GNGTS 2013 - Atti del 32° Convegno Nazionale

We then applied the method on a real dataset acquired on a glacier (Eastern Glacier of Mt. Canin (UD), Italy) with a ProEx Malå Geoscience system equipped with 250 MHz shielded antennas. A proper processing flow must be applied to the GPR data before amplitude picking, in order to satisfy the assumptions of the method and to avoid artifacts as much as possible. A bandpass filtering and a true amplitude recovery are the basic operations required before amplitude picking. In fact, most of coherent and random noise is often either outside or at the low end of the useful frequency spectrum, and can be effectively removed by means of bandpass filters. A glacier represents a particularly favorable system for GPR studies, given the low intrinsic dissipation of the frozen materials. In such a system, the signal attenuation is mainly caused by geometrical spreading, scattering and partial reflections. After the data processing, the signal attenuation is assumed to be exclusively due to partial reflections from the interfaces between the glacier’s layers. The required approximation of a multilayered model with plane parallel interfaces in the proximity of each trace position is acceptable for most real cases, particularly in a glacier were stratification is often sub-parallel except possibly in the proximity of the rock basement. Moreover the actual slopes of the layers often do not have a major influence on the reflection coefficients and, therefore, on the results. In fact, for most real situations, the incident angles are small and, in such situations, the reflection coefficient does not vary much, according to the Fresnel equations, for most real velocity contrasts. This is a very important fact, since the actual subsurface geometry is unknown, being it the ultimate outcome of any geophysical survey, and it is an essential information in calculating the actual ray paths. Therefore, this geometrical assumption can be considered a good approximation of real conditions except in some uncommon and very peculiar situations. The results of the inversion program are shown in Fig. 1. On such GPR section six layers were identified and the peak amplitudes, including their polarity, and the traveltimes were taken from the interpreted interfaces. The peak value of the initial wavelet can be easily estimated by placing the GPR system above the ground, in order to separate the airwave from the groundwave; the recorded airwave values can be in fact considered equal to the actual transmitted wave and used as reference amplitude. The velocity values of the shallow layer were obtained from single CMP gathers acquired on selected locations; a constant mean value of 24.1 cm/ns was used for the entire dataset. Other possible independent methods to estimate the velocity include direct density measures or trans-illumination experiments. It is important to note that the proposed inversion process could give unrealistic results if the used velocity values are considerably higher of lower than the real ones, because in both cases such erroneous values could interact with the picked traveltimes and amplitudes, giving in the various layers negative or imaginary values for the thickness, or clearly unrealistic velocities. On the other hand, a slightly wrong velocity input does not severely impact the results, which would be simply normalized with respect to a reference velocity, preserving the velocity contrasts at the interfaces. Moreover, the errors associated with the calculated velocity field constitute valuable information on the effect of the uncertainty of the used reference velocity. Conclusions. We implemented and validated a new method to estimate the velocity distribution from CO GPR data by using the reflection amplitudes and traveltimes picked on the interpreted interfaces in a GPR profiles. Since the method assumes the picked amplitudes to be related only to the reflection coefficients, an accurate data processing is essential before any amplitude picking, the most important steps being the amplitude recovery and the removal of scattering effects. The method also requires as input: 1) the value of the offset, 2) the velocity of the EM wave in the shallow layer, 3) the peak amplitude of the wavelet incident at the first interface. The 116 GNGTS 2013 S essione 3.2