The cation exchange capacity (CEC) is an important property with respect to soil fertility and because of the way it influences physical and chemical behavior. The need for spatial information on CEC at the field level is increasing, particularly for its applications in precision agriculture and environmental management. Traditional methods of soil survey would only provide cursory data, with resultant maps likely to contain errors and thus make interpretation and soil management difficult. Geostatistical techniques have been used with some success to improve the accuracy of spatial prediction of CEC, especially those which utilize easy-to-obtain soil variables or ancillary data. In the latter case geophysical methods such as electromagnetic (EM) induction instruments may prove useful. In this paper we developed regression relationships between CEC and various EM signal data (i.e. EM38 and EM31) and compared the results of various methods of spatial prediction models (e.g. generalized additive models, ordinary-, regression-kriging) to determine an optimal method of interpolation. We do this with respect to accuracy, precision and bias of prediction in an irrigated field in the lower Namoi valley of New South Wales, Australia. We also compare and contrast the accuracy of prediction with increasing EM survey spacing with the optimal method of interpolation. The spatial distribution of CEC is closely correlated with soil mineralogy, which has implications with regard to soil salinisation and management. This is confirmed with x-ray diffraction analysis and mineralogical modeling.