Artificial neural network versus surface polynomials for determinetion of local geoid

National heights systems throughout the world are referenced to the geoid whereas heighting by GPS directly results in ellipsoidal heights. These GPS-defined ellipsoidal heights have to be transformed into the national height systems. In practice, the surface polynomials are the most practical method to perform the transformation. Because of advances in computer science, surveyors have had an alternative method, the artificial neural networks, to the surface polynomials. In this study, these two methods are compared to each other using the data from Istanbul, Turkey. The investigations carried out show that the neural network method provides comparable results in modeling general characteristics with the conventional polynomial, but better results in modeling local characteristics.

National heights systems throughout the world are referenced to the geoid whereas heighting by GPS directly results in ellipsoidal heights. These GPS-defined ellipsoidal heights have to be transformed into the national height systems. In practice, the surface polynomials are the most practical method to perform the transformation. Because of advances in computer science, surveyors have had an alternative method, the artificial neural networks, to the surface polynomials. In this study, these two methods are compared to each other using the data from Istanbul, Turkey. The investigations carried out show that the neural network method provides comparable results in modeling general characteristics with the conventional polynomial, but better results in modeling local characteristics.