Rapid retrieval of cadmium and lead content from urban greenbelt zones using hyperspectral characteristic bands

Greenbelts around roads are an essential part of the ecosystem that can reduce heavy metal contamination from traffic and contribute to sustainable development. However, only limited studies have examined the rapid detection of heavy metal content in urban road greenbelt zones. In the present study, soil spectral reflectance was measured using a FieldSpec-3 portable handheld ground object spectrometer to measure cadmium (Cd) and lead (Pb) content from 23 roads in the municipality of Chongqing, China. After smoothing the spectral data using Savitzy-Golay (SG), the first derivative (FD), second derivative (SD), and multiple scattering correction (MSC) procedures were applied. Pearson correlation was used to select the characteristic bands. Cd and Pb amounts were inverted using multiple linear regressions (MLR), partial least squares regressions (PLS), and the backpropagation neural network models. It was found that the Cd and Pb levels in urban road greenbelt zones were higher than the soil background values. The Nemero comprehensive pollution index and potential ecological risk index indicated modest Cd and Pb contamination on sampled roads. In the Cd content inversion models, the MLR and PLS models pretreated with SG-FD provided high prediction accuracy, with determination coefficients of >0.60. For the Pb content inversion models, the PLS model processed by SG-MSC had the best prediction accuracy, with determination coefficients of >0.66. It was shown that the MLR and PLS models processed by SG-FD could be used to forecast soil Cd concentration in urban road greenbelt zones. Comparatively, the PLS model processed by SG-MSC could be used to estimate soil Pb levels. Additionally, these results demonstrate how heavy metals in urban greenbelt zones can be inverted using hyperspectral imaging globally.