Soil libraries represent an invaluable resource in terms of research, management or planning. However, the access to such libraries and selection of soils is often tedious and time consuming thus limiting their usefulness. In this study we propose an ontology-based approach for an efficient and intuitive selection and classification of soils. For this test, a soil library of 458 soils from Australia was used. An ontology was then developed to model the fundamentals concepts and relationships found in the data. The basic capabilities of the ontology are shown to select samples with certain values for a number of attributes. In addition, an inference process known as realization is tested to automatically assign individuals to concepts, in our present case to a soil texture class or soil order (the latter known as soil classification). Results show the potential of ontology approaches to select samples from large libraries in an efficient and intuitive way. In addition, and through the use of reasoning processes, we were able to classify soils from different orders and textural classes with accuracy higher than 80% in most cases. This represents an additional application of ontology approaches to produce hidden data from the original data set.