Offline loading of R-trees is useful to increase node utilization
and  improve  query  performance.  We present a new algorithm for
bulk loading R-trees which differs from previous proposals in  at
least  two important aspects: First, our algorithm partitions in-
put data into subtrees in a top-down fashion as partitions  close
to the root are likely to have the biggest impact on performance.
Second, at each tree level,  our  algorithm  considers  all  cuts
orthogonal  to the coordinate axes that result in a balanced tree
and greedily selects the ones that %minimize optimize a user sup-
plied  cost  function.  Extensive  experimental studies with both
real and synthetic data indicate that our algorithm is especially
competitive for region data, where it results in R-trees that re-
quire up to three times fewer disk accesses than previous  propo-
sals.   More  specifically,  it  is the method of choice for data
sets  with  skew  in  center point  locations,  areas,   or   as-
pect  ratios.   Such  data sets are common in various application
areas.