Random and Non-Random Uncertainties in Precision GPR Measurements: Identifying and Compensating for Instrument Drift

Abstract.

The precision of ground penetrating radar (GPR) data is of increasing interest, as the technology finds ever increasing applications to near surface geophysical studies. Our group has undertaken a series of studies to identify the precision and accuracy with which GPR traveltimes, velocities and interval properties can be estimated under controlled conditions, recently reporting that random errors in two-way traveltime and velocity are on the order of ±0.7 ns and ±0.001 m/ns, respectively, at the 95% confidence level. The high degree of precision in that dataset makes it possible to observe non-random patterns in the independent estimates of the same parameter, suggesting possible systematic biases in the data. One source of systematic error that workers often encounter is the effect of back-scattering from above ground features. Another that we documented in our previous study is that the material properties of the subsurface may change from one survey to the next. A third possibility is non-random instrumentation error, one source of which, and the focus of the present report, is a variation in the time base. We address ways to identify its presence, to assess its influence on estimating GPR parameters, and how and when to compensate for its effects. Using a static transmitter (Tx) and receiver (Rx) offset to record a series of GPR traces over a nominal period of one hour, we find that time base drift can be significant for up to 25 min after turning on the instrument. Fortunately, the type of drift that we have found to be most apparent—time zero drift—can be readily identified and compensated for, if one employs the air phase at a number of Tx–Rx offsets. An invariant condition in a GPR survey is that the true velocity of the direct air phase should be the velocity of an electromagnetic pulse in free space. Thus, upon carefully determining the observed velocity of the air phase, and using the ratio between it and the true value of 0.29979 m/ns, one can correct the entire time base of a radargram. Guidelines for when to apply the air phase compensation technique are based on whether the observed velocity of the air phase falls outside the limits of an acceptable precision under the most ideal circumstances. We illustrate the compensation procedure using two CMPs collected on different days at the same survey position. Analyzing the uncompensated data unsurprisingly yields two different subsurface velocities, however the depth estimates of the same subsurface reflector differ by 0.5 m, which for this site is physically unlikely. The difference between the observed and the true air phase velocities for both data sets exceeded the minimum expected error according to our guidelines, thus after applying our time base correction, the difference between the depth estimates improves to 0.05~m.