GOCE sensitivity studies in terms of cross-over analysis

Abstract

The GOCE (Gravity field and steady-state Ocean Circulation Explorer) satellite, launched on 17 March 2009, for the first time applies satellite gravity gradiometry (SGG) to recover the Earth's gravity field with cm accuracy at a resolution of 100km. To meet the envisaged accuracy, measurement validation at cross-over points (XOs) is necessary. Typically, validation is based on gravity gradients (GGs). However, the coefficient matrix of the gravitational tensor is dependent on orientation. In order to avoid matrix rotation, analysis based on orientation-independent invariants is possible. By applying various noise models, the goodness of XO-validation based on GGs and invariants will be studied in this thesis. First, by determining the maximum of scalar products from two tracks, the XOs can be predicted. Next, using local polynomial approximation, the geographical coordinates of XOs are calculated by solving a system of equations. Due to the orbit drift, the interpolation of height is performed separately along ascending and descending track before final comparison. Considering a sampling rate of 1Hz, GGs and invariants in all points of a one-week orbit are simulated for the further interpolation at the XOs. To determine the goodness of the selected interpolation algorithm, a closed loop test with noise-free data is investigated first. Since signal to noise ratios of GGs and invariants are all above 70dB, the same algorithm is applied in closed loop tests with noisy data. Since GOCE can only provide high accuracy for the main diagonal tensor components, various noise models, i.e. homogeneous and inhomogeneous white noise as well as homogeneous and inhomogeneous coloured noise, are added to the simulated values. The comparison of the goodness of GGs opposed to invariants is based on the signal to noise ratio (SNR). In this study, the second invariant demonstrates better SNR than GGs and the third invariant in the case of homogenous noise. However, due to the impact of inaccurate GGs, the SNR of invariants is poorer than the SNR of all GGs in the case of inhomogeneous noise.