The differences are due to what the depth and distances derates are attempting to model.
Distance based AOCVM is used to model global variation effects. It's assumed that different parts of the chip will vary more so that's why as distance increases you see more variation.
Depth based AOCVM is used to model local variation effects. Local variations tend to cancel out over the course of a long path - some delays are a little worse, some are a little better, but as the depth increases the derate approaches 1 (or some other value depending on your methodology).
In my experience, distance based is used much *less* frequently than depth based AOCVM, today.
Does that answer your question? If not I'd be happy to post a more thorough explanation.