Modeling and identification of fertility maps using artificial neural networks

Abstract

The application of agricultural fertilizers using variable rates along the field can be made through fertility maps previously elaborated or through real-time sensors. In most of the cases applies maps previously elaborated. These maps are identified from analyzes done in soil samples collected regularly (a sample for each field cell) or irregularly along the field. At the moment, mathematical interpolation methods such as nearest neighbor, local average, weighted inverse distance, contouring and kriging are used for predicting the variables involved with elaboration of fertility maps. However, some of these methods present deficiencies that can generate different fertility maps for a same data set. Moreover, such methods can generate inprecise maps to be used in precision farming. In this paper, artificial neural networks have been applied for elaboration and identification of precise fertility maps which can reduce the production costs and environmental impacts.