基于EM38的土壤剖面电导率预测研究

【目的】以海涂围垦区盐碱土为研究对象，利用EM38大地电导率仪在地表不同高度测量的土壤表征电导率预测土壤不同深度土层剖面的电导率。【方法】利用EM38电导率线性响应模型结合Tikhonov正则化能较好解决病态矩阵的线性反演问题。利用该方法来预测土壤剖面电导率，并对预测结果进行误差分析，最后通过偏差变化法增加噪声利用局部灵敏度分析法来评价模型的灵敏度。【结果】研究发现，土壤剖面平均电导率和地表不同高度平均表征电导率具有极显著相关性，能利用表征电导率较好的预测剖面平均盐分。线性模型不仅能较好的预测土壤剖面电导率的变化趋势，而且在数值上也能较好的预测电导率的大小，平均预测误差在40%左右。相对预测误差较大的土层，模型灵敏度越大，因此可通过提高EM38数据测量的稳定性来提高预测精度。【结论】利用EM38表征电导率采用线性模型结合Tikhonov正则化的方法能够较好的反演土壤剖面电导率，预测结果可以为样区土壤管理提供科学的决策依据。

Inversion of Soil Conductivity Profiles Based on EM38 Apparent Electrical Conductivity

【Objective】In this paper, apparent electrical conductivity(ECa) measured by EM38 at different heights above the soil was used to determine lateral variations of apparent electrical conductivity at different depths in a coastal saline field.【Method】Tikhonov regularization is an effective technique for the solution of ill-posed linear inverse problem. Electrical conductivity profiles were inversed through solving a nonnegative lease squares problem by combining the McNeil's linear model theory of EM38 with Tikhonov regularization. The prediction error of the depth profile of electrical conductivity was analyzed. Finally, sensitivity of the model was assessed by adding noise to the observed data from a normal distribution.【Result】There was a significant correlation between mean of electrical conductivity for the whole soil profile and mean of ECa measured at different heights above the soil. Moreover, the inversed data of soil electrical conductivity using linear model can characterize well the changing trend of electrical conductivity for the whole soil profile. It is noteworthy the average prediction error of this model is about 40%, and the model was more sensitive to the layers which having larger prediction error. It suggested that accuracy of prediction can be improved by strengthening the measurement stability of EM38 equipment in field.【Conclusion】The results demonstrated that an integrated approach of the linear model and Tikhonov regularization can be used to inverse EM38 data to the electrical conductivity profile quantitatively, which presented a great potential way for soil investigation and management.