The density covariances estimated by the conventional method is stationary and diffuse, which could result in an unfocused image by cokriging inversion. We present a compact inversion method based on cokriging to recover a 3D density model by inverting gravity data or gravity gradient data. Firstly, based on the broadly used spatial homogeneity of the density in cokriging, when the density distribution of a focused model is available in all regions we will get the theoretical covariances which should be nonstationary. To break the stationary character of the conventional covariances, we develop an algorithm, starting from a very small threshold initially. We increase the threshold gradually and execute cokriging iteratively until the maximum density in the recovered model reaches the upper bound for the density of the region. We applied the new method on two synthetic models, one is a two-prism model, and the other one is a dipping dike. The results demonstrate that the new method can obtain the well-focused image with much sharper boundaries and smaller anomalous bodies than the conventional cokriging.
展开▼