Abstarct: Geostatistics techniques have many applications in estimations made in different branch of geosciences. Estimation of groundwater table has always been of great importance. The routine method to estimate this depth is to use geophysical methods such as electrical resistivity that is an indirect method. These methods are usually relatively costly. Due to insufficient data for depth estimation with sufficient accuracy, we have to use other relevant data that have lower costs and can give us information in this regard. Between these methods, the kriging with an external drift method is an useful tool for estimation. In this study, sensitivity of the Kriging with an External Drift (KED) method are measured baxsed on its relevant parameters like as correlation coefficient between primary and secondary data and the portion of nugget effect in the structure of primary data. This study gives some thumb rules to use this method more efficiently. Two cases are studied; in first case-study, the available piezometric data as the main data and topographic data as supplementary data in Chanaran-Mashad have been used. Finally, the results obtained from cross validation of the used method(s) have compared with the Conventional ordinary kriging (OK) estimation. It has been observed that the kriging method with an external drift leads to better estimation than the conventional method of ordinary kriging. The second case consists of developing a multivariable KED to estimate the piezometric level of underground water table in Damghan-plain. The available piezometric data are assumed as primary data and topographic data and estimated depth by geo-electrical survey are taken as secondary data. In this study also the result OK has been compared with the result of KED with topography as secondary variable, and results of KED with the estimated depth as secondary variable. All of them are compared with results of multivariable KED. The results show the multivariable KED gives a better estimation than univariate forms of KED. The results of univariate forms of KED have the better results than OK. In this research the capability of KED estimators are examined and the method is applied on two real cases, in them, the results are improved in comparison with the results of routine method.