An enhancing precision method for downward continuation of gravity anomalies

Abstract

Downward continuation of the gravity field observed on the surface is a method applied to calculate the anomalous field below the surface close to deep sources. However, the downward continuation is mathematically incomplete and uncertain, and its accuracy depends on the sampling rate and coverage of the gravity data. To improve the accuracy of the downward continuation of gravity anomalies, we constructed a quasi-neural network that is made up of multiple layers of neurons. Each of the neurons analyzes the attributes of the input data to determine the best control parameters for the downstream output of the current layer, and to judge whether the results of the downstream extension are close to the main anomalous sources. We define the singularity statistic parameter and singular probability of the signal to characterize the singularity attribute of each-layer's continuation results. If the continuation data contain large singularity, they are corrected by the minimum curvature method to weaken the influence of the interference of shallow random sources. Numerical simulation tests showed that the proposed method can effectively weaken the interference of shallow random sources, and obtain more accurate information on deep-sources. A real-world example demonstrated that the method successfully delineate deep geological structures.