GNGTS 2024 - Atti del 42° Convegno Nazionale

Session 3.2 GNGTS 2024 • Which is the optmal resoluton of mult-channel GPR data for archaeological applicatons, i.e. the minimum space sampling below which there is not a great improvement in the data? • Can the data acquired by adjacent channels be stacked to improve the signal-to-noise rato, and if so, in which cases this processing step can give substantal advantages? Instrument and data processing The GPR multchannel system used in this study is the Stream C manufactured by IDS and available at Ofcina della Ricerca e della Didatca of University of Torino. It has 34 antennas, 24 of which have the dipoles oriented vertcally, and 10 horizontally, with respect to the acquisiton directon. A total of 23 traces with both the transmiter and the receiver oriented vertcally ("VV" with an horizontal separaton of about 4 cm) and 9 horizontally ("HH" with an horizontal separaton of about 10 cm) are therefore available for each acquisiton (i.e. each swath). The central frequency band is 600 MHz; however, frequency distributons centered at about 250-300 MHz were observed in most test sites. This system was adopted in diferent case studies of archaeological interest in the past few years. Two example test sites are reported in the following to test the working objectves. For both test sites the data were processed, using the Refexw sofware, according to the following processing fow: 1. subtract-mean (dewow) flter with 4 ns tmewindow; 2. automatc tme-zero correcton; 3. background removal; 4. energy decay (an automatc gain functon); 5. meanflter (a 1D moving average flter over a 10-sample window); 6. bandpass flter (100-1000 MHz); 7. subtractng average over 100 traces (= 5m); 8. f migraton with velocity estmated from hyperbola analysis; 9. energy decay; 10. Envelope; 11. normalize profles (to beter compare the depth-slices). Optonally, between step 3 and step 4, an additonal processing step consistng in the stacking of an on purpose defned number of adjacent channels was adopted. Afer those processing steps, the 2D sectons are combined to produce a 3D GPR data volume, which is then "cut horizontally" to produce depth-slices. The 3D GPR data volume was produced with a lateral resoluton of 5 cm, and a interpolaton distance of 50 cm (the interpolaton was weighted on the squared distance between the data points and the output point). First experiment: how the data quality of a GPR 3D data volume is dependent on the number of channels used. The frst dataset was acquired at the archaeological site of “Stato Ad Fines” in Malano in the municipality of Avigliana. Roman buildings and structures are atended at the test site. In this experiment, the 3D GPR data volume, and the consequent depth-slices, consider diferent subsets of the total 32 channels of the Stream C system. The results of this experiment, comparing the diferent subsets of channels, is reported in Fig. 1. Looking at the upper panels of Fig. 1, one can notce the diference between VV and HH dipoles, which can detect diferent features of the subsoil: in partcular, HH dipoles can "see" an elongated diagonal structure at the top of the picture, which VV dipoles struggle to detect. Conversely VV dipoles, can detect more clearly the rectangular features in the lower part of the area. This observaton was expected, and it is in line with previous literature (Lualdi and Lombardi, 2014).