Modelling geoid undulations with an artificial neural network

Abstract

Examines the use of a backpropagation neural network to model geoid undulations. Modelling of the Earth's gravity field, and in particular the separation between ellipsoid and geoid surface, is one of the fundamental problems in the field of geodesy. Geoid undulations are important for relating heights derived from the satellite based Global Positioning System to orthometric heights, which determine the flow of water. Modelling of geoid undulations has been traditionally done using Stokes integral, least squares collocation, or by fast Fourier transforms. The paper presents the results of preliminary investigations which suggest the backpropagation neural network provides a useful tool for geoid undulation modelling.