Abstract: In order to study the hyperspectral diagnostic method of desert soil organic matter, the hyperspectral reflectance characteristics of both no-grinded and 2 mm particle size samples as well as their correlation with organic matter content were analyzed. Results showed that the original spectral reflectance of no-grinded sample had very significantly negative correlation with soil organic matter in 365.0~449.2 nm and 449.2~567.8 nm. The reflectance of mathematical transformation expanded the scope of related band. The maximum correlation coefficient between the spectral reflectance of 2 mm sample and soil organic matter (|r|=0.420) was higher than that between the spectral reflectance of no-grinded sample and soil organic matter (|r|=0.351). By first-order differential transformation of 2 mm sample spectral reflectance, the hyperspectral prediction model of desert soil organic matter was developed in 569.2 nm and 1141.5 nm, the prediction precision of soil organic matter content was higher, but the prediction accuracy was poor when the organic matter content was less than 9 g·kg^-1.

Key words: Hyperspectral reflectance characteristics, Soil particle size, Soil organic matter, Prediction

摘要： 为研究荒漠土壤有机质的高光谱诊断技术，分析了不研磨土样与2 mm土样的高光谱反射特征及与有机质含量的相关性。结果表明：不研磨土样的原始反射光谱在365.0~449.2 nm和567.8~799.8 nm与土壤有机质成极显著负相关。反射光谱经数学变换后，扩展了与有机质极显著相关的波段范围，2 mm土样光谱反射率与有机质的最大相关系数（|r|=0.420）高于不研磨土样（|r|=0.351）。2 mm土样光谱经一阶微分变换后，在569.2 nm和1141.5 nm建立的土壤有机质含量预测模型精度较高，但是当有机质含量低于9 g·kg^-1时，预测精度较差。

关键词: 高光谱反射特征, 土壤粒径, 土壤有机质, 预测