Prediction of heavy metals in soils of an arid area based on multi-spectral data.

With the rapid and extensive development of industry and agriculture, the soil environment inevitably becomes contaminated with heavy metals, thus creating adverse environmental conditions for flora and fauna. The traditional methods for combining field sampling with laboratory analysis of soil heavy metals are limited not only because they are time-consuming and expensive, but also because they are unable to obtain adequate information about the spatial distribution characteristics of heavy metals in soil over a large area. Three hundred and ninety-four soil samples (Gobi and farmland) were collected in an arid area in Jiuquan in Northwest China and analyzed for elements concentrations. Based on these measured concentrations, as well as rapid and environmentally friendly remote sensing (multi-spectral data), stepwise multiple linear regression (SMLR) and partial least-squares regression (PLS) were combined to predict concentrations and distributions of heavy metals in the soils of the study area. Furthermore, laboratory data were used to assess the accuracy of the prediction results. Obtained results suggest that the SMLR and PLS models were able to predict the metals contents in the study area. The concentrations of Cr, Ni, V and Zn could be predicted by two regression models, while those of Cu and Mn were predicted more accurately when they were attached to the SMLR model. The spatial distribution of heavy metals derived from the two models is consistent with measured values, indicating that it is reasonable to predict the concentrations of heavy metals in the soil of the study area using the multi-spectral data.