Feasibility of Using Rice Leaves Hyperspectral Data to Estimate CaCl-extractable Concentrations of Heavy Metals in Agricultural Soil.

Heavy metals contamination is a serious problem of China. It is necessary to estimate bioavailability concentrations of heavy metals in agricultural soil for keeping the food security and human health. This study aimed to use hyperspectral data of rice (Oryza sativa) leaves as an indicator to retrieve the CaCl-extractable concentrations of heavy metals in agricultural soil. Twenty-one rice samples, soil samples and reflectance spectra of rice leaves were collected, respectively. The potential relations between hyperspectral data and CaCl-extractable heavy metals (E-HM) were explored. The partial least-squares regression (PLSR) method with leave-one-out cross-validation has been used to predict concentrations of CaCl-extractable cadmium (E-Cd) and concentrations of CaCl-extractable lead (E-Pb) in farmland soil. The results showed that the concentrations of E-Cd in soil had significant correlation with concentrations of Cd in rice leaves; the number of bands associated with E-Cd was more than that of E-Pb. Four indices (normalized difference vegetation index (NDVI), carotenoid reflectance index (CRI), photochemical reflectance index 2 (PRI2), normalized pigments chlorophyll ratio index (NPCI)) were significant (P < 0.05) and negatively related to the E-Cd concentrations. The PLSR model of E-Cd concentrations performed better than the PLSR model of E-Pb concentrations, which with R = 0.592 and RMSE = 0.046. We conclude that if the rice was sensitive to E-HM and/or the crop was stressed by the E-HM, the hyperspectral data of field rice leaves hold potentials in estimating concentration of E-HM in farmland soil. Therefore, this method provides a new insight to monitoring the E-HM content in agricultural soil.