Navigating from Magnetic Maps – Improving Reliability of Magnetic Anomaly and Gradient Reference Values from Imperfect Data

Alternative magnetic navigation (aka MagNav) depends on matching of onboard sensor readings with prior mapping of the Earth’s magnetic field. Three critical components to successful MagNav are (1) sensor accuracy, (2) magnetically clean sensor platform with compensation for any residual platform effects, and (3) reference map quality. Relatively inexpensive magnetic sensors are sufficient for successful MagNav. Established methods are sufficient to calibrate all but the noisiest platforms. The purpose of this presentation is to dig into (3); specifically, how to assess the quality of directional magnetic anomaly gradients (DMAG) from magnetic maps and deliver the most reliable reference values to nav systems. At each navigation time step a comparison is made between the measured and mapped gradients. Successful MagNav depends on the ability to quantify the level of match between these values. A critical component is an understanding of the characteristics of uncertainty in the estimation of DMAG for input to the navigation filter algorithm.

Current navigation systems ingest magnetic map data as a “stack of grids” prepared from an original survey grid by upward continuation (typically using an FFT method). Anomaly values and east-west/north-south gradients are interpolated from these grids for comparison with magnetic sensor data in the navigation filter. The reliability of these anomaly and gradient values is dependent on several factors, including the uncertainty in the original survey grid, edge effects or other artefacts from upward continuation, and method of grid interpolation. The use of an equivalent source model instead of a stack of static grids offers opportunities for uncertainty propagation and ability to query anomaly and gradient values and related uncertainty at arbitrary locations and cadence. The purpose of this presentation is to give comparative examples of DMAG evaluations using different methods applied to synthetic and actual data.