Estimation of variance in an ordinary adjustment model is straightforward, but if the model becomes unstable or ill-conditioned its solution and the variance of the solution will be very sensitive to the errors of observations. This sensitivity can be controlled by stabilizing methods but the results will be distorted due to stabilization. In this paper, stabilizing an unstable condition model using Tikhonov regularization, the estimations of variance of unit weight and variance components are investigated. It will be theoretically proved that the estimator of variance or variance components has not the minimum variance property when the model is stabilized, but unbiased estimation of variance is possible. A simple numerical example is provided to show the performance of the theory.
The Polar Regions are not covered by satellite gravity gradiometry data if the orbital inclination of the satellite is not equal to 90°. This paper investigates the feasibility of determining gravity anomaly (at sea level) by inversion of satellite gravity gradiometry data in these regions. Inversion of each element of tensor of gravitation as well as their joint inversion are investigated. Numerical studies show that gravity anomaly can be recovered with an error of 3 mGal in the north polar gap and 5 mGal in south polar gaps in the presence of 1 mE white noise in the satellite data. These errors can be reduced to 1 mGal and 3 mGal, respectively, by removing the regularization bias from the recovered gravity anomalies.
Computational time is an important matter in numerical aspects and it depends on the algorithm and computer that is used. An inappropriate algorithm can increase computation time and cost. The main goal of this paper is to present a vectorization algorithm to speed up the global gradiometric synthesis and analysis. The paper discusses details of this technique and its very high capabilities. Numerical computations show that the global gradiometric synthesis with 0.5° × 0.5° resolution can be done in a few minutes (6 minutes) by vectorization, which is considerable less compared to several hours (9 hours) by an inappropriate algorithm. The global gradiometric analysis of representation by spherical harmonics up to degree and order of 360, can be performed within one hour using vectorization, but if an inconvenient algorithm is used it can be delayed more than 1 day. Here we present the vectorization technique to gradiometric synthesis and analysis, but it can also be used in many other computational aspects and disciplines.
The topographic and atmospheric masses influence the satellite gravity gradiometry data, and it is necessary to remove these effects as precise as possible to make the computational space harmonic and simplify the downward continuation of such data. The topographic effects have been formulated based on constant density assumption for the topographic masses. However in this paper we formulate and study the effect of lateral density variation of crustal and topographic masses on the satellite gravity gradiometry data. Numerical studies over Fennoscandia and Iran show that the lateral density variation effect of the crust on GOCE data can reach to 1.5 E in Fennoscandia and 1 E in Iran. The maximum effect of lateral density variation of topography is 0.1 E and 0.05 E in Iran and Fennoscandia, respectively.
An ill-posed problem which involves heterogonous data can yield good results if the weight of observations is properly introduced into the adjustment model. Variance component estimation can be used in this respect to update and improve the weights based on the results of the adjustment. The variance component estimation will not be as simple as that is in an ordinary adjustment problem, because the result of the solution of an ill-posed problem contains a bias due to stabilizing the adjustment model. This paper investigates the variance component estimation in those ill-posed problems solved by the truncation singular value decomposition. The biases of the variance components are analyzed and the biased-corrected and the biased-corrected non-negative estimators of the variance components are developed. The derivations show that in order to estimate unbiased variance components, it suffices to estimate and remove the bias from the estimated residuals.
In solution of gradiometric boundary value problem in space a regular grid of satellite gravity gradiometry data is required. This grid is considered on a sphere with radius of the mean Earth sphere and altitude of satellite. However, the gravitational gradients are measured by a gradiometer mounted on GOCE satellite and orbital perturbations of the satellite influence GOCE observations as well. In this study we present that these effects are about 2 E on GOCE data. Also numerical studies on the gravitational gradients in orbital frame show that the perturbations of co-latitude are more significant than that of inclination. The effect of perturbed inclination is less than −9 mE while the effect of perturbed co-latitude is within −173 mE in one day revolution of GOCE.
The gravimetric model of the Moho discontinuity is usually derived based on isostatic adjustment theories considering floating crust on the viscous mantle. In computation of such a model some a priori information about the density contrast between the crust and mantle and the mean Moho depth are required. Due to our poor knowledge about them they are assumed unrealistically constant. In this paper, our idea is to improve a computed gravimetric Moho model, by the Vening Meinesz-Moritz theory, using the seismic model in Fennoscandia and estimate the error of each model through a combined adjustment with variance component estimation process. Corrective surfaces of bi-linear, bi-quadratic, bi-cubic and multi-quadric radial based function are used to model the discrepancies between the models and estimating the errors of the models. Numerical studies show that in the case of using the bi-linear surface negative variance components were come out, the bi-quadratic can model the difference better and delivers errors of 2.7 km and 1.5 km for the gravimetric and seismic models, respectively. These errors are 2.1 km and 1.6 km in the case of using the bi-cubic surface and 1 km and 1.5 km when the multi-quadric radial base function is used. The combined gravimetric models will be computed based on the estimated errors and each corrective surface.
In precise geoid modelling the combination of terrestrial gravity data and an Earth Gravitational Model (EGM) is standard. The proper combination of these data sets is of great importance, and spectral combination is one alternative utilized here. In this method data from satellite gravity gradiometry (SGG), terrestrial gravity and an EGM are combined in a least squares sense by minimizing the expected global mean square error. The spectral filtering process also allows the SGG data to be downward continued to the Earth’s surface without solving a system of equations, which is likely to be ill-conditioned. Each practical formula is presented as a combination of one or two integral formulas and the harmonic series of the EGM.Numerical studies show that the kernels of the integral part of the geoid and gravity anomaly estimators approach zero at a spherical distance of about 5°. Also shown (by the expected root mean square errors) is the necessity to combine EGM08 with local data, such as terrestrial gravimetric data, and/or SGG data to attain the 1-cm accuracy in local geoid determination.
There are numerous methods to modify Stokes’ formula with the usually common feature of reducing the truncation error committed by the lack of gravity data in the far-zone, resulting in an integral formula over the near-zone combined with an Earth Gravity Model that mainly contributes with the long-wavelength information. Here we study the reverse problem, namely to estimate the geoid height with data missing in a cap around the computation point but available in the far-zone outside the cap. Secondly, we study also the problem with gravity data available only in a spherical ring around the computation point. In both cases the modified Stokes formulas are derived using Molodensky and least squares types of solutions. The numerical studies show that the Molodensky type of modification is useless, while the latter method efficiently depresses the various errors contributing to the geoid error. The least squares methods can be used for estimating geoid heights in regions with gravity data gaps, such as in Polar Regions, over great lakes and in some developing countries with lacking gravity data.
There are different criteria for designing a geodetic network in an optimal way. An optimum network can be regarded as a network having high precision, reliability and low cost. Accordingly, corresponding to these criteria different single-objective models can be defined. Each one can be subjected to two other criteria as constraints. Sometimes the constraints can be contradictory so that some of the constraints are violated. In this contribution, these models are mathematically reviewed. It is numerically shown how to prepare these mathematical models for optimization process through a simulated network. We found that the reliability model yields small position changes between those obtained using precision respectively. Elimination of some observations may happen using precision and cost model while the reliability model tries to save number of observations. In our numerical studies, no contradictions can be seen in reliability model and this model seems to be more suitable for designing of the geodetic and deformation networks.