This paper compares the evaluation of three geostatistical interpolation methods including ordinary kriging, residual kriging and cokriging for the interpolation of long-term monthly and yearly reference crop potential evapotranspiration (ETo). This study has been conducted in a region including Fars, Booshehr, Hormozgan, and Kohgilooye-Boyrahmad provinces. Long-term mean values of monthly and yearly ETo were computed from recorded meteorological variables at 119 weather stations using the Hargreaves-Samani method. ETo estimates and estimation errors were evaluated at 19 validation stations. In general, estimates were in good agreement with observed values for residual kriging and cokriging methods. Based on mean absolute error (MAE), mean square error (MSE), mean error percent (MEP) and root mean square interpolation error (RMSIE), the best method for Farvardin (April) is kriging and for Khordad (June), Tir (July), Aban (November), and Azar (December) is cokriging. For other months and for mean annual ETo the best method is residual kriging. It should also be noted that MAE, MSE, and MEP for Mordad (August), Mehr (October), Dey (January), and Bahman (February) are very similar for cokriging and residual kriging. With the exception of Farvardin (April), Ordibehesht (May), and Shahrivar (September), for the other months and for annual ETo, the deviation of cokriging estimations from a 1:1 line is less than kriging and residual kriging. In other words, the points from these methods are more spread out around the 1:1 line, but the band of the deviation in cokriging is less than the two other methods. Therefore, the best method for estimation of monthly and yearly ETo is cokriging, except in Farvardin (April), Ordibehesht (May) and Shahrivar (September).