Hi, I have ~5000 1024x1024 image arrays I'd like to hold in a data cube. A good percentage of the pixels (maybe 30%?) are data that I don't care about. Subsequently, I'm going to be extracting various statistics that drill down the longest dimension of the cube (i.e. the temporal direction). What's the most memory efficient way to handle this? Is there a trick with sparse arrays? More generally, perhaps: How do other folks handle large time-series data cubes in IDL without simply throwing more RAM at the problem? (Or is that my only solution?) Thanks!