基于双频加权组合的土壤含水量估算方法研究

Research on Soil Water Content Estimation Method Based on Dual Frequency Weighted Combination

全球定位系统干涉反射测量可用于估算近地表土壤水分含量。用遗传BP神经网络和多元线性回归分析将L1和L2两个频点的数据进行组合,以改进反演土壤湿度的精度。经理论分析和算例表明,遗传BP神经网络双频融合反演土壤湿度,改善了不同频点信号间存在的误差,融合后数据的变化趋势也更加接近于土壤湿度变化趋势;多元线性回归反演土壤湿度,可以准确计算L1、L2频点观测相位与实测土壤湿度之间的相关程度,提高预测方程式的效果;双频融合反演结果与土壤湿度参考值之间的相关系数R 2分别为0.865、0.824、0.766、0.724,比L2频点至少提高17.5%,比L1频点提高了0.1%。

The Global Positioning System Interference Reflectance Measurement(GPS-IR)can be used to estimate the soil moisture content near the surface.The genetic BP neural network and multiple linear regression analysis are used to combine the data of the two frequency points of L1 and L2 to improve the accuracy of soil moisture inversion.The theoretical analysis and calculation examples show that the genetic BP neural network dual-frequency fusion inversion of soil moisture can improve the error between signals at different frequency points,and the change trend of the data after fusion is closer to the change trend of soil moisture.The multivariate linear regression inversion of soil moisture can accurately calculate the correlation between the observed phases of L1 and L2 frequency points and the measured soil moisture,and improve the effect of the prediction equation.The correlation coefficients R 2 between the dual-frequency fusion inversion result and the soil moisture reference value are 0.865,0.824,0.766 and 0.724 respectively,which are at least 17.5%higher than the L2 frequency point and 0.1%higher than the L1 frequency point.