The cause and treatment of anisotropic errors in near-Earth geomagnetic data

Abstract The data available to constrain models of the geomagnetic field are dominated by measurements made at or near to the Earth's surface, typically treated as independent and uncorrelated. Here we examine the effect on field modelling of error anisotropy in these data, potentially leading to correlations between different field components measured at the same location. We consider three examples. First, we examine the effect of crustal magnetisation, in general the largest source of error in main field modelling. We demonstrate that under a stochastic treatment of the crustal field, the error resulting from this source is larger by a factor of approximately 2 in Z than in X and Y. and that correct modelling of this effect reduces the variances of the field model coefficients by greater than 10%. The anisotropy leads to error correlation between measurements of inclination and field intensity. Second, we consider the calculation of cartesian component vector data from measurements of field intensity and direction. When measurement error is the dominant source of uncertainty, the error estimates for the resulting vector data can be strongly correlated. Third, we examine the proposed deployment of a set of ocean-bottom magnetic observatories. One important difficulty with this project is the determination of accurate horizontal orientation. We demonstrate that with correct modelling, orientation information of limited accuracy is sufficient to allow determination of high-quality field models.