Inversion of surface gravity data using iteratively reweighted bounded variables least squares method

We investigate an 1 L -norm based approach which produces subsurface density images with sharp boundaries for inversion of surface gravity data. The density distribution of subsurface is modeled with a uniform grid of rectangular cells. The density contrast of each cell relative to the background density, is inverted for by minimizing the 1 L -norm of the data misfit with upper and lower bounds on the density contrasts using the iteratively reweighted least squares (IRLS) with bounded variables algorithm. The 2.5 dimensional synthetic tests show that this approach retrieves structures with sharp boundaries acceptably, without the need of explicit regularization. The quality of inversion, however, depends upon a proper estimation of the minimum depth to the anomalous density body. A comparison of our results with 2 L -norm based method on a noisy synthetic data set favors the 1 L -norm based approach.