Ordinary kriging (OK) is one of the most suitable interpolation methods for the construction of thematic maps used in precision agriculture. However, the use of OK is complex. Farmers/agronomists are generally not highly trained to use geostatistical methods to produce soil and plant attribute maps for precision agriculture and thus ensure that best management approaches are used. Therefore, the objective of this work was to develop and apply computational routines using procedures and geostatistical libraries in the R software, to automatically identify the best parameters for OK interpolation. Through the implemented procedures, six different models (spherical, Gaussian, exponential, Matérn 1.0, Matérn 1.5 and Matérn 2.0) and two statistical methods of optimizing the semivariogram (OLS - ordinary least squares and WLS - weighted least squares) were analyzed. We tested 25 values for the semivariogram parameters for each of the 12 models, totalizing 300 different configurations to identify the best parameters used to measure data by OK. To validate the procedure the routines were applied to corn and soybean crop data collected for three years in an agricultural area with approximately 20.9 ha and 73 sampling points. For soybean yield, the best fit model for the experimental semivariogram was Gaussian-WLS and for corn yield was Matérn 1.5-OLS. Other models presented similar adjustments to those chosen as better, emphasizing the quality of the geostatistical procedures implemented that were able to identify, without tendencies, the best adjustments for semivariogram. The obtained parameters were used in the interpolation process by OK onto a 5-m grid (pixels of 25 m2). After the interpolation was processed, thematic maps were generated for each crop. By means of the interpolated maps generated by OK, it was possible to identify regions in which the variations in the average yield occur in both crops, thus being possible to perform differentiated management in these locations. It can be concluded that the computational routines implemented were efficient and able to identify the best fit for the semivariogram to be used when applying OK.