This paper demonstrates an iterative optimisation algorithm, where a pre-defined low-frequency geoid model is simulated based on assumptions about mass-density distributions within the upper mantle. This aims to model lowfrequency spectra of the anomalous external Earth’s gravity field to produce a realistic synthetic Earth gravity model (SEGM). All mass anomalies are represented by an envelope of 3D discrete bodies (prisms) that refer to a regular geographic grid on a spherical reference surface. The optimisation algorithm uses forward gravity field modelling techniques based on Newton’s integral to derive the gravitational potential of each 3D mass element, and subsequently its effect on the synthetic (simulated) geoid height via Bruns’s formula. Geoid height differences from a given reference model (EGM2008) are minimised by applying the mass-model optimisation algorithm that iteratively modifies the volume (prism height) of each mass element introduced. Finally, the effectiveness of the optimisation algorithm is demonstrated through a numerical example for regional-scale geoid modelling over Austria. A mass-model was created that inversely produces a similar pattern compared to the observed gravity field given by EGM20008 where the differences are within several centimetres of the geoid height.