Rapid prediction of atrazine sorption in soil using visible near-infrared spectroscopy.

Sorption is an important process for determining the fate, effects, and ecological risks of pesticides in terrestrial and aquatic environments. Within a watershed, soil properties vary greatly because of landscape and management practices, leading to spatial variation of pesticide sorption coefficients (Kd). A method for the rapid determination of the sorption variability of atrazine in soils of the Baima river catchment using visible near-infrared (Vis-NIR) spectroscopy is studied in this work. Partial least square regression (PLS) was used to build calibration models. To achieve optimum models, several methods of spectral preprocessing and variable selection were investigated. The results show that the combination of standard normal variant transform (SNV) and Monte Carlo uninformative variable elimination (MC-UVE) can significantly improve the model. For validation samples, the correlation coefficient between the predicted value and the reference value determined by high-performance liquid chromatography (HPLC) analysis is 0.8090. Moreover, positive correlations are observed between the pesticide adsorption coefficient and the organic carbon (OC) and total nitrogen (TN) contents, respectively. Prediction models for OC and TN were built. The correlation coefficients of OC and TN between the predicted values and the reference values are 0.9285 and 0.6599, respectively. The results show that Vis-NIR can be used as a rapid and simple method to predict soil composition and pesticide sorption.