土壤高光谱技术具有方便快捷、 无破坏、 成本低等优点, 已被广泛应用于估算土壤有机质含量(SOMC)。 然而, 野外测量的土壤高光谱数据因受外部环境因素(土壤湿度、 温度、 表面粗糙度等)干扰, 导致SOMC估算模型适用性有待提升。 土壤含水率(SMC)是影响野外测量高光谱的最主要的障碍因素之一, 它的变化严重影响可见-近红外(Vis-NIR)光谱反射率的观测结果。 因此, 消除SMC对高光谱数据的干扰是提高土壤高光谱估算SOMC模型预测精度的关键环节。 以江汉平原潜江市潮土样本为研究对象, 在室内人工加湿土样, 分别获取6个SMC水平的土壤高光谱数据, 采用标准正态变换(SNV)对光谱数据进行预处理, 基于外部参数正交化法(EPO)去除土壤水分对高光谱的影响, 利用偏最小二乘方法(PLSR)建立并对比EPO处理前、 后不同SMC水平SOMC反演模型。 结果表明, 土壤水分对Vis-NIR光谱反射率有显著的影响, 掩盖了SOMC的光谱吸收特征； EPO处理前不同SMC水平的光谱曲线之间的差异较为明显, 而EPO处理后的各SMC水平的光谱曲线形态基本相似； 采用EPO处理后的土壤高光谱数据建立SOMC估算模型, 预测集的R2p, RPD分别为084和250, 其精度与EPO处理前所建模型相比有较大提升, 表明EPO算法可以有效去除土壤水分的影响, 从而提升SOMC的估算精度。 对定向去除外部环境参数对土壤高光谱影响进行了实证, 为完善野外原位获取SOMC信息技术提供理论基础。

Soil hyperspectral technique was considered to be a fast, non-destructive and cost-effective alternative method for reliably analyzing soil organic matter content (SOMC). Nevertheless, hyperspectral technique challenged to use in the field, because several external environmental factors, such as soil moisture, temperature and texture, were uncontrolled, which could impact spectral reflectance seriously. Furthermore, soil moisture content (SMC) was an prime limiting aspect for hyperspectral field applications, which showed sensitive influence on the Vi-NIR optical domain. With the aim to remove the effect of SMC on the improvement of SOMC prediction, 32 fluvo-aquic soil samples at 0～20 cm depth were collected from Qianjiang in Jianghan Plain, which were rewetted in laboratory. 192 spectral reflectance from 6 levels of SMC were measured using ASD FieldSpec 3 and normalized using standard normal variate (SNV). Meanwhile, the influence of SMC on the soil spectra was analyzed and discussed. Specifically, we would like to investigate the external parameter orthogonalization (EPO) algorithm to remove the SMC effect on the spectral calibration, and the feasibility of EPO corrected spectra for estimating SOMC by comparing the partial least squares regression (PLSR) prediction results of the EPO uncorrected and corrected spectra. Results showed that the SMC had a large influence on soil spectral reflectance, which masked the subtle responses of SOMC on reflectance. The SNV transformation could not correct the differences between the dry soil spectra and the spectra obtained at various of SMC. However, the spectra at different levels of SMC were unified after EPO correction. Using the PLSR model calibrated with the EPO corrected spectra, the model accuracy was significantly improved relative to the EPO uncorrected spectra, and its values of R2, RPD for the predicted model were 084, 250, and the EPO-PLSR model could estimate SOMC comprehensively and stably, which indicated that the effects of SMC on the spectra was successfully eliminated. Thus, in the future, this approach may facilitate the proximally sensed field spectra for rapidly measuring SOMC across this study area.