Prediction of soil phosphorus sorption capacity in agricultural soils using mid-infrared spectroscopy. 

Abstract

             Advice for phosphorus (P) fertilisation based on soil testing using extractive methods but does not consider P sorption processes. Traditional soil P sorption capacity examined from a Langmuir isotherm batch experimental design, which is time-consuming, labour intensive and expensive. Mid-infrared (MIR) spectroscopy is a rapid analysis technique that can potentially replace the extractive technique traditionally used in soil analysis. The objective of this work was to predict the isothermal parameter of P sorption maximum capacity (Smax, mg·kg-1) from MIR spectroscopy.              This study created spectral libraries from benchtop (Bruker) and handheld (Agilent) MIR spectrometers by scanning samples in two particle sizes, < 0.100 mm (ball-milled) and < 2 mm. The four spectral libraries created used an archive of samples with a database of sorption parameters where soils were classified into low and high sorption capacities using a threshold value of Smax = 450.03 mg·kg-1. To assess the optimal algorithmic method with highest Smax prediction accuracy, regression models were based on the partial least squares (PLS) regression, Cubist, support vector machine (SVM) regression and random forest (RF) regression algorithms. After the first derivative Savitzky-Golay smoothing, Bruker spectroscopies with both soil particle sizes yielded ‘excellent models’, with SVM predicting Smax values with high accuracy (RPIQVal = 4.50 and 4.25 for the spectral libraries of the ball-milled and <2mm samples, respectively). In comparison, the Agilent handheld spectrometer produced spectra with more noise and less resolution than the Bruker benchtop spectrometer. Unlike Bruker, for Agilent MIR spectroscopy, more homogeneous samples after ball-milling resulted in a higher accurate Smax prediction. For Agilent spectroscopy of ball-milled samples, an ‘approximate quantitative model’ (RPIQVal = 2.74) was obtained from the raw spectra using the Cubist algorithm. However, for Agilent spectroscopy of < 2 mm samples, the best performing Cubist algorithm can only achieve a ‘fair model’ (RPIQVal = 2.23) with the potential to discriminate between high and low Smax values.              The results suggest that the Bruker bench-top spectrometer can predict the Langmuir Smax value with high accuracy without the need to ball-mill samples, highlighting the availability of the MIR spectrometer as a rapid alternative method for understanding soil P sorption capacity. However, for handheld spectrometers, the Agilent instruments can only make approximate quantitative predictions of Smax for ball milled samples. For <2mm samples, Agilent can only be used to classify low and high sorption capacity soils.