During scalar magnetic surveys, where the amplitude of the magnetic field is measured, small changes in towed sensor positions can produce complex noise-resembling signals in the data. For well-constructed measurement systems, these signals often contain valuable information, rather than noise, but it can difficult to realize their potential. We present a simple, general approach, which can be used to directly invert data from scalar magnetic surveys, regardless of dynamic or unexpected sensor position variations. The approach generalizes classic along-track gradients to an iterative, or recursive, difference, that can be applied irrespective of the amount of magnetic sensors and their positions within a dynamic measurement system, as long as these are known. The computed difference can be inverted directly, providing a versatile method with very little data pre-processing requirements, which we denote as recursive difference inversion. We explain the approach in a general setting, and expand it to provide a complete framework for Unexploded Ordnance (UXO) detection using a point-dipole model. Being an extension of classic along-track gradients, the method retains many of the same properties, which include added robustness to external time-dependent disturbances, and the ability to produce aesthetic visual data representations. In addition, the framework requires neither tie lines, data levelling, nor diurnal corrections. Only light pre-processing actions, namely initial survey trimming and data position calculation, are required. The method is demonstrated on data from a dual sensor system, conventionally referred to as a vertical gradiometer, which is towed from an Unmanned Aerial Vehicle. The system enables collection of high quality magnetic data in adverse settings, and simultaneously reduces the risk of inadvertent UXO detonations. To enable qualitative testing, we established a UXO detection test facility with several buried UXO, typical to World War II, in a magnetically complex in-land area. Data from the test facility was mainly used to evaluate inversion robustness and depth accuracy of the point-dipole model. Subsequently, we apply the method to real UXO survey data collected for the Hornsea II offshore wind farm project in the United Kingdom. This dataset was collected in a coastal setting, and subject to significant sensor position changes during flight due to varying wind conditions over multiple survey days. This makes the raw dataset challenging to interpret directly, but it can still be easily and reliably inverted for source locations through recursive difference inversion. In each of the two datasets, we attempt to recover UXO positions using recursive difference inversion on data from both a single sensor, as well as on data from two synchronized sensors, in each case inverting the difference directly for point-dipole model parameters. To seed the inversion, we propose a simple routine for picking out potential targets, based on the choice of a significant peak prominence in the time-series of computed differences. Higher order difference inversion was found to provide robust results in the magnetically complex setting, and the recovered equivalent dipole depths were found to approximate the actual UXO depths well.
- Inverse theory
- Numerical approximations and analysis
- Numerical modelling