On the optimality of clustering properties of space filling curves

Space filling curves have for long been used in the design of data structures for multidimensional data. A fundamental quality metric of a space filling curve is its "clustering number" with respect to a class of queries, which is the average number of contiguous segments on the space filling curve that a query region can be partitioned into. We present a characterization of the clustering number of a general class of space filling curves, as well as the first non-trivial lower bounds on the clustering number for any space filling curve. Our results also answer an open problem that was posed by Jagadish in 1997.