Comparison of kriging and inverse-distance methods for mapping soil parameters

Variable-rate technology may provide a means of increasing fertilizer use efficiency by matching applications to specific conditions at a given field location. Effective implementation of this technology depends on accurately characterizing the spatial variability of soil parameters used to define the application rate. Kriging and inverse-distance-squared are two commonly used techniques for characterizing this spatial variability and interpolating between sampled points. To assess the accuracy of these techniques, data sets obtained from grid sampling two field research sites were used in a prediction-validation comparison of ordinary kriging and inverse-distance methods using powers p = 1, 2, and 4. The accuracy of the inverse-distance methods tended to increase with the power of distance for data sets with a coefficient of variation less than about 25% (typical of soil organic matter). However, for data sets with greater variation (such as soil NO3-), inverse-distance prediction methods using high distance powers (2 or 4) can give very inaccurate predictions. The accuracy of predictions from kriging was generally unaffected by the coefficient of variation, and was relatively high for all of the sampling configurations considered in this study. These tendencies were also observed using 48- and 72-m subsamples, although the use of wider sampling spacings greatly reduced the information in the maps constructed by each method. Careful thought should be given to the choice of sample spacing and interpolation method to be used before data are collected. Summary statistics, and the coefficient of variation in particular, are simple measures that ran give an indication of the relative accuracy of the inverse-distance and kriging mapping approaches.