Abstract: Soil organic matter (SOM) is an important soil health and fertility indicator. Accurate estimation of SOM is crucial for sustainable agricultural development. Hyperspectral technology combined with machine learning provides the possibility for large-scale and non-destructive monitoring of SOM. However, existing studies have not fully explored the physical significance of spectral positions within the sensitive spectral range of soil active materials and the spectral effects of agricultural straw return activities. Therefore, this study proposed an innovative Pearson correlation coefficient and spectral position joint optimization algorithm (PAS) to solve this problem. The experimental results show that: In the PAS algorithm, the spectral position to correlation coefficient ratio is 3∶7, which can effectively improve the model performance.The characteristic bands near 1260 nm are a typical site of straw organic components. The characteristic bands near 1300 nm are a typical site of interaction between 1260 nm and SOM. The characteristic bands near 1320 nm are affected by 1260 nm, SOM, and 1340 nm. This study demonstrates the important influence of the spectral effect of agricultural straw return activities on SOM estimation and the approach in this study to potentially quantify the carbon sequestration effect of straw return activities.

Key words: soil organic matter, hyperspectral, soil active materials, agricultural straw return, GWR, soil carbon sequestration