Inversion of reclaimed soil moisture based on spectrum analysis of ground penetrating radar

Abstract: Precise irrigation has been a highly urgent need to realize a quick, accurate, and non-destructive acquisition for the distribution of moisture content in reclaimed soil, particularly in coal mining subsidence areas. Consequently, traditional measurement can be replaced with Ground-Penetrating Radar (GPR), a new type of near-earth microwave remote sensing with large-scale continuous detection, fast, non-destructive, and low-cost. GPR has also been widely used in the detection of water content in soil. However, GPR is rarely used in the detection of water content in reclaimed soil, due to the complex composition of reclaimed soil and the fast attenuation of electromagnetic waves in cohesive soil. Fortunately, spectrum analysis can convert the radar data (signal change) from the time domain to the frequency domain. Energy distribution of signal frequency can be used to alleviate a large wave speed error under different media. In this study, the soil of a subsided reclamation area in Huaibei City, Anhui Province of China was taken as the research object. A control group was set up in the surrounding normal areas. The Chirp-Z-transform (CZT) was used to find the characteristic frequency. Rayleigh scattering was selected to analyze the characteristic spectrum response of radar signals under volumetric water content in soil. Various regressions were used to establish inversion models of water cut. The results showed that the frequency of peak (FP) using the Rayleigh scattering had an excellent correlation with the volumetric water content of reclaimed soil, where the correlation coefficient reached 0.90, indicating that the radar signal FP was feasible to invert the soil volumetric water content of the reclaimed area. Moreover, the CZT effectively identified the characteristic frequency in a higher resolution without changing the original frequency spectrum. The correlation coefficient between FP and soil moisture content was improved from 0.77 to 0.93, compared with the fast Fourier transform (FFT). In addition, the spectrum of radar signal in the reclaimed area demonstrated a greater drift than that in the non-reclaimed area under the same change of water content. The bulk density was also one of the factors that affect the radar signal. The highest accuracy was achieved in the exponential regression, where the verification precisions were R2=0.84, and RMSE=1.97%, showing that the frequency spectrum analysis was an effective way to invert the moisture content of cultivated soil. When large construction equipment was used to compact the reconstructed soil, there will be a larger influence on the distribution of soil moisture content and the fertility of reclaimed soil. The findings can provide an insightful theoretical basis and technical support for rapidly monitoring the soil quality and precision irrigation in reclaimed areas.