Spatial data, including image data, are typically downloaded into GIS systems for processing purposes. The GIS data are optimized for establishing spatial relationships among objects. Spatial data can be produced rapidly from a variety of sources and the use of spatial data to improve agricultural management has become common [1]. However, most GIS systems are limited in their data mining capabilities. Data mining software provides advanced prediction capabilities for record-based data. The research goal of this project is to create a tool that would allow input of images and metadata, then process them using geospatial software to convert it to a record format such that data mining can be performed. This process opens the possibility of applying data mining techniques to agricultural data, for which such techniques are not yet in common usage. This paper proposes one such tool for classification of spatial data sets using J48 and Random Forest techniques.