Much of the article seems to be discussing the advantages of using time as the major dimension, without using the word "major". But an alternative to switching the major dimension would be to arrange the data along a Hilbert curve in 4 dimensions, giving good locality of access for range scans in time or geographical coordinate. That seems like it would also offer better compression from delta coding, since with better locality you should also enjoy smaller deltas.
Much of the article seems to be discussing the advantages of using time as the major dimension, without using the word "major". But an alternative to switching the major dimension would be to arrange the data along a Hilbert curve in 4 dimensions, giving good locality of access for range scans in time or geographical coordinate. That seems like it would also offer better compression from delta coding, since with better locality you should also enjoy smaller deltas.