Soil organic matter content estimation based on hyperspectral properties

Organic matter (OM) content is a suitable index of soil fertility which is widely used in field nutrient management. This study established spectral indices and derived equations for estimating soil organic matter (SOM) using hyperspectral technology. In the study, visible-NIR spectral reflectance of paddy and fluvo-aquic soils were measured using the ASD2500 device. Then by using dried soil samples from two different soil types, variations in the spectrum characteristics and sensitive wavebands in relation to changing OM content were determined. Then spectral index-based models were established for estimating SOM content. The results showed that under similar SOM content, changing trends of spectrum curves of different soil types exhibited no obvious difference, while their reflectance were different. The first derivative better described the spectrum curve peak. At sensitive wavebands of two soil types existed in similar spectral regions. The original spectral reflectance was negatively correlated with OM in visible-NIR ranges, with the highest significance at 685 nm. The first derivative spectrum had a significant negative correlation at 554 nm. Step-wise multiple regression analysis revealed that for all the calibrated samples, combined spectral bands of 800 nm, 1 398 nm and 546 nm well estimated SOM content of two types of soil. Furthermore, estimation model of SOM based on difference index (SOMDI) and the first derivatives of reflectance at 1 400 nm (R_FD_{1 400}) and 554 nm (R_FD₅₅₄) showed a better prediction performance; with a general equation of Y=4.19?12.85×(R_FD554?R_FD_{1 400}). The above monitoring models tested with independent datasets from two soil samples gave an R² = 0.79. This suggested that it was feasible to rapidly estimate SOM using hyperspectral technology.