Gravity field approximation in the area of Greece with emphasis on local characteristics

The accuracy of the gravity field approximation depends on the amount of the available data and their distribution as well as on the variation of the gravity field. The variation of the gravity field in the Greek mainland, which is the test area in this study, is very high (the variance of point free air gravity anomalies is 3191.5mgal ²). Among well known reductions used to smooth the gravity field, the complete isostatic reduction causes the best possible smoothing, however remain strong local anomalies which disturb the homogeneity of the gravity field in this area. The prediction of free air gravity anomalies using least squares collocation and regional covariance function is obtained within a ±4 ... ±19mgal accuracy depending on the local peculiarities of the free air gravity field. By taking into account the topography and its isostatic compensation with the usual remove-restore technique, the accuracy of the prediction mentioned obove was increased by about a factor of 4 and the prediction results become quite insensitive to the covariance function used (local or regional). But when predicting geoidal heights, in spite of using the smoothed field, the prediction results remain still depend on the covariance function used in such a way that differences up to about 50cm/100km result between relative geoidal heights computed with regional or local covariance functions.     

Densifying 3D GPS Networks by Accurate Transformation of Vector Components

A primary output of Global Positoning System (GPS) post-processed data is a set of non-trivial (independent) vector components and their full covariance information referred to a specific local Cartesian terrestrial frame (e. g., ITRF, WGS84) and epoch. It is important to recognize that when GPS-determined vector components are simultaneously combined into 3D geodetic network adjustments, they should always refer to a common coordinate frame and epoch. This paper uses geometric concepts to formulate rigorous matrix transformations to correct vector components for changes in coordinate systems, secular displacements due to plate rotations, and antenna centering and/or height measuring errors. Finally, the associated variance-covariance matrix of the transformed vector components is derived. ? 2001 John Wiley & Sons, Inc.     

A universal formula of maximum likelihood estimation of variance-covariance components

The derivation of a universal formula for the variance-covariance component estimation is discussed. The formula is derived adopting the universal functional model (the condition adjustment with unknown parameters and constraints among the parameters),and is based on the maximum likelihood principle. The derived formula in this paper can be applied to all adjustment models for estimating variance-covariance components, which expands the formulas given by K. Kubik (1970)and K. R. Koch (1986).Besides, it is proved that the estimator given in this paper is equivalent to that of Helmert type and best quadratic unbiased estimation (BQUE).     

On a four-dimensional integrated geodesy

In contrast to continuous global considerations of time dependent boundary value problems an attempt is made to define4D-linear observation equations in the framework of integrated geodesy for discrete, more or less regional and local applications (deformation analysis) where time variations in position and in the gravity field have to be considered. The derivation is a strict analogue and extension of the3D integrated approach. In addition the construction of time dependent covariance functions is discussed, which are necessary to solve for unknown displacements and changes in the gravity potential in the generalized least squares collocation model.     

基于等效残差的方差-协方差分量估计

首先概括性地阐述方差-协方差分量估计(VCE)理论的发展历史与研究现状;利用正交分解提取出等效残差,建立VCE的基本方程,在此基础上阐述VCE算法的局部最优特性和可能出现负定协方差阵估计结果等两大问题,并分析可能的解决方案及其复杂性;导出初值给定的Helmert,最小二乘和MINQUE VCE的线性逼近估计公式,证明基于等效残差的估计公式与已有VCE公式等价,各种估计方法在本质上是一致的;最后,用两个算例验证本文的观点.     

Best invariant covariance component estimators and its application to the generalize multivariate adjustment of heterogeneous deformation observations

Repeated measurements, like relevelling or periodically observed control networks, e.g., may be adjusted within the Simple Multivariate Gauss-Markov Model if the coefficient matrices are identical at each epoch and if the observations are homogeneous yielding proportional variance-covariance matrices between each two (not necessarily different) epochs. However, when processing data concerning recent crustal movements both conditions are violated, in general, such that the usual approach becomes inapplicable. Therefore, a procedure is proposed for this case which makes a joint adjustment possible within the Generalized Multivariate Regression Model of Heterogeneous Observations and which further coincides with the standard multivariate analysis in the special case.     

On the reliability of the VCV Matrix: A case study based on GAMIT and Bernese GPS Software

Commonly, the variance-covariance (VCV) matrix derived from GPS processing software underestimates the magnitude of the error, mainly due to the fact that physical correlations are normally neglected. The GAMIT and Bernese software packages serve the scientific community as important tools for GPS measurement processing and analyzing, especially in precise applications. Therefore, the reliability of the VCV matrices derived by the GAMIT and Bernese packages is of great importance. Formal accuracies derived from both software need to be scaled by applying a scaling factor (SF) that multiplies the software-derived formal errors. However, to the best of our knowledge, no standard approach approved by the GPS community exists. In this report, an analysis is carried out in order to test the reliability and the validity of the VCV matrices in both software, and to provide SFs needed to calculate the realistic accuracies reflecting the actual error levels. The method applied in this study allows deriving SFs for formal accuracies obtained from GAMIT and Bernese. The results attained from the time series of eight days for eight baselines (lengths of 20–415 km) indicate that the overall SF for GAMIT is more than 10 times smaller than for Bernese (1.9 and 23.0, respectively). Although no distance-dependent SF was detected in either case, the session-duration dependence was detected for the Bernese software, while no clear session-duration dependence was observed for the GAMIT. Furthermore, no receiver/antenna dependence could be deduced from the results of this analysis.     

Fractional vegetation cover estimation in heterogeneous areas by combining a radiative transfer model and a dynamic vegetation model

ABSTRACT

A fractional vegetation cover (FVC) estimation method incorporating a vegetation growth model and a radiative transfer model was previously developed, which was suitable for FVC estimation in homogeneous areas because the finer-resolution pixels corresponding to one coarse-resolution FVC pixel were all assumed to have the same vegetation growth model. However, this assumption does not hold over heterogeneous areas, meaning that the method cannot be applied to large regions. Therefore, this study proposes a finer spatial resolution FVC estimation method applicable to heterogeneous areas using Landsat 8 Operational Land Imager reflectance data and Global LAnd Surface Satellite (GLASS) FVC product. The FVC product was first decomposed according to the normalized difference vegetation index from the Landsat 8 OLI data. Then, independent dynamic vegetation models were built for each finer-resolution pixel. Finally, the dynamic vegetation model and a radiative transfer model were combined to estimate FVC at the Landsat 8 scale. Validation results indicated that the proposed method (R²?=?0.7757, RMSE?=?0.0881) performed better than either the previous method (R²?=?0.7038, RMSE?=?0.1125) or a commonly used method involving look-up table inversions of the PROSAIL model (R²?=?0.7457, RMSE?=?0.1249).     

A Pattern‐Based Approach for Matching Nodes in Heterogeneous Urban Road Networks

This article presents an approach to hierarchical matching of nodes in heterogeneous road networks in the same urban area. Heterogeneous road networks not only exist at different levels of detail (LoD), but also have different coordinate systems, leading to difficulties in matching and integrating them. To overcome these difficulties, a pattern‐based method was implemented. Based on the authors' previous work on detecting patterns of divided highways, complex road junctions, and strokes to eliminate the LoD effect of road networks, the proposed method extracts the local networks around each node in a road network and uses them as the matching units for the nodes. Second, the degree of shape similarity between the matching units is measured using a Minimum Road Edit Distance based on a transformation. Finally, the proposed method hierarchically matches the nodes in a road network using the Minimum Road Edit Distance and eliminates false matching nodes using M‐estimators. An experiment involving matching heterogeneous road networks with different LoDs and coordinate systems was carried out to verify the validity of the proposed method. The method achieves good and effective matching regardless of differences in LoDs and road‐network coordinate systems.     

Calibration of geoid error models via a combined adjustment of ellipsoidal, orthometric and gravimetric geoid height data

The well-known statistical tool of variance component estimation (VCE) is implemented in the combined least-squares (LS) adjustment of heterogeneous height data (ellipsoidal, orthometric and geoid), for the purpose of calibrating geoid error models. This general treatment of the stochastic model offers the flexibility of estimating more than one variance and/or covariance component to improve the covariance information. Specifically, the iterative minimum norm quadratic unbiased estimation (I-MINQUE) and the iterative almost unbiased estimation (I-AUE) schemes are implemented in case studies with observed height data from Switzerland and parts of Canada. The effect of correlation among measurements of the same height type and the role of the systematic effects and datum inconsistencies in the combined adjustment of ellipsoidal, geoid and orthometric heights on the estimated variance components are investigated in detail. Results give valuable insight into the usefulness of the VCE approach for calibrating geoid error models and the challenges encountered when implementing such a scheme in practice. In all cases, the estimated variance component corresponding to the geoid height data was less than or equal to 1, indicating an overall downscaling of the initial covariance (CV) matrix was necessary. It was also shown that overly optimistic CV matrices are obtained when diagonal-only cofactor matrices are implemented in the stochastic model for the observations. Finally, the divergence of the VCE solution and/or the computation of negative variance components provide insight into the selected parametric model effectiveness.     

Estimation and modelling of the local empirical covariance function using gravity and satellite altimeter data

The estimation of a local empirical covariance function from a set of observations was done in the Faeroe Islands region. Gravity and adjusted Seasat altimeter data relative to theGPM2 spherical harmonic approximation were selected holding one value in celles of1/8°×1/4° covering the area. In order to center the observations they were transformed into a locally best fitting reference system having a semimajor axis1.8 m smaller than the one ofGRS80. The variance of the data then was273 mgal ² and0.12 m ² respectively. In the calculations both the space domain method and the frequency domain method were used. Using the space domain method the auto-covariances for gravity anomalies and geoid heights and the cross-covariances between the quantities were estimated. Furthermore an empirical error estimate was derived. Using the frequency domain method the auto-covariances of gridded gravity anomalies was estimated. The gridding procedure was found to have a considerable smoothing effect, but a deconvolution made the results of the two methods to agree. The local covariance function model was represented by a Tscherning/Rapp degree-variance model,A/((i−1)(i−2)(i+24))(R _B /R _E )²ⁱ⁺², and the error degree-variances related to the potential coefficient setGPM2. This covariance function was adjusted to fit the empirical values using an iterative least squares inversion procedure adjusting the factor A, the depth to the Bjerhammar sphere(R _E −R _B ), and a scale factor associated with the error degree-variances. Three different combinations of the empirical covariance values were used. The scale factor was not well determined from the gravity anomaly covariance values, and the depth to the Bjerhammar sphere was not well determined from geoid height covariance values only. A combination of the two types of auto-covariance values resulted in a well determined model.     

Local characteristics of the gravity anomaly covariance function

Summary Using a data set of 260 000 gravity anomalies it is shown that common characteristics for a local covariance function exist in an area as large as Canada excluding the Rocky Mountains. After eliminating global features by referencing the data to the GEM-10 satellite solution, the shape of the covariance function is remarkably consistent from one sample area to the next. The determination of the essential parameters and the fitting of the covariance function are discussed in detail. To test the reliability of the derived function, deflections of the vertical are estimated at about 230 stations where astrogeodetic data are available. Results show that the standard error obtained from the discrepancies is about1″ for each component and that the error covariance matrix of least-squares collocation reflects this accuracy remarkably well.     

Uniquely and overdetermined geodetic boundary value problems by least squares

Least-squares by observation equations is applied to the solution of geodetic boundary value problems (g.b.v.p.). The procedure is explained solving the vectorial Stokes problem in spherical and constant radius approximation. The results are Stokes and Vening-Meinesz integrals and, in addition, the respective a posteriori variance-covariances. Employing the same procedure the overdeterminedg.b.v.p. has been solved for observable functions potential, scalar gravity, astronomical latitude and longitude, gravity gradients Г_xz, Г_yz, and Г_zz and three-dimensional geocentric positions. The solutions of a large variety of uniquely and overdeterminedg.b.v.p.'s can be obtained from it by specializing weights. Interesting is that the anomalous potential can be determined—up to a constant—from astronomical latitude and longitude in combination with either {Г_xz,Г_yz} or horizontal coordinate corrections Δx and Δy, or both. Dual to the formulation in terms of observation equations the overdeterminedg.b.v.p.'s can as well be solved by condition equations. Constant radius approximation can be overcome in an iterative approach. For the Stokes problem this results in the solution of the “simple” Molodenskii problem. Finally defining an error covariance model with a Krarup-type kernel first results were obtained for a posteriori variance-covariance and reliability analysis.     

Simulation of regional gravity field recovery from satellite gravity gradiometer data using collocation and FFT

When planning a satellite gravity gradiometer (SGG) mission, it is important to know the quality of the quantities to be recovered at ground level as a function of e.g. satellite altitude, data type and sampling rate, and signal variance and noise. This kind of knowledge may be provided either using the formal error estimates of wanted quantities using least-squares collocation (LSC) or by comparing simulated data at ground level with results computed by methods like LSC or Fast Fourier Transform (FFT). Results of a regional gravity field recovery in a 10^o×20^o area surrounding the Alps using LSC and FFT are reported. Data used as observations in satellite altitude (202 or161 km) and for comparison at ground level were generated using theOSU86F coefficient set, complete to degree 360. These observations are referred to points across simulated orbits. The simulated quantities were computed for a 45 days mission period and 4 s sampling. A covariance function which also included terms above degree 360 was used for prediction and error estimation. This had the effect that the formal error standard deviation for gravity anomalies were considerably larger than the standard deviations of predicted minus simulated quantities. This shows the importance of using data with frequency content above degree 360 in simulation studies. Using data at202 km altitude the standard deviation of the predicted minus simulated data was equal to8.3 mgal for gravity and0.33 m for geoid heights.     

An improved probabilistic relaxation method for matching multi-scale road networks

Matching multi-scale road networks in the same area is the first step in merging two road networks or updating one based upon the other. The quality of the merge or update depends greatly on the matching accuracy of the two road networks. We propose an improved probabilistic relaxation method, considering both local and global optimizations for matching multi-scale of road networks. The aim is to achieve local optimization, as well as to address the identification of the M:N matching pattern by means of inserting virtual nodes to achieve global optimization effects. Then, by adding two attribute-related evaluation indicators, we developed four evaluation indicators to evaluate the matching accuracy, considering both geographic and attribute information. This paper also provides instructions on how to identify the proper buffer threshold during matching procedures. Extensive experiments were conducted to compare the proposed method with the traditional approach. The results indicate that: (1) the overall matching accuracy of each evaluation indicator exceeds 90%; (2) the overall matching accuracy increases by 6–12% after an M:N matching pattern is added, and by 4–6% following the addition of topology indicators; and (3) the proper buffer threshold is about twice the average value of the closest distance from all nodes.     

The filtering effect of orbit correction on geopotential errors

Summary The geophysical interpretation of satellite tracking residuals generally ignores the filtering effect of initial orbit correction on the true errors of the model. While the filtered information is usually regarded as lost, knowing the spectral characteristics of the filter is a great aid in the detailed interpretation of residuals, especially of global data sets. In this regard, we derive the filter characteristics (admittances) of orbit correction in the presence of geopotential-caused trajectory errors. We then apply the filter to determine the likely power of the lost radial information in crossover differences of sea heights determined from satellite altimetry or in the latitude lumped coefficients derived from them. For example, we find that resonant geopotential information with periods longer than the corrected orbit's arc length is largely lost in residual crossover data. Results are given for GEOSAT, ERS-1 and TOPEX/Poseidon in their Exact Repeat Missions, using calibrated variancecovariance matrices of the harmonic geopotential coefficients of several recent Earth gravity models. To prove that filtering is important, we first employed a simple cut of all perturbing terms with periods longer than the general tracking period (4 days for GEOSAT and ERS-1, and 10 days for TOPEX). But the cut is too crude a method from a theoretical viewpoint, and thus, we developed two new filters. A comparison of their admittances explains the differences (and sometimes anomalous behaviour) between them and the cut. Many numerical examples (single-satellite crossover errors and latitude lumped coefficient errors, as projected from the variance-covariance matrices) are presented.This paper has been presented during the Panel on Satellite Dynamics, at COSPAR 1994, in Hamburg, Germany.     

Error-covariances of the estimates of spherical harmonic coefficients computed by LSC,using second-order radial derivative functionals associated with realistic GOCE orbits

Least-squares collocation may be used for the estimation of spherical harmonic coefficients and their error and error correlations from GOCE data. Due to the extremely large number of data, this requires the use of the so-called method of Fast Spherical Collocation (FSC) which requires that data is gridded equidistantly on each parallel and have the same uncorrelated noise on the parallel. A consequence of this is that error-covariances will be zero except between coefficients of the same signed order (i.e., the same order and the same coefficient type C–C or S–S). If the data distribution and the characteristics of the data noise are symmetric with respect to the equator, then, within a given order and coefficient type, the error-covariances amongst coefficients whose degrees are of different parity also vanish. The deviation from this “ideal” pattern has been studied using data-sets of second order radial derivatives of the anomalous potential. A total number of points below 17,000 were used having an equi-angular or an equal area distribution or being associated with points on a realistic GOCE orbit but close to the nodes of a grid. Also the data were considered having a correlated or an uncorrelated noise and three different signal covariance functions. Grids including data or not including data in the polar areas were used. Using the functionals associated with the data, error estimates of coefficients and error-correlations between coefficients were calculated up to a maximal degree and order equal to 90. As expected, for the data-distributions with no data in the polar areas the error-estimates were found to be larger than when the polar areas contained data. In all cases it was found that only the error-correlations between coefficients of the same order were significantly different from zero (up to 88%). Error-correlations were significantly larger when data had been regarded as having non-zero error-correlations. Also the error-correlations were largest when the covariance function with the largest signal covariance distance was used. The main finding of this study was that the correlated noise has more pronounced impact on gridded data than on data distributed on a realistic GOCE orbit. This is useful information for methods using gridded data, such as FSC.     

Generalyzed variance-covariance propagation law formulae and application to explicit least-squares adjustments

Standard formulae overlook the contribution of a number of terms in the derivation of variance-covariance matrices for parameters in nonlinear least squares adjustment. In a large class of nonlinear mathematical models, these terms can contribute to an important error in the estimation of parameter variances. Improved formulae are derived. A numerical example is given and the use of our improved formula in the case of least-squares adjustment in the explicit case (L=F(X)) is fully documented.     

Sub-pixel classification of SPOT-VEGETATION time series for the assessment of regional crop areas in Belgium

Global time series of low resolution images are available with high repeat frequency and at low cost, but their analysis is hampered by the presence of mixed pixels and the difficulty in locating detailed spatial features. This study examined the potential of sub-pixel classification for regional crop area estimation using time series of monthly NDVI-composites of the 1 km resolution sensor SPOT-VEGETATION. Belgium was selected as test zone, because of the availability of ample reference data in the form of a vectorial GIS with the boundaries and cover type of the large majority of agricultural fields. Two different methods were investigated: the linear mixture model and neural networks. Both result in area fraction images (AFIs), which contain for each 1 km pixel the estimated area proportions occupied by the different cover types (crops or other land use). Both algorithms were trained with part of the reference data and validated with the remainder. Validation was repeated at three different levels: the 1 km pixel, the municipality and the agro-statistical district. In general, the neural network outperformed the linear mixture model. For the major classes (winter wheat, maize, forest) the obtained acreage estimates showed good agreement with the true values, especially when aggregated to the level of the municipality (R² ≈ 85%) or district (R² ≈ 95%). The method seems attractive for wide-scale, regional area estimation in data-poor countries.