Geoid Computations in Taiwan

Abstract

Different geoid solutions were carried out in the Taiwan island and in the surrounding sea region. These local gravimetric geoid predictions were based on terrestrial gravity data, satellite altimeter data, and height data, using as a reference surface the OSU91A geopotential model. Gravity anomalies derived from GEOSAT and ERS-1 altimetry data and irregularly distributed gravity data derived by individual sea and land surveys were gridded on a 3’x3’ geographical grid and in this form were used in the numerical tests. The geoid height computations were done in the space domain by the least squares collocation (LSC) method, and in the frequency domain by the fast Fourier transform (FFT) technique on the plane and on the sphere, which resulted in improvements in terms of accuracy, computational efficiency and computer memory required. The computed geoid heights were compared with 26 GPS stations along a 200-km-long first-order leveling line. Results show that the differences between control and predicted geoid heights have a standard deviation (sd) at the 12 to 17 cm level, with differences between methods at 2 to 7 cm (1σ). By fitting a four-parameter level to the geoid predictions (corresponding to a GPS, leveling or geoid datum shift and scale change), long wavelength errors were substantially reduced, with all methods yielding geoid fits of about 4 cm. The relative accuracies achieved were of the order of 0.4 to 0.8 ppm for a mean baseline of 100 km. On an alternative data set of 24 GPS points in the high mountains less accurate results were obtained, in part due to lack of sufficiently accurate heights of the GPS points.