Recent studies have extended the scope of data mining from relational and transactional databases to spatial databases. Spatial data mining is a promising field, where the research work on spatial data classification is still in its initial stage. In this paper, the advantages and the disadvantages of several existing methods of spatial data classification are compared first. Then an effective three-step method, which is based on the rough set theory for spatial data classification, is proposed. The validity of this algorithm to the problem of incomplete spatial information is verified by pertinent experimental results.