Optimization of multi-channel ground-penetrating radar for quantifying field-scale soil water dynamics

Abstract

Continuous multi-offset surveys conducted with multi-channel ground-penetrating radar (GPR) systems have already shown their potential for a fast and high-resolution mapping of near-surface soil architecture and volumetric soil water content. Here, we study the accuracy of an 8-channel setup as a function of antenna separation, reflector depth and dielectric permittivity. This is done by Monte Carlo simulations that add noise to the components of the measuring process. We find that adapting the antenna setup to the particular situation is mandatory for an optimal accuracy. In the second step, we demonstrate the applicability and accuracy of our approach with a time-series of real data from a site with two pronounced reflectors. We find that the measured radargrams are highly reproducible and allow to determine reflector depths with an accuracy of about 0.1 m and soil relative dielectric permittivity with an accuracy of about 0.5. With this we quantify the effective field-scale dynamics of soil water for the layers between the ground-surface and the reflectors.

Introduction

Ground-penetrating radar is a well-established geophysical measurement technique for high-resolution imaging of the shallow subsurface (e.g., Knight, 2001, Neal, 2004, van Dam, 2012). It is being employed in a wide range of applications like characterizing the subsurface architecture in sedimentological studies (e.g., Neal, 2004) or building the basic framework of hydrogeologic models (e.g., Knight, 2001). Further, more detailed applications involve the estimation of hydraulic and state variables such as porosity (Pascale et al., 2004) and volumetric soil water content (e.g., Huisman et al., 2003), the detection of the depth of the groundwater table (e.g., Roth et al., 2004, van Overmeeren et al., 1997), or the movement of contaminants (e.g., Brewster & Annan, 1994).

In order to obtain accurate images of the subsurface structure, the signal travel times measured by GPR need to be transformed into depth information. Besides the depth of the reflection, travel times also depend on the dielectric permittivity of the soil. This depends on soil water content which may vary in space due to spatial variability of soil texture. Consequently, assuming only one dielectric permittivity for transforming travel times into reflector depths for a complete site may lead to a distorted model of the subsurface architecture. Calibration of reflector depth and dielectric permittivity can be conducted by using information from boreholes (e.g., Lunt et al., 2005) or from common midpoint measurements (CMP) at selected positions along the site (e.g., Greaves et al., 1996, Turesson, 2006). These surveys are labour intensive, however.

The Normal Move Out (NMO) method widely used in seismology has been applied in GPR applications (e.g., Campbell et al., 1995, Greaves et al., 1996, Bradford et al., 1996). Since the multifold acquisition in a GPR survey with different transmitter/receiver offsets enables dense CMP measurements, it was used by Greaves et al. (1996) for investigating lateral velocity variability. To evade some of the limits imposed by the assumptions of small spread length, small velocity gradients and planar reflecting horizons for the NMO-based processing method in GPR investigations, Bradford, 2006, Bradford, 2008 applied reflection tomography coupled with prestack depth migration, and significantly improved the accuracy of measuring vertical and lateral heterogeneities. Gerhards et al. (2008) proposed an alternative approach and used a multi-channel GPR setup with just a few channels to estimate dielectric permittivities and reflector depths. The advantage of this method is that it allows the simultaneous determination of reflector depth and average water content with a high spatial resolution and at a moderate effort. The price to pay is that the reflected signal must first be picked and that the antenna separation has to be adapted to the reflector depth in order to achieve optimal results. In addition, measurement errors occurring under field conditions also need to be taken into account. The challenge is then, to reach an optimal compromise between measurement accuracy, signal detectability, spatial resolution and reflector depth.

Based on the algorithm of Gerhards at al. (2008), we conduct and evaluate synthetic Monte Carlo simulations of multi-channel GPR travel times of the CMP gathers for various antenna separations and soil water states and we demonstrate the application with actual measurements at a study site. The objectives of this study are to (1) determine appropriate setups for a multi-channel GPR array leading to minimal uncertainties in reflector depth and permittivity estimates; and (2) investigate the reproducibility of the depth estimate and the ability of the method to detect spatial and temporal changes in soil water content by analysing a time series of field measurements performed during different hydraulic states.