I have been attempting to track ice motion in landsat 8 files. I have only been able to find the L1T files. These are terrain corrected files. Unfortunately for my study region (the Renland ice cap, East Greenland) the DEM used in this processing is of too poor quality and which introduces large geodetic errors. DEM errors can be seen as residual perspective effects [Click image on the right to expand to an animated GIF showing flipping between two images].
I would therefore really like to have the L1G product and account for the terrain correction my self (using gimpdem).
Another advantage of L1G over L1T is that there is less resampling which degrades feature tracking performance.
(This work is partly motivated by testing a new open source feature tracking & georeferencing toolbox for matlab we have been working on. I am looking forwards to putting it on github.)