Researches On Theories And Methods Of Earth's Gravity Filed Recovery Based On Satellite-to-satellite Tracking Measurements

Since the launch of Sputnik 1 in 1957,the tracking data of the orbits have been used to develop the Earth's gravity field models.Based on the science data collected by the satellite gravity missions including CHAMP,GRACE and GOCE,the quality of the Earth's gravity field models have been improved significantly.In particular,the Earth's time-varying gravity field models compiled from the GRACE KBR data,which are made with the so-called low-low satellite-to-satellite tracking mode.Although the accuracy of the GRACE time-varying gravity field models have been increased considerably during the past decade,there still remains an offset in the order of one magnitude between the error level of present gravity field solutions and the GRACE baseline accuracy,i.e.the predicted accuracy from the pre-launch simulations.Therefore,efforts are still ongoing to identify the remaining error sources.Apart from the errors caused by aliasing effects,imperfections of the background models and measurement noise,unfiltered GRACE-only solutions suffer from pronounced stripes in the spatial domain,due to the anisotropic sensitivity of KBR data.To suppress noise in these solutions,one typically applies a filter.Unfortunately,filtering not only suppress noise,but also inevitably damps signal in the models produced.This triggers an idea of an improvement of such models not by applying a filter,but by using other sources of information in a combined modelling.On the other hand,the GRACE twin satellites have been in orbit for more than 15 years and now they are not able to sustainably provide time-varying gravity solutions due to the degraded quality of the onboard battery capacity.It's a consensus that the GPS data made with the high-low satellite-to-satellite tracking mode might play an important role to bridge the gap between the current GRACE and the future GRACE Follow-On mission.However,the precision of the GPS data is at least three orders of magnitude worse than that of the KBR data,thus the GPS data are only sensitive to the time-varying signals at very low degrees.Therefore,much efforts need to be made to further improve the accuracy of the solutions derived from the GPS data.To account for those issues mentioned above,a large amount of researches have been implemented during the preparation of this dissertation.The main contents and contributions of this dissertation are listed as follows:1.A novel method(denoted as 'phase method*)of deriving average satellite accelerations from the onboard GPS phase measurements is proposed.Within the phase method,the phase measurements are processed with an epoch-difference scheme,thus systematic errors common to adjacent epochs are largely eliminated.The GPS data collected by the GOCE mission are used to demonstrate the added value of the proposed method.Comparisons are made between the phase method and the orbit method,where the average accelerations are computed from the kinematic orbits with a double differentiation scheme.The results indicate that the average accelerations derived in this way are more precise,with noise being reduced by about 20%at the cross-track component and the cumulative geoid height errors of the produced models are reduced by 15%up to degree 50.An analysis in the spatial domain shows that large errors along the geomagnetic equator,which are caused by a high electron density coupled with large short-term variations,are substantially reduced.Finally,the new method allows for a better observation of mass transport signals.In particular,sufficiently realistic mass transport signals in North America and south-west Africa are obtained.2.The fundamental issues about the dynamic approach to gravity field modelling are discussed in details and the dynamic approach are further implemented in the PANDA software.Importantly,the frequency-dependent data weighting scheme is appled in the dynamic approach for the first time to account for the colored noise in the measurements.Monthly time-varying gravity field models are produced for the period spanning from January 2005 to December 2010 based on this implementation.Their performance is assessed by confronting them with other solutions.The results demonstrate an obvious advantage of the WHU model over other models when L1B data are used.To be specific,the WHU model is of comparable signal power to other models,but with a significantly lower noise level.A comparison made in the spectral domain show that the cumulative degree error is reduced by 35%as compared to the CSR model.A further analysis of mass changes in the spatial domain demonstrates that the correlation coefficients of mass changes in large river basins from other models with respect to those from the WHU model are all above 0.96.3.It has been shown that the DMT models are very sensitive to the errors in the GRACE dynamic orbits,as the orbits are used as known quantities during the derivation of the DMT solutions.This dissertation comes up with a new method for determination of the dynamic orbits,where the KBR data are inverted together with the kinematic/reduced-dynamic orbits.The relative precision of the dynamic orbits determined in this way is considerably improved and the DMT models produced with this newly computed orbits are more precise,with noise being reduced by at least 30%.4.The GRACE KBR data are combined with the GOCE GPS data processed with the phase method to derived a combined time-varying gravity field solutions(KG model).As compared to those solutions produced with the KBR data alone(KA model),the systematic errors manifested themselves by the pronounced north-south stripes in the spatial domain are reduced by about 70%for the months with short orbit repeat period in the case of the KG model.On the other hand,the optimally filtered solutions benefit much less from the incorporation of the GOCE GPS data in general,with the differences being lower than the typical noise level of the GRACE time-varying gravity solutions.However,the impacts might be considerable over areas with significant time-varying signals,where the impacts would reach up to 25%of the signal itself.5.The phase center variations(PCVs)are important systematic errors in GPS data processing.Their impacts on time-varying gravity field modelling are analyzed in this dissertation.The results indicate that the errors caused by the unmodeled PC Vs resemble the stripes typically observed in GRACE models.While their impacts on the DMT model can be safely neglected,they could reach to the noise level of GRACE time-varying gravity solutions over areas with significant signals in the case of the WHU model.6.The receiver clock offsets can be described with a deterministic function in the case of GRACE and GOCE satellites,where an Ultra Stable Oscillator is used.The potential benefits from receiver clock modelling for gravity field modelling are presented for the first time.The results demonstrate that with a piece-wise modelling,the precision of the kinematic orbits in the radial direction could be improved by at least 10%and the improvements are 3-4%in the context of gravity field modelling.