There are numerous methods for data filling in hydrology. Most, however, are based on correlations with nearby stations in a general scheme of regionalization. These methods, though robust, fail to function when and where all the near stations are missed-data too. The Mashhad synoptic station has annual rainfall data over a 50 year period from 1951 to 2000 and historic rainfall data from 1893 to 1940, just before World War II. This time series has about 15 years of missed data which cannot be filled by usual methods. So, the techniques of geostatistics and kriging were adapted to this long-term time series as an alternative. The data showed a poor correlation at every time lag, showing that, while all the semi-variogram models performed nearly equal, there was a high correlation among each of the others. The results included with polynomial regression fits to different moving average orders, nailed at some reasonable estimates for missed rainfall values.