Optimisation of point grids in regional gravity field analysis

In this contribution, we propose a Monte Carlo algorithm in a Bayesian framework to optimise the configuration of local basis functions jointly with their scaling coefficients. The problem can be formulated as a quasi-linear model. Quasi-linear models are non-linear ones, but linearity holds for a subset of the searched-for parameters. This allows us to utilise the analytical solution of the linear model, while the unknown parameters in the design matrix are kept fix. As a particular advantage of this approach, already existing gravity recovery software can be used in a black-box manner. The new method is analysed theoretically, followed by a demonstration of its applicability and performance by examples from the field of regional gravity field recovery from GOCE satellite data.