Accuracy and Reproducibility of Mapping Surface Soil Water Content with the Ground Wave of Ground-Penetrating Radar

Abstract

Ground-penetrating radar (GPR) is an attractive technique for the non-destructive measurement of soil water content. In this paper we investigated the accuracy of volumetric water content measurements with the ground wave of GPR. In particular, we 1) examined the influence of equipment in the vicinity of the GPR antennas on the data quality and 2) compared two methods to calculate soil permittivity from the arrival time of the air and ground wave. The first method assumes that the air wave and ground wave arrive simultaneously at zero antenna offset, whereas the second method includes a correction for a time shift in arrival time at zero antenna offset. We tested the influence of the equipment on the radar by introducing the different components (e.g., sleds, odometer and vehicle to pull equipment) stepwise. The results showed that the use of sleds caused a decrease in the amplitude of the ground wave. The presence of the sleds also changed the pulse shape of the air and ground wave and caused disturbances in the air wave. However, the data quality after the addition of all components was sufficient to allow automatic picking of arrival times by standard GPR software. The importance of the time shift at zero antenna offset was assessed with 20 duplicate WARR measurements measured along a [Formula: see text] transect and a sensitivity analysis based on GPR simulations. The results showed that: 1) there was a mean time shift at zero offset of [Formula: see text] that could not be explained from the GPR simulations and 2) there was a large variation in time shift at zero offset that could be explained by considering time picking error, positional error and soil water content heterogeneity in the sensitivity analysis. We conclude that a time shift correction in the calculation of soil permittivity would improve absolute soil water content measurements with GPR, but it should be noted that the large variation in time shifts makes it difficult to accurately estimate the mean time shift at zero offset.