Statistical prediction of satellite magnetic anomalies

The geological utility of satellite magnetic observations is limited by orbital altitude variations which may be as large as a few hundred kilometres. This study investigates the use of fast and elegant statistical procedures for altitude normalization and gridding of magnetic anomaly data as an alternative to more commonly used equivalent source inversion procedures involving computationally extensive and complex least‐squares matrix methods. A standard statistical approach for gridding satellite magnetic anomalies is to recompute numerically averaged values from three‐dimensionally distributed observations which are within two standard deviations of an initially determined averaged anomaly estimate. the errors of this procedure for geological analysis are investigated using orbital anomaly simulations of lithospheric sources over a spherical earth. the simulations suggest that numerical averaging errors constitute small and relatively minor contributions to the total error‐budget of higher orbital estimates (≳400 km), whereas for lower orbital estimates the error of averaging may increase substantially. A more complex statistical procedure involving least‐squares collocation in 3‐D is found to produce substantially more accurate anomaly estimates as the elevation of prediction is decreased towards the lithospheric sources. Moreover, 3‐D collocation is computationally much more efficient and faster to apply than equivalent source inversion methods for altitude‐normalizing and gridding magnetic anomaly data. Application of this procedure to MAGSAT magnetic observations of South America demonstrates its utility for producing accurately gridded magnetic anomalies at constant elevation for geological analysis.