An experimental study on paddy soil moisture inversion based on emissive hyperspectra

To explore the potential of thermal infrared hyperspecra for retrieving soil moisture, the soil was measured using a 102F Fourier Transform Infrared Spectroradiometer(FTIR), and the soil´s emissivity spectra were then obtained with the Iterative Spectrally Smooth Temperature/Emissivity Separation Algorithm (ISSTES). Additionally, the spectral characteristics for different soil moisture were briefly analyzed through Correlation Analysis and Principal Component Analysis (PCA). Furthermore, the soil´s water contents were predicted and validated using a nonlinear modeling method - Partial Least Squares Regression(PLSR) and a linear multivariate regression method based upon the simulation of satellite channels calculated from measured spectra. The results show that the Reststrahlen Feature (RF) of SiO₂ is obvious in the emissivity spectra of soil with two large absorption valleys near 8.13μm and 9.17μm when the soil is under relatively dry condition. With Soil moisture increasing, emissivity curves become higher and rougher. This trend is more evident especially in the regions of 9.5-12μm. For different transformations of spectral variables, the PLSR model can achieve the highest predictive precision by using Baseline-transformed spectra, it´s RMSE of calibration and validation are 0.7681% and 1.2295% respectively and the R², 0.9929 and 0.9828. For the case of satellite bands´ simulation, the results are also satisfactory with RMSE of calibration and validation are 2.28% and 2.42% respectively, and the R² are 0.89 and 0.73, which means that the emissivity hyperspectra has promising potential for retrieving water content in soil.