Analysing the effects of applying agricultural lime to soils by VNIR spectral sensing: a quantitative and quick method

Abstract

The potential of visible and near-infrared (VNIR) diffuse reflectance spectra to predict the chemical properties of Ferralsols and Arenosols cultivated with maize during four crop cycles were evaluated. The study was carried out in a greenhouse and aimed to (i) evaluate soil chemistry variation induced by plants and the application of lime with different degrees of reactivity using conventional methods and proximal soil-sensing techniques, (ii) identify the wavelength ranges related to soil chemistry changes, and (iii) construct models that predict soil chemistry attributes using soil VNIR spectra. Treatments used were three lime rates applied to raise the base saturation to 40%, 60% and 80% and one control. Partial least squares regression with cross-validation was used to establish relationships between the VNIR spectra and the reference data from chemical analyses. The predicted results were evaluated based on the values of coefficient of determination (R2), the ratio of the standard deviation of the validation set to the root mean square error of cross-validation (RPD), and the root mean square of prediction. The predicted results were excellent (R2 > 0.90 and RPD > 3) for potassium and for the lime requirement calculation. Good predictions (0.81 < R2 < 0.90 and 2.5 < RPD < 3) were also obtained for pH and sum of bases. The resulting models for exchangeable calcium, cation exchangeable capacity, and base saturation had moderate predictive power (0.66 < R2 < 0.80 and 2.0 < RPD < 2.5). Our findings suggest that VNIR reflectance spectroscopy could be used as a rapid, inexpensive, and non-destructive technique to predict some soil chemistry properties for these soil types. As this methodology evolves, it may eventually permit real-time analyses of soil variability and real-time management responses via sensors installed on tractors.