Construction of dense CMP data from sparsely collected GPR CMP data for the improved estimation of soil dielectric constant profile

• Published in: Vadose zone journal Vol. 24; no. 1
• Main Authors: Oikawa, Koki, Saito, Hirotaka, Kuroda, Seiichiro, Takahashi, Kazunori
• Format: Journal Article
• Language: English
• Published: Madison John Wiley & Sons, Inc 01.01.2025; Wiley
• Subjects: Algorithms; Antennas; Construction; Convexity; Dielectric constant; dielectric properties; Fourier transforms; ground-penetrating radar; Methods; Oligodendrocyte-myelin glycoprotein; Radar; Sand; soil; Spatial discrimination; Spatial resolution; Transmitters; vadose zone; Velocity
• ISSN: 1539-1663; 1539-1663
• DOI: 10.1002/vzj2.20392

A multichannel ground‐penetrating radar unit, which consists of multiple fixed transmitters and receivers, allows one to obtain time‐lapse multi‐offset gathers (MOG) by fixing the unit in one place at the expense of spatial resolution. When applying common semblance analysis for dielectric constant estimation, it is important that MOG, such as common midpoint (CMP) data, are collected with a small antenna step size to obtain clear contrast in semblance values. Thus, a reliable computational technique for constructing spatially dense CMP data from sparse CMP data is required. An F‐K filter method based on the projection onto convex sets (POCS) algorithm was developed to interpolate missing traces. In our previous study, a novel method based on the F‐K filter was proposed by optimizing the filter zone to construct dense CMP data from sparse CMP data. The objective of this study was to formally compare the novel F‐K filter method with a method based on the POCS algorithm. The former method consists of a normal moveout correction and a fan‐shaped F‐K filter, the parameters of which are optimized by leave‐one‐out cross‐validation. The novel method was able to construct a dense reflected hyperbolic‐shaped wave signal well from the sparse CMP data collected from a sand box filled with uniform dry sand, and to improve the velocity estimation with semblance analysis. The novel method is considered more suitable for the construction of dense data from sparsely collected CMP data than the method based on the POCS algorithm. Core Ideas An optimized fan‐shaped F‐K filter was compared with the method using projection onto convex sets (POCS) for construction of dense ground‐penetrating radar (GPR) data. The fan‐shaped F‐K filter was optimized using leave‐one‐out cross validation with simulated GPR data. The optimized fan‐shaped F‐K filter method outperformed the POCS method when used for sparser CMP data. Plain Language Summary We need methods to measure underground structures and the groundwater surface without damaging the ground. This is required in a wide range of fields such as civil engineering and agriculture. We used ground‐penetrating radar (GPR) to scan underground. GPR consists of antennas to emit electromagnetic (EM) waves and to record signals of the returned ones. EM waves propagate underground and are reflected at the location where the propagation speed changes drastically. To determine the depth of such locations, time‐consuming detailed measurement is required. However, such measurements are not always possible. Therefore, this study proposed a method that allows us to make detailed data from non‐detailed data, which are obtained from quick measurements. The proposed method was verified with actual data collected from a soil tank filled with dry river sand. The proposed method allows us to avoid performing the time‐consuming measurement so that flexible fast measurements can be used.