Hoping someone can short circuit a lot of exploration
Landsat 8 values are stored as unint16 since the sensors work on a range 0 to 4096 whereas the L7+TM was 8 bit and had a digital number range 0 to 255. USGS link
Looking at the unprocessed bands in QGIS, the value tool reports values of ~8000 for band 3, whereas GRASS reports values of ~0.19 . For the same bands in L7+TM band 3 values are ~30
Understandably, the output from my old workflow in GRASS is giving nonsense. i.landsat.toar does read L8 but with the funny numbers I don't expect it is on the right track.
The end use is NDVI, I don't expect a pansharpening workflow to be the same.
Any tips on a new workflow?