Predicting nickel concentration in soil using fractional-order derivative and visible-near-infrared spectroscopy indices

Accurate monitoring and estimation of heavy metal concentrations is an important process in the prevention and treatment of soil pollution. However, the weak correlation between spectra and heavy metals in soil makes it difficult to use spectroscopy in predicting areas with a risk of heavy metal pollution. In this paper, a method for detection of Ni in soil in eastern China using the fractional-order derivative (FOD) and spectral indices was proposed. The visible-near-infrared (Vis-NIR) spectra were preprocessed using the FOD (range: 0 to 2, interval: 0.1) to solve the problems of baseline drift and overlapping peaks in the original spectra. The product index (PI), ratio index (RI), sum index (SI), difference index (DI), normalized difference index (NDI), and brightness index (BI) were applied and compared. The results showed that the spectral detail increased as the FOD increased, and the interference of the baseline drift and overlapping peaks was eliminated as the spectral reflectance decreased. Furthermore, the FOD extracted the spectral sensitivity information more effectively and improved the correlation between the Vis-NIR spectra and the Ni concentration, and the NDI had a maximum correlation coefficient (r) of 0.803 for order 1.9. The estimation model based on the NDI dataset constructed after FOD processing had the best performance, with a validation accuracy RP2 of 0.735, RMSE of 3.848, and RPD of 2.423. In addition, this method is easy to carry out and suitable for estimating other heavy metal elements in soil.