Abstract:Curve surface fitting methods have been extensively utilized for the task of conversion GPS elevation to normal elevation. However, errors of the coordinates of GPS observations maybe propagate into the transformed normal elevation during estimating the curve surface parameters. To analyze the influence of these errors in the observation data on the transformed elevation, we employed three models: the quadratic polynomial curve surface, cubic polynomial curve surface and moving quadratic curve surface, to study the accuracy of the converted normal elevation. LeastSquares and two algorithms: Total LeastSquares and Weighted Total LeastSquares to estimate the parameters of these curve surfaces. The results from numerical experiments confirm that the accuracy of normal elevation have been considerably improved when errors of the coordinates of GPS observations data were modeled during the process of transformation.