Assessing Organic Carbon of Grassland Soil in the Northern Tianshan Mountains of Xinjiang, China Using the Wavelet Decomposition of Hyperspectral Data

Soil organic carbon (SOC) represents a significant fraction of the total amount of carbon involved in the global carbon cycle. Hyperspectral remote sensing has a valuable role in the monitoring of the dynamics of SOC. This study focused upon improving the accuracy of SOC quantification by applying wavelet analysis to reflectance spectra. Spectral measurements for all soil samples (three sub-regions in the northern Tianshan Mountains, China) were performed in a controlled laboratory environment. The results demonstrated that by decomposing soil spectra, the resultant wavelet coefficients can be used to generate higher R² with SOC contents (R² >;0.95) compared to reflectance spectra (R² <;0.63) and derivative reflectance (R² ≤0.8). In addition, the selection of optimum scales and wavelengths play a key role in analyzing SOC using continuous wavelet transform. In this study, the optimum correlation between wavelet coefficients and SOC contents were appeared in scale 50-100 at around 2300 nm, and derivative reflectance may be more suitable as input to wavelet analysis than reflectance spectra. These results provided an insight for studying the global carbon cycle by predicting the changes of C in terrestrial ecosystems using hyperspectral remote sensing data.