GNGTS 2017 - 36° Convegno Nazionale

GNGTS 2017 S essione 3.2 623 (Jol, 2009). Then, we compared such values with those obtained in laboratory from dry oven samples, by applying θ v = θ g * (γ d / γ w ) (Di Matteo et al. , in press), and with ones from literature (Lambot et al. , 2004). The results show a good fit between θ v measured by Topp’s method and θ v from laboratory tests, in the range 0.05<θ v <0.17 (Fig. 2). Outside of this range (very dry and wet soils), θ v estimation progressively deviates from the Topp’s equation. Our results confirm that the Topp-model doesn’t performwell for very low θ v (Porretta and Bianchi, 2016). Moreover, at higher frequencies and WC close to saturation (θ v ~0.4), the Topp-model over-predicts the bulk ε r by up to 20% (van Dam et al. , 2005). Therefore, we suggest that a laboratory calibration is mandatory for reliable WC estimations. GPR survey on field test site. The dataset analyzed for soil S B was then used to calibrate the radargrams collected on a test site located along the Tiber river at Ponte Felcino (PG, Italy). The soil S B , here outcropping, was 2D/3D GPR surveyed to reconstruct a reliable geophysical model of the subsurface, by comparing and validating the field results with those from laboratory. In addition, in order to check the actual θ v , some soil samples were collected at some points and at different depths (max depth 1 m) over the GPR grid. Such combined approach was useful to extend our study at the site scale, constraining the pseudo-3D GPR field model, providing a more accurate geological and hydrogeological interpretation of the area. We collected total number of 28 2D CO profiles, 16 using the same 1 GHz antenna and in addition, to explore different resolutions and penetration depths, other 6 with a 500 MHz and 6 with a 1.5 GHz antenna; an odometer wheel recorded the GPR traces every 1 cm. The profiles length was 6 m each, and the crossline spacing was 20 cm for the 1 GHz antenna, whilst 60 cm for the others (Fig. 3a). By merging the 1 GHz radargrams, we generated a pseudo-3D GPR volume (Ercoli et al. , 2014). A CMP profile was also recorded using the 1.5 GHz (Tx) and 1 GHz (Rx) antennas (40 cm and 4 m the min and max offset, with a trace increment of 4 cm). The CMP was recorded to define an accurate velocity model, in addition to the hyperbola fitting analysis, mandatory in the field due to inability to use the reference metal plate adopted in laboratory. We interpreted the depth profiles after a standard processing flow and a time to depth conversion done using a mean velocity of 0.143 m/ns (ε r = 4.40, from 0-12 ns), and 0.121 m/ns more in depth (ε r = 6.14, at about 18 ns). Such velocity range was considered representative for the geological units located at the study site. GPRfieldsurveyresultsanddiscussion. GPRlaboratoryexperimentsandfieldinvestigations allowed to constrain the estimation of θ v values. On soil S B , actual θ v values were obtained by sampling the soil down to a depth of 1 m (Fig. 3a). The data show how θ v values slightly increase in depth, reaching a maximum value of about 0.16 at 0.60-0.90 m depth (Fig. 3b). The analysis of the pseudo-3D GPR data (Ercoli et al. , 2012) provide a geophysical model of the survey site, suitable to reconstruct the subsurface geology and infer hydrogeological features of unsaturated zone (Fig. 3c). The 3D GPR volume displays a general homogeneity showing sub-horizontal layers. A shallow discontinuity (below the green dashed line), gently dipping to a depth of 0.5 Fig. 2 - θ v values computed by laboratory tests vs results provided by 1 GHz GPR for both SA and SB. Figure shows also the Topp’s equation and results on sandy soil as proposed by Lambot et al. (2004).