Delineation of shallow resistivity structure around Malvan, Konkan region, Maharashtra by neural network inversion using vertical electrical sounding measurements

Modeling resistivity profiles, especially from hard rock areas, is of specific relevance for groundwater exploration. A method based on Bayesian neural network (BNN) theory using a Hybrid Monte Carlo (HMC) simulation scheme is applied to model and interpret direct current vertical electrical sounding measurements from 28 locations around the Malvan region, in the Sindhudurg district, southwest India. The modeling procedure revolves around optimizing the objective function using the HMC based sampling technique which is followed by updating each trajectory by integrating the Hamiltonian differential equations via a second order leapfrog discretization scheme. The inversion results suggest a high resistivity structure in the north-western part of the area, which correlates well with the presence of laterites. In the south-western part, a very high conductive zone is observed near the coast indicating an extensive influence of saltwater intrusion. Our results also show that the effect of intrusion of saline water diminishes from the south-western part to the north-eastern part of the region. Two dimensional modeling of four resistivity profiles shows that the groundwater flow is partly controlled by existing lineaments, fractures, and major joints. Groundwater occurs at a weathered/semi-weathered layer of laterite/clayey sand and the interface of overburden and crystalline basement. The presence of conduits is identified at a depth between 10 and 15 m along the Dhamapur–Kudal and Parule–Oros profiles, which seems to be potential zone for groundwater exploration. The NW–SE trending major lineaments and its criss-cross sections are indentified from the apparent and true resistivity surface map. The pseudo-section at different depths in the western part of the area, near Parule, shows extensive influence of saltwater intrusion and its impact reaching up to a depth of 50 m from the surface along the coastal area. Further, the deduced true electrical resistivity section against depth correlates well with available borehole lithology in the area. Present analyses suggest that HMC-based BNN method is robust for modeling resistivity data especially in hard rock terrains. These results are useful for interpreting fractures, major joints, and lineaments and crystalline basement rock and also for constraining the higher dimensional models.