Comparison and application of downward continuation methods for potential field data

Abstract

Downward continuation highlights the details of the gravity and magnetic anomalies and improves the horizontal resolution of potential field data. However, downward continuation acts as a high-pass filter. As depth increases, any high-frequency noise will be quickly amplified and mask the effective information in the continuation field. We review 13 popular downward continuation methods and categorize them into five classes. The methods are compared and discussed in terms of continued depth and noise robustness using synthetic data and then are applied to field data from the Pobei area in Xinjiang Province, China. It is found that the noise robustness of the integral equation approach is improved using the iterative Landweber form. The frequency-domain methods based on high-frequency truncation do not require iterations and have a fast calculation speed. Using the Laplacian of the Gaussian operator to calculate the second-order vertical derivative enhances the noise robustness of the Taylor series expansion-based techniques. Segmentation of the continued depth followed by downward continuation increases the accuracy of extrapolation methods. High-order derivative terms located in the denominator of the expansion series cause instability in the continuous fractional series expansion-based techniques, and the obtained results have to be further proceeded with median filtering.