The　accuracy　of　spatial　interpretation　is　close　relation　to　the　selection　of　spatial　model.　Each　model　has　its　own　advantage　or　disadvantage.　A　more　accurate　spatial　interpretation　model　can　be　obtained　by　a　linear　combination　of　some　models.　In　this　study,　first-order　spatial　autoregressive　interpretation　(SARI(1))　model,　Kriging　algorithm　interpretation　(KAI)　model　and　back-propagation　neural　network　interpretation　(BPNNI)　model　are　established　by　using　crosssection　data　or　time　series　data.　A　spatial　linear　combination　interpretation　(SLCI)　model　is　constructed　by　the　results　of　these　models.　The　weights　of　SLCI　model　obtained　by　spatial　drift.　An　empirical　research　is　carried　out　with　interpretation　some　areas＇　GDP　per　capita　in　Fujian,　2003.　It　is　found　that　the　best　one　is　the　SLCI　model.　©　2008　IEEE.