Abstract
Based on the results of Luati and Proietti (Ann Inst Stat Math 63:673–686, 2011) on an equivalence for a certain class of polynomial regressions between the diagonally weighted least squares (DWLS) and the generalized least squares (GLS) estimator, an alternative way to take correlations into account thanks to a diagonal covariance matrix is presented. The equivalent covariance matrix is much easier to compute than a diagonalization of the covariance matrix via eigenvalue decomposition which also implies a change of the least squares equations. This condensed matrix, for use in the least squares adjustment, can be seen as a diagonal or reduced version of the original matrix, its elements being simply the sums of the rows elements of the weighting matrix. The least squares results obtained with the equivalent diagonal matrices and those given by the fully populated covariance matrix are mathematically strictly equivalent for the mean estimator in terms of estimate and its a priori cofactor matrix. It is shown that this equivalence can be empirically extended to further classes of design matrices such as those used in GPS positioning (single point positioning, precise point positioning or relative positioning with double differences). Applying this new model to simulated time series of correlated observations, a significant reduction of the coordinate differences compared with the solutions computed with the commonly used diagonal elevation-dependent model was reached for the GPS relative positioning with double differences, single point positioning as well as precise point positioning cases. The estimate differences between the equivalent and classical model with fully populated covariance matrix were below the mm for all simulated GPS cases and below the sub-mm for the relative positioning with double differences. These results were confirmed by analyzing real data. Consequently, the equivalent diagonal covariance matrices, compared with the often used elevation-dependent diagonal covariance matrix is appropriate to take correlations in GPS least squares adjustment into account, yielding more accurate cofactor matrices of the unknown.
Similar content being viewed by others
References
Abramowitz M, Segun IA (1972) Handbook of mathematical functions. Dover, New York
Ammar GS, Gragg WB (1988) Superfast solution of real positive definite Toeplitz systems. SIAM J Matrix Anal Appl 9:61–76
Ataike H (1973) Block Toeplitz matrix inversion. SIAM J Appl Math 24(2):234–241
Alkhatib H, Schuh WD (2007) Integration of the Monte Carlo covariance estimation strategy into tailored solution procedures for large-scale least squares problems. JoG 81(1):53–66
Baksalary JK (1988) Criteria for the equality between ordinary least squares and best linear unbiaised estimators under certain linear models. Can J Stat 16(1):97–102
Barciss EH (1969) Numerical solution of linear equations of Toeplitz and vector Toeplitz systems. Numerische Mathematik 13:404–424
Beutler G, Bauersima I, Gurtner W, Rothacher M (1986) Correlations between simultaneous GPS double difference observations in the multistation mode: implementation considerations and first experiences. Manuscripta Geodaetica 12:40–44. Springer-Verlag, Berlin, Heidelberg, New York
Bierman GJ (1977) Factorization methods for discrete sequential estimation. Volume 123 of mathematics in science and engineering. Academic Press, New York
Bona P (2000) Precision, cross correlation, and time correlation of GPS phase and code observations. GPS Solut 4(2):3–13
Borre K, Tiberius C (2000) Time series analysis of GPS observables. In: Proceedings of ION GPS 2000, Salt Lake City, UT, USA, September 19–22, pp 1885–1894
Bottoni GP, Barzaghi R (1993) Fast collocation. Bull Géodésique 67(2):119–126
Brent R (1989) Old and new algorithms for Toeplitz systems. In: Luk FT (ed) Proceedings SPIE, volume 975, advanced algorithms and architectures for signal processing III. SPIE, Bellingham, pp 2–9
Brent RP, Gustavson FG, Yun DYY (1980) Fast solution of Toeplitz systems of equations and computation of Pade approximants. J Algorithms 1(259):295
Bruyninx C, Habrich H, Söhne W, Kenyeres A, Stangl G, Völksen C (2012) Enhancement of the EUREF Permanent Network services and products. In: Kenyon S, Pacino MC, Marti U (eds) Geodesy for Planet Earth. International association of geodesy symposia, vol 136. Springer, Berlin, Heidelberg, pp 27–35
Cressie N (1993) Statistics for spatial data. J. Wiley.&. Sons, Inc., New York
Dach R, Brockmann E, Schaer S, Beutler G, Meindl M, Prange L, Bock H, Jäggi A, Ostini L (2009) GNSS processing at CODE: status report. J Geod 83(3–4):353–365
Dach R, Hugentobler U, Fridez P, Meindl M (2007) Bernese GPS software version 5.0. Astronomical Institute, University of Bern, Switzerland
Durbin J (1960) The fitting of time series models. Rev Inst Int Stat 28:233–244
El-Rabbany A (1994) The effect of physical correlations on the ambiguity resolution and accuracy estimation in GPS differential positioning. PhD thesis, Department of Geodesy and Geomatics Engineering, University of New Brunswick, Canada
Euler HJ, Goad CC (1991) On optimal filtering of GPS dual frequency observations without using orbit information. Bull Geod 65(2):130–143
Grafarend EW (1976) Geodetic applications of stochastic processes. Phys Earth Planet Inter 12(2–3):151–179
Grafarend EW, Awange J (2012) Applications of linear and nonlinear models. Springer, Berlin
Gray RM (2006) Toeplitz and circulant matrices: a review. Found Trends Commun Inform Theory 2(3):155–239
Handcock MS, Wallis JR (1994) An approach to statistical spatial–temporal modeling of meteorological fields. J Am Stat Assoc 89(426):368–378
Hoffmann-Wellenhof B, Lichtenegger H, Collins J (2001) GPS theory and practice, 5th edn. Springer Wien, New York
Howind J, Kutterer H, Heck B (1999) Impact of temporal correlations on GPS-derived relative point positions. J Geod 73(5):246–258
Jansson P, Persson CG (2013) The effect of correlation on uncertainty estimates—with GPS examples. J Geod Sci 3(2):111–120
Jin SG, Luo O, Ren C (2010) Effects of physical correlations on long-distance GPS positioning and zenith tropospheric delay estimates. Adv Space Res 46:190–195
Journel AG, Huifbregts CJ (1978) Mining geostatistics. Academic Press, New York
Kermarrec G, Schön S (2014) On the Mátern covariance family: a proposal for modeling temporal correlations based on turbulence theory. J Geod 88:1061–1079
Klees R, Broersen P (2002) How to handle colored observation noise in large-scale least-squares problems-building the optimal filter. DUP Science, Delft University Press, Delft (30 pages)
Koch KR (1999) Parameter estimation and hypothesis testing in linear models. Springer, Berlin
Koivunen AC, Kostinski AB (1999) The feasibility of data whitening to improve performance of weather radar. J Appl Meteor 38:741–749
Krämer W (1986) Least squares regression when the independent variable follows an ARIMA process. J Am Stat Assoc 81(393):150–154. doi:10.2307/2287982
Krämer W, Donninger C (1987) Spatial autocorrelation among errors and the relative efficiency of OLS in the linear regression model. J Am Stat Assoc 82(398):577–579
Kutterer H (1999) On the sensitivity of the results of least-squares adjustments concerning the stochastic model. J Geod 73:350–361
Leandro R, Santos M, Cove K (2005) An empirical approach for the estimation of GPS covariance matrix of observations. In: Proceeding ION 61st Annual Meeting, The MITRE Corporation and Draper Laboratory, 27–29 June 2005, Cambridge, MA
Levinson N (1947) The Wiener RMS error criterion in filter design and prediction. J Math Phys 25(1–4):261–278
Luo X, Mayer M, Heck B (2012) Analysing time series of GNSS residuals by means of ARIMA processes. Int Assoc Geod Symp 137:129–134
Luo X (2012) Extending the GPS stochastic model by means of signal quality measures and ARMA processes. PhD Karlsruhe Institute of Technology
Luati A, Proietti T (2011) On the equivalence of the weighted least squares and the generalised least squares estimators, with applications to kernel smoothing. Ann Inst Stat Math 63(4):673–686
Mátern B (1960) Spatial variation-Stochastic models and their application to some problems in forest surveys and other sampling investigation. Medd Statens Skogsforskningsinstitut 49(5):144
Meier S (1981) Planar geodetic covariance functions. Rev Geophys Space Phys 19(4):673–686
Moritz H (1980) Advanced physical geodesy. Wichmann, Karlsruhe
Niemeier W (2008) Adjustment computations, 2nd edn. Walter de Gruyter, New York
Puntanen S (1987) On the relative goodness of ordinary least squares estimation in the general linear model. Acta Univ Tamper Ser A 216
Puntanen S, Styan G (1989) The equality of the ordinary least squares estimator and the best linear unbiased estimator. Am Stat 43(3):153–161. doi:10.2307/2685062
Radovanovic RS (2001) Variance-covariance modeling of carrier phase errors for rigorous adjustment of local area networks. IAG 2001 Scientific Assembly, Budapest, Hungary, September 2–7, 2001
Rao C, Toutenburg H (1999) Linear models, least-squares and alternatives, 2nd edn. Springer, New York
Santos MC, Vanicek P, Langley RB (1997) Effect of mathematical correlation on GPS network computation. J Surv Eng 123(3):101–112
Satirapod C, Wang J, Rizos C (2003) Comparing different GPS data processing techniques for modelling residual systematic errors. J Surv Eng 129(4):129–135
Schön S, Kutterer H (2006) Uncertainty in GPS networks due to remaining systematic errors: the interval approach. J Geod 80:150–162
Schön S, Brunner FK (2007) Treatment of refractivity fluctuations by fully populated variance-covariance matrices. In: Proc. 1st Colloquium Scientific and Fundamental Aspects of the Galileo Programme Toulouse Okt
Schön S, Brunner FK (2008) Atmospheric turbulence theory applied to GPS carrier-phase data. J Geod 1:47–57
Schuh WD, Krasbutter I, Kargoll B (2014) Korrelierte Messung—was nun? Neuner H (Hrsg.): Zeitabhängige Messgrößen—Ihre Daten haben (Mehr-)Wert, DVW-Schriftenreihe 74. ISBN: 978-3-89639-970-0, S. 85–101
Shkarofsky IP (1968) Generalized turbulence space-correlation and wave-number spectrum-function pairs. Can J Phys 46:2133–2153
Stein ML (1999) Interpolation of spatial data. Some theory for kriging. Springer, New York
Teunissen PJG, Kleusberg A (1998) GPS Geod. Springer, Berlin
Trench WF (1964) An algorithm for the inversion of finite Toeplitz matrices. J Soc Indus Appl Math 12(3):515–522
Trefethen LN, Bau D (1997) Numerical linear algebra. Society for Industrial and Applied Mathematics, Philadelphia. ISBN 0898713617, 9780898713619
Vennebusch M, Schön S, Weinbach U (2010) Temporal and spatial stochastic behavior of high-frequency slant tropospheric delays from simulations and real GPS data. Adv Space Res 47(10):1681–1690
Wang J, Satirapod C, Rizos C (2002) Stochastic assessment of GPS carrier phase measurements for precise static relative positioning. J Geod 76(2):95–104
Weinbach T (2012) (2012): Feasibility and impact of receiver clock modeling in precise GPS data analysis. Wissenschaftliche Arbeiten der Fachrichtung Geodäsie und Geoinformatik der Leibniz Universität Hannover ISSN 0174–1454, Nr. 303, Hannover 2013
Wieser A, Brunner FK (2000) An extended weight model for GPS phase observations. Earth Planet Space 52:777–782
Zyskind G (1967) On canonical forms, non-negative covariance matrices and best and simple least squares linear estimators in linear models. Ann Math Stat 38(4):1092–1109
Acknowledgments
Parts of the work were funded by the DFG under the label SCHO1314/1-2, this is gratefully acknowledged by the authors. The European Permanent Network and contributing agencies are thanked for providing freely GNSS data and products. Valuable comments of three anonymous reviewers helped us improve significantly the manuscript.
Author information
Authors and Affiliations
Corresponding author
Appendix: Mátern covariance family
Appendix: Mátern covariance family
First introduced by Mátern (1960), the covariance function reads:
Other parameterizations are possible as mentioned by Stein (1999), Shkarofsky (1968). Special cases arise when the smoothness factor is taken to:
-
1 / 2 :exponential covariance function
-
1: AR(1) process: autoregressive process of first order also called Markov process of first order. This process is often used in the field of geodesy to analyze gravitational fields (Meier 1981; Grafarend 1976) as well as to fit GPS covariance functions (Jansson and Persson 2013; Wang et al. 2002).
-
infinity: squared exponential covariance function. Due to its indefinitely differentiability which is difficult to explain physically, this covariance function (Stein 1999; Handcock and Wallis 1994) should be avoided. Some examples of problems that can arise using this smoothness are shown in Stein (1999).
Because of its flexibility as well as the possibility to estimate the parameters via maximum likelihood (Stein 1999; Handcock and Wallis 1994 for meteorological data field), we adopted this covariance function for all our computation of covariance matrices used in GPS least squares procedure.
Kermarrec and Schön (2014) proposed to compute the elements of the GPS covariance matrices thanks to a close formula: \(\sigma _{it}^{i( {t+\tau })} =\frac{\delta }{\sin ( {El_i ( t)})\sin ( {El_i ( {t+\tau })})}( {\alpha \tau })^1K_\nu ( {\alpha \tau })\), where \(El_i (t)\) is the elevation of satellite i at t and \(El_i (t+\tau )\) at \(t+\tau \). Other weightings than elevation-dependent model could be chosen. The ranges \(\alpha \in [0.005-0.025],\nu \in \left[ 1/4^{-1} \right] \) were proposed for GPS time series.
Rights and permissions
About this article
Cite this article
Kermarrec, G., Schön, S. Taking correlations in GPS least squares adjustments into account with a diagonal covariance matrix. J Geod 90, 793–805 (2016). https://doi.org/10.1007/s00190-016-0911-z
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00190-016-0911-z