GNGTS 2024 - Atti del 42° Convegno Nazionale

Session 3.2 GNGTS 2024 Results The results are shown in Figure 2 for the L2 line, where is also available the direct informaton. The three inverted models resultng from inversion process (Fig 2a-c) are compared to the clustering models, with the leachate levels logged superimposed. The models from traditonal approach (without clustering) show three main layers in agreement with the expected stratgraphy of a landfll: i) a shallow layer (thickness less than 10 m) showing high resistvity ( ρ > 20 Ωm) and low chargeability ( M < 2 mV/V), related to the presence of unsaturated waste and covering soil; ii) an intermediate layer with low resistvity ( ρ < 20 Ωm) and high chargeability ( M > 2 mV/V), which refects the presence of leachate. This layer seems quite heterogeneous, with lateral changes due to a diferent degree of saturaton inside the waste mass; iii) a deeper layer, extended down to the maximum depth of investgaton ( ∼ 50 m) with very high values of resistvity ( ρ >> 20 Ωm) and low chargeability ( M < 2 mV/V), associated to the presence of HDPE botom liner, which does not allow current to fow outside the landfll. However, the electrical models stll leave ambiguites on the identfcaton of leachate accumulaton zones, as confrmed by the mismatch between the high values of normalized chargeability and the leachate levels logged in P4. The results of Fuzzy C-Means analysis are shown in terms of clustered (Fig. 2d) and membership sectons (Fig. 2e), resultng from an optmal number of clusters equal to 8 computed by the Elbow method. The clustered secton shows a less ambiguous and more accurate identfcaton of the leachate accumulaton, validated by direct informaton from P4, which is in good agreement with the cluster no. 8 (lowest ρ and highest M and M n ), while the other wells do not show signifcant leachate levels according to the clustering reconstructon. In additon, thanks to membership secton we can evaluate the reliability of the reconstructon achieved by the clustering. Given the high values (µ > 0.75) displayed in the dark red cluster, we can validate the fnal reconstructon. Fig. 2 – ERT/TDIP data inversion and clustering analysis on the L2 line: resistvity model (a), chargeability model (b), normalized chargeability model (c), clustered secton (d) and membership secton (e). The leachate levels logged in piezometers P1-P4 are marked with a white area.