Geostatistical analysis of soil chemical properties and rice yield in a paddy field and application to the analysis of yield-determining factors

Abstract

To obtain basic information for rational site-specific soil management for rice production, spatial variability of soil chemical properties and grain yield of rice was evaluated in a 50 m × 100 m paddy field. One hundred surface soil samples were collected from each of the 5 m × 10 m plots before puddling to investigate the spatial variability of their chemical properties: pH, EC, total C content, total N content, C/N ratio, contents of mineralizable N, inorganic N, available P, exchangeable Ca, Mg, K, and Na. Grain yield was also measured at harvest for the corresponding 100 plots. Geostatistical analysis was carried out to examine their within-field spatial variability using semivariograms, and multivariate analysis was also carried out to evaluate yield-determining factors. Geostatistical analysis of the soil properties indicated a high to moderate spatial dependence for all the properties except for the inorganic N content. The ranges of spatial dependence were about 20–30 m for the pH, EC, total C content, total N content, content of exchangeable Na, about 40 m for the contents of available P, mineralizable N, exchangeable Ca and Mg, and about 50–60 m for the C/N ratio and content of exchangeable K. Grain yield showed a moderate spatial dependence with a range of about 50 m. The results of spatial dependence enabled to prepare kriged maps of the soil properties and yield to compare their spatial distribution in the field. Multivariate analysis further showed, in combination with geostatistics, that the soil chemical properties contributed significantly to the yield as yield-determining factors and explained as much as 65% of the spatially structured or non-random variation of the yield. In conclusion, the possible benefit of site-specific soil management or precision agriculture was demonstrated even in an almost flat paddy field.