Potential of visible-near infrared spectroscopy for the determination of three soil aggregate stability indices

Aggregate stability (AS) is an important parameter to evaluate soil resistance to erosion. The conventional determination methods to measure AS are time consuming, difficult and labour intensive. Visible (vis) and near infrared (NIR) spectroscopy could be a better alternative to the conventional determination methods of AS. This study explored the possibilities of estimating three AS indices, reflecting stability upon slow wetting (SW), fast wetting (FW) and mechanical breakdown (MB), using vis-NIR spectra data on some air-dried, non-sieved soils of the Belgian loam belt. Partial least squares regression (PLSR) was used to build calibration models for the three stability indices, using a calibration set accounting for 70% and a validation set of 30% of the total samples. Results showed that all three AS indices can be predicted to appreciable accuracies from vis-NIR-PLSR models [coefficient of determination (R2) = 0.72–0.80, residual prediction deviation (RPD) = 1.93–2.27, ratio of performance to interquartile range (RPIQ) = 2.23–4.09 and root mean square error (RMSE) = 0.29–0.52 mm]. The prediction results suggest that omission of sample pre-treatment by sieving or grinding may have very limited impact on the prediction accuracy. This opens up opportunities for the in-situ deployment of vis-NIR spectroscopy, provided that problems associated with variable soil moisture contents can be overcome.