An airborne micro-TEM and physics deep learning solution to near-surface corrections in sand-covered areas

Abstract

Sand-covered areas, such as desert environments, pose challenges to seismic imaging for resource exploration and monitoring. Aeolian sand dunes provide extreme variations of elastic parameters causing, among other effects, nonlinear velocity gradients proportional to compaction. Description and correction for sand velocities is typically performed through empirically derived sand curve functions. Shallow drilling with uphole traveltime measurements represents another established technique for characterizing the thickness of the sand layer. This defines a basal depositional surface across the investigated area to apply the sand correction for seismic processing operations. The accurate definition of the unconformable base of sand is crucial for the application of robust near-surface corrections. An airborne transient electromagnetic solution is proposed, where the acquisition instrumentation, acquisition parameters, and inversion strategies are tuned to the ultra-shallow depth of the target (2–20 m). The data inversion component makes extensive use of a recently introduced physics-driven DL inversion scheme. It enables boosted depth resolution in the resistivity model using all of the available data while achieving an acceptable data misfit to honor the propagation of the electromagnetic radiation. The imaged resistivity boundary provides an excellent match to the existing uphole data and indicates the presence of long-wavelength mismatch from the previously interpreted base of sand.