Inversion of Soil Moisture by GPS-IR Combined with Wavelet Analysis and LS-SVM

The research on the inversion of soil moisture by Global Navigation Satellite Signal-Interferometer and Reflectometry (GPS-IR) has become a hot research field in recent years. In the past few years, ground-based experiments using the US Panel Observation Program (PBO) have confirmed that GPS receivers primarily used for measurements can be used to measure changes in surface physical parameters. In this paper, we study the improvement of the separation model of satellite reflected signal in GPS signal to noise ratio (SNR) observation, and the inversion model of GPS-IR remote sensing soil moisture. Firstly, wavelet analysis is used to effectively separate satellite reflected signals. Further use Least squares support vector machine (LS-SVM) rolling predictive model for estimating soil moisture. Soil moisture was estimated using the GPS SNR provided by the P038 station of the PBO observation network. Comparative analysis of the feasibility and effectiveness of single- and multiple-GPS satellites for soil moisture rolling estimation. Theoretical analysis and experiments show that wavelet analysis can effectively improve the separation accuracy of satellite reflected signals. The LS-SVM rolling estimation results are highly consistent with the soil moisture verification data. This model fully exploits the advantages of LS-SVM and effectively integrates the satellites. Performance, improved the use of a single satellite for soil moisture estimation, the results are prone to abnormal jumps; the model requires less modeling data, the use of rolling can achieve long-term estimation, the estimation error is more stable.