Research on the soil moisture sliding estimation method using the LS-SVM based on multi-satellite fusion

Abstract

ABSTRACTGlobal Navigation Satellite System Interferometric Reflectometry (GNSS-IR) is a new remote-sensing technique; it can be used to estimate near-surface soil moisture from signal-to-noise ratio (SNR) data. Considering the advantages of multi-satellite convergence and the time and space scale of soil moisture, based on multi-satellite fusion, non-linear sliding estimation method of soil moisture is proposed. Firstly, the direct and reflection signals of Global Navigation Satellite System (GNSS) satellites are separated by means of a low-order polynomial fitting, and then, by establishing the sinusoidal fitting model of reflection signals, the relative delay phase of the SNR interferogram is obtained. Finally, a linear regression model is used to analyse and select the phase of the SNR interferogram, and a sliding estimation method of soil moisture using the least squares Support Vector Machine based on multi-satellites fusion is established. Based on the monitoring data provided by the Plate Boundary Observations network, the feasibility and effectiveness of using single and multiple GNSS satellites to slide estimate soil moisture are compared and analysed. Theoretical analysis and experimentation show that GNSS-IR can provide full play to the advantages of artificial intelligence and effectively integrate the performance of each satellite. The fitting process is stable, and the model needs less modelling data; the sliding method can achieve a long time estimation, and the estimation error is relatively stable. Based on the satellites which the correlation coefficient (r) between the phase of the SNR interferogram and the soil moisture is greater than 0.50, the fusion satellites model is established; the results are better than that of using single satellites; the r between estimated results and reference values of soil moisture is greater than or equal to 0.94, and the root mean square error and mean absolute error are less than 0.060 and 0.050, respectively. Therefore, the estimation of soil moisture can be treated as a non-linear event, and multi- satellite fusion estimation is feasible and effective.