Crustal and lithospheric mantle structure below the Indian shield based on 3-D constrained gravity inversion and Deep Neural Network approach: Geological implications

Abstract

The Indian shield region comprises of several Archean cratons, Proterozoic mobile belts, Proterozoic-Phanerozoic rift basins, and volcanic provinces and has been significantly modified by various tectono-thermal events and the supercontinental cycles. These overprinting processes and evolution can be understood, to some extent by characterizing the lithosphere structure. In this study, we carried out seismically constrained 3-D nonlinear gravity inversion using tesseroids in the spherical coordinate system to delineate the Moho geometry for the Indian shield. Further, the supervised machine learning approach using Deep Neural Network (DNN) is implemented on the geopotential data to obtain Lithosphere-Asthenosphere Boundary (LAB) across the Indian shield. The inversion results show that the Moho depth varies between 33 and 60 km below the Indian shield with a mean difference of ∼ −0.3 km and a standard deviation of ∼4.9 km relative to the Moho from receiver function data. For the DNN model, the estimated LAB depth values range between 95 and 280 km below the Indian shield. The model gave rise to the coefficient of determination R2 of 0.93, Mean Absolute Error of ∼14 km, and Root Mean Square Error of ∼22 km between the observed and predicted LAB depths. Apart from the general agreement of Moho and LAB depth estimates with previous seismological studies, the present study provides better resolved information of lithosphere structure across the Indian shield. The modelled lithosphere structure therefore will be useful to integrate with the geochronological data in order to to understand the Proterozoic tectonic evolution of the Indian Shield.