以石家庄多年污水灌溉区为对象，通过对污灌土壤的可见-近红外波段光谱信息进行测量和处理，利用相关分析和多元逐步回归分析，建立土壤镉含量与反射光谱变量之间的回归模型。结果表明，土壤镉在研究区土壤中有一定程度的积累，镉含量与其原始光谱、倒数对数光谱在多数光谱波段达到极显著相关水平，与连续统去除光谱相关性不显著。逐步回归模型中，倒数对数模型拟合精度最高，均方根误差为0.020，预测精度达到92.71%，说明倒数对数光谱变量模型可有效预测土壤镉含量，是估算研究区土壤镉含量较理想的预测模型。

The visible-near infrared spectrum curves of soil samples were collected from the sewage irrigation area in Shijiazhuang, and the regression models between soil Cd content and different spectral variables were developed using the correlation analysis and multivariate stepwise regression analysis. The results showed that cadmium has accumulated to some extent in the soil and exceeded the background value during the last 50 years of sewage irrigation, and the soil cadmium content was significantly related to the original spectrum and inverse-log spectrum in some bands, while had insignificant correlation with the continuum removal spectrum. Among the regression models, the inverse-log spectral model had the highest fitting accuracy with a root mean square error of 0.020 and prediction precision of 92.71%, which indicated that the inverse-log spectral variable model can predict the soil Cd content effectively in the study area.

河北省自然基金（D2010000867）；河北省高校重点学科建设项目资助