Estimation of fluvo-aquic soil organic matter from hyperspectral reflectance by using discrete wavelet transformation

Abstract

The estimation of soil organic matter (SOM, %) is an important issue for agricultural production, but common methods for estimating the SOM is very expensive and time-consuming. Recently, hyperspectral diagnosis technology has showed great potential in measuring SOM due to its rapid, nondestructive, reproducible and cost-effective characteristics, thus four common spectral analysis methods had been used to estimate the fluvo-aquic soil organic matter (SOM< 2%), but the highest determination coefficient (R2) was only 0.09. As a novel method, the discrete wavelet transformation was performed on each of the soil reflectance spectra. The hyperspectral reflectance removed by hull curve were decomposed into 4 scales, the modulus maxima of the detail signals in visible and near infrared bands (560nm–660nm, 940nm–1060nm) were chose as a novel spectral indice. The estimation model had been built effectively by using this spectral indice, the R2 between this indice and SOM can reach to 0.83. These results provide new insights into the role of soil absorption features in the visible and near infrared bands for the accurate spectral estimation of SOM, it may be extended to the estimation of other soil nutrient such as nitrogen, phosphate and potassium.