Innovative approach for determining polypropylene microplastics pollution in calcareous soils: Vis-NIR spectroscopy

• Polypropylene (PP) is a common type of microplastic in polluted agricultural soils • The study aimed to rapidly predict PP microplastic content in calcareous soils • Increasing soil PP microplastic content caused a decrease in NIR reflectance spectra • Vis-NIR spectroscopy accurately predicted soil PP content using PLSR, PCR, and MLR • MLR-based STF using 10 Vis-NIR spectra predicted soil PP content with R 2 CV of 0.91 Polypropylene (PP) plastic material is widely used in food packaging and agricultural-related tools, and is a major source of microplastics that degrade into agricultural soils and the environment. Generally, measuring soil microplastics content is laborious, expensive, and time-consuming. Therefore, in the present study, we attempted to indirectly predict the soil PP microplastic content using visible (Vis) and near-infrared (NIR) reflectance spectra by applying the partial least square regression (PLSR), principal component regression (PCR), multiple linear regression (MLR), and support vector regression (SVR) models. The experiment was performed using ten calcareous soils of diverse and varied ranges of initial characteristics collected from Fars Province, Iran. The soils were polluted with varying concentrations of PP microplastics (0-5%wt), based on a normal distribution to obtain ten polluted subsamples for each of the ten studied soils (total of 100 samples). Results illustrated the strong potential of Vis-NIR spectroscopy for predicting soil PP microplastic content in calcareous soils with R 2 CV (coefficient of determination related to leave-one-out cross-validation) values of 0.92 and 0.92, and RPIQ CV (ratio of performance to interquartile range related to leave-one-out cross-validation) values of 4.62 and 4.68 when, respectively, PLSR and PCR were used as predictive models. A 10-variable MLR-based spectrotransfer function, STF (which actually is a kind of pedotransfer function in which only spectral bands are considered as predictors), was derived with R 2 CV and RPIQ CV values of 0.91 and 4.31, respectively, using reflectance values at 448, 528, 1082, 1415, 1724, 1913, 2010, 2221, 2302, and 2345 nm wavelengths as effective and key spectral bands for predicting soil PP microplastic content. However, the SVR method presented lower performances with R 2 CV and RPIQ CV values of 0.89 and 3.88, respectively. Generally, the developed MLR-based STF is simple and practical, and it can be tested and applied to predict PP microplastic content in soils under various conditions.