Predicting the abundance of copper in soil using reflectance spectroscopy and GF5 hyperspectral imagery

Abstract

Heavy metals in soil are harmful environmental pollutants due to their toxicity and copper (Cu) is a key indicator. Detecting and mapping the distribution of soil property using remote sensing technique is cost-effective. In this study, GF5 hyperspectral data and field spectroscopy were used to estimate soil Cu concentration of the farmland soils around the tailing reservoir of Tongkuangyu Copper deposit, Shanxi Province, China. Sixty-six soil samples were collected and their reflectance spectra were measured for modeling the relationship between Cu concentration in soil and visible-shortwave infrared reflectance (VIS-SWIR) reflectance data. Spectral index (2209 nm/907 nm) was found sensitive to Cu contents in soil and prediction was conducted by piecewise partial least square regression (P-PLSR). The coefficient of determination (R2) and residual prediction deviation (RPD) of the model developed from laboratory soil spectra were 0.89 and 2.81. The model was then applied to the GF5 satellite hyperspectral data and the spatial distribution of Cu content was mapped with acceptable accuracy R2 (0.83) and RPD (1.56). The result confirms that it is feasible to use satellite data to map soil copper anomaly quantitatively with simple spectral index. The study also demonstrates that GF5 hyperspectral data, due to its high spectral resolution (4–7 nm), good signal-to-noise ratio (SNR 200:1–100:1) and large swath coverage (60 km), can provide an alternative data source for quick and large-scale environment monitoring to support soil remediation policies.

Keywords

• Soil Cu content
• GF5 hyperspectral data
• Spectral index
• Piecewise partial least squares regression
• Quantitative prediction