we propose a spatial filtering scheme using factorial kriging, a kind of geostatistical technique in order to separate regional and residual gravity anomaly. Spatial filtering assumes that regional anomalies have longer distance relation and residual anomalies have an effect on a smaller range. Gravity anomaly was decomposed into two variogram models depending on each influence range. And best-fitted variogram models could produce the separated regional-residual anomalies by means of factorial kriging. This algorithm was examined with synthetic gravity data, and applied to real microgravity data for figuring out the abandoned mineshaft as well.