Assessing the Precision of GPR Velocity and Vertical Two-way Travel Time Estimates

Abstract

Accurate estimates of ground penetrating radar (GPR) traveltimes and velocities are essential to a number of civil engineering, agricultural and watershed management activities. This paper reports on a study to assess confidence limits on radar parameters using a sequence of high quality common midpoint (CMP) soundings collected on eight different days over the course of a year at a single location for a range of soil moisture conditions. On each day, a “repeat” CMP was acquired leading to a total of 16 independent CMP soundings. Two prominent reflections, present in all 16 independent CMP soundings, were analyzed to provide NMO velocities and vertical two-way traveltimes along with their associated 95% confidence limits estimated using Student’s t-test, accounting for both the appropriate number of observations and estimated parameters. Interval properties were then determined using the Dix relation. The radar velocity of the [Formula: see text] interval has the largest variation observed, from [Formula: see text] for wet and dry soil moisture conditions, respectively. However, over this same time period, the estimated thickness of each interval, 0.9 [Formula: see text] and 2.9 [Formula: see text], remains constant within the respective 95% confidence limits on the 16 independent estimates and their composite average. The latter result provides a useful check on the quality of our interpretation. Analysis of the ground refraction phases also confirmed the interval velocity and thickness inferred from the reflection data. The global results of this study suggest that NMO radar velocities and vertical two-way traveltimes may be optimally resolved at the 95% confidence level with precisions of [Formula: see text] and [Formula: see text], respectively. The deviation between measurements from repeat CMPs on a given day typically fall well within the 95% confidence limit. Thus, while repeat CMP data provide a valuable check on quality control, we note that a posteriori estimates of confidence limits might provide an acceptable alternative for many applications. Regardless of the field procedure selected, this study emphasizes that an assessment of data quality is only possible by a clear statement of confidence limits and by carefully defining the basis for their estimation.