Kriging and moving window kriging on a sphere in geometric (GNSS/levelling) geoid modelling

Abstract

A comparison of kriging and moving window kriging (MWK) on a sphere is performed on GNSS/levelling data. The study was to give the answer on whether there is a significant gain in prediction accuracy when we apply an MWK instead of ‘classical’ kriging and also to what extent the use of global geopotential model EGM2008 improves prediction. The quality of prediction for all kriging and data variants has been investigated on three regions (being on the territory of the conterminous USA) characterised with a different spatial extent and density of sampling. Numerical tests revealed that in case of high-sampling density there was no accuracy gain when using MWK instead of classical kriging (cK). In contrast, for less numerous datasets and a much larger spatial extent (low-sampling density) MWK adapts itself to data much better than cK. Incorporation of EGM2008-based undulations as a long-wavelength trend for both cases (classical and moving window) significantly improved prediction quality.