Hierarchical spatial clustering for management zone delineation in precision agriculture

Abstract. Precision agriculture is concerned with the application of current technology in the field of agriculture. Huge data sets are nowa-days collected during standard farming operations. These data are fine-scale and available in high resolution, usually reflecting the heterogeneity of any agricultural field. The data may result from a multitude of sen-sors and electronic equipment and can be used for a number of purposes which affect the efficiency and effectiveness in farming operations. An important task in classic agriculture is base fertilization. This term is commonly used to describe the process to make minerals like potassium (K), phosphor (P) and magnesium (Mg) available for the planted crops. Since the field is usually heterogeneous, the question arises which part of the field should be treated and how it should be treated. This is usually associated with the concept of management zone delineation. There have been quite a number of approaches towards using fine-scale data for sub-dividing the field into a small number of zones. These approaches often require multi-year data sets and are based on high-resolution sampling methods for data acquisition. Existing research into the semi-automatic generation of management zones usually applies a variety of clustering methods to subdivide the field into homogeneous parts. Most of the ex-isting approaches use only parts of the available data for the clustering process. In this paper, we propose a novel approach which tackles the above issues. We base our work on a site-year of high-resolution spatial data from selected wheat fields in Germany. Our approach may best be described as divide-and-conquer: we split the field into homogeneous subfields and consecutively merge these subfields using a hierarchical agglomerative clustering approach with a spatial constraint