Spatial variation of soil temperature is a measurement of great help to analysis in precision agriculture. These measurements are used to create yield prediction models, diagnosis of diseases and reveal high or low concentrations of temperature. The present paper carries out a study of different methods to predict the variation of temperature in soil using geostatistical tools based on georeferenced data in a crop. Three interpolation techniques were tested to map soil properties and thus compare the accuracy of the prediction. Given the data generated, heat maps were created referring to the variables calculated by the different methods. The methods were compared using a performance criterion that included root mean square error (RMSE) and mean absolute error (ME). This comparison showed that the kriging method was the most accurate for the interpolated soil temperature, the RMSE of the method was the lowest among the others, and ME was very close to 0, suggesting that the predictions are unbiased.