Multiorder Sequential Joint Inversion of Gravity Data With Inhomogeneous Depth Weighting: From Near Surface to Basin Modeling Applications

We have established a workflow for a multiorder sequential joint inversion (MOSJI) of gravity and gravity gradients, that aims at modeling vertically stacked sources in various geological scenarios. We consider the joint inversion of the gravity data and one of the $h$ th-order derivatives of the gravity data. The first step involves separate inversions, which are fundamental to fully exploit the different wavelength-content of the two quantities to invert. The joint inversion is warranted by using the scheme of a sequential joint inversion with a cross-gradient constraint. The algorithm is able to exploit different types of a priori information, such as compactness and inhomogeneous model-weighting function. First, we test this approach on a realistic synthetic model from the SEg Advanced Modeling (SEAM) Phase I model, involving salt and mother salt structures. Then, we consider a synthetic model containing either shallower or deeper karst cavities. These tests produced a better modeling of both shallower and deeper sources, when compared to the separate unconstrained inversions. Thanks to these good results, we apply our method to a real case for cavity detection in Southern Spain. The method shows an accurate modeling of the expected sources. In all the aforementioned tests, we obtain a strong decrease of the cross-gradient values and a meaningful linearization in the scatter plots of physical parameters, both indicating the good performance of the joint inversion.