The grot is exhibited as small spots of reduced sensitivity, generally extending only over a pixel or two. Probably due to flecks of anti-reflective paint, scraped off the light baffles between the dewars, as they are forced against each other.
Notice in the Camera 1 image a "battleship" feature. This a plot of the Relative throughput across.
Above are grot masks for each of the cameras, in FITS format. These images are ratios of on-orbit flats to ground flats for all 3 cameras. A high-pass filter was applied to the ratios to remove large scale QE changes between the ground and on-orbit flats, and to remove the vignetted regions at the bottom (-Y) edge of NIC2 and NIC3. The coronographic hole was also masked out of the NIC2 ratio.
A ring filtered (low-pass) version of each ratio was subtracted from itself. Iterative sigma clipping was then done to select out the affected pixles. Previously known "bad" pixels are *not* included in these masks. These can be found in the MASKFILE calibration reference file. The masks given here just show the number of grot particles (and possibly some new dead pixels) and their locations on the detectors as of when the on-orbit flats were made. Grot affects approximately 250 pixels in each detector, or less than 1% of the total number of pixels.
The best "cure" for grot is dithering-- break your exposures into several parts, and change the pointing a bit between exposures. When recombined into a single image with CALNICB, pixels effected by grot will not be included. If you have not dithered your data, grot can be minimized in a cosmetic sense by replacing the effected pixels with the average of the surrounding unaffected ones.