Inversion of heavy metal copper content in soil-wheat systems using hyperspectral techniques and enrichment characteristics.

The growing problem of heavy metal contamination in soil will seriously threaten the China's grain safety. The development of hyperspectral remote sensing technology provides the possibility to achieve rapid and non-destructive monitoring of soil heavy metal content. In this study, we used hyperspectral techniques and enrichment characteristics to explore the potential of wheat leaf spectral inversion for heavy metal copper (Cu) content in the soil-wheat system. First, we conducted potting experiments to plant wheat on soil contaminated with varying concentrations of the heavy metal Cu. Then, we analyzed the migration characteristics, correlation characteristics and enrichment characteristics of Cu in the soil-wheat system under different soil heavy metal Cu concentration treatments. Next, we analyzed the spectral and correlation features of wheat leaves, and explored the potential of wheat leaf spectra for the inversion of Cu content in full-band and eigen-band modeling. Finally, using the estimated Cu content of wheat leaves from the best inversion model, we further conducted inversions to obtain the Cu content and precision of the grain, stem, root, total soil, and soil-available states based on the enrichment characteristics. The results showed that: (1) The accumulation pattern was root > grain > leaf > stem when the soil Cu concentration was <200 mg kg, and root > leaf > stem > grain when the soil Cu concentration was >200 mg kg. (2) The correlation coefficients between the different analyzed elements of the soil-wheat system were high, and all of them reached a highly significant level (P < 0.01). This supports the use of wheat leaves to estimate the Cu contents of soil and different parts of wheat. (3) The best inversion accuracies were obtained by modeling second derivative (SD) spectra that were pre-processed by screening the characteristic bands. The modeled R, RMSER and RMSE were 0.94, 2.72 mg kg, 0.91 and 3.64 mg kg, respectively. This indicates an excellent ability to estimate Cu content in wheat leaves. (4) Using the hyperspectral estimation of Cu content in wheat leaves and the grouped inversion of enrichment characteristics, the inversion accuracy was lower only for grains, and the R and R for stems and roots exceeded 0.90, those for total soil exceeded 0.85, and those for the soil available state exceeded 0.70. Therefore, it is possible to use the spectra of wheat leaves in combination with the inversion of enrichment characteristics to estimate the soil-wheat Cu content. This study provides guarantee and support for the detection of grain safety.