Determination of local jeoid modelling using GNSS/levelling data by various interpolation and artifical intelligence methods

GNSS surveying methods are more preferred than ground surveying methods thanks to recently developing technology in cartography activities. The third dimension of the position information called as height obtained by GNSS surveying, is defined as the ellipsoid height. However, orthometric heights are needed to be able to physically describe a point on the ground. Thus, the geoid height must be known in order to convert from ellipsoid height to orthometric height. Geoid heights can be calculated by using the existing local geoid model in the study area as suggested in the Large Scale Map and Map Information Production Regulation (BÖHHBÜY). In this thesis, a study area with four hundred and fifty-five C3 degree points are identified in Trabzon to obtain local geoid. Two models with different point density are generated for the study region moreover; multivariate adaptive regression curves, polynomials, radial based functions, kriging interpolation methods, artificial neural networks and least-squares support vector machines are used in order to model the geoid height. As a result of this study, the most accurate methods are multi-layer perceptions of artificial neural networks, least squares support vector machines and multivariate adaptive regression curves respectively in modeling geoid heights.