To　discuss　the　spatial　interpolation　based　on　panel　data　with　spatial　autocorrelation,　the　first-order　spatial　autoregressive　interpolation　model　and　the　Kriging　algorithm　interpolation　model　are　established　from　the　perspective　of　the　cross-sectional　data.　Genetic　algorithm　back-propagation　neural　network　interpolation　model　is　established　from　the　perspective　of　the　time-series　data.　A　spatial　combination　interpolation　model　is　established　by　the　results　of　these　models.　The　weights　of　the　combination　model　is　calculated　by　a　new　method　of　spatial　drift.　An　empirical　study　is　carried　out　with　interpolation　some　areas＇　GDP　per　capita　in　Fujian　2007,　China.　The　result　shows　that　the　most　effective　one　is　the　spatial　combination　interpolation　model.