Abstract

Three-dimensional (horizontal, vertical and ellipsoidal heights) coordinates of a point can be obtained using the GNSS (global navigation satellite system) systems. Although, ellipsoid heights have a geometrical meaning, height is actually a geopotential value. Therefore, determination of physical reference surface (sea level or geoid) is of particular importance in geodetic studies. Although, GNSS observations provide ellipsoidal heights (h), in practice, orthometric height (H) is used. Since the relationship between these two heights is H=h-N, geoid undulations (N) that provide transition between ellipsoidal heights and orthometric heights need to be known. In this comparison of geoid height calculation methods, a network of 126 points was chosen within the borders Adiyaman in South-eastern Turkey. Reference geoid heights were determined according to the GNSS/levelling method across the entire network. The heights of these points were then remodelled according to both the adaptive neural fuzzy inference systems (ANFIS) and the polynomial coefficients methods. In the modelling procedure, 106 points were used to develop the two models and the remaining 20 points were used to test the results obtained from each model. The geoid heights obtained via both remodelling methods were compared with reference data from the GNSS/levelling method and the results of the comparison were interpreted.

Key words: Geoid, geoid height, global positioning system (GPS)/levelling, fuzzy inference systems, surface polynomials, adaptive neural fuzzy inference systems (ANFIS).