Discriminating copper and lead contamination in crops using leaves spectra

Discrimination on the type of heavy metal contamination can contribute to the detection of diseases that caused by heavy metals, thereby to protect human health. This study aims to propose a Copper and Lead Contamination Discriminating Features (CLCDF) that constructed from leaf spectra to discriminate between copper and lead contamination in crops. Firstly, a typical agricultural corn was taken as an experiment object, and then to cultivate in soils that were contaminated with copper or lead. The soil-based contamination levels were set: 50, 100, 150, 200, 300, 400, 600, and 800 mg/kg. Secondly, an ASD FieldSpec 4 spectrometer was used to measure the spectral reflectance data of plant leaves. The continuum removal (CR) and derivative (D) were selected for spectral preprocessing (in this study, D processing with derivation window radius of 1, 2 and 3 nm was performed respectively), and obtained CRD1, CRD2 and CRD3 spectra. The Normalized Difference Copper and Lead Contamination index (NDCLCI) was finally extracted using CRD1, CRD2 and CRD3 spectra, respectively. There was a high correlation between the NDCLCI and the types of heavy metal contamination from the leaves. The CLCDF of leaves was also constructed using NDCLCI. There were significant differences in the location of CLCDF when the types of heavy metal contamination were different. A Discriminating Plane (DP) was established in the domain of CLCDF distribution area for partitioning the domain of copper and lead contaminated. According to the relationship between the position of CLCDF and DP, the rules were obtained for the visual discrimination of copper and lead contamination types. The Discriminating Distance (DD) was used to quantify the discriminating. Using the training set data, a total of 189 DPs were obtained with the discriminating correct rate of 100%. The visual discriminating rules were: the left DP was the copper contamination domain, where the domain of CLCDF corresponding to the blade contaminated with copper, while, the right DP was the lead contamination domain, where the domain of CLCDF corresponding to the blade contaminated with lead. The quantified discriminating rules were: when DD<0, the leaves corresponding to CLCDF were contaminated with copper, and when DD>0, the leaves corresponding to CLCDF were contaminated with lead. Using the validation set data to verify the effectiveness of the discriminating methods, 88 out of 189 DPs were obtained with the discriminating correct rate of 78.22%. The results showed that the CLCDF was valid and reliable to discriminate the copper and lead contamination. A discriminating error can be due to the insignificant changes in the leaf spectra, resulting from the different heavy metal contamination and small fluctuations in leaf CLCDF. The CLCDF has great potential and feasibility for other types of heavy metals contamination in crops. The complex discrimination of contamination types was turned into a simple threshold discriminating by quantifying the detection rule. This shift has made it more feasible to discriminate the contamination types of heavy metal at the canopy and pixel scales using the CLCDF combined with satellite or unmanned aerial imagery.