Application of LS-SVM to the Retrieval of Bare-surface Soil Moisture from Simulated Active and Passive Microwave Data

Active and passive microwave remote sensing data holds great potential for soil moisture estimation. Herein, the application of the Least Squares Support Vector Machines (LS-SVM) to soil moisture inversion is explored. Simulation is carried out using different sets of training and test data. The methods have been applied to two sets of data to retrieve bare-surface soil moisture, and the rms error (RMSE) as well as the correlation coefficient (R-2) is also obtained. The integral equation model (IEM) is chosen to obtain the backscattering coefficients as the simulated active microwave data. Using the IEM method, problems regarding forward scattering have been addressed, and the optimum incident angle has been determined through sensitivity analyses. Similarly, the emissivity model is applied to simulate a large range of soil moisture and surface roughness in order to acquire the brightness temperature and generate the datasets. The backscattering coefficients and brightness temperature are utilized as the informative microwave data in a manner that combines active and passive remote sensing. The frequencies of interest include 1.4 GHz (L-band), 6.9 GHz (C-band) as well as 10.7 GHz (X-band). Integrating these frequencies as well as multiple polarization states, the inversion accuracy has been improved. The effectiveness of this application is assessed by considering various input combinations (i.e., different microwave sensor frequencies, polarization status and incident angles). The soil moisture, which is retrieved by training LS-SVM, is then compared with that retrieved using back-propagation neural network (BPNN). This study demonstrates the great potential of LS-SVM in the inversion of soil moisture based on microwave remotely sensed data.