In recent years, the world's population has been growing, reaching about 7.7 billion in 2018. In the next five years, the world's population will increase to about ten billion, thus increasing the demand for food, water resources and fertile land for food production. Therefore, precision agriculture is trying to improve the quantity and quality of agricultural products by using new technologies. Precision farming is a new concept in modern agriculture and is based on the existence of heterogeneity at the farm level. One of the reasons for the growth of precision agriculture among scientists and farmers is the advancement of technology in various fields such as global coordinate system, imaging sensors and spatial information management tools. The use of remote sensing allows us to control a very large space in a very short time, which will reduce heavy costs. Along with rice and wheat, potatoes are the third most important food crop in the world, fed by more than one billion people, and the demand for this crop has increased as the population has grown; Therefore, better methods of product protection and management need to be used to improve production. In order to build the whole model, the images of the region were considered during the months of March to October in a period of two years, and pre-processing such as removing the cloud and other errors were done in order to have images with the least error. The NDVI index was then calculated using red and infrared bands (Figure 3). By calculating the NDVI index, useful information about the plant such as health, plant vigor, presence or absence of fall, and proper irrigation were obtained. On the other hand, land information, agricultural lands have been collected through questionnaires. For modeling, we deal with two sets of terrestrial and satellite data. Separate use of each of them will be associated with shortcomings and shortcomings; Therefore, for modeling, a combination of terrestrial and satellite information was used to build a model with minimal deficiencies and defects. It is common for modeling to work repetitively to obtain the best algorithm, so linear regression is the best choice for modeling. Univariate method was used for modeling and the date of 28 June was selected as the modeling interval. Because using multivariate methods, due to the fact that there was a high correlation in the information of different time periods, a lot of duplicate information was generated, and for this reason, we used the univariate method. Four machine learning algorithms were used for modeling, which Gaussian methods and support vector showed the best results. According to the production estimation model, it is possible to predict the amount of potatoes produced in each agricultural land and its health status with 50% accuracy three months before the harvest. The results of this study showed that estimation models will provide very valuable information for better management of crops to be used by agricultural planners to avoid wasting resources and prevent imports.