基于电阻抗图谱估测金丝楸茎和叶可溶性糖含量的模型

The Estimation Model of Soluble Sugar Content in Stems and Leaves of Catalpa bungei C. A. Mey Based on Electrical Impedance Spectroscopy

以金丝楸(Catalpa bungei C.A.Mey)为试材,在抗寒锻炼期间分析其茎和叶的电阻抗图谱(electrical impedance spectroscopy,EIS)、可溶性糖含量及细胞膜透性。将茎和叶的EIS参数分别与相应的可溶性糖含量进行线性和非线性回归分析,选取最佳EIS参数构建估测模型,并对模型的精度进行检验,同时,用电阻抗图谱法和电解质渗出率法对金丝楸的抗寒性进行测定。结果表明:1)随着抗寒锻炼的进程,金丝楸茎和叶的可溶性糖含量均呈上升趋势。2)抗寒锻炼期间,金丝楸茎的EIS单弧逐渐变大,弧顶电抗值呈现下降趋势。叶的EIS单弧先变大后变小,弧顶电抗值呈现先减小后增大的趋势。3)金丝楸茎高频电阻率r为估测茎可溶性糖含量的最佳EIS参数,叶弛豫时间分布系数ψ为估测叶可溶性糖含量的最佳EIS参数,分别构建最佳估测模型r=-0.0177x2+0.889x+8.5912和ψ=0.0068x3-0.2023x2+1.2192x+18.4860,对模型进行检验,预测精度分别达到91.62%和81.06%。4)采用电解质渗出率法与高频电阻率r和弛豫时间分布系数ψ测定的茎和叶的抗寒性较为接近,平均绝对误差值分别为3.01和2.83。

The Catalpa bungei C. A. Mey was used to study soluble sugar content, membrane permeability and electrical impedance spectroscopy （EIS） of stems and leaves during frost hardening. The authors analysed the correlation between soluble sugar content and parameters of the EIS in stems and leaves. Using linear and non-linear regression analysis, Models of soluble sugar content of stems and leaves during frost hardening were built. The frost hardiness of Catalpa bungei C. A. Mey was assessed by EIS and electrolyte leakage （EL） methods. The results showed that： 1） The levels of soluble sugar content of stems and leaves increased with the process of frost hardening. 2） During frost hardening the reactance value in the EIS top arc decreased in stems. In contrast, the reactance value experienced an increment in EIS top arc in the initial stage, followed by a decline in leaves. 3） The EIS parameter r was the best for assessing soluble sugar content of stems and the EIS parameter ψ for assessing soluble sugar content of leaves. The best models were r = -0.0177x2 ＋ 0.889x ＋ 8.5912, and ψ= 0.0068x3 - 0.2023x2 ＋ 1.2192x ＋ 18.486 with the prediction accuracy of 91.62% and 81.06%, respectively. 4） The frost hardiness values showed a great level of similarity when measured by r in stems and by ψ in leaves compared with the EL method. The mean absolute errors were 3.01 and 2.83, respectively.