Research On Determination Of Marine Gravity Anomalies From Multi-satellite Altimeter Data

Recovery of marine gravity field is an important part of human's cognition about the ocean, and the acquired information is widely used in many fields, such as construction of earth gravity model, inversion of bathymetry, exploration of oil or gas resources and underwater matching navigation. Obtaining the fine structure of marine gravity field is of importance for scientific research, national economy and military affairs. Traditional ways to collect marine field information are mainly based on the measuring instruments that are deployed in seabed, and boarded on submarines or ships. However, due to the drawbacks of high-cost, sparse data coverage, long measurement period and poor repeatability in the traditional ways, it is impossible to obtain the global marine gravity information in a relatively short time.The successful lauch of oceanic remote sensing satellites promotes the deveploment in detecting oceanic gravity field significantly. Due to the notable advantages on spatial and temporal scales of the satellite altimeter data, it has gradually become the major data sources for inversing marine gravity filed since the 1970s. Although the theoretical framework of inversing marine gravity field from satellite altimetry materials has been fully developed, the way of further improvement on the resolution and accuracy of derived results remains a challenge and hot topic. In recent yeays, by changing the trajectories, adopting new mode for measurement and new band of radar impulse, the satellites can complement some high quality altimeter data with consideration to both accuracy and resolution, which provides a great opportunity to explore the finer structure of marine gravity field.In this thesis, the information of vertical deflection is firstly obtained by joint analysis of multi-satellites altimeter data including Geosat, ERS-1, Envisat, T/P, Jason-1, CryoSat-2 and SARAL/AltiKa, and then the global marine gravity anomaly with 1'×1' grid cells is inversed from the vertical deflection information. The main contents and contributions of this thesis can be summarized as following:1. The relationship between satellite altimeter observations and information of geoid height or vertical deflection is described in detail. Three formulas are discussed as the theoretical foundation of inversing marine gravity field from multi-satellites altimeter data. Firstly, inverse Stokes formula that computes gravity anomaly from geoid height is deduced into one dimensional convolution form. Secondly, inverse Vening-Meinesz formula with two different kernel functions and their corresponding convolution expressions are listed. Thirdly, the equations for calculating gravity disturbance and gravity anomaly from the vertical deflection are derived from Laplace equation.2. The measuring principle of radar measurement is detailly discussed. The changing process of instantaneous illumination pattern in the footprint area from radar impulse is systematically discussed from two hypothesises of sea surface, consisting of monochromatic, unidirectional wave trains or obeying Gaussian distribution. Based on previous discussions, the average returned waveform models in both time domain and frequency domain are given, and the relationships among waveform information and corresponding parameters of two-way travel time, significant wave height or normalized radar cross section are proposed. All the principles discussed above lay the theoretical foundation for retracking altimetry waveforms.3. The mathematical theories of OCOG retracker, Davis threshold retracker, advanced threshold retracker, (3-5 retracker and Sandwell retracker for obtaining the range corrections are induced. Taking Jason-1 GM data for instance, the correlation coefficients of numerical sequences for any two retrackers mentioned above is larger than 0.91 except results from OCOG, indicating that different retrackers can get consistent and similar range corrections. The statistical results of discrepancy before and after retracking at crossover points from two numerical examples show remarkable improvement and little difference for all the retrackers except OCOG. Moreover the standard deviation of along track measurement indicates that the Sandwell retracker is more effective for reducing noise level, which is 2cm smaller than that from other retrackers. According to the statistical results of along track measurement, SARAL/AltiKa has the lowest noise level, with a standard deviation less than 3cm, and is superior to other altimetry missions.4. A simplified approach of calculating along track vertical deflection is realized from the geoid gradient based on the approximate solution of spherical distance. The alternative method needs difference information of altimeter measurement and formulas of along track velocity for satellites. Basing on the Jason-1 GM SGDR data, the comparisons between the two methods show that the root mean square of discrepancy value is about 0.02?rad. Only a few points have relatively large differences above 0.1?rad and they generally distribute in the ocean areas surrounding the islands. This phenomenon is caused by the downward accuracy of spherical distance approximation in the first method. By adding random noises with different standard deviation, the effects of noise level on calculation accuracy are explored. The accuracy decreases linearly with the increment of standard deviation of random noise.5. The effect of different altimetry corrections on the calculation of along track vertical deflection is analyzed. Despite the numerical magnitudes of instrument correction, dry tropospheric correction, solid earth tide height and geocentric pole tide height are quite different, the regular variation along track makes these four items slightly affect the computation of vertical deflection. In contrast, the content of water vapor, liquid water and charged particles in the propagation path, the reflected difference of wave crests and troughs, ocean tide and atmospheric loading can significantly affect the calculation since these items vary more irregularly along track. Therefore, radiometer wet tropospheric correction, CSR4.0 ocean tide model, inverted barometer height correction and the along track smoothed dual frequency ionosphere delay correction and sea state correction need to be carefully added through comparison for each factor. For the altimeter data with single frequency, the modeled ionospheric correction is used without any smoothing process.6. The influence of sea surface topography on calculating vertical deflection is not ignorable. The correlation coefficients of quasi-stationary sea surface topography data sequences calculated from DOT2008A and RIO05 models reached 0.99, while the results for along track vertical deflection sequences are around 0.75. Meanwhile, the standard deviation derived from DOT2008A model is much smaller than that from RIO05. Additionally, the effect of time-varying sea surface topography is considered by computing sea surface height anomaly (SSHA) in two ways. The first way fits periodic term and trend term of SSHA with 7 parameters using the AVISO monthly averaged gridding data. The other way interpolates results through space-time correlation functions based on the observed SSHA from other on-orbit altimeter satellite. According to the statistical results of discrepancies at crossover points, the first method has a slight improvement less than lcm on the discrepancy, while the second method can significantly reduce the value by around 5cm if the time interval of statistical cycles is large enough. Due to the introduced noise during the interpolation, statistical result of the second method for a same cycle will conversely increase for a few millimeters.7. Basing on direct average and distance-weighted average approach, the Jason-1 GM data with an initial sampling ratio of 20Hz are resampled into 10Hz,5Hz,2Hz and 1Hz, respectively. The 1Hz corrections are also interpolated to corresponding sampling rate. The assessment of resampling results by discrepancy statistics at crossover points show negligible difference caused by average approach and corrections interpolation, and the 5Hz observation sequence can efficiently constrain the random noise in the raw data. According to the criterion that one half gain of the filter correspond to a wavelength of 6.7km, the lengths for the FIR filters are set to 49, 99 and 199 during the resample process for altimetry data with initial sampling rates of 10Hz,20Hz and 40Hz, respectively.8. The directional components of vertical deflection are determined separately at crossover points and grid points using the Jason-1 GM and CryoSat-2 data with significant difference between their orbit inclinations. Validation with respect to the EGM2008 model shows that the east component of vertical deflection derived from Jason-1 GM data is significantly superior to CryoSat-2, indicating that the diversity in orbit inclinations of satellites can improve the whole precision of vertical deflection. Besides, taking GOCO03S model as the reference field can further improve the accuracy of the determined vertical deflection, and solves the systematic bias problem over oceans with flatly changed gravity field information.9. Based on Geosat, ERS-1, T/P, Envisat, Jason-1, CryoSat-2 and SARAL/AltiKa data before and after necessary orbit maneuvers, parameters are firstly determined for using Sandwell retracker. After the process of resampling, adding corrections to sea surface height and subtracting sea surface topography, the along track slopes are obtained for each altimeter mission. Through EGM2008 model validation and histogram statistics, different thresholds are selected and used for probing data that includes gross errors. Comparison analysis shows that Parks-McClellan filter gets a better improvement on the signal to noise ratio than Gaussian filter. The multi-satellites slopes will be further treated in the following order:subtracting reference model, calculating along track residual vertical deflection, calculating directional components of residual vertical deflection at gridding points, computing residual gravity anomaly and restoring component of the reference model. A final marine gravity anomaly with 1'×1'resolution at China sea and its adjacent areas (100°?140°E,0°?40°N) is then inversed using the above procedures. Comparison with DTU10, DTU13, V23.1 model and ship-measured data shows that the new solution has a similar accuracy level with DTU13 and V23.1, and is superior to DTU10 and EGM2008. Moreover, the calculation precision in open ocean areas is about 2mGal better than the results in coastal shallow areas. It is also concludes that the solutions with high-order gravitational field as the reference model has higher precision by comparing the solutions from 4 different reference models, including GOCO03S, EIGEN6C4, GECO and EGM2008. Furthermore, results from EGM2008 and EIGEN6C4 are slightly better than the GECO solution.10. Taking the EGM2008 model as reference model, the global marine gravity field with 1'×1' grid cells is obtained through retracking, resampling, correcting, filtering and other processing steps based on the multi-satellites altimetry data. In the selected sample areas, the root mean square of the model verification is in the range of 1.7mGa1?3.6mGal, and the validated accuracy with ship-measured data is around 4mGal.