A column generation approach for delineating rectangular agricultural management zones
In this contribution we present a column generation approach for solving a two-stage stochastic program that delineates a field into site-and-time specific rectangular management zones to be used in a crop planning problem. Given a set of potential management zones, the model provides a partition of the field seeking that the selected zones be homogeneous according to a soil property, minimizing the total number of zones and considering the soil property by a finite number of scenarios. The first-stage problem chooses a partition that minimizes the number of zones and once the realization of the soil property is known, the second-stage problem minimizes the non-compliance of a given limit of homogeneity within the partition according to each scenario. The decomposition strategy defines as Master Problem the stochastic program and considers a Subproblem that at each iteration provides a potential zone that is added to the (restricted) Master Problem until to reach the stop criteria.
Keywords: Column Generation O.R. in Agriculture Stochastic Optimization Two-stage stochastic program