Abstract: the　sort－tile－recursivestris 　a　static　r－tree 　variation with　outstanding　build　efficienc yand　acceptable 　query　efficienc y．the　onl y　problem　the 　aggre gation　of 　data．hilbert 　curve　can　ensuredata　aggre gation　howeversome　other 　information　missed．in　order 　to 　solve　deficiency　str－treeanimproved　str－tree 　spatial 　index 　al gorithm　based　on　hilbert－curve 　is 　proposed．this　paper 　focuses 　onthe　offset 　caused　by　dimensionalit y　reduction　during　the　build　process．al gorithm　sorts 　the　objects 　bythe　hilbert　value　of 　its　mbrsgenerates 　the 　temp－abstract　nodes　according　to 　the　node　capacit yandrecords 　the 　center 　of 　each　cluster．secondthe 　al gorithm　processes　hilbert　outliers　by　adding　the　dis－tance 　constraint．thirditeratin g　the 　above　process　until　hilbert　str－tree 　is 　built．both　theoretic 　anal ysis 　and　experiments 　indicate 　that 　our　al gorithm　performs　more　efficientl y　than 　the　str－tree 　querying．