GNGTS 2024 - Atti del 42° Convegno Nazionale

Session 3.2 GNGTS 2024 Denser is beter? Spatal sampling vs trace stacking in multchannel GPR data to improve sectons and depth-slices readability for archaeological prospectons. A. Vergnano 1 , C. Comina 1 1 Università degli Studi di Torino, department of Earth Sciences; Torino, Italy Introducton Among the geophysical indirect methods to preliminary assess the presence of underground structures for archaeological studies (and for many other applicatons), multchannel GPR systems have been under the spotlight for more than a decade (Corradini et al., 2022; Trinks et al., 2009; Viberg et al., 2020). Their peculiarity is to be composed of several radar antennas, both transmiter and receiver, placed one aside the other, in order to produce a spatally dense amount of data, which easily allows for full 3D surveys, with a logistcal efort “similar” to that of a standard GPR single-channel system (Novo et al., 2012). This means that very high resolutons, usually of the order of few centmeters, can be easily achieved, making it possible to detect smaller targets (Linck et al., 2022), and making more reliable the interpolaton process to produce horizontal depth- slices. Nevertheless, the data processing of multchannel GPR data can be signifcantly diferent (Trinks et al., 2018) and more tme consuming, from traditonal GPR systems, given the large amount of data created. Also, in some situatons, the increased density of spatal sampling does not correspond to an increased quality of depth-slices given that small anomalies and wave scatering can disturb the data and reduce the visibility of more relevant targets of archaeological interest (i.e. walls or structures). Moreover, notwithstanding the benefts of new multchannel GPR systems, they are stll subject to limitatons intrinsic to the GPR wave propagaton, for example the difculty of penetratng for more than a few meters into the soil, and the data quality degradaton in the presence of even small quanttes of clay materials. This makes the data processing of multchannel GPR dataset an interestng research topic and various studies are available in recent literature, which include Principal Component Analysis (Linford, 2023), studies on the signal polarizaton and orientaton (Lualdi and Lombardi, 2014), studies on various flters for artfact reducton (Verdonck et al., 2013), studies on positoning accuracy (Gabryś and Ortyl, 2020). In this context, this work aims to assess the variaton of depth-slices quality with the variaton of between-channels resoluton, and proposes an alternatve processing step for noise reducton of GPR multchannel datasets. It focuses on two objectves connected one to each other: