2016 Volume 78 Issue 5 Pages 401-415
In a previous study, we estimated multivariate regression models (MRM) for 12 soil properties, including chemical and biological (soil organic matter) properties, and soil maps were created by using a tractor-mounted soil analyzing system. The purpose of this study was to estimate MRMs for 25 soil properties, including exchangeable potassium and dry density, to evaluate the accuracy of the models by comparison with previous studies, and to create soil maps.
MRM estimations were applied to second-derivative pretreatment and partial least-squares regression analysis (Full Cross Validation) using a total of 334 data. The accuracies of 25 MRM were obtained, with R2 from 0.82 to 0.90 and residual prediction deviations from 2.35 to 3.25. We were able to show 25 soil properties on high-resolution soil maps. These included the three major nutrients in fertilizer, along with biological and physical information, enhancing the value of soil management information.