Vol. 8, No. 1, 26-39, 2009

Abstract
The aim of this work was to evaluate prediction maps created with interpolation for five common topsoil properties, namely organic matter, total CaCO₃, electric conductivity, Fe content, and clay content in a Mediterranean agricultural system. Knowledge about spatial variance of soil properties is of significant importance when implementing "good farming rules" towards sustainable rural development. A 25 ha block of fields cultivated with grape vines and olive trees in the island of Crete was targeted as a study site. 106 topsoil samples were collected on a 50 × 50 m² grid and then analysed in the laboratory. Three well-known spatial interpolation techniques, namely Ordinary Kriging (OK), Inverse Distance Weighting (IDW), and Radial Basis Functions (RBF) were applied for generating the prediction maps, which then were assessed for their accuracy and effectiveness with a new, independent set of samples. This time the samples were collected randomly. The results indicated that there was not a method clearly more accurate than the other methods for any of the tested properties. The Goodness of Prediction (G) had positive values for Fe content and total CaCO₃ for all interpolation techniques and positive values for Organic Matter only for IDW. Among G values, only the value for total CaCO₃ was big enough for all techniques, indicating that only this specific property counted for mapping. It is believed that the topography, crop alteration and farming practices significantly affected the results in the specific landscape. Future research will focus on taking account of between-field variations that are due to different crops or farming practices, incorporating segmentation techniques on cadastral data and satellite imagery.

ISSN 1729-3782