picker in gimp pulls out the same lines with less effort.
Using this, I gave each striation a distinct color in the NGRIP to NEEM product. We'll want this later to color the island-wide hypersurfaces in voxel view and cutaways (plugin --> 3D --> volumeViewer in ImageJ, resp. PovRay). The dark blue line shows an extra just done with the color picker.
The striations on this 423 km track are all continuous so obviously isochrons, but Patton properly verifies this in view of the extreme importance of these two drill sites. Here snowfall and compaction had slightly different histories on NGRIP relative to NEEM so depths and dates had to be cross-correlated (inset below, far right corner), and striations shown to fall on that curve. Which they do, red dots. There is otherwise no interpretation of striation dates in this paper.
Once an overall major striation pattern has been established, its labels can be easily transferred to a new track that doesn't tie in conveniently to NGRIP to NEEM using methods lifted from dendrochronology, tree rings just being annual isochrons wrapped into circles. There, rings on an unknown tree core (~ radar striations on an isolated flight line) are dated by best match to a sliding window (ie convolution) with the fiducial tree ring chronology (itself tiled up like flight lines). That can also be done internally on radar echoes where a striation might be missing for part of the track.
Since only long straight flight lines are really worth processing and unique additions barely run to 10,000 km/yr, I see little point to automation since a given year can be done manually in 20 hours at 500 km/hour. Between the many glitches in data acquisition and bizarre deformations, there has to be a human supervising the process anyway.
I find it exceedingly implausible that glaciologists can add to a field having already undergone 80 years of intensive methodological development (dendrochronology). True, academics wrote the book on the not-invented-here syndrome so glaciologists can write a new chapter if they choose.
<url>
http://www.igsoc.org/annals/55/67/t67A048.pdf</url>
<url>
http://www.igsoc.org/journal/60/222/t13J196.pdf</url>
The image below is at the original Cresis resolution. It takes 2047 pixels of width to tile up this 423 km flight line as an 877 kb jpeg.