5:15 PM - 6:30 PM
[AGE27-P10] Estimation of Soil Water Retention Parameters by Ground Penetrating Radar Reflection Wave Inversion
Keywords:GPR(Ground Penetrating Radar), Water Retention, Parameter Estimation
Surface ground penetrating radar (GPR) is often used to measure groundwater levels and to determine soil moisture conditions by recording reflected and refracted pulse electromagnetic (EM) waves emitted from the transmitter. When a soil profile with a shallow groundwater table is at hydrostatic equilibrium, the distribution of soil water content within the profile is determined by water retention property of the soil. As a result, the pulse EM wave emitted from the ground surface can be reflected at a point within a so-called capillary fringe, not at a depth where the pressure head is equal to zero.
In this study, we aimed to investigate numerically if such reflected waveforms can be used to inversely estimate soil water retention parameters. We developed an inverse analysis scheme with two software packages, gprMax, an open-source software package to simulates EM wave propagation, and PEST, a model independent parameter estimation software package as shown in Fig. 1. In this scheme, soil water retention parameters were optimized by fitting to waveforms simulated with known retention parameters.
In this study, we used the well-known van-Genuchten (VG) model to compute soil water distribution in a soil profile in hydrostatic equilibrium with a known groundwater table depth. We tested our scheme with a homogeneous sandy soil layer with a width of 2 m, a height of 2 m. To reduce the computational load, among the VG model parameters, we optimized the shape parameters, n and α, which strongly control the water distribution of the capillary fringe zone.
In this study, the depth to the groundwater table was set to 1.5 m. A ricker waveform with a central frequency of 400 MHz was used as a transmitted wave. The offset between the transmitter and the receiver was kept at 20 cm.
The inverse analysis scheme was first executed independently for the two parameters n and α. We found that when the initial values are below a certain value, the objective function did not converge and the optimization hardly progressed. On the other hand, the optimization proceeded to obtain parameters with an error less than 0.2 % even when an initial value that was several times larger than the true value. When an initial value above a certain value was used, while the objective function decreased significantly and optimization proceeded, it converged to a local minimum value.
When the two parameters n and a were optimized simultaneously, it was difficult to define clear limits. In some trials, even an initial value close to the true value was given, the parameter was optimized in the direction away from the true value especially when one of the parameters was way off from the true value.
With a scheme developed in this study, the inverse analysis resulted in an accurate estimation of the VG parameters for sand with an appropriate initial value. It is necessary to investigate the effectiveness of the proposed inverse scheme for different soil types and for the experimental data.
Rererences
Saintenoy, A., andHopmans, J. (201: IEEE J. Selec. Top. Appl. Earth Obs. Remote Sens., 4(4), 748– 753.
Nemes, A, Schaap, M. G., Leij, F. J., Wösten (2001): J. of Hydro., 251 (3-4), 151-162.
In this study, we aimed to investigate numerically if such reflected waveforms can be used to inversely estimate soil water retention parameters. We developed an inverse analysis scheme with two software packages, gprMax, an open-source software package to simulates EM wave propagation, and PEST, a model independent parameter estimation software package as shown in Fig. 1. In this scheme, soil water retention parameters were optimized by fitting to waveforms simulated with known retention parameters.
In this study, we used the well-known van-Genuchten (VG) model to compute soil water distribution in a soil profile in hydrostatic equilibrium with a known groundwater table depth. We tested our scheme with a homogeneous sandy soil layer with a width of 2 m, a height of 2 m. To reduce the computational load, among the VG model parameters, we optimized the shape parameters, n and α, which strongly control the water distribution of the capillary fringe zone.
In this study, the depth to the groundwater table was set to 1.5 m. A ricker waveform with a central frequency of 400 MHz was used as a transmitted wave. The offset between the transmitter and the receiver was kept at 20 cm.
The inverse analysis scheme was first executed independently for the two parameters n and α. We found that when the initial values are below a certain value, the objective function did not converge and the optimization hardly progressed. On the other hand, the optimization proceeded to obtain parameters with an error less than 0.2 % even when an initial value that was several times larger than the true value. When an initial value above a certain value was used, while the objective function decreased significantly and optimization proceeded, it converged to a local minimum value.
When the two parameters n and a were optimized simultaneously, it was difficult to define clear limits. In some trials, even an initial value close to the true value was given, the parameter was optimized in the direction away from the true value especially when one of the parameters was way off from the true value.
With a scheme developed in this study, the inverse analysis resulted in an accurate estimation of the VG parameters for sand with an appropriate initial value. It is necessary to investigate the effectiveness of the proposed inverse scheme for different soil types and for the experimental data.
Rererences
Saintenoy, A., andHopmans, J. (201: IEEE J. Selec. Top. Appl. Earth Obs. Remote Sens., 4(4), 748– 753.
Nemes, A, Schaap, M. G., Leij, F. J., Wösten (2001): J. of Hydro., 251 (3-4), 151-162.