Search

Soil organic matter not only affects the stability of agricultural ecosystems but also is closely related to the carbon cycle of the biosphere. With Fuping County as an example, this study explores the relationship between soil organic matter content in a mountainous area with hyperspectral satellite imagery. This research was based on "Zhuhai-1" hyperspectral remote sensing imagery and actual measurement data of soil sampling points, explored sensitive band information through six spectral mathematical transformations, and constructed multiple linear regression models and random forest regression models based on sensitive band information. Support vector machine regression model inversion was used to estimate the soil organic matter content in the study area. In the selected spectral transformation method, the correlation with soil organic matter from high to low involved wavelet packet decomposition of low-frequency components, SG smoothing, wavelet packet decomposition of high-frequency components, original spectrum, logarithmic transformation, and logarithmic reciprocal transformation. The first four types of mathematical transformation data were selected for modeling. Among them, the SVM model constructed based on wavelet packet decomposition of low-frequency components had the highest accuracy, whose R2 reached 0.824. The Zhuhai-1 hyperspectral satellite imagery performed well in the application of soil organic matter retrieval modeling, which supports the application of large-scale remote sensing to retrieve soil organic matter. The research results provide a reference for implementing largescale soil organic matter hyperspectral remote sensing inversion and mapping and a basis for real-time monitoring of soil organic matter content. [ABSTRACT FROM AUTHOR]