A "2D array" in
Java is always actually a 1D array of 1D arrays -- i.e., an int[10][2] is one array of length 10 that holds ten arrays of 2 ints each; this is total of 11 objects. On the other hand, you could hold the same ints in two 1D int[10]s, which would be only 2 objects. For a tiny VM where object count really mattered -- and for some Android devices this is the case according to the author of the document you link to -- then this would make the 2 1D arrays a better choice.
But for general Java development, it's terrible advice, as is much of what's in that document. In general, don't use that advice except to squeeze additional performance out of troublesome Android applications.