GNGTS 2015 - Atti del 34° Convegno Nazionale

(B). The applied processing sequence consists of DC removal, drift removal, spectral analysis and band-pass filtering, spatial filtering to remove ringing and late arrivals of the ground wave, spherical divergence correction, and time migration. In the last two steps, a constant velocity equal to 20 cm·ns –1 was used: this value was estimated by integrating the analysis of dedicated common mid-point (CMP) gathers with direct measurements obtained in a nearby snow pit dug concurrently with the GPR survey. The results of the picking and phase-based grouping method are shown in Fig. 3B, and include those horizons automatically identified as part of the first three phases of reflections with at least three identified phases, which means that each horizon must have at least two recognized sub-parallel event in order to be displayed. The complex nature of the analyzed data set offers an interesting test for the picking procedure, since the profile contains not only laterally continuous sub-horizontal reflectors, but also discontinuous reflections and interfering diffractions caused by the debris layer, some of which are not fully corrected by the applied migration algorithm. Moreover, in the deepest part of the profile there are low amplitude discontinuous reflectors, which were interpreted as the internal ice layering and the glacieret base. Although with different levels of accuracy in different parts of the profile, all the main features have been properly detected by the applied procedure. Fig. 3 – Application of the picking procedure to a glaciological 250 MHz CO GPR profile. The figure shows: A) the processed GPR profile; and B) those horizons automatically identified as one of the first three phases of reflections with at least three recognized phases. Positive amplitudes are marked in green, negative amplitudes in red. The applied labels highlight the layered firn (L); a mixed ice and debris layer (D); and the glacieret base (B), respectively. Discussion . The proposed picking method presents two main advantages with respect to other commonly used techniques: 1) it is mostly independent from the interpreter, not needing any initial manually picked seed for horizon tracking, and with just a few required control parameters and thresholds; 2) the interpreter can check the results at each step in order to evaluate the performance, manually intervene in complex areas where the algorithm might fail, GNGTS 2015 S essione 3.3 145