Analysis and interpretation of soil survey data are very important for effective management of agricultural fields. In this study, kriging and cokriging methods were applied to estimate the spatial distribution of soil properties from available large-scale survey data of Taiwan. The data were derived from soils in a 10-km(2) area divided into 250 m x 250 m node intervals. The soil properties examined included the extractable P, Ca, Mg, and Fe contents, the sum of exchangeable bases (SEE), %sand, %silt, and %clay. The sum of exchangeable bases and particle-size distribution were regarded as the primary and auxiliary variables, respectively, in the cokriging procedure. The ratio of nugget to total variation was about 57 to 80%, indicating that the spatial correlation of the tested soil properties at the large scale was moderately (cross-)dependent. The estimated spatial distributions of the soil properties by kriging, under decreasing sampling densities, all correlated significantly (P < 0.1%) with those obtained from original data. Furthermore, with the over-sampled particle-size distribution, the overall estimation of SEE quality by cokriging was superior to that by kriging. The results suggested that by kriging and cokriging, the existing sampling density could be decreased under the large-scale sampling interval by nearly half and that sufficient spatial information about the soil properties could still be retained. The information obtained could be used to improve the long-term sampling designs of soil surveys in Taiwan. It also may be useful for identifying the appropriate sampling densities for these scales of soil surveys.
