煤矿区土壤有机碳含量的高光谱预测模型

可见—近红外光谱已被证明是一种快速、及时、有效的土壤有机碳含量预测工具。利用Field Spec4对济宁鲍店矿区的104个土壤样品进行光谱测量,采用Savitzky-Golay卷积平滑(SG)、多元散射校正(MSC)及数学变换等多种方式组合对光谱预处理,并运用偏最小二乘回归分析建立土壤有机碳含量预测模型,进而探讨煤矿区土壤有机碳含量的高精度预测方法。结果表明:(1)不同的光谱预处理方法对建模结果影响差异较大,建模结果以SG加MSC预处理再结合光谱反射率的一阶微分变换最优,建模R~2=0.86,RMSE=2.0g/kg,验证R~2=0.78,RMSE=1.81g/kg,RPD=2.69。(2)倒数和倒数的对数与土壤有机碳含量的相关性曲线接近重合,与反射率曲线成反比,但是建模效果远低于反射率;光谱反射率的一阶微分能明显提高500~600nm波段相关性。(3)光谱反射率随土壤有机碳的含量减少而增大,当有机碳含量较低时,其波谱的近红外波段反射率响应能力也随之降低,反射率直接建模难度加大。

Hyperspectral Prediction Model of Soil Organic Carbon Content in Coal Mining Area

Visible-near infrared (Vis-NIR) spectroscopy has been proved to be a rapid, timely and efficient tool for predicting content of soil organic carbon (SOC). In this study, FieldSpec4 was used to measure 104 soil samples collected from the Baodian mining area of Shandong province. Vis-NIR reflectance spectra and SOC content were measured under laboratory conditions. The spectral data were first denoised using the Savitzky-Golay (SG) convolution smoothing method, the multiple scattering correction (MSC) method, after which the spectral reflectance was subjected to reciprocal, reciprocal logarithm and differential transformations to improve spectral sensitivity. Finally, regression models for estimating the SOC content by the spectral data were constructed using partial least squares regression (PLSR). The results showed that: (1) Different spectral preprocessing methods had great influence on the modeling results, and the modeling results performed best when the spectral reflectance was preprocessed by Savitzky-Golay (SG) smoothing coupled with multiple scattering correction (MSC) and first-order differential transformation (modeling R²=0.86, RMSE=2.00 g/kg, verification R²=0.78, RMSE=1.81 g/kg, RPD=2.69). (2) The correlation curve between reciprocal and SOC content was similar to the correlation curve between the logarithm of the reciprocal and SOC content. They were inversely proportional to the reflectivity curve, and the modeling effect was far lower than the reflectivity; the first-order differential of spectral reflectance could significantly improve the correlation of the 500~600 nm band. (3) The spectral reflectance increased with the decreasing of SOC content. In addition, when the SOC content was low, the sensitivity of the spectrum especially that in the near-infrared band of the original reflectance to the change of SOC content decreased, and the direct modeling difficulty of the reflectance increased.