In this article we cover the problem of generating a partition of an agricultural field into rectangular and homogeneous management zones or quarters according to a given soil property, which has variability in time that is presented as a number of possible scenarios. This problem combines aspects of precision agriculture and optimization with the purpose of achieving a site and time specific management of the field properties that is consistent and effective in time for a medium term horizon. More specifically, we propose a two stage integer stochastic linear programming model with recource that solves the problem of generating a partition facing a finite number of future scenarios, with a solution that gives satisfactory results to any possible value of the chosen soil property. We describe the proposed model, the adopted methodology and the results achieved with this methodology.