Predicting copper contamination in wheat canopy during the full growth period using hyperspectral data

The rapid and efficient determination of heavy metal content in food crops is essential for human health and environmental protection. The use of hyperspectral data has become a popular way to predict heavy metal content in plants; however, many challenges remain. One challenge is that lab conditions differ from actual agricultural production conditions. Another challenge is that spectral data characteristics are not universally applicable to all situations. Therefore, in this study, the field test method was adopted to conduct experiments during the full growth period of wheat, and the spectrum data of wheat canopy were processed by the first derivative method to screen-sensitive spectral bands as the basis for the prediction model of the copper content in wheat. The results showed that the copper content increased with an increase in the soil copper content, and there were dissimilar subtle differences in the spectral reflectance of wheat canopy under different stressed soil copper concentrations; sensitive spectral indices and wavelengths were screened based on good correlation with the copper content in the wheat canopy. Different optimal predicting models in different periods were built and verified. The established linear regression models, which were based on NDVI/SIPI and W728, were the most suitable predicting models during the tillering stage with R² = 0.669 and 0.818; Rg, W741, and multiple bands were the most suitable predicting models during the jointing stage with R² = 0.548, 0.830, and 0.868; the optimal model during the heading stage was based on W480 (R² = 0.625). This study demonstrated that the constructed models had good potential for estimating the copper content in wheat leaves during full growth periods, and this method had the potential to be applied to the actual agricultural production process.