The magnetic method is one of the two most effective geophysical techniques currently in use for unexploded ordnance (UXO) detection and discrimination. These data are acquired with sensors at some height above ground surface, and it is well known that height significantly affects magnetic anomaly shape, amplitude, and spatial extent. Consequently, anomalies due to multiple metallic targets may overlap at a given height above the ground surface, and the acquisition noise may significantly decrease the signal-to-noise ratio (SNR) of data. These adverse effects ultimately mask the true level of contamination at a site during initial wide area assessment (WAA), as well as decrease overall effectiveness of discrimination during the active clearance stage. The overall objective of this project was to develop a robust algorithm for stable downward continuation (SDC) of magnetic data acquired at some height above the ground to reconstruct the magnetic data with a higher resolution at the ground surface.
Recent developments have enabled the recovery of dipole or ellipsoidal parameters through inversion of magnetic data. The reliability of these algorithms depends on well-defined anomalies associated with individual targets and a quantitative characterization of data noise.
The signal content and relative noise characteristics in a magnetic data set are strongly affected by the observation height. The magnetic anomaly due to a dipolar source decays with distance cubed and the anomaly spreads out spatially in proportion to the distance as well. For example, the width of a magnetic dipole anomaly at one half of its maximum is approximately equal to the source depth. This means that the data resolution defined as the distance between separable anomalies decreases linearly with increasing observation height and the relative noise in the data increases with the observation height. The ideal data are those acquired with a sensor at zero height above the ground. However, because of the intrinsic limitations imposed on field data acquisition, all platforms collect data at some height above the ground surface. The magnetic sensor height in a typical man-portable system or a vehicle-towed system may be between 0.15 to 0.5 m above the ground, and a helicopter-based survey would be flown at a height about 1.5 to 2.0 m above the ground. Thus, there is a need for enhancing magnetic data to alleviate the effects of observation height and to estimate the noise characteristics.
To address this specific aspect of magnetic data, this project developed and applied an SDC method. The algorithm formulates downward continuation as an inverse problem using Tikhonov regularization and has the flexibility of incorporating the expected power spectrum of UXO anomalies. The degree of regularization for this formulation was estimated automatically using the well-established methods in linear inverse problems.
During the initial SERDP Exploratory Development (SEED) effort, researchers developed a working SDC algorithm and tested it on synthetic and field data. Applications during the first year showed that the algorithm reliably estimates the noise in UXO data and reconstructs the magnetic anomaly at ground surface within the limitation imposed by the noise. The reconstructed field at the ground surface exhibits significant enhancement compared to the original data.
In follow-on work, researchers incorporated an initial set of inversion algorithms to estimate requisite ensemble depths within the data's radial power spectrum for improved SDC. In particular, development of recursive Quenched Simulated Annealing (QSA) from SERDP project MR-1638 was adapted for this purpose. The QSA algorithm has been demonstrated to be a reliable tool for modeling the redial power spectrum of UXO magnetic data using non-linear parametric inversion. As a result, it is now possible to automatically define a model objective function for SDC to quantify the conformity of the enhanced data with expected spectral properties. In addition, the project team was able to validate the SDC enhancement package by evaluating its performance with the latest detection technology based on the extended Euler deconvolution.
Successful UXO discrimination is essential for minimizing the costs of ordnance clearance. Discrimination algorithms depend on accurate and reliable sensors for different types of data and the availability of high-quality data that these sensors provide. Much of the attention in UXO geophysics research has been devoted to the development of new sensors and new algorithms for data inversion and discrimination. However, a third and equally important aspect is the pre-processing of data and the extraction of data with high SNR. Such algorithms provide a complementary component to newly developed sensors and strengthen the foundation for advanced discrimination criteria.