Optimizing velocity estimation process of sparse array GPR CMP data for continuous determination of infiltration front

Abstract

As an array ground penetrating radar (GPR) system electrically switches any antenna combinations sequentially in milliseconds, multi-offset gather data, such as common mid-point (CMP) data, can be acquired almost seamlessly. However, the spatial resolution of the data is sacrificed due to the inflexibility of changing the antenna offset. The array GPR system has been used to track the infiltration front during the dune field infiltration experiment (Iwasaki et al., 2016) by fixing the antenna position to collect time-lapse GPR data, including CMP data. Although only a limited number of scans could be acquired for CMP data as the number of transmitting and receiving antennas is limited, electromagnetic (EM) wave velocities could be manually estimated for a given time-lapse CMP data, and the reflection position estimated by the GPR agreed reasonably well with the infiltration front depth independently measured with the soil moisture sensors installed right next to the antenna. However, we had to perform a velocity analysis of a large amount of time-lapse data manually, because the CMP data obtained by the array GPR are too sparse to perform a common semblance analysis. Our previous numerical study showed that the infiltration front depths estimated by the GPR agreed well with a calculation by HYDRUS (2D/3D). The main objective of this study was to optimize velocity estimation process using the sparse CMP data collected with array GPR during the infiltration test. In this study, we used simulated data for demonstration of the technique. An interpolation technique based on projection onto convex sets (POCS) algorithm (Yi et al., 2016) was evaluated to interpolate sparse CMP data. Then, the EM wave velocity was automatically estimated from the interpolated CMP data using common semblance analysis.