Machine learning techniques typically result from the need for intelligent solutions to practical tasks. Nowadays, large data volumes are usually involved and machine learning techniques are focused on particular tasks like classification, regression or clustering. For the latter task, clustering, quite a few algorithms have been proposed, typically tailored to particular application domains and their data sets. Recently, georeferenced (or spatial) data sets keep emerging in lots of disciplines. Therefore, algorithms which are able to handle these spatial data sets should be developed. This article shortly describes a particular application area, precision agriculture, and the spatial data sets which exist there. A particular task from this area, management zone delineation, is outlined and existing spatial clustering algorithms are evaluated for this task. Based on the experiences with those algorithms and a few requirements, HACC-SPATIAL is developed. The algorithm is based on hierarchical agglomerative clustering with a spatial constraint and it is demonstrated to produce practically advantageous results on precision agriculture data sets.