Monitoring stress levels on rice with heavy metal pollution from hyperspectral reflectance data using wavelet-fractal analysis

Remote sensing allows monitoring heavy metal pollution in crops for agricultural production and food security. This paper presents an approach towavelet-fractal analysis for exploring a set of sensitive spectral parameters to monitor the heavy metal pollution levels in rice crops from hyperspectral reflectance data. Hyperspectral and biochemical data were collected from three study farms in Changchun, Jilin Province, China. Our study explored the fractal dimension of reflectance with wavelet transform (FDWT) that demonstrated a better performance than other existing methods. Our results obtained in this study show that the red edge position (REP) was the most sensitive indicator for monitoring the heavy metal pollution levels in rice crops among common indices. As compared with REP, the FDWT is more sensitive ractal analysis to biochemical composition, namely with respect to chlorophyll concentrations, N, Cu and Cd. The established linear models showed a correlation coefficient (R2) above 0.70, model efficiency (ME) above 0.65 and a root mean square error (RMSE) below 3.5. Minimum FDWT values occurred in rice with Level II pollution followed by Level I pollution, and finally the safe level. This study suggests that wavelet transform is well suited as a spectral analysis method to eliminate noise and amplify the stress information from heavymetals. The wavelet transform in conjunction with fractal analysis is promising for detecting ss in heavy metal-induced stre