Modelling geoid height errors for local areas based on data of global models

Abstract The development of global geoid models became feasible following the launch of specialised satellite missions. Today, the root mean square deviation of the heights in global models of high degree & order varies from centimetres to decimetres across different countries. In countries where the accuracy of such models is lower, there is potential to enhance their precision by applying specific corrections. This study presents a novel methodology for locally modelling the height errors of high degree & order global geoid models using levelling sub-benchmarks for GNSS stations. The methodology is based on a combination of optimal interpolation methods, filtering, and the concept of data weighting by gravity anomaly differences. The methodology is aimed at creating a hybrid model that aligns with the local characteristics of the geoid (or quasi-geoid) derived from the traditional levelling network. The advantage of this methodology lies in its ability to reduce the residual height errors of the EGM2008 and EIGEN6C4 models to less than 1 cm when using only four control points. Such results exceed the initial values of the systematic height errors of these models by 90–96 %. For the GECO model, the residual errors are around 2 cm, while for the XGM2019e₂159 model, they reach 3 cm. These results indicate that this methodology can be applied to all global models of high degree & order, although its effectiveness may vary depending on the specifics of a particular model.