I do not see a way to make storing this more efficient (we could always leave the space unused for the smaller area but it does not hurt to use these landmarks). I.e. if we increase landmark count for Europe then we can use a smaller landmark count for areas like Turkey but the columns would be just empty and unused. To make it decoupled we would have to split the underlying graph so that we could use one landmark table (with different landmark count) per subnetwork. Or we use a HashMap instead of an array for everything but this would mean a much larger memory usage even if we safe a few entries for the smaller areas I think.