Detecting moisture content of soil by transmission-type ground penetrating radar

Abstract: Land reclamation requires the monitoring of soil moisture in order to ensure that the soil moisture meets crop growth needs. Traditionally, soil moisture is detected by destructively sampling soils, which is costly and associated with long work periods. Ground penetrating radar (GPR) technique is often used since it is fast, convenient, and nondestructive. GPR launches and receives high-frequency electromagnetic waves via antennae in detection of underground media. By calculating wave velocities, dielectric constants can be obtained. General GPR hardware and detection methods are based on reflector-type antennae. The transmitting antenna and the receiving antenna are both arranged on the surface of the media. Because of this, the reflected waves from different underground layers can mix together, causing errors in the calculation of wave velocity. The computed tomography (CT) transmission-type GPR technique is different from the reflector-type technique. Its transmitting antenna and receiving antenna are located on both sides of the media during the detection. The multiple reflected waves using the reflector-type GPR detection can recede effectively. Thus, the wave velocity can be calculated much more precisely, and thereby, the estimation of moisture content is more accurate. In this study, a 900 kHz transmitting antenna was improved for the CT GPR detection, and a physical model was established to simulate loam and sandy soil with different unsaturated moisture content in order to test the possibility of the CT GPR for soil moisture detection. After the take-off point was selected, the absolute take-off time and apparent take-off time were calculated. The difference of both was considered as actual wave travel time. According to the formula, wave velocity = wave velocity in air / root of dielectric constant, the dielectric constant could be calculated. Different models including the linear regression model, quadratic polynomial model, exponential function model, logarithmic function model, and cube polynomial model were established to describe the relationship between the dielectric constant and volumetric moisture content. Based on the determination coefficient and error, the best model was selected and used for soil moisture detection in a verification experiment. In the verification experiment, the soil samples were taken from Inner Mongolia of China and detected by the TDR and oven-drying method. The results showed that the cube polynomial in the same form as the Topp equation had the best goodness-of-fit. The equation used for sandy soil samples detection yielded closer detection results (R2=0.982) with the oven-drying method than with the TDR method (R2=0.867). The equation used for loam soil samples detection yielded closer detection results (R2=0.987) as well. This indicated that it is possible to use these relation models for accurate estimation of moisture contents of sandy soil and loam soil. For determination of moisture content in sandy soils, the relative error averaged 13.20% using the proposed method here, which was 14.34% lower than that from the TDR method. For determination of moisture content in loam soils, the relative error averaged 9.48% using the proposed method here, 15.79% lower than that from the TDR method. It suggested that the measurement accuracy of the proposed method is higher than that from the TDR method. Therefore, the CT GPR detection for soil moisture content can be an alternative method of TDR under certain conditions of soil moisture.