Kriging method is one method of data interpolation involving location information (spatial data). The existence of outliers in the spatial data causes the kriging method to be incorrect. To overcome this is required kriging method that is able to handle outliers on spatial data called robust kriging. In general, the weight of the kriging method is determined by the semivariogram that measures spatial correlation. A semivariogram in robust kriging is called a robust semivariogram. There are several robust semivariogram, three of which are robust spherical, exponential, and gaussian semivariogram. The purpose of this research are to detect the location of rainfall precipitation stations including outliers, to determine the most appropriate robust semivariogram among the three semivariograms in robust kriging and to interpolate the rainfall data in March 2015 at Malang Regency. The data used in this research is spatial data of rainfall in March 2015 at 31 rain stations in Malang Regency containing outlier data. Result of research indicate that there are three location of rain station that detected outlier that is Ngantang rain station, Wagir and Bantur (Z value> 1,96). The most suitable robust semivariogram is the exponential robust semivariogram seen from the smallest RMSE value compared to the other robust semivariogram of 24.2752. The robust kriging interpolation result with the exponential semivariogram shows that the highest rainfall occurs in Ngantang and Kedungkandang rain stations.