GNGTS 2018 - 37° Convegno Nazionale

692 GNGTS 2018 S essione 3.2 that are not common in large river dikes and that are quite dangerous for the stability of the structure. To model the water level variations in the canal in the real site, water heights of 5.5 cm and 11.5 cm from the channel base were studied at the laboratory corresponding to half of the scaled maximum height in the field and the maximum scaled, respectively. The water level was then lowered with the same procedure. Resistivity variations and accordingly the water saturation changes within the levee were continuously monitored during the test using high speed acquisitions taking about 3min. High frequency of measurement is crucial in time lapse monitoring, especially in small scale tests and with the material used in these tests, that has high permeability. In the second test, a rainfall event with different intensities was simulated at the end of the irrigation period. This test was performed also because in the field it is quite difficult to separate rainfall influence from the effect produced by changes of water level in the canal. The reason is that the managing authority lowers the water level in the canal during the rain to avoid overtopping. Rainfall event was simulated when the canal was empty, but the water saturation of the soil was estimated about 35% according to TDR measurements. Two different rain intensities of 57 mm/h and 65 mm/h were used for this test. In the third experiment, a rainfall event during winter period, when irrigation is not performed, was modelled after a short drying period. Time-lapse ERT data were measured with empty canal and volumetric water content of the soil was 24%. Two rainfall intensities of 79 mm/h and 90 mm/h were used to simulate precipitation intensities. Data processing. After performing the experiments and having a look at measured resistivity pseudosections of different times, we noticed that all datasets showed an anomalous increase of resistivity values with the depth. Fig. 2 shows two examples of such meaningless trends in measured pseudosections for the experiment simulating filling and drawdown periods. Fig. 2a is measured at T29 when the water level in the canal was 5.5cm and Fig. 2b was measured at T1.37 when the water was discharged from the canal. This amplification of resistivity values was so strong that inversion of datasets was useless ending in resistivity sections without any geological meaning. The increasing effect was first assumed to arise from the high resistivity base of the flume. 2D analytical calculations for a two-layered earth model, and forward modellings performed in Res2dmod software proved other important factors being responsible for rapid resistivity increases with the depth (Arosio et al. , 2018). The problem is born from the limitation of 2D surveys and models in assuming that the resistivity does not change in the direction perpendicular to ERT profiles. Apparent resistivity values measured along a river embankment are not only influenced by the materials directly below the ERT line, but also by severe lateral resistivity changes at the sides of the profile (river channel with varying water levels on one side and the air on the other side). Moreover, high-resistivity base and sides of the flume also Fig. 2 - Measured pseudosections at a) T29; b) T1.37.