When the particle size composition of agricultural soil can be determined by a rapid and simple method, it is possible to make site-specific management decisions and soil improvement according to soil variation, which is expected to improve yield, quality, and farmwork efficiency.
In this paper, we estimated calibration model performance for clay, sand, and silt (CLSSL), using soil diffuse reflectance spectra collected from paddy fields, and investigated the accuracy of predicted values. The accuracy of the calibration model for CLSSL was R2Cal>0.90. The predicted value accuracy for 185 unknown samples was 0.37 to 0.48 for R2. The predicted CLSSL values followed the distribution trend line of measured values, and the average value error was 0.60 to 1.41 %. Therefore, it was concluded that the variation in the relative quantitative value could be visualized in the form of a soil map. In field measurement of spectra, it is necessary to establish a method to appropriately exclude outliers from predicted values.
View full abstract