Right...this is making more sense now. You said that the fluid map has a little water everywhere. I'm going to try my old setup...I have used both mask that are cmpletely white- ie all snow on mountain set to max white, and shades of grey. I think I did check that the .ter icefield was zero in all places off the mountain. I did that in Wilbur. But with this new info, I'll take another look at this.
Basically the principle of what yuo are saying makes sense: gradation of transparency via depth of icefield height (or an image map)- either way would work but the former is surely less complicated. You can use a cutoff in TG on the imagemap shader so that we could set all levels below say 0.1m to black, )transparent), and the rest to gradate through.